# Technical Documentation: gcc-mirror/gcc > ℹ️ **Provenance:** Multi-Tier Ingestion · [CodeWiki Reference](https://codewiki.google/github.com/gcc-mirror/gcc) ### Gcc Ada Doc Gnat Style (`gcc/ada/doc/gnat-style.rst`) GNAT Coding Style: A Guide for GNAT Developers ============================================== General ------- Most of GNAT is written in Ada using a consistent style to ensure readability of the code. This document has been written to help maintain this consistent style, while having a large group of developers work on the compiler. For the coding style in the C parts of the compiler and run time, see the GNU Coding Guidelines. This document is structured after the Ada Reference Manual. Those familiar with that document should be able to quickly lookup style rules for particular constructs. Lexical Elements ---------------- Character Set and Separators **************************** .. index:: Character set .. index:: ASCII .. index:: Separators .. index:: End-of-line .. index:: Line length .. index:: Indentation * The character set used should be plain 7-bit ASCII. The only separators allowed are space and the end-of-line sequence. No other control character or format effector (such as ``HT``, ``VT``, ``FF`` ) should be used. The normal end-of-line sequence is used, which may be ``LF``, ``CR/LF`` or ``CR``, depending on the host system. An optional ``SUB`` ( ``16#1A#`` ) may be present as the last character in the file on hosts using that character as file terminator. * Files that are checked in or distributed should be in host format. * A line should never be longer than 79 characters, not counting the line separator. * Lines must not have trailing blanks. * Indentation is 3 characters per level for ``if`` statements, loops, and ``case`` statements. For exact information on required spacing between lexical elements, see file style.adb. .. index:: style.adb file Identifiers *********** * Identifiers will start with an upper case letter, and each letter following an underscore will be upper case. .. index:: Casing (for identifiers) Short acronyms may be all upper case. All other letters are lower case. An exception is for identifiers matching a foreign language. In particular, we use all lower case where appropriate for C. * Use underscores to separate words in an identifier. .. index:: Underscores * Try to limit your use of abbreviations in identifiers. It is ok to make a few abbreviations, explain what they mean, and then use them frequently, but don't use lots of obscure abbreviations. An example is the ``ALI`` word which stands for Ada Library Information and is by convention always written in upper-case when used in entity names. .. code-block:: ada procedure Find_ALI_Files; * Don't use the variable name ``I``, use ``J`` instead; ``I`` is too easily confused with ``1`` in some fonts. Similarly don't use the variable ``O``, which is too easily mistaken for the number ``0``. Numeric Literals **************** * Numeric literals should include underscores where helpful for readability. .. index:: Underscores .. code-block:: ada 1_000_000 16#8000_0000# 3.14159_26535_89793_23846 Reserved Words ************** * Reserved words use all lower case. .. index:: Casing (for reserved words) .. code-block:: ada return else * The words ``Access``, ``Delta`` and ``Digits`` are capitalized when used as attribute_designator. Comments ******** * A comment starts with ``--`` followed by two spaces. The only exception to this rule (i.e. one space is tolerated) is when the comment ends with a single space followed by ``--``. It is also acceptable to have only one space between ``--`` and the start of the comment when the comment is at the end of a line, after some Ada code. * Every sentence in a comment should start with an upper-case letter (including the first letter of the comment). .. index:: Casing (in comments) * When declarations are commented with 'hanging' comments, i.e. comments after the declaration, there is no blank line before the comment, and if it is absolutely necessary to have blank lines within the comments, e.g. to make paragraph separations within a single comment, these blank lines *do* have a ``--`` (unlike the normal rule, which is to use entirely blank lines for separating comment paragraphs). The comment starts at same level of indentation as code it is commenting. .. index:: Blank lines (in comments) .. index:: Indentation .. code-block:: ada z : Integer; -- Integer value for storing value of z -- -- The previous line was a blank line. * Comments that are dubious or incomplete, or that comment on possibly wrong or incomplete code, should be preceded or followed by ``???``. * Comments in a subprogram body must generally be surrounded by blank lines. An exception is a comment that follows a line containing a single keyword ( ``begin``, ``else``, ``loop`` ): .. code-block:: ada begin -- Comment for the next statement A := 5; -- Comment for the B statement B := 6; end; * In sequences of statements, comments at the end of the lines should be aligned. .. index:: Alignment (in comments) .. code-block:: ada My_Identifier := 5; -- First comment Other_Id := 6; -- Second comment * Short comments that fit on a single line are *not* ended with a period. Comments taking more than a line are punctuated in the normal manner. * Comments should focus on *why* instead of *what*. Descriptions of what subprograms do go with the specification. * Comments describing a subprogram spec should specifically mention the formal argument names. General rule: write a comment that does not depend on the names of things. The names are supplementary, not sufficient, as comments. * *Do not* put two spaces after periods in comments. Declarations and Types ---------------------- * In entity declarations, colons must be surrounded by spaces. Colons should be aligned. .. index:: Alignment (in declarations) .. code-block:: ada Entity1 : Integer; My_Entity : Integer; * Declarations should be grouped in a logical order. Related groups of declarations may be preceded by a header comment. * All local subprograms in a subprogram or package body should be declared before the first local subprogram body. * Do not declare local entities that hide global entities. .. index:: Hiding of outer entities * Do not declare multiple variables in one declaration that spans lines. Start a new declaration on each line, instead. * The defining_identifiers of global declarations serve as comments of a sort. So don't choose terse names, but look for names that give useful information instead. * Local names can be shorter, because they are used only within one context, where comments explain their purpose. * When starting an initialization or default expression on the line that follows the declaration line, use 2 characters for indentation. .. code-block:: ada Entity1 : Integer := Function_Name (Parameters, For_Call); * If an initialization or default expression needs to be continued on subsequent lines, the continuations should be indented from the start of the expression. .. code-block:: ada Entity1 : Integer := Long_Function_Name (parameters for call); Expressions and Names --------------------- * Every operator must be surrounded by spaces. An exception is that this rule does not apply to the exponentiation operator, for which there are no specific layout rules. The reason for this exception is that sometimes it makes clearer reading to leave out the spaces around exponentiation. .. index:: Operators .. code-block:: ada E := A * B**2 + 3 * (C - D); * Use parentheses where they clarify the intended association of operands with operators: .. index:: Parenthesization of expressions .. code-block:: ada (A / B) * C Statements ---------- Simple and Compound Statements ****************************** * Use only one statement or label per line. * A longer sequence_of_statements may be divided in logical groups or separated from surrounding code using a blank line. If Statements ************* * When the ``if``, ``elsif`` or ``else`` keywords fit on the same line with the condition and the ``then`` keyword, then the statement is formatted as follows: .. index:: Alignment (in an if statement) .. code-block:: ada if condition then ... elsif condition then ... else ... end if; When the above layout is not possible, ``then`` should be aligned with ``if``, and conditions should preferably be split before an ``and`` or ``or`` keyword a follows: .. code-block:: ada if long_condition_that_has_to_be_split and then continued_on_the_next_line then ... end if; The ``elsif``, ``else`` and ``end if`` always line up with the ``if`` keyword. The preferred location for splitting the line is before ``and`` or ``or``. The continuation of a condition is indented with two spaces or as many as needed to make nesting clear. As an exception, if conditions are closely related either of the following is allowed: .. code-block:: ada if x = lakdsjfhlkashfdlkflkdsalkhfsalkdhflkjdsahf or else x = asldkjhalkdsjfhhfd or else x = asdfadsfadsf then ... end if; if x = lakdsjfhlkashfdlkflkdsalkhfsalkdhflkjdsahf or else x = asldkjhalkdsjfhhfd or else x = asdfadsfadsf then ... end if; * Conditions should use short-circuit forms ( ``and then``, ``or else`` ), except when the operands are boolean variables or boolean constants. .. index:: Short-circuit forms * Complex conditions in ``if`` statements are indented two characters: .. index:: Indentation (in if statements) .. code-block:: ada if this_complex_condition and then that_other_one and then one_last_one then ... end if; There are some cases where complex conditionals can be laid out in manners that do not follow these rules to preserve better parallelism between branches, e.g. .. code-block:: ada if xyz.abc (gef) = 'c' or else xyz.abc (gef) = 'x' then ... end if; * Every ``if`` block is preceded and followed by a blank line, except where it begins or ends a sequence_of_statements. .. index:: Blank lines (in an if statement) .. code-block:: ada A := 5; if A = 5 then null; end if; A := 6; Case Statements *************** * Layout is as below. For long ``case`` statements, the extra indentation can be saved by aligning the ``when`` clauses with the opening ``case``. .. code-block:: ada case expression is when condition => ... when condition => ... end case; Loop Statements *************** * When possible, have ``for`` or ``while`` on one line with the condition and the ``loop`` keyword. .. code-block:: ada for J in S'Range loop ... end loop; If the condition is too long, split the condition (see 'If statements' above) and align ``loop`` with the ``for`` or ``while`` keyword. .. index:: Alignment (in a loop statement) .. code-block:: ada while long_condition_that_has_to_be_split and then continued_on_the_next_line loop ... end loop; If the loop_statement has an identifier, it is laid out as follows: .. code-block:: ada Outer : while not condition loop ... end Outer; Block Statements **************** * The ``declare`` (optional), ``begin`` and ``end`` words are aligned, except when the block_statement is named. There is a blank line before the ``begin`` keyword: .. index:: Alignment (in a block statement) .. code-block:: ada Some_Block : declare ... begin ... end Some_Block; Subprograms ----------- Subprogram Declarations *********************** * Do not write the ``in`` for parameters. .. code-block:: ada function Length (S : String) return Integer; * When the declaration line for a procedure or a function is too long to fit the entire declaration (including the keyword procedure or function) on a single line, then fold it, putting a single parameter on a line, aligning the colons, as in: .. code-block:: ada procedure Set_Heading (Source : String; Count : Natural; Pad : Character := Space; Fill : Boolean := True); In the case of a function, if the entire spec does not fit on one line, then the return may appear after the last parameter, as in: .. code-block:: ada function Head (Source : String; Count : Natural; Pad : Character := Space) return String; Or it may appear on its own as a separate line. This form is preferred when putting the return on the same line as the last parameter would result in an overlong line. The return type may optionally be aligned with the types of the parameters (usually we do this aligning if it results only in a small number of extra spaces, and otherwise we don't attempt to align). So two alternative forms for the above spec are: .. code-block:: ada function Head (Source : String; Count : Natural; Pad : Character := Space) return String; function Head (Source : String; Count : Natural; Pad : Character := Space) return String; Subprogram Bodies ***************** * Function and procedure bodies should usually be sorted alphabetically. Do not attempt to sort them in some logical order by functionality. For a sequence of subprogram specs, a general alphabetical sorting is also usually appropriate, but occasionally it makes sense to group by major function, with appropriate headers. * All subprograms have a header giving the function name, with the following format: .. code-block:: ada ----------------- -- My_Function -- ----------------- procedure My_Function is begin ... end My_Function; Note that the name in the header is preceded by a single space, not two spaces as for other comments. These headers are used on nested subprograms as well as outer level subprograms. They may also be used as headers for sections of comments, or collections of declarations that are related. * Every subprogram body must have a preceding subprogram_declaration, which includes proper client documentation so that you do not need to read the subprogram body in order to understand what the subprogram does and how to call it. All subprograms should be documented, without exceptions. .. index:: Blank lines (in subprogram bodies) * A sequence of declarations may optionally be separated from the following begin by a blank line. Just as we optionally allow blank lines in general between declarations, this blank line should be present only if it improves readability. Generally we avoid this blank line if the declarative part is small (one or two lines) and the body has no blank lines, and we include it if the declarative part is long or if the body has blank lines. * If the declarations in a subprogram contain at least one nested subprogram body, then just before the ``begin`` of the enclosing subprogram, there is a comment line and a blank line: .. code-block:: ada -- Start of processing for Enclosing_Subprogram begin ... end Enclosing_Subprogram; * When nested subprograms are present, variables that are referenced by any nested subprogram should precede the nested subprogram specs. For variables that are not referenced by nested procedures, the declarations can either also be before any of the nested subprogram specs (this is the old style, more generally used). Or then can come just before the begin, with a header. The following example shows the two possible styles: .. code-block:: ada procedure Style1 is Var_Referenced_In_Nested : Integer; Var_Referenced_Only_In_Style1 : Integer; proc Nested; -- Comments ... ------------ -- Nested -- ------------ procedure Nested is begin ... end Nested; -- Start of processing for Style1 begin ... end Style1; procedure Style2 is Var_Referenced_In_Nested : Integer; proc Nested; -- Comments ... ------------ -- Nested -- ------------ procedure Nested is begin ... end Nested; -- Local variables Var_Referenced_Only_In_Style2 : Integer; -- Start of processing for Style2 begin ... end Style2; For new code, we generally prefer Style2, but we do not insist on modifying all legacy occurrences of Style1, which is still much more common in the sources. Packages and Visibility Rules ----------------------------- * All program units and subprograms have their name at the end: .. code-block:: ada package P is ... end P; * We will use the style of ``use`` -ing ``with`` -ed packages, with the context clauses looking like: .. index:: use clauses .. code-block:: ada with A; use A; with B; use B; * Names declared in the visible part of packages should be unique, to prevent name clashes when the packages are ``use`` d. .. index:: Name clash avoidance .. code-block:: ada package Entity is type Entity_Kind is ...; ... end Entity; * After the file header comment, the context clause and unit specification should be the first thing in a program_unit. * Preelaborate, Pure and Elaborate_Body pragmas should be added right after the package name, indented an extra level and using the parameterless form: .. code-block:: ada package Preelaborate_Package is pragma Preelaborate; ... end Preelaborate_Package; Program Structure and Compilation Issues ---------------------------------------- * Every GNAT source file must be compiled with the ``-gnatg`` switch to check the coding style. (Note that you should look at style.adb to see the lexical rules enforced by ``-gnatg`` ). .. index:: -gnatg option (to gcc) .. index:: style.adb file * Each source file should contain only one compilation unit. * Filenames should be 8 or fewer characters, followed by the ``.adb`` extension for a body or ``.ads`` for a spec. .. index:: File name length * Unit names should be distinct when 'krunch'ed to 8 characters (see krunch.ads) and the filenames should match the unit name, except that they are all lower case. .. index:: krunch.ads file .. toctree:: share/gnu_free_documentation_license --- ### Gcc Ada Doc Gnat Rm (`gcc/ada/doc/gnat_rm.rst`) GNAT Reference Manual ===================== .. only:: not latex *GNAT, The GNU Ada Development Environment* .. only:: PRO *GNAT Pro Edition* | Version |version| | Date: |today| .. only:: GPL *GNAT Community Edition* | Version |version| | Date: |today| .. only:: FSF .. raw:: texinfo @include gcc-common.texi GCC version @value{version-GCC}@* AdaCore Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, with the Front-Cover Texts being "GNAT Reference Manual", and with no Back-Cover Texts. A copy of the license is included in the section entitled :ref:`gnu_fdl`. .. toctree:: :numbered: :maxdepth: 3 gnat_rm/about_this_guide gnat_rm/implementation_defined_pragmas gnat_rm/implementation_defined_aspects gnat_rm/implementation_defined_attributes gnat_rm/standard_and_implementation_defined_restrictions gnat_rm/implementation_advice gnat_rm/gnat_implementation_mode gnat_rm/implementation_defined_characteristics gnat_rm/intrinsic_subprograms gnat_rm/representation_clauses_and_pragmas gnat_rm/standard_library_routines gnat_rm/the_implementation_of_standard_i_o gnat_rm/the_gnat_library gnat_rm/interfacing_to_other_languages gnat_rm/specialized_needs_annexes gnat_rm/implementation_of_specific_ada_features gnat_rm/implementation_of_ada_2022_features gnat_rm/gnat_language_extensions gnat_rm/security_hardening_features gnat_rm/obsolescent_features gnat_rm/compatibility_and_porting_guide .. raw:: latex \appendix .. toctree:: :maxdepth: 3 share/gnu_free_documentation_license --- ### Gcc Ada Doc Gnat Ugn (`gcc/ada/doc/gnat_ugn.rst`) GNAT User's Guide for Native Platforms ====================================== .. only:: not latex *GNAT, The GNU Ada Development Environment* .. only:: PRO *GNAT Pro Edition* | Version |version| | Date: |today| .. only:: GPL *GNAT Community Edition* | Version |version| | Date: |today| .. only:: FSF .. raw:: texinfo @include gcc-common.texi GCC version @value{version-GCC}@* AdaCore Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, with the Front-Cover Texts being "GNAT User's Guide for Native Platforms", and with no Back-Cover Texts. A copy of the license is included in the section entitled :ref:`gnu_fdl`. .. toctree:: :maxdepth: 3 :numbered: gnat_ugn/about_this_guide gnat_ugn/getting_started_with_gnat gnat_ugn/the_gnat_compilation_model gnat_ugn/building_executable_programs_with_gnat gnat_ugn/gnat_utility_programs gnat_ugn/gnat_and_program_execution .. raw:: latex \appendix .. toctree:: :maxdepth: 3 A. Platform-Specific Information B. Example of Binder Output C. Elaboration Order Handling in GNAT D. Inline Assembler E. GNU Free Documentation License --- ### Gcc Doc Libgdiagnostics Index (`gcc/doc/libgdiagnostics/index.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see . libgdiagnostics =============== This document describes `libgdiagnostics `_, an API for programs to use to emit diagnostics (such as for "lint"-style checker tools), supporting: * text output similar to GCC's errors and warnings:: test-typo.c:19:13: error: unknown field 'colour' 19 | return p->colour; | ^~~~~~ quoting pertinent source code (with a cache), and underlining :doc:`points and ranges in the files being tested `, possibly with labels:: test-labelled-ranges.c:9:6: error: mismatching types: 'int' and 'const char *' 19 | 42 + "foo" | ~~ ^ ~~~~~ | | | | int const char * * emitting :doc:`fix-it hints `:: test-fix-it-hint.c:19:13: error: unknown field 'colour'; did you mean 'color' 19 | return p->colour; | ^~~~~~ | color and generating patches from them:: @@ -16,7 +16,7 @@ struct rgb get_color (struct object *p) { - return p->colour; + return p->color; } * capturing :doc:`execution paths` through code:: In function 'make_a_list_of_random_ints_badly': test-warning-with-path.c:30:5: warning: passing NULL as argument 1 to 'PyList_Append' which requires a non-NULL parameter" 30 | PyList_Append(list, item); | ^~~~~~~~~~~~~~~~~~~~~~~~~ make_a_list_of_random_ints_badly': events 1-3 26 | list = PyList_New(0); | ^~~~~~~~~~~~~ | | | (1) when 'PyList_New' fails, returning NULL 27 | 28 | for (i = 0; i < count; i++) { | ~~~~~~~~~ | | | (2) when 'i < count' 29 | item = PyLong_FromLong(random()); 30 | PyList_Append(list, item); | ~~~~~~~~~~~~~~~~~~~~~~~~~ | | | (3) when calling 'PyList_Append', passing NULL from (1) as argument 1 * support for emitting machine-readable representations of the above using the :doc:`SARIF file format ` There are actually two APIs for the library: * a pure C API: ``libgdiagnostics.h`` * a C++ wrapper API: ``libgdiagnostics+.h``. This is a header-only collection of wrapper classes around the C API to give a less verbose API. This documentation covers the C API. Contents ******** .. toctree:: :maxdepth: 2 tutorial/index.rst topics/index.rst libgdiagnostics is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. Indices and tables ****************** * :ref:`genindex` * :ref:`search` --- ### Gcc Jit Docs Index (`gcc/jit/docs/index.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see . libgccjit ========= This document describes `libgccjit `_, an API for embedding GCC inside programs and libraries. There are actually two APIs for the library: * a pure C API: ``libgccjit.h`` * a C++ wrapper API: ``libgccjit++.h``. This is a collection of "thin" wrapper classes around the C API, to save typing. Contents: .. toctree:: :maxdepth: 2 intro/index.rst topics/index.rst cp/index.rst internals/index.rst Indices and tables ================== * :ref:`genindex` * :ref:`search` --- ### Gcc Ada Doc Gnat Rm About This Guide (`gcc/ada/doc/gnat_rm/about_this_guide.rst`) .. _About_This_Guide: **************** About This Guide **************** .. only:: PRO For ease of exposition, 'GNAT Pro' will be referred to simply as 'GNAT' in the remainder of this document. This manual contains useful information in writing programs using the GNAT compiler. It includes information on implementation dependent characteristics of GNAT, including all the information required by Annex M of the Ada language standard. GNAT implements Ada 95, Ada 2005, Ada 2012 and Ada 2022, and it may also be invoked in Ada 83 compatibility mode. By default, GNAT assumes Ada 2012, but you can override with a compiler switch to explicitly specify the language version. (Please refer to the *GNAT User's Guide* for details on these switches.) Throughout this manual, references to 'Ada' without a year suffix apply to all the Ada versions of the language. Ada is designed to be highly portable. In general, a program will have the same effect even when compiled by different compilers on different platforms. However, since Ada is designed to be used in a wide variety of applications, it also contains a number of system dependent features to be used in interfacing to the external world. .. index:: Implementation-dependent features .. index:: Portability Note: Any program that makes use of implementation-dependent features may be non-portable. You should follow good programming practice and isolate and clearly document any sections of your program that make use of these features in a non-portable manner. What This Reference Manual Contains =================================== This reference manual contains the following chapters: * :ref:`Implementation_Defined_Pragmas`, lists GNAT implementation-dependent pragmas, which can be used to extend and enhance the functionality of the compiler. * :ref:`Implementation_Defined_Attributes`, lists GNAT implementation-dependent attributes, which can be used to extend and enhance the functionality of the compiler. * :ref:`Standard_and_Implementation_Defined_Restrictions`, lists GNAT implementation-dependent restrictions, which can be used to extend and enhance the functionality of the compiler. * :ref:`Implementation_Advice`, provides information on generally desirable behavior which are not requirements that all compilers must follow since it cannot be provided on all systems, or which may be undesirable on some systems. * :ref:`GNAT_Implementation_Mode` details the features and restrictions that are enabled while building the compiler and its runtime libraries. * :ref:`Implementation_Defined_Characteristics`, provides a guide to minimizing implementation dependent features. * :ref:`Intrinsic_Subprograms`, describes the intrinsic subprograms implemented by GNAT, and how they can be imported into user application programs. * :ref:`Representation_Clauses_and_Pragmas`, describes in detail the way that GNAT represents data, and in particular the exact set of representation clauses and pragmas that is accepted. * :ref:`Standard_Library_Routines`, provides a listing of packages and a brief description of the functionality that is provided by Ada's extensive set of standard library routines as implemented by GNAT. * :ref:`The_Implementation_of_Standard_I/O`, details how the GNAT implementation of the input-output facilities. * :ref:`The_GNAT_Library`, is a catalog of packages that complement the Ada predefined library. * :ref:`Interfacing_to_Other_Languages`, describes how programs written in Ada using GNAT can be interfaced to other programming languages. * :ref:`Specialized_Needs_Annexes`, describes the GNAT implementation of all of the specialized needs annexes. * :ref:`Implementation_of_Specific_Ada_Features`, discusses issues related to GNAT's implementation of machine code insertions, tasking, and several other features. * :ref:`Implementation_of_Ada_2022_Features`, describes the status of the GNAT implementation of the Ada 2022 language standard. * :ref:`Security_Hardening_Features` documents GNAT extensions aimed at security hardening. * :ref:`Obsolescent_Features` documents implementation dependent features, including pragmas and attributes, which are considered obsolescent, since there are other preferred ways of achieving the same results. These obsolescent forms are retained for backwards compatibility. * :ref:`Compatibility_and_Porting_Guide` presents some guidelines for developing portable Ada code, describes the compatibility issues that may arise between GNAT and other Ada compilation systems (including those for Ada 83), and shows how GNAT can expedite porting applications developed in other Ada environments. * :ref:`gnu_fdl` contains the license for this document. .. index:: Ada 95 Language Reference Manual .. index:: Ada 2005 Language Reference Manual This reference manual assumes a basic familiarity with the Ada 95 language, as described in the :title:`International Standard ANSI/ISO/IEC-8652:1995`. It does not require knowledge of the new features introduced by Ada 2005 or Ada 2012. All three reference manuals are included in the GNAT documentation package. Conventions =========== .. index:: Conventions, typographical .. index:: Typographical conventions Following are examples of the typographical and graphic conventions used in this guide: * ``Functions``, ``utility program names``, ``standard names``, and ``classes``. * ``Option flags`` * :file:`File names` * ``Variables`` * *Emphasis* * [optional information or parameters] * Examples are described by text :: and then shown this way. * Commands that are entered by the user are shown as preceded by a prompt string comprising the ``$`` character followed by a space. Related Information =================== See the following documents for further information on GNAT: * :title:`GNAT User's Guide for Native Platforms`, which provides information on how to use the GNAT development environment. * :title:`Ada 95 Reference Manual`, the Ada 95 programming language standard. * :title:`Ada 95 Annotated Reference Manual`, which is an annotated version of the Ada 95 standard. The annotations describe detailed aspects of the design decision, and in particular contain useful sections on Ada 83 compatibility. * :title:`Ada 2005 Reference Manual`, the Ada 2005 programming language standard. * :title:`Ada 2005 Annotated Reference Manual`, which is an annotated version of the Ada 2005 standard. The annotations describe detailed aspects of the design decision. * :title:`Ada 2012 Reference Manual`, the Ada 2012 programming language standard. * :title:`DEC Ada, Technical Overview and Comparison on DIGITAL Platforms`, which contains specific information on compatibility between GNAT and DEC Ada 83 systems. * :title:`DEC Ada, Language Reference Manual`, part number AA-PYZAB-TK, which describes in detail the pragmas and attributes provided by the DEC Ada 83 compiler system. --- ### Gcc Ada Doc Gnat Rm Compatibility And Porting Guide (`gcc/ada/doc/gnat_rm/compatibility_and_porting_guide.rst`) .. _Compatibility_and_Porting_Guide: ******************************* Compatibility and Porting Guide ******************************* This chapter presents some guidelines for developing portable Ada code, describes the compatibility issues that may arise between GNAT and other Ada compilation systems (including those for Ada 83), and shows how GNAT can expedite porting applications developed in other Ada environments. .. _Writing_Portable_Fixed-Point_Declarations: Writing Portable Fixed-Point Declarations ========================================= The Ada Reference Manual gives an implementation freedom to choose bounds that are narrower by ``Small`` from the given bounds. For example, if we write .. code-block:: ada type F1 is delta 1.0 range -128.0 .. +128.0; then the implementation is allowed to choose -128.0 .. +127.0 if it likes, but is not required to do so. This leads to possible portability problems, so let's have a closer look at this, and figure out how to avoid these problems. First, why does this freedom exist, and why would an implementation take advantage of it? To answer this, take a closer look at the type declaration for ``F1`` above. If the compiler uses the given bounds, it would need 9 bits to hold the largest positive value (and typically that means 16 bits on all machines). But if the implementation chooses the +127.0 bound then it can fit values of the type in 8 bits. Why not make the user write +127.0 if that's what is wanted? The rationale is that if you are thinking of fixed point as a kind of 'poor man's floating-point', then you don't want to be thinking about the scaled integers that are used in its representation. Let's take another example: .. code-block:: ada type F2 is delta 2.0**(-15) range -1.0 .. +1.0; Looking at this declaration, it seems casually as though it should fit in 16 bits, but again that extra positive value +1.0 has the scaled integer equivalent of 2**15 which is one too big for signed 16 bits. The implementation can treat this as: .. code-block:: ada type F2 is delta 2.0**(-15) range -1.0 .. +1.0-(2.0**(-15)); and the Ada language design team felt that this was too annoying to require. We don't need to debate this decision at this point, since it is well established (the rule about narrowing the ranges dates to Ada 83). But the important point is that an implementation is not required to do this narrowing, so we have a potential portability problem. We could imagine three types of implementation: (a) those that narrow the range automatically if they can figure out that the narrower range will allow storage in a smaller machine unit, (b) those that will narrow only if forced to by a ``'Size`` clause, and (c) those that will never narrow. Now if we are language theoreticians, we can imagine a fourth approach: to narrow all the time, e.g. to treat .. code-block:: ada type F3 is delta 1.0 range -10.0 .. +23.0; as though it had been written: .. code-block:: ada type F3 is delta 1.0 range -9.0 .. +22.0; But although technically allowed, such a behavior would be hostile and silly, and no real compiler would do this. All real compilers will fall into one of the categories (a), (b) or (c) above. So, how do you get the compiler to do what you want? The answer is give the actual bounds you want, and then use a ``'Small`` clause and a ``'Size`` clause to absolutely pin down what the compiler does. E.g., for ``F2`` above, we will write: .. code-block:: ada My_Small : constant := 2.0**(-15); My_First : constant := -1.0; My_Last : constant := +1.0 - My_Small; type F2 is delta My_Small range My_First .. My_Last; and then add .. code-block:: ada for F2'Small use my_Small; for F2'Size use 16; In practice all compilers will do the same thing here and will give you what you want, so the above declarations are fully portable. If you really want to play language lawyer and guard against ludicrous behavior by the compiler you could add .. code-block:: ada Test1 : constant := 1 / Boolean'Pos (F2'First = My_First); Test2 : constant := 1 / Boolean'Pos (F2'Last = My_Last); One or other or both are allowed to be illegal if the compiler is behaving in a silly manner, but at least the silly compiler will not get away with silently messing with your (very clear) intentions. If you follow this scheme you will be guaranteed that your fixed-point types will be portable. .. _Compatibility_with_Ada_83: Compatibility with Ada 83 ========================= .. index:: Compatibility (between Ada 83 and Ada 95 / Ada 2005 / Ada 2012 / Ada 2022) Ada 95 and the subsequent revisions Ada 2005, Ada 2012, Ada 2022 are highly upwards compatible with Ada 83. In particular, the design intention was that the difficulties associated with moving from Ada 83 to later versions of the standard should be no greater than those that occur when moving from one Ada 83 system to another. However, there are a number of points at which there are minor incompatibilities. The :title:`Ada 95 Annotated Reference Manual` contains full details of these issues as they relate to Ada 95, and should be consulted for a complete treatment. In practice the following subsections treat the most likely issues to be encountered. .. _Legal_Ada_83_programs_that_are_illegal_in_Ada_95: Legal Ada 83 programs that are illegal in Ada 95 ------------------------------------------------ Some legal Ada 83 programs are illegal (i.e., they will fail to compile) in Ada 95 and later versions of the standard: * *Character literals* Some uses of character literals are ambiguous. Since Ada 95 has introduced ``Wide_Character`` as a new predefined character type, some uses of character literals that were legal in Ada 83 are illegal in Ada 95. For example: .. code-block:: ada for Char in 'A' .. 'Z' loop ... end loop; The problem is that 'A' and 'Z' could be from either ``Character`` or ``Wide_Character``. The simplest correction is to make the type explicit; e.g.: .. code-block:: ada for Char in Character range 'A' .. 'Z' loop ... end loop; * *New reserved words* The identifiers ``abstract``, ``aliased``, ``protected``, ``requeue``, ``tagged``, and ``until`` are reserved in Ada 95. Existing Ada 83 code using any of these identifiers must be edited to use some alternative name. * *Freezing rules* The rules in Ada 95 are slightly different with regard to the point at which entities are frozen, and representation pragmas and clauses are not permitted past the freeze point. This shows up most typically in the form of an error message complaining that a representation item appears too late, and the appropriate corrective action is to move the item nearer to the declaration of the entity to which it refers. A particular case is that representation pragmas cannot be applied to a subprogram body. If necessary, a separate subprogram declaration must be introduced to which the pragma can be applied. * *Optional bodies for library packages* In Ada 83, a package that did not require a package body was nevertheless allowed to have one. This lead to certain surprises in compiling large systems (situations in which the body could be unexpectedly ignored by the binder). In Ada 95, if a package does not require a body then it is not permitted to have a body. To fix this problem, simply remove a redundant body if it is empty, or, if it is non-empty, introduce a dummy declaration into the spec that makes the body required. One approach is to add a private part to the package declaration (if necessary), and define a parameterless procedure called ``Requires_Body``, which must then be given a dummy procedure body in the package body, which then becomes required. Another approach (assuming that this does not introduce elaboration circularities) is to add an ``Elaborate_Body`` pragma to the package spec, since one effect of this pragma is to require the presence of a package body. * *Numeric_Error is the same exception as Constraint_Error* In Ada 95, the exception ``Numeric_Error`` is a renaming of ``Constraint_Error``. This means that it is illegal to have separate exception handlers for the two exceptions. The fix is simply to remove the handler for the ``Numeric_Error`` case (since even in Ada 83, a compiler was free to raise ``Constraint_Error`` in place of ``Numeric_Error`` in all cases). * *Indefinite subtypes in generics* In Ada 83, it was permissible to pass an indefinite type (e.g, ``String``) as the actual for a generic formal private type, but then the instantiation would be illegal if there were any instances of declarations of variables of this type in the generic body. In Ada 95, to avoid this clear violation of the methodological principle known as the 'contract model', the generic declaration explicitly indicates whether or not such instantiations are permitted. If a generic formal parameter has explicit unknown discriminants, indicated by using ``(<>)`` after the subtype name, then it can be instantiated with indefinite types, but no stand-alone variables can be declared of this type. Any attempt to declare such a variable will result in an illegality at the time the generic is declared. If the ``(<>)`` notation is not used, then it is illegal to instantiate the generic with an indefinite type. This is the potential incompatibility issue when porting Ada 83 code to Ada 95. It will show up as a compile time error, and the fix is usually simply to add the ``(<>)`` to the generic declaration. .. _More_deterministic_semantics: More deterministic semantics ---------------------------- * *Conversions* Conversions from real types to integer types round away from 0. In Ada 83 the conversion Integer(2.5) could deliver either 2 or 3 as its value. This implementation freedom was intended to support unbiased rounding in statistical applications, but in practice it interfered with portability. In Ada 95 the conversion semantics are unambiguous, and rounding away from 0 is required. Numeric code may be affected by this change in semantics. Note, though, that this issue is no worse than already existed in Ada 83 when porting code from one vendor to another. * *Tasking* The Real-Time Annex introduces a set of policies that define the behavior of features that were implementation dependent in Ada 83, such as the order in which open select branches are executed. .. _Changed_semantics: Changed semantics ----------------- The worst kind of incompatibility is one where a program that is legal in Ada 83 is also legal in Ada 95 but can have an effect in Ada 95 that was not possible in Ada 83. Fortunately this is extremely rare, but the one situation that you should be alert to is the change in the predefined type ``Character`` from 7-bit ASCII to 8-bit Latin-1. .. index:: Latin-1 * *Range of type ``Character``* The range of ``Standard.Character`` is now the full 256 characters of Latin-1, whereas in most Ada 83 implementations it was restricted to 128 characters. Although some of the effects of this change will be manifest in compile-time rejection of legal Ada 83 programs it is possible for a working Ada 83 program to have a different effect in Ada 95, one that was not permitted in Ada 83. As an example, the expression ``Character'Pos(Character'Last)`` returned ``127`` in Ada 83 and now delivers ``255`` as its value. In general, you should look at the logic of any character-processing Ada 83 program and see whether it needs to be adapted to work correctly with Latin-1. Note that the predefined Ada 95 API has a character handling package that may be relevant if code needs to be adapted to account for the additional Latin-1 elements. The desirable fix is to modify the program to accommodate the full character set, but in some cases it may be convenient to define a subtype or derived type of Character that covers only the restricted range. .. _Other_language_compatibility_issues: Other language compatibility issues ----------------------------------- * *-gnat83* switch All implementations of GNAT provide a switch that causes GNAT to operate in Ada 83 mode. In this mode, some but not all compatibility problems of the type described above are handled automatically. For example, the new reserved words introduced in Ada 95 and Ada 2005 are treated simply as identifiers as in Ada 83. However, in practice, it is usually advisable to make the necessary modifications to the program to remove the need for using this switch. See the ``Compiling Different Versions of Ada`` section in the :title:`GNAT User's Guide`. * Support for removed Ada 83 pragmas and attributes A number of pragmas and attributes from Ada 83 were removed from Ada 95, generally because they were replaced by other mechanisms. Ada 95 and Ada 2005 compilers are allowed, but not required, to implement these missing elements. In contrast with some other compilers, GNAT implements all such pragmas and attributes, eliminating this compatibility concern. These include ``pragma Interface`` and the floating point type attributes (``Emax``, ``Mantissa``, etc.), among other items. .. _Compatibility_between_Ada_95_and_Ada_2005: Compatibility between Ada 95 and Ada 2005 ========================================= .. index:: Compatibility between Ada 95 and Ada 2005 Although Ada 2005 was designed to be upwards compatible with Ada 95, there are a number of incompatibilities. Several are enumerated below; for a complete description please see the :title:`Annotated Ada 2005 Reference Manual`, or section 9.1.1 in :title:`Rationale for Ada 2005`. * *New reserved words.* The words ``interface``, ``overriding`` and ``synchronized`` are reserved in Ada 2005. A pre-Ada 2005 program that uses any of these as an identifier will be illegal. * *New declarations in predefined packages.* A number of packages in the predefined environment contain new declarations: ``Ada.Exceptions``, ``Ada.Real_Time``, ``Ada.Strings``, ``Ada.Strings.Fixed``, ``Ada.Strings.Bounded``, ``Ada.Strings.Unbounded``, ``Ada.Strings.Wide_Fixed``, ``Ada.Strings.Wide_Bounded``, ``Ada.Strings.Wide_Unbounded``, ``Ada.Tags``, ``Ada.Text_IO``, and ``Interfaces.C``. If an Ada 95 program does a ``with`` and ``use`` of any of these packages, the new declarations may cause name clashes. * *Access parameters.* A nondispatching subprogram with an access parameter cannot be renamed as a dispatching operation. This was permitted in Ada 95. * *Access types, discriminants, and constraints.* Rule changes in this area have led to some incompatibilities; for example, constrained subtypes of some access types are not permitted in Ada 2005. * *Aggregates for limited types.* The allowance of aggregates for limited types in Ada 2005 raises the possibility of ambiguities in legal Ada 95 programs, since additional types now need to be considered in expression resolution. * *Fixed-point multiplication and division.* Certain expressions involving '*' or '/' for a fixed-point type, which were legal in Ada 95 and invoked the predefined versions of these operations, are now ambiguous. The ambiguity may be resolved either by applying a type conversion to the expression, or by explicitly invoking the operation from package ``Standard``. * *Return-by-reference types.* The Ada 95 return-by-reference mechanism has been removed. Instead, the user can declare a function returning a value from an anonymous access type. .. _Implementation-dependent_characteristics: Implementation-dependent characteristics ======================================== Although the Ada language defines the semantics of each construct as precisely as practical, in some situations (for example for reasons of efficiency, or where the effect is heavily dependent on the host or target platform) the implementation is allowed some freedom. In porting Ada 83 code to GNAT, you need to be aware of whether / how the existing code exercised such implementation dependencies. Such characteristics fall into several categories, and GNAT offers specific support in assisting the transition from certain Ada 83 compilers. .. _Implementation-defined_pragmas: Implementation-defined pragmas ------------------------------ Ada compilers are allowed to supplement the language-defined pragmas, and these are a potential source of non-portability. All GNAT-defined pragmas are described in :ref:`Implementation_Defined_Pragmas`, and these include several that are specifically intended to correspond to other vendors' Ada 83 pragmas. For migrating from VADS, the pragma ``Use_VADS_Size`` may be useful. For compatibility with HP Ada 83, GNAT supplies the pragmas ``Extend_System``, ``Ident``, ``Inline_Generic``, ``Interface_Name``, ``Passive``, ``Suppress_All``, and ``Volatile``. Other relevant pragmas include ``External`` and ``Link_With``. Some vendor-specific Ada 83 pragmas (``Share_Generic``, ``Subtitle``, and ``Title``) are recognized, thus avoiding compiler rejection of units that contain such pragmas; they are not relevant in a GNAT context and hence are not otherwise implemented. .. _Implementation-defined_attributes: Implementation-defined attributes --------------------------------- Analogous to pragmas, the set of attributes may be extended by an implementation. All GNAT-defined attributes are described in :ref:`Implementation_Defined_Attributes`, and these include several that are specifically intended to correspond to other vendors' Ada 83 attributes. For migrating from VADS, the attribute ``VADS_Size`` may be useful. For compatibility with HP Ada 83, GNAT supplies the attributes ``Bit``, ``Machine_Size`` and ``Type_Class``. .. _Libraries: Libraries --------- Vendors may supply libraries to supplement the standard Ada API. If Ada 83 code uses vendor-specific libraries then there are several ways to manage this in Ada 95 and later versions of the standard: * If the source code for the libraries (specs and bodies) are available, then the libraries can be migrated in the same way as the application. * If the source code for the specs but not the bodies are available, then you can reimplement the bodies. * Some features introduced by Ada 95 obviate the need for library support. For example most Ada 83 vendors supplied a package for unsigned integers. The Ada 95 modular type feature is the preferred way to handle this need, so instead of migrating or reimplementing the unsigned integer package it may be preferable to retrofit the application using modular types. .. _Elaboration_order: Elaboration order ----------------- The implementation can choose any elaboration order consistent with the unit dependency relationship. This freedom means that some orders can result in Program_Error being raised due to an 'Access Before Elaboration': an attempt to invoke a subprogram before its body has been elaborated, or to instantiate a generic before the generic body has been elaborated. By default GNAT attempts to choose a safe order (one that will not encounter access before elaboration problems) by implicitly inserting ``Elaborate`` or ``Elaborate_All`` pragmas where needed. However, this can lead to the creation of elaboration circularities and a resulting rejection of the program by gnatbind. This issue is thoroughly described in the *Elaboration Order Handling in GNAT* appendix in the :title:`GNAT User's Guide`. In brief, there are several ways to deal with this situation: * Modify the program to eliminate the circularities, e.g., by moving elaboration-time code into explicitly-invoked procedures * Constrain the elaboration order by including explicit ``Elaborate_Body`` or ``Elaborate`` pragmas, and then inhibit the generation of implicit ``Elaborate_All`` pragmas either globally (as an effect of the *-gnatE* switch) or locally (by selectively suppressing elaboration checks via pragma ``Suppress(Elaboration_Check)`` when it is safe to do so). .. _Target-specific_aspects: Target-specific aspects ----------------------- Low-level applications need to deal with machine addresses, data representations, interfacing with assembler code, and similar issues. If such an Ada 83 application is being ported to different target hardware (for example where the byte endianness has changed) then you will need to carefully examine the program logic; the porting effort will heavily depend on the robustness of the original design. Moreover, Ada 95 (and thus Ada 2005, Ada 2012, and Ada 2022) are sometimes incompatible with typical Ada 83 compiler practices regarding implicit packing, the meaning of the Size attribute, and the size of access values. GNAT's approach to these issues is described in :ref:`Representation_Clauses`. .. _Compatibility_with_Other_Ada_Systems: Compatibility with Other Ada Systems ==================================== If programs avoid the use of implementation dependent and implementation defined features, as documented in the :title:`Ada Reference Manual`, there should be a high degree of portability between GNAT and other Ada systems. The following are specific items which have proved troublesome in moving Ada 95 programs from GNAT to other Ada 95 compilers, but do not affect porting code to GNAT. (As of January 2007, GNAT is the only compiler available for Ada 2005; the following issues may or may not arise for Ada 2005 programs when other compilers appear.) * *Ada 83 Pragmas and Attributes* Ada 95 compilers are allowed, but not required, to implement the missing Ada 83 pragmas and attributes that are no longer defined in Ada 95. GNAT implements all such pragmas and attributes, eliminating this as a compatibility concern, but some other Ada 95 compilers reject these pragmas and attributes. * *Specialized Needs Annexes* GNAT implements the full set of special needs annexes. At the current time, it is the only Ada 95 compiler to do so. This means that programs making use of these features may not be portable to other Ada 95 compilation systems. * *Representation Clauses* Some other Ada 95 compilers implement only the minimal set of representation clauses required by the Ada 95 reference manual. GNAT goes far beyond this minimal set, as described in the next section. .. _Representation_Clauses: Representation Clauses ====================== The Ada 83 reference manual was quite vague in describing both the minimal required implementation of representation clauses, and also their precise effects. Ada 95 (and thus also Ada 2005) are much more explicit, but the minimal set of capabilities required is still quite limited. GNAT implements the full required set of capabilities in Ada 95 and Ada 2005, but also goes much further, and in particular an effort has been made to be compatible with existing Ada 83 usage to the greatest extent possible. A few cases exist in which Ada 83 compiler behavior is incompatible with the requirements in Ada 95 (and thus also Ada 2005). These are instances of intentional or accidental dependence on specific implementation dependent characteristics of these Ada 83 compilers. The following is a list of the cases most likely to arise in existing Ada 83 code. * *Implicit Packing* Some Ada 83 compilers allowed a Size specification to cause implicit packing of an array or record. This could cause expensive implicit conversions for change of representation in the presence of derived types, and the Ada design intends to avoid this possibility. Subsequent AI's were issued to make it clear that such implicit change of representation in response to a Size clause is inadvisable, and this recommendation is represented explicitly in the Ada 95 (and Ada 2005) Reference Manuals as implementation advice that is followed by GNAT. The problem will show up as an error message rejecting the size clause. The fix is simply to provide the explicit pragma ``Pack``, or for more fine tuned control, provide a Component_Size clause. * *Meaning of Size Attribute* The Size attribute in Ada 95 (and Ada 2005) for discrete types is defined as the minimal number of bits required to hold values of the type. For example, on a 32-bit machine, the size of ``Natural`` will typically be 31 and not 32 (since no sign bit is required). Some Ada 83 compilers gave 31, and some 32 in this situation. This problem will usually show up as a compile time error, but not always. It is a good idea to check all uses of the 'Size attribute when porting Ada 83 code. The GNAT specific attribute Object_Size can provide a useful way of duplicating the behavior of some Ada 83 compiler systems. * *Size of Access Types* A common assumption in Ada 83 code is that an access type is in fact a pointer, and that therefore it will be the same size as a System.Address value. This assumption is true for GNAT in most cases with one exception. For the case of a pointer to an unconstrained array type (where the bounds may vary from one value of the access type to another), the default is to use a 'fat pointer', which is represented as two separate pointers, one to the bounds, and one to the array. This representation has a number of advantages, including improved efficiency. However, it may cause some difficulties in porting existing Ada 83 code which makes the assumption that, for example, pointers fit in 32 bits on a machine with 32-bit addressing. To get around this problem, GNAT also permits the use of 'thin pointers' for access types in this case (where the designated type is an unconstrained array type). These thin pointers are indeed the same size as a System.Address value. To specify a thin pointer, use a size clause for the type, for example: .. code-block:: ada type X is access all String; for X'Size use Standard'Address_Size; which will cause the type X to be represented using a single pointer. When using this representation, the bounds are right behind the array. This representation is slightly less efficient, and does not allow quite such flexibility in the use of foreign pointers or in using the Unrestricted_Access attribute to create pointers to non-aliased objects. But for any standard portable use of the access type it will work in a functionally correct manner and allow porting of existing code. Note that another way of forcing a thin pointer representation is to use a component size clause for the element size in an array, or a record representation clause for an access field in a record. See the documentation of Unrestricted_Access in the GNAT RM for a full discussion of possible problems using this attribute in conjunction with thin pointers. .. _Compatibility_with_HP_Ada_83: Compatibility with HP Ada 83 ============================ All the HP Ada 83 pragmas and attributes are recognized, although only a subset of them can sensibly be implemented. The description of pragmas in :ref:`Implementation_Defined_Pragmas` indicates whether or not they are applicable to GNAT. * *Default floating-point representation* In GNAT, the default floating-point format is IEEE, whereas in HP Ada 83, it is VMS format. * *System* the package System in GNAT exactly corresponds to the definition in the Ada 95 reference manual, which means that it excludes many of the HP Ada 83 extensions. However, a separate package Aux_DEC is provided that contains the additional definitions, and a special pragma, Extend_System allows this package to be treated transparently as an extension of package System. --- ### Gcc Ada Doc Gnat Rm Gnat Implementation Mode (`gcc/ada/doc/gnat_rm/gnat_implementation_mode.rst`) .. _GNAT_Implementation_Mode: ************************************************** GNAT Implementation Mode (the ``-gnatg`` switch) ************************************************** ``-gnatg`` enables **GNAT implementation mode**, the internal compilation mode used when building GNAT itself and its runtime libraries. It is not intended for user application programs. It implies ``-gnatw.ge`` (extended warning set, treated as errors) and ``-gnatyg`` (GNAT style checks). Language & Semantic Changes =========================== Ada Version and Identifier Rules --------------------------------- * **Ada version is fixed at Ada 2012**, regardless of other switches. * **Identifier character set is restricted to 7-bit ASCII**; extended characters are disallowed. * **Identifiers are allowed to start with underscore** to avoid conflicts between global symbols in the runtime and those defined in the user code. Limited Types ------------- * Copying limited objects is **permitted** (normally illegal). * Initializing limited types emits a **warning** rather than an error. * Returning objects of limited types is **permitted**. Preelaboration and Categorization ---------------------------------- * Categorization errors (pure/preelaborated unit violations) are downgraded to **warnings**. * Statements in preelaborated units emit a **warning** rather than an error. * Private objects and non-static constants/calls in preelaborated units emit **warnings** rather than errors. * Calls to non-preelaborable subprograms in preelaborated packages are demoted to **suppressible warnings** (RM 10.2.1), allowing the runtime to include elaboration code that would otherwise be forbidden. * Assignment statements to library-level variables are **excluded** from access-before-elaboration (ABE) analysis, on the assumption that the predefined runtime manages its own elaboration order correctly. Overflow -------- Overflow checks are **suppressed** but the overflow checking mode is set to **Strict** (for assertion purposes). Miscellaneous ------------- * ``Scalar_Storage_Order`` attribute applies to **generic types** (normally it does not). * Only the ``--!`` special comment sequence is recognized; all others are disallowed. * The ``Priority`` aspect/attribute on protected types accepts the unconstrained type ``Any_Priority`` rather than the subtype ``System.Priority`` (a subrange of ``Integer``), giving GNAT's own protected objects access to the full implementation-defined priority range. * ``pragma Extend_System`` is **ignored** (prevents circularities in System). Exception: predefined renaming units (``Text_IO``, ``Direct_IO``, etc.) are compiled without GNAT implementation mode so that ``Extend_System`` still works for them. * Importing a non-pure external function (``pragma Import``) into a ``Pure`` unit does **not** produce a purity warning, on the assumption that the runtime knows the purity properties of the C functions it imports. * ``pragma Style_Checks ("All_Checks")`` activates the stricter ``-gnatyg`` rule set rather than the default ``-gnaty`` set. ALI Files and Recompilation ---------------------------- * ALI files are always generated for predefined generic packages. * Compiling a unit whose top-level name is ``Ada``, ``Interfaces``, or ``System`` as the main unit is permitted; the RM reserves these as hierarchical library-unit name prefixes, but the diagnostic is suppressed to let GNAT compile its own predefined library (``Ada.Text_IO``, ``System.Storage_Elements``, etc.) as main units. --- ### Gcc Ada Doc Gnat Rm Gnat Language Extensions (`gcc/ada/doc/gnat_rm/gnat_language_extensions.rst`) .. role:: ada(code) :language: ada .. _GNAT_Language_Extensions: ************************ GNAT language extensions ************************ The GNAT compiler implements a certain number of language extensions on top of the latest Ada standard, implementing its own extended superset of Ada. There are two sets of language extensions: * The first is the curated set. The features in that set are features that we consider being worthy additions to the Ada language, and that we want to make available to users early on. * The second is the experimental set. It includes the first, but also experimental features, which are considered experimental because they're still in an early prototyping phase. These features might be removed or heavily modified at any time. How to activate the extended GNAT Ada superset ============================================== There are two ways to activate the extended GNAT Ada superset: * The :ref:`Pragma Extensions_Allowed`. To activate the curated set of extensions, you should use .. code-block:: ada pragma Extensions_Allowed (On) As a configuration pragma, you can either put it at the beginning of a source file, or in a ``.adc`` file corresponding to your project. * The ``-gnatX`` command-line option will activate the curated subset of extensions. .. attention:: You can activate the experimental set of extensions in addition by using either the ``-gnatX0`` command-line option, or the pragma ``Extensions_Allowed`` with ``All_Extensions`` as an argument. However, it is not recommended you use this subset for serious projects; it is only meant as a technology preview for use in playground experiments. .. _Curated_Language_Extensions: Curated Extensions ================== Features activated via ``-gnatX`` or ``pragma Extensions_Allowed (On)``. Local Declarations Without Block -------------------------------- A ``basic_declarative_item`` may appear at the place of any statement. This avoids the heavy syntax of block_statements just to declare something locally. The only valid kinds of declarations for now are ``object_declaration``, ``object_renaming_declaration``, ``use_package_clause``, and ``use_type_clause``. For example: .. code-block:: ada if X > 5 then X := X + 1; Squared : constant Integer := X**2; X := X + Squared; end if; It is generally a good practice to declare local variables (or constants) with as short a lifetime as possible. However, introducing a declare block to accomplish this is a relatively heavy syntactic load along with a traditional extra level of indentation. The alternative syntax supported here allows declarations in any statement sequence. The lifetime of such local declarations is until the end of the enclosing construct. The same enclosing construct cannot contain several declarations of the same defining name; however, overriding symbols from higher-level scopes works similarly to the explicit ``declare`` block. If the enclosing construct allows an exception handler (such as an accept statement, begin-except-end block or a subprogram body), declarations that appear at the place of a statement are *not* visible within the handler. Only declarations that precede the beginning of the construct with an exception handler would be visible in this handler. .. attention:: Here are a couple of examples illustrating the scoping rules described above. 1. Those declarations are not visible from the potential exception handler: .. code-block:: ada begin A : Integer ... exception when others => Put_Line (A'Image) -- ILLEGAL end; 2. The following is legal .. code-block:: ada declare A : Integer := 10; begin A : Integer := 12; end; because it is roughly expanded into .. code-block:: ada declare A : Integer := 10; begin declare A : Integer := 12; begin ... end; end; And as such the second ``A`` declaration is hiding the first one. Deep delta Aggregates --------------------- Ada 2022's delta aggregates are extended to allow deep updates. A delta aggregate may be used to specify new values for subcomponents of the copied base value, instead of only new values for direct components of the copied base value. This allows a more compact expression of updated values with a single delta aggregate, instead of multiple nested delta aggregates. The syntax of delta aggregates in the extended version is the following: Syntax ^^^^^^ .. code:: delta_aggregate ::= record_delta_aggregate | array_delta_aggregate record_delta_aggregate ::= ( base_expression with delta record_subcomponent_association_list ) record_subcomponent_association_list ::= record_subcomponent_association {, record_subcomponent_association} record_subcomponent_association ::= record_subcomponent_choice_list => expression record_subcomponent_choice_list ::= record_subcomponent_choice {'|' record_subcomponent_choice} record_subcomponent_choice ::= component_selector_name | record_subcomponent_choice (expression) | record_subcomponent_choice . component_selector_name array_delta_aggregate ::= ( base_expression with delta array_component_association_list ) | '[' base_expression with delta array_component_association_list ']' | ( base_expression with delta array_subcomponent_association_list ) | '[' base_expression with delta array_subcomponent_association_list ']' array_subcomponent_association_list ::= array_subcomponent_association {, array_subcomponent_association} array_subcomponent_association ::= array_subcomponent_choice_list => expression array_subcomponent_choice_list ::= array_subcomponent_choice {'|' array_subcomponent_choice} array_subcomponent_choice ::= ( expression ) | array_subcomponent_choice (expression) | array_subcomponent_choice . component_selector_name Legality Rules ^^^^^^^^^^^^^^ 1. For an ``array_delta_aggregate``, the discrete_choice shall not be **others**. 2. For an ``array_delta_aggregate``, the dimensionality of the type of the ``delta_aggregate`` shall be 1. 3. For an ``array_delta_aggregate``, the ``base_expression`` and each expression in every ``array_component_association`` or ``array_subcomponent_association`` shall be of a nonlimited type. 4. For a ``record_delta_aggregate``, no ``record_subcomponent_choices`` that consists of only ``component_selector_names`` shall be the same or a prefix of another record_subcomponent_choice. 5. For an ``array_subcomponent_choice`` or a ``record_subcomponent_choice`` the ``component_selector_name`` shall not be a subcomponent that depends on discriminants of an unconstrained record subtype with defaulted discriminants unless its prefix consists of only ``component_selector_names``. [Rationale: As a result of this rule, accessing the subcomponent can only lead to a discriminant check failure if the subcomponent was not present in the object denoted by the base_expression, prior to any update.] Dynamic Semantics ^^^^^^^^^^^^^^^^^ The evaluation of a ``delta_aggregate`` begins with the evaluation of the ``base_expression`` of the delta_aggregate; then that value is used to create and initialize the anonymous object of the aggregate. The bounds of the anonymous object of an ``array_delta_aggregate`` and the discriminants (if any) of the anonymous object of a ``record_delta_aggregate`` are those of the ``base_expression``. If a ``record_delta_aggregate`` is of a specific tagged type, its tag is that of the specific type; if it is of a class-wide type, its tag is that of the base_expression. For a ``delta_aggregate``, for each ``discrete_choice`` or each subcomponent associated with a ``record_subcomponent_associated``, ``array_component_association`` or ``array_subcomponent_association`` (in the order given in the enclosing ``discrete_choice_list`` or ``subcomponent_association_list``, respectively): - if the associated subcomponent belongs to a variant, a check is made that the values of the governing discriminants are such that the anonymous object has this component. The exception ``Constraint_Error`` is raised if this check fails. - if the associated subcomponent is a subcomponent of an array, then for each represented index value (in ascending order, if the ``discrete_choice`` represents a range): * the index value is converted to the index type of the array type. * a check is made that the index value belongs to the index range of the corresponding array part of the anonymous object; ``Constraint_Error`` is raised if this check fails. * the expression of the ``record_subcomponent_association``, ``array_component_association`` or ``array_subcomponent_association`` is evaluated, converted to the nominal subtype of the associated subcomponent, and assigned to the corresponding subcomponent of the anonymous object. Examples ^^^^^^^^ .. code-block:: ada :linenos: declare type Point is record X, Y : Integer; end record; type Segment is array (1 .. 2) of Point; type Triangle is array (1 .. 3) of Segment; S : Segment := (1 .. 2 => (0, 0)); T : Triangle := (1 .. 3 => (1 .. 2 => (0, 0))); begin S := (S with delta (1).X => 11, (2).Y => 12, (1).Y => 15); pragma Assert (S (1).X = 11); pragma Assert (S (2).Y = 12); pragma Assert (S (1).Y = 15); T := (T with delta (2)(1).Y => 18); pragma Assert (T (2)(1).Y = 18); end; Fixed lower bounds for array types and subtypes ----------------------------------------------- Unconstrained array types and subtypes can be specified with a lower bound that is fixed to a certain value, by writing an index range that uses the syntax `` .. <>``. This guarantees that all objects of the type or subtype will have the specified lower bound. For example, a matrix type with fixed lower bounds of zero for each dimension can be declared by the following: .. code-block:: ada type Matrix is array (Natural range 0 .. <>, Natural range 0 .. <>) of Integer; Objects of type ``Matrix`` declared with an index constraint must have index ranges starting at zero: .. code-block:: ada M1 : Matrix (0 .. 9, 0 .. 19); M2 : Matrix (2 .. 11, 3 .. 22); -- Warning about bounds; will raise CE Similarly, a subtype of ``String`` can be declared that specifies the lower bound of objects of that subtype to be ``1``: .. code-block:: ada subtype String_1 is String (1 .. <>); If a string slice is passed to a formal of subtype ``String_1`` in a call to a subprogram ``S``, the slice's bounds will "slide" so that the lower bound is ``1``. Within ``S``, the lower bound of the formal is known to be ``1``, so, unlike a normal unconstrained ``String`` formal, there is no need to worry about accounting for other possible lower-bound values. Sliding of bounds also occurs in other contexts, such as for object declarations with an unconstrained subtype with fixed lower bound, as well as in subtype conversions. Use of this feature increases safety by simplifying code, and can also improve the efficiency of indexing operations, since the compiler statically knows the lower bound of unconstrained array formals when the formal's subtype has index ranges with static fixed lower bounds. Prefixed-view notation for calls to primitive subprograms of untagged types --------------------------------------------------------------------------- When operating on an untagged type, if it has any primitive operations, and the first parameter of an operation is of the type (or is an access parameter with an anonymous type that designates the type), you may invoke these operations using an ``object.op(...)`` notation, where the parameter that would normally be the first parameter is brought out front, and the remaining parameters (if any) appear within parentheses after the name of the primitive operation. This same notation is already available for tagged types. This extension allows for untagged types. It is allowed for all primitive operations of the type independent of whether they were originally declared in a package spec, or were inherited and/or overridden as part of a derived type declaration occurring anywhere, so long as the first parameter is of the type, or an access parameter designating the type. For example: .. code-block:: ada generic type Elem_Type is private; package Vectors is type Vector is private; procedure Add_Element (V : in out Vector; Elem : Elem_Type); function Nth_Element (V : Vector; N : Positive) return Elem_Type; function Length (V : Vector) return Natural; private function Capacity (V : Vector) return Natural; -- Return number of elements that may be added without causing -- any new allocation of space type Vector is ... with Type_Invariant => Vector.Length <= Vector.Capacity; ... end Vectors; package Int_Vecs is new Vectors(Integer); V : Int_Vecs.Vector; ... V.Add_Element(42); V.Add_Element(-33); pragma Assert (V.Length = 2); pragma Assert (V.Nth_Element(1) = 42); Expression defaults for generic formal functions ------------------------------------------------ The declaration of a generic formal function is allowed to specify an expression as a default, using the syntax of an expression function. Here is an example of this feature: .. code-block:: ada generic type T is private; with function Copy (Item : T) return T is (Item); -- Defaults to Item package Stacks is type Stack is limited private; procedure Push (S : in out Stack; X : T); -- Calls Copy on X function Pop (S : in out Stack) return T; -- Calls Copy to return item private -- ... end Stacks; If Stacks is instantiated with an explicit actual for Copy, then that will be called when Copy is called in the generic body. If the default is used (i.e. there is no actual corresponding to Copy), then calls to Copy in the instance will simply return Item. String interpolation -------------------- The syntax for string literals is extended to support string interpolation. An interpolated string literal starts with ``f``, immediately before the first double-quote character. Within an interpolated string literal, an arbitrary expression, when enclosed in ``{ ... }``, is expanded at run time into the result of calling ``'Image`` on the result of evaluating the expression enclosed by the brace characters, unless it is already a string or a single character. Here is an example of this feature where the expressions ``Name`` and ``X + Y`` will be evaluated and included in the string. .. code-block:: ada procedure Test_Interpolation is X : Integer := 12; Y : Integer := 15; Name : String := "Leo"; begin Put_Line (f"The name is {Name} and the sum is {X + Y}."); end Test_Interpolation; This will print: .. code-block:: ada The name is Leo and the sum is 27. In addition, an escape character (``\``) is provided for inserting certain standard control characters (such as ``\t`` for tabulation or ``\n`` for newline) or to escape characters with special significance to the interpolated string syntax, namely ``"``, ``{``, ``}``,and ``\`` itself. ================= ================= escaped_character meaning ----------------- ----------------- ``\a`` ALERT ``\b`` BACKSPACE ``\f`` FORM FEED ``\n`` LINE FEED ``\r`` CARRIAGE RETURN ``\t`` CHARACTER TABULATION ``\v`` LINE TABULATION ``\0`` NUL ----------------- ----------------- ``\\`` ``\`` ``\"`` ``"`` ``\{`` ``{`` ``\}`` ``}`` ================= ================= Note that, unlike normal string literals, doubled double-quote characters have no special significance. So to include a double-quote or a brace character in an interpolated string, they must be preceded by a ``\``. Multiple interpolated strings are concatenated. For example: .. code-block:: ada Put_Line (f"X = {X} and Y = {Y} and X+Y = {X+Y};\n" & f" a double quote is \" and" & f" an open brace is \{"); This will print: .. code-block:: ada X = 12 and Y = 15 and X+Y = 27 a double quote is " and an open brace is { Constrained attribute for generic objects ----------------------------------------- The ``Constrained`` attribute is permitted for objects of generic types. The result indicates whether the corresponding actual is constrained. ``Static`` aspect on intrinsic functions ---------------------------------------- The Ada 202x ``Static`` aspect can be specified on Intrinsic imported functions and the compiler will evaluate some of these intrinsics statically, in particular the ``Shift_Left`` and ``Shift_Right`` intrinsics. First Controlling Parameter --------------------------- A new pragma/aspect, ``First_Controlling_Parameter``, is introduced for tagged types, altering the semantics of primitive/controlling parameters. When a tagged type is marked with this aspect, only subprograms where the first parameter is of that type will be considered dispatching primitives. This pragma/aspect applies to the entire hierarchy, starting from the specified type, without affecting inherited primitives. Here is an example of this feature: .. code-block:: ada package Example is type Root is tagged private; procedure P (V : Integer; V2 : Root); -- Primitive type Child is tagged private with First_Controlling_Parameter; private type Root is tagged null record; type Child is new Root with null record; overriding procedure P (V : Integer; V2 : Child); -- Primitive procedure P2 (V : Integer; V2 : Child); -- NOT Primitive function F return Child; -- NOT Primitive function F2 (V : Child) return Child; -- Primitive, but only controlling on the first parameter end Example; Note that ``function F2 (V : Child) return Child;`` differs from ``F2 (V : Child) return Child'Class;`` in that the return type is a specific, definite type. This is also distinct from the legacy semantics, where further derivations with added fields would require overriding the function. The option ``-gnatw_j``, that you can pass to the compiler directly, enables warnings related to this new language feature. For instance, compiling the example above without this switch produces no warnings, but compiling it with ``-gnatw_j`` generates the following warning on the declaration of procedure P2: .. code-block:: ada warning: not a dispatching primitive of tagged type "Child" warning: disallowed by First_Controlling_Parameter on "Child" For generic formal tagged types, you can specify whether the type has the First_Controlling_Parameter aspect enabled: .. code-block:: ada generic type T is tagged private with First_Controlling_Parameter; package T is type U is new T with null record; function Foo return U; -- Not a primitive end T; For tagged partial views, the value of the aspect must be consistent between the partial and full views: .. code-block:: ada package R is type T is tagged private; ... private type T is tagged null record with First_Controlling_Parameter; -- ILLEGAL end R; Restricting the position of controlling parameter offers several advantages: * Simplification of the dispatching rules improves readability of Ada programs. One doesn't need to analyze all subprogram parameters to understand if the given subprogram is a primitive of a certain tagged type. * A programmer is free to use any type, including class-wide types, on other parameters of a subprogram, without the need to consider possible effects of overriding a primitive or creating new one. * The result of a function is never a controlling result. ``Unsigned_Base_Range`` aspect ------------------------------ A new pragma/aspect, ``Unsigned_Base_Range``, is introduced to explicitly enforce the use of an unsigned base type for signed integer types. RM-3.5.4(9) mandates a symmetric base range for signed integer types. This requirement often requires the use of larger data types for arithmetic operations, potentially introducing undesirable run-time overhead and performance penalties, particularly in embedded systems. For instance, on a 64-bit architecture, a 64-bit multiplication can be performed with a single hardware instruction, whereas a 128-bit multiplication requires multiple instructions and intermediate steps. Here is an example of this feature: .. code-block:: ada type Uns_64 is range 0 .. 2 ** 64 - 1 with Size => 64, Unsigned_Base_Range => True; It ensures that arithmetic operations of type ``Uns_64`` are carried out using 64 bits. Generalized Finalization ------------------------ The ``Finalizable`` aspect can be applied to any record type, tagged or not, to specify that it provides the same level of control on the operations of initialization, finalization, and assignment of objects as the controlled types (see RM 7.6(2) for a high-level overview). The only restriction is that the record type must be a root type, in other words not a derived type. The aspect additionally makes it possible to specify relaxed semantics for the finalization operations by means of the ``Relaxed_Finalization`` setting. Here is the archetypal example: .. code-block:: ada type T is record ... end record with Finalizable => (Initialize => Initialize, Adjust => Adjust, Finalize => Finalize, Relaxed_Finalization => True); procedure Adjust (Obj : in out T); procedure Finalize (Obj : in out T); procedure Initialize (Obj : in out T); The three procedures must be primitive operations of ``T`` and have a single ``in out`` parameter. They need not be all specified by the aspect. If they are specified, they have the same dynamic semantics as for controlled types: - ``Initialize`` is called when an object of type ``T`` is declared without initialization expression. - ``Adjust`` is called after an object of type ``T`` is assigned a new value. - ``Finalize`` is called when an object of type ``T`` goes out of scope (for stack-allocated objects) or is deallocated (for heap-allocated objects). It is also called when the value is replaced by an assignment. However, when ``Relaxed_Finalization`` is either ``True`` or not explicitly specified, the following differences are implemented relative to the semantics of controlled types: * The compiler has permission to perform no automatic finalization of heap-allocated objects: ``Finalize`` is only called when such an object is explicitly deallocated, or when the designated object is assigned a new value. As a consequence, no runtime support is needed for performing implicit deallocation. In particular, no per-object header data is needed for heap-allocated objects. Heap-allocated objects allocated through a nested access type will therefore **not** be deallocated either. The result is simply that memory will be leaked in this case. * The ``Adjust`` and ``Finalize`` procedures are automatically considered as having the :ref:`No_Raise_Aspect` specified for them. In particular, the compiler has permission to enforce none of the guarantees specified by the RM 7.6.1 (14/1) and subsequent subclauses. Simple example of ref-counted type: .. code-block:: ada type T is record Value : Integer; Ref_Count : Natural := 0; end record; procedure Inc_Ref (X : in out T); procedure Dec_Ref (X : in out T); type T_Access is access all T; type T_Ref is record Value : T_Access; end record with Finalizable => (Adjust => Adjust, Finalize => Finalize); procedure Adjust (Ref : in out T_Ref) is begin Inc_Ref (Ref.Value); end Adjust; procedure Finalize (Ref : in out T_Ref) is begin Def_Ref (Ref.Value); end Finalize; Simple file handle that ensures resources are properly released: .. code-block:: ada package P is type File (<>) is limited private; function Open (Path : String) return File; procedure Close (F : in out File); private type File is limited record Handle : ...; end record with Finalizable (Finalize => Close); end P; Finalizable tagged types ^^^^^^^^^^^^^^^^^^^^^^^^ The aspect is inherited by derived types and the primitives may be overridden by the derivation. The compiler-generated calls to these operations are then dispatching whenever it makes sense, i.e. when the object in question is of a class-wide type and the class includes at least one finalizable tagged type. Composite types ^^^^^^^^^^^^^^^ When a finalizable type is used as a component of a composite type, the latter becomes finalizable as well. The three primitives are derived automatically in order to call the primitives of their components. The dynamic semantics is the same as for controlled components of composite types. Interoperability with controlled types ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Finalizable types are fully interoperable with controlled types, in particular it is possible for a finalizable type to have a controlled component and vice versa, but the stricter dynamic semantics, in other words that of controlled types, is applied in this case. .. _Experimental_Language_Extensions: Experimental Language Extensions ================================ Features activated via ``-gnatX0`` or ``pragma Extensions_Allowed (All_Extensions)``. Conditional when constructs --------------------------- This feature extends the use of ``when`` as a way to condition a control-flow related statement, to all control-flow related statements. To do a conditional return in a procedure the following syntax should be used: .. code-block:: ada procedure P (Condition : Boolean) is begin return when Condition; end P; This will return from the procedure if ``Condition`` is true. When being used in a function the conditional part comes after the return value: .. code-block:: ada function Is_Null (I : Integer) return Boolean is begin return True when I = 0; return False; end; In a similar way to the ``exit when`` a ``goto ... when`` can be employed: .. code-block:: ada procedure Low_Level_Optimized is Flags : Bitmapping; begin Do_1 (Flags); goto Cleanup when Flags (1); Do_2 (Flags); goto Cleanup when Flags (32); -- ... <> -- ... end; .. code-block To use a conditional raise construct: .. code-block:: ada procedure Foo is begin raise Error when Imported_C_Func /= 0; end; An exception message can also be added: .. code-block:: ada procedure Foo is begin raise Error with "Unix Error" when Imported_C_Func /= 0; end; Implicit With ------------- This feature allows a standalone ``use`` clause in the context clause of a compilation unit to imply an implicit ``with`` of the same library unit where an equivalent ``with`` clause would be allowed. .. code-block:: ada use Ada.Text_IO; procedure Main is begin Put_Line ("Hello"); end; Storage Model ------------- This extends Storage Pools into a more efficient model allowing higher performance, easier integration with low footprint embedded run-times and copying data between different pools of memory. The latter is especially useful when working with distributed memory models, in particular to support interactions with GPU. Aspect Storage_Model_Type ^^^^^^^^^^^^^^^^^^^^^^^^^^ A Storage model is a type with a specified ``Storage_Model_Type`` aspect, e.g.: .. code-block:: Ada type A_Model is null record with Storage_Model_Type (...); Storage_Model_Type itself accepts six parameters: - Address_Type, the type of the address managed by this model. This has to be a scalar type or derived from System.Address. - Allocate, a procedure used for allocating memory in this model - Deallocate, a procedure used for deallocating memory in this model - Copy_To, a procedure used to copy memory from native memory to this model - Copy_From, a procedure used to copy memory from this model to native memory - Storage_Size, a function returning the amount of memory left - Null_Address, a value for the null address value By default, Address_Type is System.Address, and the five subprograms perform native operations (e.g. the allocator is the native ``new`` allocator). Users can decide to specify one or more of these. When an Address_Type is specified to be other than System.Address, all of the subprograms have to be specified. The prototypes of these procedures are as follows: .. code-block:: Ada procedure Allocate (Model : in out A_Model; Storage_Address : out Address_Type; Size : Storage_Count; Alignment : Storage_Count); procedure Deallocate (Model : in out A_Model; Storage_Address : out Address_Type; Size : Storage_Count; Alignment : Storage_Count); procedure Copy_To (Model : in out A_Model; Target : Address_Type; Source : System.Address; Size : Storage_Count); procedure Copy_From (Model : in out A_Model; Target : System.Address; Source : Address_Type; Size : Storage_Count); function Storage_Size (Pool : A_Model) return Storage_Count; Here's an example of how this could be instantiated in the context of CUDA: .. code-block:: Ada package CUDA_Memory is type CUDA_Storage_Model is null record with Storage_Model_Type => ( Address_Type => CUDA_Address, Allocate => CUDA_Allocate, Deallocate => CUDA_Deallocate, Copy_To => CUDA_Copy_To, Copy_From => CUDA_Copy_From, Storage_Size => CUDA_Storage_Size, Null_Address => CUDA_Null_Address ); type CUDA_Address is new System.Address; -- We're assuming for now same address size on host and device procedure CUDA_Allocate (Model : in out CUDA_Storage_Model; Storage_Address : out CUDA_Address; Size : Storage_Count; Alignment : Storage_Count); procedure CUDA_Deallocate (Model : in out CUDA_Storage_Model; Storage_Address : out CUDA_Address; Size : Storage_Count; Alignment : Storage_Count); procedure CUDA_Copy_To (Model : in out CUDA_Storage_Model; Target : CUDA_Address; Source : System.Address; Size : Storage_Count); procedure CUDA_Copy_From (Model : in out CUDA_Storage_Model; Target : System.Address; Source : CUDA_Address; Size : Storage_Count); function CUDA_Storage_Size (Pool : CUDA_Storage_Model) return Storage_Count return Storage_Count'Last; CUDA_Null_Address : constant CUDA_Address := CUDA_Address (System.Null_Address); CUDA_Memory : CUDA_Storage_Model; end CUDA_Memory; Aspect Designated_Storage_Model ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ A new aspect, Designated_Storage_Model, allows to specify the memory model for the objects pointed by an access type. Under this aspect, allocations and deallocations will come from the specified memory model instead of the standard ones. In addition, if write operations are needed for initialization, or if there is a copy of the target object from and to a standard memory area, the Copy_To and Copy_From functions will be called. It allows to encompass the capabilities of storage pools, e.g.: .. code-block:: Ada procedure Main is type Integer_Array is array (Integer range <>) of Integer; type Host_Array_Access is access all Integer_Array; type Device_Array_Access is access Integer_Array with Designated_Storage_Model => CUDA_Memory; procedure Free is new Unchecked_Deallocation (Host_Array_Type, Host_Array_Access); procedure Free is new Unchecked_Deallocation (Device_Array_Type, Device_Array_Access); Host_Array : Host_Array_Access := new Integer_Array (1 .. 10); Device_Array : Device_Array_Access := new Host_Array (1 .. 10); -- Calls CUDA_Storage_Model.Allocate to allocate the fat pointers and -- the bounds, then CUDA_Storage_Model.Copy_In to copy the values of the -- boundaries. begin Host_Array.all := (others => 0); Device_Array.all := Host_Array.all; -- Calls CUDA_Storage_Model.Copy_To to write to the device array from the -- native memory. Host_Array.all := Device_Array.all; -- Calls CUDA_Storage_Model.Copy_From to read from the device array and -- write to native memory. Free (Host_Array); Free (Device_Array); -- Calls CUDA_Storage_Model.Deallocate; end; Taking ``'Address`` of an object with a specific memory model returns an object of the type of the address for that memory category, which may be different from System.Address. When copying is performed between two specific memory models, the native memory is used as a temporary between the two. E.g.: .. code-block:: Ada type Foo_I is access Integer with Designated_Storage_Model => Foo; type Bar_I is access Integer with Designated_Storage_Model => Bar; X : Foo_I := new Integer; Y : Bar_I := new Integer; begin X.all := Y.all; conceptually becomes: .. code-block:: Ada X : Foo_I := new Integer; T : Integer; Y : Bar_I := new Integer; begin T := Y.all; X.all := T; Legacy Storage Pools ^^^^^^^^^^^^^^^^^^^^^ Legacy Storage Pools are now replaced by a Storage_Model_Type. They are implemented as follows: .. code-block:: Ada type Root_Storage_Pool is abstract new Ada.Finalization.Limited_Controlled with private with Storage_Model_Type => ( Allocate => Allocate, Deallocate => Deallocate, Storage_Size => Storage_Size ); pragma Preelaborable_Initialization (Root_Storage_Pool); procedure Allocate (Pool : in out Root_Storage_Pool; Storage_Address : out System.Address; Size_In_Storage_Elements : System.Storage_Elements.Storage_Count; Alignment : System.Storage_Elements.Storage_Count) is abstract; procedure Deallocate (Pool : in out Root_Storage_Pool; Storage_Address : System.Address; Size_In_Storage_Elements : System.Storage_Elements.Storage_Count; Alignment : System.Storage_Elements.Storage_Count) is abstract; function Storage_Size (Pool : Root_Storage_Pool) return System.Storage_Elements.Storage_Count is abstract; The legacy notation: .. code-block:: Ada type My_Pools is new Root_Storage_Pool with record [...] My_Pool_Instance : Storage_Model_Pool.Storage_Model := My_Pools'(others => <>); type Acc is access Integer_Array with Storage_Pool => My_Pool; can still be accepted as a shortcut for the new syntax. Attribute Super --------------- .. index:: Super The ``Super`` attribute can be applied to objects of tagged types in order to obtain a view conversion to the most immediate specific parent type. It cannot be applied to objects of types without any ancestors. .. code-block:: ada type T1 is tagged null record; procedure P (V : T1); type T2 is new T1 with null record; type T3 is new T2 with null record; procedure P (V : T3); procedure Call ( V1 : T1'Class; V2 : T2'Class; V3 : T3'Class) is begin V1'Super.P; -- Illegal call as T1 doesn't have any ancestors V2'Super.P; -- Equivalent to "T1 (V).P;", a non-dispatching call -- to T1's primitive procedure P. V3'Super.P; -- Equivalent to "T2 (V).P;"; Since T2 doesn't -- override P, a non-dispatching call to T1.P is -- executed. end; Simpler Accessibility Model --------------------------- The goal of this feature is to simplify the accessibility rules by removing dynamic accessibility checks that are often difficult to understand and debug. The new rules eliminate the need for runtime accessibility checks by imposing more conservative legality rules when enabled via a new restriction (see RM 13.12), No_Dynamic_Accessibility_Checks, which prevents dangling reference problems at compile time. This restriction has no effect on the user-visible behavior of a program when executed; the only effect of this restriction is to enable additional compile-time checks (described below) which ensure statically that Ada's dynamic accessibility checks will not fail. The feature can be activated with ``pragma Restrictions (No_Dynamic_Accessibility_Checks);``. As a result, additional compile-time checks are performed; these checks pertain to stand-alone objects, subprogram parameters, and function results as described below. All of the refined rules are compatible with the [use of anonymous access types in SPARK] (http://docs.adacore.com/spark2014-docs/html/lrm/declarations-and-types.html#access-types). Stand-alone objects ^^^^^^^^^^^^^^^^^^^^ .. code-block:: ada Var : access T := ... Var_To_Cst : access constant T := ... Cst : constant access T := ... Cst_To_Cst : constant access constant T := ... In this section, we will refer to a stand-alone object of an anonymous access type as an SO. When the restriction is in effect, the "statically deeper" relationship (see RM 3.10.2(4)) does apply to the type of a SO (contrary to RM 3.10.2(19.2)) and, for the purposes of compile-time checks, the accessibility level of the type of a SO is the accessibility level of that SO. This supports many common use-cases without the employment of ``Unchecked_Access`` while still removing the need for dynamic checks. This statically disallows cases that would otherwise require a dynamic accessibility check, such as .. code-block:: ada type Ref is access all Integer; Ptr : Ref; Good : aliased Integer; procedure Proc is Bad : aliased Integer; Stand_Alone : access Integer; begin if then Stand_Alone := Good'Access; else Stand_Alone := Bad'Access; end if; Ptr := Ref (Stand_Alone); end Proc; If a No_Dynamic_Accessibility_Checks restriction is in effect, then the otherwise-legal type conversion (the right-hand side of the assignment to Ptr) becomes a violation of the RM 4.6 rule "The accessibility level of the operand type shall not be statically deeper than that of the target type ...". Subprogram parameters ^^^^^^^^^^^^^^^^^^^^^^ .. code-block:: ada procedure P (V : access T; X : access constant T); In most cases (the exceptions are described below), a No_Dynamic_Accessibility_Checks restriction means that the "statically deeper" relationship does apply to the anonymous type of an access parameter specifying an access-to-object type (contrary to RM 3.10.2(19.1)) and, for purposes of compile-time "statically deeper" checks, the accessibility level of the type of such a parameter is the accessibility level of the parameter. This change (at least as described so far) doesn't affect the caller's side, but on the callee's side it means that object designated by a non-null parameter of an anonymous access type is treated as having the same accessibility level as a local object declared immediately within the called subprogram. With the restriction in effect, the otherwise-legal type conversion in the following example becomes illegal: .. code-block:: ada type Ref is access all Integer; Ptr : Ref; procedure Proc (Param : access Integer) is begin Ptr := Ref (Param); end Proc; The aforementioned exceptions have to do with return statements from functions that either return the given parameter (in the case of a function whose result type is an anonymous access type) or return the given parameter value as an access discriminant of the function result (or of some discriminated part thereof). More specifically, the "statically deeper" changes described above do not apply for purposes of checking the "shall not be statically deeper" rule for access discriminant parts of function results (RM 6.5(5.9)) or in determining the legality of an (implicit) type conversion from the anonymous access type of a parameter of a function to an anonymous access result type of that function. In order to prevent these rule relaxations from introducing the possibility of dynamic accessibility check failures, compensating compile-time checks are performed at the call site to prevent cases where including the value of an access parameter as part of a function result could make such check failures possible (specifically, the discriminant checks of RM 6.5(21) or, in the case of an anonymous access result type, the RM 4.6(48) check performed when converting to that result type). These compile-time checks are described in the next section. From the callee's perspective, the level of anonymous access formal parameters would be between the level of the subprogram and the level of the subprogram's locals. This has the effect of formal parameters being treated as local to the callee except in: * Use as a function result * Use as a value for an access discriminant in result object * Use as an assignments between formal parameters Note that with these more restricted rules we lose track of accessibility levels when assigned to local objects thus making (in the example below) the assignment to Node2.Link from Temp below compile-time illegal. .. code-block:: ada type Node is record Data : Integer; Link : access Node; end record; procedure Swap_Links (Node1, Node2 : in out Node) is Temp : constant access Node := Node1.Link; -- We lose the "association" to Node1 begin Node1.Link := Node2.Link; -- Allowed Node2.Link := Temp; -- Not allowed end; function Identity (N : access Node) return access Node is Local : constant access Node := N; begin if True then return N; -- Allowed else return Local; -- Not allowed end if; end; Function results ^^^^^^^^^^^^^^^^ .. code-block:: ada function Get (X : Rec) return access T; If the result subtype of a function is either an anonymous access (sub)type, a class-wide (sub)type, an unconstrained subtype with an access discriminant, or a type with an unconstrained subcomponent subtype that has at least one access discriminant (this last case is only possible if the access discriminant has a default value), then we say that the function result type "might require an anonymous-access-part accessibility check". If a function has an access parameter, or a parameter whose subtype "might require an anonymous-access-part accessibility check", then we say that the each such parameter "might be used to pass in an anonymous-access value". If the first of these conditions holds for the result subtype of a function and the second condition holds for at least one parameter that function, then it is possible that a call to that function could return a result that contains anonymous-access values that were passed in via the parameter. Given a function call where the result type "might require an anonymous-access-part accessibility check" and a formal parameter of that function that "might be used to pass in an anonymous-access value", either the type of that formal parameter is an anonymous access type or it is not. If it is, and if a No_Dynamic_Access_Checks restriction is in effect, then the accessibility level of the type of the actual parameter shall be statically known to not be deeper than that of the master of the call. If it isn't, then the accessibility level of the actual parameter shall be statically known to not be deeper than that of the master of the call. Function result example: .. code-block:: ada declare type T is record Comp : aliased Integer; end record; function Identity (Param : access Integer) return access Integer is begin return Param; -- Legal end; function Identity_2 (Param : aliased Integer) return access Integer is begin return Param'Access; -- Legal end; X : access Integer; begin X := Identity (X); -- Legal declare Y : access Integer; Z : aliased Integer; begin X := Identity (Y); -- Illegal since Y is too deep X := Identity_2 (Z); -- Illegal since Z is too deep end; end; In order to avoid having to expand the definition of "might be used to pass in an anonymous-access value" to include any parameter of a tagged type, the No_Dynamic_Access_Checks restriction also imposes a requirement that a type extension cannot include the explicit definition of an access discriminant. Here is an example of one such case of an upward conversion which would lead to a memory leak: .. code-block:: ada declare type T is tagged null record; type T2 (Disc : access Integer) is new T with null record; -- Must be illegal function Identity (Param : aliased T'Class) return access Integer is begin return T2 (T'Class (Param)).Disc; -- Here P gets effectively returned and set to X end; X : access Integer; begin declare P : aliased Integer; Y : T2 (P'Access); begin X := Identity (T'Class (Y)); -- Pass local variable P (via Y's discriminant), -- leading to a memory leak. end; end; ``` Thus we need to make the following illegal to avoid such situations: ```ada package Pkg1 is type T1 is tagged null record; function Func (X1 : T1) return access Integer is (null); end; package Pkg2 is type T2 (Ptr1, Ptr2 : access Integer) is new Pkg1.T1 with null record; -- Illegal ... end; In order to prevent upward conversions of anonymous function results (like below), we also would need to assure that the level of such a result (from the callee's perspective) is statically deeper: .. code-block:: ada declare type Ref is access all Integer; Ptr : Ref; function Foo (Param : access Integer) return access Integer is begin return Result : access Integer := Param; do Ptr := Ref (Result); -- Not allowed end return; end; begin declare Local : aliased Integer; begin Foo (Local'Access).all := 123; end; end; Case pattern matching --------------------- The selector for a case statement (but not for a case expression) may be of a composite type, subject to some restrictions (described below). Aggregate syntax is used for choices of such a case statement; however, in cases where a "normal" aggregate would require a discrete value, a discrete subtype may be used instead; box notation can also be used to match all values. Consider this example: .. code-block:: ada type Rec is record F1, F2 : Integer; end record; procedure Caser_1 (X : Rec) is begin case X is when (F1 => Positive, F2 => Positive) => Do_This; when (F1 => Natural, F2 => <>) | (F1 => <>, F2 => Natural) => Do_That; when others => Do_The_Other_Thing; end case; end Caser_1; If ``Caser_1`` is called and both components of X are positive, then ``Do_This`` will be called; otherwise, if either component is nonnegative then ``Do_That`` will be called; otherwise, ``Do_The_Other_Thing`` will be called. In addition, pattern bindings are supported. This is a mechanism for binding a name to a component of a matching value for use within an alternative of a case statement. For a component association that occurs within a case choice, the expression may be followed by ``is ``. In the special case of a "box" component association, the identifier may instead be provided within the box. Either of these indicates that the given identifier denotes (a constant view of) the matching subcomponent of the case selector. .. attention:: Binding is not yet supported for arrays or subcomponents thereof. Consider this example (which uses type ``Rec`` from the previous example): .. code-block:: ada procedure Caser_2 (X : Rec) is begin case X is when (F1 => Positive is Abc, F2 => Positive) => Do_This (Abc) when (F1 => Natural is N1, F2 => ) | (F1 => , F2 => Natural is N1) => Do_That (Param_1 => N1, Param_2 => N2); when others => Do_The_Other_Thing; end case; end Caser_2; This example is the same as the previous one with respect to determining whether ``Do_This``, ``Do_That``, or ``Do_The_Other_Thing`` will be called. But for this version, ``Do_This`` takes a parameter and ``Do_That`` takes two parameters. If ``Do_This`` is called, the actual parameter in the call will be ``X.F1``. If ``Do_That`` is called, the situation is more complex because there are two choices for that alternative. If ``Do_That`` is called because the first choice matched (i.e., because ``X.F1`` is nonnegative and either ``X.F1`` or ``X.F2`` is zero or negative), then the actual parameters of the call will be (in order) ``X.F1`` and ``X.F2``. If ``Do_That`` is called because the second choice matched (and the first one did not), then the actual parameters will be reversed. Within the choice list for single alternative, each choice must define the same set of bindings and the component subtypes for for a given identifier must all statically match. Currently, the case of a binding for a nondiscrete component is not implemented. If the set of values that match the choice(s) of an earlier alternative overlaps the corresponding set of a later alternative, then the first set shall be a proper subset of the second (and the later alternative will not be executed if the earlier alternative "matches"). All possible values of the composite type shall be covered. The composite type of the selector shall be an array or record type that is neither limited nor class-wide. Currently, a "when others =>" case choice is required; it is intended that this requirement will be relaxed at some point. If a subcomponent's subtype does not meet certain restrictions, then the only value that can be specified for that subcomponent in a case choice expression is a "box" component association (which matches all possible values for the subcomponent). This restriction applies if: - the component subtype is not a record, array, or discrete type; or - the component subtype is subject to a non-static constraint or has a predicate; or: - the component type is an enumeration type that is subject to an enumeration representation clause; or - the component type is a multidimensional array type or an array type with a nonstatic index subtype. Support for casing on arrays (and on records that contain arrays) is currently subject to some restrictions. Non-positional array aggregates are not supported as (or within) case choices. Likewise for array type and subtype names. The current implementation exceeds compile-time capacity limits in some annoyingly common scenarios; the message generated in such cases is usually "Capacity exceeded in compiling case statement with composite selector type". Mutably Tagged Types with Size'Class Aspect ------------------------------------------- For a specific tagged nonformal type T that satisfies some conditions described later in this section, the universal-integer-valued type-related representation aspect ``Size'Class`` may be specified; any such specified aspect value shall be static. Specifying this aspect imposes an upper bound on the sizes of all specific descendants of T (including T itself). T'Class (but not T) is then said to be a "mutably tagged" type - meaning that T'Class is a definite subtype and that the tag of a variable of type T'Class may be modified by assignment in some cases described later in this section. An inherited ``Size'Class`` aspect value may be overridden, but not with a larger value. If the ``Size'Class`` aspect is specified for a type T, then every specific descendant of T (including T itself) * shall have a Size that does not exceed the specified value; and * shall have a (possibly inherited) ``Size'Class`` aspect that does not exceed the specified value; and * shall be undiscriminated; and * shall have no composite subcomponent whose subtype is subject to a nonstatic constraint; and * shall not have a tagged partial view other than a private extension; and * shall not be a descendant of an interface type; and * shall not have a statically deeper accessibility level than that of T. If the ``Size'Class`` aspect is not specified for a type T (either explicitly or by inheritance), then it shall not be specified for any descendant of T. Example: .. code-block:: ada type Root_Type is tagged null record with Size'Class => 16 * 8; type Derived_Type is new Root_Type with record Stuff : Some_Type; end record; -- ERROR if Derived_Type exceeds 16 bytes Because any subtype of a mutably tagged type is definite, it can be used as a component subtype for enclosing array or record types, as the subtype of a default-initialized stand-alone object, or as the subtype of an uninitialized allocator, as in this example: .. code-block:: ada Obj : Root_Type'Class; type Array_of_Roots is array (Positive range <>) of Root_Type'Class; Default initialization of an object of such a definite subtype proceeds as for the corresponding specific type, except that Program_Error is raised if the specific type is abstract. In particular, the initial tag of the object is that of the corresponding specific type. There is a general design principle that if a type has a tagged partial view, then the type's ``Size'Class`` aspect (or lack thereof) should be determinable by looking only at the partial view. That provides the motivation for the rules of the next two paragraphs. If a type has a tagged partial view, then a ``Size'Class`` aspect specification may be provided only at the point of the partial view declaration (in other words, no such aspect specification may be provided when the full view of the type is declared). All of the above rules (in particular, the rule that an overriding ``Size'Class`` aspect value shall not be larger than the overridden inherited value) are also enforced when the full view (which may have a different ancestor type than that of the partial view) is declared. If a partial view for a type inherits a ``Size'Class`` aspect value and does not override that value with an explicit aspect specification, then the (static) aspect values inherited by the partial view and by the full view shall be equal. An actual parameter of an instantiation whose corresponding formal parameter is a formal tagged private type shall not be either mutably tagged or the corresponding specific type of a mutably tagged type. For the legality rules in this section, the RM 12.3(11) rule about legality checking in the visible part and formal part of an instance is extended (in the same way that it is extended in many other places in the RM) to include the private part of an instance. An object (or a view thereof) of a tagged type is defined to be "tag-constrained" if it is * an object whose type is not mutably tagged; or * a constant object; or * a view conversion of a tag-constrained object; or * a view conversion to a type that is not a descendant of the operand's type; or * a formal in out or out parameter whose corresponding actual parameter is tag-constrained; or * a dereference of an access value. In the case of an assignment to a tagged variable that is not tag-constrained, no check is performed that the tag of the value of the expression is the same as that of the target (RM 5.2 notwithstanding). Instead, the tag of the target object becomes that of the source object of the assignment. Note that the tag of an object of a mutably tagged type MT will always be the tag of some specific type that is a descendant of MT. An assignment to a composite object similarly copies the tags of any subcomponents of the source object that have a mutably tagged type. The Constrained attribute is defined for any name denoting an object of a mutably tagged type (RM 3.7.2 notwithstanding). In this case, the Constrained attribute yields the value True if the object is tag-constrained and False otherwise. Renaming is not allowed (see RM 8.5.1) for a type conversion having an operand of a mutably tagged type MT and a target type TT such that TT (or its corresponding specific type if TT is class-wide) is not an ancestor of MT (this is sometimes called a "downward" conversion), nor for any part of such an object, nor for any slice of any part of such an object. This rule also applies in any context where a name is required to be one for which "renaming is allowed" (for example, see RM 12.4). [This is analogous to the way that renaming is not allowed for a discriminant-dependent component of an unconstrained variable.] A name denoting a view of a variable of a mutably tagged type shall not occur as an operative constituent of the prefix of a name denoting a prefixed view of a callable entity, except as the callee name in a call to the callable entity. This disallows, for example, renaming such a prefixed view, passing the prefixed view name as a generic actual parameter, or using the prefixed view name as the prefix of an attribute. The execution of a construct is erroneous if the construct has a constituent that is a name denoting a subcomponent of a tagged object and the object's tag is changed by this execution between evaluating the name and the last use (within this execution) of the subcomponent denoted by the name. This is analogous to the RM 3.7.2(4) rule about discriminant-dependent subcomponents. If the type of a formal parameter is a specific tagged type, then the execution of the call is erroneous if the tag of the actual is changed while the formal parameter exists (that is, before leaving the corresponding callable construct). This is analogous to the RM 6.4.1(18) rule about discriminated parameters. .. _No_Raise_Aspect: No_Raise aspect ---------------- The ``No_Raise`` aspect can be applied to a subprogram to declare that this subprogram is not expected to raise an exception. Should an exception still be raised during the execution of the subprogram, it is caught at the end of this execution and ``Program_Error`` is propagated to the caller. Inference of Dependent Types in Generic Instantiations ------------------------------------------------------ If a generic formal type T2 depends on another formal type T1, the actual for T1 can be inferred from the actual for T2. That is, you can give the actual for T2, and leave out the one for T1. For example, ``Ada.Unchecked_Deallocation`` has two generic formals: .. code-block:: ada generic type Object (<>) is limited private; type Name is access Object; procedure Ada.Unchecked_Deallocation (X : in out Name); where ``Name`` depends on ``Object``. With this language extension, you can leave out the actual for ``Object``, as in: .. code-block:: ada type Integer_Access is access all Integer; procedure Free is new Unchecked_Deallocation (Name => Integer_Access); The compiler will infer that the actual type for ``Object`` is ``Integer``. Note that named notation is always required when using inference. The following inferences are allowed: - For a formal access type, the designated type can be inferred. - For a formal array type, the index type(s) and the component type can be inferred. - For a formal type with discriminants, the type(s) of the discriminants can be inferred. Example for arrays: .. code-block:: ada generic type Element_Type is private; type Index_Type is (<>); type Array_Type is array (Index_Type range <>) of Element_Type; package Array_Operations is ... end Array_Operations; ... type Int_Array is array (Positive range <>) of Integer; package Int_Array_Operations is new Array_Operations (Array_Type => Int_Array); The index and component types of ``Array_Type`` are inferred from ``Int_Array``, so that the above instantiation is equivalent to the following standard-Ada instantiation: .. code-block:: ada package Int_Array_Operations is new Array_Operations (Element_Type => Integer, Index_Type => Positive, Array_Type => Int_Array); External_Initialization Aspect ------------------------------ The ``External_Initialization`` aspect provides a feature similar to Rust's ``include_bytes!`` and to C23's ``#embed``. It has the effect of initializing an object with the contents of a file specified by a file path. Only string objects and objects of type ``Ada.Streams.Stream_Element_Array`` can be subject to the ``External_Initialization`` aspect. Example: .. code-block:: ada with Ada.Streams; package P is S : constant String with External_Initialization => "foo.txt"; X : constant Ada.Streams.Stream_Element_Array with External_Initialization => "bar.bin"; end P; ``External_Initialization`` aspect accepts the following parameters: - mandatory ``Path``: the path the compiler uses to access the binary resource. If ``Path`` is a relative path, it is interpreted relatively to the directory of the file that contains the aspect specification. .. attention:: The maximum size of loaded files is limited to 2\ :sup:`31` bytes. Finally construct ----------------- The ``finally`` keyword makes it possible to have a sequence of statements be executed when another sequence of statements is completed, whether normally or abnormally. This feature is similar to the one with the same name in other languages such as Java. Note that ``finally`` is a keyword but it is not a reserved word. This is a configuration that does not exist in standard Ada. Syntax ^^^^^^ .. code-block:: text handled_sequence_of_statements ::= sequence_of_statements [exception exception_handler {exception_handler}] [finally sequence_of_statements] Legality Rules ^^^^^^^^^^^^^^ Return statements in the ``sequence_of_statements`` attached to the finally that would cause control to be transferred outside the finally part are forbidden. Goto & exit where the target is outside of the finally's ``sequence_of_statements`` are forbidden Dynamic Semantics ^^^^^^^^^^^^^^^^^ Statements in the optional ``sequence_of_statements`` contained in the ``finally`` branch will be executed unconditionally, after the main ``sequence_of_statements`` is executed, and after any potential ``exception_handler`` is executed. If an exception is raised in the finally part, it cannot be caught by the ``exception_handler``. Abort/ATC (asynchronous transfer of control) cannot interrupt a finally block, nor prevent its execution, that is the finally block must be executed in full even if the containing task is aborted, or if the control is transferred out of the block. Continue statement ------------------ The ``continue`` keyword makes it possible to stop execution of a loop iteration and continue with the next one. A continue statement has the same syntax (except "exit" is replaced with "continue"), static semantics, and legality rules as an exit statement. The difference is in the dynamic semantics: where an exit statement would cause a transfer of control that completes the (implicitly or explicitly) specified loop_statement, a continue statement would instead cause a transfer of control that completes only the current iteration of that loop_statement, like a goto statement targeting a label following the last statement in the sequence of statements of the specified loop_statement. Note that ``continue`` is a keyword but it is not a reserved word. This is a configuration that does not exist in standard Ada. Destructors ----------- The :ada:`Destructor` extension adds a new finalization mechanism that significantly differs from standard Ada in how it interacts with type derivation. New syntax is introduced to make it possible to define "destructors" for record types, tagged or untagged. Here's a simple example: .. code-block:: ada package P is type T is record ... end record; procedure T'Destructor (X : in out T); end P; .. code-block:: ada package body P is procedure T'Destructor (X : in out T) is begin ... end T'Destructor; end P; Like :ada:`Finalize` procedures, destructors are called on objects just before they are destroyed. But destructors are more flexible in how they can used with derived types. With standard Ada finalization, when you derive from a finalizable type, you must either inherit the :ada:`Finalize` procedure or override it completely. Destructors work differently. You can define a destructor for a type derived from a parent type that also has a destructor, and then when objects of the derived type are finalized, both destructors will be called. For example: .. code-block:: ada type T1 is record ... end record; procedure T1'Destructor (X : in out T1); type T2 is new T1; procedure T2'Destructor (X : in out T2); When an object of type :ada:`T2` is finalized, there will be first a call to :ada:`T2'Destructor`, and then a call to :ada:`T1'Destructor` on the object. Structural Generic Instantiation -------------------------------- The compiler implements a second kind of generic instantiation, called "structural", alongside the traditional instantiation specified by the language, which is defined as follows: the structural instantiation of a generic unit on given actual parameters is the anonymous instantiation of the generic unit on the actual parameters done in the outermost scope where it would be legal to do an identical traditional instantiation. There is at most one structural instantiation of a generic unit on given actual parameters done in a partition. Structural generic instances (the product of structural instantiation) are implicitly created whenever a reference to them is made in a place where a name is accepted by the language. Syntax ^^^^^^ .. code-block:: text name ::= { set of productions specified in the RM } | structural_generic_instance_name structural_generic_instance_name ::= name generic_actual_part Legality Rules ^^^^^^^^^^^^^^ The ``name`` in a ``structural_generic_instance_name`` shall denote a generic unit that is preelaborated. Note that, unlike in a traditional instantiation, there are no square brackets around the ``generic_actual_part`` in the second production, which means that it is mandatory and, therefore, that the generic unit shall have at least one generic formal parameter. The generic unit shall not take a generic formal object of mode ``in out``. If the generic unit takes a generic formal object of mode ``in``, then the corresponding generic actual parameter shall be a static expression. A ``structural_generic_instance_name`` for a generic package shall not be present in a library unit if the structural instance is also a library unit and has a semantic dependence on the former. For a generic subprogram, if a local entity of the enclosing library-level package is used as an actual and the structural instance would have a semantic dependence on the package, the structural instantiation is automatically demoted to a local instantiation. In this case, several instances of the generic subprogram may be present in a single partition, unless whole-partition optimization is performed (e.g., via LTO). Static Semantics ^^^^^^^^^^^^^^^^ A ``structural_generic_instance_name`` denotes the instance that is the product of the structural instantiation of a generic unit on the specified actual parameters. This instance is unique to a partition, except when a generic subprogram instantiation is automatically demoted to a local instantiation as described under Legality Rules. Example: .. code-block:: ada with Ada.Containers.Vectors; procedure P is V : Ada.Containers.Vectors(Positive,Integer).Vector; begin V.Append (1); V.Append (0); Ada.Containers.Vectors(Positive,Integer).Generic_Sorting("<").Sort (V); end; This procedure references two structural instantiations of two different generic units: ``Ada.Containers.Vectors(Positive,Integer)`` is the structural instance of the generic unit ``Ada.Containers.Vectors`` on ``Positive`` and ``Integer`` and ``Ada.Containers.Vectors(Positive,Integer).Generic_Sorting("<")`` is the structural instance of the nested generic unit ``Ada.Containers.Vectors(Positive,Integer).Generic_Sorting`` on ``"<"``. Note that the following example is illegal: .. code-block:: ada with Ada.Containers.Vectors; package Q is type T is record I : Integer; end record; V : Ada.Containers.Vectors(Positive,T).Vector; end Q; The reason is that ``Ada.Containers.Vectors``, ``Positive`` and ``Q.T`` being library-level entities, the structural instance ``Ada.Containers.Vectors(Positive,T)`` is a library unit with a dependence on ``Q`` and, therefore, cannot be referenced from within ``Q``. This restriction applies to structural instantiations of generic packages. The simple way out is to declare a traditional instantiation in this case: .. code-block:: ada with Ada.Containers.Vectors; package Q is type T is record I : Integer; end record; package Vectors_Of_T is new Ada.Containers.Vectors(Positive,T); V : Vectors_Of_T.Vector; end Q; But the following example is legal: .. code-block:: ada with Ada.Containers.Vectors; procedure P is type T is record I : Integer; end record; V : Ada.Containers.Vectors(Positive,T).Vector; end; because the structural instance ``Ada.Containers.Vectors(Positive,T)`` is not a library unit. For generic subprograms, the restriction does not apply: if a local entity of a library-level package is used as an actual, the structural instantiation is automatically demoted to a local instantiation. For example: .. code-block:: ada with Ada.Unchecked_Deallocation; package Q is type T is record I : Integer; end record; type T_Access is access T; end Q; package body Q is procedure Free_T (X : in out T_Access) is begin Ada.Unchecked_Deallocation(T, T_Access)(X); end Free_T; end Q; is legal: since ``T`` and ``T_Access`` are local entities of ``Q``, the structural instantiation of ``Ada.Unchecked_Deallocation`` is demoted to a local instantiation rather than producing an error. Note that the uniqueness guarantee no longer holds in this case and several instances of the generic subprogram may be present in a single partition. The first example can be rewritten in a less verbose manner: .. code-block:: ada with Ada.Containers.Vectors; use Ada.Containers.Vectors(Positive,Integer); procedure P is V : Vector; begin V.Append (1); V.Append (0); Generic_Sorting("<").Sort (V); end; Another example, which additionally uses the inference of dependent types: .. code-block:: ada with Ada.Unchecked_Deallocation; procedure P is type Integer_Access is access all Integer; A : Integer_Access := new Integer'(1); begin Ada.Unchecked_Deallocation(Name => Integer_Access) (A); end; Attribute Subprograms --------------------- This feature allows declaring subprograms that are associated with a named aspect or attribute of a given type, as an alternative to explicitly specifying the aspect or attribute for the type (using an aspect_specification or attribute_definition_clause), by defining the name of the subprogram using attribute syntax. In effect, this implicitly specifies the aspect or attribute for the type, and is intended to be semantically equivalent to using an explicit aspect specification, but without needing to give a specific name in the aspect and subprogram. This is allowed for the various aspects that denote subprograms (see below for the exact list of supported aspects). Here is an example of the use of this feature for defining indexing functions for a container type and an integer-literal function for the container's element type: .. code-block:: ada package My_Container is type Container is tagged ...; -- Implicit indexing aspects type Element_Type is ...; -- Implicit Integer_Literal aspect type Cursor_Type is ...; type Reference_Type is ...; function Container'Constant_Indexing (C : Container; Index : Positive) return Element_Type; function Container'Constant_Indexing (C : Container; Index : Cursor_Type) return Element_Type; function Container'Variable_Indexing (C : in out Container; Index : Positive) return Reference_Type; function Element_Type'Integer_Literal (S : String) return Element_Type; end My_Container; Syntax ^^^^^^ .. code-block:: text defining_program_unit_name ::= [parent_unit_name . ]defining_identifier | defining_attribute_subprogram_name defining_attribute_subprogram_name ::= prefix'attribute_designator Legality Rules ^^^^^^^^^^^^^^ The prefix of a ``defining_attribute_subprogram_name`` shall be either a ``direct_name`` denoting a type (more properly, a first subtype) declared immediately within the scope where the associated subprogram is declared, or else an attribute name itself whose ``attribute_designator`` is ``Class`` and whose prefix denotes a type subject to the same declaration condition. A ``defining_program_unit_name`` that is a ``defining_attribute_subprogram_name`` may only be used to define the name of a subprogram that is neither a library subprogram nor a subprogram whose body is given by a ``body_stub`` (nor can it be used to define the name of a generic subprogram). The ``attribute_designator`` given in a ``defining_attribute_subprogram_name`` shall name an aspect that is allowed to designate a subprogram. Specifically, it shall be one of the predefined aspects ``Constant_Indexing``, ``Variable_Indexing``, ``Default_Iterator``, ``Integer_Literal``, ``Real_Literal``, ``String_Literal``, ``Put_Image``, ``Read``, ``Write``, ``Input``, or ``Output``, or else one of the implementation-defined aspects ``Constructor`` or ``Destructor``. (Note that this feature is not supported for aspects that have elements that may designate a subprogram, such as the ``Aggregate`` aspect.) If the prefix of a ``defining_attribute_subprogram_name`` is an attribute name whose ``attribute_designator`` is ``Class``, then the ``attribute_designator`` of the ``defining_attribute_subprogram_name`` shall be the name of a stream attribute. An attribute subprogram for a given type and aspect is subject to the same restrictions as would be imposed on a subprogram denoted by an explicit ``aspect_specification`` or ``attribute_definition_clause`` for that type and aspect. In particular, the profile of the subprogram must satisfy any expected profile defined for the aspect. If a subprogram is declared using a ``defining_attribute_subprogram_name``, then the type named in the attribute subprogram's name shall not have an explicit ``aspect_specification`` that specifies the same aspect as that named by the subprogram, nor shall the type derive from an ancestor type that explicitly specifies that aspect. Similarly, an explicit ``aspect_specification`` shall not specify an aspect for a type that is derived from a type for which any attribute subprograms have been declared for the same aspect. --- ### Gcc Ada Doc Gnat Rm Implementation Advice (`gcc/ada/doc/gnat_rm/implementation_advice.rst`) .. _Implementation_Advice: ********************* Implementation Advice ********************* The main text of the Ada Reference Manual describes the required behavior of all Ada compilers, and the GNAT compiler conforms to these requirements. In addition, there are sections throughout the Ada Reference Manual headed by the phrase 'Implementation advice'. These sections are not normative, i.e., they do not specify requirements that all compilers must follow. Rather they provide advice on generally desirable behavior. They are not requirements, because they describe behavior that cannot be provided on all systems, or may be undesirable on some systems. As far as practical, GNAT follows the implementation advice in the Ada Reference Manual. Each such RM section corresponds to a section in this chapter whose title specifies the RM section number and paragraph number and the subject of the advice. The contents of each section consists of the RM text within quotation marks, followed by the GNAT interpretation of the advice. Most often, this simply says 'followed', which means that GNAT follows the advice. However, in a number of cases, GNAT deliberately deviates from this advice, in which case the text describes what GNAT does and why. .. index:: Error detection RM 1.1.3(20): Error Detection ============================= "If an implementation detects the use of an unsupported Specialized Needs Annex feature at run time, it should raise ``Program_Error`` if feasible." Not relevant. All specialized needs annex features are either supported, or diagnosed at compile time. .. index:: Child Units RM 1.1.3(31): Child Units ========================= "If an implementation wishes to provide implementation-defined extensions to the functionality of a language-defined library unit, it should normally do so by adding children to the library unit." Followed. .. index:: Bounded errors RM 1.1.5(12): Bounded Errors ============================ "If an implementation detects a bounded error or erroneous execution, it should raise ``Program_Error``." Followed in all cases in which the implementation detects a bounded error or erroneous execution. Not all such situations are detected at runtime. .. index:: Pragmas .. _RM_2_8_16_Pragmas: RM 2.8(16): Pragmas =================== "Normally, implementation-defined pragmas should have no semantic effect for error-free programs; that is, if the implementation-defined pragmas are removed from a working program, the program should still be legal, and should still have the same semantics." The following implementation defined pragmas are exceptions to this rule: +--------------------+-------------------+ | Pragma | Explanation | +====================+===================+ | *Abort_Defer* | Affects semantics | +--------------------+-------------------+ |*Ada_83* | Affects legality | +--------------------+-------------------+ |*Assert* | Affects semantics | +--------------------+-------------------+ |*CPP_Class* | Affects semantics | +--------------------+-------------------+ |*CPP_Constructor* | Affects semantics | +--------------------+-------------------+ |*Debug* | Affects semantics | +--------------------+-------------------+ |*Interface_Name* | Affects semantics | +--------------------+-------------------+ |*Machine_Attribute* | Affects semantics | +--------------------+-------------------+ |*Unimplemented_Unit*| Affects legality | +--------------------+-------------------+ |*Unchecked_Union* | Affects semantics | +--------------------+-------------------+ In each of the above cases, it is essential to the purpose of the pragma that this advice not be followed. For details see :ref:`Implementation_Defined_Pragmas`. RM 2.8(17-19): Pragmas ====================== "Normally, an implementation should not define pragmas that can make an illegal program legal, except as follows: * A pragma used to complete a declaration, such as a pragma ``Import``; * A pragma used to configure the environment by adding, removing, or replacing ``library_items``." See :ref:`RM_2_8_16_Pragmas`. .. index:: Character Sets .. index:: Alternative Character Sets RM 3.5.2(5): Alternative Character Sets ======================================= "If an implementation supports a mode with alternative interpretations for ``Character`` and ``Wide_Character``, the set of graphic characters of ``Character`` should nevertheless remain a proper subset of the set of graphic characters of ``Wide_Character``. Any character set 'localizations' should be reflected in the results of the subprograms defined in the language-defined package ``Characters.Handling`` (see A.3) available in such a mode. In a mode with an alternative interpretation of ``Character``, the implementation should also support a corresponding change in what is a legal ``identifier_letter``." Not all wide character modes follow this advice, in particular the JIS and IEC modes reflect standard usage in Japan, and in these encoding, the upper half of the Latin-1 set is not part of the wide-character subset, since the most significant bit is used for wide character encoding. However, this only applies to the external forms. Internally there is no such restriction. .. index:: Integer types RM 3.5.4(28): Integer Types =========================== "An implementation should support ``Long_Integer`` in addition to ``Integer`` if the target machine supports 32-bit (or longer) arithmetic. No other named integer subtypes are recommended for package ``Standard``. Instead, appropriate named integer subtypes should be provided in the library package ``Interfaces`` (see B.2)." ``Long_Integer`` is supported. Other standard integer types are supported so this advice is not fully followed. These types are supported for convenient interface to C, and so that all hardware types of the machine are easily available. RM 3.5.4(29): Integer Types =========================== "An implementation for a two's complement machine should support modular types with a binary modulus up to ``System.Max_Int*2+2``. An implementation should support a non-binary modules up to ``Integer'Last``." Followed. .. index:: Enumeration values RM 3.5.5(8): Enumeration Values =============================== "For the evaluation of a call on ``S'Pos`` for an enumeration subtype, if the value of the operand does not correspond to the internal code for any enumeration literal of its type (perhaps due to an un-initialized variable), then the implementation should raise ``Program_Error``. This is particularly important for enumeration types with noncontiguous internal codes specified by an enumeration_representation_clause." Followed. .. index:: Float types RM 3.5.7(17): Float Types ========================= "An implementation should support ``Long_Float`` in addition to ``Float`` if the target machine supports 11 or more digits of precision. No other named floating point subtypes are recommended for package ``Standard``. Instead, appropriate named floating point subtypes should be provided in the library package ``Interfaces`` (see B.2)." ``Short_Float`` and ``Long_Long_Float`` are also provided. The former provides improved compatibility with other implementations supporting this type. The latter corresponds to the highest precision floating-point type supported by the hardware. On most machines, this will be the same as ``Long_Float``, but on some machines, it will correspond to the IEEE extended form. The notable case is all x86 implementations, where ``Long_Long_Float`` corresponds to the 80-bit extended precision format supported in hardware on this processor. Note that the 128-bit format on SPARC is not supported, since this is a software rather than a hardware format. .. index:: Multidimensional arrays .. index:: Arrays, multidimensional RM 3.6.2(11): Multidimensional Arrays ===================================== "An implementation should normally represent multidimensional arrays in row-major order, consistent with the notation used for multidimensional array aggregates (see 4.3.3). However, if a pragma ``Convention`` (``Fortran``, ...) applies to a multidimensional array type, then column-major order should be used instead (see B.5, *Interfacing with Fortran*)." Followed. .. index:: Duration'Small RM 9.6(30-31): Duration'Small ============================= "Whenever possible in an implementation, the value of ``Duration'Small`` should be no greater than 100 microseconds." Followed. (``Duration'Small`` = 10**(-9)). "The time base for ``delay_relative_statements`` should be monotonic; it need not be the same time base as used for ``Calendar.Clock``." Followed. RM 10.2.1(12): Consistent Representation ======================================== "In an implementation, a type declared in a pre-elaborated package should have the same representation in every elaboration of a given version of the package, whether the elaborations occur in distinct executions of the same program, or in executions of distinct programs or partitions that include the given version." Followed, except in the case of tagged types. Tagged types involve implicit pointers to a local copy of a dispatch table, and these pointers have representations which thus depend on a particular elaboration of the package. It is not easy to see how it would be possible to follow this advice without severely impacting efficiency of execution. .. index:: Exception information RM 11.4.1(19): Exception Information ==================================== "``Exception_Message`` by default and ``Exception_Information`` should produce information useful for debugging. ``Exception_Message`` should be short, about one line. ``Exception_Information`` can be long. ``Exception_Message`` should not include the ``Exception_Name``. ``Exception_Information`` should include both the ``Exception_Name`` and the ``Exception_Message``." Followed. For each exception that doesn't have a specified ``Exception_Message``, the compiler generates one containing the location of the raise statement. This location has the form 'file_name:line', where file_name is the short file name (without path information) and line is the line number in the file. Note that in the case of the Zero Cost Exception mechanism, these messages become redundant with the Exception_Information that contains a full backtrace of the calling sequence, so they are disabled. To disable explicitly the generation of the source location message, use the Pragma ``Discard_Names``. .. index:: Suppression of checks .. index:: Checks, suppression of RM 11.5(28): Suppression of Checks ================================== "The implementation should minimize the code executed for checks that have been suppressed." Followed. .. index:: Representation clauses RM 13.1 (21-24): Representation Clauses ======================================= "The recommended level of support for all representation items is qualified as follows: An implementation need not support representation items containing nonstatic expressions, except that an implementation should support a representation item for a given entity if each nonstatic expression in the representation item is a name that statically denotes a constant declared before the entity." Followed. In fact, GNAT goes beyond the recommended level of support by allowing nonstatic expressions in some representation clauses even without the need to declare constants initialized with the values of such expressions. For example: .. code-block:: ada X : Integer; Y : Float; for Y'Address use X'Address; is accepted directly by GNAT. "An implementation need not support a specification for the ``Size`` for a given composite subtype, nor the size or storage place for an object (including a component) of a given composite subtype, unless the constraints on the subtype and its composite subcomponents (if any) are all static constraints." Followed. Size Clauses are not permitted on nonstatic components, as described above. "An aliased component, or a component whose type is by-reference, should always be allocated at an addressable location." Followed. .. index:: Packed types RM 13.2(6-8): Packed Types ========================== "If a type is packed, then the implementation should try to minimize storage allocated to objects of the type, possibly at the expense of speed of accessing components, subject to reasonable complexity in addressing calculations. The recommended level of support pragma ``Pack`` is: For a packed record type, the components should be packed as tightly as possible subject to the Sizes of the component subtypes, and subject to any *record_representation_clause* that applies to the type; the implementation may, but need not, reorder components or cross aligned word boundaries to improve the packing. A component whose ``Size`` is greater than the word size may be allocated an integral number of words." Followed. Tight packing of arrays is supported for all component sizes up to 64-bits. If the array component size is 1 (that is to say, if the component is a boolean type or an enumeration type with two values) then values of the type are implicitly initialized to zero. This happens both for objects of the packed type, and for objects that have a subcomponent of the packed type. .. index:: Address clauses RM 13.3(14-19): Address Clauses =============================== "For an array ``X``, ``X'Address`` should point at the first component of the array, and not at the array bounds." Followed. "The recommended level of support for the ``Address`` attribute is: ``X'Address`` should produce a useful result if ``X`` is an object that is aliased or of a by-reference type, or is an entity whose ``Address`` has been specified." Followed. A valid address will be produced even if none of those conditions have been met. If necessary, the object is forced into memory to ensure the address is valid. "An implementation should support ``Address`` clauses for imported subprograms." Followed. "Objects (including subcomponents) that are aliased or of a by-reference type should be allocated on storage element boundaries." Followed. "If the ``Address`` of an object is specified, or it is imported or exported, then the implementation should not perform optimizations based on assumptions of no aliases." Followed. .. index:: Alignment clauses RM 13.3(29-35): Alignment Clauses ================================= "The recommended level of support for the ``Alignment`` attribute for subtypes is: An implementation should support specified Alignments that are factors and multiples of the number of storage elements per word, subject to the following:" Followed. "An implementation need not support specified Alignments for combinations of Sizes and Alignments that cannot be easily loaded and stored by available machine instructions." Followed. "An implementation need not support specified Alignments that are greater than the maximum ``Alignment`` the implementation ever returns by default." Followed. "The recommended level of support for the ``Alignment`` attribute for objects is: Same as above, for subtypes, but in addition:" Followed. "For stand-alone library-level objects of statically constrained subtypes, the implementation should support all alignments supported by the target linker. For example, page alignment is likely to be supported for such objects, but not for subtypes." Followed. .. index:: Size clauses RM 13.3(42-43): Size Clauses ============================ "The recommended level of support for the ``Size`` attribute of objects is: A ``Size`` clause should be supported for an object if the specified ``Size`` is at least as large as its subtype's ``Size``, and corresponds to a size in storage elements that is a multiple of the object's ``Alignment`` (if the ``Alignment`` is nonzero)." Followed. RM 13.3(50-56): Size Clauses ============================ "If the ``Size`` of a subtype is specified, and allows for efficient independent addressability (see 9.10) on the target architecture, then the ``Size`` of the following objects of the subtype should equal the ``Size`` of the subtype: Aliased objects (including components)." Followed. "`Size` clause on a composite subtype should not affect the internal layout of components." Followed. But note that this can be overridden by use of the implementation pragma Implicit_Packing in the case of packed arrays. "The recommended level of support for the ``Size`` attribute of subtypes is: The ``Size`` (if not specified) of a static discrete or fixed point subtype should be the number of bits needed to represent each value belonging to the subtype using an unbiased representation, leaving space for a sign bit only if the subtype contains negative values. If such a subtype is a first subtype, then an implementation should support a specified ``Size`` for it that reflects this representation." Followed. "For a subtype implemented with levels of indirection, the ``Size`` should include the size of the pointers, but not the size of what they point at." Followed. .. index:: Component_Size clauses RM 13.3(71-73): Component Size Clauses ====================================== "The recommended level of support for the ``Component_Size`` attribute is: An implementation need not support specified ``Component_Sizes`` that are less than the ``Size`` of the component subtype." Followed. "An implementation should support specified Component_Sizes that are factors and multiples of the word size. For such Component_Sizes, the array should contain no gaps between components. For other Component_Sizes (if supported), the array should contain no gaps between components when packing is also specified; the implementation should forbid this combination in cases where it cannot support a no-gaps representation." Followed. .. index:: Enumeration representation clauses .. index:: Representation clauses, enumeration RM 13.4(9-10): Enumeration Representation Clauses ================================================= "The recommended level of support for enumeration representation clauses is: An implementation need not support enumeration representation clauses for boolean types, but should at minimum support the internal codes in the range ``System.Min_Int .. System.Max_Int``." Followed. .. index:: Record representation clauses .. index:: Representation clauses, records RM 13.5.1(17-22): Record Representation Clauses =============================================== "The recommended level of support for *record_representation_clause*\ s is: An implementation should support storage places that can be extracted with a load, mask, shift sequence of machine code, and set with a load, shift, mask, store sequence, given the available machine instructions and run-time model." Followed. "A storage place should be supported if its size is equal to the ``Size`` of the component subtype, and it starts and ends on a boundary that obeys the ``Alignment`` of the component subtype." Followed. "If the default bit ordering applies to the declaration of a given type, then for a component whose subtype's ``Size`` is less than the word size, any storage place that does not cross an aligned word boundary should be supported." Followed. "An implementation may reserve a storage place for the tag field of a tagged type, and disallow other components from overlapping that place." Followed. The storage place for the tag field is the beginning of the tagged record, and its size is Address'Size. GNAT will reject an explicit component clause for the tag field. "An implementation need not support a *component_clause* for a component of an extension part if the storage place is not after the storage places of all components of the parent type, whether or not those storage places had been specified." Followed. The above advice on record representation clauses is followed, and all mentioned features are implemented. .. index:: Storage place attributes RM 13.5.2(5): Storage Place Attributes ====================================== "If a component is represented using some form of pointer (such as an offset) to the actual data of the component, and this data is contiguous with the rest of the object, then the storage place attributes should reflect the place of the actual data, not the pointer. If a component is allocated discontinuously from the rest of the object, then a warning should be generated upon reference to one of its storage place attributes." Followed. There are no such components in GNAT. .. index:: Bit ordering RM 13.5.3(7-8): Bit Ordering ============================ "The recommended level of support for the non-default bit ordering is: The implementation should support the nondefault bit ordering in addition to the default bit ordering." Followed. .. index:: Address, as private type RM 13.7(37): Address as Private =============================== "`Address` should be of a private type." Followed. .. index:: Operations, on ``Address`` .. index:: Address, operations of RM 13.7.1(16): Address Operations ================================= "Operations in ``System`` and its children should reflect the target environment semantics as closely as is reasonable. For example, on most machines, it makes sense for address arithmetic to 'wrap around'. Operations that do not make sense should raise ``Program_Error``." Followed. Address arithmetic is modular arithmetic that wraps around. No operation raises ``Program_Error``, since all operations make sense. .. index:: Unchecked conversion RM 13.9(14-17): Unchecked Conversion ==================================== "The ``Size`` of an array object should not include its bounds; hence, the bounds should not be part of the converted data." Followed. "The implementation should not generate unnecessary run-time checks to ensure that the representation of ``S`` is a representation of the target type. It should take advantage of the permission to return by reference when possible. Restrictions on unchecked conversions should be avoided unless required by the target environment." Followed. There are no restrictions on unchecked conversion. A warning is generated if the source and target types do not have the same size since the semantics in this case may be target dependent. "The recommended level of support for unchecked conversions is: Unchecked conversions should be supported and should be reversible in the cases where this clause defines the result. To enable meaningful use of unchecked conversion, a contiguous representation should be used for elementary subtypes, for statically constrained array subtypes whose component subtype is one of the subtypes described in this paragraph, and for record subtypes without discriminants whose component subtypes are described in this paragraph." Followed. .. index:: Heap usage, implicit RM 13.11(23-25): Implicit Heap Usage ==================================== "An implementation should document any cases in which it dynamically allocates heap storage for a purpose other than the evaluation of an allocator." Followed, the only other points at which heap storage is dynamically allocated are as follows: * At initial elaboration time, to allocate dynamically sized global objects. * To allocate space for a task when a task is created. * To extend the secondary stack dynamically when needed. The secondary stack is used for returning variable length results. .. "A default (implementation-provided) storage pool for an access-to-constant type should not have overhead to support deallocation of individual objects." Followed. "A storage pool for an anonymous access type should be created at the point of an allocator for the type, and be reclaimed when the designated object becomes inaccessible." Followed. .. index:: Unchecked deallocation RM 13.11.2(17): Unchecked Deallocation ====================================== "For a standard storage pool, ``Free`` should actually reclaim the storage." Followed. .. index:: Stream oriented attributes RM 13.13.2(1.6): Stream Oriented Attributes =========================================== "If not specified, the value of Stream_Size for an elementary type should be the number of bits that corresponds to the minimum number of stream elements required by the first subtype of the type, rounded up to the nearest factor or multiple of the word size that is also a multiple of the stream element size." Followed, except that the number of stream elements is 1, 2, 3, 4 or 8. The Stream_Size may be used to override the default choice. The default implementation is based on direct binary representations and is therefore target- and endianness-dependent. To address this issue, GNAT also supplies an alternate implementation of the stream attributes ``Read`` and ``Write``, which uses the target-independent XDR standard representation for scalar types. This XDR alternative can be enabled via the binder switch -xdr. .. index:: XDR representation .. index:: Read attribute .. index:: Write attribute .. index:: Stream oriented attributes RM A.1(52): Names of Predefined Numeric Types ============================================= "If an implementation provides additional named predefined integer types, then the names should end with ``Integer`` as in ``Long_Integer``. If an implementation provides additional named predefined floating point types, then the names should end with ``Float`` as in ``Long_Float``." Followed. .. index:: Ada.Characters.Handling RM A.3.2(49): ``Ada.Characters.Handling`` ========================================= "If an implementation provides a localized definition of ``Character`` or ``Wide_Character``, then the effects of the subprograms in ``Characters.Handling`` should reflect the localizations. See also 3.5.2." Followed. GNAT provides no such localized definitions. .. index:: Bounded-length strings RM A.4.4(106): Bounded-Length String Handling ============================================= "Bounded string objects should not be implemented by implicit pointers and dynamic allocation." Followed. No implicit pointers or dynamic allocation are used. .. index:: Random number generation RM A.5.2(46-47): Random Number Generation ========================================= "Any storage associated with an object of type ``Generator`` should be reclaimed on exit from the scope of the object." Followed. "If the generator period is sufficiently long in relation to the number of distinct initiator values, then each possible value of ``Initiator`` passed to ``Reset`` should initiate a sequence of random numbers that does not, in a practical sense, overlap the sequence initiated by any other value. If this is not possible, then the mapping between initiator values and generator states should be a rapidly varying function of the initiator value." Followed. The generator period is sufficiently long for the first condition here to hold true. .. index:: Get_Immediate RM A.10.7(23): ``Get_Immediate`` ================================ "The ``Get_Immediate`` procedures should be implemented with unbuffered input. For a device such as a keyboard, input should be available if a key has already been typed, whereas for a disk file, input should always be available except at end of file. For a file associated with a keyboard-like device, any line-editing features of the underlying operating system should be disabled during the execution of ``Get_Immediate``." Followed on all targets except VxWorks. For VxWorks, there is no way to provide this functionality that does not result in the input buffer being flushed before the ``Get_Immediate`` call. A special unit ``Interfaces.Vxworks.IO`` is provided that contains routines to enable this functionality. .. index:: Containers RM A.18: ``Containers`` ================================ All implementation advice pertaining to Ada.Containers and its child units (that is, all implementation advice occurring within section A.18 and its subsections) is followed except for A.18.24(17): "Bounded ordered set objects should be implemented without implicit pointers or dynamic allocation. " The implementations of the two Reference_Preserving_Key functions of the generic package Ada.Containers.Bounded_Ordered_Sets each currently make use of dynamic allocation; other operations on bounded ordered set objects follow the implementation advice. .. index:: Export RM B.1(39-41): Pragma ``Export`` ================================ "If an implementation supports pragma ``Export`` to a given language, then it should also allow the main subprogram to be written in that language. It should support some mechanism for invoking the elaboration of the Ada library units included in the system, and for invoking the finalization of the environment task. On typical systems, the recommended mechanism is to provide two subprograms whose link names are ``adainit`` and ``adafinal``. ``adainit`` should contain the elaboration code for library units. ``adafinal`` should contain the finalization code. These subprograms should have no effect the second and subsequent time they are called." Followed. "Automatic elaboration of pre-elaborated packages should be provided when pragma ``Export`` is supported." Followed when the main program is in Ada. If the main program is in a foreign language, then ``adainit`` must be called to elaborate pre-elaborated packages. "For each supported convention *L* other than ``Intrinsic``, an implementation should support ``Import`` and ``Export`` pragmas for objects of *L*\ -compatible types and for subprograms, and pragma `Convention` for *L*\ -eligible types and for subprograms, presuming the other language has corresponding features. Pragma ``Convention`` need not be supported for scalar types." Followed. .. index:: Package Interfaces .. index:: Interfaces RM B.2(12-13): Package ``Interfaces`` ===================================== "For each implementation-defined convention identifier, there should be a child package of package Interfaces with the corresponding name. This package should contain any declarations that would be useful for interfacing to the language (implementation) represented by the convention. Any declarations useful for interfacing to any language on the given hardware architecture should be provided directly in ``Interfaces``." Followed. "An implementation supporting an interface to C, COBOL, or Fortran should provide the corresponding package or packages described in the following clauses." Followed. GNAT provides all the packages described in this section. .. index:: C, interfacing with RM B.3(63-71): Interfacing with C ================================= "An implementation should support the following interface correspondences between Ada and C." Followed. "An Ada procedure corresponds to a void-returning C function." Followed. "An Ada function corresponds to a non-void C function." Followed. "An Ada ``in`` scalar parameter is passed as a scalar argument to a C function." Followed. "An Ada ``in`` parameter of an access-to-object type with designated type ``T`` is passed as a ``t*`` argument to a C function, where ``t`` is the C type corresponding to the Ada type ``T``." Followed. "An Ada access ``T`` parameter, or an Ada ``out`` or ``in out`` parameter of an elementary type ``T``, is passed as a ``t*`` argument to a C function, where ``t`` is the C type corresponding to the Ada type ``T``. In the case of an elementary ``out`` or ``in out`` parameter, a pointer to a temporary copy is used to preserve by-copy semantics." Followed. "An Ada parameter of a record type ``T``, of any mode, is passed as a ``t*`` argument to a C function, where ``t`` is the C structure corresponding to the Ada type ``T``." Followed. This convention may be overridden by the use of the C_Pass_By_Copy pragma, or Convention, or by explicitly specifying the mechanism for a given call using an extended import or export pragma. "An Ada parameter of an array type with component type ``T``, of any mode, is passed as a ``t*`` argument to a C function, where ``t`` is the C type corresponding to the Ada type ``T``." Followed. "An Ada parameter of an access-to-subprogram type is passed as a pointer to a C function whose prototype corresponds to the designated subprogram's specification." Followed. .. index:: COBOL, interfacing with RM B.4(95-98): Interfacing with COBOL ===================================== "An Ada implementation should support the following interface correspondences between Ada and COBOL." Followed. "An Ada access ``T`` parameter is passed as a ``BY REFERENCE`` data item of the COBOL type corresponding to ``T``." Followed. "An Ada in scalar parameter is passed as a ``BY CONTENT`` data item of the corresponding COBOL type." Followed. "Any other Ada parameter is passed as a ``BY REFERENCE`` data item of the COBOL type corresponding to the Ada parameter type; for scalars, a local copy is used if necessary to ensure by-copy semantics." Followed. .. index:: Fortran, interfacing with RM B.5(22-26): Interfacing with Fortran ======================================= "An Ada implementation should support the following interface correspondences between Ada and Fortran:" Followed. "An Ada procedure corresponds to a Fortran subroutine." Followed. "An Ada function corresponds to a Fortran function." Followed. "An Ada parameter of an elementary, array, or record type ``T`` is passed as a ``T`` argument to a Fortran procedure, where ``T`` is the Fortran type corresponding to the Ada type ``T``, and where the INTENT attribute of the corresponding dummy argument matches the Ada formal parameter mode; the Fortran implementation's parameter passing conventions are used. For elementary types, a local copy is used if necessary to ensure by-copy semantics." Followed. "An Ada parameter of an access-to-subprogram type is passed as a reference to a Fortran procedure whose interface corresponds to the designated subprogram's specification." Followed. .. index:: Machine operations RM C.1(3-5): Access to Machine Operations ========================================= "The machine code or intrinsic support should allow access to all operations normally available to assembly language programmers for the target environment, including privileged instructions, if any." Followed. "The interfacing pragmas (see Annex B) should support interface to assembler; the default assembler should be associated with the convention identifier ``Assembler``." Followed. "If an entity is exported to assembly language, then the implementation should allocate it at an addressable location, and should ensure that it is retained by the linking process, even if not otherwise referenced from the Ada code. The implementation should assume that any call to a machine code or assembler subprogram is allowed to read or update every object that is specified as exported." Followed. RM C.1(10-16): Access to Machine Operations =========================================== "The implementation should ensure that little or no overhead is associated with calling intrinsic and machine-code subprograms." Followed for both intrinsics and machine-code subprograms. "It is recommended that intrinsic subprograms be provided for convenient access to any machine operations that provide special capabilities or efficiency and that are not otherwise available through the language constructs." Followed. A full set of machine operation intrinsic subprograms is provided. "Atomic read-modify-write operations---e.g., test and set, compare and swap, decrement and test, enqueue/dequeue." Followed on any target supporting such operations. "Standard numeric functions---e.g.:, sin, log." Followed on any target supporting such operations. "String manipulation operations---e.g.:, translate and test." Followed on any target supporting such operations. "Vector operations---e.g.:, compare vector against thresholds." Followed on any target supporting such operations. "Direct operations on I/O ports." Followed on any target supporting such operations. .. index:: Interrupt support RM C.3(28): Interrupt Support ============================= "If the ``Ceiling_Locking`` policy is not in effect, the implementation should provide means for the application to specify which interrupts are to be blocked during protected actions, if the underlying system allows for a finer-grain control of interrupt blocking." Followed. The underlying system does not allow for finer-grain control of interrupt blocking. .. index:: Protected procedure handlers RM C.3.1(20-21): Protected Procedure Handlers ============================================= "Whenever possible, the implementation should allow interrupt handlers to be called directly by the hardware." Followed on any target where the underlying operating system permits such direct calls. "Whenever practical, violations of any implementation-defined restrictions should be detected before run time." Followed. Compile time warnings are given when possible. .. index:: Package ``Interrupts`` .. index:: Interrupts RM C.3.2(25): Package ``Interrupts`` ==================================== "If implementation-defined forms of interrupt handler procedures are supported, such as protected procedures with parameters, then for each such form of a handler, a type analogous to ``Parameterless_Handler`` should be specified in a child package of ``Interrupts``, with the same operations as in the predefined package Interrupts." Followed. .. index:: Pre-elaboration requirements RM C.4(14): Pre-elaboration Requirements ======================================== "It is recommended that pre-elaborated packages be implemented in such a way that there should be little or no code executed at run time for the elaboration of entities not already covered by the Implementation Requirements." Followed. Executable code is generated in some cases, e.g., loops to initialize large arrays. RM C.5(8): Pragma ``Discard_Names`` =================================== "If the pragma applies to an entity, then the implementation should reduce the amount of storage used for storing names associated with that entity." Followed. .. index:: Package Task_Attributes .. index:: Task_Attributes RM C.7.2(30): The Package Task_Attributes ========================================= "Some implementations are targeted to domains in which memory use at run time must be completely deterministic. For such implementations, it is recommended that the storage for task attributes will be pre-allocated statically and not from the heap. This can be accomplished by either placing restrictions on the number and the size of the task's attributes, or by using the pre-allocated storage for the first ``N`` attribute objects, and the heap for the others. In the latter case, ``N`` should be documented." Not followed. This implementation is not targeted to such a domain. .. index:: Locking Policies RM D.3(17): Locking Policies ============================ "The implementation should use names that end with ``_Locking`` for locking policies defined by the implementation." Followed. Two implementation-defined locking policies are defined, whose names (``Inheritance_Locking`` and ``Concurrent_Readers_Locking``) follow this suggestion. .. index:: Entry queuing policies RM D.4(16): Entry Queuing Policies ================================== "Names that end with ``_Queuing`` should be used for all implementation-defined queuing policies." Followed. No such implementation-defined queuing policies exist. .. index:: Preemptive abort RM D.6(9-10): Preemptive Abort ============================== "Even though the *abort_statement* is included in the list of potentially blocking operations (see 9.5.1), it is recommended that this statement be implemented in a way that never requires the task executing the *abort_statement* to block." Followed. "On a multi-processor, the delay associated with aborting a task on another processor should be bounded; the implementation should use periodic polling, if necessary, to achieve this." Followed. .. index:: Tasking restrictions RM D.7(21): Tasking Restrictions ================================ "When feasible, the implementation should take advantage of the specified restrictions to produce a more efficient implementation." GNAT currently takes advantage of these restrictions by providing an optimized run time when the Ravenscar profile and the GNAT restricted run time set of restrictions are specified. See pragma ``Profile (Ravenscar)`` and pragma ``Profile (Restricted)`` for more details. .. index:: Time, monotonic RM D.8(47-49): Monotonic Time ============================= "When appropriate, implementations should provide configuration mechanisms to change the value of ``Tick``." Such configuration mechanisms are not appropriate to this implementation and are thus not supported. "It is recommended that ``Calendar.Clock`` and ``Real_Time.Clock`` be implemented as transformations of the same time base." Followed. "It is recommended that the best time base which exists in the underlying system be available to the application through ``Clock``. `Best` may mean highest accuracy or largest range." Followed. .. index:: Partition communication subsystem .. index:: PCS RM E.5(28-29): Partition Communication Subsystem ================================================ "Whenever possible, the PCS on the called partition should allow for multiple tasks to call the RPC-receiver with different messages and should allow them to block until the corresponding subprogram body returns." A separately supplied PCS that can be used with GNAT when combined with the PolyORB product (NB! See the note in :ref:`polyORB` regarding the lifetime of this product). "The ``Write`` operation on a stream of type ``Params_Stream_Type`` should raise ``Storage_Error`` if it runs out of space trying to write the ``Item`` into the stream." .. index:: COBOL support RM F(7): COBOL Support ====================== "If COBOL (respectively, C) is widely supported in the target environment, implementations supporting the Information Systems Annex should provide the child package ``Interfaces.COBOL`` (respectively, ``Interfaces.C``) specified in Annex B and should support a ``convention_identifier`` of COBOL (respectively, C) in the interfacing pragmas (see Annex B), thus allowing Ada programs to interface with programs written in that language." Followed. .. index:: Decimal radix support RM F.1(2): Decimal Radix Support ================================ "Packed decimal should be used as the internal representation for objects of subtype ``S`` when ``S``'Machine_Radix = 10." Not followed. GNAT ignores ``S``'Machine_Radix and always uses binary representations. .. index:: Numerics RM G: Numerics ============== "If Fortran (respectively, C) is widely supported in the target environment, implementations supporting the Numerics Annex should provide the child package ``Interfaces.Fortran`` (respectively, ``Interfaces.C``) specified in Annex B and should support a ``convention_identifier`` of Fortran (respectively, C) in the interfacing pragmas (see Annex B), thus allowing Ada programs to interface with programs written in that language." Followed. .. index:: Complex types RM G.1.1(56-58): Complex Types ============================== "Because the usual mathematical meaning of multiplication of a complex operand and a real operand is that of the scaling of both components of the former by the latter, an implementation should not perform this operation by first promoting the real operand to complex type and then performing a full complex multiplication. In systems that, in the future, support an Ada binding to IEC 559:1989, the latter technique will not generate the required result when one of the components of the complex operand is infinite. (Explicit multiplication of the infinite component by the zero component obtained during promotion yields a NaN that propagates into the final result.) Analogous advice applies in the case of multiplication of a complex operand and a pure-imaginary operand, and in the case of division of a complex operand by a real or pure-imaginary operand." Not followed. "Similarly, because the usual mathematical meaning of addition of a complex operand and a real operand is that the imaginary operand remains unchanged, an implementation should not perform this operation by first promoting the real operand to complex type and then performing a full complex addition. In implementations in which the ``Signed_Zeros`` attribute of the component type is ``True`` (and which therefore conform to IEC 559:1989 in regard to the handling of the sign of zero in predefined arithmetic operations), the latter technique will not generate the required result when the imaginary component of the complex operand is a negatively signed zero. (Explicit addition of the negative zero to the zero obtained during promotion yields a positive zero.) Analogous advice applies in the case of addition of a complex operand and a pure-imaginary operand, and in the case of subtraction of a complex operand and a real or pure-imaginary operand." Not followed. "Implementations in which ``Real'Signed_Zeros`` is ``True`` should attempt to provide a rational treatment of the signs of zero results and result components. As one example, the result of the ``Argument`` function should have the sign of the imaginary component of the parameter ``X`` when the point represented by that parameter lies on the positive real axis; as another, the sign of the imaginary component of the ``Compose_From_Polar`` function should be the same as (respectively, the opposite of) that of the ``Argument`` parameter when that parameter has a value of zero and the ``Modulus`` parameter has a nonnegative (respectively, negative) value." Followed. .. index:: Complex elementary functions RM G.1.2(49): Complex Elementary Functions ========================================== "Implementations in which ``Complex_Types.Real'Signed_Zeros`` is ``True`` should attempt to provide a rational treatment of the signs of zero results and result components. For example, many of the complex elementary functions have components that are odd functions of one of the parameter components; in these cases, the result component should have the sign of the parameter component at the origin. Other complex elementary functions have zero components whose sign is opposite that of a parameter component at the origin, or is always positive or always negative." Followed. .. index:: Accuracy requirements RM G.2.4(19): Accuracy Requirements =================================== "The versions of the forward trigonometric functions without a ``Cycle`` parameter should not be implemented by calling the corresponding version with a ``Cycle`` parameter of ``2.0*Numerics.Pi``, since this will not provide the required accuracy in some portions of the domain. For the same reason, the version of ``Log`` without a ``Base`` parameter should not be implemented by calling the corresponding version with a ``Base`` parameter of ``Numerics.e``." Followed. .. index:: Complex arithmetic accuracy .. index:: Accuracy, complex arithmetic RM G.2.6(15): Complex Arithmetic Accuracy ========================================= "The version of the ``Compose_From_Polar`` function without a ``Cycle`` parameter should not be implemented by calling the corresponding version with a ``Cycle`` parameter of ``2.0*Numerics.Pi``, since this will not provide the required accuracy in some portions of the domain." Followed. .. index:: Sequential elaboration policy RM H.6(15/2): Pragma Partition_Elaboration_Policy ================================================= "If the partition elaboration policy is ``Sequential`` and the Environment task becomes permanently blocked during elaboration then the partition is deadlocked and it is recommended that the partition be immediately terminated." Not followed. --- ### Gcc Ada Doc Gnat Rm Implementation Defined Aspects (`gcc/ada/doc/gnat_rm/implementation_defined_aspects.rst`) .. _Implementation_Defined_Aspects: ****************************** Implementation Defined Aspects ****************************** Ada defines (throughout the Ada 2012 reference manual, summarized in Annex K) a set of aspects that can be specified for certain entities. These language defined aspects are implemented in GNAT in Ada 2012 mode and work as described in the Ada 2012 Reference Manual. In addition, Ada 2012 allows implementations to define additional aspects whose meaning is defined by the implementation. GNAT provides a number of these implementation-defined aspects which can be used to extend and enhance the functionality of the compiler. This section of the GNAT reference manual describes these additional aspects. Note that any program using these aspects may not be portable to other compilers (although GNAT implements this set of aspects on all platforms). Therefore if portability to other compilers is an important consideration, you should minimize the use of these aspects. Note that for many of these aspects, the effect is essentially similar to the use of a pragma or attribute specification with the same name applied to the entity. For example, if we write: .. code-block:: ada type R is range 1 .. 100 with Value_Size => 10; then the effect is the same as: .. code-block:: ada type R is range 1 .. 100; for R'Value_Size use 10; and if we write: .. code-block:: ada type R is new Integer with Shared => True; then the effect is the same as: .. code-block:: ada type R is new Integer; pragma Shared (R); In the documentation below, such cases are simply marked as being boolean aspects equivalent to the corresponding pragma or attribute definition clause. Aspect Abstract_State ===================== .. index:: Abstract_State This aspect is equivalent to :ref:`pragma Abstract_State`. Aspect Always_Terminates ======================== .. index:: Always_Terminates This boolean aspect is equivalent to :ref:`pragma Always_Terminates`. Aspect Annotate =============== .. index:: Annotate There are three forms of this aspect (where ID is an identifier, and ARG is a general expression), corresponding to :ref:`pragma Annotate`. *Annotate => ID* Equivalent to ``pragma Annotate (ID, Entity => Name);`` *Annotate => (ID)* Equivalent to ``pragma Annotate (ID, Entity => Name);`` *Annotate => (ID ,ID {, ARG})* Equivalent to ``pragma Annotate (ID, ID {, ARG}, Entity => Name);`` Aspect Async_Readers ==================== .. index:: Async_Readers This boolean aspect is equivalent to :ref:`pragma Async_Readers`. Aspect Async_Writers ==================== .. index:: Async_Writers This boolean aspect is equivalent to :ref:`pragma Async_Writers`. Aspect Constant_After_Elaboration ================================= .. index:: Constant_After_Elaboration This aspect is equivalent to :ref:`pragma Constant_After_Elaboration`. Aspect Contract_Cases ===================== .. index:: Contract_Cases This aspect is equivalent to :ref:`pragma Contract_Cases`, the sequence of clauses being enclosed in parentheses so that syntactically it is an aggregate. Aspect Depends ============== .. index:: Depends This aspect is equivalent to :ref:`pragma Depends`. Aspect Default_Initial_Condition ================================ .. index:: Default_Initial_Condition This aspect is equivalent to :ref:`pragma Default_Initial_Condition`. Aspect Dimension ================ .. index:: Dimension The ``Dimension`` aspect is used to specify the dimensions of a given subtype of a dimensioned numeric type. The aspect also specifies a symbol used when doing formatted output of dimensioned quantities. The syntax is:: with Dimension => ([Symbol =>] SYMBOL, DIMENSION_VALUE {, DIMENSION_Value}) SYMBOL ::= STRING_LITERAL | CHARACTER_LITERAL DIMENSION_VALUE ::= RATIONAL | others => RATIONAL | DISCRETE_CHOICE_LIST => RATIONAL RATIONAL ::= [-] NUMERIC_LITERAL [/ NUMERIC_LITERAL] This aspect can only be applied to a subtype whose parent type has a ``Dimension_System`` aspect. The aspect must specify values for all dimensions of the system. The rational values are the powers of the corresponding dimensions that are used by the compiler to verify that physical (numeric) computations are dimensionally consistent. For example, the computation of a force must result in dimensions (L => 1, M => 1, T => -2). For further examples of the usage of this aspect, see package ``System.Dim.Mks``. Note that when the dimensioned type is an integer type, then any dimension value must be an integer literal. Aspect Dimension_System ======================= .. index:: Dimension_System The ``Dimension_System`` aspect is used to define a system of dimensions that will be used in subsequent subtype declarations with ``Dimension`` aspects that reference this system. The syntax is:: with Dimension_System => (DIMENSION {, DIMENSION}); DIMENSION ::= ([Unit_Name =>] IDENTIFIER, [Unit_Symbol =>] SYMBOL, [Dim_Symbol =>] SYMBOL) SYMBOL ::= CHARACTER_LITERAL | STRING_LITERAL This aspect is applied to a type, which must be a numeric derived type (typically a floating-point type), that will represent values within the dimension system. Each ``DIMENSION`` corresponds to one particular dimension. A maximum of 7 dimensions may be specified. ``Unit_Name`` is the name of the dimension (for example ``Meter``). ``Unit_Symbol`` is the shorthand used for quantities of this dimension (for example ``m`` for ``Meter``). ``Dim_Symbol`` gives the identification within the dimension system (typically this is a single letter, e.g. ``L`` standing for length for unit name ``Meter``). The ``Unit_Symbol`` is used in formatted output of dimensioned quantities. The ``Dim_Symbol`` is used in error messages when numeric operations have inconsistent dimensions. GNAT provides the standard definition of the International MKS system in the run-time package ``System.Dim.Mks``. You can easily define similar packages for cgs units or British units, and define conversion factors between values in different systems. The MKS system is characterized by the following aspect: .. code-block:: ada type Mks_Type is new Long_Long_Float with Dimension_System => ( (Unit_Name => Meter, Unit_Symbol => 'm', Dim_Symbol => 'L'), (Unit_Name => Kilogram, Unit_Symbol => "kg", Dim_Symbol => 'M'), (Unit_Name => Second, Unit_Symbol => 's', Dim_Symbol => 'T'), (Unit_Name => Ampere, Unit_Symbol => 'A', Dim_Symbol => 'I'), (Unit_Name => Kelvin, Unit_Symbol => 'K', Dim_Symbol => '@'), (Unit_Name => Mole, Unit_Symbol => "mol", Dim_Symbol => 'N'), (Unit_Name => Candela, Unit_Symbol => "cd", Dim_Symbol => 'J')); Note that in the above type definition, we use the ``at`` symbol (``@``) to represent a theta character (avoiding the use of extended Latin-1 characters in this context). See section 'Performing Dimensionality Analysis in GNAT' in the GNAT Users Guide for detailed examples of use of the dimension system. Aspect Disable_Controlled ========================= .. index:: Disable_Controlled The aspect ``Disable_Controlled`` is defined for controlled record types. If active, this aspect causes suppression of all related calls to ``Initialize``, ``Adjust``, and ``Finalize``. The intended use is for conditional compilation, where for example you might want a record to be controlled or not depending on whether some run-time check is enabled or suppressed. Aspect Effective_Reads ====================== .. index:: Effective_Reads This aspect is equivalent to :ref:`pragma Effective_Reads`. Aspect Effective_Writes ======================= .. index:: Effective_Writes This aspect is equivalent to :ref:`pragma Effective_Writes`. Aspect Exceptional_Cases ======================== .. index:: Exceptional_Cases This aspect may be specified for procedures and functions with side effects; it can be used to list exceptions that might be propagated by the subprogram with side effects in the context of its precondition, and associate them with a specific postcondition. For the syntax and semantics of this aspect, see the SPARK 2014 Reference Manual, section 6.1.9. Aspect Exit_Cases ================= .. index:: Exit_Cases This aspect may be specified for procedures and functions with side effects; it can be used to partition the input state into a list of cases and specify, for each case, how the subprogram is allowed to terminate (i.e. return normally or propagate an exception). For the syntax and semantics of this aspect, see the SPARK 2014 Reference Manual, section 6.1.10. Aspect Extended_Access ====================== This nonoverridable boolean-valued type-related representation aspect can be specified as part of a full_type_declaration for a general access type designating an unconstrained array subtype. The absence of an Extended_Access aspect specification for such a full_type_declaration is equivalent to an explicit "Extended_Access => False" specification. This implies that the aspect is never unspecified for an eligible access type. An access type for which this aspect is True is said to be an extended access type; this includes the case of a type derived from an extended access type. Similarly, a value of such a type is said to be an extended access value. The representation of an extended access value is different than that of other access values. This representation makes it possible to designate objects that cannot be designated using the usual "thin" or "fat" access representations for an access type designating an unconstrained array subtype (notably slices and array objects imported from other languages). In particular, two rules are modified in determining the legality of an Access or Unchecked_Access attribute reference if the expected access type is an extended access type: * A slice of an aliased array object of a non-bitpacked type (more precisely, of an array type having independently addressable components) is considered to be aliased (and the accessibility level of a slice of an array object is defined to be that of the array object); this also applies to renamings of such slices, slices of such renamings, etc. * The requirement that the nominal subtype of the prefix shall statically match the designated subtype of the access type need not be met. The Size aspect (and other aspects including Stream_Size, Object_Size, and Alignment) of an extended access type may depend on the properties of the designated type. Further details of this dependence are not documented. An extended access value is not convertible to a non-extended access type, although conversions in the opposite direction are allowed. We don't want to allow .. code-block:: ada type Big_Ref is access all String with Extended_Access; type Small_Ref is access all String; Obj : aliased String := "abcde"; Big_Ptr : Big_Ref := Obj (2 .. 4)'Access; -- OK Small_Ptr : Small_Ref := Small_Ref (Big_Ptr); -- ERROR: illegal conversion because there is no way to represent the result of such a conversion. A dereference of an extended access value (or a reference to a renaming thereof) shall not occur in any of the following contexts: * as an operative constituent of the prefix of an Access or Unchecked_Access attribute reference whose expected type is not extended; or * as an operative constituent of an actual parameter in a call where the corresponding formal parameter is explicitly aliased. For the same reasons that explicit conversions from an extended access type to a non-extended access type are forbidden, we also need to disallow getting the same effect via a Extended_Ptr.all'Access reference; this includes the case of passing Extended_Ptr.all as an actual parameter in a call where the corresponding formal parameter is explicitly aliased (because the callee could evaluate Formal_Parameter'Access). This goal is accomplished by adjusting the definition of the term "aliased". A dereference of an extended value occurring in one of these contexts is defined to denote a nonaliased view. This has the desired effect because these contexts require an aliased view. Continuing the preceding example, this rule disallows .. code-block:: ada Sneaky_1 : Small_Ptr := Big_Ptr.all'Access; -- ERROR: illegal 'Access prefix function Make (Str : aliased in out String) return Small_Ptr is (Str'Access); -- OK Sneaky_2 : Small_Ptr := Make (Str => Big_Ptr.all); -- ERROR: bad parameter for the same reason given above in the case of an explicit type conversion. Aspect Extensions_Visible ========================= .. index:: Extensions_Visible This aspect is equivalent to :ref:`pragma Extensions_Visible`. Aspect Favor_Top_Level ====================== .. index:: Favor_Top_Level This boolean aspect is equivalent to :ref:`pragma Favor_Top_Level`. Aspect Ghost ============= .. index:: Ghost This aspect is equivalent to :ref:`pragma Ghost`. Aspect Ghost_Predicate ====================== .. index:: Ghost_Predicate This aspect introduces a subtype predicate that can reference ghost entities. The subtype cannot appear as a subtype_mark in a membership test. For the detailed semantics of this aspect, see the entry for subtype predicates in the SPARK Reference Manual, section 3.2.4. Aspect Global ============= .. index:: Global This aspect is equivalent to :ref:`pragma Global`. Aspect Initial_Condition ======================== .. index:: Initial_Condition This aspect is equivalent to :ref:`pragma Initial_Condition`. Aspect Initializes ================== .. index:: Initializes This aspect is equivalent to :ref:`pragma Initializes`. Aspect Inline_Always ==================== .. index:: Inline_Always This boolean aspect is equivalent to :ref:`pragma Inline_Always`. Aspect Invariant ================ .. index:: Invariant This aspect is equivalent to :ref:`pragma Invariant`. It is a synonym for the language defined aspect ``Type_Invariant`` except that it is separately controllable using pragma ``Assertion_Policy``. Aspect Invariant'Class ====================== .. index:: Invariant'Class This aspect is equivalent to :ref:`pragma Type_Invariant_Class`. It is a synonym for the language defined aspect ``Type_Invariant'Class`` except that it is separately controllable using pragma ``Assertion_Policy``. Aspect Iterable =============== .. index:: Iterable This aspect provides a light-weight mechanism for loops and quantified expressions over container types, without the overhead imposed by the tampering checks of standard Ada 2012 iterators. The value of the aspect is an aggregate with six named components, of which the last four are optional: ``First``, ``Next``, ``Has_Element``, ``Element``, ``Constant_Reference``, ``Last``, and ``Previous``. When only the first three components are specified, only the ``for .. in`` form of iteration over cursors is available. When either ``Element`` or ``Constant_Reference`` is specified, both this form and the ``for .. of`` form of iteration over elements are available. ``Element`` and ``Constant_Reference`` cannot be specified together. If the last two components are specified, reverse iterations over the container can be specified (analogous to what can be done over predefined containers that support the ``Reverse_Iterator`` interface). The following is a typical example of use: .. code-block:: ada type List is private with Iterable => (First => First_Cursor, Next => Advance, Has_Element => Cursor_Has_Element [,Element => Get_Element] [,Constant_Reference => Element_Constant_Access] [,Last => Last_Cursor] [,Previous => Retreat]); * The values of ``First`` and ``Last`` are primitive operations of the container type that return a ``Cursor``, which must be a type declared in the container package or visible from it. For example: .. code-block:: ada function First_Cursor (Cont : Container) return Cursor; function Last_Cursor (Cont : Container) return Cursor; * The values of ``Next`` and ``Previous`` are primitive operations of the container type that take both a container and a cursor and yield a cursor. For example: .. code-block:: ada function Advance (Cont : Container; Position : Cursor) return Cursor; function Retreat (Cont : Container; Position : Cursor) return Cursor; * The value of ``Has_Element`` is a primitive operation of the container type that takes both a container and a cursor and yields a boolean. For example: .. code-block:: ada function Cursor_Has_Element (Cont : Container; Position : Cursor) return Boolean; * The value of ``Element`` is a primitive operation of the container type that takes both a container and a cursor and yields an ``Element_Type``, which must be a type declared in the container package or visible from it. For example: .. code-block:: ada function Get_Element (Cont : Container; Position : Cursor) return Element_Type; * The value of ``Constant_Reference`` is a primitive operation of the container type that takes both a container and a cursor and yields an anonymous access-to-constant type designating ``Element_Type`` with null exclusion, which must be a type declared in the container package or visible from it. For example: .. code-block:: ada function Element_Constant_Access (Cont : Container; Position : Cursor) return not null access constant Element_Type; This aspect is used in the GNAT-defined formal container packages. Aspect Linker_Section ===================== .. index:: Linker_Section This aspect is equivalent to :ref:`pragma Linker_Section`. Aspect Local_Restrictions ========================= .. index:: Local_Restrictions This aspect may be specified for a subprogram (and for other declarations as described below). It is used to specify that a particular subprogram does not violate one or more local restrictions, nor can it call a subprogram that is not subject to the same requirement. Positional aggregate syntax (with parentheses, not square brackets) may be used to specify more than one local restriction, as in .. code-block:: ada procedure Do_Something with Local_Restrictions => (Some_Restriction, Another_Restriction); Parentheses are currently required even in the case of specifying a single local restriction (this requirement may be relaxed in the future). Supported local restrictions currently include (only) No_Heap_Allocations and No_Secondary_Stack. No_Secondary_Stack corresponds to the GNAT-defined (global) restriction of the same name. No_Heap_Allocations corresponds to the conjunction of the Ada-defined restrictions No_Allocators and No_Implicit_Heap_Allocations. Additional requirements are imposed in order to ensure that restriction violations cannot be achieved via overriding dispatching operations, calling through an access-to-subprogram value, calling a generic formal subprogram, or calling through a subprogram renaming. For a dispatching operation, an overrider must be subject to (at least) the same restrictions as the overridden inherited subprogram; similarly, the actual subprogram corresponding to a generic formal subprogram in an instantiation must be subject to (at least) the same restrictions as the formal subprogram. A call through an access-to-subprogram value is conservatively assumed to violate all local restrictions; tasking-related constructs (notably entry calls) are treated similarly. A renaming-as-body is treated like a subprogram body containing a call to the renamed subprogram. The Local_Restrictions aspect can be specified for a package specification, in which case the aspect specification also applies to all eligible entities declared with the package. This includes types. Default initialization of an object of a given type is treated like a call to an implicitly-declared initialization subprogram. Such a "call" is subject to the same local restriction checks as any other call. If a type is subject to a local restriction, then any violations of that restriction within the default initialization expressions (if any) of the type are rejected. This may include "calls" to the default initialization subprograms of other types. Local_Restrictions aspect specifications are additive (for example, in the case of a declaration that occurs within nested packages that each have a Local_Restrictions specification). Aspect Lock_Free ================ .. index:: Lock_Free This boolean aspect is equivalent to :ref:`pragma Lock_Free`. Aspect Max_Queue_Length ======================= .. index:: Max_Queue_Length This aspect is equivalent to :ref:`pragma Max_Queue_Length`. Aspect Modifies =============== .. index:: Modifies For the syntax and semantics of this aspect, see the SPARK 2014 Reference Manual, section TBD. Aspect No_Caching ================= .. index:: No_Caching This boolean aspect is equivalent to :ref:`pragma No_Caching`. Aspect No_Elaboration_Code_All ============================== .. index:: No_Elaboration_Code_All This aspect is equivalent to :ref:`pragma No_Elaboration_Code_All` for a program unit. Aspect No_Inline ================ .. index:: No_Inline This boolean aspect is equivalent to :ref:`pragma No_Inline`. Aspect No_Raise =============== .. index:: No_Raise This boolean aspect is equivalent to :ref:`pragma No_Raise`. Aspect No_Tagged_Streams ======================== .. index:: No_Tagged_Streams This aspect is equivalent to :ref:`pragma No_Tagged_Streams` with an argument specifying a root tagged type (thus this aspect can only be applied to such a type). Aspect No_Task_Parts ======================== .. index:: No_Task_Parts Applies to a type. If True, requires that the type and any descendants do not have any task parts. The rules for this aspect are the same as for the language-defined No_Controlled_Parts aspect (see RM-H.4.1), replacing "controlled" with "task". If No_Task_Parts is True for a type T, then the compiler can optimize away certain tasking-related code that would otherwise be needed for T'Class, because descendants of T might contain tasks. Aspect Object_Size ================== .. index:: Object_Size This aspect is equivalent to :ref:`attribute Object_Size`. Aspect Obsolescent ================== .. index:: Obsolescent This aspect is equivalent to :ref:`pragma Obsolescent`. Note that the evaluation of this aspect happens at the point of occurrence, it is not delayed until the freeze point. Aspect Part_Of ============== .. index:: Part_Of This aspect is equivalent to :ref:`pragma Part_Of`. Aspect Persistent_BSS ===================== .. index:: Persistent_BSS This boolean aspect is equivalent to :ref:`pragma Persistent_BSS`. Aspect Potentially_Invalid ========================== .. index:: Potentially_Invalid For the syntax and semantics of this aspect, see the SPARK 2014 Reference Manual, section 13.9.1. Aspect Predicate ================ .. index:: Predicate This aspect is equivalent to :ref:`pragma Predicate`. It is thus similar to the language defined aspects ``Dynamic_Predicate`` and ``Static_Predicate`` except that whether the resulting predicate is static or dynamic is controlled by the form of the expression. It is also separately controllable using pragma ``Assertion_Policy``. Aspect Program_Exit =================== .. index:: Program_Exit This boolean aspect is equivalent to :ref:`pragma Program_Exit`. Aspect Pure_Function ==================== .. index:: Pure_Function This boolean aspect is equivalent to :ref:`pragma Pure_Function`. Aspect Refined_Depends ====================== .. index:: Refined_Depends This aspect is equivalent to :ref:`pragma Refined_Depends`. Aspect Refined_Global ===================== .. index:: Refined_Global This aspect is equivalent to :ref:`pragma Refined_Global`. Aspect Refined_Post =================== .. index:: Refined_Post This aspect is equivalent to :ref:`pragma Refined_Post`. Aspect Refined_State ==================== .. index:: Refined_State This aspect is equivalent to :ref:`pragma Refined_State`. Aspect Relaxed_Initialization ============================= .. index:: Refined_Initialization For the syntax and semantics of this aspect, see the SPARK 2014 Reference Manual, section 6.10. Aspect Remote_Access_Type ========================= .. index:: Remote_Access_Type This aspect is equivalent to :ref:`pragma Remote_Access_Type`. Aspect Scalar_Storage_Order =========================== .. index:: Scalar_Storage_Order This aspect is equivalent to a :ref:`attribute Scalar_Storage_Order`. Aspect Secondary_Stack_Size =========================== .. index:: Secondary_Stack_Size This aspect is equivalent to :ref:`pragma Secondary_Stack_Size`. Aspect Shared ============= .. index:: Shared This boolean aspect is equivalent to :ref:`pragma Shared` and is thus a synonym for aspect ``Atomic``. Aspect Side_Effects =================== .. index:: Side_Effects This aspect is equivalent to :ref:`pragma Side_Effects`. Aspect Simple_Storage_Pool ========================== .. index:: Simple_Storage_Pool This aspect is equivalent to :ref:`attribute Simple_Storage_Pool`. Aspect Simple_Storage_Pool_Type =============================== .. index:: Simple_Storage_Pool_Type This boolean aspect is equivalent to :ref:`pragma Simple_Storage_Pool_Type`. Aspect SPARK_Mode ================= .. index:: SPARK_Mode This aspect is equivalent to :ref:`pragma SPARK_Mode` and may be specified for either or both of the specification and body of a subprogram or package. Aspect Subprogram_Variant ========================= .. index:: Subprogram_Variant For the syntax and semantics of this aspect, see the SPARK 2014 Reference Manual, section 6.1.8. Aspect Suppress_Debug_Info ========================== .. index:: Suppress_Debug_Info This boolean aspect is equivalent to :ref:`pragma Suppress_Debug_Info`. Aspect Suppress_Initialization ============================== .. index:: Suppress_Initialization This boolean aspect is equivalent to :ref:`pragma Suppress_Initialization`. Aspect Test_Case ================ .. index:: Test_Case This aspect is equivalent to :ref:`pragma Test_Case`. Aspect Thread_Local_Storage =========================== .. index:: Thread_Local_Storage This boolean aspect is equivalent to :ref:`pragma Thread_Local_Storage`. Aspect Universal_Aliasing ========================= .. index:: Universal_Aliasing This boolean aspect is equivalent to :ref:`pragma Universal_Aliasing`. Aspect Unmodified ================= .. index:: Unmodified This boolean aspect is equivalent to :ref:`pragma Unmodified`. Aspect Unreferenced =================== .. index:: Unreferenced This boolean aspect is equivalent to :ref:`pragma Unreferenced`. When using the ``-gnat2022`` switch, this aspect is also supported on formal parameters, which is in particular the only form possible for expression functions. Aspect Unreferenced_Objects =========================== .. index:: Unreferenced_Objects This boolean aspect is equivalent to :ref:`pragma Unreferenced_Objects`. Aspect User_Aspect ================== .. index:: User_Aspect This aspect takes an argument that is the name of an aspect defined by a User_Aspect_Definition configuration pragma. A User_Aspect aspect specification is semantically equivalent to replicating the set of aspect specifications associated with the named pragma-defined aspect. Aspect Value_Size ================= .. index:: Value_Size This aspect is equivalent to :ref:`attribute Value_Size`. Aspect Volatile_Full_Access =========================== .. index:: Volatile_Full_Access This boolean aspect is equivalent to :ref:`pragma Volatile_Full_Access`. Aspect Volatile_Function =========================== .. index:: Volatile_Function This boolean aspect is equivalent to :ref:`pragma Volatile_Function`. Aspect Warnings =============== .. index:: Warnings This aspect is equivalent to the two argument form of :ref:`pragma Warnings`, where the first argument is ``ON`` or ``OFF`` and the second argument is the entity. --- ### Gcc Ada Doc Gnat Rm Implementation Defined Attributes (`gcc/ada/doc/gnat_rm/implementation_defined_attributes.rst`) .. _Implementation_Defined_Attributes: ********************************* Implementation Defined Attributes ********************************* Ada defines (throughout the Ada reference manual, summarized in Annex K), a set of attributes that provide useful additional functionality in all areas of the language. These language defined attributes are implemented in GNAT and work as described in the Ada Reference Manual. In addition, Ada allows implementations to define additional attributes whose meaning is defined by the implementation. GNAT provides a number of these implementation-dependent attributes which can be used to extend and enhance the functionality of the compiler. This section of the GNAT reference manual describes these additional attributes. It also describes additional implementation-dependent features of standard language-defined attributes. Note that any program using these attributes may not be portable to other compilers (although GNAT implements this set of attributes on all platforms). Therefore if portability to other compilers is an important consideration, you should minimize the use of these attributes. Attribute Abort_Signal ====================== .. index:: Abort_Signal ``Standard'Abort_Signal`` (``Standard`` is the only allowed prefix) provides the entity for the special exception used to signal task abort or asynchronous transfer of control. Normally this attribute should only be used in the tasking runtime (it is highly peculiar, and completely outside the normal semantics of Ada, for a user program to intercept the abort exception). Attribute Address_Size ====================== .. index:: Size of ``Address`` .. index:: Address_Size ``Standard'Address_Size`` (``Standard`` is the only allowed prefix) is a static constant giving the number of bits in an ``Address``. It is the same value as System.Address'Size, but has the advantage of being static, while a direct reference to System.Address'Size is nonstatic because Address is a private type. Attribute Asm_Input =================== .. index:: Asm_Input The ``Asm_Input`` attribute denotes a function that takes two parameters. The first is a string, the second is an expression of the type designated by the prefix. The first (string) argument is required to be a static expression, and is the constraint for the parameter, (e.g., what kind of register is required). The second argument is the value to be used as the input argument. The possible values for the constant are the same as those used in the RTL, and are dependent on the configuration file used to built the GCC back end. :ref:`Machine_Code_Insertions` Attribute Asm_Output ==================== .. index:: Asm_Output The ``Asm_Output`` attribute denotes a function that takes two parameters. The first is a string, the second is the name of a variable of the type designated by the attribute prefix. The first (string) argument is required to be a static expression and designates the constraint for the parameter (e.g., what kind of register is required). The second argument is the variable to be updated with the result. The possible values for constraint are the same as those used in the RTL, and are dependent on the configuration file used to build the GCC back end. If there are no output operands, then this argument may either be omitted, or explicitly given as ``No_Output_Operands``. :ref:`Machine_Code_Insertions` Attribute Atomic_Always_Lock_Free ================================= .. index:: Atomic_Always_Lock_Free The prefix of the ``Atomic_Always_Lock_Free`` attribute is a type. The result indicates whether atomic operations are supported by the target for the given type. Attribute Bit ============= .. index:: Bit ``obj'Bit``, where ``obj`` is any object, yields the bit offset within the storage unit (byte) that contains the first bit of storage allocated for the object. The value of this attribute is of the type *universal_integer* and is always a nonnegative number smaller than ``System.Storage_Unit``. For an object that is a variable or a constant allocated in a register, the value is zero. (The use of this attribute does not force the allocation of a variable to memory). For an object that is a formal parameter, this attribute applies to either the matching actual parameter or to a copy of the matching actual parameter. For an access object the value is zero. Note that ``obj.all'Bit`` is subject to an ``Access_Check`` for the designated object. Similarly for a record component ``X.C'Bit`` is subject to a discriminant check and ``X(I).Bit`` and ``X(I1..I2)'Bit`` are subject to index checks. This attribute is designed to be compatible with the DEC Ada 83 definition and implementation of the ``Bit`` attribute. Attribute Bit_Position ====================== .. index:: Bit_Position ``R.C'Bit_Position``, where ``R`` is a record object and ``C`` is one of the fields of the record type, yields the bit offset within the record contains the first bit of storage allocated for the object. The value of this attribute is of the type *universal_integer*. The value depends only on the field ``C`` and is independent of the alignment of the containing record ``R``. Attribute Code_Address ====================== .. index:: Code_Address .. index:: Subprogram address .. index:: Address of subprogram code The ``'Address`` attribute may be applied to subprograms in Ada 95 and Ada 2005, but the intended effect seems to be to provide an address value which can be used to call the subprogram by means of an address clause as in the following example: .. code-block:: ada procedure K is ... procedure L; for L'Address use K'Address; pragma Import (Ada, L); A call to ``L`` is then expected to result in a call to ``K``. In Ada 83, where there were no access-to-subprogram values, this was a common work-around for getting the effect of an indirect call. GNAT implements the above use of ``Address`` and the technique illustrated by the example code works correctly. However, for some purposes, it is useful to have the address of the start of the generated code for the subprogram. On some architectures, this is not necessarily the same as the ``Address`` value described above. For example, the ``Address`` value may reference a subprogram descriptor rather than the subprogram itself. The ``'Code_Address`` attribute, which can only be applied to subprogram entities, always returns the address of the start of the generated code of the specified subprogram, which may or may not be the same value as is returned by the corresponding ``'Address`` attribute. Attribute Compiler_Version ========================== .. index:: Compiler_Version ``Standard'Compiler_Version`` (``Standard`` is the only allowed prefix) yields a static string identifying the version of the compiler being used to compile the unit containing the attribute reference. Attribute Constrained ===================== .. index:: Constrained In addition to the usage of this attribute in the Ada RM, GNAT also permits the use of the ``'Constrained`` attribute in a generic template for any type, including types without discriminants. The value of this attribute in the generic instance when applied to a scalar type or a record type without discriminants is always ``True``. This usage is compatible with older Ada compilers, including notably DEC Ada. Attribute Default_Bit_Order =========================== .. index:: Big endian .. index:: Little endian .. index:: Default_Bit_Order ``Standard'Default_Bit_Order`` (``Standard`` is the only allowed prefix), provides the value ``System.Default_Bit_Order`` as a ``Pos`` value (0 for ``High_Order_First``, 1 for ``Low_Order_First``). This is used to construct the definition of ``Default_Bit_Order`` in package ``System``. Attribute Default_Scalar_Storage_Order ====================================== .. index:: Big endian .. index:: Little endian .. index:: Default_Scalar_Storage_Order ``Standard'Default_Scalar_Storage_Order`` (``Standard`` is the only allowed prefix), provides the current value of the default scalar storage order (as specified using pragma ``Default_Scalar_Storage_Order``, or equal to ``Default_Bit_Order`` if unspecified) as a ``System.Bit_Order`` value. This is a static attribute. Attribute Deref =============== .. index:: Deref The attribute ``typ'Deref(expr)`` where ``expr`` is of type ``System.Address`` yields the variable of type ``typ`` that is located at the given address. It is similar to ``(totyp (expr).all)``, where ``totyp`` is an unchecked conversion from address to a named access-to-`typ` type, except that it yields a variable, so it can be used on the left side of an assignment. Attribute Descriptor_Size ========================= .. index:: Descriptor .. index:: Dope vector .. index:: Descriptor_Size Nonstatic attribute ``Descriptor_Size`` returns the size in bits of the descriptor allocated for a type. The result is non-zero only for unconstrained array types and the returned value is of type universal integer. In GNAT, an array descriptor contains bounds information and is located immediately before the first element of the array. .. code-block:: ada type Unconstr_Array is array (Short_Short_Integer range <>) of Positive; Put_Line ("Descriptor size = " & Unconstr_Array'Descriptor_Size'Img); The attribute takes into account any padding due to the alignment of the component type. In the example above, the descriptor contains two values of type ``Short_Short_Integer`` representing the low and high bound. But, since ``Positive`` has an alignment of 4, the size of the descriptor is ``2 * Short_Short_Integer'Size`` rounded up to the next multiple of 32, which yields a size of 32 bits, i.e. including 16 bits of padding. Attribute Elaborated ==================== .. index:: Elaborated The prefix of the ``'Elaborated`` attribute must be a unit name. The value is a Boolean which indicates whether or not the given unit has been elaborated. This attribute is primarily intended for internal use by the generated code for dynamic elaboration checking, but it can also be used in user programs. The value will always be True once elaboration of all units has been completed. An exception is for units which need no elaboration, the value is always False for such units. Attribute Elab_Body =================== .. index:: Elab_Body This attribute can only be applied to a program unit name. It returns the entity for the corresponding elaboration procedure for elaborating the body of the referenced unit. This is used in the main generated elaboration procedure by the binder and is not normally used in any other context. However, there may be specialized situations in which it is useful to be able to call this elaboration procedure from Ada code, e.g., if it is necessary to do selective re-elaboration to fix some error. Attribute Elab_Spec =================== .. index:: Elab_Spec This attribute can only be applied to a program unit name. It returns the entity for the corresponding elaboration procedure for elaborating the spec of the referenced unit. This is used in the main generated elaboration procedure by the binder and is not normally used in any other context. However, there may be specialized situations in which it is useful to be able to call this elaboration procedure from Ada code, e.g., if it is necessary to do selective re-elaboration to fix some error. Attribute Elab_Subp_Body ======================== .. index:: Elab_Subp_Body This attribute can only be applied to a library level subprogram name and is only allowed in CodePeer mode. It returns the entity for the corresponding elaboration procedure for elaborating the body of the referenced subprogram unit. This is used in the main generated elaboration procedure by the binder in CodePeer mode only and is unrecognized otherwise. Attribute Emax ============== .. index:: Ada 83 attributes .. index:: Emax The ``Emax`` attribute is provided for compatibility with Ada 83. See the Ada 83 reference manual for an exact description of the semantics of this attribute. Attribute Enabled ================= .. index:: Enabled The ``Enabled`` attribute allows an application program to check at compile time to see if the designated check is currently enabled. The prefix is a simple identifier, referencing any predefined check name (other than ``All_Checks``) or a check name introduced by pragma Check_Name. If no argument is given for the attribute, the check is for the general state of the check, if an argument is given, then it is an entity name, and the check indicates whether an ``Suppress`` or ``Unsuppress`` has been given naming the entity (if not, then the argument is ignored). Note that instantiations inherit the check status at the point of the instantiation, so a useful idiom is to have a library package that introduces a check name with ``pragma Check_Name``, and then contains generic packages or subprograms which use the ``Enabled`` attribute to see if the check is enabled. A user of this package can then issue a ``pragma Suppress`` or ``pragma Unsuppress`` before instantiating the package or subprogram, controlling whether the check will be present. Attribute Enum_Rep ================== .. index:: Representation of enums .. index:: Enum_Rep Note that this attribute is now standard in Ada 202x and is available as an implementation defined attribute for earlier Ada versions. For every enumeration subtype ``S``, ``S'Enum_Rep`` denotes a function with the following spec: .. code-block:: ada function S'Enum_Rep (Arg : S'Base) return ; It is also allowable to apply ``Enum_Rep`` directly to an object of an enumeration type or to a non-overloaded enumeration literal. In this case ``S'Enum_Rep`` is equivalent to ``typ'Enum_Rep(S)`` where ``typ`` is the type of the enumeration literal or object. The function returns the representation value for the given enumeration value. This will be equal to value of the ``Pos`` attribute in the absence of an enumeration representation clause. This is a static attribute (i.e., the result is static if the argument is static). ``S'Enum_Rep`` can also be used with integer types and objects, in which case it simply returns the integer value. The reason for this is to allow it to be used for ``(<>)`` discrete formal arguments in a generic unit that can be instantiated with either enumeration types or integer types. Note that if ``Enum_Rep`` is used on a modular type whose upper bound exceeds the upper bound of the largest signed integer type, and the argument is a variable, so that the universal integer calculation is done at run time, then the call to ``Enum_Rep`` may raise ``Constraint_Error``. Attribute Enum_Val ================== .. index:: Representation of enums .. index:: Enum_Val Note that this attribute is now standard in Ada 202x and is available as an implementation defined attribute for earlier Ada versions. For every enumeration subtype ``S``, ``S'Enum_Val`` denotes a function with the following spec: .. code-block:: ada function S'Enum_Val (Arg : ) return S'Base; The function returns the enumeration value whose representation matches the argument, or raises Constraint_Error if no enumeration literal of the type has the matching value. This will be equal to value of the ``Val`` attribute in the absence of an enumeration representation clause. This is a static attribute (i.e., the result is static if the argument is static). Attribute Epsilon ================= .. index:: Ada 83 attributes .. index:: Epsilon The ``Epsilon`` attribute is provided for compatibility with Ada 83. See the Ada 83 reference manual for an exact description of the semantics of this attribute. Attribute Fast_Math =================== .. index:: Fast_Math ``Standard'Fast_Math`` (``Standard`` is the only allowed prefix) yields a static Boolean value that is True if pragma ``Fast_Math`` is active, and False otherwise. Attribute Finalization_Size =========================== .. index:: Finalization_Size The prefix of attribute ``Finalization_Size`` must be an object or a type. This attribute returns the size of any hidden data reserved by the compiler to handle finalization-related actions. The type of the attribute is *universal_integer*. ``Finalization_Size`` yields a value of zero for a type with no controlled parts, an object whose type has no controlled parts, or an object of a class-wide type whose tag denotes a type with no controlled parts. For a class-wide type, ``Finalization_Size`` yields a non-zero value except if a No_Finalization restriction is in effect, in which case it yields zero. Note that only heap-allocated objects contain finalization data. Attribute Fixed_Value ===================== .. index:: Fixed_Value For every fixed-point type ``S``, ``S'Fixed_Value`` denotes a function with the following specification: .. code-block:: ada function S'Fixed_Value (Arg : ) return S; The value returned is the fixed-point value ``V`` such that:: V = Arg * S'Small The effect is thus similar to first converting the argument to the integer type used to represent ``S``, and then doing an unchecked conversion to the fixed-point type. The difference is that there are full range checks, to ensure that the result is in range. This attribute is primarily intended for use in implementation of the input-output functions for fixed-point values. Attribute From_Address ====================== .. index:: From_Address The prefix of this attribute must be a general access-to-array type (or subtype); the attribute takes a System.Address argument and possibly some additional arguments (described below) and yields a value of the given access type that designates an array object located at the given address. In the case of a non-null array this means that the given address is the address of the first element of the array object (not the address of any sort of bounds descriptor). This allows associating bounds with an address that is, for example, passed in from C code. If the designated array subtype is unconstrained (which is the usual case), then for each dimension (in order) the attribute takes either one or two additional arguments of the corresponding index type - one if the index subtype is a fixed lower bound subtype, two (low bound first) otherwise. In this case, the access type shall be an extended access type (see the description of the Extended_Access aspect). These additional arguments specify the bounds of the designated array object. If the designated array subtype is constrained then no additional arguments are provided and the bounds of the designated object are those of the designated subtype. Roughly speaking, My_Access_Type'From_Address (Addr, Lo, Hi) is equivalent to a declare expression: .. code-block:: ada (declare Obj : aliased Designated_Array_Type (Lo .. Hi) with Import, Address => Addr; begin My_Access_Type'(Obj'Unchecked_Access) end) Attribute From_Any ================== .. index:: From_Any This internal attribute is used for the generation of remote subprogram stubs in the context of the Distributed Systems Annex. Attribute Has_Access_Values =========================== .. index:: Access values, testing for .. index:: Has_Access_Values The prefix of the ``Has_Access_Values`` attribute is a type. The result is a Boolean value which is True if the is an access type, or is a composite type with a component (at any nesting depth) that is an access type, and is False otherwise. The intended use of this attribute is in conjunction with generic definitions. If the attribute is applied to a generic private type, it indicates whether or not the corresponding actual type has access values. Attribute Has_Discriminants =========================== .. index:: Discriminants, testing for .. index:: Has_Discriminants The prefix of the ``Has_Discriminants`` attribute is a type. The result is a Boolean value which is True if the type has discriminants, and False otherwise. The intended use of this attribute is in conjunction with generic definitions. If the attribute is applied to a generic private type, it indicates whether or not the corresponding actual type has discriminants. Attribute Has_Tagged_Values =========================== .. index:: Tagged values, testing for .. index:: Has_Tagged_Values The prefix of the ``Has_Tagged_Values`` attribute is a type. The result is a Boolean value which is True if the type is a composite type (array or record) that is either a tagged type or has a subcomponent that is tagged, and is False otherwise. The intended use of this attribute is in conjunction with generic definitions. If the attribute is applied to a generic private type, it indicates whether or not the corresponding actual type has access values. Attribute Img ============= .. index:: Img The ``Img`` attribute differs from ``Image`` in that, while both can be applied directly to an object, ``Img`` cannot be applied to types. Example usage of the attribute: .. code-block:: ada Put_Line ("X = " & X'Img); which has the same meaning as the more verbose: .. code-block:: ada Put_Line ("X = " & T'Image (X)); where ``T`` is the (sub)type of the object ``X``. Note that technically, in analogy to ``Image``, ``X'Img`` returns a parameterless function that returns the appropriate string when called. This means that ``X'Img`` can be renamed as a function-returning-string, or used in an instantiation as a function parameter. Attribute Initialized ===================== .. index:: Initialized For the syntax and semantics of this attribute, see the SPARK 2014 Reference Manual, section 6.10. Attribute Integer_Value ======================= .. index:: Integer_Value For every integer type ``S``, ``S'Integer_Value`` denotes a function with the following spec: .. code-block:: ada function S'Integer_Value (Arg : ) return S; The value returned is the integer value ``V``, such that:: Arg = V * T'Small where ``T`` is the type of ``Arg``. The effect is thus similar to first doing an unchecked conversion from the fixed-point type to its corresponding implementation type, and then converting the result to the target integer type. The difference is that there are full range checks, to ensure that the result is in range. This attribute is primarily intended for use in implementation of the standard input-output functions for fixed-point values. Attribute Invalid_Value ======================= .. index:: Invalid_Value For every scalar type S, S'Invalid_Value returns an undefined value of the type. If possible this value is an invalid representation for the type. The value returned is identical to the value used to initialize an otherwise uninitialized value of the type if pragma Initialize_Scalars is used, including the ability to modify the value with the binder -Sxx flag and relevant environment variables at run time. Attribute Large =============== .. index:: Ada 83 attributes .. index:: Large The ``Large`` attribute is provided for compatibility with Ada 83. See the Ada 83 reference manual for an exact description of the semantics of this attribute. Attribute Library_Level ======================= .. index:: Library_Level ``P'Library_Level``, where P is an entity name, returns a Boolean value which is True if the entity is declared at the library level, and False otherwise. Note that within a generic instantiation, the name of the generic unit denotes the instance, which means that this attribute can be used to test if a generic is instantiated at the library level, as shown in this example: .. code-block:: ada generic ... package Gen is pragma Compile_Time_Error (not Gen'Library_Level, "Gen can only be instantiated at library level"); ... end Gen; Attribute Loop_Entry ==================== .. index:: Loop_Entry Syntax:: X'Loop_Entry [(loop_name)] The ``Loop_Entry`` attribute is used to refer to the value that an expression had upon entry to a given loop in much the same way that the ``Old`` attribute in a subprogram postcondition can be used to refer to the value an expression had upon entry to the subprogram. The relevant loop is either identified by the given loop name, or it is the innermost enclosing loop when no loop name is given. A ``Loop_Entry`` attribute can only occur within an ``Assert``, ``Assert_And_Cut``, ``Assume``, ``Loop_Variant`` or ``Loop_Invariant`` pragma. In addition, such a pragma must be one of the items in the sequence of statements of a loop body, or nested inside block statements that appear in the sequence of statements of a loop body. A common use of ``Loop_Entry`` is to compare the current value of objects with their initial value at loop entry, in a ``Loop_Invariant`` pragma. The effect of using ``X'Loop_Entry`` is the same as declaring a constant initialized with the initial value of ``X`` at loop entry. This copy is not performed if the loop is not entered, or if the corresponding pragmas are ignored or disabled. Attribute Machine_Size ====================== .. index:: Machine_Size This attribute is identical to the ``Object_Size`` attribute. It is provided for compatibility with the DEC Ada 83 attribute of this name. Attribute Mantissa ================== .. index:: Ada 83 attributes .. index:: Mantissa The ``Mantissa`` attribute is provided for compatibility with Ada 83. See the Ada 83 reference manual for an exact description of the semantics of this attribute. .. _Attribute_Maximum_Alignment: Attribute Maximum_Alignment =========================== .. index:: Alignment, maximum .. index:: Maximum_Alignment ``Standard'Maximum_Alignment`` (``Standard`` is the only allowed prefix) provides the maximum default alignment value for the target, that is to say the maximum alignment that the compiler may choose by default for a type or an object. Larger alignments are supported up to some maximum value dependent on the target, but may require specific mechanisms that are not needed up to ``Standard'Maximum_Alignment``. Attribute Max_Integer_Size ========================== .. index:: Max_Integer_Size ``Standard'Max_Integer_Size`` (``Standard`` is the only allowed prefix) provides the size of the largest supported integer type for the target. The result is a static constant. Attribute Mechanism_Code ======================== .. index:: Return values, passing mechanism .. index:: Parameters, passing mechanism .. index:: Mechanism_Code ``func'Mechanism_Code`` yields an integer code for the mechanism used for the result of function ``func``, and ``subprog'Mechanism_Code (n)`` yields the mechanism used for formal parameter number *n* (a static integer value, with 1 meaning the first parameter) of subprogram ``subprog``. The code returned is: *1* by copy (value) *2* by reference Attribute Null_Parameter ======================== .. index:: Zero address, passing .. index:: Null_Parameter A reference ``T'Null_Parameter`` denotes an imaginary object of type or subtype ``T`` allocated at machine address zero. The attribute is allowed only as the default expression of a formal parameter, or as an actual expression of a subprogram call. In either case, the subprogram must be imported. The identity of the object is represented by the address zero in the argument list, independent of the passing mechanism (explicit or default). This capability is needed to specify that a zero address should be passed for a record or other composite object passed by reference. There is no way of indicating this without the ``Null_Parameter`` attribute. .. _Attribute-Object_Size: Attribute Object_Size ===================== .. index:: Size, used for objects .. index:: Object_Size The size of an object is not necessarily the same as the size of the type of an object. This is because by default object sizes are increased to be a multiple of the alignment of the object. For example, ``Natural'Size`` is 31, but by default objects of type ``Natural`` will have a size of 32 bits. Similarly, a record containing an integer and a character: .. code-block:: ada type Rec is record I : Integer; C : Character; end record; will have a size of 40 (that is ``Rec'Size`` will be 40). The alignment will be 4, because of the integer field, and so the default size of record objects for this type will be 64 (8 bytes). If the alignment of the above record is specified to be 1, then the object size will be 40 (5 bytes). This is true by default, and also an object size of 40 can be explicitly specified in this case. A consequence of this capability is that different object sizes can be given to subtypes that would otherwise be considered in Ada to be statically matching. But it makes no sense to consider such subtypes as statically matching. Consequently, GNAT adds a rule to the static matching rules that requires object sizes to match. Consider this example: .. code-block:: ada 1. procedure BadAVConvert is 2. type R is new Integer; 3. subtype R1 is R range 1 .. 10; 4. subtype R2 is R range 1 .. 10; 5. for R1'Object_Size use 8; 6. for R2'Object_Size use 16; 7. type R1P is access all R1; 8. type R2P is access all R2; 9. R1PV : R1P := new R1'(4); 10. R2PV : R2P; 11. begin 12. R2PV := R2P (R1PV); | >>> target designated subtype not compatible with type "R1" defined at line 3 13. end; In the absence of lines 5 and 6, types ``R1`` and ``R2`` statically match and hence the conversion on line 12 is legal. But since lines 5 and 6 cause the object sizes to differ, GNAT considers that types ``R1`` and ``R2`` are not statically matching, and line 12 generates the diagnostic shown above. Similar additional checks are performed in other contexts requiring statically matching subtypes. Attribute Old ============= .. index:: Old In addition to the usage of ``Old`` defined in the Ada 2012 RM (usage within ``Post`` aspect), GNAT also permits the use of this attribute in implementation defined pragmas ``Postcondition``, ``Contract_Cases`` and ``Test_Case``. Also usages of ``Old`` which would be illegal according to the Ada 2012 RM definition are allowed under control of implementation defined pragma ``Unevaluated_Use_Of_Old``. Attribute Passed_By_Reference ============================= .. index:: Parameters, when passed by reference .. index:: Passed_By_Reference ``typ'Passed_By_Reference`` for any subtype `typ` returns a value of type ``Boolean`` value that is ``True`` if the type is normally passed by reference and ``False`` if the type is normally passed by copy in calls. For scalar types, the result is always ``False`` and is static. For non-scalar types, the result is nonstatic. Attribute Pool_Address ====================== .. index:: Pool_Address ``X'Pool_Address`` for any object ``X`` returns the address of X within its storage pool. This is the same as ``X'Address``, except that for an unconstrained array whose bounds are allocated just before the first component, ``X'Pool_Address`` returns the address of those bounds, whereas ``X'Address`` returns the address of the first component. Here, we are interpreting 'storage pool' broadly to mean ``wherever the object is allocated``, which could be a user-defined storage pool, the global heap, on the stack, or in a static memory area. For an object created by ``new``, ``Ptr.all'Pool_Address`` is what is passed to ``Allocate`` and returned from ``Deallocate``. Attribute Range_Length ====================== .. index:: Range_Length ``typ'Range_Length`` for any discrete type `typ` yields the number of values represented by the subtype (zero for a null range). The result is static for static subtypes. ``Range_Length`` applied to the index subtype of a one dimensional array always gives the same result as ``Length`` applied to the array itself. Attribute Ref ============= ``System.Address'Ref`` (``Address`` is the only permissible prefix) is equivalent to ``System'To_Address``, provided for compatibility with other compilers. Attribute Restriction_Set ========================= .. index:: Restriction_Set .. index:: Restrictions This attribute allows compile time testing of restrictions that are currently in effect. It is primarily intended for specializing code in the run-time based on restrictions that are active (e.g. don't need to save fpt registers if restriction No_Floating_Point is known to be in effect), but can be used anywhere. There are two forms: .. code-block:: ada System'Restriction_Set (partition_boolean_restriction_NAME) System'Restriction_Set (No_Dependence => library_unit_NAME); In the case of the first form, the only restriction names allowed are parameterless restrictions that are checked for consistency at bind time. For a complete list see the subtype ``System.Rident.Partition_Boolean_Restrictions``. The result returned is True if the restriction is known to be in effect, and False if the restriction is known not to be in effect. An important guarantee is that the value of a Restriction_Set attribute is known to be consistent throughout all the code of a partition. This is trivially achieved if the entire partition is compiled with a consistent set of restriction pragmas. However, the compilation model does not require this. It is possible to compile one set of units with one set of pragmas, and another set of units with another set of pragmas. It is even possible to compile a spec with one set of pragmas, and then WITH the same spec with a different set of pragmas. Inconsistencies in the actual use of the restriction are checked at bind time. In order to achieve the guarantee of consistency for the Restriction_Set pragma, we consider that a use of the pragma that yields False is equivalent to a violation of the restriction. So for example if you write .. code-block:: ada if System'Restriction_Set (No_Floating_Point) then ... else ... end if; And the result is False, so that the else branch is executed, you can assume that this restriction is not set for any unit in the partition. This is checked by considering this use of the restriction pragma to be a violation of the restriction No_Floating_Point. This means that no other unit can attempt to set this restriction (if some unit does attempt to set it, the binder will refuse to bind the partition). Technical note: The restriction name and the unit name are interpreted entirely syntactically, as in the corresponding Restrictions pragma, they are not analyzed semantically, so they do not have a type. Attribute Result ================ .. index:: Result ``function'Result`` can only be used with in a Postcondition pragma for a function. The prefix must be the name of the corresponding function. This is used to refer to the result of the function in the postcondition expression. For a further discussion of the use of this attribute and examples of its use, see the description of pragma Postcondition. Attribute Round ===================== .. index:: Round In addition to the usage of this attribute in the Ada RM, GNAT also permits the use of the ``'Round`` attribute for ordinary fixed point types. Attribute Safe_Emax =================== .. index:: Ada 83 attributes .. index:: Safe_Emax The ``Safe_Emax`` attribute is provided for compatibility with Ada 83. See the Ada 83 reference manual for an exact description of the semantics of this attribute. Attribute Safe_Large ==================== .. index:: Ada 83 attributes .. index:: Safe_Large The ``Safe_Large`` attribute is provided for compatibility with Ada 83. See the Ada 83 reference manual for an exact description of the semantics of this attribute. Attribute Safe_Small ==================== .. index:: Ada 83 attributes .. index:: Safe_Small The ``Safe_Small`` attribute is provided for compatibility with Ada 83. See the Ada 83 reference manual for an exact description of the semantics of this attribute. .. _Attribute-Scalar_Storage_Order: Attribute Scalar_Storage_Order ============================== .. index:: Endianness .. index:: Scalar storage order .. index:: Scalar_Storage_Order For every array or record type ``S``, the representation attribute ``Scalar_Storage_Order`` denotes the order in which storage elements that make up scalar components are ordered within S. The value given must be a static expression of type System.Bit_Order. The following is an example of the use of this feature: .. code-block:: ada -- Component type definitions subtype Yr_Type is Natural range 0 .. 127; subtype Mo_Type is Natural range 1 .. 12; subtype Da_Type is Natural range 1 .. 31; -- Record declaration type Date is record Years_Since_1980 : Yr_Type; Month : Mo_Type; Day_Of_Month : Da_Type; end record; -- Record representation clause for Date use record Years_Since_1980 at 0 range 0 .. 6; Month at 0 range 7 .. 10; Day_Of_Month at 0 range 11 .. 15; end record; -- Attribute definition clauses for Date'Bit_Order use System.High_Order_First; for Date'Scalar_Storage_Order use System.High_Order_First; -- If Scalar_Storage_Order is specified, it must be consistent with -- Bit_Order, so it's best to always define the latter explicitly if -- the former is used. Other properties are as for the standard representation attribute ``Bit_Order`` defined by Ada RM 13.5.3(4). The default is ``System.Default_Bit_Order``. For a record type ``T``, if ``T'Scalar_Storage_Order`` is specified explicitly, it shall be equal to ``T'Bit_Order``. Note: this means that if a ``Scalar_Storage_Order`` attribute definition clause is not confirming, then the type's ``Bit_Order`` shall be specified explicitly and set to the same value. Derived types inherit an explicitly set scalar storage order from their parent types. This may be overridden for the derived type by giving an explicit scalar storage order for it. However, for a record extension, the derived type must have the same scalar storage order as the parent type. A component of a record type that is itself a record or an array and that does not start and end on a byte boundary must have have the same scalar storage order as the record type. A component of a bit-packed array type that is itself a record or an array must have the same scalar storage order as the array type. No component of a type that has an explicit ``Scalar_Storage_Order`` attribute definition may be aliased. A confirming ``Scalar_Storage_Order`` attribute definition clause (i.e. with a value equal to ``System.Default_Bit_Order``) has no effect. If the opposite storage order is specified, then whenever the value of a scalar component of an object of type ``S`` is read, the storage elements of the enclosing machine scalar are first reversed (before retrieving the component value, possibly applying some shift and mask operations on the enclosing machine scalar), and the opposite operation is done for writes. In that case, the restrictions set forth in 13.5.1(10.3/2) for scalar components are relaxed. Instead, the following rules apply: * the underlying storage elements are those at positions ``(position + first_bit / storage_element_size) .. (position + (last_bit + storage_element_size - 1) / storage_element_size)`` * the sequence of underlying storage elements shall have a size no greater than the largest machine scalar * the enclosing machine scalar is defined as the smallest machine scalar starting at a position no greater than ``position + first_bit / storage_element_size`` and covering storage elements at least up to ``position + (last_bit + storage_element_size - 1) / storage_element_size`` * the position of the component is interpreted relative to that machine scalar. If no scalar storage order is specified for a type (either directly, or by inheritance in the case of a derived type), then the default is normally the native ordering of the target, but this default can be overridden using pragma ``Default_Scalar_Storage_Order``. If a component of ``T`` is itself of a record or array type, the specified ``Scalar_Storage_Order`` does *not* apply to that nested type: an explicit attribute definition clause must be provided for the component type as well if desired. Representation changes that explicitly or implicitly toggle the scalar storage order are not supported and may result in erroneous execution of the program, except when performed by means of an instance of ``Ada.Unchecked_Conversion``. In particular, overlays are not supported and a warning is given for them: .. code-block:: ada type Rec_LE is record I : Integer; end record; for Rec_LE use record I at 0 range 0 .. 31; end record; for Rec_LE'Bit_Order use System.Low_Order_First; for Rec_LE'Scalar_Storage_Order use System.Low_Order_First; type Rec_BE is record I : Integer; end record; for Rec_BE use record I at 0 range 0 .. 31; end record; for Rec_BE'Bit_Order use System.High_Order_First; for Rec_BE'Scalar_Storage_Order use System.High_Order_First; R_LE : Rec_LE; R_BE : Rec_BE; for R_BE'Address use R_LE'Address; ``warning: overlay changes scalar storage order [enabled by default]`` In most cases, such representation changes ought to be replaced by an instantiation of a function or procedure provided by ``GNAT.Byte_Swapping``. Note that the scalar storage order only affects the in-memory data representation. It has no effect on the representation used by stream attributes. Note that debuggers may be unable to display the correct value of scalar components of a type for which the opposite storage order is specified. .. _Attribute_Simple_Storage_Pool: Attribute Simple_Storage_Pool ============================= .. index:: Storage pool, simple .. index:: Simple storage pool .. index:: Simple_Storage_Pool For every nonformal, nonderived access-to-object type ``Acc``, the representation attribute ``Simple_Storage_Pool`` may be specified via an attribute_definition_clause (or by specifying the equivalent aspect): .. code-block:: ada My_Pool : My_Simple_Storage_Pool_Type; type Acc is access My_Data_Type; for Acc'Simple_Storage_Pool use My_Pool; The name given in an attribute_definition_clause for the ``Simple_Storage_Pool`` attribute shall denote a variable of a 'simple storage pool type' (see pragma `Simple_Storage_Pool_Type`). The use of this attribute is only allowed for a prefix denoting a type for which it has been specified. The type of the attribute is the type of the variable specified as the simple storage pool of the access type, and the attribute denotes that variable. It is illegal to specify both ``Storage_Pool`` and ``Simple_Storage_Pool`` for the same access type. If the ``Simple_Storage_Pool`` attribute has been specified for an access type, then applying the ``Storage_Pool`` attribute to the type is flagged with a warning and its evaluation raises the exception ``Program_Error``. If the Simple_Storage_Pool attribute has been specified for an access type ``S``, then the evaluation of the attribute ``S'Storage_Size`` returns the result of calling ``Storage_Size (S'Simple_Storage_Pool)``, which is intended to indicate the number of storage elements reserved for the simple storage pool. If the Storage_Size function has not been defined for the simple storage pool type, then this attribute returns zero. If an access type ``S`` has a specified simple storage pool of type ``SSP``, then the evaluation of an allocator for that access type calls the primitive ``Allocate`` procedure for type ``SSP``, passing ``S'Simple_Storage_Pool`` as the pool parameter. The detailed semantics of such allocators is the same as those defined for allocators in section 13.11 of the :title:`Ada Reference Manual`, with the term *simple storage pool* substituted for *storage pool*. If an access type ``S`` has a specified simple storage pool of type ``SSP``, then a call to an instance of the ``Ada.Unchecked_Deallocation`` for that access type invokes the primitive ``Deallocate`` procedure for type ``SSP``, passing ``S'Simple_Storage_Pool`` as the pool parameter. The detailed semantics of such unchecked deallocations is the same as defined in section 13.11.2 of the Ada Reference Manual, except that the term *simple storage pool* is substituted for *storage pool*. Attribute Small =============== .. index:: Ada 83 attributes .. index:: Small The ``Small`` attribute is defined in Ada 95 (and Ada 2005) only for fixed-point types. GNAT also allows this attribute to be applied to floating-point types for compatibility with Ada 83. See the Ada 83 reference manual for an exact description of the semantics of this attribute when applied to floating-point types. Attribute Small_Denominator =========================== .. index:: Small .. index:: Small_Denominator ``typ'Small_Denominator`` for any fixed-point subtype `typ` yields the denominator in the representation of ``typ'Small`` as a rational number with coprime factors (i.e. as an irreducible fraction). Attribute Small_Numerator ========================= .. index:: Small .. index:: Small_Numerator ``typ'Small_Numerator`` for any fixed-point subtype `typ` yields the numerator in the representation of ``typ'Small`` as a rational number with coprime factors (i.e. as an irreducible fraction). Attribute Storage_Unit ====================== .. index:: Storage_Unit ``Standard'Storage_Unit`` (``Standard`` is the only allowed prefix) provides the same value as ``System.Storage_Unit``. Attribute Stub_Type =================== .. index:: Stub_Type The GNAT implementation of remote access-to-classwide types is organized as described in AARM section E.4 (20.t): a value of an RACW type (designating a remote object) is represented as a normal access value, pointing to a "stub" object which in turn contains the necessary information to contact the designated remote object. A call on any dispatching operation of such a stub object does the remote call, if necessary, using the information in the stub object to locate the target partition, etc. For a prefix ``T`` that denotes a remote access-to-classwide type, ``T'Stub_Type`` denotes the type of the corresponding stub objects. By construction, the layout of ``T'Stub_Type`` is identical to that of type ``RACW_Stub_Type`` declared in the internal implementation-defined unit ``System.Partition_Interface``. Use of this attribute will create an implicit dependency on this unit. Attribute System_Allocator_Alignment ==================================== .. index:: Alignment, allocator .. index:: System_Allocator_Alignment ``Standard'System_Allocator_Alignment`` (``Standard`` is the only allowed prefix) provides the observable guaranteed to be honored by the system allocator (malloc). This is a static value that can be used in user storage pools based on malloc either to reject allocation with alignment too large or to enable a realignment circuitry if the alignment request is larger than this value. Attribute Target_Name ===================== .. index:: Target_Name ``Standard'Target_Name`` (``Standard`` is the only allowed prefix) provides a static string value that identifies the target for the current compilation. For GCC implementations, this is the standard gcc target name without the terminating slash (for example, GNAT 5.0 on windows yields "i586-pc-mingw32msv"). Attribute To_Address ==================== .. index:: To_Address The ``System'To_Address`` (``System`` is the only allowed prefix) denotes a function identical to ``System.Storage_Elements.To_Address`` except that it is a static attribute. This means that if its argument is a static expression, then the result of the attribute is a static expression. This means that such an expression can be used in contexts (e.g., preelaborable packages) which require a static expression and where the function call could not be used (since the function call is always nonstatic, even if its argument is static). The argument must be in the range -(2**(m-1)) .. 2**m-1, where m is the memory size (typically 32 or 64). Negative values are interpreted in a modular manner (e.g., -1 means the same as 16#FFFF_FFFF# on a 32 bits machine). Attribute To_Any ================ .. index:: To_Any This internal attribute is used for the generation of remote subprogram stubs in the context of the Distributed Systems Annex. Attribute Type_Class ==================== .. index:: Type_Class ``typ'Type_Class`` for any type or subtype `typ` yields the value of the type class for the full type of `typ`. If `typ` is a generic formal type, the value is the value for the corresponding actual subtype. The value of this attribute is of type ``System.Aux_DEC.Type_Class``, which has the following definition: .. code-block:: ada type Type_Class is (Type_Class_Enumeration, Type_Class_Integer, Type_Class_Fixed_Point, Type_Class_Floating_Point, Type_Class_Array, Type_Class_Record, Type_Class_Access, Type_Class_Task, Type_Class_Address); Protected types yield the value ``Type_Class_Task``, which thus applies to all concurrent types. This attribute is designed to be compatible with the DEC Ada 83 attribute of the same name. Attribute Type_Key ================== .. index:: Type_Key The ``Type_Key`` attribute is applicable to a type or subtype and yields a value of type Standard.String containing encoded information about the type or subtype. This provides improved compatibility with other implementations that support this attribute. Attribute TypeCode ================== .. index:: TypeCode This internal attribute is used for the generation of remote subprogram stubs in the context of the Distributed Systems Annex. Attribute Unconstrained_Array ============================= .. index:: Unconstrained_Array The ``Unconstrained_Array`` attribute can be used with a prefix that denotes any type or subtype. It is a static attribute that yields ``True`` if the prefix designates an unconstrained array, and ``False`` otherwise. In a generic instance, the result is still static, and yields the result of applying this test to the generic actual. Attribute Universal_Literal_String ================================== .. index:: Named numbers, representation of .. index:: Universal_Literal_String The prefix of ``Universal_Literal_String`` must be a named number. The static result is the string consisting of the characters of the number as defined in the original source. This allows the user program to access the actual text of named numbers without intermediate conversions and without the need to enclose the strings in quotes (which would preclude their use as numbers). For example, the following program prints the first 50 digits of pi: .. code-block:: ada with Text_IO; use Text_IO; with Ada.Numerics; procedure Pi is begin Put (Ada.Numerics.Pi'Universal_Literal_String); end; Attribute Unrestricted_Access ============================= .. index:: Access, unrestricted .. index:: Unrestricted_Access The ``Unrestricted_Access`` attribute is similar to ``Access`` except that all accessibility and aliased view checks are omitted. This is a user-beware attribute. For objects, it is similar to ``Address``, for which it is a desirable replacement where the value desired is an access type. In other words, its effect is similar to first applying the ``Address`` attribute and then doing an unchecked conversion to a desired access type. For subprograms, ``P'Unrestricted_Access`` may be used where ``P'Access`` would be illegal, to construct a value of a less-nested named access type that designates a more-nested subprogram. This value may be used in indirect calls, so long as the more-nested subprogram still exists; once the subprogram containing it has returned, such calls are erroneous. For example: .. code-block:: ada package body P is type Less_Nested is access procedure; Global : Less_Nested; procedure P1 is begin Global.all; end P1; procedure P2 is Local_Var : Integer; procedure More_Nested is begin ... Local_Var ... end More_Nested; begin Global := More_Nested'Unrestricted_Access; P1; end P2; end P; When P1 is called from P2, the call via Global is OK, but if P1 were called after P2 returns, it would be an erroneous use of a dangling pointer. For objects, it is possible to use ``Unrestricted_Access`` for any type. However, if the result is of an access-to-unconstrained array subtype, then the resulting pointer has the same scope as the context of the attribute, and must not be returned to some enclosing scope. For instance, if a function uses ``Unrestricted_Access`` to create an access-to-unconstrained-array and returns that value to the caller, the result will involve dangling pointers. In addition, it is only valid to create pointers to unconstrained arrays using this attribute if the pointer has the normal default 'fat' representation where a pointer has two components, one points to the array and one points to the bounds. If a size clause is used to force 'thin' representation for a pointer to unconstrained where there is only space for a single pointer, then the resulting pointer is not usable. In the simple case where a direct use of Unrestricted_Access attempts to make a thin pointer for a non-aliased object, the compiler will reject the use as illegal, as shown in the following example: .. code-block:: ada with System; use System; procedure SliceUA2 is type A is access all String; for A'Size use Standard'Address_Size; procedure P (Arg : A) is begin null; end P; X : String := "hello world!"; X2 : aliased String := "hello world!"; AV : A := X'Unrestricted_Access; -- ERROR | >>> illegal use of Unrestricted_Access attribute >>> attempt to generate thin pointer to unaliased object begin P (X'Unrestricted_Access); -- ERROR | >>> illegal use of Unrestricted_Access attribute >>> attempt to generate thin pointer to unaliased object P (X(7 .. 12)'Unrestricted_Access); -- ERROR | >>> illegal use of Unrestricted_Access attribute >>> attempt to generate thin pointer to unaliased object P (X2'Unrestricted_Access); -- OK end; but other cases cannot be detected by the compiler, and are considered to be erroneous. Consider the following example: .. code-block:: ada with System; use System; with System; use System; procedure SliceUA is type AF is access all String; type A is access all String; for A'Size use Standard'Address_Size; procedure P (Arg : A) is begin if Arg'Length /= 6 then raise Program_Error; end if; end P; X : String := "hello world!"; Y : AF := X (7 .. 12)'Unrestricted_Access; begin P (A (Y)); end; A normal unconstrained array value or a constrained array object marked as aliased has the bounds in memory just before the array, so a thin pointer can retrieve both the data and the bounds. But in this case, the non-aliased object ``X`` does not have the bounds before the string. If the size clause for type ``A`` were not present, then the pointer would be a fat pointer, where one component is a pointer to the bounds, and all would be well. But with the size clause present, the conversion from fat pointer to thin pointer in the call loses the bounds, and so this is erroneous, and the program likely raises a ``Program_Error`` exception. In general, it is advisable to completely avoid mixing the use of thin pointers and the use of ``Unrestricted_Access`` where the designated type is an unconstrained array. The use of thin pointers should be restricted to cases of porting legacy code that implicitly assumes the size of pointers, and such code should not in any case be using this attribute. Another erroneous situation arises if the attribute is applied to a constant. The resulting pointer can be used to access the constant, but the effect of trying to modify a constant in this manner is not well-defined. Consider this example: .. code-block:: ada P : constant Integer := 4; type R is access all Integer; RV : R := P'Unrestricted_Access; .. RV.all := 3; Here we attempt to modify the constant P from 4 to 3, but the compiler may or may not notice this attempt, and subsequent references to P may yield either the value 3 or the value 4 or the assignment may blow up if the compiler decides to put P in read-only memory. One particular case where ``Unrestricted_Access`` can be used in this way is to modify the value of an ``in`` parameter: .. code-block:: ada procedure K (S : in String) is type R is access all Character; RV : R := S (3)'Unrestricted_Access; begin RV.all := 'a'; end; In general this is a risky approach. It may appear to "work" but such uses of ``Unrestricted_Access`` are potentially non-portable, even from one version of GNAT to another, so are best avoided if possible. Attribute Update ================ .. index:: Update The ``Update`` attribute creates a copy of an array or record value with one or more modified components. The syntax is:: PREFIX'Update ( RECORD_COMPONENT_ASSOCIATION_LIST ) PREFIX'Update ( ARRAY_COMPONENT_ASSOCIATION {, ARRAY_COMPONENT_ASSOCIATION } ) PREFIX'Update ( MULTIDIMENSIONAL_ARRAY_COMPONENT_ASSOCIATION {, MULTIDIMENSIONAL_ARRAY_COMPONENT_ASSOCIATION } ) MULTIDIMENSIONAL_ARRAY_COMPONENT_ASSOCIATION ::= INDEX_EXPRESSION_LIST_LIST => EXPRESSION INDEX_EXPRESSION_LIST_LIST ::= INDEX_EXPRESSION_LIST {| INDEX_EXPRESSION_LIST } INDEX_EXPRESSION_LIST ::= ( EXPRESSION {, EXPRESSION } ) where ``PREFIX`` is the name of an array or record object, the association list in parentheses does not contain an ``others`` choice and the box symbol ``<>`` may not appear in any expression. The effect is to yield a copy of the array or record value which is unchanged apart from the components mentioned in the association list, which are changed to the indicated value. The original value of the array or record value is not affected. For example: .. code-block:: ada type Arr is Array (1 .. 5) of Integer; ... Avar1 : Arr := (1,2,3,4,5); Avar2 : Arr := Avar1'Update (2 => 10, 3 .. 4 => 20); yields a value for ``Avar2`` of 1,10,20,20,5 with ``Avar1`` begin unmodified. Similarly: .. code-block:: ada type Rec is A, B, C : Integer; ... Rvar1 : Rec := (A => 1, B => 2, C => 3); Rvar2 : Rec := Rvar1'Update (B => 20); yields a value for ``Rvar2`` of (A => 1, B => 20, C => 3), with ``Rvar1`` being unmodified. Note that the value of the attribute reference is computed completely before it is used. This means that if you write: .. code-block:: ada Avar1 := Avar1'Update (1 => 10, 2 => Function_Call); then the value of ``Avar1`` is not modified if ``Function_Call`` raises an exception, unlike the effect of a series of direct assignments to elements of ``Avar1``. In general this requires that two extra complete copies of the object are required, which should be kept in mind when considering efficiency. The ``Update`` attribute cannot be applied to prefixes of a limited type, and cannot reference discriminants in the case of a record type. The accessibility level of an Update attribute result object is defined as for an aggregate. In the record case, no component can be mentioned more than once. In the array case, two overlapping ranges can appear in the association list, in which case the modifications are processed left to right. Multi-dimensional arrays can be modified, as shown by this example: .. code-block:: ada A : array (1 .. 10, 1 .. 10) of Integer; .. A := A'Update ((1, 2) => 20, (3, 4) => 30); which changes element (1,2) to 20 and (3,4) to 30. Attribute Valid_Value ======================= .. index:: Valid_Value The ``'Valid_Value`` attribute is defined for enumeration types other than those in package Standard or types derived from those types. This attribute is a function that takes a String, and returns Boolean. ``T'Valid_Value (S)`` returns True if and only if ``T'Value (S)`` would not raise Constraint_Error. Attribute Valid_Scalars ======================= .. index:: Valid_Scalars The ``'Valid_Scalars`` attribute is intended to make it easier to check the validity of scalar subcomponents of composite objects. The attribute is defined for any prefix ``P`` which denotes an object. Prefix ``P`` can be any type except for ``Unchecked_Union`` types. The value of the attribute is of type ``Boolean``. ``P'Valid_Scalars`` yields ``True`` if and only if the evaluation of ``C'Valid`` yields ``True`` for every scalar subcomponent ``C`` of ``P``, or if ``P`` has no scalar subcomponents. Attribute ``'Valid_Scalars`` is equivalent to attribute ``'Valid`` for scalar types. It is not specified in what order the subcomponents are checked, nor whether any more are checked after any one of them is determined to be invalid. If the prefix ``P`` is of a class-wide type ``T'Class`` (where ``T`` is the associated specific type), or if the prefix ``P`` is of a specific tagged type ``T``, then only the subcomponents of ``T`` are checked; in other words, components of extensions of ``T`` are not checked even if ``T'Class (P)'Tag /= T'Tag``. The compiler will issue a warning if it can be determined at compile time that the prefix of the attribute has no scalar subcomponents. Note: ``Valid_Scalars`` can generate a lot of code, especially in the case of a large variant record. If the attribute is called in many places in the same program applied to objects of the same type, it can reduce program size to write a function with a single use of the attribute, and then call that function from multiple places. Attribute VADS_Size =================== .. index:: Size, VADS compatibility .. index:: VADS_Size The ``'VADS_Size`` attribute is intended to make it easier to port legacy code which relies on the semantics of ``'Size`` as implemented by the VADS Ada 83 compiler. GNAT makes a best effort at duplicating the same semantic interpretation. In particular, ``'VADS_Size`` applied to a predefined or other primitive type with no Size clause yields the Object_Size (for example, ``Natural'Size`` is 32 rather than 31 on typical machines). In addition ``'VADS_Size`` applied to an object gives the result that would be obtained by applying the attribute to the corresponding type. .. _Attribute-Value_Size: Attribute Value_Size ==================== .. index:: Size, setting for not-first subtype .. index:: Value_Size ``type'Value_Size`` is the number of bits required to represent a value of the given subtype. It is the same as ``type'Size``, but, unlike ``Size``, may be set for non-first subtypes. Attribute Wchar_T_Size ====================== .. index:: Wchar_T_Size ``Standard'Wchar_T_Size`` (``Standard`` is the only allowed prefix) provides the size in bits of the C ``wchar_t`` type primarily for constructing the definition of this type in package ``Interfaces.C``. The result is a static constant. Attribute Word_Size =================== .. index:: Word_Size ``Standard'Word_Size`` (``Standard`` is the only allowed prefix) provides the value ``System.Word_Size``. The result is a static constant. --- ### Gcc Ada Doc Gnat Rm Implementation Defined Characteristics (`gcc/ada/doc/gnat_rm/implementation_defined_characteristics.rst`) .. _Implementation_Defined_Characteristics: ************************************** Implementation Defined Characteristics ************************************** In addition to the implementation dependent pragmas and attributes, and the implementation advice, there are a number of other Ada features that are potentially implementation dependent and are designated as implementation-defined. These are mentioned throughout the Ada Reference Manual, and are summarized in Annex M. A requirement for conforming Ada compilers is that they provide documentation describing how the implementation deals with each of these issues. In this chapter you will find each point in Annex M listed, followed by a description of how GNAT handles the implementation dependence. You can use this chapter as a guide to minimizing implementation dependent features in your programs if portability to other compilers and other operating systems is an important consideration. The numbers in each entry below correspond to the paragraph numbers in the Ada Reference Manual. * "Whether or not each recommendation given in Implementation Advice is followed. See 1.1.2(37)." See :ref:`Implementation_Advice`. * "Capacity limitations of the implementation. See 1.1.3(3)." The complexity of programs that can be processed is limited only by the total amount of available virtual memory, and disk space for the generated object files. * "Variations from the standard that are impractical to avoid given the implementation's execution environment. See 1.1.3(6)." There are no variations from the standard. * "Which code_statements cause external interactions. See 1.1.3(10)." Any *code_statement* can potentially cause external interactions. * "The coded representation for the text of an Ada program. See 2.1(4)." See separate section on source representation. * "The semantics of an Ada program whose text is not in Normalization Form C. See 2.1(4)." See separate section on source representation. * "The representation for an end of line. See 2.2(2)." See separate section on source representation. * "Maximum supported line length and lexical element length. See 2.2(15)." The maximum line length is 255 characters and the maximum length of a lexical element is also 255 characters. This is the default setting if not overridden by the use of compiler switch *-gnaty* (which sets the maximum to 79) or *-gnatyMnn* which allows the maximum line length to be specified to be any value up to 32767. The maximum length of a lexical element is the same as the maximum line length. * "Implementation defined pragmas. See 2.8(14)." See :ref:`Implementation_Defined_Pragmas`. * "Effect of pragma ``Optimize``. See 2.8(27)." Pragma ``Optimize``, if given with a ``Time`` or ``Space`` parameter, checks that the optimization flag is set, and aborts if it is not. * "The message string associated with the Assertion_Error exception raised by the failure of a predicate check if there is no applicable Predicate_Failure aspect. See 3.2.4(31)." In the case of a Dynamic_Predicate aspect, the string is "Dynamic_Predicate failed at ", where "" might be something like "foo.adb:123". The Static_Predicate case is handled analogously. * "The predefined integer types declared in ``Standard``. See 3.5.4(25)." .. tabularcolumns:: |l|L| ========================= ======================================= Type Representation ========================= ======================================= *Short_Short_Integer* 8-bit signed *Short_Integer* 16-bit signed *Integer* 32-bit signed *Long_Integer* 64-bit signed (on most 64-bit targets, depending on the C definition of long) 32-bit signed (on all other targets) *Long_Long_Integer* 64-bit signed *Long_Long_Long_Integer* 128-bit signed (on 64-bit targets) 64-bit signed (on 32-bit targets) ========================= ======================================= * "Any nonstandard integer types and the operators defined for them. See 3.5.4(26)." There are no nonstandard integer types. * "Any nonstandard real types and the operators defined for them. See 3.5.6(8)." There are no nonstandard real types. * "What combinations of requested decimal precision and range are supported for floating point types. See 3.5.7(7)." The precision and range are defined by the IEEE Standard for Floating-Point Arithmetic (IEEE 754-2019). * "The predefined floating point types declared in ``Standard``. See 3.5.7(16)." ====================== =============================================== Type Representation ====================== =============================================== *Short_Float* IEEE Binary32 (Single) *Float* IEEE Binary32 (Single) *Long_Float* IEEE Binary64 (Double) *Long_Long_Float* IEEE Binary64 (Double) on non-x86 architectures IEEE 80-bit Extended on x86 architecture ====================== =============================================== The default rounding mode specified by the IEEE 754 Standard is assumed both for static and dynamic computations (that is, round to nearest, ties to even). The input routines yield correctly rounded values for Short_Float, Float, and Long_Float at least. The output routines can compute up to twice as many exact digits as the value of ``T'Digits`` for any type, for example 30 digits for Long_Float; if more digits are requested, zeros are printed. * "The small of an ordinary fixed point type. See 3.5.9(8)." The small is the largest power of two that does not exceed the delta. * "What combinations of small, range, and digits are supported for fixed point types. See 3.5.9(10)." For an ordinary fixed point type, on 32-bit platforms, the small must lie in 2.0**(-80) .. 2.0**80 and the range in -9.0E+36 .. 9.0E+36; any combination is permitted that does not result in a mantissa larger than 63 bits. On 64-bit platforms, the small must lie in 2.0**(-127) .. 2.0**127 and the range in -1.0E+76 .. 1.0E+76; any combination is permitted that does not result in a mantissa larger than 63 bits, and any combination is permitted that results in a mantissa between 64 and 127 bits if the small is the ratio of two integers that lie in 1 .. 2.0**127. If the small is the ratio of two integers with 64-bit magnitude on 32-bit platforms and 128-bit magnitude on 64-bit platforms, which is the case if no ``small`` clause is provided, then the operations of the fixed point type are entirely implemented by means of integer instructions. In the other cases, some operations, in particular input and output, may be implemented by means of floating-point instructions and may be affected by accuracy issues on architectures other than x86. For a decimal fixed point type, on 32-bit platforms, the small must lie in 1.0E-18 .. 1.0E+18 and the digits in 1 .. 18. On 64-bit platforms, the small must lie in 1.0E-38 .. 1.0E+38 and the digits in 1 .. 38. * "The result of ``Tags.Expanded_Name`` for types declared within an unnamed *block_statement*. See 3.9(10)." Block numbers of the form :samp:`B{nnn}`, where *nnn* is a decimal integer are allocated. * "The sequence of characters of the value returned by Tags.Expanded_Name (respectively, Tags.Wide_Expanded_Name) when some of the graphic characters of Tags.Wide_Wide_Expanded_Name are not defined in Character (respectively, Wide_Character). See 3.9(10.1)." This is handled in the same way as the implementation-defined behavior referenced in A.4.12(34). * "Implementation-defined attributes. See 4.1.4(12)." See :ref:`Implementation_Defined_Attributes`. * "The value of the parameter to Empty for some container aggregates. See 4.3.5(40)." As per the suggestion given in the Annotated Ada RM, the default value of the formal parameter is used if one exists and zero is used otherwise. * "The maximum number of chunks for a parallel reduction expression without a chunk_specification. See 4.5.10(21)." Feature unimplemented. * "Rounding of real static expressions which are exactly half-way between two machine numbers. See 4.9(38)." Round to even is used in all such cases. * "The maximum number of chunks for a parallel generalized iterator without a chunk_specification. See 5.5.2(10)." Feature unimplemented. * "The number of chunks for an array component iterator. See 5.5.2(11)." Feature unimplemented. * "Any extensions of the Global aspect. See 6.1.2(43)." Feature unimplemented. * "The circumstances the implementation passes in the null value for a view conversion of an access type used as an out parameter. See 6.4.1(19)." Difficult to characterize. * "Any extensions of the Default_Initial_Condition aspect. See 7.3.3(11)." SPARK allows specifying *null* as the Default_Initial_Condition aspect of a type. See the SPARK reference manual for further details. * "Any implementation-defined time types. See 9.6(6)." There are no implementation-defined time types. * "The time base associated with relative delays. See 9.6(20)." See 9.6(20). The time base used is that provided by the C library function ``gettimeofday``. * "The time base of the type ``Calendar.Time``. See 9.6(23)." The time base used is that provided by the C library function ``gettimeofday``. * "The time zone used for package ``Calendar`` operations. See 9.6(24)." The time zone used by package ``Calendar`` is the current system time zone setting for local time, as accessed by the C library function ``localtime``. * "Any limit on *delay_until_statements* of *select_statements*. See 9.6(29)." There are no such limits. * "The result of Calendar.Formatting.Image if its argument represents more than 100 hours. See 9.6.1(86)." Calendar.Time_Error is raised. * "Implementation-defined conflict check policies. See 9.10.1(5)." There are no implementation-defined conflict check policies. * "The representation for a compilation. See 10.1(2)." A compilation is represented by a sequence of files presented to the compiler in a single invocation of the *gcc* command. * "Any restrictions on compilations that contain multiple compilation_units. See 10.1(4)." No single file can contain more than one compilation unit, but any sequence of files can be presented to the compiler as a single compilation. * "The mechanisms for creating an environment and for adding and replacing compilation units. See 10.1.4(3)." See separate section on compilation model. * "The manner of explicitly assigning library units to a partition. See 10.2(2)." If a unit contains an Ada main program, then the Ada units for the partition are determined by recursive application of the rules in the Ada Reference Manual section 10.2(2-6). In other words, the Ada units will be those that are needed by the main program, and then this definition of need is applied recursively to those units, and the partition contains the transitive closure determined by this relationship. In short, all the necessary units are included, with no need to explicitly specify the list. If additional units are required, e.g., by foreign language units, then all units must be mentioned in the context clause of one of the needed Ada units. If the partition contains no main program, or if the main program is in a language other than Ada, then GNAT provides the binder options *-z* and *-n* respectively, and in this case a list of units can be explicitly supplied to the binder for inclusion in the partition (all units needed by these units will also be included automatically). For full details on the use of these options, refer to *GNAT Make Program gnatmake* in the :title:`GNAT User's Guide`. * "The implementation-defined means, if any, of specifying which compilation units are needed by a given compilation unit. See 10.2(2)." The units needed by a given compilation unit are as defined in the Ada Reference Manual section 10.2(2-6). There are no implementation-defined pragmas or other implementation-defined means for specifying needed units. * "The manner of designating the main subprogram of a partition. See 10.2(7)." The main program is designated by providing the name of the corresponding :file:`ALI` file as the input parameter to the binder. * "The order of elaboration of *library_items*. See 10.2(18)." The first constraint on ordering is that it meets the requirements of Chapter 10 of the Ada Reference Manual. This still leaves some implementation-dependent choices, which are resolved by analyzing the elaboration code of each unit and identifying implicit elaboration-order dependencies. * "Parameter passing and function return for the main subprogram. See 10.2(21)." The main program has no parameters. It may be a procedure, or a function returning an integer type. In the latter case, the returned integer value is the return code of the program (overriding any value that may have been set by a call to ``Ada.Command_Line.Set_Exit_Status``). * "The mechanisms for building and running partitions. See 10.2(24)." GNAT itself supports programs with only a single partition. The PolyORB product (which also includes an implementation of the PCS) provides a completely flexible method for building and running programs consisting of multiple partitions. **NB!** See the note in :ref:`polyORB` regarding the lifetime of this product. * "The details of program execution, including program termination. See 10.2(25)." See separate section on compilation model. * "The semantics of any non-active partitions supported by the implementation. See 10.2(28)." Passive partitions are supported on targets where shared memory is provided by the operating system. **NB!** See the note in :ref:`polyORB` regarding the lifetime of this product. * "The information returned by ``Exception_Message``. See 11.4.1(10)." Exception message returns the null string unless a specific message has been passed by the program. * "The result of ``Exceptions.Exception_Name`` for types declared within an unnamed *block_statement*. See 11.4.1(12)." Blocks have implementation defined names of the form :samp:`B{nnn}` where *nnn* is an integer. * "The information returned by ``Exception_Information``. See 11.4.1(13)." ``Exception_Information`` returns a string in the following format:: *Exception_Name:* nnnnn *Message:* mmmmm *PID:* ppp *Load address:* 0xhhhh *Call stack traceback locations:* 0xhhhh 0xhhhh 0xhhhh ... 0xhhh where * ``nnnn`` is the fully qualified name of the exception in all upper case letters. This line is always present. * ``mmmm`` is the message (this line present only if message is non-null) * ``ppp`` is the Process Id value as a decimal integer (this line is present only if the Process Id is nonzero). Currently we are not making use of this field. * The Load address line, the Call stack traceback locations line and the following values are present only if at least one traceback location was recorded. The Load address indicates the address at which the main executable was loaded; this line may not be present if operating system hasn't relocated the main executable. The values are given in C style format, with lower case letters for a-f, and only as many digits present as are necessary. The line terminator sequence at the end of each line, including the last line is a single ``LF`` character (``16#0A#``). * "The sequence of characters of the value returned by Exceptions.Exception_Name (respectively, Exceptions.Wide_Exception_Name) when some of the graphic characters of Exceptions.Wide_Wide_Exception_Name are not defined in Character (respectively, Wide_Character). See 11.4.1(12.1)." This is handled in the same way as the implementation-defined behavior referenced in A.4.12(34). * "The information returned by Exception_Information. See 11.4.1(13)." The exception name and the source location at which the exception was raised are included. * "Implementation-defined policy_identifiers and assertion_aspect_marks allowed in a pragma Assertion_Policy. See 11.4.2(9)." Implementation-defined assertion_aspect_marks include Assert_And_Cut, Assume, Contract_Cases, Debug, Ghost, Initial_Condition, Loop_Invariant, Loop_Variant, Postcondition, Precondition, Predicate, Refined_Post, Statement_Assertions, and Subprogram_Variant. Implementation-defined policy_identifiers include Disable and Suppressible. * "The default assertion policy. See 11.4.2(10)." The default assertion policy is Ignore, although this can be overridden via compiler switches such as "-gnata". * "Implementation-defined check names. See 11.5(27)." The implementation-defined check names include Alignment_Check, Container_Checks, Duplicated_Tag_Check, Predicate_Check, Raise_Check, Tampering_Check, and Validity_Check. In addition, a user program can add implementation-defined check names by means of the pragma Check_Name. See the description of pragma ``Suppress`` for details. * "Existence and meaning of second parameter of pragma Unsuppress. See 11.5(27.1)." The legality rules for and semantics of the second parameter of pragma Unsuppress match those for the second argument of pragma Suppress. * "The cases that cause conflicts between the representation of the ancestors of a type_declaration. See 13.1(13.1)." No such cases exist. * "The interpretation of each representation aspect. See 13.1(20)." See separate section on data representations. * "Any restrictions placed upon the specification of representation aspects. See 13.1(20)." See separate section on data representations. * "Implementation-defined aspects, including the syntax for specifying such aspects and the legality rules for such aspects. See 13.1.1(38)." See :ref:`Implementation_Defined_Aspects`. * "The set of machine scalars. See 13.3(8.1)." See separate section on data representations. * "The meaning of ``Size`` for indefinite subtypes. See 13.3(48)." The Size attribute of an indefinite subtype is not less than the Size attribute of any object of that type. * "The meaning of Object_Size for indefinite subtypes. See 13.3(58)." The Object_Size attribute of an indefinite subtype is not less than the Object_Size attribute of any object of that type. * "The default external representation for a type tag. See 13.3(75)." The default external representation for a type tag is the fully expanded name of the type in upper case letters. * "What determines whether a compilation unit is the same in two different partitions. See 13.3(76)." A compilation unit is the same in two different partitions if and only if it derives from the same source file. * "Implementation-defined components. See 13.5.1(15)." The only implementation defined component is the tag for a tagged type, which contains a pointer to the dispatching table. * "If ``Word_Size`` = ``Storage_Unit``, the default bit ordering. See 13.5.3(5)." ``Word_Size`` does not equal ``Storage_Unit`` in this implementation. * "The contents of the visible part of package ``System``. See 13.7(2)." See the definition of package System in :file:`system.ads`. Note that two declarations are added to package System. .. code-block:: ada Max_Priority : constant Positive := Priority'Last; Max_Interrupt_Priority : constant Positive := Interrupt_Priority'Last; * "The range of Storage_Elements.Storage_Offset, the modulus of Storage_Elements.Storage_Element, and the declaration of Storage_Elements.Integer_Address. See 13.7.1(11)." See the definition of package System.Storage_Elements in :file:`s-stoele.ads`. * "The contents of the visible part of package ``System.Machine_Code``, and the meaning of *code_statements*. See 13.8(7)." See the definition and documentation in file :file:`s-maccod.ads`. * "The result of unchecked conversion for instances with scalar result types whose result is not defined by the language. See 13.9(11)." Unchecked conversion between types of the same size results in an uninterpreted transmission of the bits from one type to the other. If the types are of unequal sizes, then in the case of discrete types, a shorter source is first zero or sign extended as necessary, and a shorter target is simply truncated on the left. For all non-discrete types, the source is first copied if necessary to ensure that the alignment requirements of the target are met, then a pointer is constructed to the source value, and the result is obtained by dereferencing this pointer after converting it to be a pointer to the target type. Unchecked conversions where the target subtype is an unconstrained array are not permitted. If the target alignment is greater than the source alignment, then a copy of the result is made with appropriate alignment * "The result of unchecked conversion for instances with nonscalar result types whose result is not defined by the language. See 13.9(11)." See preceding definition for the scalar result case. * "Whether or not the implementation provides user-accessible names for the standard pool type(s). See 13.11(17)." There are 3 different standard pools used by the compiler when ``Storage_Pool`` is not specified depending whether the type is local to a subprogram or defined at the library level and whether ``Storage_Size``is specified or not. See documentation in the runtime library units ``System.Pool_Global``, ``System.Pool_Size`` and ``System.Pool_Local`` in files :file:`s-poosiz.ads`, :file:`s-pooglo.ads` and :file:`s-pooloc.ads` for full details on the default pools used. All these pools are accessible by means of `with`\ ing these units. * "The meaning of ``Storage_Size`` when neither the Storage_Size nor the Storage_Pool is specified for an access type. See 13.11(18)." ``Storage_Size`` is measured in storage units, and refers to the total space available for an access type collection, or to the primary stack space for a task. * "The effect of specifying aspect Default_Storage_Pool on an instance of a language-defined generic unit. See 13.11.3(5)." Instances of language-defined generic units are treated the same as other instances with respect to the Default_Storage_Pool aspect. * "Implementation-defined restrictions allowed in a pragma ``Restrictions``. See 13.12(8.7)." See :ref:`Standard_and_Implementation_Defined_Restrictions`. * "The consequences of violating limitations on ``Restrictions`` pragmas. See 13.12(9)." Restrictions that can be checked at compile time are enforced at compile time; violations are illegal. For other restrictions, any violation during program execution results in erroneous execution. * "Implementation-defined usage profiles allowed in a pragma Profile. See 13.12(15)." See :ref:`Implementation_Defined_Pragmas`. * "The contents of the stream elements read and written by the Read and Write attributes of elementary types. See 13.13.2(9)." The representation is the in-memory representation of the base type of the type, using the number of bits corresponding to the ``type'Size`` value, and the natural ordering of the machine. * "The names and characteristics of the numeric subtypes declared in the visible part of package ``Standard``. See A.1(3)." See items describing the integer and floating-point types supported. * "The values returned by Strings.Hash. See A.4.9(3)." This hash function has predictable collisions and is subject to equivalent substring attacks. It is not suitable for construction of a hash table keyed on possibly malicious user input. * "The value returned by a call to a Text_Buffer Get procedure if any character in the returned sequence is not defined in Character. See A.4.12(34)." The contents of a buffer is represented internally as a UTF_8 string. The value return by Text_Buffer.Get is the result of passing that UTF_8 string to UTF_Encoding.Strings.Decode. * "The value returned by a call to a Text_Buffer Wide_Get procedure if any character in the returned sequence is not defined in Wide_Character. See A.4.12(34)." The contents of a buffer is represented internally as a UTF_8 string. The value return by Text_Buffer.Wide_Get is the result of passing that UTF_8 string to UTF_Encoding.Wide_Strings.Decode. * "The accuracy actually achieved by the elementary functions. See A.5.1(1)." The elementary functions correspond to the functions available in the C library. Only fast math mode is implemented. * "The sign of a zero result from some of the operators or functions in ``Numerics.Generic_Elementary_Functions``, when ``Float_Type'Signed_Zeros`` is ``True``. See A.5.1(46)." The sign of zeroes follows the requirements of the IEEE 754 standard on floating-point. * "The value of ``Numerics.Float_Random.Max_Image_Width``. See A.5.2(27)." Maximum image width is 6864, see library file :file:`s-rannum.ads`. * "The value of ``Numerics.Discrete_Random.Max_Image_Width``. See A.5.2(27)." Maximum image width is 6864, see library file :file:`s-rannum.ads`. * "The string representation of a random number generator's state. See A.5.2(38)." The value returned by the Image function is the concatenation of the fixed-width decimal representations of the 624 32-bit integers of the state vector. * "The values of the ``Model_Mantissa``, ``Model_Emin``, ``Model_Epsilon``, ``Model``, ``Safe_First``, and ``Safe_Last`` attributes, if the Numerics Annex is not supported. See A.5.3(72)." Running the compiler with *-gnatS* to produce a listing of package ``Standard`` displays the values of these attributes. * "The value of ``Buffer_Size`` in ``Storage_IO``. See A.9(10)." All type representations are contiguous, and the ``Buffer_Size`` is the value of ``type'Size`` rounded up to the next storage unit boundary. * "External files for standard input, standard output, and standard error See A.10(5)." These files are mapped onto the files provided by the C streams libraries. See source file :file:`i-cstrea.ads` for further details. * "The accuracy of the value produced by ``Put``. See A.10.9(36)." If more digits are requested in the output than are represented by the precision of the value, zeroes are output in the corresponding least significant digit positions. * "Current size for a stream file for which positioning is not supported. See A.12.1(1.1)." Positioning is supported. * "The meaning of ``Argument_Count``, ``Argument``, and ``Command_Name``. See A.15(1)." These are mapped onto the ``argv`` and ``argc`` parameters of the main program in the natural manner. * "The interpretation of file names and directory names. See A.16(46)." These names are interpreted consistently with the underlying file system. * "The maximum value for a file size in Directories. See A.16(87)." Directories.File_Size'Last is equal to Long_Long_Integer'Last . * "The result for Directories.Size for a directory or special file. See A.16(93)." Name_Error is raised. * "The result for Directories.Modification_Time for a directory or special file. See A.16(93)." Name_Error is raised. * "The interpretation of a nonnull search pattern in Directories. See A.16(104)." When the ``Pattern`` parameter is not the null string, it is interpreted according to the syntax of regular expressions as defined in the ``GNAT.Regexp`` package. See :ref:`GNAT.Regexp_(g-regexp.ads)`. * "The results of a Directories search if the contents of the directory are altered while a search is in progress. See A.16(110)." The effect of a call to Get_Next_Entry is determined by the current state of the directory. * "The definition and meaning of an environment variable. See A.17(1)." This definition is determined by the underlying operating system. * "The circumstances where an environment variable cannot be defined. See A.17(16)." There are no such implementation-defined circumstances. * "Environment names for which Set has the effect of Clear. See A.17(17)." There are no such names. * "The value of Containers.Hash_Type'Modulus. The value of Containers.Count_Type'Last. See A.18.1(7)." Containers.Hash_Type'Modulus is 2**32. Containers.Count_Type'Last is 2**31 - 1. * "Implementation-defined convention names. See B.1(11)." The following convention names are supported .. tabularcolumns:: |l|L| ======================= ============================================================================== Convention Name Interpretation ======================= ============================================================================== *Ada* Ada *Ada_Pass_By_Copy* Allowed for any types except by-reference types such as limited records. Compatible with convention Ada, but causes any parameters with this convention to be passed by copy. *Ada_Pass_By_Reference* Allowed for any types except by-copy types such as scalars. Compatible with convention Ada, but causes any parameters with this convention to be passed by reference. *Assembler* Assembly language *Asm* Synonym for Assembler *Assembly* Synonym for Assembler *C* C *C_Pass_By_Copy* Allowed only for record types, like C, but also notes that record is to be passed by copy rather than reference. *COBOL* COBOL *C_Plus_Plus (or CPP)* C++ *Default* Treated the same as C *External* Treated the same as C *Fortran* Fortran *Intrinsic* For support of pragma ``Import`` with convention Intrinsic, see separate section on Intrinsic Subprograms. *Stdcall* Stdcall (used for Windows implementations only). This convention correspond to the WINAPI (previously called Pascal convention) C/C++ convention under Windows. A routine with this convention cleans the stack before exit. This pragma cannot be applied to a dispatching call. *DLL* Synonym for Stdcall *Win32* Synonym for Stdcall *Stubbed* Stubbed is a special convention used to indicate that the body of the subprogram will be entirely ignored. Any call to the subprogram is converted into a raise of the ``Program_Error`` exception. If a pragma ``Import`` specifies convention ``stubbed`` then no body need be present at all. This convention is useful during development for the inclusion of subprograms whose body has not yet been written. In addition, all otherwise unrecognized convention names are also treated as being synonymous with convention C. In all implementations, use of such other names results in a warning. ======================= ============================================================================== * "The meaning of link names. See B.1(36)." Link names are the actual names used by the linker. * "The manner of choosing link names when neither the link name nor the address of an imported or exported entity is specified. See B.1(36)." The default linker name is that which would be assigned by the relevant external language, interpreting the Ada name as being in all lower case letters. * "The effect of pragma ``Linker_Options``. See B.1(37)." The string passed to ``Linker_Options`` is presented uninterpreted as an argument to the link command, unless it contains ASCII.NUL characters. NUL characters if they appear act as argument separators, so for example .. code-block:: ada pragma Linker_Options ("-labc" & ASCII.NUL & "-ldef"); causes two separate arguments ``-labc`` and ``-ldef`` to be passed to the linker. The order of linker options is preserved for a given unit. The final list of options passed to the linker is in reverse order of the elaboration order. For example, linker options for a body always appear before the options from the corresponding package spec. * "The contents of the visible part of package ``Interfaces`` and its language-defined descendants. See B.2(1)." See files with prefix :file:`i-` in the distributed library. * "Implementation-defined children of package ``Interfaces``. The contents of the visible part of package ``Interfaces``. See B.2(11)." See files with prefix :file:`i-` in the distributed library. * "The definitions of certain types and constants in Interfaces.C. See B.3(41)." See source file :file:`i-c.ads`. * "The types ``Floating``, ``Long_Floating``, ``Binary``, ``Long_Binary``, ``Decimal_ Element``, and ``COBOL_Character``; and the initialization of the variables ``Ada_To_COBOL`` and ``COBOL_To_Ada``, in ``Interfaces.COBOL``. See B.4(50)." ===================== ==================================== COBOL Ada ===================== ==================================== *Floating* Float *Long_Floating* (Floating) Long_Float *Binary* Integer *Long_Binary* Long_Long_Integer *Decimal_Element* Character *COBOL_Character* Character ===================== ==================================== For initialization, see the file :file:`i-cobol.ads` in the distributed library. * "The types Fortran_Integer, Real, Double_Precision, and Character_Set in Interfaces.Fortran. See B.5(17)." See source file :file:`i-fortra.ads`. These types are derived, respectively, from Integer, Float, Long_Float, and Character. * "Implementation-defined intrinsic subprograms. See C.1(1)." See separate section on Intrinsic Subprograms. * "Any restrictions on a protected procedure or its containing type when an aspect Attach_handler or Interrupt_Handler is specified. See C.3.1(17)." There are no such restrictions. * "Any other forms of interrupt handler supported by the Attach_Handler and Interrupt_Handler aspects. See C.3.1(19)." There are no such forms. * "The semantics of some attributes and functions of an entity for which aspect Discard_Names is True. See C.5(7)." If Discard_Names is True for an enumeration type, the Image attribute provides the image of the Pos of the literal, and Value accepts Pos values. If both of the aspects``Discard_Names`` and ``No_Tagged_Streams`` are true for a tagged type, its Expanded_Name and External_Tag values are empty strings. This is useful to avoid exposing entity names at binary level. * "The modulus and size of Test_and_Set_Flag. See C.6.3(8)." The modulus is 2**8. The size is 8. * "The value used to represent the set value for Atomic_Test_and_Set. See C.6.3(10)." The value is 1. * "The result of the ``Task_Identification.Image`` attribute. See C.7.1(7)." The result of this attribute is a string that identifies the object or component that denotes a given task. If a variable ``Var`` has a task type, the image for this task will have the form :samp:`Var_{XXXXXXXX}`, where the suffix *XXXXXXXX* is the hexadecimal representation of the virtual address of the corresponding task control block. If the variable is an array of tasks, the image of each task will have the form of an indexed component indicating the position of a given task in the array, e.g., :samp:`Group(5)_{XXXXXXX}`. If the task is a component of a record, the image of the task will have the form of a selected component. These rules are fully recursive, so that the image of a task that is a subcomponent of a composite object corresponds to the expression that designates this task. If a task is created by an allocator, its image depends on the context. If the allocator is part of an object declaration, the rules described above are used to construct its image, and this image is not affected by subsequent assignments. If the allocator appears within an expression, the image includes only the name of the task type. If the configuration pragma Discard_Names is present, or if the restriction No_Implicit_Heap_Allocation is in effect, the image reduces to the numeric suffix, that is to say the hexadecimal representation of the virtual address of the control block of the task. * "The value of ``Current_Task`` when in a protected entry or interrupt handler. See C.7.1(17)." Protected entries or interrupt handlers can be executed by any convenient thread, so the value of ``Current_Task`` is undefined. * "Granularity of locking for Task_Attributes. See C.7.2(16)." No locking is needed if the formal type Attribute has the size and alignment of either Integer or System.Address and the bit representation of Initial_Value is all zeroes. Otherwise, locking is performed. * "The declarations of ``Any_Priority`` and ``Priority``. See D.1(11)." See declarations in file :file:`system.ads`. * "Implementation-defined execution resources. See D.1(15)." There are no implementation-defined execution resources. * "Whether, on a multiprocessor, a task that is waiting for access to a protected object keeps its processor busy. See D.2.1(3)." On a multi-processor, a task that is waiting for access to a protected object does not keep its processor busy. * "The affect of implementation defined execution resources on task dispatching. See D.2.1(9)." Tasks map to threads in the threads package used by GNAT. Where possible and appropriate, these threads correspond to native threads of the underlying operating system. * "Implementation-defined task dispatching policies. See D.2.2(3)." There are no implementation-defined task dispatching policies. * "The value of Default_Quantum in Dispatching.Round_Robin. See D.2.5(4)." The value is 10 milliseconds. * "Implementation-defined *policy_identifiers* allowed in a pragma ``Locking_Policy``. See D.3(4)." The two implementation defined policies permitted in GNAT are ``Inheritance_Locking`` and ``Concurrent_Readers_Locking``. On targets that support the ``Inheritance_Locking`` policy, locking is implemented by inheritance, i.e., the task owning the lock operates at a priority equal to the highest priority of any task currently requesting the lock. On targets that support the ``Concurrent_Readers_Locking`` policy, locking is implemented with a read/write lock allowing multiple protected object functions to enter concurrently. * "Default ceiling priorities. See D.3(10)." The ceiling priority of protected objects of the type ``System.Interrupt_Priority'Last`` as described in the Ada Reference Manual D.3(10), * "The ceiling of any protected object used internally by the implementation. See D.3(16)." The ceiling priority of internal protected objects is ``System.Priority'Last``. * "Implementation-defined queuing policies. See D.4(1)." There are no implementation-defined queuing policies. * "Implementation-defined admission policies. See D.4.1(1)." There are no implementation-defined admission policies. * "Any operations that implicitly require heap storage allocation. See D.7(8)." The only operation that implicitly requires heap storage allocation is task creation. * "When restriction No_Dynamic_CPU_Assignment applies to a partition, the processor on which a task with a CPU value of a Not_A_Specific_CPU will execute. See D.7(10)." Unknown. * "When restriction No_Task_Termination applies to a partition, what happens when a task terminates. See D.7(15.1)." Execution is erroneous in that case. * "The behavior when restriction Max_Storage_At_Blocking is violated. See D.7(17)." Execution is erroneous in that case. * "The behavior when restriction Max_Asynchronous_Select_Nesting is violated. See D.7(18)." Execution is erroneous in that case. * "The behavior when restriction Max_Tasks is violated. See D.7(19)." Execution is erroneous in that case. * "Whether the use of pragma Restrictions results in a reduction in program code or data size or execution time. See D.7(20)." Yes it can, but the precise circumstances and properties of such reductions are difficult to characterize. * "The value of Barrier_Limit'Last in Synchronous_Barriers. See D.10.1(4)." Synchronous_Barriers.Barrier_Limit'Last is Integer'Last . * "When an aborted task that is waiting on a Synchronous_Barrier is aborted. See D.10.1(13)." Difficult to characterize. * "The value of Min_Handler_Ceiling in Execution_Time.Group_Budgets. See D.14.2(7)." See source file :file:`a-etgrbu.ads`. * "The value of CPU_Range'Last in System.Multiprocessors. See D.16(4)." See source file :file:`s-multip.ads`. * "The processor on which the environment task executes in the absence of a value for the aspect CPU. See D.16(13)." Unknown. * "The means for creating and executing distributed programs. See E(5)." The PolyORB product provides means creating and executing distributed programs. **NB!** See the note in :ref:`polyORB` regarding the lifetime of this product. * "Any events that can result in a partition becoming inaccessible. See E.1(7)." See the PolyORB user guide for full details on such events. **NB!** Consider the note in :ref:`polyORB` regarding the lifetime of this product. * "The scheduling policies, treatment of priorities, and management of shared resources between partitions in certain cases. See E.1(11)." See the PolyORB user guide for full details on these aspects of multi-partition execution. **NB!** Consider the note in :ref:`polyORB` regarding the lifetime of this product. * "Whether the execution of the remote subprogram is immediately aborted as a result of cancellation. See E.4(13)." See the PolyORB user guide for details on the effect of abort in a distributed application. **NB!** Consider the note in :ref:`polyORB` regarding the lifetime of this product. * "The range of type System.RPC.Partition_Id. See E.5(14)." System.RPC.Partition_ID'Last is Integer'Last. See source file :file:`s-rpc.ads`. * "Implementation-defined interfaces in the PCS. See E.5(26)." See the PolyORB user guide for a full description of all implementation defined interfaces. **NB!** See the note in :ref:`polyORB` regarding the lifetime of this product. * "The values of named numbers in the package ``Decimal``. See F.2(7)." ==================== ========== Named Number Value ==================== ========== *Max_Scale* +18 *Min_Scale* -18 *Min_Delta* 1.0E-18 *Max_Delta* 1.0E+18 *Max_Decimal_Digits* 18 ==================== ========== * "The value of ``Max_Picture_Length`` in the package ``Text_IO.Editing``. See F.3.3(16)." 64 * "The value of ``Max_Picture_Length`` in the package ``Wide_Text_IO.Editing``. See F.3.4(5)." 64 * "The accuracy actually achieved by the complex elementary functions and by other complex arithmetic operations. See G.1(1)." Standard library functions are used for the complex arithmetic operations. Only fast math mode is currently supported. * "The sign of a zero result (or a component thereof) from any operator or function in ``Numerics.Generic_Complex_Types``, when ``Real'Signed_Zeros`` is True. See G.1.1(53)." The signs of zero values are as recommended by the relevant implementation advice. * "The sign of a zero result (or a component thereof) from any operator or function in ``Numerics.Generic_Complex_Elementary_Functions``, when ``Real'Signed_Zeros`` is ``True``. See G.1.2(45)." The signs of zero values are as recommended by the relevant implementation advice. * "Whether the strict mode or the relaxed mode is the default. See G.2(2)." The strict mode is the default. There is no separate relaxed mode. GNAT provides a highly efficient implementation of strict mode. * "The result interval in certain cases of fixed-to-float conversion. See G.2.1(10)." For cases where the result interval is implementation dependent, the accuracy is that provided by performing all operations in 64-bit IEEE floating-point format. * "The result of a floating point arithmetic operation in overflow situations, when the ``Machine_Overflows`` attribute of the result type is ``False``. See G.2.1(13)." Infinite and NaN values are produced as dictated by the IEEE floating-point standard. Note that on machines that are not fully compliant with the IEEE floating-point standard, such as Alpha, the *-mieee* compiler flag must be used for achieving IEEE conforming behavior (although at the cost of a significant performance penalty), so infinite and NaN values are properly generated. * "The result interval for division (or exponentiation by a negative exponent), when the floating point hardware implements division as multiplication by a reciprocal. See G.2.1(16)." Not relevant, division is IEEE exact. * "The definition of close result set, which determines the accuracy of certain fixed point multiplications and divisions. See G.2.3(5)." Operations in the close result set are performed using IEEE long format floating-point arithmetic. The input operands are converted to floating-point, the operation is done in floating-point, and the result is converted to the target type. * "Conditions on a *universal_real* operand of a fixed point multiplication or division for which the result shall be in the perfect result set. See G.2.3(22)." The result is only defined to be in the perfect result set if the result can be computed by a single scaling operation involving a scale factor representable in 64 bits. * "The result of a fixed point arithmetic operation in overflow situations, when the ``Machine_Overflows`` attribute of the result type is ``False``. See G.2.3(27)." Not relevant, ``Machine_Overflows`` is ``True`` for fixed-point types. * "The result of an elementary function reference in overflow situations, when the ``Machine_Overflows`` attribute of the result type is ``False``. See G.2.4(4)." IEEE infinite and Nan values are produced as appropriate. * "The value of the angle threshold, within which certain elementary functions, complex arithmetic operations, and complex elementary functions yield results conforming to a maximum relative error bound. See G.2.4(10)." Information on this subject is not yet available. * "The accuracy of certain elementary functions for parameters beyond the angle threshold. See G.2.4(10)." Information on this subject is not yet available. * "The result of a complex arithmetic operation or complex elementary function reference in overflow situations, when the ``Machine_Overflows`` attribute of the corresponding real type is ``False``. See G.2.6(5)." IEEE infinite and Nan values are produced as appropriate. * "The accuracy of certain complex arithmetic operations and certain complex elementary functions for parameters (or components thereof) beyond the angle threshold. See G.2.6(8)." Information on those subjects is not yet available. * "The accuracy requirements for the subprograms Solve, Inverse, Determinant, Eigenvalues and Eigensystem for type Real_Matrix. See G.3.1(81)." Information on those subjects is not yet available. * "The accuracy requirements for the subprograms Solve, Inverse, Determinant, Eigenvalues and Eigensystem for type Complex_Matrix. See G.3.2(149)." Information on those subjects is not yet available. * "The consequences of violating No_Hidden_Indirect_Globals. See H.4(23.9)." Execution is erroneous in that case. --- ### Gcc Ada Doc Gnat Rm Implementation Defined Pragmas (`gcc/ada/doc/gnat_rm/implementation_defined_pragmas.rst`) .. role:: switch(samp) .. _Implementation_Defined_Pragmas: ****************************** Implementation Defined Pragmas ****************************** Ada defines a set of pragmas that can be used to supply additional information to the compiler. These language defined pragmas are implemented in GNAT and work as described in the Ada Reference Manual. In addition, Ada allows implementations to define additional pragmas whose meaning is defined by the implementation. GNAT provides a number of these implementation-defined pragmas, which can be used to extend and enhance the functionality of the compiler. This section of the GNAT Reference Manual describes these additional pragmas. Note that any program using these pragmas might not be portable to other compilers (although GNAT implements this set of pragmas on all platforms). Therefore if portability to other compilers is an important consideration, the use of these pragmas should be minimized. Pragma Abort_Defer ================== .. index:: Deferring aborts Syntax: .. code-block:: ada pragma Abort_Defer; This pragma must appear at the start of the statement sequence of a handled sequence of statements (right after the ``begin``). It has the effect of deferring aborts for the sequence of statements (but not for the declarations or handlers, if any, associated with this statement sequence). This can also be useful for adding a polling point in Ada code, where asynchronous abort of tasks is checked when leaving the statement sequence, and is lighter than, for example, using ``delay 0.0;``, since with zero-cost exception handling, propagating exceptions (implicitly used to implement task abort) cannot be done reliably in an asynchronous way. An example of usage would be: .. code-block:: ada -- Add a polling point to check for task aborts begin pragma Abort_Defer; end; .. _Pragma-Abstract_State: Pragma Abstract_State ===================== Syntax: .. code-block:: ada pragma Abstract_State (ABSTRACT_STATE_LIST); ABSTRACT_STATE_LIST ::= null | STATE_NAME_WITH_OPTIONS | (STATE_NAME_WITH_OPTIONS {, STATE_NAME_WITH_OPTIONS} ) STATE_NAME_WITH_OPTIONS ::= STATE_NAME | (STATE_NAME with OPTION_LIST) OPTION_LIST ::= OPTION {, OPTION} OPTION ::= SIMPLE_OPTION | NAME_VALUE_OPTION SIMPLE_OPTION ::= Ghost | Synchronous NAME_VALUE_OPTION ::= Part_Of => ABSTRACT_STATE | External [=> EXTERNAL_PROPERTY_LIST] EXTERNAL_PROPERTY_LIST ::= EXTERNAL_PROPERTY | (EXTERNAL_PROPERTY {, EXTERNAL_PROPERTY} ) EXTERNAL_PROPERTY ::= Async_Readers [=> static_boolean_EXPRESSION] | Async_Writers [=> static_boolean_EXPRESSION] | Effective_Reads [=> static_boolean_EXPRESSION] | Effective_Writes [=> static_boolean_EXPRESSION] others => static_boolean_EXPRESSION STATE_NAME ::= defining_identifier ABSTRACT_STATE ::= name For the semantics of this pragma, see the entry for aspect ``Abstract_State`` in the SPARK 2014 Reference Manual, section 7.1.4. Pragma Ada_83 ============= Syntax: .. code-block:: ada pragma Ada_83; A configuration pragma that establishes Ada 83 mode for the unit to which it applies, regardless of the mode set by the command line switches. In Ada 83 mode, GNAT attempts to be as compatible with the syntax and semantics of Ada 83, as defined in the original Ada 83 Reference Manual as possible. In particular, the keywords added by Ada 95 and Ada 2005 are not recognized, optional package bodies are allowed, and generics may name types with unknown discriminants without using the ``(<>)`` notation. In addition, some but not all of the additional restrictions of Ada 83 are enforced. Like all configuration pragmas, if the pragma is placed before a library level package specification it is not propagated to the corresponding package body (see RM 10.1.5(8)); it must be added explicitly to the package body. Ada 83 mode is intended for two purposes. Firstly, it allows existing Ada 83 code to be compiled and adapted to GNAT with less effort. Secondly, it aids in keeping code backwards compatible with Ada 83. However, there is no guarantee that code that is processed correctly by GNAT in Ada 83 mode will in fact compile and execute with an Ada 83 compiler, since GNAT does not enforce all the additional checks required by Ada 83. Pragma Ada_95 ============= Syntax: .. code-block:: ada pragma Ada_95; A configuration pragma that establishes Ada 95 mode for the unit to which it applies, regardless of the mode set by the command line switches. This mode is set automatically for the ``Ada`` and ``System`` packages and their children, so you need not specify it in these contexts. This pragma is useful when writing a reusable component that itself uses Ada 95 features, but which is intended to be usable from either Ada 83 or Ada 95 programs. Like all configuration pragmas, if the pragma is placed before a library level package specification it is not propagated to the corresponding package body (see RM 10.1.5(8)); it must be added explicitly to the package body. Pragma Ada_05 ============= Syntax: .. code-block:: ada pragma Ada_05; pragma Ada_05 (local_NAME); A configuration pragma that establishes Ada 2005 mode for the unit to which it applies, regardless of the mode set by the command line switches. This pragma is useful when writing a reusable component that itself uses Ada 2005 features, but which is intended to be usable from either Ada 83 or Ada 95 programs. Like all configuration pragmas, if the pragma is placed before a library level package specification it is not propagated to the corresponding package body (see RM 10.1.5(8)); it must be added explicitly to the package body. The one argument form (which is not a configuration pragma) is used for managing the transition from Ada 95 to Ada 2005 in the run-time library. If an entity is marked as Ada_2005 only, then referencing the entity in Ada_83 or Ada_95 mode will generate a warning. In addition, in Ada_83 or Ada_95 mode, a preference rule is established which does not choose such an entity unless it is unambiguously specified. This avoids extra subprograms marked this way from generating ambiguities in otherwise legal pre-Ada_2005 programs. The one argument form is intended for exclusive use in the GNAT run-time library. Pragma Ada_2005 =============== Syntax: .. code-block:: ada pragma Ada_2005; This configuration pragma is a synonym for pragma Ada_05 and has the same syntax and effect. Pragma Ada_12 ============= Syntax: .. code-block:: ada pragma Ada_12; pragma Ada_12 (local_NAME); A configuration pragma that establishes Ada 2012 mode for the unit to which it applies, regardless of the mode set by the command line switches. This mode is set automatically for the ``Ada`` and ``System`` packages and their children, so you need not specify it in these contexts. This pragma is useful when writing a reusable component that itself uses Ada 2012 features, but which is intended to be usable from Ada 83, Ada 95, or Ada 2005 programs. Like all configuration pragmas, if the pragma is placed before a library level package specification it is not propagated to the corresponding package body (see RM 10.1.5(8)); it must be added explicitly to the package body. The one argument form, which is not a configuration pragma, is used for managing the transition from Ada 2005 to Ada 2012 in the run-time library. If an entity is marked as Ada_2012 only, then referencing the entity in any pre-Ada_2012 mode will generate a warning. In addition, in any pre-Ada_2012 mode, a preference rule is established which does not choose such an entity unless it is unambiguously specified. This avoids extra subprograms marked this way from generating ambiguities in otherwise legal pre-Ada_2012 programs. The one argument form is intended for exclusive use in the GNAT run-time library. Pragma Ada_2012 =============== Syntax: .. code-block:: ada pragma Ada_2012; This configuration pragma is a synonym for pragma Ada_12 and has the same syntax and effect. Pragma Ada_2022 =============== Syntax: .. code-block:: ada pragma Ada_2022; pragma Ada_2022 (local_NAME); A configuration pragma that establishes Ada 2022 mode for the unit to which it applies, regardless of the mode set by the command line switches. This mode is set automatically for the ``Ada`` and ``System`` packages and their children, so you need not specify it in these contexts. This pragma is useful when writing a reusable component that itself uses Ada 2022 features, but which is intended to be usable from Ada 83, Ada 95, Ada 2005 or Ada 2012 programs. Like all configuration pragmas, if the pragma is placed before a library level package specification it is not propagated to the corresponding package body (see RM 10.1.5(8)); it must be added explicitly to the package body. The one argument form, which is not a configuration pragma, is used for managing the transition from Ada 2012 to Ada 2022 in the run-time library. If an entity is marked as Ada_2022 only, then referencing the entity in any pre-Ada_2022 mode will generate a warning. In addition, in any pre-Ada_2012 mode, a preference rule is established which does not choose such an entity unless it is unambiguously specified. This avoids extra subprograms marked this way from generating ambiguities in otherwise legal pre-Ada_2022 programs. The one argument form is intended for exclusive use in the GNAT run-time library. Pragma Aggregate_Individually_Assign ==================================== Syntax: .. code-block:: ada pragma Aggregate_Individually_Assign; Where possible, GNAT will store the binary representation of a record aggregate in memory for space and performance reasons. This configuration pragma changes this behavior so that record aggregates are instead always converted into individual assignment statements. Pragma Allow_Integer_Address ============================ Syntax: .. code-block:: ada pragma Allow_Integer_Address; In almost all versions of GNAT, ``System.Address`` is a private type in accordance with the implementation advice in the RM. This means that integer values, in particular integer literals, are not allowed as address values. If the configuration pragma ``Allow_Integer_Address`` is given, then integer expressions may be used anywhere a value of type ``System.Address`` is required. The effect is to introduce an implicit unchecked conversion from the integer value to type ``System.Address``. The reverse case of using an address where an integer type is required is handled analogously. The following example compiles without errors: .. code-block:: ada pragma Allow_Integer_Address; with System; use System; package AddrAsInt is X : Integer; Y : Integer; for X'Address use 16#1240#; for Y use at 16#3230#; m : Address := 16#4000#; n : constant Address := 4000; p : constant Address := Address (X + Y); v : Integer := y'Address; w : constant Integer := Integer (Y'Address); type R is new integer; RR : R := 1000; Z : Integer; for Z'Address use RR; end AddrAsInt; Note that pragma ``Allow_Integer_Address`` is ignored if ``System.Address`` is not a private type. In implementations of ``GNAT`` where System.Address is a visible integer type, this pragma serves no purpose but is ignored rather than rejected to allow common sets of sources to be used in the two situations. .. _Pragma-Always_Terminates: Pragma Always_Terminates ======================== Syntax: .. code-block:: ada pragma Always_Terminates [ (boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Always_Terminates`` in the SPARK 2014 Reference Manual, section 6.1.11. .. _Pragma-Annotate: Pragma Annotate =============== Syntax:: pragma Annotate (IDENTIFIER [, IDENTIFIER {, ARG}] [, entity => local_NAME]); ARG ::= NAME | EXPRESSION This pragma is used to annotate programs. IDENTIFIER identifies the type of annotation. GNAT verifies that it is an identifier, but does not otherwise analyze it. The second optional identifier is also left unanalyzed, and by convention is used to control the action of the tool to which the annotation is addressed. The remaining ARG arguments can be either string literals or more generally expressions. String literals (and concatenations of string literals) are assumed to be either of type ``Standard.String`` or else ``Wide_String`` or ``Wide_Wide_String`` depending on the character literals they contain. All other kinds of arguments are analyzed as expressions, and must be unambiguous. The last argument if present must have the identifier ``Entity`` and GNAT verifies that a local name is given. The analyzed pragma is retained in the tree, but not otherwise processed by any part of the GNAT compiler, except to generate corresponding note lines in the generated ALI file. For the format of these note lines, see the compiler source file lib-writ.ads. This pragma is intended for use by external tools. The use of pragma Annotate does not affect the compilation process in any way. This pragma may be used as a configuration pragma. Pragma Assert ============= Syntax:: pragma Assert ( boolean_EXPRESSION [, string_EXPRESSION]); The effect of this pragma depends on whether the corresponding command line switch is set to activate assertions. The pragma expands into code equivalent to the following: .. code-block:: ada if assertions-enabled then if not boolean_EXPRESSION then System.Assertions.Raise_Assert_Failure (string_EXPRESSION); end if; end if; The string argument, if given, is the message that will be associated with the exception occurrence if the exception is raised. If no second argument is given, the default message is ``file``:``nnn``, where ``file`` is the name of the source file containing the assert, and ``nnn`` is the line number of the assert. Note that, as with the ``if`` statement to which it is equivalent, the type of the expression is either ``Standard.Boolean``, or any type derived from this standard type. Assert checks can be either checked or ignored. By default they are ignored. They will be checked if either the command line switch *-gnata* is used, or if an ``Assertion_Policy`` or ``Check_Policy`` pragma is used to enable ``Assert_Checks``. If assertions are ignored, then there is no run-time effect (and in particular, any side effects from the expression will not occur at run time). (The expression is still analyzed at compile time, and may cause types to be frozen if they are mentioned here for the first time). If assertions are checked, then the given expression is tested, and if it is ``False`` then ``System.Assertions.Raise_Assert_Failure`` is called which results in the raising of ``Assert_Failure`` with the given message. You should generally avoid side effects in the expression arguments of this pragma, because these side effects will turn on and off with the setting of the assertions mode, resulting in assertions that have an effect on the program. However, the expressions are analyzed for semantic correctness whether or not assertions are enabled, so turning assertions on and off cannot affect the legality of a program. Note that the implementation defined policy ``DISABLE``, given in a pragma ``Assertion_Policy``, can be used to suppress this semantic analysis. Note: this is a standard language-defined pragma in versions of Ada from 2005 on. In GNAT, it is implemented in all versions of Ada, and the DISABLE policy is an implementation-defined addition. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Assert_And_Cut ===================== Syntax:: pragma Assert_And_Cut ( boolean_EXPRESSION [, string_EXPRESSION]); The effect of this pragma is identical to that of pragma ``Assert``, except that in an ``Assertion_Policy`` pragma, the identifier ``Assert_And_Cut`` is used to control whether it is ignored or checked (or disabled). The intention is that this be used within a subprogram when the given test expression sums up all the work done so far in the subprogram, so that the rest of the subprogram can be verified (informally or formally) using only the entry preconditions, and the expression in this pragma. This allows dividing up a subprogram into sections for the purposes of testing or formal verification. The pragma also serves as useful documentation. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Assertion_Level ======================= Syntax:: pragma Assertion_Level (LEVEL_IDENTIFIER [, depends => DEPENDENCY_DESCRIPTOR]); DEPENDENCY_DESCRIPTOR ::= LEVEL_IDENTIFIER | LEVEL_IDENTIFIER_LIST LEVEL_IDENTIFIER_LIST ::= '[' LEVEL_IDENTIFIER {, LEVEL_IDENTIFIER} ']' For the semantics of this pragma, see the SPARK 2014 Reference Manual, section 11.4.3. Pragma Assertion_Policy ======================= Syntax:: pragma Assertion_Policy (CHECK | DISABLE | IGNORE | SUPPRESSIBLE); pragma Assertion_Policy ( ASSERTION_KIND => POLICY_IDENTIFIER {, ASSERTION_KIND => POLICY_IDENTIFIER}); ASSERTION_KIND ::= RM_ASSERTION_KIND | ID_ASSERTION_KIND | ASSERTION_LEVEL RM_ASSERTION_KIND ::= Assert | Static_Predicate | Dynamic_Predicate | Pre | Pre'Class | Post | Post'Class | Type_Invariant | Type_Invariant'Class | Default_Initial_Condition ID_ASSERTION_KIND ::= Assertions | Assert_And_Cut | Assume | Contract_Cases | Debug | Ghost | Initial_Condition | Invariant | Invariant'Class | Loop_Invariant | Loop_Variant | Postcondition | Precondition | Predicate | Refined_Post | Statement_Assertions | Subprogram_Variant POLICY_IDENTIFIER ::= Check | Disable | Ignore | Suppressible This is a standard Ada 2012 pragma that is available as an implementation-defined pragma in earlier versions of Ada. The assertion kinds ``RM_ASSERTION_KIND`` are those defined in the Ada standard. The assertion kinds ``ID_ASSERTION_KIND`` are implementation defined additions recognized by the GNAT compiler. Additionally the pragma can apply to an assertion level defined by the ``Assertion_Level`` pragma. For more details see the SPARK 2014 Reference Manual, section 11.4.2. The pragma applies in both cases to pragmas and aspects with matching names, e.g. ``Pre`` applies to the Pre aspect, and ``Precondition`` applies to both the ``Precondition`` pragma and the aspect ``Precondition``. Note that the identifiers for pragmas Pre_Class and Post_Class are Pre'Class and Post'Class (not Pre_Class and Post_Class), since these pragmas are intended to be identical to the corresponding aspects. If the policy is ``CHECK``, then assertions are enabled, i.e. the corresponding pragma or aspect is activated. If the policy is ``IGNORE``, then assertions are ignored, i.e. the corresponding pragma or aspect is deactivated. This pragma overrides the effect of the *-gnata* switch on the command line. If the policy is ``SUPPRESSIBLE``, then assertions are enabled by default, however, if the *-gnatp* switch is specified all assertions are ignored. The implementation defined policy ``DISABLE`` is like ``IGNORE`` except that it completely disables semantic checking of the corresponding pragma or aspect. This is useful when the pragma or aspect argument references subprograms in a with'ed package which is replaced by a dummy package for the final build. The implementation defined assertion kind ``Assertions`` applies to all assertion kinds. The form with no assertion kind given implies this choice, so it applies to all assertion kinds (RM defined, and implementation defined). The implementation defined assertion kind ``Statement_Assertions`` applies to ``Assert``, ``Assert_And_Cut``, ``Assume``, ``Loop_Invariant``, and ``Loop_Variant``. Pragma Assume ============= Syntax: :: pragma Assume ( boolean_EXPRESSION [, string_EXPRESSION]); The effect of this pragma is identical to that of pragma ``Assert``, except that in an ``Assertion_Policy`` pragma, the identifier ``Assume`` is used to control whether it is ignored or checked (or disabled). The intention is that this be used for assumptions about the external environment. So you cannot expect to verify formally or informally that the condition is met, this must be established by examining things outside the program itself. For example, we may have code that depends on the size of ``Long_Long_Integer`` being at least 64. So we could write: .. code-block:: ada pragma Assume (Long_Long_Integer'Size >= 64); This assumption cannot be proved from the program itself, but it acts as a useful run-time check that the assumption is met, and documents the need to ensure that it is met by reference to information outside the program. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Assume_No_Invalid_Values =============================== .. index:: Invalid representations .. index:: Invalid values Syntax: .. code-block:: ada pragma Assume_No_Invalid_Values (On | Off); This is a configuration pragma that controls the assumptions made by the compiler about the occurrence of invalid representations (invalid values) in the code. The default behavior (corresponding to an Off argument for this pragma), is to assume that values may in general be invalid unless the compiler can prove they are valid. Consider the following example: .. code-block:: ada V1 : Integer range 1 .. 10; V2 : Integer range 11 .. 20; ... for J in V2 .. V1 loop ... end loop; if V1 and V2 have valid values, then the loop is known at compile time not to execute since the lower bound must be greater than the upper bound. However in default mode, no such assumption is made, and the loop may execute. If ``Assume_No_Invalid_Values (On)`` is given, the compiler will assume that any occurrence of a variable other than in an explicit ``'Valid`` test always has a valid value, and the loop above will be optimized away. The use of ``Assume_No_Invalid_Values (On)`` is appropriate if you know your code is free of uninitialized variables and other possible sources of invalid representations, and may result in more efficient code. A program that accesses an invalid representation with this pragma in effect is erroneous, so no guarantees can be made about its behavior. It is peculiar though permissible to use this pragma in conjunction with validity checking (-gnatVa). In such cases, accessing invalid values will generally give an exception, though formally the program is erroneous so there are no guarantees that this will always be the case, and it is recommended that these two options not be used together. .. _Pragma-Async_Readers: Pragma Async_Readers ==================== Syntax: .. code-block:: ada pragma Async_Readers [ (static_boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Async_Readers`` in the SPARK 2014 Reference Manual, section 7.1.2. .. _Pragma-Async_Writers: Pragma Async_Writers ==================== Syntax: .. code-block:: ada pragma Async_Writers [ (static_boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Async_Writers`` in the SPARK 2014 Reference Manual, section 7.1.2. Pragma Attribute_Definition =========================== Syntax: :: pragma Attribute_Definition ([Attribute =>] ATTRIBUTE_DESIGNATOR, [Entity =>] LOCAL_NAME, [Expression =>] EXPRESSION | NAME); If ``Attribute`` is a known attribute name, this pragma is equivalent to the attribute definition clause: .. code-block:: ada for Entity'Attribute use Expression; If ``Attribute`` is not a recognized attribute name, the pragma is ignored, and a warning is emitted. This allows source code to be written that takes advantage of some new attribute, while remaining compilable with earlier compilers. Pragma C_Pass_By_Copy ===================== .. index:: Passing by copy Syntax: :: pragma C_Pass_By_Copy ([Max_Size =>] static_integer_EXPRESSION); Normally the default mechanism for passing C convention records to C convention subprograms is to pass them by reference, as suggested by RM B.3(69). Use the configuration pragma ``C_Pass_By_Copy`` to change this default, by requiring that record formal parameters be passed by copy if all of the following conditions are met: * The size of the record type does not exceed the value specified for ``Max_Size``. * The record type has ``Convention C``. * The formal parameter has this record type, and the subprogram has a foreign (non-Ada) convention. If these conditions are met the argument is passed by copy; i.e., in a manner consistent with what C expects if the corresponding formal in the C prototype is a struct (rather than a pointer to a struct). You can also pass records by copy by specifying the convention ``C_Pass_By_Copy`` for the record type, or by using the extended ``Import`` and ``Export`` pragmas, which allow specification of passing mechanisms on a parameter by parameter basis. Pragma Check ============ .. index:: Assertions .. index:: Named assertions Syntax: :: pragma Check ( [Name =>] CHECK_KIND, [Check =>] Boolean_EXPRESSION [, [Message =>] string_EXPRESSION] ); CHECK_KIND ::= IDENTIFIER | Pre'Class | Post'Class | Type_Invariant'Class | Invariant'Class This pragma is similar to the predefined pragma ``Assert`` except that an extra identifier argument is present. In conjunction with pragma ``Check_Policy``, this can be used to define groups of assertions that can be independently controlled. The identifier ``Assertion`` is special, it refers to the normal set of pragma ``Assert`` statements. Checks introduced by this pragma are normally deactivated by default. They can be activated either by the command line option *-gnata*, which turns on all checks, or individually controlled using pragma ``Check_Policy``. The identifiers ``Assertions`` and ``Statement_Assertions`` are not permitted as check kinds, since this would cause confusion with the use of these identifiers in ``Assertion_Policy`` and ``Check_Policy`` pragmas, where they are used to refer to sets of assertions. Pragma Check_Float_Overflow =========================== .. index:: Floating-point overflow Syntax: .. code-block:: ada pragma Check_Float_Overflow; In Ada, the predefined floating-point types (``Short_Float``, ``Float``, ``Long_Float``, ``Long_Long_Float``) are defined to be *unconstrained*. This means that even though each has a well-defined base range, an operation that delivers a result outside this base range is not required to raise an exception. This implementation permission accommodates the notion of infinities in IEEE floating-point, and corresponds to the efficient execution mode on most machines. GNAT will not raise overflow exceptions on these machines; instead it will generate infinities and NaN's as defined in the IEEE standard. Generating infinities, although efficient, is not always desirable. Often the preferable approach is to check for overflow, even at the (perhaps considerable) expense of run-time performance. This can be accomplished by defining your own constrained floating-point subtypes -- i.e., by supplying explicit range constraints -- and indeed such a subtype can have the same base range as its base type. For example: .. code-block:: ada subtype My_Float is Float range Float'Range; Here ``My_Float`` has the same range as ``Float`` but is constrained, so operations on ``My_Float`` values will be checked for overflow against this range. This style will achieve the desired goal, but it is often more convenient to be able to simply use the standard predefined floating-point types as long as overflow checking could be guaranteed. The ``Check_Float_Overflow`` configuration pragma achieves this effect. If a unit is compiled subject to this configuration pragma, then all operations on predefined floating-point types including operations on base types of these floating-point types will be treated as though those types were constrained, and overflow checks will be generated. The ``Constraint_Error`` exception is raised if the result is out of range. This mode can also be set by use of the compiler switch *-gnateF*. Pragma Check_Name ================= .. index:: Defining check names .. index:: Check names, defining Syntax: .. code-block:: ada pragma Check_Name (check_name_IDENTIFIER); This is a configuration pragma that defines a new implementation defined check name (unless IDENTIFIER matches one of the predefined check names, in which case the pragma has no effect). Check names are global to a partition, so if two or more configuration pragmas are present in a partition mentioning the same name, only one new check name is introduced. An implementation defined check name introduced with this pragma may be used in only three contexts: ``pragma Suppress``, ``pragma Unsuppress``, and as the prefix of a ``Check_Name'Enabled`` attribute reference. For any of these three cases, the check name must be visible. A check name is visible if it is in the configuration pragmas applying to the current unit, or if it appears at the start of any unit that is part of the dependency set of the current unit (e.g., units that are mentioned in ``with`` clauses). Check names introduced by this pragma are subject to control by compiler switches (in particular -gnatp) in the usual manner. Pragma Check_Policy =================== .. index:: Controlling assertions .. index:: Assertions, control .. index:: Check pragma control .. index:: Named assertions Syntax: :: pragma Check_Policy ([Name =>] CHECK_KIND, [Policy =>] POLICY_IDENTIFIER); pragma Check_Policy ( CHECK_KIND => POLICY_IDENTIFIER {, CHECK_KIND => POLICY_IDENTIFIER}); ASSERTION_KIND ::= RM_ASSERTION_KIND | ID_ASSERTION_KIND CHECK_KIND ::= IDENTIFIER | Pre'Class | Post'Class | Type_Invariant'Class | Invariant'Class The identifiers Name and Policy are not allowed as CHECK_KIND values. This avoids confusion between the two possible syntax forms for this pragma. POLICY_IDENTIFIER ::= ON | OFF | CHECK | DISABLE | IGNORE This pragma is used to set the checking policy for assertions (specified by aspects or pragmas), the ``Debug`` pragma, or additional checks to be checked using the ``Check`` pragma. It may appear either as a configuration pragma, or within a declarative part of package. In the latter case, it applies from the point where it appears to the end of the declarative region (like pragma ``Suppress``). The ``Check_Policy`` pragma is similar to the predefined ``Assertion_Policy`` pragma, and if the check kind corresponds to one of the assertion kinds that are allowed by ``Assertion_Policy``, then the effect is identical. If the first argument is Debug, then the policy applies to Debug pragmas, disabling their effect if the policy is ``OFF``, ``DISABLE``, or ``IGNORE``, and allowing them to execute with normal semantics if the policy is ``ON`` or ``CHECK``. In addition if the policy is ``DISABLE``, then the procedure call in ``Debug`` pragmas will be totally ignored and not analyzed semantically. Finally the first argument may be some other identifier than the above possibilities, in which case it controls a set of named assertions that can be checked using pragma ``Check``. For example, if the pragma: .. code-block:: ada pragma Check_Policy (Critical_Error, OFF); is given, then subsequent ``Check`` pragmas whose first argument is also ``Critical_Error`` will be disabled. The check policy is ``OFF`` to turn off corresponding checks, and ``ON`` to turn on corresponding checks. The default for a set of checks for which no ``Check_Policy`` is given is ``OFF`` unless the compiler switch *-gnata* is given, which turns on all checks by default. The check policy settings ``CHECK`` and ``IGNORE`` are recognized as synonyms for ``ON`` and ``OFF``. These synonyms are provided for compatibility with the standard ``Assertion_Policy`` pragma. The check policy setting ``DISABLE`` causes the second argument of a corresponding ``Check`` pragma to be completely ignored and not analyzed. Pragma Comment ============== Syntax: .. code-block:: ada pragma Comment (static_string_EXPRESSION); This is almost identical in effect to pragma ``Ident``. It allows the placement of a comment into the object file and hence into the executable file if the operating system permits such usage. The difference is that ``Comment``, unlike ``Ident``, has no limitations on placement of the pragma (it can be placed anywhere in the main source unit), and if more than one pragma is used, all comments are retained. Pragma Common_Object ==================== Syntax: :: pragma Common_Object ( [Internal =>] LOCAL_NAME [, [External =>] EXTERNAL_SYMBOL] [, [Size =>] EXTERNAL_SYMBOL] ); EXTERNAL_SYMBOL ::= IDENTIFIER | static_string_EXPRESSION This pragma enables the shared use of variables stored in overlaid linker areas corresponding to the use of ``COMMON`` in Fortran. The single object ``LOCAL_NAME`` is assigned to the area designated by the ``External`` argument. You may define a record to correspond to a series of fields. The ``Size`` argument is syntax checked in GNAT, but otherwise ignored. ``Common_Object`` is not supported on all platforms. If no support is available, then the code generator will issue a message indicating that the necessary attribute for implementation of this pragma is not available. .. _Compile_Time_Error: Pragma Compile_Time_Error ========================= Syntax: .. code-block:: ada pragma Compile_Time_Error (boolean_EXPRESSION, static_string_EXPRESSION); This pragma can be used to generate additional compile time error messages. It is particularly useful in generics, where errors can be issued for specific problematic instantiations. The first parameter is a boolean expression. The pragma ensures that the value of an expression is known at compile time, and has the value False. The set of expressions whose values are known at compile time includes all static boolean expressions, and also other values which the compiler can determine at compile time (e.g., the size of a record type set by an explicit size representation clause, or the value of a variable which was initialized to a constant and is known not to have been modified). If these conditions are not met, an error message is generated using the value given as the second argument. This string value may contain embedded ASCII.LF characters to break the message into multiple lines. Pragma Compile_Time_Warning =========================== Syntax: .. code-block:: ada pragma Compile_Time_Warning (boolean_EXPRESSION, static_string_EXPRESSION); Same as pragma Compile_Time_Error, except a warning is issued instead of an error message. If switch *-gnatw_C* is used, a warning is only issued if the value of the expression is known to be True at compile time, not when the value of the expression is not known at compile time. Note that if this pragma is used in a package that is with'ed by a client, the client will get the warning even though it is issued by a with'ed package (normally warnings in with'ed units are suppressed, but this is a special exception to that rule). One typical use is within a generic where compile time known characteristics of formal parameters are tested, and warnings given appropriately. Another use with a first parameter of True is to warn a client about use of a package, for example that it is not fully implemented. In previous versions of the compiler, combining *-gnatwe* with Compile_Time_Warning resulted in a fatal error. Now the compiler always emits a warning. You can use :ref:`Compile_Time_Error` to force the generation of an error. Pragma Complete_Representation ============================== Syntax: .. code-block:: ada pragma Complete_Representation; This pragma must appear immediately within a record representation clause. Typical placements are before the first component clause or after the last component clause. The effect is to give an error message if any component is missing a component clause. This pragma may be used to ensure that a record representation clause is complete, and that this invariant is maintained if fields are added to the record in the future. Pragma Complex_Representation ============================= Syntax: :: pragma Complex_Representation ([Entity =>] LOCAL_NAME); The ``Entity`` argument must be the name of a record type which has two fields of the same floating-point type. The effect of this pragma is to force gcc to use the special internal complex representation form for this record, which may be more efficient. Note that this may result in the code for this type not conforming to standard ABI (application binary interface) requirements for the handling of record types. For example, in some environments, there is a requirement for passing records by pointer, and the use of this pragma may result in passing this type in floating-point registers. Pragma Component_Alignment ========================== .. index:: Alignments of components .. index:: Pragma Component_Alignment Syntax: :: pragma Component_Alignment ( [Form =>] ALIGNMENT_CHOICE [, [Name =>] type_LOCAL_NAME]); ALIGNMENT_CHOICE ::= Component_Size | Component_Size_4 | Storage_Unit | Default Specifies the alignment of components in array or record types. The meaning of the ``Form`` argument is as follows: .. index:: Component_Size (in pragma Component_Alignment) *Component_Size* Aligns scalar components and subcomponents of the array or record type on boundaries appropriate to their inherent size (naturally aligned). For example, 1-byte components are aligned on byte boundaries, 2-byte integer components are aligned on 2-byte boundaries, 4-byte integer components are aligned on 4-byte boundaries and so on. These alignment rules correspond to the normal rules for C compilers on all machines except the VAX. .. index:: Component_Size_4 (in pragma Component_Alignment) *Component_Size_4* Naturally aligns components with a size of four or fewer bytes. Components that are larger than 4 bytes are placed on the next 4-byte boundary. .. index:: Storage_Unit (in pragma Component_Alignment) *Storage_Unit* Specifies that array or record components are byte aligned, i.e., aligned on boundaries determined by the value of the constant ``System.Storage_Unit``. .. index:: Default (in pragma Component_Alignment) *Default* Specifies that array or record components are aligned on default boundaries, appropriate to the underlying hardware or operating system or both. The ``Default`` choice is the same as ``Component_Size`` (natural alignment). If the ``Name`` parameter is present, ``type_LOCAL_NAME`` must refer to a local record or array type, and the specified alignment choice applies to the specified type. The use of ``Component_Alignment`` together with a pragma ``Pack`` causes the ``Component_Alignment`` pragma to be ignored. The use of ``Component_Alignment`` together with a record representation clause is only effective for fields not specified by the representation clause. If the ``Name`` parameter is absent, the pragma can be used as either a configuration pragma, in which case it applies to one or more units in accordance with the normal rules for configuration pragmas, or it can be used within a declarative part, in which case it applies to types that are declared within this declarative part, or within any nested scope within this declarative part. In either case it specifies the alignment to be applied to any record or array type which has otherwise standard representation. If the alignment for a record or array type is not specified (using pragma ``Pack``, pragma ``Component_Alignment``, or a record rep clause), the GNAT uses the default alignment as described previously. .. _Pragma-Constant_After_Elaboration: Pragma Constant_After_Elaboration ================================= Syntax: .. code-block:: ada pragma Constant_After_Elaboration [ (static_boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Constant_After_Elaboration`` in the SPARK 2014 Reference Manual, section 3.3.1. .. _Pragma-Contract_Cases: Pragma Contract_Cases ===================== .. index:: Contract cases Syntax: .. code-block:: ada pragma Contract_Cases (CONTRACT_CASE {, CONTRACT_CASE}); CONTRACT_CASE ::= CASE_GUARD => CONSEQUENCE CASE_GUARD ::= boolean_EXPRESSION | others CONSEQUENCE ::= boolean_EXPRESSION The ``Contract_Cases`` pragma allows defining fine-grain specifications that can complement or replace the contract given by a precondition and a postcondition. Additionally, the ``Contract_Cases`` pragma can be used by testing and formal verification tools. The compiler checks its validity and, depending on the assertion policy at the point of declaration of the pragma, it may insert a check in the executable. For code generation, the contract cases .. code-block:: ada pragma Contract_Cases ( Cond1 => Pred1, Cond2 => Pred2); are equivalent to .. code-block:: ada C1 : constant Boolean := Cond1; -- evaluated at subprogram entry C2 : constant Boolean := Cond2; -- evaluated at subprogram entry pragma Precondition ((C1 and not C2) or (C2 and not C1)); pragma Postcondition (if C1 then Pred1); pragma Postcondition (if C2 then Pred2); The precondition ensures that one and only one of the case guards is satisfied on entry to the subprogram. The postcondition ensures that for the case guard that was True on entry, the corresponding consequence is True on exit. Other consequence expressions are not evaluated. A precondition ``P`` and postcondition ``Q`` can also be expressed as contract cases: .. code-block:: ada pragma Contract_Cases (P => Q); The placement and visibility rules for ``Contract_Cases`` pragmas are identical to those described for preconditions and postconditions. The compiler checks that boolean expressions given in case guards and consequences are valid, where the rules for case guards are the same as the rule for an expression in ``Precondition`` and the rules for consequences are the same as the rule for an expression in ``Postcondition``. In particular, attributes ``'Old`` and ``'Result`` can only be used within consequence expressions. The case guard for the last contract case may be ``others``, to denote any case not captured by the previous cases. The following is an example of use within a package spec: .. code-block:: ada package Math_Functions is ... function Sqrt (Arg : Float) return Float; pragma Contract_Cases (((Arg in 0.0 .. 99.0) => Sqrt'Result < 10.0, Arg >= 100.0 => Sqrt'Result >= 10.0, others => Sqrt'Result = 0.0)); ... end Math_Functions; The meaning of contract cases is that only one case should apply at each call, as determined by the corresponding case guard evaluating to True, and that the consequence for this case should hold when the subprogram returns. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Convention_Identifier ============================ .. index:: Conventions, synonyms Syntax: :: pragma Convention_Identifier ( [Name =>] IDENTIFIER, [Convention =>] convention_IDENTIFIER); This pragma provides a mechanism for supplying synonyms for existing convention identifiers. The ``Name`` identifier can subsequently be used as a synonym for the given convention in other pragmas (including for example pragma ``Import`` or another ``Convention_Identifier`` pragma). As an example of the use of this, suppose you had legacy code which used Fortran77 as the identifier for Fortran. Then the pragma: .. code-block:: ada pragma Convention_Identifier (Fortran77, Fortran); would allow the use of the convention identifier ``Fortran77`` in subsequent code, avoiding the need to modify the sources. As another example, you could use this to parameterize convention requirements according to systems. Suppose you needed to use ``Stdcall`` on windows systems, and ``C`` on some other system, then you could define a convention identifier ``Library`` and use a single ``Convention_Identifier`` pragma to specify which convention would be used system-wide. Pragma CPP_Class ================ .. index:: Interfacing with C++ Syntax: :: pragma CPP_Class ([Entity =>] LOCAL_NAME); The argument denotes an entity in the current declarative region that is declared as a record type. It indicates that the type corresponds to an externally declared C++ class type, and is to be laid out the same way that C++ would lay out the type. If the C++ class has virtual primitives then the record must be declared as a tagged record type. Types for which ``CPP_Class`` is specified do not have assignment or equality operators defined (such operations can be imported or declared as subprograms as required). Initialization is allowed only by constructor functions (see pragma ``CPP_Constructor``). Such types are implicitly limited if not explicitly declared as limited or derived from a limited type, and an error is issued in that case. See :ref:`Interfacing_to_C++` for related information. Note: Pragma ``CPP_Class`` is currently obsolete. It is supported for backward compatibility but its functionality is available using pragma ``Import`` with ``Convention`` = ``CPP``. Pragma CPP_Constructor ====================== .. index:: Interfacing with C++ Syntax: :: pragma CPP_Constructor ([Entity =>] LOCAL_NAME [, [External_Name =>] static_string_EXPRESSION ] [, [Link_Name =>] static_string_EXPRESSION ]); This pragma identifies an imported function (imported in the usual way with pragma ``Import``) as corresponding to a C++ constructor. If ``External_Name`` and ``Link_Name`` are not specified then the ``Entity`` argument is a name that must have been previously mentioned in a pragma ``Import`` with ``Convention`` = ``CPP``. Such name must be of one of the following forms: * **function** ``Fname`` **return** T` * **function** ``Fname`` **return** T'Class * **function** ``Fname`` (...) **return** T` * **function** ``Fname`` (...) **return** T'Class where ``T`` is a limited record type imported from C++ with pragma ``Import`` and ``Convention`` = ``CPP``. The first two forms import the default constructor, used when an object of type ``T`` is created on the Ada side with no explicit constructor. The latter two forms cover all the non-default constructors of the type. See the GNAT User's Guide for details. If no constructors are imported, it is impossible to create any objects on the Ada side and the type is implicitly declared abstract. Pragma ``CPP_Constructor`` is intended primarily for automatic generation using an automatic binding generator tool (such as the :switch:`-fdump-ada-spec` GCC switch). See :ref:`Interfacing_to_C++` for more related information. Note: The use of functions returning class-wide types for constructors is currently obsolete. They are supported for backward compatibility. The use of functions returning the type T leave the Ada sources more clear because the imported C++ constructors always return an object of type T; that is, they never return an object whose type is a descendant of type T. Pragma CPP_Virtual ================== .. index:: Interfacing to C++ This pragma is now obsolete and, other than generating a warning if warnings on obsolescent features are enabled, is completely ignored. It is retained for compatibility purposes. It used to be required to ensure compatibility with C++, but is no longer required for that purpose because GNAT generates the same object layout as the G++ compiler by default. See :ref:`Interfacing_to_C++` for related information. Pragma CPP_Vtable ================= .. index:: Interfacing with C++ This pragma is now obsolete and, other than generating a warning if warnings on obsolescent features are enabled, is completely ignored. It used to be required to ensure compatibility with C++, but is no longer required for that purpose because GNAT generates the same object layout as the G++ compiler by default. See :ref:`Interfacing_to_C++` for related information. Pragma CPU ========== Syntax: .. code-block:: ada pragma CPU (EXPRESSION); This pragma is standard in Ada 2012, but is available in all earlier versions of Ada as an implementation-defined pragma. See Ada 2012 Reference Manual for details. Pragma Deadline_Floor ===================== Syntax: .. code-block:: ada pragma Deadline_Floor (time_span_EXPRESSION); This pragma applies only to protected types and specifies the floor deadline inherited by a task when the task enters a protected object. It is effective only when the EDF scheduling policy is used. Pragma Debug ============ Syntax: :: pragma Debug ([CONDITION, ]PROCEDURE_CALL_WITHOUT_SEMICOLON); PROCEDURE_CALL_WITHOUT_SEMICOLON ::= PROCEDURE_NAME | PROCEDURE_PREFIX ACTUAL_PARAMETER_PART The procedure call argument has the syntactic form of an expression, meeting the syntactic requirements for pragmas. If debug pragmas are not enabled or if the condition is present and evaluates to False, this pragma has no effect. If debug pragmas are enabled, the semantics of the pragma is exactly equivalent to the procedure call statement corresponding to the argument with a terminating semicolon. Pragmas are permitted in sequences of declarations, so you can use pragma ``Debug`` to intersperse calls to debug procedures in the middle of declarations. Debug pragmas can be enabled either by use of the command line switch *-gnata* or by use of the pragma ``Check_Policy`` with a first argument of ``Debug``. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Debug_Policy =================== Syntax: .. code-block:: ada pragma Debug_Policy (CHECK | DISABLE | IGNORE | ON | OFF); This pragma is equivalent to a corresponding ``Check_Policy`` pragma with a first argument of ``Debug``. It is retained for historical compatibility reasons. .. _Pragma-Default_Initial_Condition: Pragma Default_Initial_Condition ================================ Syntax: .. code-block:: ada pragma Default_Initial_Condition [ (null | boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Default_Initial_Condition`` in the SPARK 2014 Reference Manual, section 7.3.3. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Default_Scalar_Storage_Order =================================== .. index:: Default_Scalar_Storage_Order .. index:: Scalar_Storage_Order Syntax: .. code-block:: ada pragma Default_Scalar_Storage_Order (High_Order_First | Low_Order_First); Normally if no explicit ``Scalar_Storage_Order`` is given for a record type or array type, then the scalar storage order defaults to the ordinary default for the target. But this default may be overridden using this pragma. The pragma may appear as a configuration pragma, or locally within a package spec or declarative part. In the latter case, it applies to all subsequent types declared within that package spec or declarative part. The following example shows the use of this pragma: .. code-block:: ada pragma Default_Scalar_Storage_Order (High_Order_First); with System; use System; package DSSO1 is type H1 is record a : Integer; end record; type L2 is record a : Integer; end record; for L2'Scalar_Storage_Order use Low_Order_First; type L2a is new L2; package Inner is type H3 is record a : Integer; end record; pragma Default_Scalar_Storage_Order (Low_Order_First); type L4 is record a : Integer; end record; end Inner; type H4a is new Inner.L4; type H5 is record a : Integer; end record; end DSSO1; In this example record types with names starting with *L* have `Low_Order_First` scalar storage order, and record types with names starting with *H* have ``High_Order_First``. Note that in the case of ``H4a``, the order is not inherited from the parent type. Only an explicitly set ``Scalar_Storage_Order`` gets inherited on type derivation. If this pragma is used as a configuration pragma which appears within a configuration pragma file (as opposed to appearing explicitly at the start of a single unit), then the binder will require that all units in a partition be compiled in a similar manner, other than run-time units, which are not affected by this pragma. Note that the use of this form is discouraged because it may significantly degrade the run-time performance of the software, instead the default scalar storage order ought to be changed only on a local basis. Pragma Default_Storage_Pool =========================== .. index:: Default_Storage_Pool Syntax: .. code-block:: ada pragma Default_Storage_Pool (storage_pool_NAME | null); This pragma is standard in Ada 2012, but is available in all earlier versions of Ada as an implementation-defined pragma. See Ada 2012 Reference Manual for details. .. _Pragma-Depends: Pragma Depends ============== Syntax: .. code-block:: ada pragma Depends (DEPENDENCY_RELATION); DEPENDENCY_RELATION ::= null | (DEPENDENCY_CLAUSE {, DEPENDENCY_CLAUSE}) DEPENDENCY_CLAUSE ::= OUTPUT_LIST =>[+] INPUT_LIST | NULL_DEPENDENCY_CLAUSE NULL_DEPENDENCY_CLAUSE ::= null => INPUT_LIST OUTPUT_LIST ::= OUTPUT | (OUTPUT {, OUTPUT}) INPUT_LIST ::= null | INPUT | (INPUT {, INPUT}) OUTPUT ::= NAME | FUNCTION_RESULT INPUT ::= NAME where FUNCTION_RESULT is a function Result attribute_reference For the semantics of this pragma, see the entry for aspect ``Depends`` in the SPARK 2014 Reference Manual, section 6.1.5. Pragma Detect_Blocking ====================== Syntax: .. code-block:: ada pragma Detect_Blocking; This is a standard pragma in Ada 2005, that is available in all earlier versions of Ada as an implementation-defined pragma. This is a configuration pragma that forces the detection of potentially blocking operations within a protected operation, and to raise Program_Error if that happens. Pragma Disable_Atomic_Synchronization ===================================== .. index:: Atomic Synchronization Syntax: :: pragma Disable_Atomic_Synchronization [(Entity)]; pragma Enable_Atomic_Synchronization [(Entity)]; Ada requires that accesses (reads or writes) of an atomic variable be regarded as synchronization points in the case of multiple tasks. Particularly in the case of multi-processors this may require special handling, e.g. the generation of memory barriers. This synchronization is performed by default, but can be turned off using pragma ``Disable_Atomic_Synchronization``. The ``Enable_Atomic_Synchronization`` pragma turns it back on. The placement and scope rules for these pragmas are the same as those for ``pragma Suppress``. In particular they can be used as configuration pragmas, or in a declaration sequence where they apply until the end of the scope. If an ``Entity`` argument is present, the action applies only to that entity. Pragma Dispatching_Domain ========================= Syntax: .. code-block:: ada pragma Dispatching_Domain (EXPRESSION); This pragma is standard in Ada 2012, but is available in all earlier versions of Ada as an implementation-defined pragma. See Ada 2012 Reference Manual for details. .. _Pragma-Effective_Reads: Pragma Effective_Reads ====================== Syntax: .. code-block:: ada pragma Effective_Reads [ (static_boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Effective_Reads`` in the SPARK 2014 Reference Manual, section 7.1.2. .. _Pragma-Effective_Writes: Pragma Effective_Writes ======================= Syntax: .. code-block:: ada pragma Effective_Writes [ (static_boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Effective_Writes`` in the SPARK 2014 Reference Manual, section 7.1.2. Pragma Elaboration_Checks ========================= .. index:: Elaboration control Syntax: .. code-block:: ada pragma Elaboration_Checks (Dynamic | Static); This is a configuration pragma which specifies the elaboration model to be used during compilation. For more information on the elaboration models of GNAT, consult the chapter on elaboration order handling in the *GNAT User's Guide*. The pragma may appear in the following contexts: * Configuration pragmas file * Prior to the context clauses of a compilation unit's initial declaration Any other placement of the pragma will result in a warning and the effects of the offending pragma will be ignored. If the pragma argument is ``Dynamic``, then the dynamic elaboration model is in effect. If the pragma argument is ``Static``, then the static elaboration model is in effect. Pragma Eliminate ================ .. index:: Elimination of unused subprograms Syntax: :: pragma Eliminate ( [ Unit_Name => ] IDENTIFIER | SELECTED_COMPONENT , [ Entity => ] IDENTIFIER | SELECTED_COMPONENT | STRING_LITERAL [, Source_Location => SOURCE_TRACE ] ); SOURCE_TRACE ::= STRING_LITERAL This pragma indicates that the given entity is not used in the program to be compiled and built, thus allowing the compiler to eliminate the code or data associated with the named entity. Any reference to an eliminated entity causes a compile-time or link-time error. The pragma has the following semantics, where ``U`` is the unit specified by the ``Unit_Name`` argument and ``E`` is the entity specified by the ``Entity`` argument: * ``E`` must be a subprogram that is explicitly declared either: * Within ``U``, or * Within a generic package that is instantiated in ``U``, or * As an instance of generic subprogram instantiated in ``U``. Otherwise the pragma is ignored. * If ``E`` is overloaded within ``U`` then, in the absence of a ``Source_Location`` argument, all overloadings are eliminated. * If ``E`` is overloaded within ``U`` and only some overloadings are to be eliminated, then each overloading to be eliminated must be specified in a corresponding pragma ``Eliminate`` with a ``Source_Location`` argument identifying the line where the declaration appears, as described below. * If ``E`` is declared as the result of a generic instantiation, then a ``Source_Location`` argument is needed, as described below. Pragma ``Eliminate`` allows a program to be compiled in a system-independent manner, so that unused entities are eliminated but without needing to modify the source text. Normally the required set of ``Eliminate`` pragmas is constructed automatically using the ``gnatelim`` tool. Any source file change that removes, splits, or adds lines may make the set of ``Eliminate`` pragmas invalid because their ``Source_Location`` argument values may get out of date. Pragma ``Eliminate`` may be used where the referenced entity is a dispatching operation. In this case all the subprograms to which the given operation can dispatch are considered to be unused (are never called as a result of a direct or a dispatching call). The string literal given for the source location specifies the line number of the declaration of the entity, using the following syntax for ``SOURCE_TRACE``: :: SOURCE_TRACE ::= SOURCE_REFERENCE [ LBRACKET SOURCE_TRACE RBRACKET ] LBRACKET ::= '[' RBRACKET ::= ']' SOURCE_REFERENCE ::= FILE_NAME : LINE_NUMBER LINE_NUMBER ::= DIGIT {DIGIT} Spaces around the colon in a ``SOURCE_REFERENCE`` are optional. The source trace that is given as the ``Source_Location`` must obey the following rules (or else the pragma is ignored), where ``U`` is the unit ``U`` specified by the ``Unit_Name`` argument and ``E`` is the subprogram specified by the ``Entity`` argument: * ``FILE_NAME`` is the short name (with no directory information) of the Ada source file for ``U``, using the required syntax for the underlying file system (e.g. case is significant if the underlying operating system is case sensitive). If ``U`` is a package and ``E`` is a subprogram declared in the package specification and its full declaration appears in the package body, then the relevant source file is the one for the package specification; analogously if ``U`` is a generic package. * If ``E`` is not declared in a generic instantiation (this includes generic subprogram instances), the source trace includes only one source line reference. ``LINE_NUMBER`` gives the line number of the occurrence of the declaration of ``E`` within the source file (as a decimal literal without an exponent or point). * If ``E`` is declared by a generic instantiation, its source trace (from left to right) starts with the source location of the declaration of ``E`` in the generic unit and ends with the source location of the instantiation, given in square brackets. This approach is applied recursively with nested instantiations: the rightmost (nested most deeply in square brackets) element of the source trace is the location of the outermost instantiation, and the leftmost element (that is, outside of any square brackets) is the location of the declaration of ``E`` in the generic unit. Examples: .. code-block:: ada pragma Eliminate (Pkg0, Proc); -- Eliminate (all overloadings of) Proc in Pkg0 pragma Eliminate (Pkg1, Proc, Source_Location => "pkg1.ads:8"); -- Eliminate overloading of Proc at line 8 in pkg1.ads -- Assume the following file contents: -- gen_pkg.ads -- 1: generic -- 2: type T is private; -- 3: package Gen_Pkg is -- 4: procedure Proc(N : T); -- ... ... -- ... end Gen_Pkg; -- -- q.adb -- 1: with Gen_Pkg; -- 2: procedure Q is -- 3: package Inst_Pkg is new Gen_Pkg(Integer); -- ... -- No calls on Inst_Pkg.Proc -- ... end Q; -- The following pragma eliminates Inst_Pkg.Proc from Q pragma Eliminate (Q, Proc, Source_Location => "gen_pkg.ads:4[q.adb:3]"); Pragma Enable_Atomic_Synchronization ==================================== .. index:: Atomic Synchronization Syntax: :: pragma Enable_Atomic_Synchronization [(Entity)]; Reenables atomic synchronization; see ``pragma Disable_Atomic_Synchronization`` for details. Pragma Exceptional_Cases ======================== .. index:: Exceptional_Cases Syntax: :: pragma Exceptional_Cases (EXCEPTIONAL_CASE_LIST); EXCEPTIONAL_CASE_LIST ::= EXCEPTIONAL_CASE {, EXCEPTIONAL_CASE} EXCEPTIONAL_CASE ::= exception_choice {'|' exception_choice} => CONSEQUENCE CONSEQUENCE ::= Boolean_expression For the semantics of this aspect, see the SPARK 2014 Reference Manual, section 6.1.9. Pragma Exit_Cases ================= .. index:: Exit_Cases Syntax: :: pragma Exit_Cases (EXIT_CASE_LIST); EXIT_CASE_LIST ::= EXIT_CASE {, EXIT_CASE} EXIT_CASE ::= GUARD => EXIT_KIND EXIT_KIND ::= Normal_Return | Exception_Raised | (Exception_Raised => exception_name) | Program_Exit GUARD ::= Boolean_expression For the semantics of this aspect, see the SPARK 2014 Reference Manual, section 6.1.10. Pragma Export_Function ====================== .. index:: Argument passing mechanisms Syntax: :: pragma Export_Function ( [Internal =>] LOCAL_NAME [, [External =>] EXTERNAL_SYMBOL] [, [Parameter_Types =>] PARAMETER_TYPES] [, [Result_Type =>] result_SUBTYPE_MARK] [, [Mechanism =>] MECHANISM] [, [Result_Mechanism =>] MECHANISM_NAME]); EXTERNAL_SYMBOL ::= IDENTIFIER | static_string_EXPRESSION | "" PARAMETER_TYPES ::= null | TYPE_DESIGNATOR {, TYPE_DESIGNATOR} TYPE_DESIGNATOR ::= subtype_NAME | subtype_Name ' Access MECHANISM ::= MECHANISM_NAME | (MECHANISM_ASSOCIATION {, MECHANISM_ASSOCIATION}) MECHANISM_ASSOCIATION ::= [formal_parameter_NAME =>] MECHANISM_NAME MECHANISM_NAME ::= Value | Reference Use this pragma to make a function externally callable and optionally provide information on mechanisms to be used for passing parameter and result values. We recommend, for the purposes of improving portability, this pragma always be used in conjunction with a separate pragma ``Export``, which must precede the pragma ``Export_Function``. GNAT does not require a separate pragma ``Export``, but if none is present, ``Convention Ada`` is assumed, which is usually not what is wanted, so it is usually appropriate to use this pragma in conjunction with a ``Export`` or ``Convention`` pragma that specifies the desired foreign convention. Pragma ``Export_Function`` (and ``Export``, if present) must appear in the same declarative region as the function to which they apply. The ``internal_name`` must uniquely designate the function to which the pragma applies. If more than one function name exists of this name in the declarative part you must use the ``Parameter_Types`` and ``Result_Type`` parameters to achieve the required unique designation. The `subtype_mark`\ s in these parameters must exactly match the subtypes in the corresponding function specification, using positional notation to match parameters with subtype marks. The form with an ``'Access`` attribute can be used to match an anonymous access parameter. .. index:: Suppressing external name Special treatment is given if the EXTERNAL is an explicit null string or a static string expressions that evaluates to the null string. In this case, no external name is generated. This form still allows the specification of parameter mechanisms. Pragma Export_Object ==================== Syntax: :: pragma Export_Object ( [Internal =>] LOCAL_NAME [, [External =>] EXTERNAL_SYMBOL] [, [Size =>] EXTERNAL_SYMBOL]); EXTERNAL_SYMBOL ::= IDENTIFIER | static_string_EXPRESSION This pragma designates an object as exported, and apart from the extended rules for external symbols, is identical in effect to the use of the normal ``Export`` pragma applied to an object. You may use a separate Export pragma (and you probably should from the point of view of portability), but it is not required. ``Size`` is syntax checked, but otherwise ignored by GNAT. Pragma Export_Procedure ======================= Syntax: :: pragma Export_Procedure ( [Internal =>] LOCAL_NAME [, [External =>] EXTERNAL_SYMBOL] [, [Parameter_Types =>] PARAMETER_TYPES] [, [Mechanism =>] MECHANISM]); EXTERNAL_SYMBOL ::= IDENTIFIER | static_string_EXPRESSION | "" PARAMETER_TYPES ::= null | TYPE_DESIGNATOR {, TYPE_DESIGNATOR} TYPE_DESIGNATOR ::= subtype_NAME | subtype_Name ' Access MECHANISM ::= MECHANISM_NAME | (MECHANISM_ASSOCIATION {, MECHANISM_ASSOCIATION}) MECHANISM_ASSOCIATION ::= [formal_parameter_NAME =>] MECHANISM_NAME MECHANISM_NAME ::= Value | Reference This pragma is identical to ``Export_Function`` except that it applies to a procedure rather than a function and the parameters ``Result_Type`` and ``Result_Mechanism`` are not permitted. GNAT does not require a separate pragma ``Export``, but if none is present, ``Convention Ada`` is assumed, which is usually not what is wanted, so it is usually appropriate to use this pragma in conjunction with a ``Export`` or ``Convention`` pragma that specifies the desired foreign convention. .. index:: Suppressing external name Special treatment is given if the EXTERNAL is an explicit null string or a static string expressions that evaluates to the null string. In this case, no external name is generated. This form still allows the specification of parameter mechanisms. Pragma Export_Valued_Procedure ============================== Syntax: :: pragma Export_Valued_Procedure ( [Internal =>] LOCAL_NAME [, [External =>] EXTERNAL_SYMBOL] [, [Parameter_Types =>] PARAMETER_TYPES] [, [Mechanism =>] MECHANISM]); EXTERNAL_SYMBOL ::= IDENTIFIER | static_string_EXPRESSION | "" PARAMETER_TYPES ::= null | TYPE_DESIGNATOR {, TYPE_DESIGNATOR} TYPE_DESIGNATOR ::= subtype_NAME | subtype_Name ' Access MECHANISM ::= MECHANISM_NAME | (MECHANISM_ASSOCIATION {, MECHANISM_ASSOCIATION}) MECHANISM_ASSOCIATION ::= [formal_parameter_NAME =>] MECHANISM_NAME MECHANISM_NAME ::= Value | Reference This pragma is identical to ``Export_Procedure`` except that the first parameter of ``LOCAL_NAME``, which must be present, must be of mode ``out``, and externally the subprogram is treated as a function with this parameter as the result of the function. GNAT provides for this capability to allow the use of ``out`` and ``in out`` parameters in interfacing to external functions (which are not permitted in Ada functions). GNAT does not require a separate pragma ``Export``, but if none is present, ``Convention Ada`` is assumed, which is almost certainly not what is wanted since the whole point of this pragma is to interface with foreign language functions, so it is usually appropriate to use this pragma in conjunction with a ``Export`` or ``Convention`` pragma that specifies the desired foreign convention. .. index:: Suppressing external name Special treatment is given if the EXTERNAL is an explicit null string or a static string expressions that evaluates to the null string. In this case, no external name is generated. This form still allows the specification of parameter mechanisms. Pragma Extend_System ==================== .. index:: System, extending .. index:: DEC Ada 83 Syntax: :: pragma Extend_System ([Name =>] IDENTIFIER); This pragma is used to provide backwards compatibility with other implementations that extend the facilities of package ``System``. In GNAT, ``System`` contains only the definitions that are present in the Ada RM. However, other implementations, notably the DEC Ada 83 implementation, provide many extensions to package ``System``. For each such implementation accommodated by this pragma, GNAT provides a package :samp:`Aux_{xxx}`, e.g., ``Aux_DEC`` for the DEC Ada 83 implementation, which provides the required additional definitions. You can use this package in two ways. You can ``with`` it in the normal way and access entities either by selection or using a ``use`` clause. In this case no special processing is required. However, if existing code contains references such as :samp:`System.{xxx}` where *xxx* is an entity in the extended definitions provided in package ``System``, you may use this pragma to extend visibility in ``System`` in a non-standard way that provides greater compatibility with the existing code. Pragma ``Extend_System`` is a configuration pragma whose single argument is the name of the package containing the extended definition (e.g., ``Aux_DEC`` for the DEC Ada case). A unit compiled under control of this pragma will be processed using special visibility processing that looks in package :samp:`System.Aux_{xxx}` where :samp:`Aux_{xxx}` is the pragma argument for any entity referenced in package ``System``, but not found in package ``System``. You can use this pragma either to access a predefined ``System`` extension supplied with the compiler, for example ``Aux_DEC`` or you can construct your own extension unit following the above definition. Note that such a package is a child of ``System`` and thus is considered part of the implementation. To compile it you will have to use the *-gnatg* switch for compiling System units, as explained in the GNAT User's Guide. .. _Pragma_Extensions_Allowed: Pragma Extensions_Allowed ========================= .. index:: Ada Extensions .. index:: GNAT Extensions Syntax: .. code-block:: ada pragma Extensions_Allowed (On | Off | All_Extensions); This configuration pragma enables (via the "On" or "All_Extensions" argument) or disables (via the "Off" argument) the implementation extension mode; the pragma takes precedence over the ``-gnatX`` and ``-gnatX0`` command switches. If an argument of ``"On"`` is specified, the latest version of the Ada language is implemented (currently Ada 2022) and, in addition, a curated set of GNAT specific extensions are recognized. (See the list here :ref:`here`) An argument of ``"All_Extensions"`` has the same effect except that some extra experimental extensions are enabled (See the list here :ref:`here`) .. _Pragma-Extensions_Visible: Pragma Extensions_Visible ========================= Syntax: .. code-block:: ada pragma Extensions_Visible [ (static_boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Extensions_Visible`` in the SPARK 2014 Reference Manual, section 6.1.7. Pragma External =============== Syntax: :: pragma External ( [ Convention =>] convention_IDENTIFIER, [ Entity =>] LOCAL_NAME [, [External_Name =>] static_string_EXPRESSION ] [, [Link_Name =>] static_string_EXPRESSION ]); This pragma is identical in syntax and semantics to pragma ``Export`` as defined in the Ada Reference Manual. It is provided for compatibility with some Ada 83 compilers that used this pragma for exactly the same purposes as pragma ``Export`` before the latter was standardized. Pragma External_Name_Casing =========================== .. index:: Dec Ada 83 casing compatibility .. index:: External Names, casing .. index:: Casing of External names Syntax: :: pragma External_Name_Casing ( Uppercase | Lowercase [, Uppercase | Lowercase | As_Is]); This pragma provides control over the casing of external names associated with Import and Export pragmas. There are two cases to consider: * Implicit external names Implicit external names are derived from identifiers. The most common case arises when a standard Ada Import or Export pragma is used with only two arguments, as in: .. code-block:: ada pragma Import (C, C_Routine); Since Ada is a case-insensitive language, the spelling of the identifier in the Ada source program does not provide any information on the desired casing of the external name, and so a convention is needed. In GNAT the default treatment is that such names are converted to all lower case letters. This corresponds to the normal C style in many environments. The first argument of pragma ``External_Name_Casing`` can be used to control this treatment. If ``Uppercase`` is specified, then the name will be forced to all uppercase letters. If ``Lowercase`` is specified, then the normal default of all lower case letters will be used. This same implicit treatment is also used in the case of extended DEC Ada 83 compatible Import and Export pragmas where an external name is explicitly specified using an identifier rather than a string. * Explicit external names Explicit external names are given as string literals. The most common case arises when a standard Ada Import or Export pragma is used with three arguments, as in: .. code-block:: ada pragma Import (C, C_Routine, "C_routine"); In this case, the string literal normally provides the exact casing required for the external name. The second argument of pragma ``External_Name_Casing`` may be used to modify this behavior. If ``Uppercase`` is specified, then the name will be forced to all uppercase letters. If ``Lowercase`` is specified, then the name will be forced to all lowercase letters. A specification of ``As_Is`` provides the normal default behavior in which the casing is taken from the string provided. This pragma may appear anywhere that a pragma is valid. In particular, it can be used as a configuration pragma in the :file:`gnat.adc` file, in which case it applies to all subsequent compilations, or it can be used as a program unit pragma, in which case it only applies to the current unit, or it can be used more locally to control individual Import/Export pragmas. It was primarily intended for use with OpenVMS systems, where many compilers convert all symbols to upper case by default. For interfacing to such compilers (e.g., the DEC C compiler), it may be convenient to use the pragma: .. code-block:: ada pragma External_Name_Casing (Uppercase, Uppercase); to enforce the upper casing of all external symbols. Pragma Fast_Math ================ Syntax: .. code-block:: ada pragma Fast_Math; This is a configuration pragma which activates a mode in which speed is considered more important for floating-point operations than absolutely accurate adherence to the requirements of the standard. Currently the following operations are affected: *Complex Multiplication* The normal simple formula for complex multiplication can result in intermediate overflows for numbers near the end of the range. The Ada standard requires that this situation be detected and corrected by scaling, but in Fast_Math mode such cases will simply result in overflow. Note that to take advantage of this you must instantiate your own version of ``Ada.Numerics.Generic_Complex_Types`` under control of the pragma, rather than use the preinstantiated versions. .. _Pragma-Favor_Top_Level: Pragma Favor_Top_Level ====================== Syntax: .. code-block:: ada pragma Favor_Top_Level (type_LOCAL_NAME); The argument of pragma ``Favor_Top_Level`` must be a named access-to-subprogram type. This pragma is an efficiency hint to the compiler, regarding the use of ``'Access`` or ``'Unrestricted_Access`` on nested (non-library-level) subprograms. The pragma means that nested subprograms are not used with this type, or are rare, so that the generated code should be efficient in the top-level case. When this pragma is used, dynamically generated trampolines may be used on some targets for nested subprograms. See restriction ``No_Implicit_Dynamic_Code``. Pragma Finalize_Storage_Only ============================ Syntax: .. code-block:: ada pragma Finalize_Storage_Only (first_subtype_LOCAL_NAME); The argument of pragma ``Finalize_Storage_Only`` must denote a local type which is derived from ``Ada.Finalization.Controlled`` or ``Limited_Controlled``. The pragma suppresses the call to ``Finalize`` for declared library-level objects of the argument type. This is mostly useful for types where finalization is only used to deal with storage reclamation since in most environments it is not necessary to reclaim memory just before terminating execution, hence the name. Note that this pragma does not suppress Finalize calls for library-level heap-allocated objects (see pragma ``No_Heap_Finalization``). Pragma Float_Representation =========================== Syntax:: pragma Float_Representation (FLOAT_REP[, float_type_LOCAL_NAME]); FLOAT_REP ::= VAX_Float | IEEE_Float In the one argument form, this pragma is a configuration pragma which allows control over the internal representation chosen for the predefined floating point types declared in the packages ``Standard`` and ``System``. This pragma is only provided for compatibility and has no effect. The two argument form specifies the representation to be used for the specified floating-point type. The argument must be ``IEEE_Float`` to specify the use of IEEE format, as follows: * For a digits value of 6, 32-bit IEEE short format will be used. * For a digits value of 15, 64-bit IEEE long format will be used. * No other value of digits is permitted. .. _Pragma-Ghost: Pragma Ghost ============ Syntax: .. code-block:: ada pragma Ghost [ (static_boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Ghost`` in the SPARK 2014 Reference Manual, section 6.9. .. _Pragma-Global: Pragma Global ============= Syntax: .. code-block:: ada pragma Global (GLOBAL_SPECIFICATION); GLOBAL_SPECIFICATION ::= null | (GLOBAL_LIST) | (MODED_GLOBAL_LIST {, MODED_GLOBAL_LIST}) MODED_GLOBAL_LIST ::= MODE_SELECTOR => GLOBAL_LIST MODE_SELECTOR ::= In_Out | Input | Output | Proof_In GLOBAL_LIST ::= GLOBAL_ITEM | (GLOBAL_ITEM {, GLOBAL_ITEM}) GLOBAL_ITEM ::= NAME For the semantics of this pragma, see the entry for aspect ``Global`` in the SPARK 2014 Reference Manual, section 6.1.4. Pragma Ident ============ Syntax: .. code-block:: ada pragma Ident (static_string_EXPRESSION); This pragma is identical in effect to pragma ``Comment``. It is provided for compatibility with other Ada compilers providing this pragma. Pragma Ignore_Pragma ==================== Syntax: .. code-block:: ada pragma Ignore_Pragma (pragma_IDENTIFIER); This is a configuration pragma that takes a single argument that is a simple identifier. Any subsequent use of a pragma whose pragma identifier matches this argument will be silently ignored. Any preceding use of a pragma whose pragma identifier matches this argument will be parsed and then ignored. This may be useful when legacy code or code intended for compilation with some other compiler contains pragmas that match the name, but not the exact implementation, of a GNAT pragma. The use of this pragma allows such pragmas to be ignored, which may be useful in CodePeer mode, or during porting of legacy code. Pragma Implementation_Defined ============================= Syntax: .. code-block:: ada pragma Implementation_Defined (local_NAME); This pragma marks a previously declared entity as implementation-defined. For an overloaded entity, applies to the most recent homonym. .. code-block:: ada pragma Implementation_Defined; The form with no arguments appears anywhere within a scope, most typically a package spec, and indicates that all entities that are defined within the package spec are Implementation_Defined. This pragma is used within the GNAT runtime library to identify implementation-defined entities introduced in language-defined units, for the purpose of implementing the No_Implementation_Identifiers restriction. Pragma Implemented ================== Syntax: :: pragma Implemented (procedure_LOCAL_NAME, implementation_kind); implementation_kind ::= By_Entry | By_Protected_Procedure | By_Any This is an Ada 2012 representation pragma which applies to protected, task and synchronized interface primitives. The use of pragma Implemented provides a way to impose a static requirement on the overriding operation by adhering to one of the three implementation kinds: entry, protected procedure or any of the above. This pragma is available in all earlier versions of Ada as an implementation-defined pragma. .. code-block:: ada type Synch_Iface is synchronized interface; procedure Prim_Op (Obj : in out Iface) is abstract; pragma Implemented (Prim_Op, By_Protected_Procedure); protected type Prot_1 is new Synch_Iface with procedure Prim_Op; -- Legal end Prot_1; protected type Prot_2 is new Synch_Iface with entry Prim_Op; -- Illegal end Prot_2; task type Task_Typ is new Synch_Iface with entry Prim_Op; -- Illegal end Task_Typ; When applied to the procedure_or_entry_NAME of a requeue statement, pragma Implemented determines the runtime behavior of the requeue. Implementation kind By_Entry guarantees that the action of requeueing will proceed from an entry to another entry. Implementation kind By_Protected_Procedure transforms the requeue into a dispatching call, thus eliminating the chance of blocking. Kind By_Any shares the behavior of By_Entry and By_Protected_Procedure depending on the target's overriding subprogram kind. Pragma Implicit_Packing ======================= .. index:: Rational Profile Syntax: .. code-block:: ada pragma Implicit_Packing; This is a configuration pragma that requests implicit packing for packed arrays for which a size clause is given but no explicit pragma Pack or specification of Component_Size is present. It also applies to records where no record representation clause is present. Consider this example: .. code-block:: ada type R is array (0 .. 7) of Boolean; for R'Size use 8; In accordance with the recommendation in the RM (RM 13.3(53)), a Size clause does not change the layout of a composite object. So the Size clause in the above example is normally rejected, since the default layout of the array uses 8-bit components, and thus the array requires a minimum of 64 bits. If this declaration is compiled in a region of code covered by an occurrence of the configuration pragma Implicit_Packing, then the Size clause in this and similar examples will cause implicit packing and thus be accepted. For this implicit packing to occur, the type in question must be an array of small components whose size is known at compile time, and the Size clause must specify the exact size that corresponds to the number of elements in the array multiplied by the size in bits of the component type (both single and multi-dimensioned arrays can be controlled with this pragma). .. index:: Array packing Similarly, the following example shows the use in the record case .. code-block:: ada type r is record a, b, c, d, e, f, g, h : boolean; chr : character; end record; for r'size use 16; Without a pragma Pack, each Boolean field requires 8 bits, so the minimum size is 72 bits, but with a pragma Pack, 16 bits would be sufficient. The use of pragma Implicit_Packing allows this record declaration to compile without an explicit pragma Pack. Pragma Import_Function ====================== Syntax: :: pragma Import_Function ( [Internal =>] LOCAL_NAME, [, [External =>] EXTERNAL_SYMBOL] [, [Parameter_Types =>] PARAMETER_TYPES] [, [Result_Type =>] SUBTYPE_MARK] [, [Mechanism =>] MECHANISM] [, [Result_Mechanism =>] MECHANISM_NAME]); EXTERNAL_SYMBOL ::= IDENTIFIER | static_string_EXPRESSION PARAMETER_TYPES ::= null | TYPE_DESIGNATOR {, TYPE_DESIGNATOR} TYPE_DESIGNATOR ::= subtype_NAME | subtype_Name ' Access MECHANISM ::= MECHANISM_NAME | (MECHANISM_ASSOCIATION {, MECHANISM_ASSOCIATION}) MECHANISM_ASSOCIATION ::= [formal_parameter_NAME =>] MECHANISM_NAME MECHANISM_NAME ::= Value | Reference This pragma is used in conjunction with a pragma ``Import`` to specify additional information for an imported function. The pragma ``Import`` (or equivalent pragma ``Interface``) must precede the ``Import_Function`` pragma and both must appear in the same declarative part as the function specification. The ``Internal`` argument must uniquely designate the function to which the pragma applies. If more than one function name exists of this name in the declarative part you must use the ``Parameter_Types`` and ``Result_Type`` parameters to achieve the required unique designation. Subtype marks in these parameters must exactly match the subtypes in the corresponding function specification, using positional notation to match parameters with subtype marks. The form with an ``'Access`` attribute can be used to match an anonymous access parameter. You may optionally use the ``Mechanism`` and ``Result_Mechanism`` parameters to specify passing mechanisms for the parameters and result. If you specify a single mechanism name, it applies to all parameters. Otherwise you may specify a mechanism on a parameter by parameter basis using either positional or named notation. If the mechanism is not specified, the default mechanism is used. Pragma Import_Object ==================== Syntax: :: pragma Import_Object ( [Internal =>] LOCAL_NAME [, [External =>] EXTERNAL_SYMBOL] [, [Size =>] EXTERNAL_SYMBOL]); EXTERNAL_SYMBOL ::= IDENTIFIER | static_string_EXPRESSION This pragma designates an object as imported, and apart from the extended rules for external symbols, is identical in effect to the use of the normal ``Import`` pragma applied to an object. Unlike the subprogram case, you need not use a separate ``Import`` pragma, although you may do so (and probably should do so from a portability point of view). ``size`` is syntax checked, but otherwise ignored by GNAT. Pragma Import_Procedure ======================= Syntax: :: pragma Import_Procedure ( [Internal =>] LOCAL_NAME [, [External =>] EXTERNAL_SYMBOL] [, [Parameter_Types =>] PARAMETER_TYPES] [, [Mechanism =>] MECHANISM]); EXTERNAL_SYMBOL ::= IDENTIFIER | static_string_EXPRESSION PARAMETER_TYPES ::= null | TYPE_DESIGNATOR {, TYPE_DESIGNATOR} TYPE_DESIGNATOR ::= subtype_NAME | subtype_Name ' Access MECHANISM ::= MECHANISM_NAME | (MECHANISM_ASSOCIATION {, MECHANISM_ASSOCIATION}) MECHANISM_ASSOCIATION ::= [formal_parameter_NAME =>] MECHANISM_NAME MECHANISM_NAME ::= Value | Reference This pragma is identical to ``Import_Function`` except that it applies to a procedure rather than a function and the parameters ``Result_Type`` and ``Result_Mechanism`` are not permitted. Pragma Import_Valued_Procedure ============================== Syntax: :: pragma Import_Valued_Procedure ( [Internal =>] LOCAL_NAME [, [External =>] EXTERNAL_SYMBOL] [, [Parameter_Types =>] PARAMETER_TYPES] [, [Mechanism =>] MECHANISM]); EXTERNAL_SYMBOL ::= IDENTIFIER | static_string_EXPRESSION PARAMETER_TYPES ::= null | TYPE_DESIGNATOR {, TYPE_DESIGNATOR} TYPE_DESIGNATOR ::= subtype_NAME | subtype_Name ' Access MECHANISM ::= MECHANISM_NAME | (MECHANISM_ASSOCIATION {, MECHANISM_ASSOCIATION}) MECHANISM_ASSOCIATION ::= [formal_parameter_NAME =>] MECHANISM_NAME MECHANISM_NAME ::= Value | Reference This pragma is identical to ``Import_Procedure`` except that the first parameter of ``LOCAL_NAME``, which must be present, must be of mode ``out``, and externally the subprogram is treated as a function with this parameter as the result of the function. The purpose of this capability is to allow the use of ``out`` and ``in out`` parameters in interfacing to external functions (which are not permitted in Ada functions). You may optionally use the ``Mechanism`` parameters to specify passing mechanisms for the parameters. If you specify a single mechanism name, it applies to all parameters. Otherwise you may specify a mechanism on a parameter by parameter basis using either positional or named notation. If the mechanism is not specified, the default mechanism is used. Note that it is important to use this pragma in conjunction with a separate pragma Import that specifies the desired convention, since otherwise the default convention is Ada, which is almost certainly not what is required. Pragma Independent ================== Syntax: .. code-block:: ada pragma Independent (component_LOCAL_NAME); This pragma is standard in Ada 2012 mode (which also provides an aspect of the same name). It is also available as an implementation-defined pragma in all earlier versions. It specifies that the designated object or all objects of the designated type must be independently addressable. This means that separate tasks can safely manipulate such objects. For example, if two components of a record are independent, then two separate tasks may access these two components. This may place constraints on the representation of the object (for instance prohibiting tight packing). Pragma Independent_Components ============================= Syntax: .. code-block:: ada pragma Independent_Components (Local_NAME); This pragma is standard in Ada 2012 mode (which also provides an aspect of the same name). It is also available as an implementation-defined pragma in all earlier versions. It specifies that the components of the designated object, or the components of each object of the designated type, must be independently addressable. This means that separate tasks can safely manipulate separate components in the composite object. This may place constraints on the representation of the object (for instance prohibiting tight packing). .. _Pragma-Initial_Condition: Pragma Initial_Condition ======================== Syntax: .. code-block:: ada pragma Initial_Condition (boolean_EXPRESSION); For the semantics of this pragma, see the entry for aspect ``Initial_Condition`` in the SPARK 2014 Reference Manual, section 7.1.6. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Initialize_Scalars ========================= .. index:: debugging with Initialize_Scalars Syntax: .. code-block:: ada pragma Initialize_Scalars [ ( TYPE_VALUE_PAIR {, TYPE_VALUE_PAIR} ) ]; TYPE_VALUE_PAIR ::= SCALAR_TYPE => static_EXPRESSION SCALAR_TYPE := Short_Float | Float | Long_Float | Long_Long_Flat | Signed_8 | Signed_16 | Signed_32 | Signed_64 | Unsigned_8 | Unsigned_16 | Unsigned_32 | Unsigned_64 This pragma is similar to ``Normalize_Scalars`` conceptually but has two important differences. First, there is no requirement for the pragma to be used uniformly in all units of a partition. In particular, it is fine to use this just for some or all of the application units of a partition, without needing to recompile the run-time library. In the case where some units are compiled with the pragma, and some without, then a declaration of a variable where the type is defined in package Standard or is locally declared will always be subject to initialization, as will any declaration of a scalar variable. For composite variables, whether the variable is initialized may also depend on whether the package in which the type of the variable is declared is compiled with the pragma. The other important difference is that the programmer can control the value used for initializing scalar objects. This effect can be achieved in several different ways: * At compile time, the programmer can specify the invalid value for a particular family of scalar types using the optional arguments of the pragma. The compile-time approach is intended to optimize the generated code for the pragma, by possibly using fast operations such as ``memset``. Note that such optimizations require using values where the bytes all have the same binary representation. * At bind time, the programmer has several options: * Initialization with invalid values (similar to Normalize_Scalars, though for Initialize_Scalars it is not always possible to determine the invalid values in complex cases like signed component fields with nonstandard sizes). * Initialization with high values. * Initialization with low values. * Initialization with a specific bit pattern. See the GNAT User's Guide for binder options for specifying these cases. The bind-time approach is intended to provide fast turnaround for testing with different values, without having to recompile the program. * At execution time, the programmer can specify the invalid values using an environment variable. See the GNAT User's Guide for details. The execution-time approach is intended to provide fast turnaround for testing with different values, without having to recompile and rebind the program. Note that pragma ``Initialize_Scalars`` is particularly useful in conjunction with the enhanced validity checking that is now provided in GNAT, which checks for invalid values under more conditions. Using this feature (see description of the *-gnatV* flag in the GNAT User's Guide) in conjunction with pragma ``Initialize_Scalars`` provides a powerful new tool to assist in the detection of problems caused by uninitialized variables. Note: the use of ``Initialize_Scalars`` has a fairly extensive effect on the generated code. This may cause your code to be substantially larger. It may also cause an increase in the amount of stack required, so it is probably a good idea to turn on stack checking (see description of stack checking in the GNAT User's Guide) when using this pragma. .. _Pragma-Initializes: Pragma Initializes ================== Syntax: .. code-block:: ada pragma Initializes (INITIALIZATION_LIST); INITIALIZATION_LIST ::= null | (INITIALIZATION_ITEM {, INITIALIZATION_ITEM}) INITIALIZATION_ITEM ::= name [=> INPUT_LIST] INPUT_LIST ::= null | INPUT | (INPUT {, INPUT}) INPUT ::= name For the semantics of this pragma, see the entry for aspect ``Initializes`` in the SPARK 2014 Reference Manual, section 7.1.5. .. _Pragma-Inline_Always: Pragma Inline_Always ==================== Syntax: :: pragma Inline_Always (NAME [, NAME]); Similar to pragma ``Inline`` except that inlining is unconditional. Inline_Always instructs the compiler to inline every direct call to the subprogram or else to emit a compilation error, independently of any option, in particular *-gnatn* or *-gnatN* or the optimization level. It is an error to take the address or access of ``NAME``. It is also an error to apply this pragma to a primitive operation of a tagged type. Thanks to such restrictions, the compiler is allowed to remove the out-of-line body of ``NAME``. Pragma Inline_Generic ===================== Syntax: :: pragma Inline_Generic (GNAME {, GNAME}); GNAME ::= generic_unit_NAME | generic_instance_NAME This pragma is provided for compatibility with Dec Ada 83. It has no effect in GNAT (which always inlines generics), other than to check that the given names are all names of generic units or generic instances. Pragma Interface ================ Syntax: :: pragma Interface ( [Convention =>] convention_identifier, [Entity =>] local_NAME [, [External_Name =>] static_string_expression] [, [Link_Name =>] static_string_expression]); This pragma is identical in syntax and semantics to the standard Ada pragma ``Import``. It is provided for compatibility with Ada 83. The definition is upwards compatible both with pragma ``Interface`` as defined in the Ada 83 Reference Manual, and also with some extended implementations of this pragma in certain Ada 83 implementations. The only difference between pragma ``Interface`` and pragma ``Import`` is that there is special circuitry to allow both pragmas to appear for the same subprogram entity (normally it is illegal to have multiple ``Import`` pragmas). This is useful in maintaining Ada 83/Ada 95 compatibility and is compatible with other Ada 83 compilers. Pragma Interface_Name ===================== Syntax: :: pragma Interface_Name ( [Entity =>] LOCAL_NAME [, [External_Name =>] static_string_EXPRESSION] [, [Link_Name =>] static_string_EXPRESSION]); This pragma provides an alternative way of specifying the interface name for an interfaced subprogram, and is provided for compatibility with Ada 83 compilers that use the pragma for this purpose. You must provide at least one of ``External_Name`` or ``Link_Name``. Pragma Interrupt_Handler ======================== Syntax: .. code-block:: ada pragma Interrupt_Handler (procedure_LOCAL_NAME); This program unit pragma is supported for parameterless protected procedures as described in Annex C of the Ada Reference Manual. Pragma Interrupt_State ====================== Syntax: :: pragma Interrupt_State ([Name =>] value, [State =>] SYSTEM | RUNTIME | USER); Normally certain interrupts are reserved to the implementation. Any attempt to attach an interrupt causes Program_Error to be raised, as described in RM C.3.2(22). A typical example is the ``SIGINT`` interrupt used in many systems for an :kbd:`Ctrl-C` interrupt. Normally this interrupt is reserved to the implementation, so that :kbd:`Ctrl-C` can be used to interrupt execution. Additionally, signals such as ``SIGSEGV``, ``SIGABRT``, ``SIGFPE`` and ``SIGILL`` are often mapped to specific Ada exceptions, or used to implement run-time functions such as the ``abort`` statement and stack overflow checking. Pragma ``Interrupt_State`` provides a general mechanism for overriding such uses of interrupts. It subsumes the functionality of pragma ``Unreserve_All_Interrupts``. Pragma ``Interrupt_State`` is not available on Windows. On all other platforms than VxWorks, it applies to signals; on VxWorks, it applies to vectored hardware interrupts and may be used to mark interrupts required by the board support package as reserved. Interrupts can be in one of three states: * System The interrupt is reserved (no Ada handler can be installed), and the Ada run-time may not install a handler. As a result you are guaranteed standard system default action if this interrupt is raised. This also allows installing a low level handler via C APIs such as sigaction(), outside of Ada control. * Runtime The interrupt is reserved (no Ada handler can be installed). The run time is allowed to install a handler for internal control purposes, but is not required to do so. * User The interrupt is unreserved. The user may install an Ada handler via Ada.Interrupts and pragma Interrupt_Handler or Attach_Handler to provide some other action. These states are the allowed values of the ``State`` parameter of the pragma. The ``Name`` parameter is a value of the type ``Ada.Interrupts.Interrupt_ID``. Typically, it is a name declared in ``Ada.Interrupts.Names``. This is a configuration pragma, and the binder will check that there are no inconsistencies between different units in a partition in how a given interrupt is specified. It may appear anywhere a pragma is legal. The effect is to move the interrupt to the specified state. By declaring interrupts to be SYSTEM, you guarantee the standard system action, such as a core dump. By declaring interrupts to be USER, you guarantee that you can install a handler. Note that certain signals on many operating systems cannot be caught and handled by applications. In such cases, the pragma is ignored. See the operating system documentation, or the value of the array ``Reserved`` declared in the spec of package ``System.OS_Interface``. Overriding the default state of signals used by the Ada runtime may interfere with an application's runtime behavior in the cases of the synchronous signals, and in the case of the signal used to implement the ``abort`` statement. Pragma Interrupts_System_By_Default =================================== Syntax: :: pragma Interrupts_System_By_Default; Default all interrupts to the System state as defined above in pragma ``Interrupt_State``. This is a configuration pragma. .. _Pragma-Invariant: Pragma Invariant ================ Syntax: :: pragma Invariant ([Entity =>] private_type_LOCAL_NAME, [Check =>] EXPRESSION [,[Message =>] String_Expression]); This pragma provides exactly the same capabilities as the Type_Invariant aspect defined in AI05-0146-1, and in the Ada 2012 Reference Manual. The Type_Invariant aspect is fully implemented in Ada 2012 mode, but since it requires the use of the aspect syntax, which is not available except in 2012 mode, it is not possible to use the Type_Invariant aspect in earlier versions of Ada. However the Invariant pragma may be used in any version of Ada. Also note that the aspect Invariant is a synonym in GNAT for the aspect Type_Invariant, but there is no pragma Type_Invariant. The pragma must appear within the visible part of the package specification, after the type to which its Entity argument appears. As with the Invariant aspect, the Check expression is not analyzed until the end of the visible part of the package, so it may contain forward references. The Message argument, if present, provides the exception message used if the invariant is violated. If no Message parameter is provided, a default message that identifies the line on which the pragma appears is used. It is permissible to have multiple Invariants for the same type entity, in which case they are and'ed together. It is permissible to use this pragma in Ada 2012 mode, but you cannot have both an invariant aspect and an invariant pragma for the same entity. For further details on the use of this pragma, see the Ada 2012 documentation of the Type_Invariant aspect. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Keep_Names ================= Syntax: :: pragma Keep_Names ([On =>] enumeration_first_subtype_LOCAL_NAME); The ``LOCAL_NAME`` argument must refer to an enumeration first subtype in the current declarative part. The effect is to retain the enumeration literal names for use by ``Image`` and ``Value`` even if a global ``Discard_Names`` pragma applies. This is useful when you want to generally suppress enumeration literal names and for example you therefore use a ``Discard_Names`` pragma in the :file:`gnat.adc` file, but you want to retain the names for specific enumeration types. Pragma License ============== .. index:: License checking Syntax: .. code-block:: ada pragma License (Unrestricted | GPL | Modified_GPL | Restricted); This pragma is provided to allow automated checking for appropriate license conditions with respect to the standard and modified GPL. A pragma ``License``, which is a configuration pragma that typically appears at the start of a source file or in a separate :file:`gnat.adc` file, specifies the licensing conditions of a unit as follows: * Unrestricted This is used for a unit that can be freely used with no license restrictions. Examples of such units are public domain units, and units from the Ada Reference Manual. * GPL This is used for a unit that is licensed under the unmodified GPL, and which therefore cannot be ``with``\ ed by a restricted unit. * Modified_GPL This is used for a unit licensed under the GNAT modified GPL that includes a special exception paragraph that specifically permits the inclusion of the unit in programs without requiring the entire program to be released under the GPL. * Restricted This is used for a unit that is restricted in that it is not permitted to depend on units that are licensed under the GPL. Typical examples are proprietary code that is to be released under more restrictive license conditions. Note that restricted units are permitted to ``with`` units which are licensed under the modified GPL (this is the whole point of the modified GPL). Normally a unit with no ``License`` pragma is considered to have an unknown license, and no checking is done. However, standard GNAT headers are recognized, and license information is derived from them as follows. A GNAT license header starts with a line containing 78 hyphens. The following comment text is searched for the appearance of any of the following strings. If the string 'GNU General Public License' is found, then the unit is assumed to have GPL license, unless the string 'As a special exception' follows, in which case the license is assumed to be modified GPL. If one of the strings 'This specification is adapted from the Ada Semantic Interface' or 'This specification is derived from the Ada Reference Manual' is found then the unit is assumed to be unrestricted. These default actions means that a program with a restricted license pragma will automatically get warnings if a GPL unit is inappropriately ``with``\ ed. For example, the program: .. code-block:: ada with Sem_Ch3; with GNAT.Sockets; procedure Secret_Stuff is ... end Secret_Stuff if compiled with pragma ``License`` (``Restricted``) in a :file:`gnat.adc` file will generate the warning:: 1. with Sem_Ch3; | >>> license of withed unit "Sem_Ch3" is incompatible 2. with GNAT.Sockets; 3. procedure Secret_Stuff is Here we get a warning on ``Sem_Ch3`` since it is part of the GNAT compiler and is licensed under the GPL, but no warning for ``GNAT.Sockets`` which is part of the GNAT run time, and is therefore licensed under the modified GPL. Pragma Link_With ================ Syntax: :: pragma Link_With (static_string_EXPRESSION {,static_string_EXPRESSION}); This pragma is provided for compatibility with certain Ada 83 compilers. It has exactly the same effect as pragma ``Linker_Options`` except that spaces occurring within one of the string expressions are treated as separators. For example, in the following case: .. code-block:: ada pragma Link_With ("-labc -ldef"); results in passing the strings ``-labc`` and ``-ldef`` as two separate arguments to the linker. In addition pragma Link_With allows multiple arguments, with the same effect as successive pragmas. Pragma Linker_Alias =================== Syntax: :: pragma Linker_Alias ( [Entity =>] LOCAL_NAME, [Target =>] static_string_EXPRESSION); ``LOCAL_NAME`` must refer to an object that is declared at the library level. This pragma establishes the given entity as a linker alias for the given target. It is equivalent to ``__attribute__((alias))`` in GNU C and causes ``LOCAL_NAME`` to be emitted as an alias for the symbol ``static_string_EXPRESSION`` in the object file, that is to say no space is reserved for ``LOCAL_NAME`` by the assembler and it will be resolved to the same address as ``static_string_EXPRESSION`` by the linker. The actual linker name for the target must be used (e.g., the fully encoded name with qualification in Ada, or the mangled name in C++), or it must be declared using the C convention with ``pragma Import`` or ``pragma Export``. Not all target machines support this pragma. On some of them it is accepted only if ``pragma Weak_External`` has been applied to ``LOCAL_NAME``. .. code-block:: ada -- Example of the use of pragma Linker_Alias package p is i : Integer := 1; pragma Export (C, i); new_name_for_i : Integer; pragma Linker_Alias (new_name_for_i, "i"); end p; Pragma Linker_Constructor ========================= Syntax: .. code-block:: ada pragma Linker_Constructor (procedure_LOCAL_NAME); ``procedure_LOCAL_NAME`` must refer to a parameterless procedure that is declared at the library level. A procedure to which this pragma is applied will be treated as an initialization routine by the linker. It is equivalent to ``__attribute__((constructor))`` in GNU C and causes ``procedure_LOCAL_NAME`` to be invoked before the entry point of the executable is called (or immediately after the shared library is loaded if the procedure is linked in a shared library), in particular before the Ada run-time environment is set up. Because of these specific contexts, the set of operations such a procedure can perform is very limited and the type of objects it can manipulate is essentially restricted to the elementary types. In particular, it must only contain code to which pragma Restrictions (No_Elaboration_Code) applies. This pragma is used by GNAT to implement auto-initialization of shared Stand Alone Libraries, which provides a related capability without the restrictions listed above. Where possible, the use of Stand Alone Libraries is preferable to the use of this pragma. Pragma Linker_Destructor ======================== Syntax: .. code-block:: ada pragma Linker_Destructor (procedure_LOCAL_NAME); ``procedure_LOCAL_NAME`` must refer to a parameterless procedure that is declared at the library level. A procedure to which this pragma is applied will be treated as a finalization routine by the linker. It is equivalent to ``__attribute__((destructor))`` in GNU C and causes ``procedure_LOCAL_NAME`` to be invoked after the entry point of the executable has exited (or immediately before the shared library is unloaded if the procedure is linked in a shared library), in particular after the Ada run-time environment is shut down. See ``pragma Linker_Constructor`` for the set of restrictions that apply because of these specific contexts. .. _Pragma-Linker_Section: Pragma Linker_Section ===================== Syntax: :: pragma Linker_Section ( [Entity =>] LOCAL_NAME, [Section =>] static_string_EXPRESSION); ``LOCAL_NAME`` must refer to an object, type, or subprogram that is declared at the library level. This pragma specifies the name of the linker section for the given entity. It is equivalent to ``__attribute__((section))`` in GNU C and causes ``LOCAL_NAME`` to be placed in the ``static_string_EXPRESSION`` section of the executable (assuming the linker doesn't rename the section). GNAT also provides an implementation defined aspect of the same name. In the case of specifying this aspect for a type, the effect is to specify the corresponding section for all library-level objects of the type that do not have an explicit linker section set. Note that this only applies to whole objects, not to components of composite objects. In the case of a subprogram, the linker section applies to all previously declared matching overloaded subprograms in the current declarative part which do not already have a linker section assigned. The linker section aspect is useful in this case for specifying different linker sections for different elements of such an overloaded set. Note that an empty string specifies that no linker section is specified. This is not quite the same as omitting the pragma or aspect, since it can be used to specify that one element of an overloaded set of subprograms has the default linker section, or that one object of a type for which a linker section is specified should has the default linker section. The compiler normally places library-level entities in standard sections depending on the class: procedures and functions generally go in the ``.text`` section, initialized variables in the ``.data`` section and uninitialized variables in the ``.bss`` section. Other, special sections may exist on given target machines to map special hardware, for example I/O ports or flash memory. This pragma is a means to defer the final layout of the executable to the linker, thus fully working at the symbolic level with the compiler. Some file formats do not support arbitrary sections so not all target machines support this pragma. The use of this pragma may cause a program execution to be erroneous if it is used to place an entity into an inappropriate section (e.g., a modified variable into the ``.text`` section). See also ``pragma Persistent_BSS``. .. code-block:: ada -- Example of the use of pragma Linker_Section package IO_Card is Port_A : Integer; pragma Volatile (Port_A); pragma Linker_Section (Port_A, ".bss.port_a"); Port_B : Integer; pragma Volatile (Port_B); pragma Linker_Section (Port_B, ".bss.port_b"); type Port_Type is new Integer with Linker_Section => ".bss"; PA : Port_Type with Linker_Section => ".bss.PA"; PB : Port_Type; -- ends up in linker section ".bss" procedure Q with Linker_Section => "Qsection"; end IO_Card; .. _Pragma-Lock_Free: Pragma Lock_Free ================ Syntax: .. code-block:: ada pragma Lock_Free [ (static_boolean_EXPRESSION) ]; This pragma may be specified for protected types or objects. It specifies that the implementation of protected operations must be implemented without locks. Compilation fails if the compiler cannot generate lock-free code for the operations. The current conditions required to support this pragma are: * Protected type declarations may not contain entries * Protected subprogram declarations may not have nonelementary parameters In addition, each protected subprogram body must satisfy: * May reference only one protected component * May not reference nonconstant entities outside the protected subprogram scope * May not contain address representation items, allocators, or quantified expressions * May not contain delay, goto, loop, or procedure-call statements * May not contain exported and imported entities * May not dereferenced access values * Function calls and attribute references must be static If the Lock_Free aspect is specified to be True for a protected unit and the Ceiling_Locking locking policy is in effect, then the run-time actions associated with the Ceiling_Locking locking policy (described in Ada RM D.3) are not performed when a protected operation of the protected unit is executed. Pragma Loop_Invariant ===================== Syntax: .. code-block:: ada pragma Loop_Invariant ( boolean_EXPRESSION ); The effect of this pragma is similar to that of pragma ``Assert``, except that in an ``Assertion_Policy`` pragma, the identifier ``Loop_Invariant`` is used to control whether it is ignored or checked (or disabled). ``Loop_Invariant`` can only appear as one of the items in the sequence of statements of a loop body, or nested inside block statements that appear in the sequence of statements of a loop body. The intention is that it be used to represent a "loop invariant" assertion, i.e. something that is true each time through the loop, and which can be used to show that the loop is achieving its purpose. Multiple ``Loop_Invariant`` and ``Loop_Variant`` pragmas that apply to the same loop should be grouped in the same sequence of statements. To aid in writing such invariants, the special attribute ``Loop_Entry`` may be used to refer to the value of an expression on entry to the loop. This attribute can only be used within the expression of a ``Loop_Invariant`` pragma. For full details, see documentation of attribute ``Loop_Entry``. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Loop_Optimize ==================== Syntax: :: pragma Loop_Optimize (OPTIMIZATION_HINT {, OPTIMIZATION_HINT}); OPTIMIZATION_HINT ::= Ivdep | No_Unroll | Unroll | No_Vector | Vector This pragma must appear immediately within a loop statement. It allows the programmer to specify optimization hints for the enclosing loop. The hints are not mutually exclusive and can be freely mixed, but not all combinations will yield a sensible outcome. There are five supported optimization hints for a loop: * Ivdep The programmer asserts that there are no loop-carried dependencies which would prevent consecutive iterations of the loop from being executed simultaneously. * No_Unroll The loop must not be unrolled. This is a strong hint: the compiler will not unroll a loop marked with this hint. * Unroll The loop should be unrolled. This is a weak hint: the compiler will try to apply unrolling to this loop preferably to other optimizations, notably vectorization, but there is no guarantee that the loop will be unrolled. * No_Vector The loop must not be vectorized. This is a strong hint: the compiler will not vectorize a loop marked with this hint. * Vector The loop should be vectorized. This is a weak hint: the compiler will try to apply vectorization to this loop preferably to other optimizations, notably unrolling, but there is no guarantee that the loop will be vectorized. These hints do not remove the need to pass the appropriate switches to the compiler in order to enable the relevant optimizations, that is to say *-funroll-loops* for unrolling and *-ftree-vectorize* for vectorization. Pragma Loop_Variant =================== Syntax: :: pragma Loop_Variant ( LOOP_VARIANT_ITEM {, LOOP_VARIANT_ITEM } ); LOOP_VARIANT_ITEM ::= CHANGE_DIRECTION => discrete_EXPRESSION CHANGE_DIRECTION ::= Increases | Decreases ``Loop_Variant`` can only appear as one of the items in the sequence of statements of a loop body, or nested inside block statements that appear in the sequence of statements of a loop body. It allows the specification of quantities which must always decrease or increase in successive iterations of the loop. In its simplest form, just one expression is specified, whose value must increase or decrease on each iteration of the loop. In a more complex form, multiple arguments can be given which are interpreted in a nesting lexicographic manner. For example: .. code-block:: ada pragma Loop_Variant (Increases => X, Decreases => Y); specifies that each time through the loop either X increases, or X stays the same and Y decreases. A ``Loop_Variant`` pragma ensures that the loop is making progress. It can be useful in helping to show informally or prove formally that the loop always terminates. ``Loop_Variant`` is an assertion whose effect can be controlled using an ``Assertion_Policy`` with a check name of ``Loop_Variant``. The policy can be ``Check`` to enable the loop variant check, ``Ignore`` to ignore the check (in which case the pragma has no effect on the program), or ``Disable`` in which case the pragma is not even checked for correct syntax. Multiple ``Loop_Invariant`` and ``Loop_Variant`` pragmas that apply to the same loop should be grouped in the same sequence of statements. The ``Loop_Entry`` attribute may be used within the expressions of the ``Loop_Variant`` pragma to refer to values on entry to the loop. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Machine_Attribute ======================== Syntax: :: pragma Machine_Attribute ( [Entity =>] LOCAL_NAME, [Attribute_Name =>] static_string_EXPRESSION [, [Info =>] static_EXPRESSION {, static_EXPRESSION}] ); Machine-dependent attributes can be specified for types and/or declarations. This pragma is semantically equivalent to :samp:`__attribute__(({attribute_name}))` (if ``info`` is not specified) or :samp:`__attribute__(({attribute_name(info})))` or :samp:`__attribute__(({attribute_name(info,...})))` in GNU C, where *attribute_name* is recognized by the compiler middle-end or the ``TARGET_ATTRIBUTE_TABLE`` machine specific macro. Note that a string literal for the optional parameter ``info`` or the following ones is transformed by default into an identifier, which may make this pragma unusable for some attributes. For further information see :title:`GNU Compiler Collection (GCC) Internals`. Pragma Main =========== Syntax:: pragma Main (MAIN_OPTION [, MAIN_OPTION]); MAIN_OPTION ::= [Stack_Size =>] static_integer_EXPRESSION | [Task_Stack_Size_Default =>] static_integer_EXPRESSION | [Time_Slicing_Enabled =>] static_boolean_EXPRESSION This pragma is provided for compatibility with OpenVMS VAX Systems. It has no effect in GNAT, other than being syntax checked. Pragma Main_Storage =================== Syntax:: pragma Main_Storage (MAIN_STORAGE_OPTION [, MAIN_STORAGE_OPTION]); MAIN_STORAGE_OPTION ::= [WORKING_STORAGE =>] static_SIMPLE_EXPRESSION | [TOP_GUARD =>] static_SIMPLE_EXPRESSION This pragma is provided for compatibility with OpenVMS VAX Systems. It has no effect in GNAT, other than being syntax checked. .. _Pragma-Max_Queue_Length: Pragma Max_Queue_Length ======================= Syntax:: pragma Max_Queue_Length (static_integer_EXPRESSION); This pragma is used to specify the maximum callers per entry queue for individual protected entries and entry families. It accepts a single integer (-1 or more) as a parameter and must appear after the declaration of an entry. A value of -1 represents no additional restriction on queue length. Pragma Modifies =============== .. index:: Modifies Syntax: :: pragma Modifies (MODIFIES_SPECIFICATION); MODIFIES_SPECIFICATION ::= MODIFIES_CLAUSE | (MODIFIES_CLAUSE {, MODIFIES_CLAUSE}); MODIFIES_CLAUSE ::= MODIFIED_OBJECTS { when GUARD } GUARD ::= boolean_EXPRESSION MODIFIED_OBJECTS ::= MODIFIED_OBJECT | (MODIFIED_OBJECT {, MODIFIED_OBJECT}) MODIFIED_OBJECT ::= name | MODIFIED_OBJECT . all | MODIFIED_OBJECT . component_selector_name | MODIFIED_OBJECT (expression {, expression}) The ``Modifies`` pragma is intended to be an exact replacement for the implementation-defined ``Modifies`` aspect, and shares its restrictions and semantics. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section TBD. Pragma No_Body ============== Syntax: .. code-block:: ada pragma No_Body; There are a number of cases in which a package spec does not require a body, and in fact a body is not permitted. GNAT will not permit the spec to be compiled if there is a body around. The pragma No_Body allows you to provide a body file, even in a case where no body is allowed. The body file must contain only comments and a single No_Body pragma. This is recognized by the compiler as indicating that no body is logically present. This is particularly useful during maintenance when a package is modified in such a way that a body needed before is no longer needed. The provision of a dummy body with a No_Body pragma ensures that there is no interference from earlier versions of the package body. .. _Pragma-No_Caching: Pragma No_Caching ================= Syntax: .. code-block:: ada pragma No_Caching [ (static_boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``No_Caching`` in the SPARK 2014 Reference Manual, section 7.1.2. Pragma No_Component_Reordering ============================== Syntax: :: pragma No_Component_Reordering [([Entity =>] type_LOCAL_NAME)]; ``type_LOCAL_NAME`` must refer to a record type declaration in the current declarative part. The effect is to preclude any reordering of components for the layout of the record, i.e. the record is laid out by the compiler in the order in which the components are declared textually. The form with no argument is a configuration pragma which applies to all record types declared in units to which the pragma applies and there is a requirement that this pragma be used consistently within a partition. .. _Pragma-No_Elaboration_Code_All: Pragma No_Elaboration_Code_All ============================== Syntax: :: pragma No_Elaboration_Code_All [(program_unit_NAME)]; This is a program unit pragma (there is also an equivalent aspect of the same name) that establishes the restriction ``No_Elaboration_Code`` for the current unit and any extended main source units (body and subunits). It also has the effect of enforcing a transitive application of this aspect, so that if any unit is implicitly or explicitly with'ed by the current unit, it must also have the `No_Elaboration_Code_All` aspect set. It may be applied to package or subprogram specs or their generic versions. Pragma No_Heap_Finalization =========================== Syntax: :: pragma No_Heap_Finalization [ (first_subtype_LOCAL_NAME) ]; Pragma ``No_Heap_Finalization`` may be used as a configuration pragma or as a type-specific pragma. In its configuration form, the pragma must appear within a configuration file such as gnat.adc, without an argument. The pragma suppresses the call to ``Finalize`` for heap-allocated objects created through library-level named access-to-object types in cases where the designated type requires finalization actions. In its type-specific form, the argument of the pragma must denote a library-level named access-to-object type. The pragma suppresses the call to ``Finalize`` for heap-allocated objects created through the specific access type in cases where the designated type requires finalization actions. It is still possible to finalize such heap-allocated objects by explicitly deallocating them. A library-level named access-to-object type declared within a generic unit will lose its ``No_Heap_Finalization`` pragma when the corresponding instance does not appear at the library level. .. _Pragma-No_Inline: Pragma No_Inline ================ Syntax: :: pragma No_Inline (NAME {, NAME}); This pragma suppresses inlining for the callable entity or the instances of the generic subprogram designated by ``NAME``, including inlining that results from the use of pragma ``Inline``. This pragma is always active, in particular it is not subject to the use of option *-gnatn* or *-gnatN*. It is illegal to specify both pragma ``No_Inline`` and pragma ``Inline_Always`` for the same ``NAME``. .. _Pragma-No_Raise: Pragma No_Raise =============== Syntax: :: pragma No_Raise (subprogram_LOCAL_NAME {, subprogram_LOCAL_NAME}); Each ``subprogram_LOCAL_NAME`` argument must refer to one or more subprogram declarations in the current declarative part. A subprogram to which this pragma is applied may not raise an exception that is not caught within it. An implementation-defined check named `Raise_Check` is associated with the pragma, and `Program_Error` is raised upon its failure (see RM 11.5(19/5)). Pragma No_Return ================ Syntax: :: pragma No_Return (procedure_LOCAL_NAME {, procedure_LOCAL_NAME}); Each ``procedure_LOCAL_NAME`` argument must refer to one or more procedure declarations in the current declarative part. A procedure to which this pragma is applied may not contain any explicit ``return`` statements. In addition, if the procedure contains any implicit returns from falling off the end of a statement sequence, then execution of that implicit return will cause Program_Error to be raised. One use of this pragma is to identify procedures whose only purpose is to raise an exception. Another use of this pragma is to suppress incorrect warnings about missing returns in functions, where the last statement of a function statement sequence is a call to such a procedure. Note that in Ada 2005 mode, this pragma is part of the language. It is available in all earlier versions of Ada as an implementation-defined pragma. Pragma No_Strict_Aliasing ========================= Syntax: :: pragma No_Strict_Aliasing [([Entity =>] type_LOCAL_NAME)]; ``type_LOCAL_NAME`` must refer to an access type declaration in the current declarative part. The effect is to inhibit strict aliasing optimization for the given type. The form with no arguments is a configuration pragma which applies to all access types declared in units to which the pragma applies. For a detailed description of the strict aliasing optimization, and the situations in which it must be suppressed, see the section on Optimization and Strict Aliasing in the :title:`GNAT User's Guide`. This pragma currently has no effects on access to unconstrained array types. .. _Pragma-No_Tagged_Streams: Pragma No_Tagged_Streams ======================== Syntax: :: pragma No_Tagged_Streams [([Entity =>] tagged_type_LOCAL_NAME)]; Normally when a tagged type is introduced using a full type declaration, part of the processing includes generating stream access routines to be used by stream attributes referencing the type (or one of its subtypes or derived types). This can involve the generation of significant amounts of code which is wasted space if stream routines are not needed for the type in question. The ``No_Tagged_Streams`` pragma causes the generation of these stream routines to be skipped, and any attempt to use stream operations on types subject to this pragma will be statically rejected as illegal. There are two forms of the pragma. The form with no arguments must appear in a declarative sequence or in the declarations of a package spec. This pragma affects all subsequent root tagged types declared in the declaration sequence, and specifies that no stream routines be generated. The form with an argument (for which there is also a corresponding aspect) specifies a single root tagged type for which stream routines are not to be generated. Once the pragma has been given for a particular root tagged type, all subtypes and derived types of this type inherit the pragma automatically, so the effect applies to a complete hierarchy (this is necessary to deal with the class-wide dispatching versions of the stream routines). When pragmas ``Discard_Names`` and ``No_Tagged_Streams`` are simultaneously applied to a tagged type its Expanded_Name and External_Tag are initialized with empty strings. This is useful to avoid exposing entity names at binary level but has a negative impact on the debuggability of tagged types. Pragma Normalize_Scalars ======================== Syntax: .. code-block:: ada pragma Normalize_Scalars; This is a language defined pragma which is fully implemented in GNAT. The effect is to cause all scalar objects that are not otherwise initialized to be initialized. The initial values are implementation dependent and are as follows: *Standard.Character* Objects whose root type is Standard.Character are initialized to Character'Last unless the subtype range excludes NUL (in which case NUL is used). This choice will always generate an invalid value if one exists. *Standard.Wide_Character* Objects whose root type is Standard.Wide_Character are initialized to Wide_Character'Last unless the subtype range excludes NUL (in which case NUL is used). This choice will always generate an invalid value if one exists. *Standard.Wide_Wide_Character* Objects whose root type is Standard.Wide_Wide_Character are initialized to the invalid value 16#FFFF_FFFF# unless the subtype range excludes NUL (in which case NUL is used). This choice will always generate an invalid value if one exists. *Integer types* Objects of an integer type are treated differently depending on whether negative values are present in the subtype. If no negative values are present, then all one bits is used as the initial value except in the special case where zero is excluded from the subtype, in which case all zero bits are used. This choice will always generate an invalid value if one exists. For subtypes with negative values present, the largest negative number is used, except in the unusual case where this largest negative number is in the subtype, and the largest positive number is not, in which case the largest positive value is used. This choice will always generate an invalid value if one exists. *Floating-Point Types* Objects of all floating-point types are initialized to all 1-bits. For standard IEEE format, this corresponds to a NaN (not a number) which is indeed an invalid value. *Fixed-Point Types* Objects of all fixed-point types are treated as described above for integers, with the rules applying to the underlying integer value used to represent the fixed-point value. *Modular types* Objects of a modular type are initialized to all one bits, except in the special case where zero is excluded from the subtype, in which case all zero bits are used. This choice will always generate an invalid value if one exists. *Enumeration types* Objects of an enumeration type are initialized to all one-bits, i.e., to the value ``2 ** typ'Size - 1`` unless the subtype excludes the literal whose Pos value is zero, in which case a code of zero is used. This choice will always generate an invalid value if one exists. .. _Pragma_Obsolescent: Pragma Obsolescent ================== Syntax: :: pragma Obsolescent; pragma Obsolescent ( [Message =>] static_string_EXPRESSION [,[Version =>] Ada_05]); pragma Obsolescent ( [Entity =>] NAME [,[Message =>] static_string_EXPRESSION [,[Version =>] Ada_05]]); This pragma can occur immediately following a declaration of an entity, including the case of a record component. If no Entity argument is present, then this declaration is the one to which the pragma applies. If an Entity parameter is present, it must either match the name of the entity in this declaration, or alternatively, the pragma can immediately follow an enumeration type declaration, where the Entity argument names one of the enumeration literals. This pragma is used to indicate that the named entity is considered obsolescent and should not be used. Typically this is used when an API must be modified by eventually removing or modifying existing subprograms or other entities. The pragma can be used at an intermediate stage when the entity is still present, but will be removed later. The effect of this pragma is to output a warning message on a reference to an entity thus marked that the subprogram is obsolescent if the appropriate warning option in the compiler is activated. If the ``Message`` parameter is present, then a second warning message is given containing this text. In addition, a reference to the entity is considered to be a violation of pragma ``Restrictions (No_Obsolescent_Features)``. This pragma can also be used as a program unit pragma for a package, in which case the entity name is the name of the package, and the pragma indicates that the entire package is considered obsolescent. In this case a client ``with``\ ing such a package violates the restriction, and the ``with`` clause is flagged with warnings if the warning option is set. If the ``Version`` parameter is present (which must be exactly the identifier ``Ada_05``, no other argument is allowed), then the indication of obsolescence applies only when compiling in Ada 2005 mode. This is primarily intended for dealing with the situations in the predefined library where subprograms or packages have become defined as obsolescent in Ada 2005 (e.g., in ``Ada.Characters.Handling``), but may be used anywhere. The following examples show typical uses of this pragma: .. code-block:: ada package p is pragma Obsolescent (p, Message => "use pp instead of p"); end p; package q is procedure q2; pragma Obsolescent ("use q2new instead"); type R is new integer; pragma Obsolescent (Entity => R, Message => "use RR in Ada 2005", Version => Ada_05); type M is record F1 : Integer; F2 : Integer; pragma Obsolescent; F3 : Integer; end record; type E is (a, bc, 'd', quack); pragma Obsolescent (Entity => bc) pragma Obsolescent (Entity => 'd') function "+" (a, b : character) return character; pragma Obsolescent (Entity => "+"); end; Note that, as for all pragmas, if you use a pragma argument identifier, then all subsequent parameters must also use a pragma argument identifier. So if you specify ``Entity =>`` for the ``Entity`` argument, and a ``Message`` argument is present, it must be preceded by ``Message =>``. Pragma Optimize_Alignment ========================= .. index:: Alignment, default settings Syntax: .. code-block:: ada pragma Optimize_Alignment (TIME | SPACE | OFF); This is a configuration pragma which affects the choice of default alignments for types and objects where no alignment is explicitly specified. There is a time/space trade-off in the selection of these values. Large alignments result in more efficient code, at the expense of larger data space, since sizes have to be increased to match these alignments. Smaller alignments save space, but the access code is slower. The normal choice of default alignments for types and individual alignment promotions for objects (which is what you get if you do not use this pragma, or if you use an argument of OFF), tries to balance these two requirements. Specifying SPACE causes smaller default alignments to be chosen in two cases. First any packed record is given an alignment of 1. Second, if a size is given for the type, then the alignment is chosen to avoid increasing this size. For example, consider: .. code-block:: ada type R is record X : Integer; Y : Character; end record; for R'Size use 5*8; In the default mode, this type gets an alignment of 4, so that access to the Integer field X are efficient. But this means that objects of the type end up with a size of 8 bytes. This is a valid choice, since sizes of objects are allowed to be bigger than the size of the type, but it can waste space if for example fields of type R appear in an enclosing record. If the above type is compiled in ``Optimize_Alignment (Space)`` mode, the alignment is set to 1. However, there is one case in which SPACE is ignored. If a variable length record (that is a discriminated record with a component which is an array whose length depends on a discriminant), has a pragma Pack, then it is not in general possible to set the alignment of such a record to one, so the pragma is ignored in this case (with a warning). Specifying SPACE also disables alignment promotions for standalone objects, which occur when the compiler increases the alignment of a specific object without changing the alignment of its type. Specifying SPACE also disables component reordering in unpacked record types, which can result in larger sizes in order to meet alignment requirements. Specifying TIME causes larger default alignments to be chosen in the case of small types with sizes that are not a power of 2. For example, consider: .. code-block:: ada type R is record A : Character; B : Character; C : Boolean; end record; pragma Pack (R); for R'Size use 17; The default alignment for this record is normally 1, but if this type is compiled in ``Optimize_Alignment (Time)`` mode, then the alignment is set to 4, which wastes space for objects of the type, since they are now 4 bytes long, but results in more efficient access when the whole record is referenced. As noted above, this is a configuration pragma, and there is a requirement that all units in a partition be compiled with a consistent setting of the optimization setting. This would normally be achieved by use of a configuration pragma file containing the appropriate setting. The exception to this rule is that units with an explicit configuration pragma in the same file as the source unit are excluded from the consistency check, as are all predefined units. The latter are compiled by default in pragma Optimize_Alignment (Off) mode if no pragma appears at the start of the file. Pragma Ordered ============== Syntax: .. code-block:: ada pragma Ordered (enumeration_first_subtype_LOCAL_NAME); Most enumeration types are from a conceptual point of view unordered. For example, consider: .. code-block:: ada type Color is (Red, Blue, Green, Yellow); By Ada semantics ``Blue > Red`` and ``Green > Blue``, but really these relations make no sense; the enumeration type merely specifies a set of possible colors, and the order is unimportant. For unordered enumeration types, it is generally a good idea if clients avoid comparisons (other than equality or inequality) and explicit ranges. (A *client* is a unit where the type is referenced, other than the unit where the type is declared, its body, and its subunits.) For example, if code buried in some client says: .. code-block:: ada if Current_Color < Yellow then ... if Current_Color in Blue .. Green then ... then the client code is relying on the order, which is undesirable. It makes the code hard to read and creates maintenance difficulties if entries have to be added to the enumeration type. Instead, the code in the client should list the possibilities, or an appropriate subtype should be declared in the unit that declares the original enumeration type. E.g., the following subtype could be declared along with the type ``Color``: .. code-block:: ada subtype RBG is Color range Red .. Green; and then the client could write: .. code-block:: ada if Current_Color in RBG then ... if Current_Color = Blue or Current_Color = Green then ... However, some enumeration types are legitimately ordered from a conceptual point of view. For example, if you declare: .. code-block:: ada type Day is (Mon, Tue, Wed, Thu, Fri, Sat, Sun); then the ordering imposed by the language is reasonable, and clients can depend on it, writing for example: .. code-block:: ada if D in Mon .. Fri then ... if D < Wed then ... The pragma *Ordered* is provided to mark enumeration types that are conceptually ordered, alerting the reader that clients may depend on the ordering. GNAT provides a pragma to mark enumerations as ordered rather than one to mark them as unordered, since in our experience, the great majority of enumeration types are conceptually unordered. The types ``Boolean``, ``Character``, ``Wide_Character``, and ``Wide_Wide_Character`` are considered to be ordered types, so each is declared with a pragma ``Ordered`` in package ``Standard``. Normally pragma ``Ordered`` serves only as documentation and a guide for coding standards, but GNAT provides a warning switch *-gnatw.u* that requests warnings for inappropriate uses (comparisons and explicit subranges) for unordered types. If this switch is used, then any enumeration type not marked with pragma ``Ordered`` will be considered as unordered, and will generate warnings for inappropriate uses. Note that generic types are not considered ordered or unordered (since the template can be instantiated for both cases), so we never generate warnings for the case of generic enumerated types. For additional information please refer to the description of the *-gnatw.u* switch in the GNAT User's Guide. Pragma Overflow_Mode ==================== Syntax: :: pragma Overflow_Mode ( [General =>] MODE [,[Assertions =>] MODE]); MODE ::= STRICT | MINIMIZED | ELIMINATED This pragma sets the current overflow mode to the given setting. For details of the meaning of these modes, please refer to the 'Overflow Check Handling in GNAT' appendix in the GNAT User's Guide. If only the ``General`` parameter is present, the given mode applies to all expressions. If both parameters are present, the ``General`` mode applies to expressions outside assertions, and the ``Eliminated`` mode applies to expressions within assertions. The case of the ``MODE`` parameter is ignored, so ``MINIMIZED``, ``Minimized`` and ``minimized`` all have the same effect. The ``Overflow_Mode`` pragma has the same scoping and placement rules as pragma ``Suppress``, so it can occur either as a configuration pragma, specifying a default for the whole program, or in a declarative scope, where it applies to the remaining declarations and statements in that scope. The pragma ``Suppress (Overflow_Check)`` suppresses overflow checking, but does not affect the overflow mode. The pragma ``Unsuppress (Overflow_Check)`` unsuppresses (enables) overflow checking, but does not affect the overflow mode. Pragma Overriding_Renamings =========================== .. index:: Rational profile .. index:: Rational compatibility Syntax: .. code-block:: ada pragma Overriding_Renamings; This is a GNAT configuration pragma to simplify porting legacy code accepted by the Rational Ada compiler. In the presence of this pragma, a renaming declaration that renames an inherited operation declared in the same scope is legal if selected notation is used as in: .. code-block:: ada pragma Overriding_Renamings; ... package R is function F (..); ... function F (..) renames R.F; end R; even though RM 8.3 (15) stipulates that an overridden operation is not visible within the declaration of the overriding operation. .. _Pragma-Part_Of: Pragma Part_Of ============== Syntax: .. code-block:: ada pragma Part_Of (ABSTRACT_STATE); ABSTRACT_STATE ::= NAME For the semantics of this pragma, see the entry for aspect ``Part_Of`` in the SPARK 2014 Reference Manual, section 7.2.6. Pragma Partition_Elaboration_Policy =================================== Syntax: :: pragma Partition_Elaboration_Policy (POLICY_IDENTIFIER); POLICY_IDENTIFIER ::= Concurrent | Sequential This pragma is standard in Ada 2005, but is available in all earlier versions of Ada as an implementation-defined pragma. See Ada 2012 Reference Manual for details. Pragma Passive ============== Syntax: :: pragma Passive [(Semaphore | No)]; Syntax checked, but otherwise ignored by GNAT. This is recognized for compatibility with DEC Ada 83 implementations, where it is used within a task definition to request that a task be made passive. If the argument ``Semaphore`` is present, or the argument is omitted, then DEC Ada 83 treats the pragma as an assertion that the containing task is passive and that optimization of context switch with this task is permitted and desired. If the argument ``No`` is present, the task must not be optimized. GNAT does not attempt to optimize any tasks in this manner (since protected objects are available in place of passive tasks). For more information on the subject of passive tasks, see the section 'Passive Task Optimization' in the GNAT Users Guide. .. _Pragma-Persistent_BSS: Pragma Persistent_BSS ===================== Syntax: :: pragma Persistent_BSS [(object_LOCAL_NAME)] This pragma allows selected objects to be placed in the ``.persistent_bss`` section. On some targets the linker and loader provide for special treatment of this section, allowing a program to be reloaded without affecting the contents of this data (hence the name persistent). There are two forms of usage. If an argument is given, it must be the local name of a library-level object, with no explicit initialization and whose type is potentially persistent. If no argument is given, then the pragma is a configuration pragma, and applies to all library-level objects with no explicit initialization of potentially persistent types. A potentially persistent type is a scalar type, or an untagged, non-discriminated record, all of whose components have no explicit initialization and are themselves of a potentially persistent type, or an array, all of whose constraints are static, and whose component type is potentially persistent. If this pragma is used on a target where this feature is not supported, then the pragma will be ignored. See also ``pragma Linker_Section``. Pragma Post =========== .. index:: Post .. index:: Checks, postconditions Syntax: .. code-block:: ada pragma Post (Boolean_Expression); The ``Post`` pragma is intended to be an exact replacement for the language-defined ``Post`` aspect, and shares its restrictions and semantics. It must appear either immediately following the corresponding subprogram declaration (only other pragmas may intervene), or if there is no separate subprogram declaration, then it can appear at the start of the declarations in a subprogram body (preceded only by other pragmas). This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Postcondition ==================== .. index:: Postcondition .. index:: Checks, postconditions Syntax: :: pragma Postcondition ( [Check =>] Boolean_Expression [,[Message =>] String_Expression]); The ``Postcondition`` pragma allows specification of automatic postcondition checks for subprograms. These checks are similar to assertions, but are automatically inserted just prior to the return statements of the subprogram with which they are associated (including implicit returns at the end of procedure bodies and associated exception handlers). In addition, the boolean expression which is the condition which must be true may contain references to function'Result in the case of a function to refer to the returned value. ``Postcondition`` pragmas may appear either immediately following the (separate) declaration of a subprogram, or at the start of the declarations of a subprogram body. Only other pragmas may intervene (that is appear between the subprogram declaration and its postconditions, or appear before the postcondition in the declaration sequence in a subprogram body). In the case of a postcondition appearing after a subprogram declaration, the formal arguments of the subprogram are visible, and can be referenced in the postcondition expressions. The postconditions are collected and automatically tested just before any return (implicit or explicit) in the subprogram body. A postcondition is only recognized if postconditions are active at the time the pragma is encountered. The compiler switch *gnata* turns on all postconditions by default, and pragma ``Check_Policy`` with an identifier of ``Postcondition`` can also be used to control whether postconditions are active. The general approach is that postconditions are placed in the spec if they represent functional aspects which make sense to the client. For example we might have: .. code-block:: ada function Direction return Integer; pragma Postcondition (Direction'Result = +1 or else Direction'Result = -1); which serves to document that the result must be +1 or -1, and will test that this is the case at run time if postcondition checking is active. Postconditions within the subprogram body can be used to check that some internal aspect of the implementation, not visible to the client, is operating as expected. For instance if a square root routine keeps an internal counter of the number of times it is called, then we might have the following postcondition: .. code-block:: ada Sqrt_Calls : Natural := 0; function Sqrt (Arg : Float) return Float is pragma Postcondition (Sqrt_Calls = Sqrt_Calls'Old + 1); ... end Sqrt As this example, shows, the use of the ``Old`` attribute is often useful in postconditions to refer to the state on entry to the subprogram. Note that postconditions are only checked on normal returns from the subprogram. If an abnormal return results from raising an exception, then the postconditions are not checked. If a postcondition fails, then the exception ``System.Assertions.Assert_Failure`` is raised. If a message argument was supplied, then the given string will be used as the exception message. If no message argument was supplied, then the default message has the form "Postcondition failed at file_name:line". The exception is raised in the context of the subprogram body, so it is possible to catch postcondition failures within the subprogram body itself. Within a package spec, normal visibility rules in Ada would prevent forward references within a postcondition pragma to functions defined later in the same package. This would introduce undesirable ordering constraints. To avoid this problem, all postcondition pragmas are analyzed at the end of the package spec, allowing forward references. The following example shows that this even allows mutually recursive postconditions as in: .. code-block:: ada package Parity_Functions is function Odd (X : Natural) return Boolean; pragma Postcondition (Odd'Result = (x = 1 or else (x /= 0 and then Even (X - 1)))); function Even (X : Natural) return Boolean; pragma Postcondition (Even'Result = (x = 0 or else (x /= 1 and then Odd (X - 1)))); end Parity_Functions; There are no restrictions on the complexity or form of conditions used within ``Postcondition`` pragmas. The following example shows that it is even possible to verify performance behavior. .. code-block:: ada package Sort is Performance : constant Float; -- Performance constant set by implementation -- to match target architecture behavior. procedure Treesort (Arg : String); -- Sorts characters of argument using N*logN sort pragma Postcondition (Float (Clock - Clock'Old) <= Float (Arg'Length) * log (Float (Arg'Length)) * Performance); end Sort; Note: postcondition pragmas associated with subprograms that are marked as Inline_Always, or those marked as Inline with front-end inlining (-gnatN option set) are accepted and legality-checked by the compiler, but are ignored at run-time even if postcondition checking is enabled. Note that pragma ``Postcondition`` differs from the language-defined ``Post`` aspect (and corresponding ``Post`` pragma) in allowing multiple occurrences, allowing occurrences in the body even if there is a separate spec, and allowing a second string parameter, and the use of the pragma identifier ``Check``. Historically, pragma ``Postcondition`` was implemented prior to the development of Ada 2012, and has been retained in its original form for compatibility purposes. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Post_Class ================= .. index:: Post .. index:: Checks, postconditions Syntax: .. code-block:: ada pragma Post_Class (Boolean_Expression); The ``Post_Class`` pragma is intended to be an exact replacement for the language-defined ``Post'Class`` aspect, and shares its restrictions and semantics. It must appear either immediately following the corresponding subprogram declaration (only other pragmas may intervene), or if there is no separate subprogram declaration, then it can appear at the start of the declarations in a subprogram body (preceded only by other pragmas). Note: This pragma is called ``Post_Class`` rather than ``Post'Class`` because the latter would not be strictly conforming to the allowed syntax for pragmas. The motivation for providing pragmas equivalent to the aspects is to allow a program to be written using the pragmas, and then compiled if necessary using an Ada compiler that does not recognize the pragmas or aspects, but is prepared to ignore the pragmas. The assertion policy that controls this pragma is ``Post'Class``, not ``Post_Class``. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Pre ========== .. index:: Pre .. index:: Checks, preconditions Syntax: .. code-block:: ada pragma Pre (Boolean_Expression); The ``Pre`` pragma is intended to be an exact replacement for the language-defined ``Pre`` aspect, and shares its restrictions and semantics. It must appear either immediately following the corresponding subprogram declaration (only other pragmas may intervene), or if there is no separate subprogram declaration, then it can appear at the start of the declarations in a subprogram body (preceded only by other pragmas). This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Precondition =================== .. index:: Preconditions .. index:: Checks, preconditions Syntax: :: pragma Precondition ( [Check =>] Boolean_Expression [,[Message =>] String_Expression]); The ``Precondition`` pragma is similar to ``Postcondition`` except that the corresponding checks take place immediately upon entry to the subprogram, and if a precondition fails, the exception is raised in the context of the caller, and the attribute 'Result cannot be used within the precondition expression. Otherwise, the placement and visibility rules are identical to those described for postconditions. The following is an example of use within a package spec: .. code-block:: ada package Math_Functions is ... function Sqrt (Arg : Float) return Float; pragma Precondition (Arg >= 0.0) ... end Math_Functions; ``Precondition`` pragmas may appear either immediately following the (separate) declaration of a subprogram, or at the start of the declarations of a subprogram body. Only other pragmas may intervene (that is appear between the subprogram declaration and its postconditions, or appear before the postcondition in the declaration sequence in a subprogram body). Note: precondition pragmas associated with subprograms that are marked as Inline_Always, or those marked as Inline with front-end inlining (-gnatN option set) are accepted and legality-checked by the compiler, but are ignored at run-time even if precondition checking is enabled. Note that pragma ``Precondition`` differs from the language-defined ``Pre`` aspect (and corresponding ``Pre`` pragma) in allowing multiple occurrences, allowing occurrences in the body even if there is a separate spec, and allowing a second string parameter, and the use of the pragma identifier ``Check``. Historically, pragma ``Precondition`` was implemented prior to the development of Ada 2012, and has been retained in its original form for compatibility purposes. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. .. _Pragma-Predicate: Pragma Predicate ================ Syntax: :: pragma Predicate ([Entity =>] type_LOCAL_NAME, [Check =>] EXPRESSION); This pragma (available in all versions of Ada in GNAT) encompasses both the ``Static_Predicate`` and ``Dynamic_Predicate`` aspects in Ada 2012. A predicate is regarded as static if it has an allowed form for ``Static_Predicate`` and is otherwise treated as a ``Dynamic_Predicate``. Otherwise, predicates specified by this pragma behave exactly as described in the Ada 2012 reference manual. For example, if we have .. code-block:: ada type R is range 1 .. 10; subtype S is R; pragma Predicate (Entity => S, Check => S not in 4 .. 6); subtype Q is R pragma Predicate (Entity => Q, Check => F(Q) or G(Q)); the effect is identical to the following Ada 2012 code: .. code-block:: ada type R is range 1 .. 10; subtype S is R with Static_Predicate => S not in 4 .. 6; subtype Q is R with Dynamic_Predicate => F(Q) or G(Q); Note that there are no pragmas ``Dynamic_Predicate`` or ``Static_Predicate``. That is because these pragmas would affect legality and semantics of the program and thus do not have a neutral effect if ignored. The motivation behind providing pragmas equivalent to corresponding aspects is to allow a program to be written using the pragmas, and then compiled with a compiler that will ignore the pragmas. That doesn't work in the case of static and dynamic predicates, since if the corresponding pragmas are ignored, then the behavior of the program is fundamentally changed (for example a membership test ``A in B`` would not take into account a predicate defined for subtype B). When following this approach, the use of predicates should be avoided. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Predicate_Failure ======================== Syntax: :: pragma Predicate_Failure ([Entity =>] type_LOCAL_NAME, [Message =>] String_Expression); The ``Predicate_Failure`` pragma is intended to be an exact replacement for the language-defined ``Predicate_Failure`` aspect, and shares its restrictions and semantics. Pragma Preelaborable_Initialization =================================== Syntax: .. code-block:: ada pragma Preelaborable_Initialization (DIRECT_NAME); This pragma is standard in Ada 2005, but is available in all earlier versions of Ada as an implementation-defined pragma. See Ada 2012 Reference Manual for details. Pragma Prefix_Exception_Messages ================================ .. index:: Prefix_Exception_Messages .. index:: exception .. index:: Exception_Message Syntax: .. code-block:: ada pragma Prefix_Exception_Messages; This is an implementation-defined configuration pragma that affects the behavior of raise statements with a message given as a static string constant (typically a string literal). In such cases, the string will be automatically prefixed by the name of the enclosing entity (giving the package and subprogram containing the raise statement). This helps to identify where messages are coming from, and this mode is automatic for the run-time library. The pragma has no effect if the message is computed with an expression other than a static string constant, since the assumption in this case is that the program computes exactly the string it wants. If you still want the prefixing in this case, you can always call ``GNAT.Source_Info.Enclosing_Entity`` and prepend the string manually. Pragma Pre_Class ================ .. index:: Pre_Class .. index:: Checks, preconditions Syntax: .. code-block:: ada pragma Pre_Class (Boolean_Expression); The ``Pre_Class`` pragma is intended to be an exact replacement for the language-defined ``Pre'Class`` aspect, and shares its restrictions and semantics. It must appear either immediately following the corresponding subprogram declaration (only other pragmas may intervene), or if there is no separate subprogram declaration, then it can appear at the start of the declarations in a subprogram body (preceded only by other pragmas). Note: This pragma is called ``Pre_Class`` rather than ``Pre'Class`` because the latter would not be strictly conforming to the allowed syntax for pragmas. The motivation for providing pragmas equivalent to the aspects is to allow a program to be written using the pragmas, and then compiled if necessary using an Ada compiler that does not recognize the pragmas or aspects, but is prepared to ignore the pragmas. The assertion policy that controls this pragma is ``Pre'Class``, not ``Pre_Class``. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Priority_Specific_Dispatching ==================================== Syntax: :: pragma Priority_Specific_Dispatching ( POLICY_IDENTIFIER, first_priority_EXPRESSION, last_priority_EXPRESSION) POLICY_IDENTIFIER ::= EDF_Across_Priorities | FIFO_Within_Priorities | Non_Preemptive_Within_Priorities | Round_Robin_Within_Priorities This pragma is standard in Ada 2005, but is available in all earlier versions of Ada as an implementation-defined pragma. See Ada 2012 Reference Manual for details. Pragma Profile ============== Syntax: .. code-block:: ada pragma Profile (Ravenscar | Restricted | Rational | Jorvik | GNAT_Extended_Ravenscar | GNAT_Ravenscar_EDF ); This pragma is standard in Ada 2005, but is available in all earlier versions of Ada as an implementation-defined pragma. This is a configuration pragma that establishes a set of configuration pragmas that depend on the argument. ``Ravenscar`` is standard in Ada 2005. ``Jorvik`` is standard in Ada 202x. The other possibilities (``Restricted``, ``Rational``, ``GNAT_Extended_Ravenscar``, ``GNAT_Ravenscar_EDF``) are implementation-defined. ``GNAT_Extended_Ravenscar`` is an alias for ``Jorvik``. The set of configuration pragmas is defined in the following sections. * Pragma Profile (Ravenscar) The ``Ravenscar`` profile is standard in Ada 2005, but is available in all earlier versions of Ada as an implementation-defined pragma. This profile establishes the following set of configuration pragmas: * ``Task_Dispatching_Policy (FIFO_Within_Priorities)`` [RM D.2.2] Tasks are dispatched following a preemptive priority-ordered scheduling policy. * ``Locking_Policy (Ceiling_Locking)`` [RM D.3] While tasks and interrupts execute a protected action, they inherit the ceiling priority of the corresponding protected object. * ``Detect_Blocking`` This pragma forces the detection of potentially blocking operations within a protected operation, and to raise Program_Error if that happens. plus the following set of restrictions: * ``Max_Entry_Queue_Length => 1`` No task can be queued on a protected entry. * ``Max_Protected_Entries => 1`` * ``Max_Task_Entries => 0`` No rendezvous statements are allowed. * ``No_Abort_Statements`` * ``No_Dynamic_Attachment`` * ``No_Dynamic_Priorities`` * ``No_Implicit_Heap_Allocations`` * ``No_Local_Protected_Objects`` * ``No_Local_Timing_Events`` * ``No_Protected_Type_Allocators`` * ``No_Relative_Delay`` * ``No_Requeue_Statements`` * ``No_Select_Statements`` * ``No_Specific_Termination_Handlers`` * ``No_Task_Allocators`` * ``No_Task_Hierarchy`` * ``No_Task_Termination`` * ``Simple_Barriers`` The Ravenscar profile also includes the following restrictions that specify that there are no semantic dependencies on the corresponding predefined packages: * ``No_Dependence => Ada.Asynchronous_Task_Control`` * ``No_Dependence => Ada.Calendar`` * ``No_Dependence => Ada.Execution_Time.Group_Budget`` * ``No_Dependence => Ada.Execution_Time.Timers`` * ``No_Dependence => Ada.Task_Attributes`` * ``No_Dependence => System.Multiprocessors.Dispatching_Domains`` This set of configuration pragmas and restrictions correspond to the definition of the 'Ravenscar Profile' for limited tasking, devised and published by the :title:`International Real-Time Ada Workshop, 1997`. A description is also available at `http://www-users.cs.york.ac.uk/~burns/ravenscar.ps `_. The original definition of the profile was revised at subsequent IRTAW meetings. It has been included in the ISO :title:`Guide for the Use of the Ada Programming Language in High Integrity Systems`, and was made part of the Ada 2005 standard. The formal definition given by the Ada Rapporteur Group (ARG) can be found in two Ada Issues (AI-249 and AI-305) available at `http://www.ada-auth.org/cgi-bin/cvsweb.cgi/ais/ai-00249.txt `_ and `http://www.ada-auth.org/cgi-bin/cvsweb.cgi/ais/ai-00305.txt `_. The above set is a superset of the restrictions provided by pragma ``Profile (Restricted)``, it includes six additional restrictions (``Simple_Barriers``, ``No_Select_Statements``, ``No_Calendar``, ``No_Implicit_Heap_Allocations``, ``No_Relative_Delay`` and ``No_Task_Termination``). This means that pragma ``Profile (Ravenscar)``, like the pragma ``Profile (Restricted)``, automatically causes the use of a simplified, more efficient version of the tasking run-time library. * Pragma Profile (Jorvik) ``Jorvik`` is the new profile added to the Ada 202x draft standard, previously implemented under the name ``GNAT_Extended_Ravenscar``. The ``No_Implicit_Heap_Allocations`` restriction has been replaced by ``No_Implicit_Task_Allocations`` and ``No_Implicit_Protected_Object_Allocations``. The ``Simple_Barriers`` restriction has been replaced by ``Pure_Barriers``. The ``Max_Protected_Entries``, ``Max_Entry_Queue_Length``, and ``No_Relative_Delay`` restrictions have been removed. Details on the rationale for ``Jorvik`` and implications for use may be found in :title:`A New Ravenscar-Based Profile` by P. Rogers, J. Ruiz, T. Gingold and P. Bernardi, in :title:`Reliable Software Technologies -- Ada Europe 2017`, Springer-Verlag Lecture Notes in Computer Science, Number 10300. * Pragma Profile (GNAT_Ravenscar_EDF) This profile corresponds to the Ravenscar profile but using EDF_Across_Priority as the Task_Scheduling_Policy. * Pragma Profile (Restricted) This profile corresponds to the GNAT restricted run time. It establishes the following set of restrictions: * ``No_Abort_Statements`` * ``No_Entry_Queue`` * ``No_Task_Hierarchy`` * ``No_Task_Allocators`` * ``No_Dynamic_Priorities`` * ``No_Terminate_Alternatives`` * ``No_Dynamic_Attachment`` * ``No_Protected_Type_Allocators`` * ``No_Local_Protected_Objects`` * ``No_Requeue_Statements`` * ``No_Task_Attributes_Package`` * ``Max_Asynchronous_Select_Nesting = 0`` * ``Max_Task_Entries = 0`` * ``Max_Protected_Entries = 1`` * ``Max_Select_Alternatives = 0`` This set of restrictions causes the automatic selection of a simplified version of the run time that provides improved performance for the limited set of tasking functionality permitted by this set of restrictions. * Pragma Profile (Rational) The Rational profile is intended to facilitate porting legacy code that compiles with the Rational APEX compiler, even when the code includes non- conforming Ada constructs. The profile enables the following three pragmas: * ``pragma Implicit_Packing`` * ``pragma Overriding_Renamings`` * ``pragma Use_VADS_Size`` Pragma Profile_Warnings ======================= Syntax: .. code-block:: ada pragma Profile_Warnings (Ravenscar | Restricted | Rational); This is an implementation-defined pragma that is similar in effect to ``pragma Profile`` except that instead of generating ``Restrictions`` pragmas, it generates ``Restriction_Warnings`` pragmas. The result is that violations of the profile generate warning messages instead of error messages. .. _Pragma-Program_Exit: Pragma Program_Exit =================== Syntax: .. code-block:: ada pragma Program_Exit [ (boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Program_Exit`` in the SPARK 2014 Reference Manual, section 6.1.10. Pragma Propagate_Exceptions =========================== .. index:: Interfacing to C++ Syntax: .. code-block:: ada pragma Propagate_Exceptions; This pragma is now obsolete and, other than generating a warning if warnings on obsolescent features are enabled, is ignored. It is retained for compatibility purposes. It used to be used in connection with optimization of a now-obsolete mechanism for implementation of exceptions. Pragma Provide_Shift_Operators ============================== .. index:: Shift operators Syntax: .. code-block:: ada pragma Provide_Shift_Operators (integer_first_subtype_LOCAL_NAME); This pragma can be applied to a first subtype local name that specifies either an unsigned or signed type. It has the effect of providing the five shift operators (Shift_Left, Shift_Right, Shift_Right_Arithmetic, Rotate_Left and Rotate_Right) for the given type. It is similar to including the function declarations for these five operators, together with the pragma Import (Intrinsic, ...) statements. Pragma Psect_Object =================== Syntax: :: pragma Psect_Object ( [Internal =>] LOCAL_NAME, [, [External =>] EXTERNAL_SYMBOL] [, [Size =>] EXTERNAL_SYMBOL]); EXTERNAL_SYMBOL ::= IDENTIFIER | static_string_EXPRESSION This pragma is identical in effect to pragma ``Common_Object``. .. _Pragma-Pure_Function: Pragma Pure_Function ==================== Syntax: :: pragma Pure_Function ([Entity =>] function_LOCAL_NAME); This pragma appears in the same declarative part as a function declaration (or a set of function declarations if more than one overloaded declaration exists, in which case the pragma applies to all entities). It specifies that the function ``Entity`` is to be considered pure for the purposes of code generation. This means that the compiler can assume that there are no side effects, and in particular that two identical calls produce the same result in the same context. It also means that the function can be used in an address clause. Note that, quite deliberately, there are no static checks to try to ensure that this promise is met, so ``Pure_Function`` can be used with functions that are conceptually pure, even if they do modify global variables. For example, a square root function that is instrumented to count the number of times it is called is still conceptually pure, and can still be optimized, even though it modifies a global variable (the count). Memo functions are another example (where a table of previous calls is kept and consulted to avoid re-computation). Note also that the normal rules excluding optimization of subprograms in pure units (when parameter types are descended from System.Address, or when the full view of a parameter type is limited), do not apply for the Pure_Function case. If you explicitly specify Pure_Function, the compiler may optimize away calls with identical arguments, and if that results in unexpected behavior, the proper action is not to use the pragma for subprograms that are not (conceptually) pure. Note: Most functions in a ``Pure`` package are automatically pure, and there is no need to use pragma ``Pure_Function`` for such functions. One exception is any function that has at least one formal of type ``System.Address`` or a type derived from it. Such functions are not considered pure by default, since the compiler assumes that the ``Address`` parameter may be functioning as a pointer and that the referenced data may change even if the address value does not. Similarly, imported functions are not considered to be pure by default, since there is no way of checking that they are in fact pure. The use of pragma ``Pure_Function`` for such a function will override these default assumption, and cause the compiler to treat a designated subprogram as pure in these cases. Note: If pragma ``Pure_Function`` is applied to a renamed function, it applies to the underlying renamed function. This can be used to disambiguate cases of overloading where some but not all functions in a set of overloaded functions are to be designated as pure. If pragma ``Pure_Function`` is applied to a library-level function, the function is also considered pure from an optimization point of view, but the unit is not a Pure unit in the categorization sense. So for example, a function thus marked is free to ``with`` non-pure units. Pragma Rational =============== Syntax: .. code-block:: ada pragma Rational; This pragma is considered obsolescent, but is retained for compatibility purposes. It is equivalent to: .. code-block:: ada pragma Profile (Rational); Pragma Ravenscar ================ Syntax: .. code-block:: ada pragma Ravenscar; This pragma is considered obsolescent, but is retained for compatibility purposes. It is equivalent to: .. code-block:: ada pragma Profile (Ravenscar); which is the preferred method of setting the ``Ravenscar`` profile. .. _Pragma-Refined_Depends: Pragma Refined_Depends ====================== Syntax: .. code-block:: ada pragma Refined_Depends (DEPENDENCY_RELATION); DEPENDENCY_RELATION ::= null | (DEPENDENCY_CLAUSE {, DEPENDENCY_CLAUSE}) DEPENDENCY_CLAUSE ::= OUTPUT_LIST =>[+] INPUT_LIST | NULL_DEPENDENCY_CLAUSE NULL_DEPENDENCY_CLAUSE ::= null => INPUT_LIST OUTPUT_LIST ::= OUTPUT | (OUTPUT {, OUTPUT}) INPUT_LIST ::= null | INPUT | (INPUT {, INPUT}) OUTPUT ::= NAME | FUNCTION_RESULT INPUT ::= NAME where FUNCTION_RESULT is a function Result attribute_reference For the semantics of this pragma, see the entry for aspect ``Refined_Depends`` in the SPARK 2014 Reference Manual, section 6.1.5. .. _Pragma-Refined_Global: Pragma Refined_Global ===================== Syntax: .. code-block:: ada pragma Refined_Global (GLOBAL_SPECIFICATION); GLOBAL_SPECIFICATION ::= null | (GLOBAL_LIST) | (MODED_GLOBAL_LIST {, MODED_GLOBAL_LIST}) MODED_GLOBAL_LIST ::= MODE_SELECTOR => GLOBAL_LIST MODE_SELECTOR ::= In_Out | Input | Output | Proof_In GLOBAL_LIST ::= GLOBAL_ITEM | (GLOBAL_ITEM {, GLOBAL_ITEM}) GLOBAL_ITEM ::= NAME For the semantics of this pragma, see the entry for aspect ``Refined_Global`` in the SPARK 2014 Reference Manual, section 6.1.4. .. _Pragma-Refined_Post: Pragma Refined_Post =================== Syntax: .. code-block:: ada pragma Refined_Post (boolean_EXPRESSION); For the semantics of this pragma, see the entry for aspect ``Refined_Post`` in the SPARK 2014 Reference Manual, section 7.2.7. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. .. _Pragma-Refined_State: Pragma Refined_State ==================== Syntax: .. code-block:: ada pragma Refined_State (REFINEMENT_LIST); REFINEMENT_LIST ::= (REFINEMENT_CLAUSE {, REFINEMENT_CLAUSE}) REFINEMENT_CLAUSE ::= state_NAME => CONSTITUENT_LIST CONSTITUENT_LIST ::= null | CONSTITUENT | (CONSTITUENT {, CONSTITUENT}) CONSTITUENT ::= object_NAME | state_NAME For the semantics of this pragma, see the entry for aspect ``Refined_State`` in the SPARK 2014 Reference Manual, section 7.2.2. Pragma Relative_Deadline ======================== Syntax: .. code-block:: ada pragma Relative_Deadline (time_span_EXPRESSION); This pragma is standard in Ada 2005, but is available in all earlier versions of Ada as an implementation-defined pragma. See Ada 2012 Reference Manual for details. .. _Pragma-Remote_Access_Type: Pragma Remote_Access_Type ========================= Syntax: :: pragma Remote_Access_Type ([Entity =>] formal_access_type_LOCAL_NAME); This pragma appears in the formal part of a generic declaration. It specifies an exception to the RM rule from E.2.2(17/2), which forbids the use of a remote access to class-wide type as actual for a formal access type. When this pragma applies to a formal access type ``Entity``, that type is treated as a remote access to class-wide type in the generic. It must be a formal general access type, and its designated type must be the class-wide type of a formal tagged limited private type from the same generic declaration. In the generic unit, the formal type is subject to all restrictions pertaining to remote access to class-wide types. At instantiation, the actual type must be a remote access to class-wide type. Pragma Rename_Pragma ============================ .. index:: Pragmas, synonyms Syntax: :: pragma Rename_Pragma ( [New_Name =>] IDENTIFIER, [Renamed =>] pragma_IDENTIFIER); This pragma provides a mechanism for supplying new names for existing pragmas. The ``New_Name`` identifier can subsequently be used as a synonym for the Renamed pragma. For example, suppose you have code that was originally developed on a compiler that supports Inline_Only as an implementation defined pragma. And suppose the semantics of pragma Inline_Only are identical to (or at least very similar to) the GNAT implementation defined pragma Inline_Always. You could globally replace Inline_Only with Inline_Always. However, to avoid that source modification, you could instead add a configuration pragma: .. code-block:: ada pragma Rename_Pragma ( New_Name => Inline_Only, Renamed => Inline_Always); Then GNAT will treat "pragma Inline_Only ..." as if you had written "pragma Inline_Always ...". Pragma Inline_Only will not necessarily mean the same thing as the other Ada compiler; it's up to you to make sure the semantics are close enough. Pragma Restricted_Run_Time ========================== Syntax: .. code-block:: ada pragma Restricted_Run_Time; This pragma is considered obsolescent, but is retained for compatibility purposes. It is equivalent to: .. code-block:: ada pragma Profile (Restricted); which is the preferred method of setting the restricted run time profile. Pragma Restriction_Warnings =========================== Syntax: :: pragma Restriction_Warnings (restriction_IDENTIFIER {, restriction_IDENTIFIER}); This pragma allows a series of restriction identifiers to be specified (the list of allowed identifiers is the same as for pragma ``Restrictions``). For each of these identifiers the compiler checks for violations of the restriction, but generates a warning message rather than an error message if the restriction is violated. One use of this is in situations where you want to know about violations of a restriction, but you want to ignore some of these violations. Consider this example, where you want to set Ada_95 mode and enable style checks, but you want to know about any other use of implementation pragmas: .. code-block:: ada pragma Restriction_Warnings (No_Implementation_Pragmas); pragma Warnings (Off, "violation of No_Implementation_Pragmas"); pragma Ada_95; pragma Style_Checks ("2bfhkM160"); pragma Warnings (On, "violation of No_Implementation_Pragmas"); By including the above lines in a configuration pragmas file, the Ada_95 and Style_Checks pragmas are accepted without generating a warning, but any other use of implementation defined pragmas will cause a warning to be generated. Pragma Reviewable ================= Syntax: .. code-block:: ada pragma Reviewable; This pragma is an RM-defined standard pragma, but has no effect on the program being compiled, or on the code generated for the program. To obtain the required output specified in RM H.3.1, the compiler must be run with various special switches as follows: * *Where compiler-generated run-time checks remain* The switch *-gnatGL* may be used to list the expanded code in pseudo-Ada form. Runtime checks show up in the listing either as explicit checks or operators marked with {} to indicate a check is present. * *An identification of known exceptions at compile time* If the program is compiled with *-gnatwa*, the compiler warning messages will indicate all cases where the compiler detects that an exception is certain to occur at run time. * *Possible reads of uninitialized variables* The compiler warns of many such cases, but its output is incomplete. .. only:: PRO or GPL The CodePeer analysis tool may be used to obtain a comprehensive list of all possible points at which uninitialized data may be read. .. only:: FSF A supplemental static analysis tool may be used to obtain a comprehensive list of all possible points at which uninitialized data may be read. * *Where run-time support routines are implicitly invoked* In the output from *-gnatGL*, run-time calls are explicitly listed as calls to the relevant run-time routine. * *Object code listing* This may be obtained either by using the *-S* switch, or the objdump utility. * *Constructs known to be erroneous at compile time* These are identified by warnings issued by the compiler (use *-gnatwa*). * *Stack usage information* Static stack usage data (maximum per-subprogram) can be obtained via the *-fstack-usage* switch to the compiler. Dynamic stack usage data (per task) can be obtained via the *-u* switch to gnatbind .. only:: PRO or GPL The gnatstack utility can be used to provide additional information on stack usage. * *Object code listing of entire partition* This can be obtained by compiling the partition with *-S*, or by applying objdump to all the object files that are part of the partition. * *A description of the run-time model* The full sources of the run-time are available, and the documentation of these routines describes how these run-time routines interface to the underlying operating system facilities. * *Control and data-flow information* .. only:: PRO or GPL The CodePeer tool may be used to obtain complete control and data-flow information, as well as comprehensive messages identifying possible problems based on this information. .. only:: FSF A supplemental static analysis tool may be used to obtain complete control and data-flow information, as well as comprehensive messages identifying possible problems based on this information. .. _Pragma-Secondary_Stack_Size: Pragma Secondary_Stack_Size =========================== Syntax: .. code-block:: ada pragma Secondary_Stack_Size (integer_EXPRESSION); This pragma appears within the task definition of a single task declaration or a task type declaration (like pragma ``Storage_Size``) and applies to all task objects of that type. The argument specifies the size of the secondary stack to be used by these task objects, and must be of an integer type. The secondary stack is used to handle functions that return a variable-sized result, for example a function returning an unconstrained String. Note this pragma only applies to targets using fixed secondary stacks, like VxWorks 653 and bare board targets, where a fixed block for the secondary stack is allocated from the primary stack of the task. By default, these targets assign a percentage of the primary stack for the secondary stack, as defined by ``System.Parameter.Sec_Stack_Percentage``. With this pragma, an ``integer_EXPRESSION`` of bytes is assigned from the primary stack instead. For most targets, the pragma does not apply as the secondary stack grows on demand: allocated as a chain of blocks in the heap. The default size of these blocks can be modified via the :switch:`-D` binder option as described in :title:`GNAT User's Guide`. Note that no check is made to see if the secondary stack can fit inside the primary stack. Note the pragma cannot appear when the restriction ``No_Secondary_Stack`` is in effect. Pragma Share_Generic ==================== Syntax: :: pragma Share_Generic (GNAME {, GNAME}); GNAME ::= generic_unit_NAME | generic_instance_NAME This pragma is provided for compatibility with Dec Ada 83. It has no effect in GNAT (which does not implement shared generics), other than to check that the given names are all names of generic units or generic instances. .. _Pragma-Shared: Pragma Shared ============= This pragma is provided for compatibility with Ada 83. The syntax and semantics are identical to pragma Atomic. Pragma Short_Circuit_And_Or =========================== Syntax: .. code-block:: ada pragma Short_Circuit_And_Or; This configuration pragma causes the predefined AND and OR operators of type Standard.Boolean to have short-circuit semantics. That is, they behave like AND THEN and OR ELSE; the right-hand side is not evaluated if the left-hand side determines the result. This may be useful in the context of certification protocols requiring the use of short-circuited logical operators. There is no requirement that all units in a partition use this option. However, mixing of short-circuit and non-short-circuit semantics can be confusing. Therefore, the recommended use is to put the pragma in a configuration file that applies to the whole program. Alternatively, if you have a legacy library that should not use this pragma, you can put it in a separate library project that does not use the pragma. In any case, fine-grained mixing of the different semantics is not recommended. If pragma ``Short_Circuit_And_Or`` is specified, then it is illegal to rename the predefined Boolean AND and OR, or to pass them to generic formal functions; this corresponds to the fact that AND THEN and OR ELSE cannot be renamed nor passed as generic formal functions. Note that this pragma has no effect on other logical operators -- predefined operators of modular types, array-of-boolean types and types derived from Standard.Boolean, nor user-defined operators. See also the pragma ``Unevaluated_Use_Of_Old`` and the restriction ``No_Direct_Boolean_Operators``, which may be useful in conjunction with ``Short_Circuit_And_Or``. Pragma Short_Descriptors ======================== Syntax: .. code-block:: ada pragma Short_Descriptors; This pragma is provided for compatibility with other Ada implementations. It is recognized but ignored by all current versions of GNAT. .. _Pragma-Side_Effects: Pragma Side_Effects =================== Syntax: .. code-block:: ada pragma Side_Effects [ (static_boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Side_Effects`` in the SPARK Reference Manual, section 6.1.12. .. _Pragma-Simple_Storage_Pool_Type: Pragma Simple_Storage_Pool_Type =============================== .. index:: Storage pool, simple .. index:: Simple storage pool Syntax: .. code-block:: ada pragma Simple_Storage_Pool_Type (type_LOCAL_NAME); A type can be established as a 'simple storage pool type' by applying the representation pragma ``Simple_Storage_Pool_Type`` to the type. A type named in the pragma must be a library-level immutably limited record type or limited tagged type declared immediately within a package declaration. The type can also be a limited private type whose full type is allowed as a simple storage pool type. For a simple storage pool type ``SSP``, nonabstract primitive subprograms ``Allocate``, ``Deallocate``, and ``Storage_Size`` can be declared that are subtype conformant with the following subprogram declarations: .. code-block:: ada procedure Allocate (Pool : in out SSP; Storage_Address : out System.Address; Size_In_Storage_Elements : System.Storage_Elements.Storage_Count; Alignment : System.Storage_Elements.Storage_Count); procedure Deallocate (Pool : in out SSP; Storage_Address : System.Address; Size_In_Storage_Elements : System.Storage_Elements.Storage_Count; Alignment : System.Storage_Elements.Storage_Count); function Storage_Size (Pool : SSP) return System.Storage_Elements.Storage_Count; Procedure ``Allocate`` must be declared, whereas ``Deallocate`` and ``Storage_Size`` are optional. If ``Deallocate`` is not declared, then applying an unchecked deallocation has no effect other than to set its actual parameter to null. If ``Storage_Size`` is not declared, then the ``Storage_Size`` attribute applied to an access type associated with a pool object of type SSP returns zero. Additional operations can be declared for a simple storage pool type (such as for supporting a mark/release storage-management discipline). An object of a simple storage pool type can be associated with an access type by specifying the attribute :ref:`Simple_Storage_Pool `. For example: .. code-block:: ada My_Pool : My_Simple_Storage_Pool_Type; type Acc is access My_Data_Type; for Acc'Simple_Storage_Pool use My_Pool; See attribute :ref:`Simple_Storage_Pool ` for further details. .. _Pragma_Source_File_Name: Pragma Source_File_Name ======================= Syntax: :: pragma Source_File_Name ( [Unit_Name =>] unit_NAME, Spec_File_Name => STRING_LITERAL, [Index => INTEGER_LITERAL]); pragma Source_File_Name ( [Unit_Name =>] unit_NAME, Body_File_Name => STRING_LITERAL, [Index => INTEGER_LITERAL]); Use this to override the normal naming convention. It is a configuration pragma, and so has the usual applicability of configuration pragmas (i.e., it applies to either an entire partition, or to all units in a compilation, or to a single unit, depending on how it is used). ``unit_name`` is mapped to ``file_name_literal``. The identifier for the second argument is required, and indicates whether this is the file name for the spec or for the body. The optional Index argument should be used when a file contains multiple units, and when you do not want to use ``gnatchop`` to separate then into multiple files (which is the recommended procedure to limit the number of recompilations that are needed when some sources change). For instance, if the source file :file:`source.ada` contains .. code-block:: ada package B is ... end B; with B; procedure A is begin .. end A; you could use the following configuration pragmas: .. code-block:: ada pragma Source_File_Name (B, Spec_File_Name => "source.ada", Index => 1); pragma Source_File_Name (A, Body_File_Name => "source.ada", Index => 2); Note that the ``gnatname`` utility can also be used to generate those configuration pragmas. Another form of the ``Source_File_Name`` pragma allows the specification of patterns defining alternative file naming schemes to apply to all files. :: pragma Source_File_Name ( [Spec_File_Name =>] STRING_LITERAL [,[Casing =>] CASING_SPEC] [,[Dot_Replacement =>] STRING_LITERAL]); pragma Source_File_Name ( [Body_File_Name =>] STRING_LITERAL [,[Casing =>] CASING_SPEC] [,[Dot_Replacement =>] STRING_LITERAL]); pragma Source_File_Name ( [Subunit_File_Name =>] STRING_LITERAL [,[Casing =>] CASING_SPEC] [,[Dot_Replacement =>] STRING_LITERAL]); CASING_SPEC ::= Lowercase | Uppercase | Mixedcase The first argument is a pattern that contains a single asterisk indicating the point at which the unit name is to be inserted in the pattern string to form the file name. The second argument is optional. If present it specifies the casing of the unit name in the resulting file name string. The default is lower case. Finally the third argument allows for systematic replacement of any dots in the unit name by the specified string literal. Note that Source_File_Name pragmas should not be used if you are using project files. The reason for this rule is that the project manager is not aware of these pragmas, and so other tools that use the project file would not be aware of the intended naming conventions. If you are using project files, file naming is controlled by Source_File_Name_Project pragmas, which are usually supplied automatically by the project manager. A pragma Source_File_Name cannot appear after a :ref:`Pragma_Source_File_Name_Project`. For more details on the use of the ``Source_File_Name`` pragma, see the sections on `Using Other File Names` and `Alternative File Naming Schemes` in the :title:`GNAT User's Guide`. .. _Pragma_Source_File_Name_Project: Pragma Source_File_Name_Project =============================== This pragma has the same syntax and semantics as pragma Source_File_Name. It is only allowed as a stand-alone configuration pragma. It cannot appear after a :ref:`Pragma_Source_File_Name`, and most importantly, once pragma Source_File_Name_Project appears, no further Source_File_Name pragmas are allowed. The intention is that Source_File_Name_Project pragmas are always generated by the Project Manager in a manner consistent with the naming specified in a project file, and when naming is controlled in this manner, it is not permissible to attempt to modify this naming scheme using Source_File_Name or Source_File_Name_Project pragmas (which would not be known to the project manager). Pragma Source_Reference ======================= Syntax: .. code-block:: ada pragma Source_Reference (INTEGER_LITERAL, STRING_LITERAL); This pragma must appear as the first line of a source file. ``integer_literal`` is the logical line number of the line following the pragma line (for use in error messages and debugging information). ``string_literal`` is a static string constant that specifies the file name to be used in error messages and debugging information. This is most notably used for the output of ``gnatchop`` with the *-r* switch, to make sure that the original unchopped source file is the one referred to. The second argument must be a string literal, it cannot be a static string expression other than a string literal. This is because its value is needed for error messages issued by all phases of the compiler. .. _Pragma-SPARK_Mode: Pragma SPARK_Mode ================= Syntax: :: pragma SPARK_Mode [(On | Off)] ; In general a program can have some parts that are in SPARK 2014 (and follow all the rules in the SPARK Reference Manual), and some parts that are full Ada 2012. The SPARK_Mode pragma is used to identify which parts are in SPARK 2014 (by default programs are in full Ada). The SPARK_Mode pragma can be used in the following places: * As a configuration pragma, in which case it sets the default mode for all units compiled with this pragma. * Immediately following a library-level subprogram spec * Immediately within a library-level package body * Immediately following the ``private`` keyword of a library-level package spec * Immediately following the ``begin`` keyword of a library-level package body * Immediately within a library-level subprogram body Normally a subprogram or package spec/body inherits the current mode that is active at the point it is declared. But this can be overridden by pragma within the spec or body as above. The basic consistency rule is that you can't turn SPARK_Mode back ``On``, once you have explicitly (with a pragma) turned if ``Off``. So the following rules apply: If a subprogram spec has SPARK_Mode ``Off``, then the body must also have SPARK_Mode ``Off``. For a package, we have four parts: * the package public declarations * the package private part * the body of the package * the elaboration code after ``begin`` For a package, the rule is that if you explicitly turn SPARK_Mode ``Off`` for any part, then all the following parts must have SPARK_Mode ``Off``. Note that this may require repeating a pragma SPARK_Mode (``Off``) in the body. For example, if we have a configuration pragma SPARK_Mode (``On``) that turns the mode on by default everywhere, and one particular package spec has pragma SPARK_Mode (``Off``), then that pragma will need to be repeated in the package body. Pragma Static_Elaboration_Desired ================================= Syntax: .. code-block:: ada pragma Static_Elaboration_Desired; This pragma is used to indicate that the compiler should attempt to initialize statically the objects declared in the library unit to which the pragma applies, when these objects are initialized (explicitly or implicitly) by an aggregate. In the absence of this pragma, aggregates in object declarations are expanded into assignments and loops, even when the aggregate components are static constants. When the aggregate is present the compiler builds a static expression that requires no run-time code, so that the initialized object can be placed in read-only data space. If the components are not static, or the aggregate has more that 100 components, the compiler emits a warning that the pragma cannot be obeyed. (See also the restriction No_Implicit_Loops, which supports static construction of larger aggregates with static components that include an others choice.) Pragma Stream_Convert ===================== Syntax: :: pragma Stream_Convert ( [Entity =>] type_LOCAL_NAME, [Read =>] function_NAME, [Write =>] function_NAME); This pragma provides an efficient way of providing user-defined stream attributes. Not only is it simpler to use than specifying the attributes directly, but more importantly, it allows the specification to be made in such a way that the predefined unit Ada.Streams is not loaded unless it is actually needed (i.e. unless the stream attributes are actually used); the use of the Stream_Convert pragma adds no overhead at all, unless the stream attributes are actually used on the designated type. The first argument specifies the type for which stream functions are provided. The second parameter provides a function used to read values of this type. It must name a function whose argument type may be any subtype, and whose returned type must be the type given as the first argument to the pragma. The meaning of the ``Read`` parameter is that if a stream attribute directly or indirectly specifies reading of the type given as the first parameter, then a value of the type given as the argument to the Read function is read from the stream, and then the Read function is used to convert this to the required target type. Similarly the ``Write`` parameter specifies how to treat write attributes that directly or indirectly apply to the type given as the first parameter. It must have an input parameter of the type specified by the first parameter, and the return type must be the same as the input type of the Read function. The effect is to first call the Write function to convert to the given stream type, and then write the result type to the stream. The Read and Write functions must not be overloaded subprograms. If necessary renamings can be supplied to meet this requirement. The usage of this attribute is best illustrated by a simple example, taken from the GNAT implementation of package Ada.Strings.Unbounded: .. code-block:: ada function To_Unbounded (S : String) return Unbounded_String renames To_Unbounded_String; pragma Stream_Convert (Unbounded_String, To_Unbounded, To_String); The specifications of the referenced functions, as given in the Ada Reference Manual are: .. code-block:: ada function To_Unbounded_String (Source : String) return Unbounded_String; function To_String (Source : Unbounded_String) return String; The effect is that if the value of an unbounded string is written to a stream, then the representation of the item in the stream is in the same format that would be used for ``Standard.String'Output``, and this same representation is expected when a value of this type is read from the stream. Note that the value written always includes the bounds, even for Unbounded_String'Write, since Unbounded_String is not an array type. Note that the ``Stream_Convert`` pragma is not effective in the case of a derived type of a non-limited tagged type. If such a type is specified then the pragma is silently ignored, and the default implementation of the stream attributes is used instead. Pragma Style_Checks =================== Syntax: :: pragma Style_Checks (string_LITERAL | ALL_CHECKS | On | Off [, LOCAL_NAME]); This pragma is used in conjunction with compiler switches to control the built in style checking provided by GNAT. The compiler switches, if set, provide an initial setting for the switches, and this pragma may be used to modify these settings, or the settings may be provided entirely by the use of the pragma. This pragma can be used anywhere that a pragma is legal, including use as a configuration pragma (including use in the :file:`gnat.adc` file). The form with a string literal specifies which style options are to be activated. These are additive, so they apply in addition to any previously set style check options. The codes for the options are the same as those used in the *-gnaty* switch to *gcc* or *gnatmake*. For example the following two methods can be used to enable layout checking and to change the maximum nesting level value: * .. code-block:: ada -- switch on layout checks pragma Style_Checks ("l"); -- set the number of maximum allowed nesting levels to 15 pragma Style_Checks ("L15"); * :: gcc -c -gnatyl -gnatyL15 ... The string literal values can be cumulatively switched on and off by prefixing the value with ``+`` or ``-``, where: * ``+`` is equivalent to no prefix. It applies the check referenced by the literal value; * ``-`` switches the referenced check off. .. code-block:: ada :linenos: :emphasize-lines: 15 -- allow misaligned block by disabling layout check pragma Style_Checks ("-l"); declare msg : constant String := "Hello"; begin Put_Line (msg); end; -- enable the layout check again pragma Style_Checks ("l"); declare msg : constant String := "Hello"; begin Put_Line (msg); end; The code above contains two layout errors, however, only the last line is picked up by the compiler. Similarly, the switches containing a numeric value can be applied in sequence. In the example below, the permitted nesting level is reduced in in the middle block and the compiler raises a warning on the highlighted line. .. code-block:: ada :linenos: :emphasize-lines: 15 -- Permit 3 levels of nesting pragma Style_Checks ("L3"); procedure Main is begin if True then if True then null; end if; end if; -- Reduce permitted nesting levels to 2. -- Note that "+L2" and "L2" are equivalent. pragma Style_Checks ("+L2"); if True then if True then null; end if; end if; -- Disable checking permitted nesting levels. -- Note that the number after "-L" is insignificant, -- "-L", "-L3" and "-Lx" are all equivalent. pragma Style_Checks ("-L3"); if True then if True then null; end if; end if; end Main; The form ``ALL_CHECKS`` activates all standard checks (its use is equivalent to the use of the :switch:`gnaty` switch with no options. See the :title:`GNAT User's Guide` for details.) Note: the behavior is slightly different in GNAT mode (:switch:`-gnatg` used). In this case, ``ALL_CHECKS`` implies the standard set of GNAT mode style check options (i.e. equivalent to :switch:`-gnatyg`). The forms with ``Off`` and ``On`` can be used to temporarily disable style checks as shown in the following example: .. code-block:: ada pragma Style_Checks ("k"); -- requires keywords in lower case pragma Style_Checks (Off); -- turn off style checks NULL; -- this will not generate an error message pragma Style_Checks (On); -- turn style checks back on NULL; -- this will generate an error message Finally the two argument form is allowed only if the first argument is ``On`` or ``Off``. The effect is to turn of semantic style checks for the specified entity, as shown in the following example: .. code-block:: ada pragma Style_Checks ("r"); -- require consistency of identifier casing Arg : Integer; Rf1 : Integer := ARG; -- incorrect, wrong case pragma Style_Checks (Off, Arg); Rf2 : Integer := ARG; -- OK, no error Pragma Subprogram_Variant ========================= .. index:: Subprogram_Variant Syntax: :: pragma Subprogram_Variant (SUBPROGRAM_VARIANT_LIST); SUBPROGRAM_VARIANT_LIST ::= STRUCTURAL_SUBPROGRAM_VARIANT_ITEM | NUMERIC_SUBPROGRAM_VARIANT_ITEMS NUMERIC_SUBPROGRAM_VARIANT_ITEMS ::= NUMERIC_SUBPROGRAM_VARIANT_ITEM {, NUMERIC_SUBPROGRAM_VARIANT_ITEM} NUMERIC_SUBPROGRAM_VARIANT_ITEM ::= CHANGE_DIRECTION => EXPRESSION STRUCTURAL_SUBPROGRAM_VARIANT_ITEM ::= STRUCTURAL => EXPRESSION CHANGE_DIRECTION ::= Increases | Decreases The ``Subprogram_Variant`` pragma is intended to be an exact replacement for the implementation-defined ``Subprogram_Variant`` aspect, and shares its restrictions and semantics. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Subtitle =============== Syntax: :: pragma Subtitle ([Subtitle =>] STRING_LITERAL); This pragma is recognized for compatibility with other Ada compilers but is ignored by GNAT. Pragma Suppress =============== Syntax: :: pragma Suppress (Identifier [, [On =>] Name]); This is a standard pragma, and supports all the check names required in the RM. It is included here because GNAT recognizes some additional check names that are implementation defined (as permitted by the RM): * ``Alignment_Check`` can be used to suppress alignment checks on addresses used in address clauses. Note that ``Alignment_Check`` is suppressed by default on non-strict alignment machines (such as the x86). * ``Atomic_Synchronization`` can be used to suppress the special memory synchronization instructions that are normally generated for access to ``Atomic`` variables to ensure correct synchronization between tasks that use such variables for synchronization purposes. * ``Duplicated_Tag_Check`` Can be used to suppress the check that is generated for a duplicated tag value when a tagged type is declared. * ``Container_Checks`` Can be used to suppress all checks within Ada.Containers and instances of its children, including Tampering_Check. * ``Tampering_Check`` Can be used to suppress tampering check in the containers. * ``Predicate_Check`` can be used to control whether predicate checks are active. It is applicable only to predicates for which the policy is ``Check``. Unlike ``Assertion_Policy``, which determines if a given predicate is ignored or checked for the whole program, the use of ``Suppress`` and ``Unsuppress`` with this check name allows a given predicate to be turned on and off at specific points in the program. * ``Validity_Check`` can be used specifically to control validity checks. If ``Suppress`` is used to suppress validity checks, then no validity checks are performed, including those specified by the appropriate compiler switch or the ``Validity_Checks`` pragma. * Additional check names previously introduced by use of the ``Check_Name`` pragma are also allowed. Note that pragma Suppress gives the compiler permission to omit checks, but does not require the compiler to omit checks. The compiler will generate checks if they are essentially free, even when they are suppressed. In particular, if the compiler can prove that a certain check will necessarily fail, it will generate code to do an unconditional 'raise', even if checks are suppressed. The compiler warns in this case. Of course, run-time checks are omitted whenever the compiler can prove that they will not fail, whether or not checks are suppressed. Pragma Suppress_All =================== Syntax: .. code-block:: ada pragma Suppress_All; This pragma can appear anywhere within a unit. The effect is to apply ``Suppress (All_Checks)`` to the unit in which it appears. This pragma is implemented for compatibility with DEC Ada 83 usage where it appears at the end of a unit, and for compatibility with Rational Ada, where it appears as a program unit pragma. The use of the standard Ada pragma ``Suppress (All_Checks)`` as a normal configuration pragma is the preferred usage in GNAT. .. _Pragma-Suppress_Debug_Info: Pragma Suppress_Debug_Info ========================== Syntax: :: pragma Suppress_Debug_Info ([Entity =>] LOCAL_NAME); This pragma can be used to suppress generation of debug information for the specified entity. It is intended primarily for use in debugging the debugger, and navigating around debugger problems. Pragma Suppress_Exception_Locations =================================== Syntax: .. code-block:: ada pragma Suppress_Exception_Locations; In normal mode, a raise statement for an exception by default generates an exception message giving the file name and line number for the location of the raise. This is useful for debugging and logging purposes, but this entails extra space for the strings for the messages. The configuration pragma ``Suppress_Exception_Locations`` can be used to suppress the generation of these strings, with the result that space is saved, but the exception message for such raises is null. This configuration pragma may appear in a global configuration pragma file, or in a specific unit as usual. It is not required that this pragma be used consistently within a partition, so it is fine to have some units within a partition compiled with this pragma and others compiled in normal mode without it. .. _Pragma-Suppress_Initialization: Pragma Suppress_Initialization ============================== .. index:: Suppressing initialization .. index:: Initialization, suppression of Syntax: :: pragma Suppress_Initialization ([Entity =>] variable_or_subtype_LOCAL_NAME); Here variable_or_subtype_LOCAL_NAME is the name introduced by a type declaration or subtype declaration or the name of a variable introduced by an object declaration. In the case of a type or subtype this pragma suppresses any implicit or explicit initialization for all variables of the given type or subtype, including initialization resulting from the use of pragmas Normalize_Scalars or Initialize_Scalars. This is considered a representation item, so it cannot be given after the type is frozen. It applies to all subsequent object declarations, and also any allocator that creates objects of the type. If the pragma is given for the first subtype, then it is considered to apply to the base type and all its subtypes. If the pragma is given for other than a first subtype, then it applies only to the given subtype. The pragma may not be given after the type is frozen. Note that this includes eliminating initialization of discriminants for discriminated types, and tags for tagged types. In these cases, you will have to use some non-portable mechanism (e.g. address overlays or unchecked conversion) to achieve required initialization of these fields before accessing any object of the corresponding type. For the variable case, implicit initialization for the named variable is suppressed, just as though its subtype had been given in a pragma Suppress_Initialization, as described above. Pragma Task_Name ================ Syntax .. code-block:: ada pragma Task_Name (string_EXPRESSION); This pragma appears within a task definition (like pragma ``Priority``) and applies to the task in which it appears. The argument must be of type String, and provides a name to be used for the task instance when the task is created. Note that this expression is not required to be static, and in particular, it can contain references to task discriminants. This facility can be used to provide different names for different tasks as they are created, as illustrated in the example below. The task name is recorded internally in the run-time structures and is accessible to tools like the debugger. In addition the routine ``Ada.Task_Identification.Image`` will return this string, with a unique task address appended. .. code-block:: ada -- Example of the use of pragma Task_Name with Ada.Task_Identification; use Ada.Task_Identification; with Text_IO; use Text_IO; procedure t3 is type Astring is access String; task type Task_Typ (Name : access String) is pragma Task_Name (Name.all); end Task_Typ; task body Task_Typ is Nam : constant String := Image (Current_Task); begin Put_Line ("-->" & Nam (1 .. 14) & "<--"); end Task_Typ; type Ptr_Task is access Task_Typ; Task_Var : Ptr_Task; begin Task_Var := new Task_Typ (new String'("This is task 1")); Task_Var := new Task_Typ (new String'("This is task 2")); end; Pragma Task_Storage =================== Syntax: :: pragma Task_Storage ( [Task_Type =>] LOCAL_NAME, [Top_Guard =>] static_integer_EXPRESSION); This pragma specifies the length of the guard area for tasks. The guard area is an additional storage area allocated to a task. A value of zero means that either no guard area is created or a minimal guard area is created, depending on the target. This pragma can appear anywhere a ``Storage_Size`` attribute definition clause is allowed for a task type. .. _Pragma-Test_Case: Pragma Test_Case ================ .. index:: Test cases Syntax: :: pragma Test_Case ( [Name =>] static_string_Expression ,[Mode =>] (Nominal | Robustness) [, Requires => Boolean_Expression] [, Ensures => Boolean_Expression]); The ``Test_Case`` pragma allows defining fine-grain specifications for use by testing tools. The compiler checks the validity of the ``Test_Case`` pragma, but its presence does not lead to any modification of the code generated by the compiler. ``Test_Case`` pragmas may only appear immediately following the (separate) declaration of a subprogram in a package declaration, inside a package spec unit. Only other pragmas may intervene (that is appear between the subprogram declaration and a test case). The compiler checks that boolean expressions given in ``Requires`` and ``Ensures`` are valid, where the rules for ``Requires`` are the same as the rule for an expression in ``Precondition`` and the rules for ``Ensures`` are the same as the rule for an expression in ``Postcondition``. In particular, attributes ``'Old`` and ``'Result`` can only be used within the ``Ensures`` expression. The following is an example of use within a package spec: .. code-block:: ada package Math_Functions is ... function Sqrt (Arg : Float) return Float; pragma Test_Case (Name => "Test 1", Mode => Nominal, Requires => Arg < 10000.0, Ensures => Sqrt'Result < 10.0); ... end Math_Functions; The meaning of a test case is that there is at least one context where ``Requires`` holds such that, if the associated subprogram is executed in that context, then ``Ensures`` holds when the subprogram returns. Mode ``Nominal`` indicates that the input context should also satisfy the precondition of the subprogram, and the output context should also satisfy its postcondition. Mode ``Robustness`` indicates that the precondition and postcondition of the subprogram should be ignored for this test case. .. _Pragma-Thread_Local_Storage: Pragma Thread_Local_Storage =========================== .. index:: Task specific storage .. index:: TLS (Thread Local Storage) .. index:: Task_Attributes Syntax: :: pragma Thread_Local_Storage ([Entity =>] LOCAL_NAME); This pragma specifies that the specified entity, which must be a variable declared in a library-level package, is to be marked as "Thread Local Storage" (``TLS``). On systems supporting this (which include Windows, Solaris, GNU/Linux, and VxWorks), this causes each thread (and hence each Ada task) to see a distinct copy of the variable. The variable must not have default initialization, and if there is an explicit initialization, it must be either ``null`` for an access variable, a static expression for a scalar variable, or a fully static aggregate for a composite type, that is to say, an aggregate all of whose components are static, and which does not include packed or discriminated components. This provides a low-level mechanism similar to that provided by the ``Ada.Task_Attributes`` package, but much more efficient and is also useful in writing interface code that will interact with foreign threads. If this pragma is used on a system where ``TLS`` is not supported, then an error message will be generated and the program will be rejected. Pragma Time_Slice ================= Syntax: .. code-block:: ada pragma Time_Slice (static_duration_EXPRESSION); For implementations of GNAT on operating systems where it is possible to supply a time slice value, this pragma may be used for this purpose. It is ignored if it is used in a system that does not allow this control, or if it appears in other than the main program unit. Pragma Title ============ Syntax: :: pragma Title (TITLING_OPTION [, TITLING OPTION]); TITLING_OPTION ::= [Title =>] STRING_LITERAL, | [Subtitle =>] STRING_LITERAL Syntax checked but otherwise ignored by GNAT. This is a listing control pragma used in DEC Ada 83 implementations to provide a title and/or subtitle for the program listing. The program listing generated by GNAT does not have titles or subtitles. Unlike other pragmas, the full flexibility of named notation is allowed for this pragma, i.e., the parameters may be given in any order if named notation is used, and named and positional notation can be mixed following the normal rules for procedure calls in Ada. Pragma Type_Invariant ===================== Syntax: :: pragma Type_Invariant ([Entity =>] type_LOCAL_NAME, [Check =>] EXPRESSION); The ``Type_Invariant`` pragma is intended to be an exact replacement for the language-defined ``Type_Invariant`` aspect, and shares its restrictions and semantics. It differs from the language defined ``Invariant`` pragma in that it does not permit a string parameter, and it is controlled by the assertion identifier ``Type_Invariant`` rather than ``Invariant``. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. .. _Pragma-Type_Invariant_Class: Pragma Type_Invariant_Class =========================== Syntax: :: pragma Type_Invariant_Class ([Entity =>] type_LOCAL_NAME, [Check =>] EXPRESSION); The ``Type_Invariant_Class`` pragma is intended to be an exact replacement for the language-defined ``Type_Invariant'Class`` aspect, and shares its restrictions and semantics. Note: This pragma is called ``Type_Invariant_Class`` rather than ``Type_Invariant'Class`` because the latter would not be strictly conforming to the allowed syntax for pragmas. The motivation for providing pragmas equivalent to the aspects is to allow a program to be written using the pragmas, and then compiled if necessary using an Ada compiler that does not recognize the pragmas or aspects, but is prepared to ignore the pragmas. The assertion policy that controls this pragma is ``Type_Invariant'Class``, not ``Type_Invariant_Class``. This is an assertion kind pragma that can associate a set of its arguments with an assertion level. See SPARK 2014 Reference Manual, section 11.4.2. Pragma Unchecked_Union ====================== .. index:: Unions in C Syntax: .. code-block:: ada pragma Unchecked_Union (first_subtype_LOCAL_NAME); This pragma is used to specify a representation of a record type that is equivalent to a C union. It was introduced as a GNAT implementation defined pragma in the GNAT Ada 95 mode. Ada 2005 includes an extended version of this pragma, making it language defined, and GNAT fully implements this extended version in all language modes (Ada 83, Ada 95, and Ada 2005). For full details, consult the Ada 2012 Reference Manual, section B.3.3. Pragma Unevaluated_Use_Of_Old ============================= .. index:: Attribute Old .. index:: Attribute Loop_Entry .. index:: Unevaluated_Use_Of_Old Syntax: .. code-block:: ada pragma Unevaluated_Use_Of_Old (Error | Warn | Allow); This pragma controls the processing of attributes Old and Loop_Entry. If either of these attributes is used in a potentially unevaluated expression (e.g. the then or else parts of an if expression), then normally this usage is considered illegal if the prefix of the attribute is other than an entity name. The language requires this behavior for Old, and GNAT copies the same rule for Loop_Entry. The reason for this rule is that otherwise, we can have a situation where we save the Old value, and this results in an exception, even though we might not evaluate the attribute. Consider this example: .. code-block:: ada package UnevalOld is K : Character; procedure U (A : String; C : Boolean) -- ERROR with Post => (if C then A(1)'Old = K else True); end; If procedure U is called with a string with a lower bound of 2, and C false, then an exception would be raised trying to evaluate A(1) on entry even though the value would not be actually used. Although the rule guarantees against this possibility, it is sometimes too restrictive. For example if we know that the string has a lower bound of 1, then we will never raise an exception. The pragma ``Unevaluated_Use_Of_Old`` can be used to modify this behavior. If the argument is ``Error`` then an error is given (this is the default RM behavior). If the argument is ``Warn`` then the usage is allowed as legal but with a warning that an exception might be raised. If the argument is ``Allow`` then the usage is allowed as legal without generating a warning. This pragma may appear as a configuration pragma, or in a declarative part or package specification. In the latter case it applies to uses up to the end of the corresponding statement sequence or sequence of package declarations. Pragma User_Aspect_Definition ============================= Syntax: .. code-block:: ada pragma User_Aspect_Definition (Identifier {, Identifier [(Identifier {, Identifier})]}); This configuration pragma defines a new aspect, making it available for subsequent use in a `User_Aspect` aspect specification. The first identifier is the name of the new aspect. Any subsequent arguments specify the names of other aspects. A subsequent name for which no parenthesized arguments are given shall denote either a Boolean-valued non-representation aspect or an aspect that has been defined by another `User_Aspect_Definition` pragma. A name for which one or more arguments are given shall be either `Annotate` or `Local_Restrictions` (and the arguments shall be appropriate for the named aspect). This pragma, together with the `User_Aspect` aspect, provides a mechanism for avoiding textual duplication if some set of aspect specifications is needed in multiple places. This is somewhat analogous to how profiles allow avoiding duplication of `Restrictions` pragmas. The visibility rules for an aspect defined by a `User_Aspect_Definition` pragma are the same as for a check name introduced by a `Check_Name` pragma. If multiple definitions are visible for some aspect at some point, then the definitions must agree. A predefined aspect cannot be redefined. Pragma Unimplemented_Unit ========================= Syntax: .. code-block:: ada pragma Unimplemented_Unit; If this pragma occurs in a unit that is processed by the compiler, GNAT aborts with the message :samp:`xxx is not supported in this configuration`, where ``xxx`` is the name of the current compilation unit. This pragma is intended to allow the compiler to handle unimplemented library units in a clean manner. The abort only happens if code is being generated. Thus you can use specs of unimplemented packages in syntax or semantic checking mode. .. _Pragma-Universal_Aliasing: Pragma Universal_Aliasing ========================= Syntax: :: pragma Universal_Aliasing [([Entity =>] type_LOCAL_NAME)]; ``type_LOCAL_NAME`` must refer to a type declaration in the current declarative part. The effect is to inhibit strict type-based aliasing optimizations for the given type. For a detailed description of the strict type-based aliasing optimizations and the situations in which they need to be suppressed, see the section on ``Optimization and Strict Aliasing`` in the :title:`GNAT User's Guide`. .. _Pragma-Unmodified: Pragma Unmodified ================= .. index:: Warnings, unmodified Syntax: :: pragma Unmodified (LOCAL_NAME {, LOCAL_NAME}); This pragma signals that the assignable entities (variables, ``out`` parameters, ``in out`` parameters) whose names are listed are deliberately not assigned in the current source unit. This suppresses warnings about the entities being referenced but not assigned, and in addition a warning will be generated if one of these entities is in fact assigned in the same unit as the pragma (or in the corresponding body, or one of its subunits). This is particularly useful for clearly signaling that a particular parameter is not modified, even though the spec suggests that it might be. For the variable case, warnings are never given for unreferenced variables whose name contains one of the substrings ``DISCARD, DUMMY, IGNORE, JUNK, UNUSE, TMP, TEMP`` in any casing. Such names are typically to be used in cases where such warnings are expected. Thus it is never necessary to use ``pragma Unmodified`` for such variables, though it is harmless to do so. .. _Pragma-Unreferenced: Pragma Unreferenced =================== .. index:: Warnings, unreferenced Syntax: :: pragma Unreferenced (LOCAL_NAME {, LOCAL_NAME}); pragma Unreferenced (library_unit_NAME {, library_unit_NAME}); This pragma signals that the entities whose names are listed are deliberately not referenced in the current source unit after the occurrence of the pragma. This suppresses warnings about the entities being unreferenced, and in addition a warning will be generated if one of these entities is in fact subsequently referenced in the same unit as the pragma (or in the corresponding body, or one of its subunits). This is particularly useful for clearly signaling that a particular parameter is not referenced in some particular subprogram implementation and that this is deliberate. It can also be useful in the case of objects declared only for their initialization or finalization side effects. If ``LOCAL_NAME`` identifies more than one matching homonym in the current scope, then the entity most recently declared is the one to which the pragma applies. Note that in the case of accept formals, the pragma Unreferenced may appear immediately after the keyword ``do`` which allows the indication of whether or not accept formals are referenced or not to be given individually for each accept statement. The left hand side of an assignment does not count as a reference for the purpose of this pragma. Thus it is fine to assign to an entity for which pragma Unreferenced is given. However, use of an entity as an actual for an out parameter does count as a reference unless warnings for unread output parameters are enabled via :switch:`-gnatw.o`. Note that if a warning is desired for all calls to a given subprogram, regardless of whether they occur in the same unit as the subprogram declaration, then this pragma should not be used (calls from another unit would not be flagged); pragma Obsolescent can be used instead for this purpose, see :ref:`Pragma_Obsolescent`. The second form of pragma ``Unreferenced`` is used within a context clause. In this case the arguments must be unit names of units previously mentioned in ``with`` clauses (similar to the usage of pragma ``Elaborate_All``). The effect is to suppress warnings about unreferenced units and unreferenced entities within these units. For the variable case, warnings are never given for unreferenced variables whose name contains one of the substrings ``DISCARD, DUMMY, IGNORE, JUNK, UNUSE, TMP, TEMP`` in any casing. Such names are typically to be used in cases where such warnings are expected. Thus it is never necessary to use ``pragma Unreferenced`` for such variables, though it is harmless to do so. .. _Pragma-Unreferenced_Objects: Pragma Unreferenced_Objects =========================== .. index:: Warnings, unreferenced Syntax: :: pragma Unreferenced_Objects (local_subtype_NAME {, local_subtype_NAME}); This pragma signals that for the types or subtypes whose names are listed, objects which are declared with one of these types or subtypes may not be referenced, and if no references appear, no warnings are given. This is particularly useful for objects which are declared solely for their initialization and finalization effect. Such variables are sometimes referred to as RAII variables (Resource Acquisition Is Initialization). Using this pragma on the relevant type (most typically a limited controlled type), the compiler will automatically suppress unwanted warnings about these variables not being referenced. Pragma Unreserve_All_Interrupts =============================== Syntax: .. code-block:: ada pragma Unreserve_All_Interrupts; Normally certain interrupts are reserved to the implementation. Any attempt to attach an interrupt causes Program_Error to be raised, as described in RM C.3.2(22). A typical example is the ``SIGINT`` interrupt used in many systems for a :kbd:`Ctrl-C` interrupt. Normally this interrupt is reserved to the implementation, so that :kbd:`Ctrl-C` can be used to interrupt execution. If the pragma ``Unreserve_All_Interrupts`` appears anywhere in any unit in a program, then all such interrupts are unreserved. This allows the program to handle these interrupts, but disables their standard functions. For example, if this pragma is used, then pressing :kbd:`Ctrl-C` will not automatically interrupt execution. However, a program can then handle the ``SIGINT`` interrupt as it chooses. For a full list of the interrupts handled in a specific implementation, see the source code for the spec of ``Ada.Interrupts.Names`` in file :file:`a-intnam.ads`. This is a target dependent file that contains the list of interrupts recognized for a given target. The documentation in this file also specifies what interrupts are affected by the use of the ``Unreserve_All_Interrupts`` pragma. For a more general facility for controlling what interrupts can be handled, see pragma ``Interrupt_State``, which subsumes the functionality of the ``Unreserve_All_Interrupts`` pragma. Pragma Unsuppress ================= Syntax: :: pragma Unsuppress (IDENTIFIER [, [On =>] NAME]); This pragma undoes the effect of a previous pragma ``Suppress``. If there is no corresponding pragma ``Suppress`` in effect, it has no effect. The range of the effect is the same as for pragma ``Suppress``. The meaning of the arguments is identical to that used in pragma ``Suppress``. One important application is to ensure that checks are on in cases where code depends on the checks for its correct functioning, so that the code will compile correctly even if the compiler switches are set to suppress checks. For example, in a program that depends on external names of tagged types and wants to ensure that the duplicated tag check occurs even if all run-time checks are suppressed by a compiler switch, the following configuration pragma will ensure this test is not suppressed: .. code-block:: ada pragma Unsuppress (Duplicated_Tag_Check); This pragma is standard in Ada 2005. It is available in all earlier versions of Ada as an implementation-defined pragma. Note that in addition to the checks defined in the Ada RM, GNAT recognizes a number of implementation-defined check names. See the description of pragma ``Suppress`` for full details. .. _Pragma-Unused: Pragma Unused ============= .. index:: Warnings, unused Syntax: :: pragma Unused (LOCAL_NAME {, LOCAL_NAME}); This pragma signals that the assignable entities (variables, ``out`` parameters, and ``in out`` parameters) whose names are listed deliberately do not get assigned or referenced in the current source unit after the occurrence of the pragma in the current source unit. This suppresses warnings about the entities that are unreferenced and/or not assigned, and, in addition, a warning will be generated if one of these entities gets assigned or subsequently referenced in the same unit as the pragma (in the corresponding body or one of its subunits). This is particularly useful for clearly signaling that a particular parameter is not modified or referenced, even though the spec suggests that it might be. For the variable case, warnings are never given for unreferenced variables whose name contains one of the substrings ``DISCARD, DUMMY, IGNORE, JUNK, UNUSE, TMP, TEMP`` in any casing. Such names are typically to be used in cases where such warnings are expected. Thus it is never necessary to use ``pragma Unused`` for such variables, though it is harmless to do so. Pragma Use_VADS_Size ==================== .. index:: Size, VADS compatibility .. index:: Rational profile Syntax: .. code-block:: ada pragma Use_VADS_Size; This is a configuration pragma. In a unit to which it applies, any use of the 'Size attribute is automatically interpreted as a use of the 'VADS_Size attribute. Note that this may result in incorrect semantic processing of valid Ada 95 or Ada 2005 programs. This is intended to aid in the handling of existing code which depends on the interpretation of Size as implemented in the VADS compiler. See description of the VADS_Size attribute for further details. Pragma Validity_Checks ====================== Syntax: .. code-block:: ada pragma Validity_Checks (string_LITERAL | ALL_CHECKS | On | Off); This pragma is used in conjunction with compiler switches to control the built-in validity checking provided by GNAT. The compiler switches, if set provide an initial setting for the switches, and this pragma may be used to modify these settings, or the settings may be provided entirely by the use of the pragma. This pragma can be used anywhere that a pragma is legal, including use as a configuration pragma (including use in the :file:`gnat.adc` file). The form with a string literal specifies which validity options are to be activated. The validity checks are first set to include only the default reference manual settings, and then a string of letters in the string specifies the exact set of options required. The form of this string is exactly as described for the *-gnatVx* compiler switch (see the GNAT User's Guide for details). For example the following two methods can be used to enable validity checking for mode ``in`` and ``in out`` subprogram parameters: * .. code-block:: ada pragma Validity_Checks ("im"); * .. code-block:: sh $ gcc -c -gnatVim ... The form ALL_CHECKS activates all standard checks (its use is equivalent to the use of the :switch:`gnatVa` switch). The forms with ``Off`` and ``On`` can be used to temporarily disable validity checks as shown in the following example: .. code-block:: ada pragma Validity_Checks ("c"); -- validity checks for copies pragma Validity_Checks (Off); -- turn off validity checks A := B; -- B will not be validity checked pragma Validity_Checks (On); -- turn validity checks back on A := C; -- C will be validity checked .. _Pragma-Volatile: Pragma Volatile =============== Syntax: .. code-block:: ada pragma Volatile (LOCAL_NAME); This pragma is defined by the Ada Reference Manual, and the GNAT implementation is fully conformant with this definition. The reason it is mentioned in this section is that a pragma of the same name was supplied in some Ada 83 compilers, including DEC Ada 83. The Ada 95 / Ada 2005 implementation of pragma Volatile is upwards compatible with the implementation in DEC Ada 83. .. _Pragma-Volatile_Full_Access: Pragma Volatile_Full_Access =========================== Syntax: .. code-block:: ada pragma Volatile_Full_Access (LOCAL_NAME); This is similar in effect to pragma Volatile, except that any reference to the object is guaranteed to be done only with instructions that read or write all the bits of the object. Furthermore, if the object is of a composite type, then any reference to a subcomponent of the object is guaranteed to read and/or write all the bits of the object. The intention is that this be suitable for use with memory-mapped I/O devices on some machines. Note that there are two important respects in which this is different from ``pragma Atomic``. First a reference to a ``Volatile_Full_Access`` object is not a sequential action in the RM 9.10 sense and, therefore, does not create a synchronization point. Second, in the case of ``pragma Atomic``, there is no guarantee that all the bits will be accessed if the reference is not to the whole object; the compiler is allowed (and generally will) access only part of the object in this case. .. _Pragma-Volatile_Function: Pragma Volatile_Function ======================== Syntax: .. code-block:: ada pragma Volatile_Function [ (static_boolean_EXPRESSION) ]; For the semantics of this pragma, see the entry for aspect ``Volatile_Function`` in the SPARK 2014 Reference Manual, section 7.1.2. Pragma Warning_As_Error ======================= Syntax: .. code-block:: ada pragma Warning_As_Error (static_string_EXPRESSION); This configuration pragma allows the programmer to specify a set of warnings that will be treated as errors. Any warning that matches the pattern given by the pragma argument will be treated as an error. This gives more precise control than -gnatwe, which treats warnings as errors. This pragma can apply to regular warnings (messages enabled by -gnatw) and to style warnings (messages that start with "(style)", enabled by -gnaty). The pattern may contain asterisks, which match zero or more characters in the message. For example, you can use ``pragma Warning_As_Error ("bits of*unused")`` to treat the warning message ``warning: 960 bits of "a" unused`` as an error. All characters other than asterisk are treated as literal characters in the match. The match is case insensitive; for example XYZ matches xyz. Note that the pattern matches if it occurs anywhere within the warning message string (it is not necessary to put an asterisk at the start and the end of the message, since this is implied). Another possibility for the static_string_EXPRESSION which works whether or not error tags are enabled (*-gnatw.d*) is to use a single *-gnatw* tag string, enclosed in brackets, as shown in the example below, to treat one category of warnings as errors. Note that if you want to treat multiple categories of warnings as errors, you can use multiple pragma Warning_As_Error. The above use of patterns to match the message applies only to warning messages generated by the front end. This pragma can also be applied to warnings provided by the back end and mentioned in :ref:`Pragma_Warnings`. By using a single full *-Wxxx* switch in the pragma, such warnings can also be treated as errors. The pragma can appear either in a global configuration pragma file (e.g. :file:`gnat.adc`), or at the start of a file. Given a global configuration pragma file containing: .. code-block:: ada pragma Warning_As_Error ("[-gnatwj]"); which will treat all obsolescent feature warnings as errors, the following program compiles as shown (compile options here are *-gnatwa.d -gnatl -gnatj55*). :: 1. pragma Warning_As_Error ("*never assigned*"); 2. function Warnerr return String is 3. X : Integer; | >>> error: variable "X" is never read and never assigned [-gnatwv] [warning-as-error] 4. Y : Integer; | >>> warning: variable "Y" is assigned but never read [-gnatwu] 5. begin 6. Y := 0; 7. return %ABC%; | >>> error: use of "%" is an obsolescent feature (RM J.2(4)), use """ instead [-gnatwj] [warning-as-error] 8. end; 8 lines: No errors, 3 warnings (2 treated as errors) Note that this pragma does not affect the set of warnings issued in any way, it merely changes the effect of a matching warning if one is produced as a result of other warnings options. As shown in this example, if the pragma results in a warning being treated as an error, the tag is changed from "warning:" to "error:" and the string "[warning-as-error]" is appended to the end of the message. .. _Pragma_Warnings: Pragma Warnings =============== Syntax: .. code-block:: ada pragma Warnings ([TOOL_NAME,] DETAILS [, REASON]); DETAILS ::= On | Off DETAILS ::= On | Off, local_NAME DETAILS ::= static_string_EXPRESSION DETAILS ::= On | Off, static_string_EXPRESSION TOOL_NAME ::= GNAT | GNATprove REASON ::= Reason => STRING_LITERAL {& STRING_LITERAL} Note: in Ada 83 mode, a string literal may be used in place of a static string expression (which does not exist in Ada 83). Note if the second argument of ``DETAILS`` is a ``local_NAME`` then the second form is always understood. If the intention is to use the fourth form, then you can write ``NAME & ""`` to force the interpretation as a *static_string_EXPRESSION*. Note: if the first argument is a valid ``TOOL_NAME``, it will be interpreted that way. The use of the ``TOOL_NAME`` argument is relevant only to users of SPARK and GNATprove, see last part of this section for details. Normally warnings are enabled, with the output being controlled by the command line switch. Warnings (``Off``) turns off generation of warnings until a Warnings (``On``) is encountered or the end of the current unit. If generation of warnings is turned off using this pragma, then some or all of the warning messages are suppressed, regardless of the setting of the command line switches. The ``Reason`` parameter may optionally appear as the last argument in any of the forms of this pragma. It is intended purely for the purposes of documenting the reason for the ``Warnings`` pragma. The compiler will check that the argument is a static string but otherwise ignore this argument. Other tools may provide specialized processing for this string. The form with a single argument (or two arguments if Reason present), where the first argument is ``ON`` or ``OFF`` may be used as a configuration pragma. If the ``LOCAL_NAME`` parameter is present, warnings are suppressed for the specified entity. This suppression is effective from the point where it occurs till the end of the extended scope of the variable (similar to the scope of ``Suppress``). This form cannot be used as a configuration pragma. In the case where the first argument is other than ``ON`` or ``OFF``, the third form with a single static_string_EXPRESSION argument (and possible reason) provides more precise control over which warnings are active. The string is a list of letters specifying which warnings are to be activated and which deactivated. The code for these letters is the same as the string used in the command line switch controlling warnings. For a brief summary, use the gnatmake command with no arguments, which will generate usage information containing the list of warnings switches supported. For full details see the section on ``Warning Message Control`` in the :title:`GNAT User's Guide`. This form can also be used as a configuration pragma. The warnings controlled by the :switch:`-gnatw` switch are generated by the front end of the compiler. The GCC back end can provide additional warnings and they are controlled by the :switch:`-W` switch. Such warnings can be identified by the appearance of a string of the form ``[-W{xxx}]`` in the message which designates the :switch:`-W{xxx}` switch that controls the message. The form with a single *static_string_EXPRESSION* argument also works for these warnings, but the string must be a single full :switch:`-W{xxx}` switch in this case. The above reference lists a few examples of these additional warnings. The specified warnings will be in effect until the end of the program or another pragma ``Warnings`` is encountered. The effect of the pragma is cumulative. Initially the set of warnings is the standard default set as possibly modified by compiler switches. Then each pragma Warning modifies this set of warnings as specified. This form of the pragma may also be used as a configuration pragma. The fourth form, with an ``On|Off`` parameter and a string, is used to control individual messages, based on their text. The string argument is a pattern that is used to match against the text of individual warning messages (not including the initial "warning: " tag). The pattern may contain asterisks, which match zero or more characters in the message. For example, you can use ``pragma Warnings (Off, "bits of*unused")`` to suppress the warning message ``warning: 960 bits of "a" unused``. No other regular expression notations are permitted. All characters other than asterisk in these three specific cases are treated as literal characters in the match. The match is case insensitive, for example XYZ matches xyz. Note that the pattern matches if it occurs anywhere within the warning message string (it is not necessary to put an asterisk at the start and the end of the message, since this is implied). The above use of patterns to match the message applies only to warning messages generated by the front end. This form of the pragma with a string argument can also be used to control warnings provided by the back end and mentioned above. By using a single full :switch:`-W{xxx}` switch in the pragma, such warnings can be turned on and off. There are two ways to use the pragma in this form. The OFF form can be used as a configuration pragma. The effect is to suppress all warnings (if any) that match the pattern string throughout the compilation (or match the -W switch in the back end case). The second usage is to suppress a warning locally, and in this case, two pragmas must appear in sequence: .. code-block:: ada pragma Warnings (Off, Pattern); ... code where given warning is to be suppressed pragma Warnings (On, Pattern); In this usage, the pattern string must match in the Off and On pragmas, and (if *-gnatw.w* is given) at least one matching warning must be suppressed. Note: if the ON form is not found, then the effect of the OFF form extends until the end of the file (pragma Warnings is purely textual, so its effect does not stop at the end of the enclosing scope). Note: to write a string that will match any warning, use the string ``"***"``. It will not work to use a single asterisk or two asterisks since this looks like an operator name. This form with three asterisks is similar in effect to specifying ``pragma Warnings (Off)`` except (if :switch:`-gnatw.w` is given) that a matching ``pragma Warnings (On, "***")`` will be required. This can be helpful in avoiding forgetting to turn warnings back on. Note: the debug flag :switch:`-gnatd.i` can be used to cause the compiler to entirely ignore all WARNINGS pragmas. This can be useful in checking whether obsolete pragmas in existing programs are hiding real problems. Note: pragma Warnings does not affect the processing of style messages. See separate entry for pragma Style_Checks for control of style messages. Users of the formal verification tool GNATprove for the SPARK subset of Ada may use the version of the pragma with a ``TOOL_NAME`` parameter. If present, ``TOOL_NAME`` is the name of a tool, currently either ``GNAT`` for the compiler or ``GNATprove`` for the formal verification tool. A given tool only takes into account pragma Warnings that do not specify a tool name, or that specify the matching tool name. This makes it possible to disable warnings selectively for each tool, and as a consequence to detect useless pragma Warnings with switch :switch:`-gnatw.w`. Pragma Weak_External ==================== Syntax: .. code-block:: ada pragma Weak_External ([Entity =>] LOCAL_NAME); ``LOCAL_NAME`` must refer to an object that is declared at the library level. This pragma specifies that the given entity should be marked as a weak symbol for the linker. It is equivalent to ``__attribute__((weak))`` in GNU C and causes ``LOCAL_NAME`` to be emitted as a weak symbol instead of a regular symbol, that is to say a symbol that does not have to be resolved by the linker if used in conjunction with a pragma Import. When a weak symbol is not resolved by the linker, its address is set to zero. This is useful in writing interfaces to external modules that may or may not be linked in the final executable, for example depending on configuration settings. If a program references at run time an entity to which this pragma has been applied, and the corresponding symbol was not resolved at link time, then the execution of the program is erroneous. It is not erroneous to take the Address of such an entity, for example to guard potential references, as shown in the example below. Some file formats do not support weak symbols so not all target machines support this pragma. .. code-block:: ada -- Example of the use of pragma Weak_External package External_Module is key : Integer; pragma Import (C, key); pragma Weak_External (key); function Present return boolean; end External_Module; with System; use System; package body External_Module is function Present return boolean is begin return key'Address /= System.Null_Address; end Present; end External_Module; Pragma Wide_Character_Encoding ============================== Syntax: .. code-block:: ada pragma Wide_Character_Encoding (IDENTIFIER | CHARACTER_LITERAL); This pragma specifies the wide character encoding to be used in program source text appearing subsequently. It is a configuration pragma, but may also be used at any point that a pragma is allowed, and it is permissible to have more than one such pragma in a file, allowing multiple encodings to appear within the same file. However, note that the pragma cannot immediately precede the relevant wide character, because then the previous encoding will still be in effect, causing "illegal character" errors. The argument can be an identifier or a character literal. In the identifier case, it is one of ``HEX``, ``UPPER``, ``SHIFT_JIS``, ``EUC``, ``UTF8``, or ``BRACKETS``. In the character literal case it is correspondingly one of the characters :kbd:`h`, :kbd:`u`, :kbd:`s`, :kbd:`e`, :kbd:`8`, or :kbd:`b`. Note that when the pragma is used within a file, it affects only the encoding within that file, and does not affect withed units, specs, or subunits. --- ### Gcc Ada Doc Gnat Rm Implementation Of Ada 2022 Features (`gcc/ada/doc/gnat_rm/implementation_of_ada_2022_features.rst`) .. _Implementation_of_Ada_2022_Features: *********************************** Implementation of Ada 2022 Features *********************************** .. index:: Ada 2022 implementation status .. index:: -gnat22 option (gcc) .. index:: pragma Ada_2022 .. index:: configuration pragma Ada_2022 .. index:: Ada_2022 configuration pragma This chapter contains a complete list of Ada 2022 features that have been implemented. Generally, these features are only available if the *-gnat22* (Ada 2022 features enabled) option is set, or if the configuration pragma ``Ada_2022`` is used. However, new pragmas, attributes, and restrictions are unconditionally available, since the Ada standard allows the addition of new pragmas, attributes, and restrictions (there are exceptions, which are documented in the individual descriptions), and also certain packages were made available in earlier versions of Ada. An ISO date (YYYY-MM-DD) appears in parentheses on the description line. This date shows the implementation date of the feature. Any wavefront subsequent to this date will contain the indicated feature, as will any subsequent releases. A date of 0000-00-00 means that GNAT has always implemented the feature, or implemented it as soon as it appeared as a binding interpretation. Each feature corresponds to an Ada Issue ('AI') approved by the Ada standardization group (ISO/IEC JTC1/SC22/WG9) for inclusion in Ada 2022. The section "RM references" lists all modified paragraphs in the Ada 2012 reference manual. The details of each modification as well as a complete description of the AIs may be found in http://www.ada-auth.org/AI12-SUMMARY.HTML. .. index:: AI12-0001 (Ada 2022 feature) * *AI12-0001 Independence and Representation clauses for atomic objects (2019-11-27)* The compiler accepts packing clauses in all cases, even if they have effectively no influence on the layout. Types, where packing is essentially infeasible are, for instance atomic, aliased and by-reference types. RM references: 13.02 (6.1/2) 13.02 (7) 13.02 (8) 13.02 (9/3) C.06 (8.1/3) C.06 (10) C.06 (11) C.06 (21) C.06 (24) .. index:: AI12-0003 (Ada 2022 feature) * *AI12-0003 Specifying the standard storage pool (2020-06-25)* Allows the standard storage pool being specified with a ``Default_Storage_Pool`` pragma or aspect. RM references: 8.02 (11) 13.11.03 (1/3) 13.11.03 (3.1/3) 13.11.03 (4/3) 13.11.03 (4.1/3) 13.11.03 (5/3) 13.11.03 (6.2/3) 13.11.03 (6.3/3) .. index:: AI12-0004 (Ada 2022 feature) * *AI12-0004 Normalization and allowed characters for identifiers (2020-06-11)* This AI clarifies that Ada identifiers containing characters which are not allowed in Normalization Form KC are illegal. RM references: 2.01 (4.1/3) 2.03 (4/3) A.03.02 (4/3) A.03.02 (32.5/3) A.03.05 (18/3) A.03.05 (51/3) .. index:: AI12-0020 (Ada 2022 feature) * *AI12-0020 'Image for all types (2020-03-30)* Put_Image prints out a human-readable representation of an object. The functionality in Ada2022 RM is fully implemented except the support for types in the ``Remote_Types`` packages. RM references: 4.10 (0) 3.05 (27.1/2) 3.05 (27.2/2) 3.05 (27.3/2) 3.05 (27.4/2) 3.05 (27.5/2) 3.05 (27.6/2) 3.05 (27.7/2) 3.05 (28) 3.05 (29) 3.05 (30/3) 3.05 (31) 3.05 (32) 3.05 (33/3) 3.05 (37.1/2) 3.05 (38) 3.05 (39) 3.05 (43/3) 3.05 (55/3) 3.05 (55.1/5) 3.05 (55.2/4) 3.05 (55.3/4) 3.05 (55.4/4) 3.05 (59) H.04 (23) H.04 (23.8/2) .. index:: AI12-0022 (Ada 2022 feature) * *AI12-0022 Raise_Expressions (2013-01-27)* This feature allows you to write "raise NAME [with STRING]" in an expression to rise given exception. It is particularly useful in the case of assertions such as preconditions allowing to specify which exception a precondition raises if it fails. RM references: 4.04 (3/3) 11.02 (6) 11.03 (2/2) 11.03 (3) 11.03 (3.1/2) 11.03 (4/2) 11.04.01 (10.1/3) .. index:: AI12-0027 (Ada 2022 feature) * *AI12-0027 Access values should never designate unaliased components (2020-06-15)* AI12-0027 adds a requirement for a value conversion that converts from an array of unaliased components to an array of aliased components to make a copy. It defines such conversions to have a local accessibility, effectively preventing the possibility of unsafe accesses to unaliased components. RM references: 4.06 (24.17/3) 4.06 (24.21/2) 4.06 (58) 6.02 (10/3) 3.10.02 (10/3) .. index:: AI12-0028 (Ada 2022 feature) * *AI12-0028 Import of variadic C functions (2020-03-03)* Ada programs can now properly call variadic C functions by means of the conventions C_Variadic_, for small integer values . RM references: B.03 (1/3) B.03 (60.15/3) B.03 (75) .. index:: AI12-0030 (Ada 2022 feature) * *AI12-0030 Formal derived types and stream attribute availability (2020-08-21)* Corner cases involving streaming operations for formal derived limited types that are now defined to raise Program_Error. Before, behavior in these cases was undefined. Stream attribute availability is more precisely computed in cases where a derived type declaration occurs ahead of a streaming attribute specification for the parent type. RM references: 12.05.01 (21/3) 13.13.02 (49/2) .. index:: AI12-0031 (Ada 2022 feature) * *AI12-0031 All_Calls_Remote and indirect calls (0000-00-00)* Remote indirect calls (i.e., calls through a remote access-to-subprogram type) behave the same as remote direct calls. RM references: E.02.03 (19/3) .. index:: AI12-0032 (Ada 2022 feature) * *AI12-0032 Questions on 'Old (2020-04-24)* AI12-0032 resolves several issues related to the 'Old attribute. The GNAT compiler already implemented what the AI requires in most of those cases, but two having to do with static and dynamic checking of the accessibility level of the constant object implicitly declared for an 'Old attribute reference were not yet implemented. Accessibility checking for these constants is now implemented as defined in the AI. RM references: 4.01.03 (9/3) 6.01.01 (22/3) 6.01.01 (26/3) 6.01.01 (35/3) .. index:: AI12-0033 (Ada 2022 feature) * *AI12-0033 Sets of CPUs when defining dispatching domains (0000-00-00)* The set of CPUs associated with a dispatching domain is no longer required to be a contiguous range of CPU values. RM references: D.16.01 (7/3) D.16.01 (9/3) D.16.01 (20/3) D.16.01 (23/3) D.16.01 (24/3) D.16.01 (26/3) .. index:: AI12-0035 (Ada 2022 feature) * *AI12-0035 Accessibility checks for indefinite elements of containers (0000-00-00)* If the element type for an instance of one of the indefinite container generics has an access discriminant, then accessibility checks (at run-time) prevent inserting a value into a container object if the value's discriminant designates an object that is too short-lived (that is, if the designated object has an accessibility level that is deeper than that of the instance). Without this check, dangling references would be possible. GNAT handled this correctly already before this AI was issued. RM references: A.18 (5/3) A.18.11 (8/2) A.18.12 (7/2) A.18.13 (8/2) A.18.14 (8/2) A.18.15 (4/2) A.18.16 (4/2) A.18.17 (7/3) A.18.18 (39/3) A.18.18 (47/3) .. index:: AI12-0036 (Ada 2022 feature) * *AI12-0036 The actual for an untagged formal derived type cannot be tagged (2019-10-21)* AI12-0036 is a binding interpretation that adds the following legality rule: The actual type for a formal derived type shall be tagged if and only if the formal derived type is a private extension. The check is implemented for all Ada dialects, not just Ada 2022. RM references: 12.05.01 (5.1/3) .. index:: AI12-0037 (Ada 2022 feature) * *AI12-0037 New types in Ada.Locales can't be converted to/from strings (2016-09-10)* The type definitions for Language_Code and Country_Code are now using dynamic predicates. RM references: A.19 (4/3) .. index:: AI12-0039 (Ada 2022 feature) * *AI12-0039 Ambiguity in syntax for membership expression removed (0000-00-00)* An ambiguity in the syntax for membership expressions was resolved. For example, "A in B and C" can be parsed in only one way because of this AI. RM references: 4.04 (3/3) 4.04 (3.2/3) 4.05.02 (3.1/3) 4.05.02 (4) 4.05.02 (4.1/3) 4.05.02 (27/3) 4.05.02 (27.1/3) 4.05.02 (28.1/3) 4.05.02 (28.2/3) 4.05.02 (29/3) 4.05.02 (30/3) 4.05.02 (30.1/3) 4.05.02 (30.2/3) 4.05.02 (30.3/3) 4.09 (11/3) 4.09 (32.6/3) 8.06 (27.1/3) 3.02.04 (17/3) .. index:: AI12-0040 (Ada 2022 feature) * *AI12-0040 Resolving the selecting_expression of a case_expression (0000-00-00)* The definition of "complete context" is corrected so that selectors of case expressions and of case statements are treated uniformly. RM references: 8.06 (9) .. index:: AI12-0041 (Ada 2022 feature) * *AI12-0041 Type_Invariant'Class for interface types (2016-12-12)* Subprogram calls within class-wide type invariant expressions get resolved as primitive operations instead of being dynamically dispatched. RM references: 7.03.02 (1/3) 7.03.02 (3/3) .. index:: AI12-0042 (Ada 2022 feature) * *AI12-0042 Type invariant checking rules (2020-06-05)* AI12-0042 adds rules for type invariants. Specifically, when inheriting a private dispatching operation when the ancestor operation is visible at the point of the type extension, the operation must be abstract or else overridden. In addition, for a class-wide view conversion from an object of a specific type T to which a type invariant applies, an invariant check is performed when the conversion is within the immediate scope of T. RM references: 7.03.02 (6/3) 7.03.02 (17/3) 7.03.02 (18/3) 7.03.02 (19/3) 7.03.02 (20/3) .. index:: AI12-0043 (Ada 2022 feature) * *AI12-0043 Details of the storage pool used when Storage_Size is specified (0000-00-00)* Clarify that a Storage_Size specification for an access type specifies both an upper bound and a lower bound (not just a lower bound) of the amount of storage allowed for allocated objects. RM references: 13.11 (18) .. index:: AI12-0044 (Ada 2022 feature) * *AI12-0044 Calling visible functions from type invariant expressions (2020-05-11)* AI05-0289-1 extends invariant checking to `in` parameters. However, this makes it impossible to call a public function of the type from an invariant expression, as that public function will attempt to check the invariant, resulting in an infinite recursion. This AI specifies, that type-invariant checking is performed on parameters of mode `in` upon return from procedure calls, but not of `in`-mode parameters in functions. RM references: 7.03.02 (19/3) .. index:: AI12-0045 (Ada 2022 feature) * *AI12-0045 Pre- and Postconditions are allowed on generic subprograms (2015-03-17)* The SPARK toolset now supports contracts on generic subprograms, packages and their respective bodies. RM references: 6.01.01 (1/3) .. index:: AI12-0046 (Ada 2022 feature) * *AI12-0046 Enforcing legality for anonymous access components in record aggregates (0000-00-00)* For a record aggregate of the form (X | Y => ....), any relevant legality rules are checked for both for X and Y. For example, .. code:: X : aliased constant String := ... ; type R is record F1 : access constant String; F2 : access String; end record; Obj : R := (F1 | F2 => X'Access); -- ok for F1, but illegal for F2 RM references: 4.03.01 (16/3) .. index:: AI12-0047 (Ada 2022 feature) * *AI12-0047 Generalized iterators and discriminant-dependent components (0000-00-00)* Iterating over the elements of an array is subject to the same legality checks as renaming the array. For example, if an assignment to an enclosing discriminated object could cause an array object to cease to exist then we don't allow renaming the array. So it is similarly not allowed to iterate over the elements of such an array. RM references: 5.05.02 (6/3) .. index:: AI12-0048 (Ada 2022 feature) * *AI12-0048 Default behavior of tasks on a multiprocessor with a specified dispatching policy (0000-00-00)* Clarify that if the user does not impose requirements about what CPUs a given task might execute on, then the implementation does not get to impose such requirements. This avoids potential problems with priority inversion. RM references: D.16.01 (30/3) .. index:: AI12-0049 (Ada 2022 feature) * *AI12-0049 Invariants need to be checked on the initialization of deferred constants (0000-00-00)* Invariant checking for deferred constants (and subcomponents thereof) is performed. Corrects a clear oversight in the previous RM wording. RM references: 7.03.02 (10/3) .. index:: AI12-0050 (Ada 2022 feature) * *AI12-0050 Conformance of quantified expressions (2016-07-22)* Compiler rejects a subprogram body when an expression for a boolean formal parameter includes a quantified expression, and the subprogram declaration contains a textual copy of the same. RM references: 6.03.01 (20) 6.03.01 (21) .. index:: AI12-0051 (Ada 2022 feature) * *AI12-0051 The Priority aspect can be specified when Attach_Handler is specified (0000-00-00)* Previous RM wording had two contradictory rules for determining (in some cases) the priority of a protected subprogram that is attached to an interrupt. The AI clarifies which one of the rules takes precedence. RM references: D.03 (10/3) .. index:: AI12-0052 (Ada 2022 feature) * *AI12-0052 Implicit objects are considered overlapping (0000-00-00)* Clarify that the rules about unsynchronized concurrent access apply as one would expect in the case of predefined routines that access Text_IO's default input and default output files. There was no compiler changes needed to implement this. RM references: A (3/2) A.10.03 (21) .. index:: AI12-0054-2 (Ada 2022 feature) * *AI12-0054-2 Aspect Predicate_Failure (0000-00-00)* New aspect Predicate_Failure is defined. A solution for the problem that a predicate like .. code:: subtype Open_File is File with Dynamic_Predicate =\> Is_Open (Open_File) or else (raise File_Not_Open); does the wrong thing in the case of a membership test. RM references: 3.02.04 (14/3) 3.02.04 (31/3) 3.02.04 (35/3) .. index:: AI12-0055 (Ada 2022 feature) * *AI12-0055 All properties of a usage profile are defined by pragmas (2020-06-09)* AI12-0055 allows the use of the No_Dynamic_CPU_Assignment restriction in pragmas Restrictions and Restrictions_Warnings. RM references: D.07 (10/3) D.13 (6/3) D.13 (8/3) D.13 (10/3) .. index:: AI12-0059 (Ada 2022 feature) * *AI12-0059 Object_Size attribute (2019-12-02)* AI12-0059 brings GNAT-defined attribute Object_Size to Ada standard and clarifies its semantics. Given that the attribute already existed in GNAT compiler, the feature is supported for all language versions. RM references: 4.09.01 (2/3) 13.01 (14) 13.01 (23) 13.03 (9/3) 13.03 (50/2) 13.03 (51) 13.03 (52) 13.03 (58) .. index:: AI12-0061 (Ada 2022 feature) * *AI12-0061 Iterated component associations in array aggregates (2016-09-01)* Ada issue AI12-061 introduces a new construct in array aggregates allowing component associations to be parameterized by a loop variable, for example: .. code:: array (1 .. 10) of Integer := (for I in 1 .. 10 => I ** 2); type Matrix is array (Positive range <>, Positive range <>) of Float; G : constant Matrix := (for I in 1 .. 4 => (for J in 1 .. 4 => (if I=J then 1.0 else 0.0))); -- Identity matrix The expression in such an association can also be a function that returns a limited type, and the range can be specified by the 'others' choice. RM references: 4.03.03 (5/2) 4.03.03 (6) 4.03.03 (17/3) 4.03.03 (20) 4.03.03 (23.1/4) 4.03.03 (32/3) 4.03.03 (43) 3.01 (6/3) 3.03 (6) 3.03 (18.1/3) 3.03.01 (23/3) 5.05 (6) 8.01 (2.1/4) .. index:: AI12-0062 (Ada 2022 feature) * *AI12-0062 Raise exception with failing string function (0000-00-00)* Clarify that if raising exception E1 is accompanied with a String-valued expression whose evaluation raises exception E2, then E2 is what gets propagated. RM references: 11.03 (4/2) .. index:: AI12-0065 (Ada 2022 feature) * *AI12-0065 Descendants of incomplete views (0000-00-00)* This AI is a clarification of potentially confusing wording. GNAT correctly handles the example given in AARM 7.3.1(5.b-5.d), which illustrates the topic of this AI. RM references: 7.03.01 (5.2/3) .. index:: AI12-0067 (Ada 2022 feature) * *AI12-0067 Accessibility level of explicitly aliased parameters of procedures and entries (0000-00-00)* The AI fixes a case where the intent was fairly obvious but the RM wording failed to mention a case (with the result that the accessibility level of an explicitly aliased parameter of a procedure or entry was undefined even though the intent was clear). RM references: 3.10.02 (7/3) .. index:: AI12-0068 (Ada 2022 feature) * *AI12-0068 Predicates and the current instance of a subtype (2020-05-06)* AI12-0068 is a binding interpretation that defines the current instance name in a type or subtype aspect to be a value rather than an object. This affects attributes whose prefix is a current instance in predicates, type invariants, and ``Default_Initial_Condition`` aspects. In particular, in the case of the ``Constrained`` attribute the value will always be True, and formerly legal attributes that require an object as their prefix (such as ``Size``, ``Access``, ``Address``, etc.) are illegal when applied to a current instance in type and subtype aspects. RM references: 8.06 (17/3) .. index:: AI12-0069 (Ada 2022 feature) * *AI12-0069 Inconsistency in Tree container definition (0000-00-00)* The description of how iteration over a Tree container's elements was contradictory in some cases regarding whether a cursor designating the Root node is included in the iteration. This contradiction was resolved. In the "!ACATS Test" section of the AI, it says that if an implementation were to get this wrong then almost any attempt to iterate over any tree would fail at runtime. RM references: A.18.10 (153/3) A.18.10 (155/3) A.18.10 (157/3) A.18.10 (159/3) .. index:: AI12-0070 (Ada 2022 feature) * *AI12-0070 9.3(2) does not work for anonymous access types (0000-00-00)* The RM contained some old wording about the master of an allocated object that only made sense for named access types. The AI clarifies the wording to clearly state the scope of validity and ensures that the paragraph does not contradict 3.10.2's rules for anonymous access types. RM references: 3.10.02 (13.1/3) 9.03 (2) .. index:: AI12-0071 (Ada 2022 feature) * *AI12-0071 Order of evaluation when multiple predicates apply (2015-08-10)* AI12-0071 specifies the semantics of multiple/inherited predicates on a single subtype. RM references: 3.02.04 (4/3) 3.02.04 (6/3) 3.02.04 (30/3) 3.02.04 (31/3) 3.02.04 (32/3) 3.02.04 (33/3) 3.02.04 (35/3) 3.05.05 (7.1/3) 3.05.05 (7.2/3) 3.05.05 (7.3/3) 3.08.01 (10.1/3) 3.08.01 (15/3) 4.05.02 (29/3) 4.05.02 (30/3) 4.06 (51/3) 4.09.01 (10/3) 5.04 (7/3) 5.05 (9/3) 13.09.02 (3/3) 13.09.02 (12) .. index:: AI12-0072 (Ada 2022 feature) * *AI12-0072 Missing rules for Discard_Names aspect (0000-00-00)* Clarify that Discard_Names is an aspect, not just a pragma. RM references: C.05 (1) C.05 (5) C.05 (7/2) C.05 (8) .. index:: AI12-0073 (Ada 2022 feature) * *AI12-0073 Synchronous Barriers are not allowed with Ravenscar (2020-02-24)* Ada 2022 adds (as a binding interpretation) a ``No_Dependence => Ada.Synchronous_Barriers`` restriction to the Ravenscar profile. RM references: D.13 (6/3) .. index:: AI12-0074 (Ada 2022 feature) * *AI12-0074 View conversions and out parameters passed by copy (2020-03-26)* This Ada 2022 AI makes illegal some cases of out parameters whose type has a ``Default_Value`` aspect. RM references: 4.06 (56) 6.04.01 (6.25/3) 6.04.01 (13.1/3) .. index:: AI12-0075 (Ada 2022 feature) * *AI12-0075 Static expression functions (2020-04-13)* Ada 2022 defines a new aspect ``Static`` that can be specified on expression functions. Such an expression function can be called in contexts requiring static expressions when the actual parameters are all static, allowing for greater abstraction in complex static expressions. RM references: 4.09 (21) 6.08 (3/4) 6.08 (5/4) 6.08 (6/4) 7.03.02 (8.2/5) 7.03.02 (15/4) 7.03.02 (16/4) 7.03.02 (17/4) 7.03.02 (19/4) 7.03.02 (20/5) .. index:: AI12-0076 (Ada 2022 feature) * *AI12-0076 Variable state in pure packages (0000-00-00)* Defines an obscure constant-modifying construct to be erroneous. The issue is that the current instance of a type is a variable object, so the following is legal: .. code:: type T; type T_Ref (Access_To_Variable : access T) is null record; type T is limited record Self : T_Ref (T'Access); Int : Integer; end record; Obj : constant T := (Self => <>, Int => 123); begin Obj.Self.Access_To_Variable.Int := 456; -- modifying a component of a constant In cases where constancy is really needed (e.g., for an object declared in a Pure context), such a case needs to be erroneous. RM references: 10.02.01 (17/3) E.02.02 (17/2) .. index:: AI12-0077 (Ada 2022 feature) * *AI12-0077 Has_Same_Storage on objects of size zero (2020-03-30)* This binding interpretation requires the Has_Same_Storage attribute to return always `false` for objects that have a size of zero. RM references: 13.03 (73.4/3) .. index:: AI12-0078 (Ada 2022 feature) * *AI12-0078 Definition of node for tree container is confusing (0000-00-00)* Clarifies the expected behavior in processing tree containers. RM references: A.18.10 (2/3) A.18.10 (3/3) .. index:: AI12-0081 (Ada 2022 feature) * *AI12-0081 Real-time aspects need to specify when they are evaluated (0000-00-00)* Clarify the point at which Priority and Interrupt_Priority aspect expressions are evaluated. RM references: D.01 (17/3) D.16 (9/3) .. index:: AI12-0084 (Ada 2022 feature) * *AI12-0084 Box expressions in array aggregates (2014-12-15)* This AI addresses an issue where compiler used to fail to initialize components of a multidimensional aggregates with box initialization when scalar components have a specified default value. The AI clarifies that in an array aggregate with box (i.e., ``<>``) component values, the ``Default_Component_Value`` of the array type (if any) should not be ignored. RM references: 4.03.03 (23.1/2) .. index:: AI12-0085 (Ada 2022 feature) * *AI12-0085 Missing aspect cases for Remote_Types (0000-00-00)* A distributed systems annex (Annex E) clarification. Aspect specifications that are forbidden using attribute definition clause syntax are also forbidden using aspect_specification syntax. RM references: E.02.02 (17/2) .. index:: AI12-0086 (Ada 2022 feature) * *AI12-0086 Aggregates and variant parts (2019-08-14)* In Ada 2012, a discriminant value that governs an active variant part in an aggregate had to be static. AI12-0086 relaxes this restriction: If the subtype of the discriminant value is a static subtype all of whose values select the same variant, then the expression for the discriminant is allowed to be nonstatic. RM references: 4.03.01 (17/3) 4.03.01 (19/3) .. index:: AI12-0088 (Ada 2022 feature) * *AI12-0088 UTF_Encoding.Conversions and overlong characters on input (0000-00-00)* Clarify that overlong characters are acceptable on input even if we never generate them as output. RM references: A.04.11 (54/3) A.04.11 (55/3) .. index:: AI12-0089 (Ada 2022 feature) * *AI12-0089 Accessibility rules need to take into account that a generic function is not a (0000-00-00)* Fix cases in RM wording where the accessibility rules for a function failed to take into account the fact that a generic function is not a function. For example, a generic function with an explicitly aliased parameter should be able to return references to that parameter in the same ways that a (non-generic) function can. The previous wording did not allow that. RM references: 3.10.02 (7/3) 3.10.02 (19.2/3) 3.10.02 (19.3/3) 6.05 (4/3) .. index:: AI12-0093 (Ada 2022 feature) * *AI12-0093 Iterator with indefinite cursor (0000-00-00)* A clarification that confirms what GNAT is already doing. RM references: 5.05.02 (8/3) 5.05.02 (10/3) .. index:: AI12-0094 (Ada 2022 feature) * *AI12-0094 An access_definition should be a declarative region (0000-00-00)* Fixes wording omission in the RM, confirming that the behaviour of GNAT is correct. RM references: 8.03 (2) 8.03 (26/3) .. index:: AI12-0095 (Ada 2022 feature) * *AI12-0095 Generic formal types and constrained partial views (0000-00-00)* Deciding whether an actual parameter corresponding to an explicitly aliased formal parameter is legal depends on (among other things) whether the parameter type has a constrained partial view. The AI clarifies how this compile-time checking works in the case of a generic formal type (assume the best in the spec and recheck each instance, assume the worst in a generic body). RM references: 3.10.02 (27.2/3) 4.06 (24.16/2) 6.04.01 (6.2/3) 12.05.01 (15) .. index:: AI12-0096 (Ada 2022 feature) * *AI12-0096 The exception raised when a subtype conversion fails a predicate check (0000-00-00)* Clarify that the Predicate_Failure aspect works the same in a subtype conversion as in any other context. RM references: 4.06 (57/3) .. index:: AI12-0097 (Ada 2022 feature) * *AI12-0097 Tag of the return object of a simple return expression (0000-00-00)* Clarify wording about the tag of a function result in the case of a simple (i.e. not extended) return statement in a function with a class-wide result type. RM references: 6.05 (8/3) .. index:: AI12-0098 (Ada 2022 feature) * *AI12-0098 Problematic examples for ATC (0000-00-00)* The AI clarifies reference manual examples, there is no compiler impact. RM references: 9.07.04 (13) .. index:: AI12-0099 (Ada 2022 feature) * *AI12-0099 Wording problems with predicates (2020-05-04)* When extending a task or protected type from an ancestor interface subtype with a predicate, a link error can occur due to the compiler failing to generate the predicate-checking function. This AI clarifies the requirement for such predicate inheritance for concurrent types. RM references: 3.02.04 (4/4) 3.02.04 (12/3) 3.02.04 (20/3) .. index:: AI12-0100 (Ada 2022 feature) * *AI12-0100 A qualified expression makes a predicate check (2020-02-17)* The compiler now enforces predicate checks on qualified expressions when the qualifying subtype imposes a predicate. RM references: 4.07 (4) .. index:: AI12-0101 (Ada 2022 feature) * *AI12-0101 Incompatibility of hidden untagged record equality (2019-10-31)* AI12-0101 is a binding interpretation that removes a legality rule that prohibited the declaration of a primitive equality function for a private type in the private part of its enclosing package (either before or after the completion of the type) when the type is completed as an untagged record type. Such declarations are now accepted in Ada 2012 and later Ada versions. As a consequence of this work, some cases where the implementation of AI05-0123 was incomplete were corrected. More specifically, if a user-defined equality operator is present for an untagged record type in an Ada 2012 program, that user-defined equality operator will be (correctly) executed in some difficult-to-characterize cases where the predefined component-by-component comparison was previously being (incorrectly) executed. This can arise, for example, in the case of the predefined equality operation for an enclosing composite type that has a component of the user-defined primitive equality op's operand type. This correction means that the impact of this change is not limited solely to code that was previously rejected at compile time. RM references: 4.05.02 (9.8/3) .. index:: AI12-0102 (Ada 2022 feature) * *AI12-0102 Stream_IO.File_Type has Preelaborable_Initialization (0000-00-00)* Modifies the declaration of one type in a predefined package. GNAT's version of ``Ada.Streams.Stream_IO`` already had this modification (the ``Preelaborable__Initialization`` pragma). RM references: A.12.01 (5) .. index:: AI12-0103 (Ada 2022 feature) * *AI12-0103 Expression functions that are completions in package specifications (0000-00-00)* Clarifies that expression functions that are completions do not cause "general" freeze-everybody-in-sight freezing like a subprogram body. RM references: 13.14 (3/3) 13.14 (5/3) .. index:: AI12-0104 (Ada 2022 feature) * *AI12-0104 Overriding an aspect is undefined (0000-00-00)* A clarification of the wording in RM, no compiler impact. RM references: 4.01.06 (4/3) 4.01.06 (17/3) .. index:: AI12-0105 (Ada 2022 feature) * *AI12-0105 Pre and Post are not allowed on any subprogram completion (0000-00-00)* Language-defined aspects (e.g., ``Post``) cannot be specified as part of the completion of a subprogram declaration. Fix a hole in the RM wording to clarify that this general rule applies even in the special cases where the completion is either an expression function or a null procedure. RM references: 13.01.01 (18/3) .. index:: AI12-0106 (Ada 2022 feature) * *AI12-0106 Write'Class aspect (0000-00-00)* Clarify that the syntax used in an ACATS test BDD2005 for specifying a class-wide streaming aspect is correct. RM references: 13.01.01 (28/3) 13.13.02 (38/3) .. index:: AI12-0107 (Ada 2022 feature) * *AI12-0107 A prefixed view of a By_Protected_Procedure interface has convention protected (2020-06-05)* A prefixed view of a subprogram with aspect Synchronization set to By_Protected_Procedure has convention protected. RM references: 6.03.01 (10.1/2) 6.03.01 (12) 6.03.01 (13) .. index:: AI12-0109 (Ada 2022 feature) * *AI12-0109 Representation of untagged derived types (2019-11-12)* Ada disallows a nonconforming specification of a type-related representation aspect of an untagged by-reference type. The motivation for this rule is to ensure that a parent type and a later type derived from the parent agree with respect to such aspects. AI12-0109 disallows a construct that otherwise could be used to get around this rule: an aspect specification for the parent type that occurs after the declaration of the derived type. RM references: 13.01 (10/3) .. index:: AI12-0110 (Ada 2022 feature) * *AI12-0110 Tampering checks are performed first (2020-04-14)* AI12-0110 requires tampering checks in the containers library to be performed first, before any other checks. RM references: A.18.02 (97.1/3) A.18.03 (69.1/3) A.18.04 (15.1/3) A.18.07 (14.1/3) A.18.10 (90/3) A.18.18 (35/3) .. index:: AI12-0112 (Ada 2022 feature) * *AI12-0112 Contracts for container operations (0000-00-00)* A representation change replacing english descriptions of contracts for operations on predefined container types with pre/post-conditions. No compiler impact. RM references: A.18.02 (99/3) 11.04.02 (23.1/3) 11.05 (23) 11.05 (26) A (4) A.18 (10) .. index:: AI12-0114 (Ada 2022 feature) * *AI12-0114 Overlapping objects designated by access parameters are not thread-safe (0000-00-00)* There are rules saying that concurrent calls to predefined subprograms don't interfere with each other unless actual parameters overlap. The AI clarifies that such an interference is also possible if overlapping objects are reachable via access dereferencing from actual parameters of the two calls. RM references: A (3/2) .. index:: AI12-0116 (Ada 2022 feature) * *AI12-0116 Private types and predicates (0000-00-00)* Clarify that the same aspect cannot be specified twice for the same type. ``Dynamic_Predicate``, for example, can be specified on either the partial view of a type or on the completion in the private part, but not on both. RM references: 13.01 (9/3) 13.01 (9.1/3) .. index:: AI12-0117 (Ada 2022 feature) * *AI12-0117 Restriction No_Tasks_Unassigned_To_CPU (2020-06-12)* This AI adds a restriction No_Tasks_Unassigned_To_CPU to provide safe use of Ravenscar. The CPU aspect is specified for the environment task. No CPU aspect is specified to be statically equal to ``Not_A_Specific_CPU``. If aspect CPU is specified (dynamically) to the value ``Not_A_Specific_CPU``, then Program_Error is raised. If Set_CPU or ``Delay_Until_And_Set_CPU`` are called with the CPU parameter equal to ``Not_A_Specific_CPU``, then ``Program_Error`` is raised. RM references: D.07 (10.8/3) .. index:: AI12-0120 (Ada 2022 feature) * *AI12-0120 Legality and exceptions of generalized loop iteration (0000-00-00)* Clarify that the expansion-based definition of generalized loop iteration includes legality checking. If the expansion would be illegal (for example, because of passing a constant actual parameter in a call when the mode of the corresponding formal parameter is in-out), then the loop is illegal too. RM references: 5.05.02 (6.1/4) 5.05.02 (10/3) 5.05.02 (13/3) .. index:: AI12-0121 (Ada 2022 feature) * *AI12-0121 Stream-oriented aspects (0000-00-00)* Clarify that streaming-oriented aspects (e.g., Read) can be specified using aspect_specification syntax, not just via an attribute definition clause. RM references: 13.13.02 (38/3) .. index:: AI12-0124 (Ada 2022 feature) * *AI12-0124 Add Object'Image (2017-03-24)* The corrigendum of Ada 2012 extends attribute ``'Image following`` the syntax for the GNAT ``'Img`` attribute. This AI fixes a gap in the earlier implementation, which did not recognize function calls and attributes that are functions as valid object prefixes. RM references: 3.05 (55/3) .. index:: AI12-0125-3 (Ada 2022 feature) * *AI12-0125-3 Add @ as an abbreviation for the LHS of an assignment (2016-11-11)* This AI introduces the use of the character '@' as an abbreviation for the left-hand side of an assignment statement, usable anywhere within the expression on the right-hand side. To use this feature the compilation flag -gnat2022 must be specified. RM references: 5.02.01 (0) 2.02 (9) 3.03 (21.1/3) 4.01 (2/3) 8.06 (9/4) .. index:: AI12-0127 (Ada 2022 feature) * *AI12-0127 Partial aggregate notation (2016-10-12)* This AI describes a new constructor for aggregates, in terms of an existing record or array object, and a series of component-wise modifications of its value, given by named associations for the modified components. To use this feature the compilation flag ``-gnat2022`` must be specified. RM references: 4.03 (2) 4.03 (3/2) 4.03 (4) 4.03.01 (9) 4.03.01 (15/3) 4.03.01 (16/4) 4.03.01 (17/5) 4.03.01 (17.1/2) 4.03.03 (4) 4.03.03 (14) 4.03.03 (17/5) 4.03.04 (0) 7.05 (2.6/2) .. index:: AI12-0128 (Ada 2022 feature) * *AI12-0128 Exact size access to parts of composite atomic objects (2019-11-24)* According to this AI, the compiler generates full access to atomic composite objects even if the access is only partial in the source code. To use this feature the compilation flag ``-gnat2022`` must be specified. RM references: C.06 (13.2/3) C.06 (19) C.06 (20) C.06 (22/2) C.06 (25/4) .. index:: AI12-0129 (Ada 2022 feature) * *AI12-0129 Make protected objects more protecting (2020-07-01)* A new aspect Exclusive_Functions has been added to the language to force the use of read/write locks on protected functions when needed. RM references: 9.05.01 (2) 9.05.01 (4) 9.05.01 (5) 9.05.01 (7) 9.05.03 (15) 9.05.03 (23) .. index:: AI12-0130 (Ada 2022 feature) * *AI12-0130 All I/O packages should have Flush (2016-07-03)* The Flush routine has been added for the ``Sequential_IO`` and ``Direct_IO`` standard packages in the Ada 2012 COR.1:2016. The Flush routine here is equivalent to the one found in ``Text_IO``. The ``Flush`` procedure synchronizes the external file with the internal file (by flushing any internal buffers) without closing the file. RM references: A.08.01 (10) A.08.02 (28/3) A.08.04 (10) A.10.03 (21) A.12.01 (28/2) A.12.01 (28.6/1) .. index:: AI12-0131 (Ada 2022 feature) * *AI12-0131 Inherited Pre'Class when unspecified on initial subprogram (0000-00-00)* If T1 is a tagged type with a primitive P that has no class-wide precondition, and if T2 is an extension of T1 which overrides the inherited primitive P, then that overriding P is not allowed to have a class-wide precondition. Allowing it would be ineffective except in corner cases where it would be confusing. RM references: 6.01.01 (17/3) 6.01.01 (18/3) .. index:: AI12-0132 (Ada 2022 feature) * *AI12-0132 Freezing of renames-as-body (2020-06-13)* This AI clarifies that a renames-as-body freezes the expression of any expression function that it renames. RM references: 13.14 (5/3) .. index:: AI12-0133 (Ada 2022 feature) * *AI12-0133 Type invariants and default initialized objects (0000-00-00)* Clarify that invariant checking for a default-initialized object is performed regardless of where the object is declared (in particular, even when the full view of the type is visible). RM references: 7.03.02 (10.3/3) .. index:: AI12-0135 (Ada 2022 feature) * *AI12-0135 Enumeration types should be eligible for convention C (0000-00-00)* Ada previously allowed but did not require supporting specifying convention C for an enumeration type. Now it is required that an implementation shall support it. RM references: B.01 (14/3) B.01 (41/3) B.03 (65) .. index:: AI12-0136 (Ada 2022 feature) * *AI12-0136 Language-defined packages and aspect Default_Storage_Pool (0000-00-00)* Clarify that the effect of specifying Default_Storage_Pool for an instance of a predefined generic is implementation-defined. No compiler impact. RM references: 13.11.03 (5/3) .. index:: AI12-0137 (Ada 2022 feature) * *AI12-0137 Incomplete views and access to class-wide types (0000-00-00)* If the designated type of an access type is incomplete when the access type is declared, then we have rules about whether we get a complete view when a value of the access type is dereferenced. Clarify that analogous rules apply if the designated type is class-wide. RM references: 3.10.01 (2.1/2) .. index:: AI12-0138 (Ada 2022 feature) * *AI12-0138 Iterators of formal derived types (2021-02-11)* AI12-0138 specifies the legality rules for confirming specifications of nonoverridable aspects. This completes the legality checks for aspect ``Implicit_Dereference`` and simplifies the checks for those aspects that are inherited operations. RM references: 13.01.01 (18/4) 13.01.01 (34/3) 4.01.05 (6/3) 4.01.06 (5/3) 4.01.06 (6/3) 4.01.06 (7/3) 4.01.06 (8/3) 4.01.06 (9/3) 5.05.01 (11/3) .. index:: AI12-0140 (Ada 2022 feature) * *AI12-0140 Access to unconstrained partial view when full view is constrained (0000-00-00)* Clarify some confusion about about whether what matters when checking whether designated subtypes statically match is the view of the designated type that is currently available v.s. the view that was available when the access type was declared. RM references: 3.02 (7/2) 7.03.01 (5/1) .. index:: AI12-0143 (Ada 2022 feature) * *AI12-0143 Using an entry index of a family in a precondition (2022-04-05)* Ada 2022 adds the ``Index`` attribute, which allows the use of the entry family index of an entry call within preconditions and post-conditions. RM references: 6.01.01 (30/3) 9.05.04 (5/3) .. index:: AI12-0144 (Ada 2022 feature) * *AI12-0144 Make Discrete_Random more flexible (2020-01-31)* A new function Random with First/Last parameters is provided in the ``Ada.Numerics.Discrete_Random`` package. RM references: A.05.02 (20) A.05.02 (32) A.05.02 (41) A.05.02 (42) .. index:: AI12-0145 (Ada 2022 feature) * *AI12-0145 Pool_of_Subpool returns null when called too early (0000-00-00)* Clarify that if you ask for the pool of a subpool (by calling ``Pool_Of_Subpool``) before ``Set_Pool_of_Subpool`` is called, then the result is null. RM references: 13.11.04 (20/3) .. index:: AI12-0147 (Ada 2022 feature) * *AI12-0147 Expression functions and null procedures can be declared in a protected_body (2015-03-05)* AI12-0147 specifies that null procedures and expression functions are now allowed in protected bodies. RM references: 9.04 (8/1) .. index:: AI12-0149 (Ada 2022 feature) * *AI12-0149 Type invariants are checked for functions returning access-to-type (0000-00-00)* Extend the rule saying that ``Type_Invariant`` checks are performed for access-to-T parameters (where T has a specified ``Type_Invariant``) so that the rule also applies to function results. RM references: 7.03.02 (19.3/4) .. index:: AI12-0150 (Ada 2022 feature) * *AI12-0150 Class-wide type invariants and statically bound calls (0000-00-00)* The same approach used in AI12-0113 to ensure that contract-related calls associated with a call to a subprogram "match" with respect to dispatching also applies to ``Type_Invariant`` checking. RM references: 7.03.02 (3/3) 7.03.02 (5/3) 7.03.02 (9/3) 7.03.02 (22/3) .. index:: AI12-0154 (Ada 2022 feature) * *AI12-0154 Aspects of library units (0000-00-00)* Clarify that an aspect_specification for a library unit is equivalent to a corresponding aspect-specifying pragma. RM references: 13.01.01 (32/3) .. index:: AI12-0156 (Ada 2022 feature) * *AI12-0156 Use subtype_indication in generalized iterators (0000-00-00)* For iterating over an array, we already allow (but do not require) explicitly providing a subtype indication in an iterator_specification. Tee AI generalizes this to handle the case where the element type of the array is of an anonymous access type. This also allows (but does not require) explicitly naming the cursor subtype in a generalized iterator. The main motivation for allowing these new cases is improving readability by making it easy to infer the (sub)type of the iteration object just by looking at the loop. RM references: 5.05.02 (2/3) 5.05.02 (5/4) 5.05.02 (7/3) 3.10.02 (11.1/2) .. index:: AI12-0157 (Ada 2022 feature) * *AI12-0157 Missing rules for expression functions (0000-00-00)* Clarify that an expression function behaves like a single-return-statement function in more cases: it can return an aggregate without extra parens, the expression has an applicable index constraint, and the same accessibility rules apply in both cases. For instance, the code below is legal: .. code:: subtype S is String (1 .. 10); function f return S is (others => '?'); RM references: 3.10.02 (19.2/4) 3.10.02 (19.3/4) 4.03.03 (11/2) 6.08 (2/3) 6.08 (3/3) 6.08 (5/3) 6.08 (6/3) 6.08 (7/3) 7.05 (2.9/3) 13.14 (5.1/4) 13.14 (5.2/4) 13.14 (8/3) 13.14 (10.1/3) 13.14 (10.2/3) 13.14 (10.3/3) .. index:: AI12-0160 (Ada 2022 feature) * *AI12-0160 Adding an indexing aspect to an indexable container type (0000-00-00)* If the parent type of a derived type has exactly one of the two indexing aspects (that is, constant_indexing and variable_indexing) specified, then the derived type cannot have a specification for the other one. RM references: 4.01.06 (6/4) 4.01.06 (9/4) 3.06 (22.2/3) .. index:: AI12-0162 (Ada 2022 feature) * *AI12-0162 Memberships and Unchecked_Unions (0000-00-00)* Clarify that membership tests for unchecked_union types work consistently when testing membership in more than one subtype (X in AA | BB | CC) as when testing for one. RM references: B.03.03 (25/2) .. index:: AI12-0164 (Ada 2022 feature) * *AI12-0164 Max_Entry_Queue_Length aspect for entries (2019-06-11)* AI12-0164 defines pragma and aspect ``Max_Entry_Queue_Length`` in addition to the GNAT-specific equivalents ``Max_Queue_Length`` and ``Max_Entry_Queue_Depth``. RM references: D.04 (16) .. index:: AI12-0165 (Ada 2022 feature) * *AI12-0165 Operations of class-wide types and formal abstract subprograms (2021-10-19)* Ada 2022 specifies that when the controlling type of a formal abstract subprogram declaration is a formal type, and the actual type is a class-wide type T'Class, the actual subprogram can be an implicitly declared subprogram corresponding to a primitive operation of type T. RM references: 12.06 (8.5/2) .. index:: AI12-0166 (Ada 2022 feature) * *AI12-0166 External calls to protected functions that appear to be internal calls (2016-11-15)* According to this AI, the compiler rejects a call to a protected operation when the call appears within a precondition for another protected operation. RM references: 6.01.01 (34/3) 9.05 (3/3) 9.05 (7.1/3) .. index:: AI12-0167 (Ada 2022 feature) * *AI12-0167 Type_Invariants and tagged-type View Conversions (0000-00-00)* This AI clarifies that no invariant check is performed in a case where an invariant-violating value is assigned to a component. This confirms the current compiler behavior. RM references: 7.03.02 (9/4) .. index:: AI12-0168 (Ada 2022 feature) * *AI12-0168 Freezing of generic instantiations of generics with bodies (0000-00-00)* Adjust freezing rules to be compatible with AI12-0103-1. The change confirms the current compiler behavior. RM references: 13.14 (3/4) .. index:: AI12-0169 (Ada 2022 feature) * *AI12-0169 Aspect specifications for entry bodies (0000-00-00)* Change syntax to allow aspect specifications for implementation-defined aspects on entry bodies. The change doesn't influence any of the language-defined aspects and is solely required for SPARK. RM references: 9.05.02 (5) .. index:: AI12-0170 (Ada 2022 feature) * *AI12-0170 Abstract subprogram calls in class-wide precondition expressions (2020-07-06)* This AI specifies rules for calls to abstract functions within class-wide preconditions and postconditions. RM references: 3.09.03 (7) 6.01.01 (7/4) 6.01.01 (18/4) 6.01.01 (18.2/4) .. index:: AI12-0172 (Ada 2022 feature) * *AI12-0172 Raise expressions in limited contexts (2019-07-29)* The compiler has been enhanced to support the use of raise expressions in limited contexts. RM references: 7.05 (2.1/3) .. index:: AI12-0173 (Ada 2022 feature) * *AI12-0173 Expression of an extended return statement (0000-00-00)* Fix the wording related to expression of an extended return statement that was made ambiguous by changes of syntax in other AI's. No compiler changes involved. RM references: 6.05 (3/2) 6.05 (5/3) .. index:: AI12-0174 (Ada 2022 feature) * *AI12-0174 Aggregates of Unchecked_Unions using named notation (0000-00-00)* In many cases, it is illegal to name a discriminant of an unchecked_union type. Relax this rule to allow the use of named notation in an aggregate of an unchecked_union type. RM references: B.03.03 (9/3) .. index:: AI12-0175 (Ada 2022 feature) * *AI12-0175 Preelaborable packages with address clauses (2020-03-20)* The compiler nows accepts calls to certain functions that are essentially unchecked conversions in preelaborated library units. To use this feature the compilation flag ``-gnat2022`` must be specified. RM references: 10.02.01 (7) .. index:: AI12-0179 (Ada 2022 feature) * *AI12-0179 Failure of postconditions of language-defined units (0000-00-00)* A clarification that expressing postconditions for predefined units via RM wording or via ``Post`` aspect specifications are equivalent. In particular, the expression in such a ``Post`` aspect specification should not yield False. No implementation changes needed. RM references: 1.01.03 (17/3) 11.04.02 (23.1/3) .. index:: AI12-0180 (Ada 2022 feature) * *AI12-0180 Using protected subprograms and entries within an invariant (2020-06-22)* AI12-0180 makes entries and protected subprograms directly visible within Invariant aspects of a task or protected type. RM references: 13.01.01 (12/3) .. index:: AI12-0181 (Ada 2022 feature) * *AI12-0181 Self-referencing representation aspects (0000-00-00)* Clarify that a name or expression which freezes an entity cannot occur in an aspect specification for that entity. RM references: 13.01 (9/4) 13.01 (9.1/4) 13.14 (19) .. index:: AI12-0182 (Ada 2022 feature) * *AI12-0182 Pre'Class and protected operations (0000-00-00)* Confirm that Pre'Class and Post'Class cannot be specified for a protected operation. No language change. RM references: 13.01.01 (16/3) .. index:: AI12-0184 (Ada 2022 feature) * *AI12-0184 Long Long C Data Types (2020-01-30)* Two new types ``long_long`` and ``unsigned_long_long`` are introduced in the package ``Interfaces.C``. RM references: B.03 (71.3/3) .. index:: AI12-0185 (Ada 2022 feature) * *AI12-0185 Resolution of postcondition-specific attributes (0000-00-00)* Clarify resolution rules for ``'Old`` and ``'Result`` attribute references to match original intent. RM references: 6.01.01 (7/4) 6.01.01 (8/3) 6.01.01 (26.10/4) 6.01.01 (29/3) .. index:: AI12-0186 (Ada 2022 feature) * *AI12-0186 Profile freezing for the Access attribute (0000-00-00)* Clarify that the use of Some_Subprogram'Access does not freeze the profile of Some_Subprogram. RM references: 13.14 (15) .. index:: AI12-0187 (Ada 2022 feature) * *AI12-0187 Stable properties of abstract data types (2020-11-04)* Ada 2022 defines a new aspect, ``Stable_Properties``, for use in generating additional postcondition checks for subprograms. RM references: 7.03.04 (0) 13.01.01 (4/3) .. index:: AI12-0191 (Ada 2022 feature) * *AI12-0191 Clarify "part" for type invariants (0000-00-00)* Clarify that for purposes of determining whether an invariant check is required for a "part" of an object, we do not look at "parts" which do not correspond to "parts" of the nominal type of the object. For example, if we have a parameter Param of a tagged type T1 (or equivalently of type T1'Class), and type T2 is an extension of T1 which declares a component Foo, and T1'Class (Param)'Tag = T2'Tag, then no invariant check is performed for Param's Foo component (or any subcomponent thereof). RM references: 3.03 (23/5) 3.09.01 (4.1/2) 6.08 (5.8/5) 7.03.02 (8.3/5) 7.03.02 (8.4/5) 7.03.02 (8.5/5) 7.03.02 (8.6/5) 7.03.02 (8.7/5) 7.03.02 (8.8/5) 7.03.02 (8.9/5) 7.03.02 (8.10/5) 7.03.02 (8.11/5) 7.03.02 (8.12/5) 7.03.02 (10.1/4) 7.03.02 (15/5) 7.03.02 (17/4) 7.03.02 (18/4) 7.03.02 (19/4) 13.13.02 (9/3) .. index:: AI12-0192 (Ada 2022 feature) * *AI12-0192 "requires late initialization" and protected types (2020-03-11)* This AI clarifies that components of a protected type require late initialization when their initialization references (implicitly) the current instance of the type. RM references: 3.03.01 (8.1/2) .. index:: AI12-0194 (Ada 2022 feature) * *AI12-0194 Language-defined aspects and entry bodies (0000-00-00)* The AI Includes entry bodies on the list of bodies for which no language-defined aspects can be specified (although specifying an implementation-defined aspect may be allowed). A wording change, no implementation impact. RM references: 13.01.01 (17/3) .. index:: AI12-0195 (Ada 2022 feature) * *AI12-0195 Inheriting body but overriding precondition or postcondition (2021-08-11)* Ada 2022 specifies that if a primitive with a class-wide precondition or postcondition is inherited, and some primitive function called in the class-wide precondition or postcondition is overridden, then a dispatching call to the first primitive with a controlling operand that has the tag of the overriding type is required to check both the interpretation using the overriding function and the interpretation using the original overridden function. RM references: 6.01.01 (38/4) .. index:: AI12-0196 (Ada 2022 feature) * *AI12-0196 Concurrent access to Ada container libraries (0000-00-00)* Clarify that parallel execution of operations which use cursors to refer to different elements of the same container does not violate the rules about erroneous concurrent access in some cases. That is, if C1 and C2 are cursors that refer to different elements of some container, then it is ok to concurrently execute an operation that is passed C1 and which accesses one element of the container, with another operation (perhaps the same operation, perhaps not) that is passed C2 and which accesses another element of the container. RM references: A.18 (2/2) A.18.02 (125/2) A.18.02 (133/3) A.18.02 (135/3) A.18.03 (81/3) A.18.04 (36/3) A.18.07 (34/2) A.18.10 (116/3) .. index:: AI12-0198 (Ada 2022 feature) * *AI12-0198 Potentially unevaluated components of array aggregates (2020-05-13)* Ada 2022 enforces the detection of components that belong to a nonstatic or null range of index values of an array aggregate. RM references: 6.01.01 (22.1/4) .. index:: AI12-0199 (Ada 2022 feature) * *AI12-0199 Abstract subprogram calls in class-wide invariant expressions (0000-00-00)* Class-wide type invariants do not apply to abstract types, to avoid various problems. Define the notion of a "corresponding expression" for a class-wide type invariant, replacing references to components as appropriate, taking into account rules for corresponding and specified discriminants when applying them to a nonabstract descendant. RM references: 7.03.02 (5/4) 7.03.02 (8/3) .. index:: AI12-0201 (Ada 2022 feature) * *AI12-0201 Missing operations of static string types (2020-02-25)* Relational operators and type conversions of static string types are now static in Ada 2022. RM references: 4.09 (9) 4.09 (19) 4.09 (20) 4.09 (24) .. index:: AI12-0203 (Ada 2022 feature) * *AI12-0203 Overriding a nonoverridable aspect (0000-00-00)* A corner case wording clarification that has no impact on compilers. RM references: 4.01.05 (5.1/4) 4.01.05 (7/3) .. index:: AI12-0204 (Ada 2022 feature) * *AI12-0204 Renaming of a prefixed view (2020-02-24)* AI12-0204 clarifies that the prefix of a prefixed view that is renamed or passed as a formal subprogram must be renameable as an object. RM references: 8.05.04 (5.2/2) 12.06 (8.3/2) 4.01.03 (13.1/2) 4.01.06 (9/5) .. index:: AI12-0205 (Ada 2022 feature) * *AI12-0205 Defaults for generic formal types (2021-04-01)* AI12-0205 specifies syntax and semantics that provide defaults for formal types of generic units. The legality rules guarantee that the default subtype_mark that is specified for a formal type would be a legal actual in any instantiation of the generic unit. RM references: 12.03 (7/3) 12.03 (10) 12.05 (2.1/3) 12.05 (2.2/3) 12.05 (7/2) .. index:: AI12-0206 (Ada 2022 feature) * *AI12-0206 Nonoverridable should allow arbitrary kinds of aspects (0000-00-00)* A non-overridable aspect can have a value other than a name; for example, ``Max_Entry_Queue_Length`` is non-overridable and it has a scalar value. Part of adding support for ``Max_Entry_Queue_Length`` (which is already supported by GNAT). RM references: 13.01.01 (18.2/4) 13.01.01 (18.3/4) 13.01.01 (18.6/4) .. index:: AI12-0207 (Ada 2022 feature) * *AI12-0207 Convention of anonymous access types (2020-02-01)* The convention of anonymous access elements of arrays now have the same convention as the array instead of convention Ada. RM references: 6.03.01 (13.1/3) B.01 (19) B.01 (21/3) .. index:: AI12-0208 (Ada 2022 feature) * *AI12-0208 Predefined Big numbers support (0000-00-00)* Add predefined package ``Ada.Numerics.Big_Numbers``. RM references: A.05.05 (0) A.05.06 (0) A.05.07 (0) .. index:: AI12-0211 (Ada 2022 feature) * *AI12-0211 Interface types and inherited nonoverridable aspects (2020-08-24)* AI12-0211 introduces two new legality rules for Ada 2022. The first says that if a nonoverridable aspect is explicitly specified for a type that also inherits that aspect from another type (an ancestor or a progenitor), then the explicit aspect specification shall be confirming. The second says that if a type inherits a nonoverridable aspect from two different sources (this can only occur if at least one of the two is an interface type), then the two sources shall agree with respect to the given aspect. This AI is a binding interpretation, so these checks are performed even for earlier Ada versions. Because of compatibility concerns, an escape mechanism for suppressing these legality checks is provided: these new checks always pass if the ``-gnatd.M`` switch (relaxed RM semantics) is specified. RM references: 13.01.01 (18.3/5) 13.01.01 (18.4/4) .. index:: AI12-0212 (Ada 2022 feature) * *AI12-0212 Container aggregates; generalized array aggregates (0000-00-00)* The AI defines a new feature: generalized array aggregates that already exists in GNAT. RM references: 4.03.05 (0) 1.01.04 (12) 1.01.04 (13) 2.01 (15) 2.02 (9/5) 3.07.01 (3) 3.08.01 (4) 4.03 (2/5) 4.03 (3/5) 4.03.01 (5) 4.03.03 (3/2) 4.03.03 (4/5) 4.03.03 (5.1/5) 4.03.03 (9) 4.03.03 (17/5) 4.03.03 (21) 4.03.03 (23.2/5) 4.03.03 (26) 4.03.03 (27) 4.03.03 (31) 4.03.04 (4/5) 4.04 (3.1/3) 11.02 (3) 13.01.01 (5/3) 13.01.01 (7/3) A.18.02 (8/3) A.18.02 (14/2) A.18.02 (47/2) A.18.02 (175/2) A.18.03 (6/3) A.18.05 (3/3) A.18.06 (4/3) A.18.08 (3/3) A.18.09 (4/3) .. index:: AI12-0216 (Ada 2022 feature) * *AI12-0216 6.4.1(6.16-17/3) should never apply to composite objects (0000-00-00)* Fix wording so that parameter passing cases where there isn't really any aliasing problems or evaluation order dependency are classified as acceptable. No compiler impact. RM references: 6.04.01 (6.17/3) .. index:: AI12-0217 (Ada 2022 feature) * *AI12-0217 Rules regarding restrictions on the use of the Old attribute are too strict (2020-03-25)* AI12-0217 loosens the rules regarding what is allowed as the prefix of a 'Old attribute reference. In particular, a prefix is now only required to "statically name" (as opposed to the previous "statically denote") an object. This means that components of composite objects that previously would have been illegal are now legal prefixes. RM references: 6.01.01 (24/3) 6.01.01 (27/3) .. index:: AI12-0220 (Ada 2022 feature) * *AI12-0220 Pre/Post for access-to-subprogram types (2020-04-14)* Contract aspects can now be specified for access-to-subprogram types, as defined for Ada 2022 in this AI. RM references: 6.01.01 (1/4) 6.01.01 (2/3) 6.01.01 (4/3) 6.01.01 (19/3) 6.01.01 (28/3) 6.01.01 (29/3) 6.01.01 (39/3) 13.01.01 (12/5) .. index:: AI12-0222 (Ada 2022 feature) * *AI12-0222 Representation aspects and private types (0000-00-00)* Clarify that the rule against specifying a representation aspect for a type before the type is completely defined also applies in the case where aspect_specification syntax is used (not just in the case where a pragma or some other kind of representation item is used). GNAT already implements this. RM references: 13.01 (9/5) 13.01 (9.1/4) 13.01 (9.2/5) .. index:: AI12-0225 (Ada 2022 feature) * *AI12-0225 Prefix of Obj'Image (0000-00-00)* Clarify some Object vs. Value corner cases to allow names that do not denote objects in more contexts, such as a qualified expression as a prefix of an Image attribute. RM references: 3.05 (55.1/4) .. index:: AI12-0226 (Ada 2022 feature) * *AI12-0226 Make objects more consistent (0000-00-00)* Allow value conversions as objects. For instance this example becomes legal: ``Long_Integer (Duration'Last)'Image``. RM references: 3.03 (11.1/3) 3.03 (21.1/3) 3.03 (23.8/5) 4.06 (58.1/4) 4.06 (58.3/4) .. index:: AI12-0227 (Ada 2022 feature) * *AI12-0227 Evaluation of nonstatic universal expressions when no operators are involved (0000-00-00)* Nonstatic universal integer expressions are always evaluated at runtime as values of type root_integer; similarly, nonstatic universal real expressions are always evaluated at runtime as values of type root_real. This AI corrects a wording oversight. Previously, the above was only true if a call to operator was involved. With this change it is true in all cases. No compiler impact. RM references: 4.04 (10) 8.06 (29) .. index:: AI12-0228 (Ada 2022 feature) * *AI12-0228 Properties of qualified expressions used as names (2020-02-19)* This AI clarifies that properties of a qualified object pass through a qualified expression used as a name. Specifically, "aliased" and "known to be constrained" are not changed by a qualified expression. RM references: 3.03 (23.7/3) 3.10 (9/3) .. index:: AI12-0231 (Ada 2022 feature) * *AI12-0231 Null_Task_Id and Activation_Is_Complete (0000-00-00)* Add ``Activation_Is_Complete`` to the list of functions that raise P_E if passed ``Null_Task_Id``, correcting an oversight. RM references: C.07.01 (15) .. index:: AI12-0232 (Ada 2022 feature) * *AI12-0232 Rules for pure generic bodies (0000-00-00)* Clarify the rules for a generic body nested in a pure library unit. RM references: 10.02.01 (9/3) 10.02.01 (15.1/3) 10.02.01 (15.5/3) .. index:: AI12-0233 (Ada 2022 feature) * *AI12-0233 Pre'Class for hidden operations of private types (0000-00-00)* Clarify how ``Pre'Class`` checking interacts with private-part overriding of inherited subprograms. A class-wide precondition can be checked at runtime even if it is specified in a private part that the caller cannot see into. RM references: 6.01.01 (38/4) .. index:: AI12-0234 (Ada 2022 feature) * *AI12-0234 Compare-and-swap for atomic objects (0000-00-00)* New predefined units for atomic operations (``System.Atomic_Operations`` and child units thereof). RM references: C.06.01 (0) C.06.02 (0) .. index:: AI12-0235 (Ada 2022 feature) * *AI12-0235 System.Storage_Pools should be pure (0000-00-00)* Change the predefined package System.Storage_Pools from preelaborated to pure. RM references: 13.11 (5) .. index:: AI12-0236 (Ada 2022 feature) * *AI12-0236 declare expressions (2020-04-08)* A ``declare expression`` allows constant objects and renamings to be declared within an expression. RM references: 2.08 (6) 3.09.02 (3) 3.10.02 (9.1/3) 3.10.02 (16.1/3) 3.10.02 (32.2/3) 4.03.02 (5.4/3) 4.03.03 (15.1/3) 4.04 (7/3) 4.05.09 (0) 6.02 (10/4) 7.05 (2.1/5) 8.01 (2.1/4) .. index:: AI12-0237 (Ada 2022 feature) * *AI12-0237 Getting the representation of an enumeration value (2020-01-31)* The GNAT-specific attributes ``Enum_Rep`` and ``Enum_Val`` have been standardized and are now also supported as Ada 2022 attributes. RM references: 13.04 (10) 13.04 (11/3) .. index:: AI12-0242 (Ada 2022 feature) * *AI12-0242 Shorthand Reduction Expressions for Objects (0000-00-00)* Allow reduction expressions to iterate over an an array or an iterable object without having to explicitly create a value sequence. This allows, for instance, writing ``A'Reduce("+", 0)`` instead of the equivalent (but more verbose) ``[for Value of A => Value]'Reduce("+", 0);``. RM references: 4.05.10 (0) 4.01.04 (6) .. index:: AI12-0247 (Ada 2022 feature) * *AI12-0247 Potentially Blocking goes too far for Detect_Blocking (0000-00-00)* During a protected action, a call on a subprogram that contains a potentially blocking operation is considered a bounded error (so raising P_E is optional). This rule imposed an unreasonable implementation burden. The new rule introduced by this AI allows ignoring (i.e., not detecting) the problem until execution of a potentially blocking operation is actually attempted. RM references: 9.05 (55/5) 9.05 (56/5) 9.05.01 (18/5) H.05 (5/2) .. index:: AI12-0249 (Ada 2022 feature) * *AI12-0249 User-defined numeric literals (2020-04-07)* Compiler support is added for three new aspects (``Integer_Literal``, ``Real_Literal``, and ``String_Literal``) as described in AI12-0249 (for ``Integer_Literal`` and ``Real_Literal``), AI12-0295 (for ``String_Literal``), and in two follow-up AIs (AI12-0325 and AI12-0342). For pre-Ada 2022 versions of Ada, these are treated as implementation-defined aspects. Some implementation work remains, particularly in the interactions between these aspects and tagged types. RM references: 4.02 (9) 4.02.01 (0) 4.09 (3) .. index:: AI12-0250 (Ada 2022 feature) * *AI12-0250 Iterator Filters (2020-05-19)* This AI defines Ada 2022 feature of iterator filters, which can be applied to loop parameter specifications and iterator specifications. RM references: 4.03.03 (21) 4.03.03 (26) 4.03.03 (31) 4.03.05 (0) 4.05.10 (0) 5.05 (4) 5.05 (7) 5.05 (9/4) 5.05 (9.1/4) 5.05 (10) 5.05.02 (2/3) 5.05.02 (10/3) 5.05.02 (11/3) .. index:: AI12-0252 (Ada 2022 feature) * *AI12-0252 Duplicate interrupt handlers under Ravenscar (2018-07-05)* Ada Issue AI12-0252 requires that the runtime shall terminate with a Program_Error when more than one interrupt handler is attached to the same interrupt and the restriction No_Dynamic_Attachment is in effect. RM references: C.03.01 (13) .. index:: AI12-0254 (Ada 2022 feature) * *AI12-0254 Bounded_Indefinite_Holders (2025-12-15)* This AI defines a new package ``Ada.Containers.Indefinite_Holders`` which is fully implemented by GNAT. RM references: A.18.32 (0) .. index:: AI12-0256 (Ada 2022 feature) * *AI12-0256 Aspect No_Controlled_Parts (2021-01-26)* The compiler now supports the Ada 2022 aspect No_Controlled_Parts (see AI12-0256). When specified for a type, this aspect requires that the type and any of its ancestors must not have any controlled parts. RM references: H.04.01 (0) 13.01.01 (18.7/5) .. index:: AI12-0258 (Ada 2022 feature) * *AI12-0258 Containers and controlled element types (0000-00-00)* Most predefined containers are allowed to defer finalization of container elements until the finalization of the container. This allows implementation flexibility but causes problems in some cases. AI12-0258 tightens up the rules for the indefinite containers to say that finalization happens earlier - if a client needs the tighter finalization guarantees, then it can use the indefinite containers (even if the element subtype in question is definite). Other solutions involving the holder generic are also possible. GNAT implements these tighter element finalization requirements for instances of the indefinite container generics. RM references: A.18 (10/4) .. index:: AI12-0259 (Ada 2022 feature) * *AI12-0259 Lower bound of strings returned from Ada.Command_Line (0000-00-00)* Specify that the low-bound of a couple of predefined String-valued functions will always be one. RM references: A.15 (14) A.15 (16/3) .. index:: AI12-0260 (Ada 2022 feature) * *AI12-0260 Functions Is_Basic and To_Basic in Wide_Characters.Handling (2020-04-01)* AI12-0260 is implemented for Ada 2022, providing the new functions ``Is_Basic`` and ``To_Basic`` in package ``Ada.Wide_Characters.Handling``. RM references: 1.02 (8/3) A.03.05 (8/3) A.03.05 (20/3) A.03.05 (21/3) A.03.05 (33/3) A.03.05 (61/3) .. index:: AI12-0261 (Ada 2022 feature) * *AI12-0261 Conflict in "private with" rules (0000-00-00)* If a library unit is only visible at some point because of a "private with", there are legality rules about a name denoting that entity. The AI cleans up the wording so that it captures the intent in a corner case involving a private-child library-unit subprogram. The previous wording incorrectly caused this case to be illegal. RM references: 10.01.02 (12/3) 10.01.02 (13/2) 10.01.02 (14/2) 10.01.02 (15/2) 10.01.02 (16/2) .. index:: AI12-0262 (Ada 2022 feature) * *AI12-0262 Map-Reduce attribute (0000-00-00)* The AI defines Reduction Expressions to allow the programmer to apply the Map-Reduce paradigm to map/transform a set of values to a new set of values, and then summarize/reduce the transformed values into a single result value. RM references: 4.01.04 (1) 4.01.04 (6) 4.01.04 (11) 4.05.10 (0) .. index:: AI12-0263 (Ada 2022 feature) * *AI12-0263 Update references to ISO/IEC 10646 (0000-00-00)* Change RM references to ISO/IEC 10646:2011 to instead refer to ISO/IEC 10646:2017. No compiler impact. RM references: 1.01.04 (14.2/3) 2.01 (1/3) 2.01 (3.1/3) 2.01 (4/3) 2.01 (4.1/5) 2.01 (5/3) 2.01 (15/3) 2.01 (4.1/5) 2.01 (5/3) 2.03 (4.1/5) 2.03 (5/3) 3.05.02 (2/3) 3.05.02 (3/3) 3.05.02 (4/3) A.01 (36.1/3) A.01 (36.2/3) A.03.02 (32.6/5) A.03.05 (51.2/5) A.03.05 (55/3) A.03.05 (59/3) A.04.10 (3/3) B.05 (21/5) .. index:: AI12-0264 (Ada 2022 feature) * *AI12-0264 Overshifting and overrotating (0000-00-00)* Clarify Shift and Rotate op behavior with large shift/rotate amounts. RM references: B.02 (9) .. index:: AI12-0265 (Ada 2022 feature) * *AI12-0265 Default_Initial_Condition for types (2020-11-13)* The aspect ``Default_Initial_Condition``, originally proposed by SPARK and supported in GNAT, is now also included in Ada 2022. One change from the original implementation is that when the aspect is specified on ancestor types of a derived type, the ancestors' check expressions also apply to the derived type. ``Default_Initial_Condition`` checks are also now applied in cases of default initialization of components, allocators, ancestor parts of extension aggregates, and box associations of aggregates. RM references: 7.03.03 (0) 1.01.03 (17.1/5) 11.04.02 (23.2/5) 11.04.02 (23.3/5) .. index:: AI12-0269 (Ada 2022 feature) * *AI12-0269 Aspect No_Return for functions reprise (2020-03-19)* This amendment has been implemented under the ``-gnat2022`` switch, and the compiler now accepts the aspect/pragma No_Return for functions and generic functions. RM references: 6.05.01 (0) 6.05.01 (1/3) 6.05.01 (3.1/3) 6.05.01 (3.4/3) 6.05.01 (5/2) 6.05.01 (6/2) 6.05.01 (7/2) J.15.02 (2/3) J.15.02 (3/3) J.15.02 (4/3) .. index:: AI12-0272 (Ada 2022 feature) * *AI12-0272 (part 1) Pre/Postconditions for formal subprograms (0000-00-00)* Pre and Post aspects can be specified for a generic formal subprogram. ``Default_Initial_Condition`` can be specified for a generic formal private type. GNAT implements this with an exception of the part related to ``Default_Initial_Condition``. RM references: 6.01.01 (1/5) 6.01.01 (39/5) 7.03.03 (1/5) 7.03.03 (2/5) 7.03.03 (8/5) 7.03.04 (5/5) F.01 (1) .. index:: AI12-0275 (Ada 2022 feature) * *AI12-0275 Make subtype_mark optional in object renames (2020-01-28)* AI12-0275 allows object renamings to be declared without an explicit subtype_mark or access_definition. This feature can be used by compiling with the switch ``-gnat2022``. RM references: 8.05.01 (2/3) 8.05.01 (3/2) .. index:: AI12-0277 (Ada 2022 feature) * *AI12-0277 The meaning of "accessibility level of the body of F" (2020-09-09)* Clarify that the only time that an explicitly aliased formal parameter has different accessibility properties than an aliased part of a "normal" parameter is for the accessibility checking associated with a return statement. RM references: 3.10.02 (19.2/4) .. index:: AI12-0278 (Ada 2022 feature) * *AI12-0278 Implicit conversions of anonymous return types (0000-00-00)* If a call to a function with an anonymous-access-type result is converted to a named access type, it doesn't matter whether the conversion is implicit or explicit. the AI fixes hole where the previous rules didn't cover the implicit conversion case. RM references: 3.10.02 (10.3/3) .. index:: AI12-0279 (Ada 2022 feature) * *AI12-0279 Nonpreemptive dispatching needs more dispatching points (2020-04-17)* Ada 2022 defines a new aspect `Yield` that can be specified in the declaration of a noninstance subprogram (including a generic formal subprogram), a generic subprogram, or an entry, to ensure that the associated subprogram has at least one task dispatching point during each invocation. RM references: D.02.01 (1.5/2) D.02.01 (7/5) .. index:: AI12-0280-2 (Ada 2022 feature) * *AI12-0280-2 Making 'Old more flexible (2020-07-24)* For Ada 2022, AI12-0280-2 relaxes Ada's restrictions on 'Old attribute references whose attribute prefix does not statically name an entity. Previously, it was required that such an attribute reference must be unconditionally evaluated when the postcondition is evaluated; with the new rule, conditional evaluation is permitted if the relevant conditions can be evaluated upon entry to the subprogram with the same results as evaluation at the time of the postcondition's evaluation. In this case, the 'Old attribute prefix is evaluated conditionally (more specifically, the prefix is evaluated only if the result of that evaluation is going to be referenced later when the postcondition is evaluated). RM references: 6.01.01 (20/3) 6.01.01 (21/3) 6.01.01 (22/3) 6.01.01 (22.1/4) 6.01.01 (22.2/5) 6.01.01 (23/3) 6.01.01 (24/3) 6.01.01 (26/4) 6.01.01 (27/5) 6.01.01 (39/5) .. index:: AI12-0282 (Ada 2022 feature) * *AI12-0282 Atomic, Volatile, and Independent generic formal types (0000-00-00)* The AI specifies that the aspects ``Atomic``, ``Volatile``, ``Independent``, ``Atomic_Components``, ``Volatile_Components``, and ``Independent_Components`` are specifiable for generic formal types. The actual type must have a matching specification. RM references: C.06 (6.1/3) C.06 (6.3/3) C.06 (6.5/3) C.06 (6.8/3) C.06 (12/3) C.06 (12.1/3) C.06 (21/4) .. index:: AI12-0285 (Ada 2022 feature) * *AI12-0285 Syntax for Stable_Properties aspects (0000-00-00)* The AI establishes the required named notation for a Stable_Properties aspect specification in order to avoid syntactic ambiguities. With the old syntax, an example like .. code:: type Ugh is ... with Stable_Properties =\> Foo, Bar, Nonblocking, Pack; was problematic; ``Nonblocking`` and ``Pack`` are other aspects, while ``Foo`` and ``Bar`` are ``Stable_Properties`` functions. With the clarified syntax, the example above shall be written as: .. code:: type Ugh is ... with Stable_Properties => (Foo, Bar), Nonblocking, Pack; RM references: 7.03.04 (2/5) 7.03.04 (3/5) 7.03.04 (4/5) 7.03.04 (6/5) 7.03.04 (7/5) 7.03.04 (9/5) 7.03.04 (10/5) 7.03.04 (14/5) 13.01.01 (4/5) .. index:: AI12-0287 (Ada 2022 feature) * *AI12-0287 Legality Rules for null exclusions in renaming are too fierce (2020-02-17)* The null exclusion legality rules for generic formal object matching and object renaming now only apply to generic formal objects with mode in out. RM references: 8.05.01 (4.4/2) 8.05.01 (4.5/2) 8.05.01 (4.6/2) 8.05.04 (4.2/2) 12.04 (8.3/2) 12.04 (8.4/2) 12.04 (8.5/2) 12.04 (8.2/5) 12.06 (8.2/5) .. index:: AI12-0289 (Ada 2022 feature) * *AI12-0289 Implicitly null excluding anonymous access types and conformance (2020-06-09)* AI12-0289 is implemented for Ada 2022, allowing safer use of access parameters when the partial view of the designated type is untagged, but the full view is tagged. RM references: 3.10 (26) .. index:: AI12-0290 (Ada 2022 feature) * *AI12-0290 Restriction Pure_Barriers (2020-02-18)* The GNAT implementation of the Pure_Barriers restriction has been updated to match the Ada RM's definition as specified in this AI. Some constructs that were accepted by the previous implementation are now rejected, and vice versa. In particular, the use of a component of a component of a protected record in a barrier expression, as in "when Some_Component.Another_Component =>", formerly was (at least in some cases) not considered to be a violation of the Pure_Barriers restriction; that is no longer the case. RM references: D.07 (2) D.07 (10.10/4) .. index:: AI12-0291 (Ada 2022 feature) * *AI12-0291 Jorvik Profile (2020-02-19)* The Jorvik profile is now implemented, as defined in this AI. For Ada 2012 and earlier versions of Ada, Jorvik is an implementation-defined profile whose definition matches its Ada 2022 definition. RM references: D.13 (0) D.13 (1/3) D.13 (4/3) D.13 (6/4) D.13 (9/3) D.13 (10/3) D.13 (11/4) D.13 (12/4) .. index:: AI12-0293 (Ada 2022 feature) * *AI12-0293 Add predefined FIFO_Streams packages (0000-00-00)* The AI adds ``Ada.Streams.Storage`` and its two subunits ``Bounded`` and ``Unbounded``. RM references: 13.13.01 (1) 13.13.01 (9) 13.13.01 (9.1/1) .. index:: AI12-0295 (Ada 2022 feature) * *AI12-0295 User-defined string (2020-04-07)* Compiler support is added for three new aspects (``Integer_Literal``, ``Real_Literal``, and ``String_Literal``) as described in AI12-0249 (for ``Integer_Literal`` and ``Real_Literal``), AI12-0295 (for ``String_Literal``), and in two follow-up AIs (AI12-0325 and AI12-0342). For pre-Ada 2022 versions of Ada, these are treated as implementation-defined aspects. Some implementation work remains, particularly in the interactions between these aspects and tagged types. RM references: 4.02 (6) 4.02 (10) 4.02 (11) 3.06.03 (1) 4.02.01 (0) 4.09 (26/3) .. index:: AI12-0301 (Ada 2022 feature) * *AI12-0301 Predicates should be checked like constraints for types with Default_Value (2020-02-25)* This AI clarifies that predicate checks apply for objects that are initialized by default and that are of a type that has any components whose subtypes specify ``Default_Value`` or ``Default_Component_Value``. RM references: 3.02.04 (31/4) .. index:: AI12-0304 (Ada 2022 feature) * *AI12-0304 Image attributes of language-defined types (2020-07-07)* According to this AI, ``Put_Image`` (and therefore ``'Image``) is provided for the containers and for unbounded strings. RM references: 4.10 (0) .. index:: AI12-0306 (Ada 2022 feature) * *AI12-0306 Split null array aggregates from positional array aggregates (0000-00-00)* The AI clarifies the wording of the references RM paragraphs without introducing any language changes. RM references: 4.03.03 (2) 4.03.03 (3/2) 4.03.03 (9/5) 4.03.03 (26/5) 4.03.03 (26.1/5) 4.03.03 (33/3) 4.03.03 (38) 4.03.03 (39) 4.03.03 (42) .. index:: AI12-0307 (Ada 2022 feature) * *AI12-0307 Resolution of aggregates (2020-08-13)* The proposed new syntax for aggregates in Ada 2022 uses square brackets as delimiters, and in particular allows ``[]`` as a notation for empty array and container aggregates. This syntax is currently available as an experimental feature under the ``-gnatX`` flag. RM references: 4.03 (3/5) .. index:: AI12-0309 (Ada 2022 feature) * *AI12-0309 Missing checks for pragma Suppress (0000-00-00)* The AI includes some previously overlooked run-time checks in the list of checks that are potentially suppressed via a pragma ``Suppress``. For example, AI12-0251-1 adds a check that the number of chunks in a chunk_specification is not zero or negative. Clarify that suppressing ``Program_Error_Check`` suppresses that check too. RM references: 11.05 (10) 11.05 (19) 11.05 (20) 11.05 (22) 11.05 (24) .. index:: AI12-0311 (Ada 2022 feature) * *AI12-0311 Suppressing client-side assertions for language-defined units (0000-00-00)* The AI defines some new assertion policies that can be given as arguments in a Suppress pragma (e.g., Calendar_Assertion_Check). GNAT recognizes and ignores those new policies, the checks are not implemented. RM references: 11.04.02 (23.5/5) 11.05 (23) 11.05 (26) .. index:: AI12-0315 (Ada 2022 feature) * *AI12-0315 Image Attributes subclause improvements (0000-00-00)* Clarify that a named number or similar can be the prefix of an Image attribute reference. RM references: 4.10 (0) .. index:: AI12-0318 (Ada 2022 feature) * *AI12-0318 No_IO should apply to Ada.Directories (2020-01-31)* The restriction No_IO now applies to and prevents the use of the ``Ada.Directories package``. RM references: H.04 (20/2) H.04 (24/3) .. index:: AI12-0321 (Ada 2022 feature) * *AI12-0321 Support for Arithmetic Atomic Operations and Test and Set (0000-00-00)* The AI adds some predefined atomic operations, e.g. package System.``Atomic_Operations.Test_And_Set``. RM references: C.06.03 (0) C.06.04 (0) .. index:: AI12-0325 (Ada 2022 feature) * *AI12-0325 Various issues with user-defined literals (2020-04-07)* Compiler support is added for three new aspects (``Integer_Literal``, ``Real_Literal``, and ``String_Literal``) as described in AI12-0249 (for ``Integer_Literal`` and ``Real_Literal``), AI12-0295 (for ``String_Literal``), and in two follow-up AIs (AI12-0325 and AI12-0342). For pre-Ada 2022 versions of Ada, these are treated as implementation-defined aspects. Some implementation work remains, particularly in the interactions between these aspects and tagged types. RM references: 4.02 (6) 4.02 (10) 4.02 (11) 4.02.01 (0) .. index:: AI12-0329 (Ada 2022 feature) * *AI12-0329 Naming of FIFO_Streams packages (0000-00-00)* The AI changes the name of predefined package ``Ada.Streams.FIFO_Streams`` to ``Ada.Streams.Storage``. RM references: 13.13.01 (9/5) 13.13.01 (9.1/5) .. index:: AI12-0331 (Ada 2022 feature) * *AI12-0331 Order of finalization of a subpool (0000-00-00)* Clarify that when a subpool is being finalized, objects allocated from that subpool are finalized before (not after) they cease to exist (i.e. object's storage has been reclaimed). RM references: 13.11.05 (5/3) 13.11.05 (6/3) 13.11.05 (7/3) 13.11.05 (7.1/4) 13.11.05 (8/3) 13.11.05 (9/3) .. index:: AI12-0333 (Ada 2022 feature) * *AI12-0333 Predicate checks on out parameters (0000-00-00)* If a view conversion is passed as an actual parameter corresponding to an out-mode formal parameter, and if the subtype of the formal parameter has a predicate, then no predicate check associated with the conversion is performed. RM references: 3.02.04 (31/5) 4.06 (51/4) 6.04.01 (14) .. index:: AI12-0335 (Ada 2022 feature) * *AI12-0335 Dynamic accessibility check needed for some requeue targets (0000-00-00)* Define a new runtime accessibility check for a corner case involving requeue statements. RM references: 9.05.04 (7/4) .. index:: AI12-0336 (Ada 2022 feature) * *AI12-0336 Meaning of Time_Offset (0000-00-00)* The AI introduces changes to the predefined package ``Ada.Calendar.Time_Zones``. RM references: 9.06.01 (6/2) 9.06.01 (35/2) 9.06.01 (40/2) 9.06.01 (41/2) 9.06.01 (42/3) 9.06.01 (90/2) 9.06.01 (91/2) .. index:: AI12-0337 (Ada 2022 feature) * *AI12-0337 Simple_Name("/") in Ada.Directories (0000-00-00)* Clarify behavior of subprograms in the predefined package ``Ada.Directories``. In particular, Simple_Name ("/") should return "/" on Unix-like systems. RM references: A.16 (47/2) A.16 (74/2) A.16 (82/3) .. index:: AI12-0338 (Ada 2022 feature) * *AI12-0338 Type invariant checking and incomplete types (0000-00-00)* Clarify that type invariants for type T are not checked for incomplete types whose completion is not available, even if that completion has components of type T. RM references: 7.03.02 (20/5) .. index:: AI12-0339 (Ada 2022 feature) * *AI12-0339 Empty function for Container aggregates (2020-08-06)* To provide uniform support for container aggregates, all standard container libraries have been enhanced with a function Empty, to be used when initializing an aggregate prior to inserting the specified elements in the object being constructed. All products have been updated to remove the ambiguities that may have arisen from previous uses of entities named Empty in our sources, and the expansion of container aggregates uses Empty wherever needed. RM references: A.18.02 (8/5) A.18.02 (12.3/5) A.18.02 (78.2/5) A.18.02 (98.6/5) A.18.03 (6/5) A.18.03 (10.2/5) A.18.03 (50.2/5) A.18.05 (3/5) A.18.05 (7.2/5) A.18.05 (37.3/5) A.18.05 (46/2) A.18.06 (4/5) A.18.06 (8.2/5) A.18.06 (51.4/5) A.18.08 (3/5) A.18.08 (8.1/5) A.18.08 (59.2/5) A.18.08 (68/2) A.18.09 (4/5) A.18.09 (9.1/5) A.18.09 (74.2/5) A.18.10 (15.2/5) A.18.18 (8.1/5) A.18.19 (6.1/5) A.18.20 (6/3) A.18.21 (6/3) A.18.22 (6/3) A.18.23 (6/3) A.18.24 (6/3) A.18.25 (8/3) .. index:: AI12-0340 (Ada 2022 feature) * *AI12-0340 Put_Image should use a Text_Buffer (0000-00-00)* Add a new predefined package Ada.Strings.Text_Buffers (along with child units) and change the definition of Put_Image attribute to refer to it. RM references: A.04.12 (0) 4.10 (3.1/5) 4.10 (3.2/5) 4.10 (6/5) 4.10 (25.2/5) 4.10 (28/5) 4.10 (31/5) 4.10 (41/5) H.04 (23.2/5) H.04 (23.11/5) .. index:: AI12-0342 (Ada 2022 feature) * *AI12-0342 Various issues with user-defined literals (part 2) (2020-04-07)* Compiler support is added for three new aspects (``Integer_Literal``, ``Real_Literal``, and ``String_Literal``) as described in AI12-0249 (for ``Integer_Literal`` and ``Real_Literal``), AI12-0295 (for ``String_Literal``), and in two follow-up AIs (AI12-0325 and AI12-0342). For pre-Ada 2022 versions of Ada, these are treated as implementation-defined aspects. Some implementation work remains, particularly in the interactions between these aspects and tagged types. RM references: 4.02.01 (0) 3.09.02 (1/2) 6.03.01 (22) .. index:: AI12-0343 (Ada 2022 feature) * *AI12-0343 Return Statement Checks (2020-04-02)* This binding interpretation has been implemented and the accessibility, predicate, and tag checks prescribed by RM 6.5 are now performed at the appropriate points, as required by this AI. RM references: 6.05 (5.12/5) 6.05 (8/4) 6.05 (8.1/3) 6.05 (21/3) .. index:: AI12-0345 (Ada 2022 feature) * *AI12-0345 Dynamic accessibility of explicitly aliased parameters (2020-09-09)* Further clarify (after AI12-0277) accessibility rules for explicitly aliased parameters. RM references: 3.10.02 (5) 3.10.02 (7/4) 3.10.02 (10.5/3) 3.10.02 (13.4/4) 3.10.02 (19.2/5) 3.10.02 (21) .. index:: AI12-0350 (Ada 2022 feature) * *AI12-0350 Swap for Indefinite_Holders (2025-12-15)* Package ``Ada.Containers.Indefinite_Holders`` is implemented in GNAT, comprising the support for ``Swap`` as specified by this AI. RM references: A.18.18 (22/5) A.18.18 (67/5) A.18.18 (73/3) A.18.32 (13/5) .. index:: AI12-0351 (Ada 2022 feature) * *AI12-0351 Matching for actuals for formal derived types (2020-04-03)* This binding interpretation requires the compiler to checks that an actual subtype in a generic parameter association of an instantiation is statically compatible (even when the actual is unconstrained) with the ancestor of an associated nondiscriminated generic formal derived type. RM references: 12.05.01 (7) 12.05.01 (8) .. index:: AI12-0352 (Ada 2022 feature) * *AI12-0352 Early derivation and equality of untagged types (2020-07-09)* AI12-0352 clarifies that declaring a user-defined primitive equality operation for a record type T is illegal if it occurs after a type has been derived from T. RM references: 4.05.02 (9.8/4) .. index:: AI12-0356 (Ada 2022 feature) * *AI12-0356 Root_Storage_Pool_With_Subpools should have Preelaborable_Initialization (0000-00-00)* Add Preelaborable_Initialization pragmas for predefined types ``Root_Storage_Pool_With_Subpools`` and ``Root_Subpool``. RM references: 13.11.04 (4/3) 13.11.04 (5/3) .. index:: AI12-0363 (Ada 2022 feature) * *AI12-0363 Fixes for Atomic and Volatile (2020-09-08)* This amendment has been implemented under the ``-gnat2022`` switch and the compiler now supports the ``Full_Access_Only`` aspect, which is mostly equivalent to GNAT's ``Volatile_Full_Access``. RM references: 3.10.02 (26/3) 9.10 (1/5) C.06 (6.4/3) C.06 (6.10/3) C.06 (8.1/4) C.06 (12/5) C.06 (12.1/5) C.06 (13.3/5) C.06 (19.1/5) .. index:: AI12-0364 (Ada 2022 feature) * *AI12-0364 Add a modular atomic arithmetic package (0000-00-00)* Generalize support for atomic integer operations to extend to modular types. Add new predefined generic package, ``System.Atomic_Operations.Modular_Arithmetic``. RM references: C.06.05 (0) C.06.04 (1/5) C.06.04 (2/5) C.06.04 (3/5) C.06.04 (9/5) .. index:: AI12-0366 (Ada 2022 feature) * *AI12-0366 Changes to Big_Integer and Big_Real (0000-00-00)* Simplify ``Big_Integer ``and ``Big_Real`` specs by eliminating explicit support for creating "invalid" values. No more ``Optional_Big_[Integer,Real]`` types. RM references: A.05.06 (0) A.05.07 (0) .. index:: AI12-0367 (Ada 2022 feature) * *AI12-0367 Glitches in aspect specifications (0000-00-00)* The AI clarifies a few wording omissions. For example, a specified Small value for a fixed point type has to be positive. RM references: 3.05.09 (8/2) 3.05.10 (2/1) 13.01 (9.1/5) 13.14 (10) .. index:: AI12-0368 (Ada 2022 feature) * *AI12-0368 Declare expressions can be static (2020-05-30)* AI12-0368 allows declare expressions to be static in Ada 2022. RM references: 4.09 (8) 4.09 (12.1/3) 4.09 (17) 6.01.01 (24.2/5) 6.01.01 (24.3/5) 6.01.01 (24.4/5) 6.01.01 (24.5/5) C.04 (9) .. index:: AI12-0369 (Ada 2022 feature) * *AI12-0369 Relaxing barrier restrictions (2020-03-25)* The definitions of the ``Simple_Barriers`` and ``Pure_Barriers`` restrictions were modified by this AI, replacing uses of "statically denotes" with "statically names". This means that in many cases (but not all) a barrier expression that references a subcomponent of a component of the protected type while subject to either of the two restrictions is now allowed; with the previous restriction definitions, such a barrier expression would not have been legal. RM references: D.07 (1.3/5) D.07 (10.12/5) .. index:: AI12-0372 (Ada 2022 feature) * *AI12-0372 Static accessibility of "master of the call" (2020-09-09)* Add an extra compile-time accessibility check for explicitly aliased parameters needed to prevent dangling references. RM references: 3.10.02 (10.5/5) 3.10.02 (19.3/4) 6.04.01 (6.4/3) .. index:: AI12-0373 (Ada 2022 feature) * *AI12-0373 Bunch of fixes (0000-00-00)* Small clarifications to various RM entries with minor impact on compiler implementation. RM references: 3.01 (1) 4.02 (4) 4.02 (8/2) 4.02.01 (3/5) 4.02.01 (4/5) 4.02.01 (5/5) 4.09 (17.3/5) 6.01.01 (41/5) 8.05.04 (4/3) 13.01.01 (4/3) 13.01.01 (11/3) 13.14 (3/5) .. index:: AI12-0376 (Ada 2022 feature) * *AI12-0376 Representation changes finally allowed for untagged derived types (0000-00-00)* A change of representation for a derived type is allowed in some previously-illegal cases where a change of representation is required to implement a call to a derived subprogram. RM references: 13.01 (10/4) .. index:: AI12-0377 (Ada 2022 feature) * *AI12-0377 View conversions and out parameters of types with Default_Value revisited (2020-06-17)* This AI clarifies that an actual of an out parameter that is a view conversion is illegal if either the target or operand type has Default_Value specified while the other does not. RM references: 6.04.01 (5.1/4) 6.04.01 (5.2/4) 6.04.01 (5.3/4) 6.04.01 (13.1/4) 6.04.01 (13.2/4) 6.04.01 (13.3/4) 6.04.01 (13.4/4) 6.04.01 (15/3) .. index:: AI12-0381 (Ada 2022 feature) * *AI12-0381 Tag of a delta aggregate (0000-00-00)* In the case of a delta aggregate of a specific tagged type, the tag of the aggregate comes from the specific type (as opposed to somehow from the base object). RM references: 4.03.04 (14/5) .. index:: AI12-0382 (Ada 2022 feature) * *AI12-0382 Loosen type-invariant overriding requirement of AI12-0042-1 (0000-00-00)* The AI relaxes some corner-case legality rules about type invariants that were added by AI12-0042-1. RM references: 7.3.2(6.1/4) .. index:: AI12-0383 (Ada 2022 feature) * *AI12-0383 Renaming values (2020-06-17)* This AI allow names that denote values rather than objects to nevertheless be renamed using an object renaming. RM references: 8.05.01 (1) 8.05.01 (4) 8.05.01 (4.1/2) 8.05.01 (6/2) 8.05.01 (8) .. index:: AI12-0384-2 (Ada 2022 feature) * *AI12-0384-2 Fixups for Put_Image and Text_Buffers (2021-04-29)* In GNAT's initial implementation of the Ada 2022 ``Put_Image`` aspect and attribute, buffering was performed using a GNAT-defined package, ``Ada.Strings.Text_Output``. Ada 2022 requires a different package, Ada.``Strings.Text_Buffers``, for this role, and that package is now provided, and the older package is eliminated. RM references: 4.10 (0) A.04.12 (0) .. index:: AI12-0385 (Ada 2022 feature) * *AI12-0385 Predefined shifts and rotates should be static (0000-00-00)* This AI allows Shift and Rotate operations in static expressions. GNAT implements this AI partially. RM references: 4.09 (20) .. index:: AI12-0389 (Ada 2022 feature) * *AI12-0389 Ignoring unrecognized aspects (2020-10-08)* Two new restrictions, ``No_Unrecognized_Aspects`` and ``No_Unrecognized_Pragmas``, are available to make the compiler emit error messages on unrecognized pragmas and aspects. RM references: 13.01.01 (38/3) 13.12.01 (6.3/3) .. index:: AI12-0394 (Ada 2022 feature) * *AI12-0394 Named Numbers and User-Defined Numeric Literals (2020-10-05)* Ada 2022 allows using integer named numbers with types that have an ``Integer_Literal`` aspect. Similarly, real named numbers may now be used with types that have a ``Real_Literal`` aspect with an overloading that takes two strings, to be used in particular with ``Ada.Numerics.Big_Numbers.Big_Reals``. RM references: 3.03.02 (3) 4.02.01 (4/5) 4.02.01 (8/5) 4.02.01 (12/5) 4.02.01 (13/5) 4.09 (5) .. index:: AI12-0395 (Ada 2022 feature) * *AI12-0395 Allow aspect_specifications on formal parameters (0000-00-00)* Change syntax rules to allow aspect_specifications on formal parameters, if an implementation if an implementation wants to define one. Currently, GNAT doesn't define any such aspect_specifications. RM references: 6.01 (15/3) .. index:: AI12-0397 (Ada 2022 feature) * *AI12-0397 Default_Initial_Condition applied to derived type (2020-12-09)* The compiler now implements the rules for resolving ``Default_Initial_Condition`` expressions that involve references to the current instance of types with the aspect, as specified by this AI. The type of the current instance is defined to be like a formal derived type, so for a derived type that inherits the aspect, a call passing the current instance to a primitive means that the call will resolve to invoke the corresponding primitive of the descendant type. This also now permits calls to abstract primitives to occur within the aspect expression of an abstract type. RM references: 7.03.03 (3/5) 7.03.03 (6/5) 7.03.03 (8/5) .. index:: AI12-0398 (Ada 2022 feature) * *AI12-0398 Most declarations should have aspect specifications (2020-11-19)* It is now possible to specify aspects for discriminant specifications, extended return object declarations, and entry index specifications. This is an extension added for Ada 2022 by this AI. RM references: 3.07 (5/2) 6.03.01 (25) 6.05 (2.1/3) 9.05.02 (8) .. index:: AI12-0399 (Ada 2022 feature) * *AI12-0399 Aspect specification for Preelaborable_Initialization (0000-00-00)* Semantics-preserving presentation change. Replace ``Preelaborable_Initialization`` pragmas with equivalent aspect specs in the listed predefined packages. GNAT follows the guidance of this AI partially. RM references: 9.05 (53/5) 3.09 (6/5) 7.06 (5/2) 7.06 (7/2) 11.04.01 (2/5) 11.04.01 (3/2) 13.11 (6/2) 13.11.04 (4/5) 13.11.04 (5/5) 13.13.01 (3/2) A.04.02 (4/2) A.04.02 (20/2) A.04.05 (4/2) A.04.07 (4/2) A.04.07 (20/2) A.04.08 (4/2) A.04.08 (20/2) A.12.01 (5/4) A.18.02 (8/5) A.18.02 (9/2) A.18.02 (79.2/5) A.18.02 (79.3/5) A.18.03 (6/5) A.18.03 (7/2) A.18.03 (50.2/5) A.18.03 (50.3/5) A.18.05 (3/5) A.18.05 (4/2) A.18.05 (37.3/5) A.18.05 (37.4/5) A.18.06 (4/5) A.18.06 (5/2) A.18.06 (51.4/5) A.18.06 (51.5/5) A.18.08 (3/5) A.18.08 (4/2) A.18.08 (58.2/5) A.18.08 (58.3/5) A.18.09 (4/5) A.18.09 (5/2) A.18.09 (74.2/5) A.18.09 (74.3/5) A.18.10 (8/5) A.18.10 (9/3) A.18.10 (70.2/5) A.18.10 (70.3/5) A.18.18 (6/5) B.03.01 (5/2) C.07.01 (2/5) G.01.01 (4/2) .. index:: AI12-0400 (Ada 2022 feature) * *AI12-0400 Ambiguities associated with Vector Append and container aggregates (0000-00-00)* Change the names of subprograms in the predefined Vector containers from ``Append`` to ``Append_Vector`` and from ``Prepend`` to ``Prepend_Vector`` in order to resolve some ambiguity problems. GNAT adds the subprograms with new names but also keeps the old ones for backward compatibility. RM references: A.18.02 (8/5) A.18.02 (36/5) A.18.02 (37/5) A.18.02 (38/5) A.18.02 (44/5) A.18.02 (46/5) A.18.02 (47/5) A.18.02 (58/5) A.18.02 (79.2/5) A.18.02 (150/5) A.18.02 (151/5) A.18.02 (152/5) A.18.02 (153/5) A.18.02 (154/5) A.18.02 (155/5) A.18.02 (156/5) A.18.02 (168/5) A.18.02 (169/5) A.18.02 (172/5) A.18.02 (173/5) A.18.02 (174/5) A.18.02 (175.1/5) A.18.03 (23/5) A.18.03 (23.1/5) A.18.03 (58.2/5) A.18.03 (96/5) A.18.03 (97.1/5) .. index:: AI12-0401 (Ada 2022 feature) * *AI12-0401 Renaming of qualified expression of variable (2020-10-31)* Ada 2022 AI12-0401 restricts renaming of a qualified expression to cases where the operand is a constant, or the target subtype statically matches the nominal subtype of the operand, or is unconstrained with no predicates, to prevent setting variables to values outside their range or constraints. RM references: 3.03 (23.2/3) 8.05.01 (4.7/5) 8.05.01 (5/3) .. index:: AI12-0402 (Ada 2022 feature) * *AI12-0402 Master of a function call with elementary result type (2026-01-07)* This AI restricts the rule that makes the master of a function call, which is used to directly initialize part of an object, that of the object being initialized, to the case where the result type of the function is composite. RM references: 3.10.2 (10.2/5) .. index:: AI12-0409 (Ada 2022 feature) * *AI12-0409 Preelaborable_Initialization and bounded containers (2021-06-23)* As defined by this AI, the ``Preelaborable_Initializatio`` aspect now has a corresponding attribute of the same name. Types declared within a generic package specification are permitted to specify the expression of a ``Prelaborable_Initialization`` aspect by including one or more references to the attribute applied to a formal private or formal derived type conjoined by ``and`` operators. This permits the full type of a private type with such an aspect expression to have components of the named formal types, and such a type will have preelaborable initialization in an instance when the actual types for all referenced formal types have preelaborable initialization. RM references: 10.02.01 (4.1/2) 10.02.01 (4.2/2) 10.02.01 (11.1/2) 10.02.01 (11.2/2) 10.02.01 (11.6/2) 10.02.01 (11.7/2) 10.02.01 (11.8/2) 13.01 (11/3) A.18.19 (5/5) A.18.20 (5/5) A.18.21 (5/5) A.18.22 (5/5) A.18.23 (5/5) A.18.24 (5/5) A.18.25 (5/5) A.18.32 (6/5) J.15.14 (0) .. index:: AI12-0411 (Ada 2022 feature) * *AI12-0411 Add "bool" to Interfaces.C (0000-00-00)* RM references: B.03 (13) B.03 (43/2) B.03 (65.1/4) .. index:: AI12-0412 (Ada 2022 feature) * *AI12-0412 Abstract Pre/Post'Class on primitive of abstract type (2021-05-19)* In Ada 2022, by AI12-0412, it's legal to specify Pre'Class and Post'Class aspects on nonabstract primitive subprograms of an abstract type, but if the expression of such an aspect is nonstatic, then it's illegal to make a nondispatching call to such a primitive, to apply ``'Access`` to it, or to pass such a primitive as an actual subprogram for a concrete formal subprogram in a generic instantiation. RM references: 6.01.01 (18.2/4) .. index:: AI12-0413 (Ada 2022 feature) * *AI12-0413 Reemergence of "=" when defined to be abstract (0000-00-00)* The AI clarifies rules about operator reemergence in instances, and nondispatching calls to abstract subprograms. RM references: 3.09.03 (7) 4.05.02 (14.1/3) 4.05.02 (24.1/3) 12.05 (8/3) .. index:: AI12-0423 (Ada 2022 feature) * *AI12-0423 Aspect inheritance fixups (0000-00-00)* Clarify that the No_Return aspect behaves as one would expect for an inherited subprogram and that inheritance works as one would expect for a multi-part aspect whose value is specified via an aggregate (e.g., the Aggregate aspect). RM references: 6.05.01 (3.3/3) 13.01 (15.7/5) 13.01 (15.8/5) .. index:: AI12-0432 (Ada 2022 feature) * *AI12-0432 View conversions of assignments and predicate checks (2021-05-05)* When a predicate applies to a tagged type, a view conversion to that type normally requires a predicate check. However, as specified by AI12-0432, when the view conversion appears as the target of an assignment, a predicate check is not applied to the object in the conversion. RM references: 3.02.04 (31/5) 4.06 (51.1/5) --- ### Gcc Ada Doc Gnat Rm Implementation Of Specific Ada Features (`gcc/ada/doc/gnat_rm/implementation_of_specific_ada_features.rst`) .. _Implementation_of_Specific_Ada_Features: *************************************** Implementation of Specific Ada Features *************************************** This chapter describes the GNAT implementation of several Ada language facilities. .. _Machine_Code_Insertions: Machine Code Insertions ======================= .. index:: Machine Code insertions Package ``Machine_Code`` provides machine code support as described in the Ada Reference Manual in two separate forms: * Machine code statements, consisting of qualified expressions that fit the requirements of RM section 13.8. * An intrinsic callable procedure, providing an alternative mechanism of including machine instructions in a subprogram. The two features are similar, and both are closely related to the mechanism provided by the asm instruction in the GNU C compiler. Full understanding and use of the facilities in this package requires understanding the asm instruction, see the section on Extended Asm in :title:`Using_the_GNU_Compiler_Collection_(GCC)`. Calls to the function ``Asm`` and the procedure ``Asm`` have identical semantic restrictions and effects as described below. Both are provided so that the procedure call can be used as a statement, and the function call can be used to form a code_statement. Consider this C ``asm`` instruction: :: asm ("fsinx %1 %0" : "=f" (result) : "f" (angle)); The equivalent can be written for GNAT as: .. code-block:: ada Asm ("fsinx %1 %0", My_Float'Asm_Output ("=f", result), My_Float'Asm_Input ("f", angle)); The first argument to ``Asm`` is the assembler template, and is identical to what is used in GNU C. This string must be a static expression. The second argument is the output operand list. It is either a single ``Asm_Output`` attribute reference, or a list of such references enclosed in parentheses (technically an array aggregate of such references). The ``Asm_Output`` attribute denotes a function that takes two parameters. The first is a string, the second is the name of a variable of the type designated by the attribute prefix. The first (string) argument is required to be a static expression and designates the constraint (see the section on Constraints in :title:`Using_the_GNU_Compiler_Collection_(GCC)`) for the parameter; e.g., what kind of register is required. The second argument is the variable to be written or updated with the result. The possible values for constraint are the same as those used in the RTL, and are dependent on the configuration file used to build the GCC back end. If there are no output operands, then this argument may either be omitted, or explicitly given as ``No_Output_Operands``. No support is provided for GNU C's symbolic names for output parameters. The second argument of ``my_float'Asm_Output`` functions as though it were an ``out`` parameter, which is a little curious, but all names have the form of expressions, so there is no syntactic irregularity, even though normally functions would not be permitted ``out`` parameters. The third argument is the list of input operands. It is either a single ``Asm_Input`` attribute reference, or a list of such references enclosed in parentheses (technically an array aggregate of such references). The ``Asm_Input`` attribute denotes a function that takes two parameters. The first is a string, the second is an expression of the type designated by the prefix. The first (string) argument is required to be a static expression, and is the constraint for the parameter, (e.g., what kind of register is required). The second argument is the value to be used as the input argument. The possible values for the constraint are the same as those used in the RTL, and are dependent on the configuration file used to built the GCC back end. No support is provided for GNU C's symbolic names for input parameters. If there are no input operands, this argument may either be omitted, or explicitly given as ``No_Input_Operands``. The fourth argument, not present in the above example, is a list of register names, called the *clobber* argument. This argument, if given, must be a static string expression, and is a space or comma separated list of names of registers that must be considered destroyed as a result of the ``Asm`` call. If this argument is the null string (the default value), then the code generator assumes that no additional registers are destroyed. In addition to registers, the special clobbers ``memory`` and ``cc`` as described in the GNU C docs are both supported. The fifth argument, not present in the above example, called the *volatile* argument, is by default ``False``. It can be set to the literal value ``True`` to indicate to the code generator that all optimizations with respect to the instruction specified should be suppressed, and in particular an instruction that has outputs will still be generated, even if none of the outputs are used. See :title:`Using_the_GNU_Compiler_Collection_(GCC)` for the full description. Generally it is strongly advisable to use Volatile for any ASM statement that is missing either input or output operands or to avoid unwanted optimizations. A warning is generated if this advice is not followed. No support is provided for GNU C's ``asm goto`` feature. The ``Asm`` subprograms may be used in two ways. First the procedure forms can be used anywhere a procedure call would be valid, and correspond to what the RM calls 'intrinsic' routines. Such calls can be used to intersperse machine instructions with other Ada statements. Second, the function forms, which return a dummy value of the limited private type ``Asm_Insn``, can be used in code statements, and indeed this is the only context where such calls are allowed. Code statements appear as aggregates of the form: .. code-block:: ada Asm_Insn'(Asm (...)); Asm_Insn'(Asm_Volatile (...)); In accordance with RM rules, such code statements are allowed only within subprograms whose entire body consists of such statements. It is not permissible to intermix such statements with other Ada statements. Typically the form using intrinsic procedure calls is more convenient and more flexible. The code statement form is provided to meet the RM suggestion that such a facility should be made available. The following is the exact syntax of the call to ``Asm``. As usual, if named notation is used, the arguments may be given in arbitrary order, following the normal rules for use of positional and named arguments: :: ASM_CALL ::= Asm ( [Template =>] static_string_EXPRESSION [,[Outputs =>] OUTPUT_OPERAND_LIST ] [,[Inputs =>] INPUT_OPERAND_LIST ] [,[Clobber =>] static_string_EXPRESSION ] [,[Volatile =>] static_boolean_EXPRESSION] ) OUTPUT_OPERAND_LIST ::= [PREFIX.]No_Output_Operands | OUTPUT_OPERAND_ATTRIBUTE | (OUTPUT_OPERAND_ATTRIBUTE {,OUTPUT_OPERAND_ATTRIBUTE}) OUTPUT_OPERAND_ATTRIBUTE ::= SUBTYPE_MARK'Asm_Output (static_string_EXPRESSION, NAME) INPUT_OPERAND_LIST ::= [PREFIX.]No_Input_Operands | INPUT_OPERAND_ATTRIBUTE | (INPUT_OPERAND_ATTRIBUTE {,INPUT_OPERAND_ATTRIBUTE}) INPUT_OPERAND_ATTRIBUTE ::= SUBTYPE_MARK'Asm_Input (static_string_EXPRESSION, EXPRESSION) The identifiers ``No_Input_Operands`` and ``No_Output_Operands`` are declared in the package ``Machine_Code`` and must be referenced according to normal visibility rules. In particular if there is no ``use`` clause for this package, then appropriate package name qualification is required. .. _GNAT_Implementation_of_Tasking: GNAT Implementation of Tasking ============================== This chapter outlines the basic GNAT approach to tasking (in particular, a multi-layered library for portability) and discusses issues related to compliance with the Real-Time Systems Annex. .. _Mapping_Ada_Tasks_onto_the_Underlying_Kernel_Threads: Mapping Ada Tasks onto the Underlying Kernel Threads ---------------------------------------------------- GNAT's run-time support comprises two layers: * GNARL (GNAT Run-time Layer) * GNULL (GNAT Low-level Library) In GNAT, Ada's tasking services rely on a platform and OS independent layer known as GNARL. This code is responsible for implementing the correct semantics of Ada's task creation, rendezvous, protected operations etc. GNARL decomposes Ada's tasking semantics into simpler lower level operations such as create a thread, set the priority of a thread, yield, create a lock, lock/unlock, etc. The spec for these low-level operations constitutes GNULLI, the GNULL Interface. This interface is directly inspired from the POSIX real-time API. If the underlying executive or OS implements the POSIX standard faithfully, the GNULL Interface maps as is to the services offered by the underlying kernel. Otherwise, some target dependent glue code maps the services offered by the underlying kernel to the semantics expected by GNARL. Whatever the underlying OS (VxWorks, UNIX, Windows, etc.) the key point is that each Ada task is mapped on a thread in the underlying kernel. For example, in the case of VxWorks, one Ada task = one VxWorks task. In addition Ada task priorities map onto the underlying thread priorities. Mapping Ada tasks onto the underlying kernel threads has several advantages: * The underlying scheduler is used to schedule the Ada tasks. This makes Ada tasks as efficient as kernel threads from a scheduling standpoint. * Interaction with code written in C containing threads is eased since at the lowest level Ada tasks and C threads map onto the same underlying kernel concept. * When an Ada task is blocked during I/O the remaining Ada tasks are able to proceed. * On multiprocessor systems Ada tasks can execute in parallel. Some threads libraries offer a mechanism to fork a new process, with the child process duplicating the threads from the parent. GNAT does not support this functionality when the parent contains more than one task. .. index:: Forking a new process .. _Ensuring_Compliance_with_the_Real-Time_Annex: Ensuring Compliance with the Real-Time Annex -------------------------------------------- .. index:: Real-Time Systems Annex compliance Although mapping Ada tasks onto the underlying threads has significant advantages, it does create some complications when it comes to respecting the scheduling semantics specified in the real-time annex (Annex D). For instance the Annex D requirement for the ``FIFO_Within_Priorities`` scheduling policy states: *When the active priority of a ready task that is not running changes, or the setting of its base priority takes effect, the task is removed from the ready queue for its old active priority and is added at the tail of the ready queue for its new active priority, except in the case where the active priority is lowered due to the loss of inherited priority, in which case the task is added at the head of the ready queue for its new active priority.* While most kernels do put tasks at the end of the priority queue when a task changes its priority, (which respects the main FIFO_Within_Priorities requirement), almost none keep a thread at the beginning of its priority queue when its priority drops from the loss of inherited priority. As a result most vendors have provided incomplete Annex D implementations. The GNAT run-time, has a nice cooperative solution to this problem which ensures that accurate FIFO_Within_Priorities semantics are respected. The principle is as follows. When an Ada task T is about to start running, it checks whether some other Ada task R with the same priority as T has been suspended due to the loss of priority inheritance. If this is the case, T yields and is placed at the end of its priority queue. When R arrives at the front of the queue it executes. Note that this simple scheme preserves the relative order of the tasks that were ready to execute in the priority queue where R has been placed at the end. .. Support_for_Locking_Policies Support for Locking Policies ---------------------------- This section specifies which policies specified by pragma Locking_Policy are supported on which platforms. GNAT supports the standard ``Ceiling_Locking`` policy, and the implementation defined ``Inheritance_Locking`` and ``Concurrent_Readers_Locking`` policies. ``Ceiling_Locking`` is supported on all platforms if the operating system supports it. In particular, ``Ceiling_Locking`` is not supported on VxWorks. ``Inheritance_Locking`` is supported on Linux, Darwin (Mac OS X), LynxOS 178, and VxWorks. ``Concurrent_Readers_Locking`` is supported on Linux. Notes about ``Ceiling_Locking`` on Linux: If the process is running as 'root', ceiling locking is used. If the capabilities facility is installed ("sudo apt-get --assume-yes install libcap-dev" on Ubuntu, for example), and the program is linked against that library ("-largs -lcap"), and the executable file has the cap_sys_nice capability ("sudo /sbin/setcap cap_sys_nice=ep executable_file_name"), then ceiling locking is used. Otherwise, the ``Ceiling_Locking`` policy is ignored. .. _GNAT_Implementation_of_Shared_Passive_Packages: GNAT Implementation of Shared Passive Packages ============================================== .. index:: Shared passive packages GNAT fully implements the :index:`pragma ` ``Shared_Passive`` for the purpose of designating shared passive packages. This allows the use of passive partitions in the context described in the Ada Reference Manual; i.e., for communication between separate partitions of a distributed application using the features in Annex E. .. index:: Annex E .. index:: Distribution Systems Annex However, the implementation approach used by GNAT provides for more extensive usage as follows: *Communication between separate programs* This allows separate programs to access the data in passive partitions, using protected objects for synchronization where needed. The only requirement is that the two programs have a common shared file system. It is even possible for programs running on different machines with different architectures (e.g., different endianness) to communicate via the data in a passive partition. *Persistence between program runs* The data in a passive package can persist from one run of a program to another, so that a later program sees the final values stored by a previous run of the same program. The implementation approach used is to store the data in files. A separate stream file is created for each object in the package, and an access to an object causes the corresponding file to be read or written. .. index:: SHARED_MEMORY_DIRECTORY environment variable The environment variable ``SHARED_MEMORY_DIRECTORY`` should be set to the directory to be used for these files. The files in this directory have names that correspond to their fully qualified names. For example, if we have the package .. code-block:: ada package X is pragma Shared_Passive (X); Y : Integer; Z : Float; end X; and the environment variable is set to ``/stemp/``, then the files created will have the names: :: /stemp/x.y /stemp/x.z These files are created when a value is initially written to the object, and the files are retained until manually deleted. This provides the persistence semantics. If no file exists, it means that no partition has assigned a value to the variable; in this case the initial value declared in the package will be used. This model ensures that there are no issues in synchronizing the elaboration process, since elaboration of passive packages elaborates the initial values, but does not create the files. The files are written using normal ``Stream_IO`` access. If you want to be able to communicate between programs or partitions running on different architectures, then you should use the XDR versions of the stream attribute routines, since these are architecture independent. If active synchronization is required for access to the variables in the shared passive package, then as described in the Ada Reference Manual, the package may contain protected objects used for this purpose. In this case a lock file (whose name is :file:`___lock`, with three underscores) is created in the shared memory directory. .. index:: ___lock file (for shared passive packages) This is used to provide the required locking semantics for proper protected object synchronization. .. _Code_Generation_for_Array_Aggregates: Code Generation for Array Aggregates ==================================== Aggregates have a rich syntax and allow the user to specify the values of complex data structures by means of a single construct. As a result, the code generated for aggregates can be quite complex and involve loops, case statements and multiple assignments. In the simplest cases, however, the compiler will recognize aggregates whose components and constraints are fully static, and in those cases the compiler will generate little or no executable code. The following is an outline of the code that GNAT generates for various aggregate constructs. For further details, you will find it useful to examine the output produced by the -gnatG flag to see the expanded source that is input to the code generator. You may also want to examine the assembly code generated at various levels of optimization. The code generated for aggregates depends on the context, the component values, and the type. In the context of an object declaration the code generated is generally simpler than in the case of an assignment. As a general rule, static component values and static subtypes also lead to simpler code. .. _Static_constant_aggregates_with_static_bounds: Static constant aggregates with static bounds --------------------------------------------- For the declarations: .. code-block:: ada type One_Dim is array (1..10) of integer; ar0 : constant One_Dim := (1, 2, 3, 4, 5, 6, 7, 8, 9, 0); GNAT generates no executable code: the constant ar0 is placed in static memory. The same is true for constant aggregates with named associations: .. code-block:: ada Cr1 : constant One_Dim := (4 => 16, 2 => 4, 3 => 9, 1 => 1, 5 .. 10 => 0); Cr3 : constant One_Dim := (others => 7777); The same is true for multidimensional constant arrays such as: .. code-block:: ada type two_dim is array (1..3, 1..3) of integer; Unit : constant two_dim := ( (1,0,0), (0,1,0), (0,0,1)); The same is true for arrays of one-dimensional arrays: the following are static: .. code-block:: ada type ar1b is array (1..3) of boolean; type ar_ar is array (1..3) of ar1b; None : constant ar1b := (others => false); -- fully static None2 : constant ar_ar := (1..3 => None); -- fully static However, for multidimensional aggregates with named associations, GNAT will generate assignments and loops, even if all associations are static. The following two declarations generate a loop for the first dimension, and individual component assignments for the second dimension: .. code-block:: ada Zero1: constant two_dim := (1..3 => (1..3 => 0)); Zero2: constant two_dim := (others => (others => 0)); .. _Constant_aggregates_with_unconstrained_nominal_types: Constant aggregates with unconstrained nominal types ---------------------------------------------------- In such cases the aggregate itself establishes the subtype, so that associations with ``others`` cannot be used. GNAT determines the bounds for the actual subtype of the aggregate, and allocates the aggregate statically as well. No code is generated for the following: .. code-block:: ada type One_Unc is array (natural range <>) of integer; Cr_Unc : constant One_Unc := (12,24,36); .. _Aggregates_with_static_bounds: Aggregates with static bounds ----------------------------- In all previous examples the aggregate was the initial (and immutable) value of a constant. If the aggregate initializes a variable, then code is generated for it as a combination of individual assignments and loops over the target object. The declarations .. code-block:: ada Cr_Var1 : One_Dim := (2, 5, 7, 11, 0, 0, 0, 0, 0, 0); Cr_Var2 : One_Dim := (others > -1); generate the equivalent of .. code-block:: ada Cr_Var1 (1) := 2; Cr_Var1 (2) := 3; Cr_Var1 (3) := 5; Cr_Var1 (4) := 11; for I in Cr_Var2'range loop Cr_Var2 (I) := -1; end loop; .. _Aggregates_with_nonstatic_bounds: Aggregates with nonstatic bounds --------------------------------- If the bounds of the aggregate are not statically compatible with the bounds of the nominal subtype of the target, then constraint checks have to be generated on the bounds. For a multidimensional array, constraint checks may have to be applied to sub-arrays individually, if they do not have statically compatible subtypes. .. _Aggregates_in_assignment_statements: Aggregates in assignment statements ----------------------------------- In general, aggregate assignment requires the construction of a temporary, and a copy from the temporary to the target of the assignment. This is because it is not always possible to convert the assignment into a series of individual component assignments. For example, consider the simple case: .. code-block:: ada A := (A(2), A(1)); This cannot be converted into: .. code-block:: ada A(1) := A(2); A(2) := A(1); So the aggregate has to be built first in a separate location, and then copied into the target. GNAT recognizes simple cases where this intermediate step is not required, and the assignments can be performed in place, directly into the target. The following sufficient criteria are applied: * The bounds of the aggregate are static, and the associations are static. * The components of the aggregate are static constants, names of simple variables that are not renamings, or expressions not involving indexed components whose operands obey these rules. If any of these conditions are violated, the aggregate will be built in a temporary (created either by the front-end or the code generator) and then that temporary will be copied onto the target. .. _The_Size_of_Discriminated_Records_with_Default_Discriminants: The Size of Discriminated Records with Default Discriminants ============================================================ If a discriminated type ``T`` has discriminants with default values, it is possible to declare an object of this type without providing an explicit constraint: .. code-block:: ada type Size is range 1..100; type Rec (D : Size := 15) is record Name : String (1..D); end T; Word : Rec; Such an object is said to be *unconstrained*. The discriminant of the object can be modified by a full assignment to the object, as long as it preserves the relation between the value of the discriminant, and the value of the components that depend on it: .. code-block:: ada Word := (3, "yes"); Word := (5, "maybe"); Word := (5, "no"); -- raises Constraint_Error In order to support this behavior efficiently, an unconstrained object is given the maximum size that any value of the type requires. In the case above, ``Word`` has storage for the discriminant and for a ``String`` of length 100. It is important to note that unconstrained objects do not require dynamic allocation. It would be an improper implementation to place on the heap those components whose size depends on discriminants. (This improper implementation was used by some Ada83 compilers, where the ``Name`` component above would have been stored as a pointer to a dynamic string). Following the principle that dynamic storage management should never be introduced implicitly, an Ada compiler should reserve the full size for an unconstrained declared object, and place it on the stack. This maximum size approach has been a source of surprise to some users, who expect the default values of the discriminants to determine the size reserved for an unconstrained object: "If the default is 15, why should the object occupy a larger size?" The answer, of course, is that the discriminant may be later modified, and its full range of values must be taken into account. This is why the declaration: .. code-block:: ada type Rec (D : Positive := 15) is record Name : String (1..D); end record; Too_Large : Rec; is flagged by the compiler with a warning: an attempt to create ``Too_Large`` will raise ``Storage_Error``, because the required size includes ``Positive'Last`` bytes. As the first example indicates, the proper approach is to declare an index type of 'reasonable' range so that unconstrained objects are not too large. One final wrinkle: if the object is declared to be ``aliased``, or if it is created in the heap by means of an allocator, then it is *not* unconstrained: it is constrained by the default values of the discriminants, and those values cannot be modified by full assignment. This is because in the presence of aliasing all views of the object (which may be manipulated by different tasks, say) must be consistent, so it is imperative that the object, once created, remain invariant. .. _Image_Values_For_Nonscalar_Types: Image Values For Nonscalar Types ================================ Ada 2022 defines the Image, Wide_Image, and Wide_Wide image attributes for nonscalar types; earlier Ada versions defined these attributes only for scalar types. Ada RM 4.10 provides some general guidance regarding the default implementation of these attributes and the GNAT compiler follows that guidance. However, beyond that the precise details of the image text generated in these cases are deliberately not documented and are subject to change. In particular, users should not rely on formatting details (such as spaces or line breaking), record field order, image values for access types, image values for types that have ancestor or subcomponent types declared in non-Ada2022 code, image values for predefined types, or the compiler's choices regarding the implementation permissions described in Ada RM 4.10. This list is not intended to be exhaustive. If more precise control of image text is required for some type T, then T'Put_Image should be explicitly specified. .. _Strict_Conformance_to_the_Ada_Reference_Manual: Strict Conformance to the Ada Reference Manual ============================================== The dynamic semantics defined by the Ada Reference Manual impose a set of run-time checks to be generated. By default, the GNAT compiler will insert many run-time checks into the compiled code, including most of those required by the Ada Reference Manual. However, there are two checks that are not enabled in the default mode for efficiency reasons: checks for access before elaboration on subprogram calls, and stack overflow checking (most operating systems do not perform this check by default). Strict conformance to the Ada Reference Manual can be achieved by adding two compiler options for dynamic checks for access-before-elaboration on subprogram calls and generic instantiations (*-gnatE*), and stack overflow checking (*-fstack-check*). Note that the result of a floating point arithmetic operation in overflow and invalid situations, when the ``Machine_Overflows`` attribute of the result type is ``False``, is to generate IEEE NaN and infinite values. This is the case for machines compliant with the IEEE floating-point standard, but on machines that are not fully compliant with this standard, such as Alpha, the *-mieee* compiler flag must be used for achieving IEEE confirming behavior (although at the cost of a significant performance penalty), so infinite and NaN values are properly generated. --- ### Gcc Ada Doc Gnat Rm Interfacing To Other Languages (`gcc/ada/doc/gnat_rm/interfacing_to_other_languages.rst`) .. _Interfacing_to_Other_Languages: ****************************** Interfacing to Other Languages ****************************** The facilities in Annex B of the Ada Reference Manual are fully implemented in GNAT, and in addition, a full interface to C++ is provided. .. _Interfacing_to_C: Interfacing to C ================ Interfacing to C with GNAT can use one of two approaches: * The types in the package ``Interfaces.C`` may be used. * Standard Ada types may be used directly. This may be less portable to other compilers, but will work on all GNAT compilers, which guarantee correspondence between the C and Ada types. Pragma ``Convention C`` may be applied to Ada types, but mostly has no effect, since this is the default. The following table shows the correspondence between Ada scalar types and the corresponding C types. ======================== ================================================================== Ada Type C Type ======================== ================================================================== ``Integer`` ``int`` ``Short_Integer`` ``short`` ``Short_Short_Integer`` ``signed char`` ``Long_Integer`` ``long`` ``Long_Long_Integer`` ``long long`` ``Short_Float`` ``float`` ``Float`` ``float`` ``Long_Float`` ``double`` ``Long_Long_Float`` This is the longest floating-point type supported by the hardware. ======================== ================================================================== Additionally, there are the following general correspondences between Ada and C types: * Ada enumeration types map to C enumeration types directly if pragma ``Convention C`` is specified, which causes them to have a length of 32 bits, except for boolean types which map to C99 ``bool`` and for which the length is 8 bits. Without pragma ``Convention C``, Ada enumeration types map to 8, 16, or 32 bits (i.e., C types ``signed char``, ``short``, ``int``, respectively) depending on the number of values passed. This is the only case in which pragma ``Convention C`` affects the representation of an Ada type. * Ada access types map to C pointers, except for the case of pointers to unconstrained types in Ada, which have no direct C equivalent. * Ada arrays map directly to C arrays. * Ada records map directly to C structures. * Packed Ada records map to C structures where all members are bit fields of the length corresponding to the ``type'Size`` value in Ada. .. _Interfacing_to_C++: Interfacing to C++ ================== The interface to C++ makes use of the following pragmas, which are primarily intended to be constructed automatically using a binding generator tool, although it is possible to construct them by hand. Using these pragmas it is possible to achieve complete inter-operability between Ada tagged types and C++ class definitions. See :ref:`Implementation_Defined_Pragmas`, for more details. :samp:`pragma CPP_Class ([Entity =>] {LOCAL_NAME})` The argument denotes an entity in the current declarative region that is declared as a tagged or untagged record type. It indicates that the type corresponds to an externally declared C++ class type, and is to be laid out the same way that C++ would lay out the type. Note: Pragma ``CPP_Class`` is currently obsolete. It is supported for backward compatibility but its functionality is available using pragma ``Import`` with ``Convention`` = ``CPP``. :samp:`pragma CPP_Constructor ([Entity =>] {LOCAL_NAME})` This pragma identifies an imported function (imported in the usual way with pragma ``Import``) as corresponding to a C++ constructor. A few restrictions are placed on the use of the ``Access`` attribute in conjunction with subprograms subject to convention ``CPP``: the attribute may be used neither on primitive operations of a tagged record type with convention ``CPP``, imported or not, nor on subprograms imported with pragma ``CPP_Constructor``. In addition, C++ exceptions are propagated and can be handled in an ``others`` choice of an exception handler. The corresponding Ada occurrence has no message, and the simple name of the exception identity contains ``Foreign_Exception``. Finalization and awaiting dependent tasks works properly when such foreign exceptions are propagated. It is also possible to import a C++ exception using the following syntax: :: LOCAL_NAME : exception; pragma Import (Cpp, [Entity =>] LOCAL_NAME, [External_Name =>] static_string_EXPRESSION); The ``External_Name`` is the name of the C++ RTTI symbol. You can then cover a specific C++ exception in an exception handler. If the string ends with "'Class", as if referencing the Class attribute of the C++ type, that enables "class-wide" type matching, i.e., instances of C++ classes derived from the one denoted by the RTTI symbol, that would be caught by C++ handlers for that type, will also be caught by Ada handlers for ``Entity``. For non-class-wide RTTI symbols imported from C++, only exact type matches will be handled. C++ rethrown (dependent) exceptions are not distinguishable from the corresponding primary exceptions: they are handled exactly as if the primary exception had been raised. With imported exceptions, especially with base-type matching, a single handled_sequence_of_statements may have exception handlers with choices that cover the same C++ types in ways that GNAT cannot detect. For example, C++ classes ``base`` and ``derived`` may be imported as exceptions with base-type matching, but GNAT does not know that they are related by inheritance, only the runtime will know it. Given: :: exception when Derived_Exception => null; when Base_Exception => null; when others => null; the earliest handler that matches the type of the raised object will be selected. If an instance of ``derived`` or a further derived type is raised, the first handler will be used; if an instance of ``base`` that is not an instance of ``derived`` is raised, the second handler will be used; raised objects that are not instances of ``base`` will be handled by the ``others`` handler. However, if the handlers were reordered (``others`` must remain last), the ``Derived_Exception`` handler would never be used, because ``Base_Exception`` would match any instances of ``derived`` before ``Derived_Exception`` or ``others`` handlers were considered. Mixing exact-type and base-type matching exceptions may also involve overlapping handlers that GNAT will not reject: an exact-type ``Base_Only_Exception`` handler placed before ``Base_Exception`` will handle instances of ``base``, whereas instances of derived types will be handled by ``Base_Exception``. Swapping them will cause ``Base_Exception`` to handle all instances of ``base`` and derived types, so that a subsequent handler for ``Base_Only_Exception`` will never be selected. The C++ object associated with a C++ ``Exception_Occurrence`` may be obtained by calling the ``GNAT.CPP_Exceptions.Get_Object_Address`` function. There are convenience generic wrappers named ``Get_Object``, ``Get_Access_To_Object``, and ``Get_Access_To_Tagged_Object``, parameterized on the expected Ada type. Note that, for exceptions imported from C++, the address of the object is that of the subobject of the type associated with the exception, which may have a different address from that of the full object; for C++ exceptions handled by ``others`` handlers, however, the address of the full object is returned. E.g., if the imported exception uses the RTTI symbol for the base class, followed by "'Class", and the C++ code raises (throws) an instance of a derived class, a handler for that imported exception will catch this ``Exception_Occurrence``, and ``Get_Object_Address`` will return the address of the base subobject of the raised derived object; ``Get_Object``, ``Get_Access_To_Object`` and ``Get_Access_To_Tagged_Object`` only convert that address to the parameterized type, so the specified type ought to be a type that imports the C++ type whose RTTI symbol was named in the declared exception, i.e., base, not derived or any other type. GNAT cannot detect or report if a type is named that does not match the handler's RTTI-specified type. For ``others`` handlers, and for exact type matches, the full object is obtained. The ``Get_Type_Info`` function that takes an ``Exception_Occurrence`` argument can be used to verify the type of the C++ object raised as an exception. The other ``Get_Type_Info`` function, that takes an ``Exception_Id``, obtains the type expected by the handler, and no such type exists for ``others`` handlers. ``GNAT.CPP.Std.Name`` can then convert the opaque ``GNAT.CPP.Std.Type_Info_Ptr`` access to ``std::type_info`` objects, returned by either ``Get_Type_Info`` function, to a C++ mangled type name. If an ``Exception_Occurrence`` was raised from C++, or following C++ conventions, ``GNAT.Exception_Actions.Exception_Language`` will return ``EL_Cpp``, whether the exception handler is an imported C++ exception or ``others``. ``GNAT.Exception_Actions.Is_Foreign_Exception`` returns True for all of these, as well as for any case in which ``Exception_Language`` is not ``EL_Ada``. :: -- Given the following partial package specification: Base_Exception : exception; pragma Import (Cpp, Base_Exception, "_ZTI4base'Class"); -- Handle instances of base, and of subclasses. type Base is limited tagged record [...] end record; pragma Import (Cpp, Base); type Derived is limited tagged record [...] end record; pragma Import (Cpp, Derived); type Unrelated is access procedure (B : Boolean); function Get_Base_Obj_Acc is new Get_Access_To_Tagged_Object (Base); function Get_Derived_Obj_Acc is new Get_Access_To_Tagged_Object (Derived); function Get_Unrelated_Obj_Acc is new Get_Access_To_Object (Unrelated); procedure Raise_Derived; -- Raises an instance of derived (with a base subobject). -- The comments next to each statement indicate the behavior of -- the following pseudocode blocks: begin Raise_Derived; exception when BEx : Base_Exception => ?? := Is_Foreign_Exception (BEx); -- True ?? := Exception_Language (BEx); -- EL_Cpp ?? := Name (Get_Type_Info (BEx)); -- "7derived" ?? := Name (Get_Type_Info (Exception_Identity (BEx))); -- "4base" ?? := Get_Object_Address (BEx); -- base subobject in derived object ?? := Get_Base_Obj_Acc (BEx): -- ditto, as access to Base ?? := Get_Derived_Obj_Acc (BEx): -- ditto, NO ERROR DETECTED! ?? := Get_Unrelated_Obj_Acc (BEx): -- ditto, NO ERROR DETECTED! end; begin Raise_Derived; exception when BEx : others => ?? := Is_Foreign_Exception (BEx); -- True ?? := Exception_Language (BEx); -- EL_Cpp ?? := Name (Get_Type_Info (BEx)); -- "7derived" ?? := Get_Type_Info (Exception_Identity (BEx)); -- null ?? := Get_Object_Address (BEx); -- full derived object ?? := Get_Derived_Obj_Acc (BEx): -- ditto, as access to Derived ?? := Get_Base_Obj_Acc (BEx): -- ditto, NO ERROR DETECTED! ?? := Get_Unrelated_Obj_Acc (BEx): -- ditto, NO ERROR DETECTED! end; The calls marked with ``NO ERROR DETECTED!`` will compile successfully, even though the types specified in the specializations of the generic function do not match the type of the exception object that the function is expected to return. Mismatches between derived and base types are particularly relevant because they will appear to work as long as there isn't any offset between pointers to these types. This may hold in many cases, but is subject to change with various possible changes to the derived class. The ``GNAT.CPP.Std`` package offers interfaces corresponding to the C++ standard type ``std::type_info``. Function ``To_Type_Info_Ptr`` builds an opaque ``Type_Info_Ptr`` to reference a ``std::type_info`` object at a given ``System.Address``. .. _Interfacing_to_COBOL: Interfacing to COBOL ==================== Interfacing to COBOL is achieved as described in section B.4 of the Ada Reference Manual. .. _Interfacing_to_Fortran: Interfacing to Fortran ====================== Interfacing to Fortran is achieved as described in section B.5 of the Ada Reference Manual. The pragma ``Convention Fortran``, applied to a multi-dimensional array causes the array to be stored in column-major order as required for convenient interface to Fortran. .. _Interfacing_to_non-GNAT_Ada_code: Interfacing to non-GNAT Ada code ================================ It is possible to specify the convention ``Ada`` in a pragma ``Import`` or pragma ``Export``. However this refers to the calling conventions used by GNAT, which may or may not be similar enough to those used by some other Ada 83 / Ada 95 / Ada 2005 compiler to allow interoperation. If arguments types are kept simple, and if the foreign compiler generally follows system calling conventions, then it may be possible to integrate files compiled by other Ada compilers, provided that the elaboration issues are adequately addressed (for example by eliminating the need for any load time elaboration). In particular, GNAT running on VMS is designed to be highly compatible with the DEC Ada 83 compiler, so this is one case in which it is possible to import foreign units of this type, provided that the data items passed are restricted to simple scalar values or simple record types without variants, or simple array types with fixed bounds. --- ### Gcc Ada Doc Gnat Rm Intrinsic Subprograms (`gcc/ada/doc/gnat_rm/intrinsic_subprograms.rst`) .. _Intrinsic_Subprograms: ********************* Intrinsic Subprograms ********************* .. index:: Intrinsic Subprograms GNAT allows a user application program to write the declaration: .. code-block:: ada pragma Import (Intrinsic, name); providing that the name corresponds to one of the implemented intrinsic subprograms in GNAT, and that the parameter profile of the referenced subprogram meets the requirements. This chapter describes the set of implemented intrinsic subprograms, and the requirements on parameter profiles. Note that no body is supplied; as with other uses of pragma Import, the body is supplied elsewhere (in this case by the compiler itself). Note that any use of this feature is potentially non-portable, since the Ada standard does not require Ada compilers to implement this feature. .. _Intrinsic_Operators: Intrinsic Operators =================== .. index:: Intrinsic operator All the predefined numeric operators in package Standard in ``pragma Import (Intrinsic,..)`` declarations. In the binary operator case, the operands must have the same size. The operand or operands must also be appropriate for the operator. For example, for addition, the operands must both be floating-point or both be fixed-point, and the right operand for ``"**"`` must have a root type of ``Standard.Integer'Base``. You can use an intrinsic operator declaration as in the following example: .. code-block:: ada type Int1 is new Integer; type Int2 is new Integer; function "+" (X1 : Int1; X2 : Int2) return Int1; function "+" (X1 : Int1; X2 : Int2) return Int2; pragma Import (Intrinsic, "+"); This declaration would permit 'mixed mode' arithmetic on items of the differing types ``Int1`` and ``Int2``. It is also possible to specify such operators for private types, if the full views are appropriate arithmetic types. .. _Compilation_ISO_Date: Compilation_ISO_Date ==================== .. index:: Compilation_ISO_Date This intrinsic subprogram is used in the implementation of the library package ``GNAT.Source_Info``. The only useful use of the intrinsic import in this case is the one in this unit, so an application program should simply call the function ``GNAT.Source_Info.Compilation_ISO_Date`` to obtain the date of the current compilation (in local time format YYYY-MM-DD). .. _Compilation_Date: Compilation_Date ================ .. index:: Compilation_Date Same as Compilation_ISO_Date, except the string is in the form MMM DD YYYY. .. _Compilation_Time: Compilation_Time ================ .. index:: Compilation_Time This intrinsic subprogram is used in the implementation of the library package ``GNAT.Source_Info``. The only useful use of the intrinsic import in this case is the one in this unit, so an application program should simply call the function ``GNAT.Source_Info.Compilation_Time`` to obtain the time of the current compilation (in local time format HH:MM:SS). .. _Enclosing_Entity: Enclosing_Entity ================ .. index:: Enclosing_Entity This intrinsic subprogram is used in the implementation of the library package ``GNAT.Source_Info``. The only useful use of the intrinsic import in this case is the one in this unit, so an application program should simply call the function ``GNAT.Source_Info.Enclosing_Entity`` to obtain the name of the current subprogram, package, task, entry, or protected subprogram. .. _Exception_Information: Exception_Information ===================== .. index:: Exception_Information' This intrinsic subprogram is used in the implementation of the library package ``GNAT.Current_Exception``. The only useful use of the intrinsic import in this case is the one in this unit, so an application program should simply call the function ``GNAT.Current_Exception.Exception_Information`` to obtain the exception information associated with the current exception. .. _Exception_Message: Exception_Message ================= .. index:: Exception_Message This intrinsic subprogram is used in the implementation of the library package ``GNAT.Current_Exception``. The only useful use of the intrinsic import in this case is the one in this unit, so an application program should simply call the function ``GNAT.Current_Exception.Exception_Message`` to obtain the message associated with the current exception. .. _Exception_Name: Exception_Name ============== .. index:: Exception_Name This intrinsic subprogram is used in the implementation of the library package ``GNAT.Current_Exception``. The only useful use of the intrinsic import in this case is the one in this unit, so an application program should simply call the function ``GNAT.Current_Exception.Exception_Name`` to obtain the name of the current exception. .. _File: File ==== .. index:: File This intrinsic subprogram is used in the implementation of the library package ``GNAT.Source_Info``. The only useful use of the intrinsic import in this case is the one in this unit, so an application program should simply call the function ``GNAT.Source_Info.File`` to obtain the name of the current file. .. _Line: Line ==== .. index:: Line This intrinsic subprogram is used in the implementation of the library package ``GNAT.Source_Info``. The only useful use of the intrinsic import in this case is the one in this unit, so an application program should simply call the function ``GNAT.Source_Info.Line`` to obtain the number of the current source line. .. _Shifts_and_Rotates: Shifts and Rotates ================== .. index:: Shift_Left .. index:: Shift_Right .. index:: Shift_Right_Arithmetic .. index:: Rotate_Left .. index:: Rotate_Right In standard Ada, the shift and rotate functions are available only for the predefined modular types in package ``Interfaces``. However, in GNAT it is possible to define these functions for any integer type (signed or modular), as in this example: .. code-block:: ada function Shift_Left (Value : T; Amount : Natural) return T with Import, Convention => Intrinsic; The function name must be one of Shift_Left, Shift_Right, Shift_Right_Arithmetic, Rotate_Left, or Rotate_Right. T must be an integer type. T'Size must be 8, 16, 32 or 64 bits; if T is modular, the modulus must be 2**8, 2**16, 2**32 or 2**64. The result type must be the same as the type of ``Value``. The shift amount must be Natural. The formal parameter names can be anything. A more convenient way of providing these shift operators is to use the Provide_Shift_Operators pragma, which provides the function declarations and corresponding pragma Import's for all five shift functions. For signed types the semantics of these operators is to interpret the bitwise result of the corresponding operator for modular type. In particular, shifting a negative number may change its sign bit to positive. .. _Source_Location: Source_Location =============== .. index:: Source_Location This intrinsic subprogram is used in the implementation of the library routine ``GNAT.Source_Info``. The only useful use of the intrinsic import in this case is the one in this unit, so an application program should simply call the function ``GNAT.Source_Info.Source_Location`` to obtain the current source file location. --- ### Gcc Ada Doc Gnat Rm Obsolescent Features (`gcc/ada/doc/gnat_rm/obsolescent_features.rst`) .. _Obsolescent_Features: ******************** Obsolescent Features ******************** This chapter describes features that are provided by GNAT, but are considered obsolescent since there are other, more appropriate, ways of achieving the same effect. These features are provided solely for historical compatibility purposes. .. _polyORB: PolyORB ======== PolyORB is a deprecated product. It will be baselined with the GNAT Pro release 28. After this release, there will be no new versions of this product. Contact AdaCore support to get recommendations for replacements. .. _pragma_No_Run_Time: pragma No_Run_Time ================== The pragma ``No_Run_Time`` is used to achieve an affect similar to the use of the "Zero Foot Print" configurable run time, but without requiring a specially configured run time. The result of using this pragma, which must be used for all units in a partition, is to restrict the use of any language features requiring run-time support code. The preferred usage is to use an appropriately configured run-time that includes just those features that are to be made accessible. .. _pragma_Ravenscar: pragma Ravenscar ================ The pragma ``Ravenscar`` has exactly the same effect as pragma ``Profile (Ravenscar)``. The latter usage is preferred since it is part of the new Ada 2005 standard. .. _pragma_Restricted_Run_Time: pragma Restricted_Run_Time ========================== The pragma ``Restricted_Run_Time`` has exactly the same effect as pragma ``Profile (Restricted)``. The latter usage is preferred since the Ada 2005 pragma ``Profile`` is intended for this kind of implementation dependent addition. .. _pragma_Task_Info: pragma Task_Info ================ The functionality provided by pragma ``Task_Info`` is now part of the Ada language. The ``CPU`` aspect and the package ``System.Multiprocessors`` offer a less system-dependent way to specify task affinity or to query the number of processors. Syntax .. code-block:: ada pragma Task_Info (EXPRESSION); This pragma appears within a task definition (like pragma ``Priority``) and applies to the task in which it appears. The argument must be of type ``System.Task_Info.Task_Info_Type``. The ``Task_Info`` pragma provides system dependent control over aspects of tasking implementation, for example, the ability to map tasks to specific processors. For details on the facilities available for the version of GNAT that you are using, see the documentation in the spec of package System.Task_Info in the runtime library. .. _package_System_Task_Info: package System.Task_Info (:file:`s-tasinf.ads`) =============================================== This package provides target dependent functionality that is used to support the ``Task_Info`` pragma. The predefined Ada package ``System.Multiprocessors`` and the ``CPU`` aspect now provide a standard replacement for GNAT's ``Task_Info`` functionality. .. raw:: latex \appendix --- ### Gcc Ada Doc Gnat Rm Representation Clauses And Pragmas (`gcc/ada/doc/gnat_rm/representation_clauses_and_pragmas.rst`) .. role:: switch(samp) .. role:: ada(code) :language: ada .. _Representation_Clauses_and_Pragmas: ********************************** Representation Clauses and Pragmas ********************************** .. index:: Representation Clauses .. index:: Representation Clause .. index:: Representation Pragma .. index:: Pragma, representation This section describes the representation clauses accepted by GNAT, and their effect on the representation of corresponding data objects. GNAT fully implements Annex C (Systems Programming). This means that all the implementation advice sections in chapter 13 are fully implemented. However, these sections only require a minimal level of support for representation clauses. GNAT provides much more extensive capabilities, and this section describes the additional capabilities provided. .. _Alignment_Clauses: Alignment Clauses ================= .. index:: Alignment Clause GNAT requires that all alignment clauses specify 0 or a power of 2, and all default alignments are always a power of 2. Specifying 0 is the same as specifying 1. The default alignment values are as follows: * *Elementary Types*. For elementary types, the alignment is the minimum of the actual size of objects of the type divided by ``Storage_Unit``, and the maximum default alignment supported by the target. (This maximum default alignment is given by the GNAT-specific attribute ``Standard'Maximum_Alignment``; see :ref:`Attribute_Maximum_Alignment`.) .. index:: Maximum_Alignment attribute For example, for type ``Long_Float``, the object size is 8 bytes, and the default alignment will be 8 on any target that supports alignments this large, but on some targets, the maximum alignment may be smaller than 8, in which case objects of type ``Long_Float`` will be maximally aligned. * *Arrays*. For arrays, the alignment is equal to the alignment of the component type for the normal case where no packing or component size is given. If the array is packed, and the packing is effective (see separate section on packed arrays), then the alignment will be either 4, 2, or 1 for long packed arrays or arrays whose length is not known at compile time, depending on whether the component size is divisible by 4, 2, or is odd. For short packed arrays, which are handled internally as modular types, the alignment will be as described for elementary types, e.g. a packed array of length 31 bits will have an object size of four bytes, and an alignment of 4. * *Records*. For the normal unpacked case, the alignment of a record is equal to the maximum alignment of any of its components. For tagged records, this includes the implicit access type used for the tag. If a pragma ``Pack`` is used and all components are packable (see separate section on pragma ``Pack``), then the resulting alignment is 1, unless the layout of the record makes it profitable to increase it. A special case is when: * the size of the record is given explicitly, or a full record representation clause is given, and * the size of the record is 2, 4, or 8 bytes. In this case, an alignment is chosen to match the size of the record. For example, if we have: .. code-block:: ada type Small is record A, B : Character; end record; for Small'Size use 16; then the default alignment of the record type ``Small`` is 2, not 1. This leads to more efficient code when the record is treated as a unit, and also allows the type to specified as ``Atomic`` on architectures requiring strict alignment. An alignment clause may specify a larger alignment than the default value up to some maximum value dependent on the target. It may also specify a smaller alignment than the default value for enumeration, integer and fixed point types, as well as for record types, for example .. code-block:: ada type V is record A : Integer; end record; for V'alignment use 1; .. index:: Alignment, default The default alignment for the type ``V`` is 4, as a result of the Integer field in the record, but it is permissible, as shown, to override the default alignment of the record with a smaller value. .. index:: Alignment, subtypes Note that according to the Ada standard, an alignment clause applies only to the first named subtype. If additional subtypes are declared, then the compiler is allowed to choose any alignment it likes, and there is no way to control this choice. Consider: .. code-block:: ada type R is range 1 .. 10_000; for R'Alignment use 1; subtype RS is R range 1 .. 1000; The alignment clause specifies an alignment of 1 for the first named subtype ``R`` but this does not necessarily apply to ``RS``. When writing portable Ada code, you should avoid writing code that explicitly or implicitly relies on the alignment of such subtypes. For the GNAT compiler, if an explicit alignment clause is given, this value is also used for any subsequent subtypes. So for GNAT, in the above example, you can count on the alignment of ``RS`` being 1. But this assumption is non-portable, and other compilers may choose different alignments for the subtype ``RS``. .. _Size_Clauses: Size Clauses ============ .. index:: Size Clause The default size for a type ``T`` is obtainable through the language-defined attribute ``T'Size`` and also through the equivalent GNAT-defined attribute ``T'Value_Size``. For objects of type ``T``, GNAT will generally increase the type size so that the object size (obtainable through the GNAT-defined attribute ``T'Object_Size``) is a multiple of ``T'Alignment * Storage_Unit``. For example: .. code-block:: ada type Smallint is range 1 .. 6; type Rec is record Y1 : integer; Y2 : boolean; end record; In this example, ``Smallint'Size`` = ``Smallint'Value_Size`` = 3, as specified by the RM rules, but objects of this type will have a size of 8 (``Smallint'Object_Size`` = 8), since objects by default occupy an integral number of storage units. On some targets, notably older versions of the Digital Alpha, the size of stand alone objects of this type may be 32, reflecting the inability of the hardware to do byte load/stores. Similarly, the size of type ``Rec`` is 40 bits (``Rec'Size`` = ``Rec'Value_Size`` = 40), but the alignment is 4, so objects of this type will have their size increased to 64 bits so that it is a multiple of the alignment (in bits). This decision is in accordance with the specific Implementation Advice in RM 13.3(43): "A ``Size`` clause should be supported for an object if the specified ``Size`` is at least as large as its subtype's ``Size``, and corresponds to a size in storage elements that is a multiple of the object's ``Alignment`` (if the ``Alignment`` is nonzero)." An explicit size clause may be used to override the default size by increasing it. For example, if we have: .. code-block:: ada type My_Boolean is new Boolean; for My_Boolean'Size use 32; then values of this type will always be 32-bit long. In the case of discrete types, the size can be increased up to 64 bits on 32-bit targets and 128 bits on 64-bit targets, with the effect that the entire specified field is used to hold the value, sign- or zero-extended as appropriate. If more than 64 bits or 128 bits resp. is specified, then padding space is allocated after the value, and a warning is issued that there are unused bits. Similarly the size of records and arrays may be increased, and the effect is to add padding bits after the value. This also causes a warning message to be generated. The largest Size value permitted in GNAT is 2**31-1. Since this is a Size in bits, this corresponds to an object of size 256 megabytes (minus one). This limitation is true on all targets. The reason for this limitation is that it improves the quality of the code in many cases if it is known that a Size value can be accommodated in an object of type Integer. .. _Storage_Size_Clauses: Storage_Size Clauses ==================== .. index:: Storage_Size Clause For tasks, the ``Storage_Size`` clause specifies the amount of space to be allocated for the task stack. This cannot be extended, and if the stack is exhausted, then ``Storage_Error`` will be raised (if stack checking is enabled). Use a ``Storage_Size`` attribute definition clause, or a ``Storage_Size`` pragma in the task definition to set the appropriate required size. A useful technique is to include in every task definition a pragma of the form: .. code-block:: ada pragma Storage_Size (Default_Stack_Size); Then ``Default_Stack_Size`` can be defined in a global package, and modified as required. Any tasks requiring stack sizes different from the default can have an appropriate alternative reference in the pragma. You can also use the *-d* binder switch to modify the default stack size. For access types, the ``Storage_Size`` clause specifies the maximum space available for allocation of objects of the type. If this space is exceeded then ``Storage_Error`` will be raised by an allocation attempt. In the case where the access type is declared local to a subprogram, the use of a ``Storage_Size`` clause triggers automatic use of a special predefined storage pool (``System.Pool_Size``) that ensures that all space for the pool is automatically reclaimed on exit from the scope in which the type is declared. A special case recognized by the compiler is the specification of a ``Storage_Size`` of zero for an access type. This means that no items can be allocated from the pool, and this is recognized at compile time, and all the overhead normally associated with maintaining a fixed size storage pool is eliminated. Consider the following example: .. code-block:: ada procedure p is type R is array (Natural) of Character; type P is access all R; for P'Storage_Size use 0; -- Above access type intended only for interfacing purposes y : P; procedure g (m : P); pragma Import (C, g); -- ... begin -- ... y := new R; end; As indicated in this example, these dummy storage pools are often useful in connection with interfacing where no object will ever be allocated. If you compile the above example, you get the warning: :: p.adb:16:09: warning: allocation from empty storage pool p.adb:16:09: warning: Storage_Error will be raised at run time Of course in practice, there will not be any explicit allocators in the case of such an access declaration. .. _Size_of_Variant_Record_Objects: Size of Variant Record Objects ============================== .. index:: Size, variant record objects .. index:: Variant record objects, size In the case of variant record objects, there is a question whether Size gives information about a particular variant, or the maximum size required for any variant. Consider the following program .. code-block:: ada with Text_IO; use Text_IO; procedure q is type R1 (A : Boolean := False) is record case A is when True => X : Character; when False => null; end case; end record; V1 : R1 (False); V2 : R1; begin Put_Line (Integer'Image (V1'Size)); Put_Line (Integer'Image (V2'Size)); end q; Here we are dealing with a variant record, where the True variant requires 16 bits, and the False variant requires 8 bits. In the above example, both V1 and V2 contain the False variant, which is only 8 bits long. However, the result of running the program is: :: 8 16 The reason for the difference here is that the discriminant value of V1 is fixed, and will always be False. It is not possible to assign a True variant value to V1, therefore 8 bits is sufficient. On the other hand, in the case of V2, the initial discriminant value is False (from the default), but it is possible to assign a True variant value to V2, therefore 16 bits must be allocated for V2 in the general case, even fewer bits may be needed at any particular point during the program execution. As can be seen from the output of this program, the ``'Size`` attribute applied to such an object in GNAT gives the actual allocated size of the variable, which is the largest size of any of the variants. The Ada Reference Manual is not completely clear on what choice should be made here, but the GNAT behavior seems most consistent with the language in the RM. In some cases, it may be desirable to obtain the size of the current variant, rather than the size of the largest variant. This can be achieved in GNAT by making use of the fact that in the case of a subprogram parameter, GNAT does indeed return the size of the current variant (because a subprogram has no way of knowing how much space is actually allocated for the actual). Consider the following modified version of the above program: .. code-block:: ada with Text_IO; use Text_IO; procedure q is type R1 (A : Boolean := False) is record case A is when True => X : Character; when False => null; end case; end record; V2 : R1; function Size (V : R1) return Integer is begin return V'Size; end Size; begin Put_Line (Integer'Image (V2'Size)); Put_Line (Integer'Image (Size (V2))); V2 := (True, 'x'); Put_Line (Integer'Image (V2'Size)); Put_Line (Integer'Image (Size (V2))); end q; The output from this program is :: 16 8 16 16 Here we see that while the ``'Size`` attribute always returns the maximum size, regardless of the current variant value, the ``Size`` function does indeed return the size of the current variant value. .. _Biased_Representation: Biased Representation ===================== .. index:: Size for biased representation .. index:: Biased representation In the case of scalars with a range starting at other than zero, it is possible in some cases to specify a size smaller than the default minimum value, and in such cases, GNAT uses an unsigned biased representation, in which zero is used to represent the lower bound, and successive values represent successive values of the type. For example, suppose we have the declaration: .. code-block:: ada type Small is range -7 .. -4; for Small'Size use 2; Although the default size of type ``Small`` is 4, the ``Size`` clause is accepted by GNAT and results in the following representation scheme: :: -7 is represented as 2#00# -6 is represented as 2#01# -5 is represented as 2#10# -4 is represented as 2#11# Biased representation is only used if the specified ``Size`` clause cannot be accepted in any other manner. These reduced sizes that force biased representation can be used for all discrete types except for enumeration types for which a representation clause is given. .. _Value_Size_and_Object_Size_Clauses: Value_Size and Object_Size Clauses ================================== .. index:: Value_Size .. index:: Object_Size .. index:: Size, of objects In Ada 95 and Ada 2005, ``T'Size`` for a type ``T`` is the minimum number of bits required to hold values of type ``T``. Although this interpretation was allowed in Ada 83, it was not required, and this requirement in practice can cause some significant difficulties. For example, in most Ada 83 compilers, ``Natural'Size`` was 32. However, in Ada 95 and Ada 2005, ``Natural'Size`` is typically 31. This means that code may change in behavior when moving from Ada 83 to Ada 95 or Ada 2005. For example, consider: .. code-block:: ada type Rec is record A : Natural; B : Natural; end record; for Rec use record A at 0 range 0 .. Natural'Size - 1; B at 0 range Natural'Size .. 2 * Natural'Size - 1; end record; In the above code, since the typical size of ``Natural`` objects is 32 bits and ``Natural'Size`` is 31, the above code can cause unexpected inefficient packing in Ada 95 and Ada 2005, and in general there are cases where the fact that the object size can exceed the size of the type causes surprises. To help get around this problem GNAT provides two implementation defined attributes, ``Value_Size`` and ``Object_Size``. When applied to a type, these attributes yield the size of the type (corresponding to the RM defined size attribute), and the size of objects of the type respectively. The ``Object_Size`` is used for determining the default size of objects and components. This size value can be referred to using the ``Object_Size`` attribute. The phrase 'is used' here means that it is the basis of the determination of the size. The backend is free to pad this up if necessary for efficiency, e.g., an 8-bit stand-alone character might be stored in 32 bits on a machine with no efficient byte access instructions such as the Alpha. The default rules for the value of ``Object_Size`` for discrete types are as follows: * The ``Object_Size`` for base subtypes reflect the natural hardware size in bits (run the compiler with *-gnatS* to find those values for numeric types). Enumeration types and fixed-point base subtypes have 8, 16, 32, or 64 bits for this size, depending on the range of values to be stored. * The ``Object_Size`` of a subtype is the same as the ``Object_Size`` of the type from which it is obtained. * The ``Object_Size`` of a derived base type is copied from the parent base type, and the ``Object_Size`` of a derived first subtype is copied from the parent first subtype. The ``Value_Size`` attribute is the (minimum) number of bits required to store a value of the type. This value is used to determine how tightly to pack records or arrays with components of this type, and also affects the semantics of unchecked conversion (unchecked conversions where the ``Value_Size`` values differ generate a warning, and are potentially target dependent). The default rules for the value of ``Value_Size`` are as follows: * The ``Value_Size`` for a base subtype is the minimum number of bits required to store all values of the type (including the sign bit only if negative values are possible). * If a subtype statically matches the first subtype of a given type, then it has by default the same ``Value_Size`` as the first subtype. (This is a consequence of RM 13.1(14): "if two subtypes statically match, then their subtype-specific aspects are the same".) * All other subtypes have a ``Value_Size`` corresponding to the minimum number of bits required to store all values of the subtype. For dynamic bounds, it is assumed that the value can range down or up to the corresponding bound of the ancestor The RM defined attribute ``Size`` corresponds to the ``Value_Size`` attribute. The ``Size`` attribute may be defined for a first-named subtype. This sets the ``Value_Size`` of the first-named subtype to the given value, and the ``Object_Size`` of this first-named subtype to the given value padded up to an appropriate boundary. It is a consequence of the default rules above that this ``Object_Size`` will apply to all further subtypes. On the other hand, ``Value_Size`` is affected only for the first subtype, any dynamic subtypes obtained from it directly, and any statically matching subtypes. The ``Value_Size`` of any other static subtypes is not affected. ``Value_Size`` and ``Object_Size`` may be explicitly set for any subtype using an attribute definition clause. Note that the use of these attributes can cause the RM 13.1(14) rule to be violated. If two access types reference aliased objects whose subtypes have differing ``Object_Size`` values as a result of explicit attribute definition clauses, then it is illegal to convert from one access subtype to the other. For a more complete description of this additional legality rule, see the description of the ``Object_Size`` attribute. To get a feel for the difference, consider the following examples (note that in each case the base is ``Short_Short_Integer`` with a size of 8): +------------------------------------------------+-------------+-------------+ |Type or subtype declaration | Object_Size | Value_Size| +================================================+=============+=============+ |:ada:`type X1 is range 0 .. 5;` | 8 | 3 | +------------------------------------------------+-------------+-------------+ |:ada:`type X2 is range 0 .. 5; | 16 | 12 | |for X2'Size use 12;` | | | +------------------------------------------------+-------------+-------------+ |:ada:`subtype X3 is X2 range 0 .. 3;` | 16 | 2 | +------------------------------------------------+-------------+-------------+ |:ada:`subtype X4 is X2'Base range 0 .. 10;` | 8 | 4 | +------------------------------------------------+-------------+-------------+ |:ada:`Dynamic : X2'Base range -64 .. +63;` | | | +------------------------------------------------+-------------+-------------+ |:ada:`subtype X5 is X2 range 0 .. Dynamic;` | 16 | 3* | +------------------------------------------------+-------------+-------------+ |:ada:`subtype X6 is X2'Base range 0 .. Dynamic;`| 8 | 7* | +------------------------------------------------+-------------+-------------+ Note: the entries marked '*' are not actually specified by the Ada Reference Manual, which has nothing to say about size in the dynamic case. What GNAT does is to allocate sufficient bits to accommodate any possible dynamic values for the bounds at run-time. So far, so good, but GNAT has to obey the RM rules, so the question is under what conditions must the RM ``Size`` be used. The following is a list of the occasions on which the RM ``Size`` must be used: * Component size for packed arrays or records * Value of the attribute ``Size`` for a type * Warning about sizes not matching for unchecked conversion For record types, the ``Object_Size`` is always a multiple of the alignment of the type (this is true for all types). In some cases the ``Value_Size`` can be smaller. Consider: .. code-block:: ada type R is record X : Integer; Y : Character; end record; On a typical 32-bit architecture, the X component will occupy four bytes and the Y component will occupy one byte, for a total of 5 bytes. As a result ``R'Value_Size`` will be 40 (bits) since this is the minimum size required to store a value of this type. For example, it is permissible to have a component of type R in an array whose component size is specified to be 40 bits. However, ``R'Object_Size`` will be 64 (bits). The difference is due to the alignment requirement for objects of the record type. The X component will require four-byte alignment because that is what type Integer requires, whereas the Y component, a Character, will only require 1-byte alignment. Since the alignment required for X is the greatest of all the components' alignments, that is the alignment required for the enclosing record type, i.e., 4 bytes or 32 bits. As indicated above, the actual object size must be rounded up so that it is a multiple of the alignment value. Therefore, 40 bits rounded up to the next multiple of 32 yields 64 bits. For all other types, the ``Object_Size`` and ``Value_Size`` are the same (and equivalent to the RM attribute ``Size``). Only ``Size`` may be specified for such types. Note that ``Value_Size`` can be used to force biased representation for a particular subtype. Consider this example: .. code-block:: ada type R is (A, B, C, D, E, F); subtype RAB is R range A .. B; subtype REF is R range E .. F; By default, ``RAB`` has a size of 1 (sufficient to accommodate the representation of ``A`` and ``B``, 0 and 1), and ``REF`` has a size of 3 (sufficient to accommodate the representation of ``E`` and ``F``, 4 and 5). But if we add the following ``Value_Size`` attribute definition clause: .. code-block:: ada for REF'Value_Size use 1; then biased representation is forced for ``REF``, and 0 will represent ``E`` and 1 will represent ``F``. A warning is issued when a ``Value_Size`` attribute definition clause forces biased representation. This warning can be turned off using :switch:`-gnatw.B`. .. _Component_Size_Clauses: Component_Size Clauses ====================== .. index:: Component_Size Clause Normally, the value specified in a component size clause must be consistent with the subtype of the array component with regard to size and alignment. In other words, the value specified must be at least equal to the size of this subtype, and must be a multiple of the alignment value. In addition, component size clauses are allowed which cause the array to be packed, by specifying a smaller value. A first case is for component size values in the range 1 through 63 on 32-bit targets, and 1 through 127 on 64-bit targets. The value specified may not be smaller than the Size of the subtype. GNAT will accurately honor all packing requests in this range. For example, if we have: .. code-block:: ada type r is array (1 .. 8) of Natural; for r'Component_Size use 31; then the resulting array has a length of 31 bytes (248 bits = 8 * 31). Of course access to the components of such an array is considerably less efficient than if the natural component size of 32 is used. A second case is when the subtype of the component is a record type padded because of its default alignment. For example, if we have: .. code-block:: ada type r is record i : Integer; j : Integer; b : Boolean; end record; type a is array (1 .. 8) of r; for a'Component_Size use 72; then the resulting array has a length of 72 bytes, instead of 96 bytes if the alignment of the record (4) was obeyed. Note that there is no point in giving both a component size clause and a pragma Pack for the same array type. if such duplicate clauses are given, the pragma Pack will be ignored. .. _Bit_Order_Clauses: Bit_Order Clauses ================= .. index:: Bit_Order Clause .. index:: bit ordering .. index:: ordering, of bits For record subtypes, GNAT permits the specification of the ``Bit_Order`` attribute. The specification may either correspond to the default bit order for the target, in which case the specification has no effect and places no additional restrictions, or it may be for the non-standard setting (that is the opposite of the default). In the case where the non-standard value is specified, the effect is to renumber bits within each byte, but the ordering of bytes is not affected. There are certain restrictions placed on component clauses as follows: * Components fitting within a single storage unit. These are unrestricted, and the effect is merely to renumber bits. For example if we are on a little-endian machine with ``Low_Order_First`` being the default, then the following two declarations have exactly the same effect: :: type R1 is record A : Boolean; B : Integer range 1 .. 120; end record; for R1 use record A at 0 range 0 .. 0; B at 0 range 1 .. 7; end record; type R2 is record A : Boolean; B : Integer range 1 .. 120; end record; for R2'Bit_Order use High_Order_First; for R2 use record A at 0 range 7 .. 7; B at 0 range 0 .. 6; end record; The useful application here is to write the second declaration with the ``Bit_Order`` attribute definition clause, and know that it will be treated the same, regardless of whether the target is little-endian or big-endian. * Components occupying an integral number of bytes. These are components that exactly fit in two or more bytes. Such component declarations are allowed, but have no effect, since it is important to realize that the ``Bit_Order`` specification does not affect the ordering of bytes. In particular, the following attempt at getting an endian-independent integer does not work: :: type R2 is record A : Integer; end record; for R2'Bit_Order use High_Order_First; for R2 use record A at 0 range 0 .. 31; end record; This declaration will result in a little-endian integer on a little-endian machine, and a big-endian integer on a big-endian machine. If byte flipping is required for interoperability between big- and little-endian machines, this must be explicitly programmed. This capability is not provided by ``Bit_Order``. * Components that are positioned across byte boundaries. but do not occupy an integral number of bytes. Given that bytes are not reordered, such fields would occupy a non-contiguous sequence of bits in memory, requiring non-trivial code to reassemble. They are for this reason not permitted, and any component clause specifying such a layout will be flagged as illegal by GNAT. Since the misconception that Bit_Order automatically deals with all endian-related incompatibilities is a common one, the specification of a component field that is an integral number of bytes will always generate a warning. This warning may be suppressed using ``pragma Warnings (Off)`` if desired. The following section contains additional details regarding the issue of byte ordering. .. _Effect_of_Bit_Order_on_Byte_Ordering: Effect of Bit_Order on Byte Ordering ==================================== .. index:: byte ordering .. index:: ordering, of bytes In this section we will review the effect of the ``Bit_Order`` attribute definition clause on byte ordering. Briefly, it has no effect at all, but a detailed example will be helpful. Before giving this example, let us review the precise definition of the effect of defining ``Bit_Order``. The effect of a non-standard bit order is described in section 13.5.3 of the Ada Reference Manual: "2 A bit ordering is a method of interpreting the meaning of the storage place attributes." To understand the precise definition of storage place attributes in this context, we visit section 13.5.1 of the manual: "13 A record_representation_clause (without the mod_clause) specifies the layout. The storage place attributes (see 13.5.2) are taken from the values of the position, first_bit, and last_bit expressions after normalizing those values so that first_bit is less than Storage_Unit." The critical point here is that storage places are taken from the values after normalization, not before. So the ``Bit_Order`` interpretation applies to normalized values. The interpretation is described in the later part of the 13.5.3 paragraph: "2 A bit ordering is a method of interpreting the meaning of the storage place attributes. High_Order_First (known in the vernacular as 'big endian') means that the first bit of a storage element (bit 0) is the most significant bit (interpreting the sequence of bits that represent a component as an unsigned integer value). Low_Order_First (known in the vernacular as 'little endian') means the opposite: the first bit is the least significant." Note that the numbering is with respect to the bits of a storage unit. In other words, the specification affects only the numbering of bits within a single storage unit. We can make the effect clearer by giving an example. Suppose that we have an external device which presents two bytes, the first byte presented, which is the first (low addressed byte) of the two byte record is called Master, and the second byte is called Slave. The left most (most significant) bit is called Control for each byte, and the remaining 7 bits are called V1, V2, ... V7, where V7 is the rightmost (least significant) bit. On a big-endian machine, we can write the following representation clause .. code-block:: ada type Data is record Master_Control : Bit; Master_V1 : Bit; Master_V2 : Bit; Master_V3 : Bit; Master_V4 : Bit; Master_V5 : Bit; Master_V6 : Bit; Master_V7 : Bit; Slave_Control : Bit; Slave_V1 : Bit; Slave_V2 : Bit; Slave_V3 : Bit; Slave_V4 : Bit; Slave_V5 : Bit; Slave_V6 : Bit; Slave_V7 : Bit; end record; for Data use record Master_Control at 0 range 0 .. 0; Master_V1 at 0 range 1 .. 1; Master_V2 at 0 range 2 .. 2; Master_V3 at 0 range 3 .. 3; Master_V4 at 0 range 4 .. 4; Master_V5 at 0 range 5 .. 5; Master_V6 at 0 range 6 .. 6; Master_V7 at 0 range 7 .. 7; Slave_Control at 1 range 0 .. 0; Slave_V1 at 1 range 1 .. 1; Slave_V2 at 1 range 2 .. 2; Slave_V3 at 1 range 3 .. 3; Slave_V4 at 1 range 4 .. 4; Slave_V5 at 1 range 5 .. 5; Slave_V6 at 1 range 6 .. 6; Slave_V7 at 1 range 7 .. 7; end record; Now if we move this to a little endian machine, then the bit ordering within the byte is backwards, so we have to rewrite the record rep clause as: .. code-block:: ada for Data use record Master_Control at 0 range 7 .. 7; Master_V1 at 0 range 6 .. 6; Master_V2 at 0 range 5 .. 5; Master_V3 at 0 range 4 .. 4; Master_V4 at 0 range 3 .. 3; Master_V5 at 0 range 2 .. 2; Master_V6 at 0 range 1 .. 1; Master_V7 at 0 range 0 .. 0; Slave_Control at 1 range 7 .. 7; Slave_V1 at 1 range 6 .. 6; Slave_V2 at 1 range 5 .. 5; Slave_V3 at 1 range 4 .. 4; Slave_V4 at 1 range 3 .. 3; Slave_V5 at 1 range 2 .. 2; Slave_V6 at 1 range 1 .. 1; Slave_V7 at 1 range 0 .. 0; end record; It is a nuisance to have to rewrite the clause, especially if the code has to be maintained on both machines. However, this is a case that we can handle with the ``Bit_Order`` attribute if it is implemented. Note that the implementation is not required on byte addressed machines, but it is indeed implemented in GNAT. This means that we can simply use the first record clause, together with the declaration .. code-block:: ada for Data'Bit_Order use High_Order_First; and the effect is what is desired, namely the layout is exactly the same, independent of whether the code is compiled on a big-endian or little-endian machine. The important point to understand is that byte ordering is not affected. A ``Bit_Order`` attribute definition never affects which byte a field ends up in, only where it ends up in that byte. To make this clear, let us rewrite the record rep clause of the previous example as: .. code-block:: ada for Data'Bit_Order use High_Order_First; for Data use record Master_Control at 0 range 0 .. 0; Master_V1 at 0 range 1 .. 1; Master_V2 at 0 range 2 .. 2; Master_V3 at 0 range 3 .. 3; Master_V4 at 0 range 4 .. 4; Master_V5 at 0 range 5 .. 5; Master_V6 at 0 range 6 .. 6; Master_V7 at 0 range 7 .. 7; Slave_Control at 0 range 8 .. 8; Slave_V1 at 0 range 9 .. 9; Slave_V2 at 0 range 10 .. 10; Slave_V3 at 0 range 11 .. 11; Slave_V4 at 0 range 12 .. 12; Slave_V5 at 0 range 13 .. 13; Slave_V6 at 0 range 14 .. 14; Slave_V7 at 0 range 15 .. 15; end record; This is exactly equivalent to saying (a repeat of the first example): .. code-block:: ada for Data'Bit_Order use High_Order_First; for Data use record Master_Control at 0 range 0 .. 0; Master_V1 at 0 range 1 .. 1; Master_V2 at 0 range 2 .. 2; Master_V3 at 0 range 3 .. 3; Master_V4 at 0 range 4 .. 4; Master_V5 at 0 range 5 .. 5; Master_V6 at 0 range 6 .. 6; Master_V7 at 0 range 7 .. 7; Slave_Control at 1 range 0 .. 0; Slave_V1 at 1 range 1 .. 1; Slave_V2 at 1 range 2 .. 2; Slave_V3 at 1 range 3 .. 3; Slave_V4 at 1 range 4 .. 4; Slave_V5 at 1 range 5 .. 5; Slave_V6 at 1 range 6 .. 6; Slave_V7 at 1 range 7 .. 7; end record; Why are they equivalent? Well take a specific field, the ``Slave_V2`` field. The storage place attributes are obtained by normalizing the values given so that the ``First_Bit`` value is less than 8. After normalizing the values (0,10,10) we get (1,2,2) which is exactly what we specified in the other case. Now one might expect that the ``Bit_Order`` attribute might affect bit numbering within the entire record component (two bytes in this case, thus affecting which byte fields end up in), but that is not the way this feature is defined, it only affects numbering of bits, not which byte they end up in. Consequently it never makes sense to specify a starting bit number greater than 7 (for a byte addressable field) if an attribute definition for ``Bit_Order`` has been given, and indeed it may be actively confusing to specify such a value, so the compiler generates a warning for such usage. If you do need to control byte ordering then appropriate conditional values must be used. If in our example, the slave byte came first on some machines we might write: .. code-block:: ada Master_Byte_First constant Boolean := ...; Master_Byte : constant Natural := 1 - Boolean'Pos (Master_Byte_First); Slave_Byte : constant Natural := Boolean'Pos (Master_Byte_First); for Data'Bit_Order use High_Order_First; for Data use record Master_Control at Master_Byte range 0 .. 0; Master_V1 at Master_Byte range 1 .. 1; Master_V2 at Master_Byte range 2 .. 2; Master_V3 at Master_Byte range 3 .. 3; Master_V4 at Master_Byte range 4 .. 4; Master_V5 at Master_Byte range 5 .. 5; Master_V6 at Master_Byte range 6 .. 6; Master_V7 at Master_Byte range 7 .. 7; Slave_Control at Slave_Byte range 0 .. 0; Slave_V1 at Slave_Byte range 1 .. 1; Slave_V2 at Slave_Byte range 2 .. 2; Slave_V3 at Slave_Byte range 3 .. 3; Slave_V4 at Slave_Byte range 4 .. 4; Slave_V5 at Slave_Byte range 5 .. 5; Slave_V6 at Slave_Byte range 6 .. 6; Slave_V7 at Slave_Byte range 7 .. 7; end record; Now to switch between machines, all that is necessary is to set the boolean constant ``Master_Byte_First`` in an appropriate manner. .. _Pragma_Pack_for_Arrays: Pragma Pack for Arrays ====================== .. index:: Pragma Pack (for arrays) Pragma ``Pack`` applied to an array has an effect that depends upon whether the component type is *packable*. For a component type to be *packable*, it must be one of the following cases: * Any elementary type. * Any small packed array type with a static size. * Any small simple record type with a static size. For all these cases, if the component subtype size is in the range 1 through 63 on 32-bit targets, and 1 through 127 on 64-bit targets, then the effect of the pragma ``Pack`` is exactly as though a component size were specified giving the component subtype size. All other types are non-packable, they occupy an integral number of storage units and the only effect of pragma Pack is to remove alignment gaps. For example if we have: .. code-block:: ada type r is range 0 .. 17; type ar is array (1 .. 8) of r; pragma Pack (ar); Then the component size of ``ar`` will be set to 5 (i.e., to ``r'size``, and the size of the array ``ar`` will be exactly 40 bits). Note that in some cases this rather fierce approach to packing can produce unexpected effects. For example, in Ada 95 and Ada 2005, subtype ``Natural`` typically has a size of 31, meaning that if you pack an array of ``Natural``, you get 31-bit close packing, which saves a few bits, but results in far less efficient access. Since many other Ada compilers will ignore such a packing request, GNAT will generate a warning on some uses of pragma ``Pack`` that it guesses might not be what is intended. You can easily remove this warning by using an explicit ``Component_Size`` setting instead, which never generates a warning, since the intention of the programmer is clear in this case. GNAT treats packed arrays in one of two ways. If the size of the array is known at compile time and is at most 64 bits on 32-bit targets, and at most 128 bits on 64-bit targets, then internally the array is represented as a single modular type, of exactly the appropriate number of bits. If the length is greater than 64 bits on 32-bit targets, and greater than 128 bits on 64-bit targets, or is not known at compile time, then the packed array is represented as an array of bytes, and its length is always a multiple of 8 bits. Note that to represent a packed array as a modular type, the alignment must be suitable for the modular type involved. For example, on typical machines a 32-bit packed array will be represented by a 32-bit modular integer with an alignment of four bytes. If you explicitly override the default alignment with an alignment clause that is too small, the modular representation cannot be used. For example, consider the following set of declarations: .. code-block:: ada type R is range 1 .. 3; type S is array (1 .. 31) of R; for S'Component_Size use 2; for S'Size use 62; for S'Alignment use 1; If the alignment clause were not present, then a 62-bit modular representation would be chosen (typically with an alignment of 4 or 8 bytes depending on the target). But the default alignment is overridden with the explicit alignment clause. This means that the modular representation cannot be used, and instead the array of bytes representation must be used, meaning that the length must be a multiple of 8. Thus the above set of declarations will result in a diagnostic rejecting the size clause and noting that the minimum size allowed is 64. .. index:: Pragma Pack (for type Natural) .. index:: Pragma Pack warning One special case that is worth noting occurs when the base type of the component size is 8/16/32 and the subtype is one bit less. Notably this occurs with subtype ``Natural``. Consider: .. code-block:: ada type Arr is array (1 .. 32) of Natural; pragma Pack (Arr); In all commonly used Ada 83 compilers, this pragma Pack would be ignored, since typically ``Natural'Size`` is 32 in Ada 83, and in any case most Ada 83 compilers did not attempt 31 bit packing. In Ada 95 and Ada 2005, ``Natural'Size`` is required to be 31. Furthermore, GNAT really does pack 31-bit subtype to 31 bits. This may result in a substantial unintended performance penalty when porting legacy Ada 83 code. To help prevent this, GNAT generates a warning in such cases. If you really want 31 bit packing in a case like this, you can set the component size explicitly: .. code-block:: ada type Arr is array (1 .. 32) of Natural; for Arr'Component_Size use 31; Here 31-bit packing is achieved as required, and no warning is generated, since in this case the programmer intention is clear. .. _Pragma_Pack_for_Records: Pragma Pack for Records ======================= .. index:: Pragma Pack (for records) Pragma ``Pack`` applied to a record will pack the components to reduce wasted space from alignment gaps and by reducing the amount of space taken by components. We distinguish between *packable* components and *non-packable* components. Components of the following types are considered packable: * Components of an elementary type are packable unless they are aliased, independent or atomic. * Small packed arrays, where the size is statically known, are represented internally as modular integers, and so they are also packable. * Small simple records, where the size is statically known, are also packable. For all these cases, if the ``'Size`` value is in the range 1 through 64 on 32-bit targets, and 1 through 128 on 64-bit targets, the components occupy the exact number of bits corresponding to this value and are packed with no padding bits, i.e. they can start on an arbitrary bit boundary. All other types are non-packable, they occupy an integral number of storage units and the only effect of pragma ``Pack`` is to remove alignment gaps. For example, consider the record .. code-block:: ada type Rb1 is array (1 .. 13) of Boolean; pragma Pack (Rb1); type Rb2 is array (1 .. 65) of Boolean; pragma Pack (Rb2); type AF is new Float with Atomic; type X2 is record L1 : Boolean; L2 : Duration; L3 : AF; L4 : Boolean; L5 : Rb1; L6 : Rb2; end record; pragma Pack (X2); The representation for the record ``X2`` is as follows on 32-bit targets: .. code-block:: ada for X2'Size use 224; for X2 use record L1 at 0 range 0 .. 0; L2 at 0 range 1 .. 64; L3 at 12 range 0 .. 31; L4 at 16 range 0 .. 0; L5 at 16 range 1 .. 13; L6 at 18 range 0 .. 71; end record; Studying this example, we see that the packable fields ``L1`` and ``L2`` are of length equal to their sizes, and placed at specific bit boundaries (and not byte boundaries) to eliminate padding. But ``L3`` is of a non-packable float type (because it is aliased), so it is on the next appropriate alignment boundary. The next two fields are fully packable, so ``L4`` and ``L5`` are minimally packed with no gaps. However, type ``Rb2`` is a packed array that is longer than 64 bits, so it is itself non-packable on 32-bit targets. Thus the ``L6`` field is aligned to the next byte boundary, and takes an integral number of bytes, i.e., 72 bits. .. _Record_Representation_Clauses: Record Representation Clauses ============================= .. index:: Record Representation Clause Record representation clauses may be given for all record types, including types obtained by record extension. Component clauses are allowed for any static component. The restrictions on component clauses depend on the type of the component. .. index:: Component Clause For all components of an elementary type, the only restriction on component clauses is that the size must be at least the ``'Size`` value of the type (actually the Value_Size). There are no restrictions due to alignment, and such components may freely cross storage boundaries. Packed arrays with a size up to and including 64 bits on 32-bit targets, and up to and including 128 bits on 64-bit targets, are represented internally using a modular type with the appropriate number of bits, and thus the same lack of restriction applies. For example, if you declare: .. code-block:: ada type R is array (1 .. 49) of Boolean; pragma Pack (R); for R'Size use 49; then a component clause for a component of type ``R`` may start on any specified bit boundary, and may specify a value of 49 bits or greater. For packed bit arrays that are longer than 64 bits on 32-bit targets, and longer than 128 bits on 64-bit targets, there are two cases. If the component size is a power of 2 (1,2,4,8,16,32,64 bits), including the important case of single bits or boolean values, then there are no limitations on placement of such components, and they may start and end at arbitrary bit boundaries. If the component size is not a power of 2 (e.g., 3 or 5), then an array of this type must always be placed on on a storage unit (byte) boundary and occupy an integral number of storage units (bytes). Any component clause that does not meet this requirement will be rejected. Any aliased component, or component of an aliased type, must have its normal alignment and size. A component clause that does not meet this requirement will be rejected. The tag field of a tagged type always occupies an address sized field at the start of the record. No component clause may attempt to overlay this tag. When a tagged type appears as a component, the tag field must have proper alignment In the case of a record extension ``T1``, of a type ``T``, no component clause applied to the type ``T1`` can specify a storage location that would overlap the first ``T'Object_Size`` bits of the record. For all other component types, including non-bit-packed arrays, the component can be placed at an arbitrary bit boundary, so for example, the following is permitted: .. code-block:: ada type R is array (1 .. 10) of Boolean; for R'Size use 80; type Q is record G, H : Boolean; L, M : R; end record; for Q use record G at 0 range 0 .. 0; H at 0 range 1 .. 1; L at 0 range 2 .. 81; R at 0 range 82 .. 161; end record; .. _Handling_of_Records_with_Holes: Handling of Records with Holes ============================== .. index:: Handling of Records with Holes As a result of alignment considerations, records may contain "holes" or gaps which do not correspond to the data bits of any of the components. Record representation clauses can also result in holes in records. GNAT does not attempt to clear these holes, so in record objects, they should be considered to hold undefined rubbish. The generated equality routine just tests components so does not access these undefined bits, and assignment and copy operations may or may not preserve the contents of these holes (for assignments, the holes in the target will in practice contain either the bits that are present in the holes in the source, or the bits that were present in the target before the assignment). If it is necessary to ensure that holes in records have all zero bits, then record objects for which this initialization is desired should be explicitly set to all zero values using Unchecked_Conversion or address overlays. For example .. code-block:: ada type HRec is record C : Character; I : Integer; end record; On typical machines, integers need to be aligned on a four-byte boundary, resulting in three bytes of undefined rubbish following the 8-bit field for C. To ensure that the hole in a variable of type HRec is set to all zero bits, you could for example do: .. code-block:: ada type Base is record Dummy1, Dummy2 : Integer := 0; end record; BaseVar : Base; RealVar : Hrec; for RealVar'Address use BaseVar'Address; Now the 8-bytes of the value of RealVar start out containing all zero bits. A safer approach is to just define dummy fields, avoiding the holes, as in: .. code-block:: ada type HRec is record C : Character; Dummy1 : Short_Short_Integer := 0; Dummy2 : Short_Short_Integer := 0; Dummy3 : Short_Short_Integer := 0; I : Integer; end record; And to make absolutely sure that the intent of this is followed, you can use representation clauses: .. code-block:: ada for Hrec use record C at 0 range 0 .. 7; Dummy1 at 1 range 0 .. 7; Dummy2 at 2 range 0 .. 7; Dummy3 at 3 range 0 .. 7; I at 4 range 0 .. 31; end record; for Hrec'Size use 64; .. _Enumeration_Clauses: Enumeration Clauses =================== The only restriction on enumeration clauses is that the range of values must be representable. For the signed case, if one or more of the representation values are negative, all values must be in the range: .. code-block:: ada System.Min_Int .. System.Max_Int For the unsigned case, where all values are nonnegative, the values must be in the range: .. code-block:: ada 0 .. System.Max_Binary_Modulus; A *confirming* representation clause is one in which the values range from 0 in sequence, i.e., a clause that confirms the default representation for an enumeration type. Such a confirming representation is permitted by these rules, and is specially recognized by the compiler so that no extra overhead results from the use of such a clause. If an array has an index type which is an enumeration type to which an enumeration clause has been applied, then the array is stored in a compact manner. Consider the declarations: .. code-block:: ada type r is (A, B, C); for r use (A => 1, B => 5, C => 10); type t is array (r) of Character; The array type t corresponds to a vector with exactly three elements and has a default size equal to ``3*Character'Size``. This ensures efficient use of space, but means that accesses to elements of the array will incur the overhead of converting representation values to the corresponding positional values, (i.e., the value delivered by the ``Pos`` attribute). .. _Address_Clauses: Address Clauses =============== .. index:: Address Clause The reference manual allows a general restriction on representation clauses, as found in RM 13.1(22): "An implementation need not support representation items containing nonstatic expressions, except that an implementation should support a representation item for a given entity if each nonstatic expression in the representation item is a name that statically denotes a constant declared before the entity." In practice this is applicable only to address clauses, since this is the only case in which a nonstatic expression is permitted by the syntax. As the AARM notes in sections 13.1 (22.a-22.h): 22.a Reason: This is to avoid the following sort of thing: 22.b X : Integer := F(...); Y : Address := G(...); for X'Address use Y; 22.c In the above, we have to evaluate the initialization expression for X before we know where to put the result. This seems like an unreasonable implementation burden. 22.d The above code should instead be written like this: 22.e Y : constant Address := G(...); X : Integer := F(...); for X'Address use Y; 22.f This allows the expression 'Y' to be safely evaluated before X is created. 22.g The constant could be a formal parameter of mode in. 22.h An implementation can support other nonstatic expressions if it wants to. Expressions of type Address are hardly ever static, but their value might be known at compile time anyway in many cases. GNAT does indeed permit many additional cases of nonstatic expressions. In particular, if the type involved is elementary there are no restrictions (since in this case, holding a temporary copy of the initialization value, if one is present, is inexpensive). In addition, if there is no implicit or explicit initialization, then there are no restrictions. GNAT will reject only the case where all three of these conditions hold: * The type of the item is non-elementary (e.g., a record or array). * There is explicit or implicit initialization required for the object. Note that access values are always implicitly initialized. * The address value is nonstatic. Here GNAT is more permissive than the RM, and allows the address value to be the address of a previously declared stand-alone variable, as long as it does not itself have an address clause. :: Anchor : Some_Initialized_Type; Overlay : Some_Initialized_Type; for Overlay'Address use Anchor'Address; However, the prefix of the address clause cannot be an array component, or a component of a discriminated record. As noted above in section 22.h, address values are typically nonstatic. In particular the To_Address function, even if applied to a literal value, is a nonstatic function call. To avoid this minor annoyance, GNAT provides the implementation defined attribute 'To_Address. The following two expressions have identical values: .. index:: Attribute .. index:: To_Address .. code-block:: ada To_Address (16#1234_0000#) System'To_Address (16#1234_0000#); except that the second form is considered to be a static expression, and thus when used as an address clause value is always permitted. Additionally, GNAT treats as static an address clause that is an unchecked_conversion of a static integer value. This simplifies the porting of legacy code, and provides a portable equivalent to the GNAT attribute ``To_Address``. Another issue with address clauses is the interaction with alignment requirements. When an address clause is given for an object, the address value must be consistent with the alignment of the object (which is usually the same as the alignment of the type of the object). If an address clause is given that specifies an inappropriately aligned address value, then the program execution is erroneous. Since this source of erroneous behavior can have unfortunate effects on machines with strict alignment requirements, GNAT checks (at compile time if possible, generating a warning, or at execution time with a run-time check) that the alignment is appropriate. If the run-time check fails, then ``Program_Error`` is raised. This run-time check is suppressed if the GNAT check Alignment_Check is suppressed, or if ``pragma Restrictions (No_Elaboration_Code)`` is in effect. It is also suppressed by default on non-strict alignment machines (such as the x86). In some cases, GNAT does not support an address specification (using either form of aspect specification syntax) for the declaration of an object that has an indefinite nominal subtype. An object declaration has an indefinite nominal subtype if it takes its bounds (for an array type), discriminant values (for a discriminated type whose discriminants lack defaults), or tag (for a class-wide type) from its initial value, as in .. code-block:: ada X : String := Some_Function_Call; -- String has no constraint, so bounds for X come from function call This restriction does not apply if the size of the object's initial value is known at compile time and the type of the object is not class-wide. .. index:: Export An address clause cannot be given for an exported object. More understandably the real restriction is that objects with an address clause cannot be exported. This is because such variables are not defined by the Ada program, so there is no external object to export. .. index:: Import It is permissible to give an address clause and a pragma Import for the same object. In this case, the variable is not really defined by the Ada program, so there is no external symbol to be linked. The link name and the external name are ignored in this case. The reason that we allow this combination is that it provides a useful idiom to avoid unwanted initializations on objects with address clauses. When an address clause is given for an object that has implicit or explicit initialization, then by default initialization takes place. This means that the effect of the object declaration is to overwrite the memory at the specified address. This is almost always not what the programmer wants, so GNAT will output a warning: :: with System; package G is type R is record M : Integer := 0; end record; Ext : R; for Ext'Address use System'To_Address (16#1234_1234#); | >>> warning: implicit initialization of "Ext" may modify overlaid storage >>> warning: use pragma Import for "Ext" to suppress initialization (RM B(24)) end G; As indicated by the warning message, the solution is to use a (dummy) pragma Import to suppress this initialization. The pragma tell the compiler that the object is declared and initialized elsewhere. The following package compiles without warnings (and the initialization is suppressed): .. code-block:: ada with System; package G is type R is record M : Integer := 0; end record; Ext : R; for Ext'Address use System'To_Address (16#1234_1234#); pragma Import (Ada, Ext); end G; A final issue with address clauses involves their use for overlaying variables, as in the following example: .. index:: Overlaying of objects .. code-block:: ada A : Integer; B : Integer; for B'Address use A'Address; or alternatively, using the form recommended by the RM: .. code-block:: ada A : Integer; Addr : constant Address := A'Address; B : Integer; for B'Address use Addr; In both of these cases, ``A`` and ``B`` become aliased to one another via the address clause. This use of address clauses to overlay variables, achieving an effect similar to unchecked conversion was erroneous in Ada 83, but in Ada 95 and Ada 2005 the effect is implementation defined. Furthermore, the Ada RM specifically recommends that in a situation like this, ``B`` should be subject to the following implementation advice (RM 13.3(19)): "19 If the Address of an object is specified, or it is imported or exported, then the implementation should not perform optimizations based on assumptions of no aliases." GNAT follows this recommendation, and goes further by also applying this recommendation to the overlaid variable (``A`` in the above example) in this case. This means that the overlay works "as expected", in that a modification to one of the variables will affect the value of the other. More generally, GNAT interprets this recommendation conservatively for address clauses: in the cases other than overlays, it considers that the object is effectively subject to pragma ``Volatile`` and implements the associated semantics. Note that when address clause overlays are used in this way, there is an issue of unintentional initialization, as shown by this example: :: package Overwrite_Record is type R is record A : Character := 'C'; B : Character := 'A'; end record; X : Short_Integer := 3; Y : R; for Y'Address use X'Address; | >>> warning: default initialization of "Y" may modify "X", use pragma Import for "Y" to suppress initialization (RM B.1(24)) end Overwrite_Record; Here the default initialization of ``Y`` will clobber the value of ``X``, which justifies the warning. The warning notes that this effect can be eliminated by adding a ``pragma Import`` which suppresses the initialization: .. code-block:: ada package Overwrite_Record is type R is record A : Character := 'C'; B : Character := 'A'; end record; X : Short_Integer := 3; Y : R; for Y'Address use X'Address; pragma Import (Ada, Y); end Overwrite_Record; Note that the use of ``pragma Initialize_Scalars`` may cause variables to be initialized when they would not otherwise have been in the absence of the use of this pragma. This may cause an overlay to have this unintended clobbering effect. The compiler avoids this for scalar types, but not for composite objects (where in general the effect of ``Initialize_Scalars`` is part of the initialization routine for the composite object): :: pragma Initialize_Scalars; with Ada.Text_IO; use Ada.Text_IO; procedure Overwrite_Array is type Arr is array (1 .. 5) of Integer; X : Arr := (others => 1); A : Arr; for A'Address use X'Address; | >>> warning: default initialization of "A" may modify "X", use pragma Import for "A" to suppress initialization (RM B.1(24)) begin if X /= Arr'(others => 1) then Put_Line ("X was clobbered"); else Put_Line ("X was not clobbered"); end if; end Overwrite_Array; The above program generates the warning as shown, and at execution time, prints ``X was clobbered``. If the ``pragma Import`` is added as suggested: .. code-block:: ada pragma Initialize_Scalars; with Ada.Text_IO; use Ada.Text_IO; procedure Overwrite_Array is type Arr is array (1 .. 5) of Integer; X : Arr := (others => 1); A : Arr; for A'Address use X'Address; pragma Import (Ada, A); begin if X /= Arr'(others => 1) then Put_Line ("X was clobbered"); else Put_Line ("X was not clobbered"); end if; end Overwrite_Array; then the program compiles without the warning and when run will generate the output ``X was not clobbered``. .. _Use_of_Address_Clauses_for_Memory-Mapped_I/O: Use of Address Clauses for Memory-Mapped I/O ============================================ .. index:: Memory-mapped I/O A common pattern is to use an address clause to map an atomic variable to a location in memory that corresponds to a memory-mapped I/O operation or operations, for example: .. code-block:: ada type Mem_Word is record A,B,C,D : Byte; end record; pragma Atomic (Mem_Word); for Mem_Word_Size use 32; Mem : Mem_Word; for Mem'Address use some-address; ... Temp := Mem; Temp.A := 32; Mem := Temp; For a full access (reference or modification) of the variable (Mem) in this case, as in the above examples, GNAT guarantees that the entire atomic word will be accessed, in accordance with the RM C.6(15) clause. A problem arises with a component access such as: .. code-block:: ada Mem.A := 32; Note that the component A is not declared as atomic. This means that it is not clear what this assignment means. It could correspond to full word read and write as given in the first example, or on architectures that supported such an operation it might be a single byte store instruction. The RM does not have anything to say in this situation, and GNAT does not make any guarantee. The code generated may vary from target to target. GNAT will issue a warning in such a case: :: Mem.A := 32; | >>> warning: access to non-atomic component of atomic array, may cause unexpected accesses to atomic object It is best to be explicit in this situation, by either declaring the components to be atomic if you want the byte store, or explicitly writing the full word access sequence if that is what the hardware requires. Alternatively, if the full word access sequence is required, GNAT also provides the pragma ``Volatile_Full_Access`` which can be used in lieu of pragma ``Atomic`` and will give the additional guarantee. .. _Effect_of_Convention_on_Representation: Effect of Convention on Representation ====================================== .. index:: Convention, effect on representation Normally the specification of a foreign language convention for a type or an object has no effect on the chosen representation. In particular, the representation chosen for data in GNAT generally meets the standard system conventions, and for example records are laid out in a manner that is consistent with C. This means that specifying convention C (for example) has no effect. There are three exceptions to this general rule: * *Convention Fortran and array subtypes*. If pragma Convention Fortran is specified for an array subtype, then in accordance with the implementation advice in section 3.6.2(11) of the Ada Reference Manual, the array will be stored in a Fortran-compatible column-major manner, instead of the normal default row-major order. * *Convention C and enumeration types* GNAT normally stores enumeration types in 8, 16, or 32 bits as required to accommodate all values of the type. For example, for the enumeration type declared by: :: type Color is (Red, Green, Blue); 8 bits is sufficient to store all values of the type, so by default, objects of type ``Color`` will be represented using 8 bits. However, normal C convention is to use 32 bits for all enum values in C, since enum values are essentially of type int. If pragma ``Convention C`` is specified for an Ada enumeration type, then the size is modified as necessary (usually to 32 bits) to be consistent with the C convention for enum values. Note that this treatment applies only to types. If Convention C is given for an enumeration object, where the enumeration type is not Convention C, then Object_Size bits are allocated. For example, for a normal enumeration type, with less than 256 elements, only 8 bits will be allocated for the object. Since this may be a surprise in terms of what C expects, GNAT will issue a warning in this situation. The warning can be suppressed by giving an explicit size clause specifying the desired size. * *Convention C/Fortran and Boolean types* In C, the usual convention for boolean values, that is values used for conditions, is that zero represents false, and nonzero values represent true. In Ada, the normal convention is that two specific values, typically 0/1, are used to represent false/true respectively. Fortran has a similar convention for ``LOGICAL`` values (any nonzero value represents true). To accommodate the Fortran and C conventions, if a pragma Convention specifies C or Fortran convention for a derived Boolean, as in the following example: :: type C_Switch is new Boolean; pragma Convention (C, C_Switch); then the GNAT generated code will treat any nonzero value as true. For truth values generated by GNAT, the conventional value 1 will be used for True, but when one of these values is read, any nonzero value is treated as True. .. _Conventions_and_Anonymous_Access_Types: Conventions and Anonymous Access Types ====================================== .. index:: Anonymous access types .. index:: Convention for anonymous access types The RM is not entirely clear on convention handling in a number of cases, and in particular, it is not clear on the convention to be given to anonymous access types in general, and in particular what is to be done for the case of anonymous access-to-subprogram. In GNAT, we decide that if an explicit Convention is applied to an object or component, and its type is such an anonymous type, then the convention will apply to this anonymous type as well. This seems to make sense since it is anomalous in any case to have a different convention for an object and its type, and there is clearly no way to explicitly specify a convention for an anonymous type, since it doesn't have a name to specify! Furthermore, we decide that if a convention is applied to a record type, then this convention is inherited by any of its components that are of an anonymous access type which do not have an explicitly specified convention. The following program shows these conventions in action: :: package ConvComp is type Foo is range 1 .. 10; type T1 is record A : access function (X : Foo) return Integer; B : Integer; end record; pragma Convention (C, T1); type T2 is record A : access function (X : Foo) return Integer; pragma Convention (C, A); B : Integer; end record; pragma Convention (COBOL, T2); type T3 is record A : access function (X : Foo) return Integer; pragma Convention (COBOL, A); B : Integer; end record; pragma Convention (C, T3); type T4 is record A : access function (X : Foo) return Integer; B : Integer; end record; pragma Convention (COBOL, T4); function F (X : Foo) return Integer; pragma Convention (C, F); function F (X : Foo) return Integer is (13); TV1 : T1 := (F'Access, 12); -- OK TV2 : T2 := (F'Access, 13); -- OK TV3 : T3 := (F'Access, 13); -- ERROR | >>> subprogram "F" has wrong convention >>> does not match access to subprogram declared at line 17 38. TV4 : T4 := (F'Access, 13); -- ERROR | >>> subprogram "F" has wrong convention >>> does not match access to subprogram declared at line 24 39. end ConvComp; .. _Determining_the_Representations_chosen_by_GNAT: Determining the Representations chosen by GNAT ============================================== .. index:: Representation, determination of .. index:: -gnatR (gcc) Although the descriptions in this section are intended to be complete, it is often easier to simply experiment to see what GNAT accepts and what the effect is on the layout of types and objects. As required by the Ada RM, if a representation clause is not accepted, then it must be rejected as illegal by the compiler. However, when a representation clause or pragma is accepted, there can still be questions of what the compiler actually does. For example, if a partial record representation clause specifies the location of some components and not others, then where are the non-specified components placed? Or if pragma ``Pack`` is used on a record, then exactly where are the resulting fields placed? The section on pragma ``Pack`` in this chapter can be used to answer the second question, but it is often easier to just see what the compiler does. For this purpose, GNAT provides the option *-gnatR*. If you compile with this option, then the compiler will output information on the actual representations chosen, in a format similar to source representation clauses. For example, if we compile the package: .. code-block:: ada package q is type r (x : boolean) is tagged record case x is when True => S : String (1 .. 100); when False => null; end case; end record; type r2 is new r (false) with record y2 : integer; end record; for r2 use record y2 at 16 range 0 .. 31; end record; type x is record y : character; end record; type x1 is array (1 .. 10) of x; for x1'component_size use 11; type ia is access integer; type Rb1 is array (1 .. 13) of Boolean; pragma Pack (rb1); type Rb2 is array (1 .. 65) of Boolean; pragma Pack (rb2); type x2 is record l1 : Boolean; l2 : Duration; l3 : Float; l4 : Boolean; l5 : Rb1; l6 : Rb2; end record; pragma Pack (x2); end q; using the switch *-gnatR* we obtain the following output: .. code-block:: ada Representation information for unit q ------------------------------------- for r'Size use ??; for r'Alignment use 4; for r use record x at 4 range 0 .. 7; _tag at 0 range 0 .. 31; s at 5 range 0 .. 799; end record; for r2'Size use 160; for r2'Alignment use 4; for r2 use record x at 4 range 0 .. 7; _tag at 0 range 0 .. 31; _parent at 0 range 0 .. 63; y2 at 16 range 0 .. 31; end record; for x'Size use 8; for x'Alignment use 1; for x use record y at 0 range 0 .. 7; end record; for x1'Size use 112; for x1'Alignment use 1; for x1'Component_Size use 11; for rb1'Size use 13; for rb1'Alignment use 2; for rb1'Component_Size use 1; for rb2'Size use 72; for rb2'Alignment use 1; for rb2'Component_Size use 1; for x2'Size use 224; for x2'Alignment use 4; for x2 use record l1 at 0 range 0 .. 0; l2 at 0 range 1 .. 64; l3 at 12 range 0 .. 31; l4 at 16 range 0 .. 0; l5 at 16 range 1 .. 13; l6 at 18 range 0 .. 71; end record; The Size values are actually the Object_Size, i.e., the default size that will be allocated for objects of the type. The ``??`` size for type r indicates that we have a variant record, and the actual size of objects will depend on the discriminant value. The Alignment values show the actual alignment chosen by the compiler for each record or array type. The record representation clause for type r shows where all fields are placed, including the compiler generated tag field (whose location cannot be controlled by the programmer). The record representation clause for the type extension r2 shows all the fields present, including the parent field, which is a copy of the fields of the parent type of r2, i.e., r1. The component size and size clauses for types rb1 and rb2 show the exact effect of pragma ``Pack`` on these arrays, and the record representation clause for type x2 shows how pragma `Pack` affects this record type. In some cases, it may be useful to cut and paste the representation clauses generated by the compiler into the original source to fix and guarantee the actual representation to be used. --- ### Gcc Ada Doc Gnat Rm Security Hardening Features (`gcc/ada/doc/gnat_rm/security_hardening_features.rst`) .. _Security_Hardening_Features: *************************** Security Hardening Features *************************** This chapter describes Ada extensions aimed at security hardening that are provided by GNAT. The features in this chapter are currently experimental and subject to change. These features are supported only by the GCC back end, not by LLVM. .. Register Scrubbing: Register Scrubbing ================== GNAT can generate code to zero-out hardware registers before returning from a subprogram. It can be enabled with the :switch:`-fzero-call-used-regs={choice}` command-line option, to affect all subprograms in a compilation, and with a :samp:`Machine_Attribute` pragma, to affect only specific subprograms. .. code-block:: ada procedure Foo; pragma Machine_Attribute (Foo, "zero_call_used_regs", "used"); -- Before returning, Foo scrubs only call-clobbered registers -- that it uses itself. function Bar return Integer; pragma Machine_Attribute (Bar, "zero_call_used_regs", "all"); -- Before returning, Bar scrubs all call-clobbered registers. function Baz return Integer; pragma Machine_Attribute (Bar, "zero_call_used_regs", "leafy"); -- Before returning, Bar scrubs call-clobbered registers, either -- those it uses itself, if it can be identified as a leaf -- function, or all of them otherwise. For usage and more details on the command-line option, on the ``zero_call_used_regs`` attribute, and on their use with other programming languages, see :title:`Using the GNU Compiler Collection (GCC)`. .. Stack Scrubbing: Stack Scrubbing =============== GNAT can generate code to zero-out stack frames used by subprograms. It can be activated with the :samp:`Machine_Attribute` pragma, on specific subprograms and variables, or their types. (This attribute always applies to a type, even when it is associated with a subprogram or a variable.) .. code-block:: ada function Foo returns Integer; pragma Machine_Attribute (Foo, "strub"); -- Foo and its callers are modified so as to scrub the stack -- space used by Foo after it returns. Shorthand for: -- pragma Machine_Attribute (Foo, "strub", "at-calls"); procedure Bar; pragma Machine_Attribute (Bar, "strub", "internal"); -- Bar is turned into a wrapper for its original body, -- and they scrub the stack used by the original body. Var : Integer; pragma Machine_Attribute (Var, "strub"); -- Reading from Var in a subprogram enables stack scrubbing -- of the stack space used by the subprogram. Furthermore, if -- Var is declared within a subprogram, this also enables -- scrubbing of the stack space used by that subprogram. Given these declarations, Foo has its type and body modified as follows: .. code-block:: ada function Foo ( : in out System.Address) returns Integer is -- ... begin <__strub_update> (); -- Updates the stack WaterMark. -- ... end; whereas its callers are modified from: .. code-block:: ada X := Foo; to: .. code-block:: ada declare : System.Address; begin <__strub_enter> (); -- Initialize . X := Foo (); <__strub_leave> (); -- Scrubs stack up to . end; As for Bar, because it is strubbed in internal mode, its callers are not modified. Its definition is modified roughly as follows: .. code-block:: ada procedure Bar is : System.Address; procedure Strubbed_Bar ( : in out System.Address) is begin <__strub_update> (); -- Updates the stack WaterMark. -- original Bar body. end Strubbed_Bar; begin <__strub_enter> (); -- Initialize . Strubbed_Bar (); <__strub_leave> (); -- Scrubs stack up to . end Bar; There are also :switch:`-fstrub={choice}` command-line options to control default settings. For usage and more details on the command-line options, on the ``strub`` attribute, and their use with other programming languages, see :title:`Using the GNU Compiler Collection (GCC)`. Note that Ada secondary stacks are not scrubbed. The restriction ``No_Secondary_Stack`` avoids their use, and thus their accidental preservation of data that should be scrubbed. Attributes ``Access`` and ``Unconstrained_Access`` of variables and constants with ``strub`` enabled require types with ``strub`` enabled; there is no way to express an access-to-strub type otherwise. ``Unchecked_Access`` bypasses this constraint, but the resulting access type designates a non-strub type. .. code-block:: ada VI : aliased Integer; pragma Machine_Attribute (VI, "strub"); XsVI : access Integer := VI'Access; -- Error. UXsVI : access Integer := VI'Unchecked_Access; -- OK, -- UXsVI does *not* enable strub in subprograms that -- dereference it to obtain the UXsVI.all value. type Strub_Int is new Integer; pragma Machine_Attribute (Strub_Int, "strub"); VSI : aliased Strub_Int; XsVSI : access Strub_Int := VSI'Access; -- OK, -- VSI and XsVSI.all both enable strub in subprograms that -- read their values. Every access-to-subprogram type, renaming, and overriding and overridden dispatching operations that may refer to a subprogram with an attribute-modified interface must be annotated with the same interface-modifying attribute. Access-to-subprogram types can be explicitly converted to different strub modes, as long as they are interface-compatible (i.e., adding or removing ``at-calls`` is not allowed). For example, a ``strub``-``disabled`` subprogram can be turned ``callable`` through such an explicit conversion: .. code-block:: ada type TBar is access procedure; type TBar_Callable is access procedure; pragma Machine_Attribute (TBar_Callable, "strub", "callable"); -- The attribute modifies the procedure type, rather than the -- access type, because of the extra argument after "strub", -- only applicable to subprogram types. Bar_Callable_Ptr : constant TBar_Callable := TBar_Callable (TBar'(Bar'Access)); procedure Bar_Callable renames Bar_Callable_Ptr.all; pragma Machine_Attribute (Bar_Callable, "strub", "callable"); Note that the renaming declaration is expanded to a full subprogram body, it won't be just an alias. Only if it is inlined will it be as efficient as a call by dereferencing the access-to-subprogram constant Bar_Callable_Ptr. .. Hardened Conditionals: Hardened Conditionals ===================== GNAT can harden conditionals to protect against control-flow attacks. This is accomplished by two complementary transformations, each activated by a separate command-line option. The option :switch:`-fharden-compares` enables hardening of compares that compute results stored in variables, adding verification that the reversed compare yields the opposite result, turning: .. code-block:: ada B := X = Y; into: .. code-block:: ada B := X = Y; declare NotB : Boolean := X /= Y; -- Computed independently of B. begin if B = NotB then <__builtin_trap>; end if; end; The option :switch:`-fharden-conditional-branches` enables hardening of compares that guard conditional branches, adding verification of the reversed compare to both execution paths, turning: .. code-block:: ada if X = Y then X := Z + 1; else Y := Z - 1; end if; into: .. code-block:: ada if X = Y then if X /= Y then -- Computed independently of X = Y. <__builtin_trap>; end if; X := Z + 1; else if X /= Y then -- Computed independently of X = Y. null; else <__builtin_trap>; end if; Y := Z - 1; end if; These transformations are introduced late in the compilation pipeline, long after boolean expressions are decomposed into separate compares, each one turned into either a conditional branch or a compare whose result is stored in a boolean variable or temporary. Compiler optimizations, if enabled, may also turn conditional branches into stored compares, and vice-versa, or into operations with implied conditionals (e.g. MIN and MAX). Conditionals may also be optimized out entirely, if their value can be determined at compile time, and occasionally multiple compares can be combined into one. It is thus difficult to predict which of these two options will affect a specific compare operation expressed in source code. Using both options ensures that every compare that is neither optimized out nor optimized into implied conditionals will be hardened. The addition of reversed compares can be observed by enabling the dump files of the corresponding passes, through command-line options :switch:`-fdump-tree-hardcmp` and :switch:`-fdump-tree-hardcbr`, respectively. They are separate options, however, because of the significantly different performance impact of the hardening transformations. For usage and more details on the command-line options, see :title:`Using the GNU Compiler Collection (GCC)`. These options can be used with other programming languages supported by GCC. .. Hardened Booleans: Hardened Booleans ================= Ada has built-in support for introducing boolean types with alternative representations, using representation clauses: .. code-block:: ada type HBool is new Boolean; for HBool use (16#5a#, 16#a5#); for HBool'Size use 8; When validity checking is enabled, the compiler will check that variables of such types hold values corresponding to the selected representations. There are multiple strategies for where to introduce validity checking (see :switch:`-gnatV` options). Their goal is to guard against various kinds of programming errors, and GNAT strives to omit checks when program logic rules out an invalid value, and optimizers may further remove checks found to be redundant. For additional hardening, the ``hardbool`` :samp:`Machine_Attribute` pragma can be used to annotate boolean types with representation clauses, so that expressions of such types used as conditions are checked even when compiling with :switch:`-gnatVT`: .. code-block:: ada pragma Machine_Attribute (HBool, "hardbool"); function To_Boolean (X : HBool) returns Boolean is (Boolean (X)); is compiled roughly like: .. code-block:: ada function To_Boolean (X : HBool) returns Boolean is begin if X not in True | False then raise Constraint_Error; elsif X in True then return True; else return False; end if; end To_Boolean; Note that :switch:`-gnatVn` will disable even ``hardbool`` testing. Analogous behavior is available as a GCC extension to the C and Objective C programming languages, through the ``hardbool`` attribute, with the difference that, instead of raising a Constraint_Error exception, when a hardened boolean variable is found to hold a value that stands for neither True nor False, the program traps. For usage and more details on that attribute, see :title:`Using the GNU Compiler Collection (GCC)`. .. Control Flow Redundancy: Control Flow Redundancy ======================= GNAT can guard against unexpected execution flows, such as branching into the middle of subprograms, as in Return Oriented Programming exploits. In units compiled with :switch:`-fharden-control-flow-redundancy`, subprograms are instrumented so that, every time they are called, basic blocks take note as control flows through them, and, before returning, subprograms verify that the taken notes are consistent with the control-flow graph. The performance impact of verification on leaf subprograms can be much higher, while the averted risks are much lower on them. Instrumentation can be disabled for leaf subprograms with :switch:`-fhardcfr-skip-leaf`. Functions with too many basic blocks, or with multiple return points, call a run-time function to perform the verification. Other functions perform the verification inline before returning. Optimizing the inlined verification can be quite time consuming, so the default upper limit for the inline mode is set at 16 blocks. Command-line option :switch:`--param hardcfr-max-inline-blocks=` can override it. Even though typically sparse control-flow graphs exhibit run-time verification time nearly proportional to the block count of a subprogram, it may become very significant for generated subprograms with thousands of blocks. Command-line option :switch:`--param hardcfr-max-blocks=` can set an upper limit for instrumentation. For each block that is marked as visited, the mechanism checks that at least one of its predecessors, and at least one of its successors, are also marked as visited. Verification is performed just before a subprogram returns. The following fragment: .. code-block:: ada if X then Y := F (Z); return; end if; gets turned into: .. code-block:: ada type Visited_Bitmap is array (1..N) of Boolean with Pack; Visited : aliased Visited_Bitmap := (others => False); -- Bitmap of visited blocks. N is the basic block count. [...] -- Basic block #I Visited(I) := True; if X then -- Basic block #J Visited(J) := True; Y := F (Z); CFR.Check (N, Visited'Access, CFG'Access); -- CFR is a hypothetical package whose Check procedure calls -- libgcc's __hardcfr_check, that traps if the Visited bitmap -- does not hold a valid path in CFG, the run-time -- representation of the control flow graph in the enclosing -- subprogram. return; end if; -- Basic block #K Visited(K) := True; Verification would also be performed before tail calls, if any front-ends marked them as mandatory or desirable, but none do. Regular calls are optimized into tail calls too late for this transformation to act on it. In order to avoid adding verification after potential tail calls, which would prevent tail-call optimization, we recognize returning calls, i.e., calls whose result, if any, is returned by the calling subprogram to its caller immediately after the call returns. Verification is performed before such calls, whether or not they are ultimately optimized to tail calls. This behavior is enabled by default whenever sibcall optimization is enabled (see :switch:`-foptimize-sibling-calls`); it may be disabled with :switch:`-fno-hardcfr-check-returning-calls`, or enabled with :switch:`-fhardcfr-check-returning-calls`, regardless of the optimization, but the lack of other optimizations may prevent calls from being recognized as returning calls: .. code-block:: ada -- CFR.Check here, with -fhardcfr-check-returning-calls. P (X); -- CFR.Check here, with -fno-hardcfr-check-returning-calls. return; or: .. code-block:: ada -- CFR.Check here, with -fhardcfr-check-returning-calls. R := F (X); -- CFR.Check here, with -fno-hardcfr-check-returning-calls. return R; Any subprogram from which an exception may escape, i.e., that may raise or propagate an exception that isn't handled internally, is conceptually enclosed by a cleanup handler that performs verification, unless this is disabled with :switch:`-fno-hardcfr-check-exceptions`. With this feature enabled, a subprogram body containing: .. code-block:: ada -- ... Y := F (X); -- May raise exceptions. -- ... raise E; -- Not handled internally. -- ... gets modified as follows: .. code-block:: ada begin -- ... Y := F (X); -- May raise exceptions. -- ... raise E; -- Not handled internally. -- ... exception when others => CFR.Check (N, Visited'Access, CFG'Access); raise; end; Verification may also be performed before No_Return calls, whether all of them, with :switch:`-fhardcfr-check-noreturn-calls=always`; all but internal subprograms involved in exception-raising or -reraising or subprograms explicitly marked with both :samp:`No_Return` and :samp:`Machine_Attribute` :samp:`expected_throw` pragmas, with :switch:`-fhardcfr-check-noreturn-calls=no-xthrow` (default); only nothrow ones, with :switch:`-fhardcfr-check-noreturn-calls=nothrow`; or none, with :switch:`-fhardcfr-check-noreturn-calls=never`. When a No_Return call returns control to its caller through an exception, verification may have already been performed before the call, if :switch:`-fhardcfr-check-noreturn-calls=always` or :switch:`-fhardcfr-check-noreturn-calls=no-xthrow` is in effect. The compiler arranges for already-checked No_Return calls without a preexisting handler to bypass the implicitly-added cleanup handler and thus the redundant check, but a local exception or cleanup handler, if present, will modify the set of visited blocks, and checking will take place again when the caller reaches the next verification point, whether it is a return or reraise statement after the exception is otherwise handled, or even another No_Return call. The instrumentation for hardening with control flow redundancy can be observed in dump files generated by the command-line option :switch:`-fdump-tree-hardcfr`. For more details on the control flow redundancy command-line options, see :title:`Using the GNU Compiler Collection (GCC)`. These options can be used with other programming languages supported by GCC. --- ### Gcc Ada Doc Gnat Rm Specialized Needs Annexes (`gcc/ada/doc/gnat_rm/specialized_needs_annexes.rst`) .. _Specialized_Needs_Annexes: ************************* Specialized Needs Annexes ************************* Ada 95, Ada 2005, Ada 2012, and Ada 2022 define a number of Specialized Needs Annexes, which are not required in all implementations. However, as described in this chapter, GNAT implements all of these annexes: *Systems Programming (Annex C)* The Systems Programming Annex is fully implemented. *Real-Time Systems (Annex D)* The Real-Time Systems Annex is fully implemented. *Distributed Systems (Annex E)* Stub generation is fully implemented in the GNAT compiler. In addition, a complete compatible PCS is available as part of ``PolyORB``, a separate product. **NB!** See the note in :ref:`polyORB` regarding the lifetime of this product. *Information Systems (Annex F)* The Information Systems annex is fully implemented. *Numerics (Annex G)* The Numerics Annex is fully implemented. *Safety and Security / High-Integrity Systems (Annex H)* The Safety and Security Annex (termed the High-Integrity Systems Annex in Ada 2005) is fully implemented. --- ### Gcc Ada Doc Gnat Rm Standard And Implementation Defined Restrictions (`gcc/ada/doc/gnat_rm/standard_and_implementation_defined_restrictions.rst`) .. _Standard_and_Implementation_Defined_Restrictions: ************************************************ Standard and Implementation Defined Restrictions ************************************************ All Ada Reference Manual-defined Restriction identifiers are implemented: * language-defined restrictions (see 13.12.1) * tasking restrictions (see D.7) * high integrity restrictions (see H.4) GNAT implements additional restriction identifiers. All restrictions, whether language defined or GNAT-specific, are listed in the following. .. _Partition-Wide_Restrictions: Partition-Wide Restrictions =========================== There are two separate lists of restriction identifiers. The first set requires consistency throughout a partition (in other words, if the restriction identifier is used for any compilation unit in the partition, then all compilation units in the partition must obey the restriction). Immediate_Reclamation --------------------- .. index:: Immediate_Reclamation [RM H.4] This restriction ensures that, except for storage occupied by objects created by allocators and not deallocated via unchecked deallocation, any storage reserved at run time for an object is immediately reclaimed when the object no longer exists. Max_Asynchronous_Select_Nesting ------------------------------- .. index:: Max_Asynchronous_Select_Nesting [RM D.7] Specifies the maximum dynamic nesting level of asynchronous selects. Violations of this restriction with a value of zero are detected at compile time. Violations of this restriction with values other than zero cause Storage_Error to be raised. Max_Entry_Queue_Length ---------------------- .. index:: Max_Entry_Queue_Length [RM D.7] This restriction is a declaration that any protected entry compiled in the scope of the restriction has at most the specified number of tasks waiting on the entry at any one time, and so no queue is required. Note that this restriction is checked at run time. Violation of this restriction results in the raising of Program_Error exception at the point of the call. .. index:: Max_Entry_Queue_Depth The restriction ``Max_Entry_Queue_Depth`` is recognized as a synonym for ``Max_Entry_Queue_Length``. This is retained for historical compatibility purposes (and a warning will be generated for its use if warnings on obsolescent features are activated). Max_Protected_Entries --------------------- .. index:: Max_Protected_Entries [RM D.7] Specifies the maximum number of entries per protected type. The bounds of every entry family of a protected unit shall be static, or shall be defined by a discriminant of a subtype whose corresponding bound is static. Max_Select_Alternatives ----------------------- .. index:: Max_Select_Alternatives [RM D.7] Specifies the maximum number of alternatives in a selective accept. Max_Storage_At_Blocking ----------------------- .. index:: Max_Storage_At_Blocking [RM D.7] Specifies the maximum portion (in storage elements) of a task's Storage_Size that can be retained by a blocked task. A violation of this restriction causes Storage_Error to be raised. Max_Task_Entries ---------------- .. index:: Max_Task_Entries [RM D.7] Specifies the maximum number of entries per task. The bounds of every entry family of a task unit shall be static, or shall be defined by a discriminant of a subtype whose corresponding bound is static. Max_Tasks --------- .. index:: Max_Tasks [RM D.7] Specifies the maximum number of tasks that may be created, not counting the creation of the environment task. Violations of this restriction with a value of zero are detected at compile time in all contexts. Violations of this restriction with a value greater than zero are detected for library-level tasks at compile time, but are not detected for local tasks at all. No_Abort_Statements ------------------- .. index:: No_Abort_Statements [RM D.7] There are no abort_statements, and there are no calls to Task_Identification.Abort_Task. No_Access_Parameter_Allocators ------------------------------ .. index:: No_Access_Parameter_Allocators [RM H.4] This restriction ensures at compile time that there are no occurrences of an allocator as the actual parameter to an access parameter. No_Access_Subprograms --------------------- .. index:: No_Access_Subprograms [RM H.4] This restriction ensures at compile time that there are no declarations of access-to-subprogram types. No_Allocators ------------- .. index:: No_Allocators [RM H.4] This restriction ensures at compile time that there are no occurrences of an allocator. No_Anonymous_Allocators ----------------------- .. index:: No_Anonymous_Allocators [RM H.4] This restriction ensures at compile time that there are no occurrences of an allocator of anonymous access type. No_Asynchronous_Control ----------------------- .. index:: No_Asynchronous_Control [RM J.13] This restriction ensures at compile time that there are no semantic dependences on the predefined package Asynchronous_Task_Control. No_Calendar ----------- .. index:: No_Calendar [GNAT] This restriction ensures at compile time that there are no semantic dependences on package Calendar. No_Coextensions --------------- .. index:: No_Coextensions [RM H.4] This restriction ensures at compile time that there are no coextensions. See 3.10.2. No_Default_Initialization ------------------------- .. index:: No_Default_Initialization [GNAT] This restriction prohibits any instance of default initialization of variables or components. The binder implements a consistency check that prevents any unit without the restriction from with'ing a unit with the restriction (this allows the generation of initialization procedures to be skipped, since you can be sure that no call is ever generated to an initialization procedure in a unit with the restriction active). If used in conjunction with Initialize_Scalars or Normalize_Scalars, the effect is to prohibit all cases of variables declared without a specific initializer (including the case of OUT scalar parameters). No_Delay -------- .. index:: No_Delay [RM H.4] This restriction ensures at compile time that there are no delay statements and no semantic dependences on package Calendar. No_Dependence ------------- .. index:: No_Dependence [RM 13.12.1] This restriction ensures at compile time that there are no dependences on a library unit. For GNAT, this includes implicit implementation dependences on units of the runtime library that are created by the compiler to support specific constructs of the language. Here are some examples: * ``System.Arith_64``: 64-bit arithmetics for 32-bit platforms, * ``System.Arith_128``: 128-bit arithmetics for 64-bit platforms, * ``System.Memory``: heap memory allocation routines, * ``System.Memory_Compare``: memory comparison routine (aka ``memcmp`` for C), * ``System.Memory_Copy``: memory copy routine (aka ``memcpy`` for C), * ``System.Memory_Move``: memory move routine (aka ``memmove`` for C), * ``System.Memory_Set``: memory set routine (aka ``memset`` for C), * ``System.Stack_Checking[.Operations]``: stack checking without MMU, * ``System.GCC``: support routines from the GCC library. No_Direct_Boolean_Operators --------------------------- .. index:: No_Direct_Boolean_Operators [GNAT] This restriction ensures that no logical operators (and/or/xor) are used on operands of type Boolean (or any type derived from Boolean). This is intended for use in safety critical programs where the certification protocol requires the use of short-circuit (and then, or else) forms for all composite boolean operations. No_Dispatch ----------- .. index:: No_Dispatch [RM H.4] This restriction ensures at compile time that there are no occurrences of ``T'Class``, for any (tagged) subtype ``T``. No_Dispatching_Calls -------------------- .. index:: No_Dispatching_Calls [GNAT] This restriction ensures at compile time that the code generated by the compiler involves no dispatching calls. The use of this restriction allows the safe use of record extensions, classwide membership tests and other classwide features not involving implicit dispatching. This restriction ensures that the code contains no indirect calls through a dispatching mechanism. Note that this includes internally-generated calls created by the compiler, for example in the implementation of class-wide objects assignments. The membership test is allowed in the presence of this restriction, because its implementation requires no dispatching. This restriction is comparable to the official Ada restriction ``No_Dispatch`` except that it is a bit less restrictive in that it allows all classwide constructs that do not imply dispatching. The following example indicates constructs that violate this restriction. .. code-block:: ada package Pkg is type T is tagged record Data : Natural; end record; procedure P (X : T); type DT is new T with record More_Data : Natural; end record; procedure Q (X : DT); end Pkg; with Pkg; use Pkg; procedure Example is procedure Test (O : T'Class) is N : Natural := O'Size; -- Error: Dispatching call C : T'Class := O; -- Error: implicit Dispatching Call begin if O in DT'Class then -- OK : Membership test Q (DT (O)); -- OK : Type conversion plus direct call else P (O); -- Error: Dispatching call end if; end Test; Obj : DT; begin P (Obj); -- OK : Direct call P (T (Obj)); -- OK : Type conversion plus direct call P (T'Class (Obj)); -- Error: Dispatching call Test (Obj); -- OK : Type conversion if Obj in T'Class then -- OK : Membership test null; end if; end Example; No_Dynamic_Attachment --------------------- .. index:: No_Dynamic_Attachment [RM D.7] This restriction ensures that there is no call to any of the operations defined in package Ada.Interrupts (Is_Reserved, Is_Attached, Current_Handler, Attach_Handler, Exchange_Handler, Detach_Handler, and Reference). .. index:: No_Dynamic_Interrupts The restriction ``No_Dynamic_Interrupts`` is recognized as a synonym for ``No_Dynamic_Attachment``. This is retained for historical compatibility purposes (and a warning will be generated for its use if warnings on obsolescent features are activated). No_Dynamic_Priorities --------------------- .. index:: No_Dynamic_Priorities [RM D.7] There are no semantic dependencies on the package Dynamic_Priorities. No_Entry_Calls_In_Elaboration_Code ---------------------------------- .. index:: No_Entry_Calls_In_Elaboration_Code [GNAT] This restriction ensures at compile time that no task or protected entry calls are made during elaboration code. As a result of the use of this restriction, the compiler can assume that no code past an accept statement in a task can be executed at elaboration time. No_Enumeration_Maps ------------------- .. index:: No_Enumeration_Maps [GNAT] This restriction ensures at compile time that no operations requiring enumeration maps are used (that is Image and Value attributes applied to enumeration types). No_Exception_Handlers --------------------- .. index:: No_Exception_Handlers [GNAT] This restriction ensures at compile time that there are no explicit exception handlers. It also indicates that no exception propagation will be provided. In this mode, exceptions may be raised but will result in an immediate call to the last chance handler, a routine that the user must define with the following profile: .. code-block:: ada procedure Last_Chance_Handler (Source_Location : System.Address; Line : Integer); pragma Export (C, Last_Chance_Handler, "__gnat_last_chance_handler"); The ``Source_Location`` parameter is a C null-terminated string representing a message to be associated with the exception (typically the source location of the raise statement generated by the compiler). The Line parameter when nonzero represents the line number in the source program where the raise occurs. No_Exception_Propagation ------------------------ .. index:: No_Exception_Propagation [GNAT] This restriction guarantees that exceptions are never propagated to an outer subprogram scope. The only case in which an exception may be raised is when the handler is statically in the same subprogram, so that the effect of a raise is essentially like a goto statement. Any other raise statement (implicit or explicit) will be considered unhandled. Exception handlers are allowed, but may not contain an exception occurrence identifier (exception choice). In addition, use of the package GNAT.Current_Exception is not permitted, and reraise statements (raise with no operand) are not permitted. No_Exception_Registration ------------------------- .. index:: No_Exception_Registration [GNAT] This restriction ensures at compile time that no stream operations for types Exception_Id or Exception_Occurrence are used. This also makes it impossible to pass exceptions to or from a partition with this restriction in a distributed environment. If this restriction is active, the generated code is simplified by omitting the otherwise-required global registration of exceptions when they are declared. No_Exceptions ------------- .. index:: No_Exceptions [RM H.4] This restriction ensures at compile time that there are no raise statements and no exception handlers and also suppresses the generation of language-defined run-time checks. No_Finalization --------------- .. index:: No_Finalization [GNAT] This restriction disables the language features described in chapter 7.6 of the Ada 2005 RM as well as all form of code generation performed by the compiler to support these features. The following types are no longer considered controlled when this restriction is in effect: * ``Ada.Finalization.Controlled`` * ``Ada.Finalization.Limited_Controlled`` * Derivations from ``Controlled`` or ``Limited_Controlled`` * Class-wide types * Protected types * Task types * Array and record types with controlled components The compiler no longer generates code to initialize, finalize or adjust an object or a nested component, either declared on the stack or on the heap. The deallocation of a controlled object no longer finalizes its contents. No_Fixed_Point -------------- .. index:: No_Fixed_Point [RM H.4] This restriction ensures at compile time that there are no occurrences of fixed point types and operations. No_Floating_Point ----------------- .. index:: No_Floating_Point [RM H.4] This restriction ensures at compile time that there are no occurrences of floating point types and operations. No_Implicit_Conditionals ------------------------ .. index:: No_Implicit_Conditionals [GNAT] This restriction ensures that the generated code does not contain any implicit conditionals, either by modifying the generated code where possible, or by rejecting any construct that would otherwise generate an implicit conditional. Note that this check does not include run time constraint checks, which on some targets may generate implicit conditionals as well. To control the latter, constraint checks can be suppressed in the normal manner. Constructs generating implicit conditionals include comparisons of composite objects and the Max/Min attributes. No_Implicit_Dynamic_Code ------------------------ .. index:: No_Implicit_Dynamic_Code .. index:: trampoline [GNAT] This restriction prevents the compiler from building 'trampolines'. This is a structure that is built on the stack and contains dynamic code to be executed at run time. On some targets, a trampoline is built for the following features: ``Access``, ``Unrestricted_Access``, or ``Address`` of a nested subprogram; nested task bodies; primitive operations of nested tagged types. Trampolines do not work on machines that prevent execution of stack data. For example, on windows systems, enabling DEP (data execution protection) will cause trampolines to raise an exception. Trampolines are also quite slow at run time. On many targets, trampolines have been largely eliminated. Look at the version of system.ads for your target --- if it has Always_Compatible_Rep equal to False, then trampolines are largely eliminated. In particular, a trampoline is built for the following features: ``Address`` of a nested subprogram; ``Access`` or ``Unrestricted_Access`` of a nested subprogram, but only if pragma Favor_Top_Level applies, or the access type has a foreign-language convention; primitive operations of nested tagged types. No_Implicit_Heap_Allocations ---------------------------- .. index:: No_Implicit_Heap_Allocations [RM D.7] No constructs are allowed to cause implicit heap allocation. No_Implicit_Protected_Object_Allocations ---------------------------------------- .. index:: No_Implicit_Protected_Object_Allocations [GNAT] No constructs are allowed to cause implicit heap allocation of a protected object. No_Implicit_Task_Allocations ---------------------------- .. index:: No_Implicit_Task_Allocations [GNAT] No constructs are allowed to cause implicit heap allocation of a task. No_Initialize_Scalars --------------------- .. index:: No_Initialize_Scalars [GNAT] This restriction ensures that no unit in the partition is compiled with pragma Initialize_Scalars. This allows the generation of more efficient code, and in particular eliminates dummy null initialization routines that are otherwise generated for some record and array types. No_IO ----- .. index:: No_IO [RM H.4] This restriction ensures at compile time that there are no dependences on any of the library units Sequential_IO, Direct_IO, Text_IO, Wide_Text_IO, Wide_Wide_Text_IO, or Stream_IO. No_Local_Allocators ------------------- .. index:: No_Local_Allocators [RM H.4] This restriction ensures at compile time that there are no occurrences of an allocator in subprograms, generic subprograms, tasks, and entry bodies. No_Local_Protected_Objects -------------------------- .. index:: No_Local_Protected_Objects [RM D.7] This restriction ensures at compile time that protected objects are only declared at the library level. No_Local_Tagged_Types --------------------- .. index:: No_Local_Tagged_Types [GNAT] This restriction ensures at compile time that tagged types are only declared at the library level. No_Local_Timing_Events ---------------------- .. index:: No_Local_Timing_Events [RM D.7] All objects of type Ada.Real_Time.Timing_Events.Timing_Event are declared at the library level. No_Long_Long_Integers --------------------- .. index:: No_Long_Long_Integers [GNAT] This partition-wide restriction forbids any explicit reference to type Standard.Long_Long_Integer, and also forbids declaring range types whose implicit base type is Long_Long_Integer, and modular types whose size exceeds Long_Integer'Size. No_Multiple_Elaboration ----------------------- .. index:: No_Multiple_Elaboration [GNAT] When this restriction is active and the static elaboration model is used, and -fpreserve-control-flow is not used, the compiler is allowed to suppress the elaboration counter normally associated with the unit, even if the unit has elaboration code. This counter is typically used to check for access before elaboration and to control multiple elaboration attempts. If the restriction is used, then the situations in which multiple elaboration is possible, including non-Ada main programs and Stand Alone libraries, are not permitted and will be diagnosed by the binder. No_Nested_Finalization ---------------------- .. index:: No_Nested_Finalization [RM D.7] All objects requiring finalization are declared at the library level. No_Protected_Type_Allocators ---------------------------- .. index:: No_Protected_Type_Allocators [RM D.7] This restriction ensures at compile time that there are no allocator expressions that attempt to allocate protected objects. No_Protected_Types ------------------ .. index:: No_Protected_Types [RM H.4] This restriction ensures at compile time that there are no declarations of protected types or protected objects. No_Recursion ------------ .. index:: No_Recursion [RM H.4] A program execution is erroneous if a subprogram is invoked as part of its execution. No_Reentrancy ------------- .. index:: No_Reentrancy [RM H.4] A program execution is erroneous if a subprogram is executed by two tasks at the same time. No_Relative_Delay ----------------- .. index:: No_Relative_Delay [RM D.7] This restriction ensures at compile time that there are no delay relative statements and prevents expressions such as ``delay 1.23;`` from appearing in source code. No_Requeue_Statements --------------------- .. index:: No_Requeue_Statements [RM D.7] This restriction ensures at compile time that no requeue statements are permitted and prevents keyword ``requeue`` from being used in source code. .. index:: No_Requeue The restriction ``No_Requeue`` is recognized as a synonym for ``No_Requeue_Statements``. This is retained for historical compatibility purposes (and a warning will be generated for its use if warnings on oNobsolescent features are activated). No_Secondary_Stack ------------------ .. index:: No_Secondary_Stack [GNAT] This restriction ensures at compile time that the generated code does not contain any reference to the secondary stack. The secondary stack is used to implement functions returning unconstrained objects (arrays or records) on some targets. Suppresses the allocation of secondary stacks for tasks (excluding the environment task) at run time. No_Select_Statements -------------------- .. index:: No_Select_Statements [RM D.7] This restriction ensures at compile time no select statements of any kind are permitted, that is the keyword ``select`` may not appear. No_Specific_Termination_Handlers -------------------------------- .. index:: No_Specific_Termination_Handlers [RM D.7] There are no calls to Ada.Task_Termination.Set_Specific_Handler or to Ada.Task_Termination.Specific_Handler. No_Specification_of_Aspect -------------------------- .. index:: No_Specification_of_Aspect [RM 13.12.1] This restriction checks at compile time that no aspect specification, attribute definition clause, or pragma is given for a given aspect. No_Standard_Allocators_After_Elaboration ---------------------------------------- .. index:: No_Standard_Allocators_After_Elaboration [RM D.7] Specifies that an allocator using a standard storage pool should never be evaluated at run time after the elaboration of the library items of the partition has completed. Otherwise, Storage_Error is raised. No_Standard_Storage_Pools ------------------------- .. index:: No_Standard_Storage_Pools [GNAT] This restriction ensures at compile time that no access types use the standard default storage pool. Any access type declared must have an explicit Storage_Pool attribute defined specifying a user-defined storage pool. No_Stream_Optimizations ----------------------- .. index:: No_Stream_Optimizations [GNAT] This restriction affects the performance of stream operations on types ``String``, ``Wide_String`` and ``Wide_Wide_String``. By default, the compiler uses block reads and writes when manipulating ``String`` objects due to their superior performance. When this restriction is in effect, the compiler performs all IO operations on a per-character basis. No_Streams ---------- .. index:: No_Streams [GNAT] This restriction ensures at compile/bind time that there are no stream objects created and no use of stream attributes. This restriction does not forbid dependences on the package ``Ada.Streams``. So it is permissible to with ``Ada.Streams`` (or another package that does so itself) as long as no actual stream objects are created and no stream attributes are used. Note that the use of restriction allows optimization of tagged types, since they do not need to worry about dispatching stream operations. To take maximum advantage of this space-saving optimization, any unit declaring a tagged type should be compiled with the restriction, though this is not required. When pragmas ``Discard_Names`` and ``Restrictions (No_Streams)`` simultaneously apply to a tagged type, its Expanded_Name and External_Tag are also initialized with empty strings. In particular, both these pragmas can be applied as configuration pragmas to avoid exposing entity names at binary level for the entire partition. No_Tagged_Type_Registration --------------------------- .. index:: No_Tagged_Type_Registration [GNAT] If this restriction is active, then class-wide streaming attributes are not supported. In addition, the subprograms in Ada.Tags are not supported. If this restriction is active, the generated code is simplified by omitting the otherwise-required global registration of tagged types when they are declared. This restriction may be necessary in order to also apply the No_Elaboration_Code restriction. No_Task_Allocators ------------------ .. index:: No_Task_Allocators [RM D.7] There are no allocators for task types or types containing task subcomponents. No_Task_At_Interrupt_Priority ----------------------------- .. index:: No_Task_At_Interrupt_Priority [GNAT] This restriction ensures at compile time that there is no Interrupt_Priority aspect or pragma for a task or a task type. As a consequence, the tasks are always created with a priority below that an interrupt priority. No_Task_Attributes_Package -------------------------- .. index:: No_Task_Attributes_Package [GNAT] This restriction ensures at compile time that there are no implicit or explicit dependencies on the package ``Ada.Task_Attributes``. .. index:: No_Task_Attributes The restriction ``No_Task_Attributes`` is recognized as a synonym for ``No_Task_Attributes_Package``. This is retained for historical compatibility purposes (and a warning will be generated for its use if warnings on obsolescent features are activated). No_Task_Hierarchy ----------------- .. index:: No_Task_Hierarchy [RM D.7] All (non-environment) tasks depend directly on the environment task of the partition. No_Task_Termination ------------------- .. index:: No_Task_Termination [RM D.7] Tasks that terminate are erroneous. No_Tasking ---------- .. index:: No_Tasking [GNAT] This restriction prevents the declaration of tasks or task types throughout the partition. It is similar in effect to the use of ``Max_Tasks => 0`` except that violations are caught at compile time and cause an error message to be output either by the compiler or binder. No_Terminate_Alternatives ------------------------- .. index:: No_Terminate_Alternatives [RM D.7] There are no selective accepts with terminate alternatives. No_Unchecked_Access ------------------- .. index:: No_Unchecked_Access [RM H.4] This restriction ensures at compile time that there are no occurrences of the Unchecked_Access attribute. No_Unchecked_Conversion ----------------------- .. index:: No_Unchecked_Conversion [RM J.13] This restriction ensures at compile time that there are no semantic dependences on the predefined generic function Unchecked_Conversion. No_Unchecked_Deallocation ------------------------- .. index:: No_Unchecked_Deallocation [RM J.13] This restriction ensures at compile time that there are no semantic dependences on the predefined generic procedure Unchecked_Deallocation. No_Use_Of_Attribute ------------------- .. index:: No_Use_Of_Attribute [RM 13.12.1] This is a standard Ada 2012 restriction that is GNAT defined in earlier versions of Ada. No_Use_Of_Entity ---------------- .. index:: No_Use_Of_Entity [GNAT] This restriction ensures at compile time that there are no references to the entity given in the form :: No_Use_Of_Entity => Name where ``Name`` is the fully qualified entity, for example :: No_Use_Of_Entity => Ada.Text_IO.Put_Line No_Use_Of_Pragma ---------------- .. index:: No_Use_Of_Pragma [RM 13.12.1] This is a standard Ada 2012 restriction that is GNAT defined in earlier versions of Ada. Pure_Barriers ------------- .. index:: Pure_Barriers [GNAT] This restriction ensures at compile time that protected entry barriers are restricted to: * components of the protected object (excluding selection from dereferences), * constant declarations, * named numbers, * enumeration literals, * integer literals, * real literals, * character literals, * implicitly defined comparison operators, * uses of the Standard."not" operator, * short-circuit operator, * the Count attribute This restriction is a relaxation of the Simple_Barriers restriction, but still ensures absence of side effects, exceptions, and recursion during the evaluation of the barriers. Simple_Barriers --------------- .. index:: Simple_Barriers [RM D.7] This restriction ensures at compile time that barriers in entry declarations for protected types are restricted to either static boolean expressions or references to simple boolean variables defined in the private part of the protected type. No other form of entry barriers is permitted. .. index:: Boolean_Entry_Barriers The restriction ``Boolean_Entry_Barriers`` is recognized as a synonym for ``Simple_Barriers``. This is retained for historical compatibility purposes (and a warning will be generated for its use if warnings on obsolescent features are activated). Static_Priorities ----------------- .. index:: Static_Priorities [GNAT] This restriction ensures at compile time that all priority expressions are static, and that there are no dependences on the package ``Ada.Dynamic_Priorities``. Static_Storage_Size ------------------- .. index:: Static_Storage_Size [GNAT] This restriction ensures at compile time that any expression appearing in a Storage_Size pragma or attribute definition clause is static. .. _Program_Unit_Level_Restrictions: Program Unit Level Restrictions =============================== The second set of restriction identifiers does not require partition-wide consistency. The restriction may be enforced for a single compilation unit without any effect on any of the other compilation units in the partition. No_Elaboration_Code ------------------- .. index:: No_Elaboration_Code [GNAT] This restriction ensures at compile time that no elaboration code is generated. Note that this is not the same condition as is enforced by pragma ``Preelaborate``. There are cases in which pragma ``Preelaborate`` still permits code to be generated (e.g., code to initialize a large array to all zeroes), and there are cases of units which do not meet the requirements for pragma ``Preelaborate``, but for which no elaboration code is generated. Generally, it is the case that preelaborable units will meet the restrictions, with the exception of large aggregates initialized with an others_clause, and exception declarations (which generate calls to a run-time registry procedure). This restriction is enforced on a unit by unit basis, it need not be obeyed consistently throughout a partition. In the case of aggregates with others, if the aggregate has a dynamic size, there is no way to eliminate the elaboration code (such dynamic bounds would be incompatible with ``Preelaborate`` in any case). If the bounds are static, then use of this restriction actually modifies the code choice of the compiler to avoid generating a loop, and instead generate the aggregate statically if possible, no matter how many times the data for the others clause must be repeatedly generated. It is not possible to precisely document the constructs which are compatible with this restriction, since, unlike most other restrictions, this is not a restriction on the source code, but a restriction on the generated object code. For example, if the source contains a declaration: .. code-block:: ada Val : constant Integer := X; where X is not a static constant, it may be possible, depending on complex optimization circuitry, for the compiler to figure out the value of X at compile time, in which case this initialization can be done by the loader, and requires no initialization code. It is not possible to document the precise conditions under which the optimizer can figure this out. Note that this the implementation of this restriction requires full code generation. If it is used in conjunction with "semantics only" checking, then some cases of violations may be missed. When this restriction is active, we are not requesting control-flow preservation with -fpreserve-control-flow, and the static elaboration model is used, the compiler is allowed to suppress the elaboration counter normally associated with the unit. This counter is typically used to check for access before elaboration and to control multiple elaboration attempts. No_Dynamic_Accessibility_Checks ------------------------------- .. index:: No_Dynamic_Accessibility_Checks [GNAT] No dynamic accessibility checks are generated when this restriction is in effect. Instead, dangling references are prevented via more conservative compile-time checking. More specifically, existing compile-time checks are enforced but with more conservative assumptions about the accessibility levels of the relevant entities. These conservative assumptions eliminate the need for dynamic accessibility checks. These new rules for computing (at compile-time) the accessibility level of an anonymous access type T are as follows: * If T is a function result type then, from the caller's perspective, its level is that of the innermost master enclosing the function call. From the callee's perspective, the level of parameters and local variables of the callee is statically deeper than the level of T. For any other accessibility level L such that the level of parameters and local variables of the callee is statically deeper than L, the level of T (from the callee's perspective) is also statically deeper than L. * If T is the type of a formal parameter then, from the caller's perspective, its level is at least as deep as that of the type of the corresponding actual parameter (whatever that actual parameter might be). From the callee's perspective, the level of parameters and local variables of the callee is statically deeper than the level of T. * If T is the type of a discriminant then its level is that of the discriminated type. * If T is the type of a stand-alone object then its level is the level of the object. * In all other cases, the level of T is as defined by the existing rules of Ada. No_Dynamic_Sized_Objects ------------------------ .. index:: No_Dynamic_Sized_Objects [GNAT] This restriction disallows certain constructs that might lead to the creation of dynamic-sized composite objects (or array or discriminated type). An array subtype indication is illegal if the bounds are not static or references to discriminants of an enclosing type. A discriminated subtype indication is illegal if the type has discriminant-dependent array components or a variant part, and the discriminants are not static. In addition, array and record aggregates are illegal in corresponding cases. Note that this restriction does not forbid access discriminants. It is often a good idea to combine this restriction with No_Secondary_Stack. No_Entry_Queue -------------- .. index:: No_Entry_Queue [GNAT] This restriction is a declaration that any protected entry compiled in the scope of the restriction has at most one task waiting on the entry at any one time, and so no queue is required. This restriction is not checked at compile time. A program execution is erroneous if an attempt is made to queue a second task on such an entry. No_Implementation_Aspect_Specifications --------------------------------------- .. index:: No_Implementation_Aspect_Specifications [RM 13.12.1] This restriction checks at compile time that no GNAT-defined aspects are present. With this restriction, the only aspects that can be used are those defined in the Ada Reference Manual. No_Implementation_Attributes ---------------------------- .. index:: No_Implementation_Attributes [RM 13.12.1] This restriction checks at compile time that no GNAT-defined attributes are present. With this restriction, the only attributes that can be used are those defined in the Ada Reference Manual. No_Implementation_Identifiers ----------------------------- .. index:: No_Implementation_Identifiers [RM 13.12.1] This restriction checks at compile time that no implementation-defined identifiers (marked with pragma Implementation_Defined) occur within language-defined packages. No_Implementation_Pragmas ------------------------- .. index:: No_Implementation_Pragmas [RM 13.12.1] This restriction checks at compile time that no GNAT-defined pragmas are present. With this restriction, the only pragmas that can be used are those defined in the Ada Reference Manual. No_Implementation_Restrictions ------------------------------ .. index:: No_Implementation_Restrictions [GNAT] This restriction checks at compile time that no GNAT-defined restriction identifiers (other than ``No_Implementation_Restrictions`` itself) are present. With this restriction, the only other restriction identifiers that can be used are those defined in the Ada Reference Manual. No_Implementation_Units ----------------------- .. index:: No_Implementation_Units [RM 13.12.1] This restriction checks at compile time that there is no mention in the context clause of any implementation-defined descendants of packages Ada, Interfaces, or System. No_Implicit_Aliasing -------------------- .. index:: No_Implicit_Aliasing [GNAT] This restriction, which is not required to be partition-wide consistent, requires an explicit aliased keyword for an object to which 'Access, 'Unchecked_Access, or 'Address is applied, and forbids entirely the use of the 'Unrestricted_Access attribute for objects. Note: the reason that Unrestricted_Access is forbidden is that it would require the prefix to be aliased, and in such cases, it can always be replaced by the standard attribute Unchecked_Access which is preferable. No_Implicit_Loops ----------------- .. index:: No_Implicit_Loops [GNAT] This restriction ensures that the generated code of the unit marked with this restriction does not contain any implicit ``for`` loops, either by modifying the generated code where possible, or by rejecting any construct that would otherwise generate an implicit ``for`` loop. If this restriction is active, it is possible to build large array aggregates with all static components without generating an intermediate temporary, and without generating a loop to initialize individual components. Otherwise, a loop is created for arrays larger than about 5000 scalar components. Note that if this restriction is set in the spec of a package, it will not apply to its body. No_Obsolescent_Features ----------------------- .. index:: No_Obsolescent_Features [RM 13.12.1] This restriction checks at compile time that no obsolescent features are used, as defined in Annex J of the Ada Reference Manual. No_Wide_Characters ------------------ .. index:: No_Wide_Characters [GNAT] This restriction ensures at compile time that no uses of the types ``Wide_Character`` or ``Wide_String`` or corresponding wide wide types appear, and that no wide or wide wide string or character literals appear in the program (that is literals representing characters not in type ``Character``). Static_Dispatch_Tables ---------------------- .. index:: Static_Dispatch_Tables [GNAT] This restriction checks at compile time that all the artifacts associated with dispatch tables can be placed in read-only memory. SPARK_05 -------- .. index:: SPARK_05 [GNAT] This restriction no longer has any effect and is superseded by SPARK 2014, whose restrictions are checked by the tool GNATprove. To check that a codebase respects SPARK 2014 restrictions, mark the code with pragma or aspect ``SPARK_Mode``, and run the tool GNATprove at Stone assurance level, as follows:: gnatprove -P project.gpr --mode=stone or equivalently:: gnatprove -P project.gpr --mode=check_all --- ### Gcc Ada Doc Gnat Rm Standard Library Routines (`gcc/ada/doc/gnat_rm/standard_library_routines.rst`) .. _Standard_Library_Routines: ************************* Standard Library Routines ************************* The Ada Reference Manual contains in Annex A a full description of an extensive set of standard library routines that can be used in any Ada program, and which must be provided by all Ada compilers. They are analogous to the standard C library used by C programs. GNAT implements all of the facilities described in annex A, and for most purposes the description in the Ada Reference Manual, or appropriate Ada text book, will be sufficient for making use of these facilities. In the case of the input-output facilities, :ref:`The_Implementation_of_Standard_I/O`, gives details on exactly how GNAT interfaces to the file system. For the remaining packages, the Ada Reference Manual should be sufficient. The following is a list of the packages included, together with a brief description of the functionality that is provided. For completeness, references are included to other predefined library routines defined in other sections of the Ada Reference Manual (these are cross-indexed from Annex A). For further details see the relevant package declarations in the run-time library. In particular, a few units are not implemented, as marked by the presence of pragma Unimplemented_Unit, and in this case the package declaration contains comments explaining why the unit is not implemented. ``Ada`` *(A.2)* This is a parent package for all the standard library packages. It is usually included implicitly in your program, and itself contains no useful data or routines. ``Ada.Assertions`` *(11.4.2)* ``Assertions`` provides the ``Assert`` subprograms, and also the declaration of the ``Assertion_Error`` exception. ``Ada.Asynchronous_Task_Control`` *(D.11)* ``Asynchronous_Task_Control`` provides low level facilities for task synchronization. It is typically not implemented. See package spec for details. ``Ada.Calendar`` *(9.6)* ``Calendar`` provides time of day access, and routines for manipulating times and durations. ``Ada.Calendar.Arithmetic`` *(9.6.1)* This package provides additional arithmetic operations for ``Calendar``. ``Ada.Calendar.Formatting`` *(9.6.1)* This package provides formatting operations for ``Calendar``. ``Ada.Calendar.Time_Zones`` *(9.6.1)* This package provides additional ``Calendar`` facilities for handling time zones. ``Ada.Characters`` *(A.3.1)* This is a dummy parent package that contains no useful entities ``Ada.Characters.Conversions`` *(A.3.2)* This package provides character conversion functions. ``Ada.Characters.Handling`` *(A.3.2)* This package provides some basic character handling capabilities, including classification functions for classes of characters (e.g., test for letters, or digits). ``Ada.Characters.Latin_1`` *(A.3.3)* This package includes a complete set of definitions of the characters that appear in type CHARACTER. It is useful for writing programs that will run in international environments. For example, if you want an upper case E with an acute accent in a string, it is often better to use the definition of ``UC_E_Acute`` in this package. Then your program will print in an understandable manner even if your environment does not support these extended characters. ``Ada.Command_Line`` *(A.15)* This package provides access to the command line parameters and the name of the current program (analogous to the use of ``argc`` and ``argv`` in C), and also allows the exit status for the program to be set in a system-independent manner. ``Ada.Complex_Text_IO`` *(G.1.3)* This package provides text input and output of complex numbers. ``Ada.Containers`` *(A.18.1)* A top level package providing a few basic definitions used by all the following specific child packages that provide specific kinds of containers. ``Ada.Containers.Bounded_Priority_Queues`` *(A.18.31)* ``Ada.Containers.Bounded_Synchronized_Queues`` *(A.18.29)* ``Ada.Containers.Doubly_Linked_Lists`` *(A.18.3)* ``Ada.Containers.Generic_Array_Sort`` *(A.18.26)* ``Ada.Containers.Generic_Constrained_Array_Sort`` *(A.18.26)* ``Ada.Containers.Generic_Sort`` *(A.18.26)* ``Ada.Containers.Hashed_Maps`` *(A.18.5)* ``Ada.Containers.Hashed_Sets`` *(A.18.8)* ``Ada.Containers.Indefinite_Doubly_Linked_Lists`` *(A.18.12)* ``Ada.Containers.Indefinite_Hashed_Maps`` *(A.18.13)* ``Ada.Containers.Indefinite_Hashed_Sets`` *(A.18.15)* ``Ada.Containers.Indefinite_Holders`` *(A.18.18)* ``Ada.Containers.Indefinite_Multiway_Trees`` *(A.18.17)* ``Ada.Containers.Indefinite_Ordered_Maps`` *(A.18.14)* ``Ada.Containers.Indefinite_Ordered_Sets`` *(A.18.16)* ``Ada.Containers.Indefinite_Vectors`` *(A.18.11)* ``Ada.Containers.Multiway_Trees`` *(A.18.10)* ``Ada.Containers.Ordered_Maps`` *(A.18.6)* ``Ada.Containers.Ordered_Sets`` *(A.18.9)* ``Ada.Containers.Synchronized_Queue_Interfaces`` *(A.18.27)* ``Ada.Containers.Unbounded_Priority_Queues`` *(A.18.30)* ``Ada.Containers.Unbounded_Synchronized_Queues`` *(A.18.28)* ``Ada.Containers.Vectors`` *(A.18.2)* ``Ada.Directories`` *(A.16)* This package provides operations on directories. ``Ada.Directories.Hierarchical_File_Names`` *(A.16.1)* This package provides additional directory operations handling hierarchical file names. ``Ada.Directories.Information`` *(A.16)* This is an implementation defined package for additional directory operations, which is not implemented in GNAT. ``Ada.Decimal`` *(F.2)* This package provides constants describing the range of decimal numbers implemented, and also a decimal divide routine (analogous to the COBOL verb DIVIDE ... GIVING ... REMAINDER ...) ``Ada.Direct_IO`` *(A.8.4)* This package provides input-output using a model of a set of records of fixed-length, containing an arbitrary definite Ada type, indexed by an integer record number. ``Ada.Dispatching`` *(D.2.1)* A parent package containing definitions for task dispatching operations. ``Ada.Dispatching.EDF`` *(D.2.6)* Not implemented in GNAT. ``Ada.Dispatching.Non_Preemptive`` *(D.2.4)* Not implemented in GNAT. ``Ada.Dispatching.Round_Robin`` *(D.2.5)* Not implemented in GNAT. ``Ada.Dynamic_Priorities`` *(D.5)* This package allows the priorities of a task to be adjusted dynamically as the task is running. ``Ada.Environment_Variables`` *(A.17)* This package provides facilities for accessing environment variables. ``Ada.Exceptions`` *(11.4.1)* This package provides additional information on exceptions, and also contains facilities for treating exceptions as data objects, and raising exceptions with associated messages. ``Ada.Execution_Time`` *(D.14)* This package provides CPU clock functionalities. It is not implemented on all targets (see package spec for details). ``Ada.Execution_Time.Group_Budgets`` *(D.14.2)* Not implemented in GNAT. ``Ada.Execution_Time.Timers`` *(D.14.1)'* Not implemented in GNAT. ``Ada.Finalization`` *(7.6)* This package contains the declarations and subprograms to support the use of controlled types, providing for automatic initialization and finalization (analogous to the constructors and destructors of C++). ``Ada.Float_Text_IO`` *(A.10.9)* A library level instantiation of Text_IO.Float_IO for type Float. ``Ada.Float_Wide_Text_IO`` *(A.10.9)* A library level instantiation of Wide_Text_IO.Float_IO for type Float. ``Ada.Float_Wide_Wide_Text_IO`` *(A.10.9)* A library level instantiation of Wide_Wide_Text_IO.Float_IO for type Float. ``Ada.Integer_Text_IO`` *(A.10.9)* A library level instantiation of Text_IO.Integer_IO for type Integer. ``Ada.Integer_Wide_Text_IO`` *(A.10.9)* A library level instantiation of Wide_Text_IO.Integer_IO for type Integer. ``Ada.Integer_Wide_Wide_Text_IO`` *(A.10.9)* A library level instantiation of Wide_Wide_Text_IO.Integer_IO for type Integer. ``Ada.Interrupts`` *(C.3.2)* This package provides facilities for interfacing to interrupts, which includes the set of signals or conditions that can be raised and recognized as interrupts. ``Ada.Interrupts.Names`` *(C.3.2)* This package provides the set of interrupt names (actually signal or condition names) that can be handled by GNAT. ``Ada.IO_Exceptions`` *(A.13)* This package defines the set of exceptions that can be raised by use of the standard IO packages. ``Ada.Iterator_Interfaces`` *(5.5.1)* This package provides a generic interface to generalized iterators. ``Ada.Locales`` *(A.19)* This package provides declarations providing information (Language and Country) about the current locale. ``Ada.Numerics`` This package contains some standard constants and exceptions used throughout the numerics packages. Note that the constants pi and e are defined here, and it is better to use these definitions than rolling your own. ``Ada.Numerics.Complex_Arrays`` *(G.3.2)* Provides operations on arrays of complex numbers. ``Ada.Numerics.Complex_Elementary_Functions`` Provides the implementation of standard elementary functions (such as log and trigonometric functions) operating on complex numbers using the standard ``Float`` and the ``Complex`` and ``Imaginary`` types created by the package ``Numerics.Complex_Types``. ``Ada.Numerics.Complex_Types`` This is a predefined instantiation of ``Numerics.Generic_Complex_Types`` using ``Standard.Float`` to build the type ``Complex`` and ``Imaginary``. ``Ada.Numerics.Discrete_Random`` This generic package provides a random number generator suitable for generating uniformly distributed values of a specified discrete subtype. It should not be used as a cryptographic pseudo-random source. ``Ada.Numerics.Float_Random`` This package provides a random number generator suitable for generating uniformly distributed floating point values in the unit interval. It should not be used as a cryptographic pseudo-random source. ``Ada.Numerics.Generic_Complex_Elementary_Functions`` This is a generic version of the package that provides the implementation of standard elementary functions (such as log and trigonometric functions) for an arbitrary complex type. The following predefined instantiations of this package are provided: * ``Short_Float`` ``Ada.Numerics.Short_Complex_Elementary_Functions`` * ``Float`` ``Ada.Numerics.Complex_Elementary_Functions`` * ``Long_Float`` ``Ada.Numerics.Long_Complex_Elementary_Functions`` ``Ada.Numerics.Generic_Complex_Types`` This is a generic package that allows the creation of complex types, with associated complex arithmetic operations. The following predefined instantiations of this package exist * ``Short_Float`` ``Ada.Numerics.Short_Complex_Complex_Types`` * ``Float`` ``Ada.Numerics.Complex_Complex_Types`` * ``Long_Float`` ``Ada.Numerics.Long_Complex_Complex_Types`` ``Ada.Numerics.Generic_Elementary_Functions`` This is a generic package that provides the implementation of standard elementary functions (such as log an trigonometric functions) for an arbitrary float type. The following predefined instantiations of this package exist * ``Short_Float`` ``Ada.Numerics.Short_Elementary_Functions`` * ``Float`` ``Ada.Numerics.Elementary_Functions`` * ``Long_Float`` ``Ada.Numerics.Long_Elementary_Functions`` ``Ada.Numerics.Generic_Real_Arrays`` *(G.3.1)* Generic operations on arrays of reals ``Ada.Numerics.Real_Arrays`` *(G.3.1)* Preinstantiation of Ada.Numerics.Generic_Real_Arrays (Float). ``Ada.Real_Time`` *(D.8)* This package provides facilities similar to those of ``Calendar``, but operating with a finer clock suitable for real time control. Note that annex D requires that there be no backward clock jumps, and GNAT generally guarantees this behavior, but of course if the external clock on which the GNAT runtime depends is deliberately reset by some external event, then such a backward jump may occur. ``Ada.Real_Time.Timing_Events`` *(D.15)* This package allows procedures to be executed at a specified time without the use of a task or a delay statement. ``Ada.Sequential_IO`` *(A.8.1)* This package provides input-output facilities for sequential files, which can contain a sequence of values of a single type, which can be any Ada type, including indefinite (unconstrained) types. ``Ada.Storage_IO`` *(A.9)* This package provides a facility for mapping arbitrary Ada types to and from a storage buffer. It is primarily intended for the creation of new IO packages. ``Ada.Streams`` *(13.13.1)* This is a generic package that provides the basic support for the concept of streams as used by the stream attributes (``Input``, ``Output``, ``Read`` and ``Write``). ``Ada.Streams.Stream_IO`` *(A.12.1)* This package is a specialization of the type ``Streams`` defined in package ``Streams`` together with a set of operations providing Stream_IO capability. The Stream_IO model permits both random and sequential access to a file which can contain an arbitrary set of values of one or more Ada types. ``Ada.Strings`` *(A.4.1)* This package provides some basic constants used by the string handling packages. ``Ada.Strings.Bounded`` *(A.4.4)* This package provides facilities for handling variable length strings. The bounded model requires a maximum length. It is thus somewhat more limited than the unbounded model, but avoids the use of dynamic allocation or finalization. ``Ada.Strings.Bounded.Equal_Case_Insensitive`` *(A.4.10)* Provides case-insensitive comparisons of bounded strings ``Ada.Strings.Bounded.Hash`` *(A.4.9)* This package provides a generic hash function for bounded strings ``Ada.Strings.Bounded.Hash_Case_Insensitive`` *(A.4.9)* This package provides a generic hash function for bounded strings that converts the string to be hashed to lower case. ``Ada.Strings.Bounded.Less_Case_Insensitive`` *(A.4.10)* This package provides a comparison function for bounded strings that works in a case insensitive manner by converting to lower case before the comparison. ``Ada.Strings.Fixed`` *(A.4.3)* This package provides facilities for handling fixed length strings. ``Ada.Strings.Fixed.Equal_Case_Insensitive`` *(A.4.10)* This package provides an equality function for fixed strings that compares the strings after converting both to lower case. ``Ada.Strings.Fixed.Hash_Case_Insensitive`` *(A.4.9)* This package provides a case insensitive hash function for fixed strings that converts the string to lower case before computing the hash. ``Ada.Strings.Fixed.Less_Case_Insensitive`` *(A.4.10)* This package provides a comparison function for fixed strings that works in a case insensitive manner by converting to lower case before the comparison. ``Ada.Strings.Hash`` *(A.4.9)* This package provides a hash function for strings. ``Ada.Strings.Hash_Case_Insensitive`` *(A.4.9)* This package provides a hash function for strings that is case insensitive. The string is converted to lower case before computing the hash. ``Ada.Strings.Less_Case_Insensitive`` *(A.4.10)* This package provides a comparison function for\\strings that works in a case insensitive manner by converting to lower case before the comparison. ``Ada.Strings.Maps`` *(A.4.2)* This package provides facilities for handling character mappings and arbitrarily defined subsets of characters. For instance it is useful in defining specialized translation tables. ``Ada.Strings.Maps.Constants`` *(A.4.6)* This package provides a standard set of predefined mappings and predefined character sets. For example, the standard upper to lower case conversion table is found in this package. Note that upper to lower case conversion is non-trivial if you want to take the entire set of characters, including extended characters like E with an acute accent, into account. You should use the mappings in this package (rather than adding 32 yourself) to do case mappings. ``Ada.Strings.Unbounded`` *(A.4.5)* This package provides facilities for handling variable length strings. The unbounded model allows arbitrary length strings, but requires the use of dynamic allocation and finalization. ``Ada.Strings.Unbounded.Equal_Case_Insensitive`` *(A.4.10)* Provides case-insensitive comparisons of unbounded strings ``Ada.Strings.Unbounded.Hash`` *(A.4.9)* This package provides a generic hash function for unbounded strings ``Ada.Strings.Unbounded.Hash_Case_Insensitive`` *(A.4.9)* This package provides a generic hash function for unbounded strings that converts the string to be hashed to lower case. ``Ada.Strings.Unbounded.Less_Case_Insensitive`` *(A.4.10)* This package provides a comparison function for unbounded strings that works in a case insensitive manner by converting to lower case before the comparison. ``Ada.Strings.UTF_Encoding`` *(A.4.11)* This package provides basic definitions for dealing with UTF-encoded strings. ``Ada.Strings.UTF_Encoding.Conversions`` *(A.4.11)* This package provides conversion functions for UTF-encoded strings. ``Ada.Strings.UTF_Encoding.Strings`` *(A.4.11)* ``Ada.Strings.UTF_Encoding.Wide_Strings`` *(A.4.11)* ``Ada.Strings.UTF_Encoding.Wide_Wide_Strings`` *(A.4.11)* These packages provide facilities for handling UTF encodings for Strings, Wide_Strings and Wide_Wide_Strings. ``Ada.Strings.Wide_Bounded`` *(A.4.7)* ``Ada.Strings.Wide_Fixed`` *(A.4.7)* ``Ada.Strings.Wide_Maps`` *(A.4.7)* ``Ada.Strings.Wide_Unbounded`` *(A.4.7)* These packages provide analogous capabilities to the corresponding packages without ``Wide_`` in the name, but operate with the types ``Wide_String`` and ``Wide_Character`` instead of ``String`` and ``Character``. Versions of all the child packages are available. ``Ada.Strings.Wide_Wide_Bounded`` *(A.4.7)* ``Ada.Strings.Wide_Wide_Fixed`` *(A.4.7)* ``Ada.Strings.Wide_Wide_Maps`` *(A.4.7)* ``Ada.Strings.Wide_Wide_Unbounded`` *(A.4.7)* These packages provide analogous capabilities to the corresponding packages without ``Wide_`` in the name, but operate with the types ``Wide_Wide_String`` and ``Wide_Wide_Character`` instead of ``String`` and ``Character``. ``Ada.Synchronous_Barriers`` *(D.10.1)* This package provides facilities for synchronizing tasks at a low level with barriers. ``Ada.Synchronous_Task_Control`` *(D.10)* This package provides some standard facilities for controlling task communication in a synchronous manner. ``Ada.Synchronous_Task_Control.EDF`` *(D.10)* Not implemented in GNAT. ``Ada.Tags`` This package contains definitions for manipulation of the tags of tagged values. ``Ada.Tags.Generic_Dispatching_Constructor`` *(3.9)* This package provides a way of constructing tagged class-wide values given only the tag value. ``Ada.Task_Attributes`` *(C.7.2)* This package provides the capability of associating arbitrary task-specific data with separate tasks. ``Ada.Task_Identification`` *(C.7.1)* This package provides capabilities for task identification. ``Ada.Task_Termination`` *(C.7.3)* This package provides control over task termination. ``Ada.Text_IO`` This package provides basic text input-output capabilities for character, string and numeric data. The subpackages of this package are listed next. Note that although these are defined as subpackages in the RM, they are actually transparently implemented as child packages in GNAT, meaning that they are only loaded if needed. ``Ada.Text_IO.Decimal_IO`` Provides input-output facilities for decimal fixed-point types ``Ada.Text_IO.Enumeration_IO`` Provides input-output facilities for enumeration types. ``Ada.Text_IO.Fixed_IO`` Provides input-output facilities for ordinary fixed-point types. ``Ada.Text_IO.Float_IO`` Provides input-output facilities for float types. The following predefined instantiations of this generic package are available: * ``Short_Float`` ``Short_Float_Text_IO`` * ``Float`` ``Float_Text_IO`` * ``Long_Float`` ``Long_Float_Text_IO`` ``Ada.Text_IO.Integer_IO`` Provides input-output facilities for integer types. The following predefined instantiations of this generic package are available: * ``Short_Short_Integer`` ``Ada.Short_Short_Integer_Text_IO`` * ``Short_Integer`` ``Ada.Short_Integer_Text_IO`` * ``Integer`` ``Ada.Integer_Text_IO`` * ``Long_Integer`` ``Ada.Long_Integer_Text_IO`` * ``Long_Long_Integer`` ``Ada.Long_Long_Integer_Text_IO`` ``Ada.Text_IO.Modular_IO`` Provides input-output facilities for modular (unsigned) types. ``Ada.Text_IO.Bounded_IO (A.10.11)`` Provides input-output facilities for bounded strings. ``Ada.Text_IO.Complex_IO (G.1.3)`` This package provides basic text input-output capabilities for complex data. ``Ada.Text_IO.Editing (F.3.3)`` This package contains routines for edited output, analogous to the use of pictures in COBOL. The picture formats used by this package are a close copy of the facility in COBOL. ``Ada.Text_IO.Text_Streams (A.12.2)`` This package provides a facility that allows Text_IO files to be treated as streams, so that the stream attributes can be used for writing arbitrary data, including binary data, to Text_IO files. ``Ada.Text_IO.Unbounded_IO (A.10.12)`` This package provides input-output facilities for unbounded strings. ``Ada.Unchecked_Conversion (13.9)`` This generic package allows arbitrary conversion from one type to another of the same size, providing for breaking the type safety in special circumstances. If the types have the same Size (more accurately the same Value_Size), then the effect is simply to transfer the bits from the source to the target type without any modification. This usage is well defined, and for simple types whose representation is typically the same across all implementations, gives a portable method of performing such conversions. If the types do not have the same size, then the result is implementation defined, and thus may be non-portable. The following describes how GNAT handles such unchecked conversion cases. If the types are of different sizes, and are both discrete types, then the effect is of a normal type conversion without any constraint checking. In particular if the result type has a larger size, the result will be zero or sign extended. If the result type has a smaller size, the result will be truncated by ignoring high order bits. If the types are of different sizes, and are not both discrete types, then the conversion works as though pointers were created to the source and target, and the pointer value is converted. The effect is that bits are copied from successive low order storage units and bits of the source up to the length of the target type. A warning is issued if the lengths differ, since the effect in this case is implementation dependent, and the above behavior may not match that of some other compiler. A pointer to one type may be converted to a pointer to another type using unchecked conversion. The only case in which the effect is undefined is when one or both pointers are pointers to unconstrained array types. In this case, the bounds information may get incorrectly transferred, and in particular, GNAT uses double size pointers for such types, and it is meaningless to convert between such pointer types. GNAT will issue a warning if the alignment of the target designated type is more strict than the alignment of the source designated type (since the result may be unaligned in this case). A pointer other than a pointer to an unconstrained array type may be converted to and from System.Address. Such usage is common in Ada 83 programs, but note that Ada.Address_To_Access_Conversions is the preferred method of performing such conversions in Ada 95 and Ada 2005. Neither unchecked conversion nor Ada.Address_To_Access_Conversions should be used in conjunction with pointers to unconstrained objects, since the bounds information cannot be handled correctly in this case. ``Ada.Unchecked_Deallocation`` *(13.11.2)* This generic package allows explicit freeing of storage previously allocated by use of an allocator. ``Ada.Wide_Text_IO`` *(A.11)* This package is similar to ``Ada.Text_IO``, except that the external file supports wide character representations, and the internal types are ``Wide_Character`` and ``Wide_String`` instead of ``Character`` and ``String``. The corresponding set of nested packages and child packages are defined. ``Ada.Wide_Wide_Text_IO`` *(A.11)* This package is similar to ``Ada.Text_IO``, except that the external file supports wide character representations, and the internal types are ``Wide_Character`` and ``Wide_String`` instead of ``Character`` and ``String``. The corresponding set of nested packages and child packages are defined. For packages in Interfaces and System, all the RM defined packages are available in GNAT, see the Ada 2012 RM for full details. --- ### Gcc Ada Doc Gnat Rm The Gnat Library (`gcc/ada/doc/gnat_rm/the_gnat_library.rst`) .. role:: switch(samp) .. _The_GNAT_Library: **************** The GNAT Library **************** The GNAT library contains a number of general and special purpose packages. It represents functionality that the GNAT developers have found useful, and which is made available to GNAT users. The packages described here are fully supported, and upwards compatibility will be maintained in future releases, so you can use these facilities with the confidence that the same functionality will be available in future releases. The chapter here simply gives a brief summary of the facilities available. The full documentation is found in the spec file for the package. The full sources of these library packages, including both spec and body, are provided with all GNAT releases. For example, to find out the full specifications of the SPITBOL pattern matching capability, including a full tutorial and extensive examples, look in the :file:`g-spipat.ads` file in the library. For each entry here, the package name (as it would appear in a ``with`` clause) is given, followed by the name of the corresponding spec file in parentheses. The packages are children in four hierarchies, ``Ada``, ``Interfaces``, ``System``, and ``GNAT``, the latter being a GNAT-specific hierarchy. Note that an application program should only use packages in one of these four hierarchies if the package is defined in the Ada Reference Manual, or is listed in this section of the GNAT Programmers Reference Manual. All other units should be considered internal implementation units and should not be directly ``with``\ ed by application code. The use of a ``with`` clause that references one of these internal implementation units makes an application potentially dependent on changes in versions of GNAT, and will generate a warning message. .. _`Ada.Characters.Latin_9_(a-chlat9.ads)`: ``Ada.Characters.Latin_9`` (:file:`a-chlat9.ads`) ================================================= .. index:: Ada.Characters.Latin_9 (a-chlat9.ads) .. index:: Latin_9 constants for Character This child of ``Ada.Characters`` provides a set of definitions corresponding to those in the RM-defined package ``Ada.Characters.Latin_1`` but with the few modifications required for ``Latin-9``. The provision of such a package is specifically authorized by the Ada Reference Manual (RM A.3.3(27)). .. _`Ada.Characters.Wide_Latin_1_(a-cwila1.ads)`: ``Ada.Characters.Wide_Latin_1`` (:file:`a-cwila1.ads`) ====================================================== .. index:: Ada.Characters.Wide_Latin_1 (a-cwila1.ads) .. index:: Latin_1 constants for Wide_Character This child of ``Ada.Characters`` provides a set of definitions corresponding to those in the RM-defined package ``Ada.Characters.Latin_1`` but with the types of the constants being ``Wide_Character`` instead of ``Character``. The provision of such a package is specifically authorized by the Ada Reference Manual (RM A.3.3(27)). .. _`Ada.Characters.Wide_Latin_9_(a-cwila9.ads)`: ``Ada.Characters.Wide_Latin_9`` (:file:`a-cwila9.ads`) ====================================================== .. index:: Ada.Characters.Wide_Latin_9 (a-cwila9.ads) .. index:: Latin_9 constants for Wide_Character This child of ``Ada.Characters`` provides a set of definitions corresponding to those in the GNAT defined package ``Ada.Characters.Latin_9`` but with the types of the constants being ``Wide_Character`` instead of ``Character``. The provision of such a package is specifically authorized by the Ada Reference Manual (RM A.3.3(27)). .. _`Ada.Characters.Wide_Wide_Latin_1_(a-chzla1.ads)`: ``Ada.Characters.Wide_Wide_Latin_1`` (:file:`a-chzla1.ads`) =========================================================== .. index:: Ada.Characters.Wide_Wide_Latin_1 (a-chzla1.ads) .. index:: Latin_1 constants for Wide_Wide_Character This child of ``Ada.Characters`` provides a set of definitions corresponding to those in the RM-defined package ``Ada.Characters.Latin_1`` but with the types of the constants being ``Wide_Wide_Character`` instead of ``Character``. The provision of such a package is specifically authorized by the Ada Reference Manual (RM A.3.3(27)). .. _`Ada.Characters.Wide_Wide_Latin_9_(a-chzla9.ads)`: ``Ada.Characters.Wide_Wide_Latin_9`` (:file:`a-chzla9.ads`) =========================================================== .. index:: Ada.Characters.Wide_Wide_Latin_9 (a-chzla9.ads) .. index:: Latin_9 constants for Wide_Wide_Character This child of ``Ada.Characters`` provides a set of definitions corresponding to those in the GNAT defined package ``Ada.Characters.Latin_9`` but with the types of the constants being ``Wide_Wide_Character`` instead of ``Character``. The provision of such a package is specifically authorized by the Ada Reference Manual (RM A.3.3(27)). .. _`Ada.Containers.Bounded_Holders_(a-coboho.ads)`: ``Ada.Containers.Bounded_Holders`` (:file:`a-coboho.ads`) ========================================================= .. index:: Ada.Containers.Bounded_Holders (a-coboho.ads) .. index:: Formal container for vectors This child of ``Ada.Containers`` defines a modified version of Indefinite_Holders that avoids heap allocation. .. _`Ada.Command_Line.Environment_(a-colien.ads)`: ``Ada.Command_Line.Environment`` (:file:`a-colien.ads`) ======================================================= .. index:: Ada.Command_Line.Environment (a-colien.ads) .. index:: Environment entries This child of ``Ada.Command_Line`` provides a mechanism for obtaining environment values on systems where this concept makes sense. .. _`Ada.Command_Line.Remove_(a-colire.ads)`: ``Ada.Command_Line.Remove`` (:file:`a-colire.ads`) ================================================== .. index:: Ada.Command_Line.Remove (a-colire.ads) .. index:: Removing command line arguments .. index:: Command line, argument removal This child of ``Ada.Command_Line`` provides a mechanism for logically removing arguments from the argument list. Once removed, an argument is not visible to further calls to the subprograms in ``Ada.Command_Line``. These calls will not see the removed argument. .. _`Ada.Command_Line.Response_File_(a-clrefi.ads)`: ``Ada.Command_Line.Response_File`` (:file:`a-clrefi.ads`) ========================================================= .. index:: Ada.Command_Line.Response_File (a-clrefi.ads) .. index:: Response file for command line .. index:: Command line, response file .. index:: Command line, handling long command lines This child of ``Ada.Command_Line`` provides a mechanism facilities for getting command line arguments from a text file, called a "response file". Using a response file allow passing a set of arguments to an executable longer than the maximum allowed by the system on the command line. .. _`Ada.Direct_IO.C_Streams_(a-diocst.ads)`: ``Ada.Direct_IO.C_Streams`` (:file:`a-diocst.ads`) ================================================== .. index:: Ada.Direct_IO.C_Streams (a-diocst.ads) .. index:: C Streams, Interfacing with Direct_IO This package provides subprograms that allow interfacing between C streams and ``Direct_IO``. The stream identifier can be extracted from a file opened on the Ada side, and an Ada file can be constructed from a stream opened on the C side. .. _`Ada.Exceptions.Is_Null_Occurrence_(a-einuoc.ads)`: ``Ada.Exceptions.Is_Null_Occurrence`` (:file:`a-einuoc.ads`) ============================================================ .. index:: Ada.Exceptions.Is_Null_Occurrence (a-einuoc.ads) .. index:: Null_Occurrence, testing for This child subprogram provides a way of testing for the null exception occurrence (``Null_Occurrence``) without raising an exception. .. _`Ada.Exceptions.Last_Chance_Handler_(a-elchha.ads)`: ``Ada.Exceptions.Last_Chance_Handler`` (:file:`a-elchha.ads`) ============================================================= .. index:: Ada.Exceptions.Last_Chance_Handler (a-elchha.ads) .. index:: Null_Occurrence, testing for This child subprogram is used for handling otherwise unhandled exceptions (hence the name last chance), and perform clean ups before terminating the program. Note that this subprogram never returns. .. _`Ada.Exceptions.Traceback_(a-exctra.ads)`: ``Ada.Exceptions.Traceback`` (:file:`a-exctra.ads`) =================================================== .. index:: Ada.Exceptions.Traceback (a-exctra.ads) .. index:: Traceback for Exception Occurrence This child package provides the subprogram (``Tracebacks``) to give a traceback array of addresses based on an exception occurrence. .. _`Ada.Sequential_IO.C_Streams_(a-siocst.ads)`: ``Ada.Sequential_IO.C_Streams`` (:file:`a-siocst.ads`) ====================================================== .. index:: Ada.Sequential_IO.C_Streams (a-siocst.ads) .. index:: C Streams, Interfacing with Sequential_IO This package provides subprograms that allow interfacing between C streams and ``Sequential_IO``. The stream identifier can be extracted from a file opened on the Ada side, and an Ada file can be constructed from a stream opened on the C side. .. _`Ada.Streams.Stream_IO.C_Streams_(a-ssicst.ads)`: ``Ada.Streams.Stream_IO.C_Streams`` (:file:`a-ssicst.ads`) ========================================================== .. index:: Ada.Streams.Stream_IO.C_Streams (a-ssicst.ads) .. index:: C Streams, Interfacing with Stream_IO This package provides subprograms that allow interfacing between C streams and ``Stream_IO``. The stream identifier can be extracted from a file opened on the Ada side, and an Ada file can be constructed from a stream opened on the C side. .. _`Ada.Strings.Unbounded.Text_IO_(a-suteio.ads)`: ``Ada.Strings.Unbounded.Text_IO`` (:file:`a-suteio.ads`) ======================================================== .. index:: Ada.Strings.Unbounded.Text_IO (a-suteio.ads) .. index:: Unbounded_String, IO support .. index:: Text_IO, extensions for unbounded strings This package provides subprograms for Text_IO for unbounded strings, avoiding the necessity for an intermediate operation with ordinary strings. .. _`Ada.Strings.Wide_Unbounded.Wide_Text_IO_(a-swuwti.ads)`: ``Ada.Strings.Wide_Unbounded.Wide_Text_IO`` (:file:`a-swuwti.ads`) ================================================================== .. index:: Ada.Strings.Wide_Unbounded.Wide_Text_IO (a-swuwti.ads) .. index:: Unbounded_Wide_String, IO support .. index:: Text_IO, extensions for unbounded wide strings This package provides subprograms for Text_IO for unbounded wide strings, avoiding the necessity for an intermediate operation with ordinary wide strings. .. _`Ada.Strings.Wide_Wide_Unbounded.Wide_Wide_Text_IO_(a-szuzti.ads)`: ``Ada.Strings.Wide_Wide_Unbounded.Wide_Wide_Text_IO`` (:file:`a-szuzti.ads`) ============================================================================ .. index:: Ada.Strings.Wide_Wide_Unbounded.Wide_Wide_Text_IO (a-szuzti.ads) .. index:: Unbounded_Wide_Wide_String, IO support .. index:: Text_IO, extensions for unbounded wide wide strings This package provides subprograms for Text_IO for unbounded wide wide strings, avoiding the necessity for an intermediate operation with ordinary wide wide strings. .. _`Ada.Task_Initialization_(a-tasini.ads)`: ``Ada.Task_Initialization`` (:file:`a-tasini.ads`) ================================================== .. index:: Ada.Task_Initialization (a-tasini.ads) This package provides a way to set a global initialization handler that is automatically invoked whenever a task is activated. Handlers are parameterless procedures. Note that such a handler is only invoked for those tasks activated after the handler is set. .. _`Ada.Text_IO.C_Streams_(a-tiocst.ads)`: ``Ada.Text_IO.C_Streams`` (:file:`a-tiocst.ads`) ================================================ .. index:: Ada.Text_IO.C_Streams (a-tiocst.ads) .. index:: C Streams, Interfacing with ``Text_IO`` This package provides subprograms that allow interfacing between C streams and ``Text_IO``. The stream identifier can be extracted from a file opened on the Ada side, and an Ada file can be constructed from a stream opened on the C side. .. _`Ada.Text_IO.Reset_Standard_Files_(a-tirsfi.ads)`: ``Ada.Text_IO.Reset_Standard_Files`` (:file:`a-tirsfi.ads`) =========================================================== .. index:: Ada.Text_IO.Reset_Standard_Files (a-tirsfi.ads) .. index:: Text_IO resetting standard files This procedure is used to reset the status of the standard files used by Ada.Text_IO. This is useful in a situation (such as a restart in an embedded application) where the status of the files may change during execution (for example a standard input file may be redefined to be interactive). .. _`Ada.Wide_Characters.Unicode_(a-wichun.ads)`: ``Ada.Wide_Characters.Unicode`` (:file:`a-wichun.ads`) ====================================================== .. index:: Ada.Wide_Characters.Unicode (a-wichun.ads) .. index:: Unicode categorization, Wide_Character This package provides subprograms that allow categorization of Wide_Character values according to Unicode categories. .. _`Ada.Wide_Text_IO.C_Streams_(a-wtcstr.ads)`: ``Ada.Wide_Text_IO.C_Streams`` (:file:`a-wtcstr.ads`) ===================================================== .. index:: Ada.Wide_Text_IO.C_Streams (a-wtcstr.ads) .. index:: C Streams, Interfacing with ``Wide_Text_IO`` This package provides subprograms that allow interfacing between C streams and ``Wide_Text_IO``. The stream identifier can be extracted from a file opened on the Ada side, and an Ada file can be constructed from a stream opened on the C side. .. _`Ada.Wide_Text_IO.Reset_Standard_Files_(a-wrstfi.ads)`: ``Ada.Wide_Text_IO.Reset_Standard_Files`` (:file:`a-wrstfi.ads`) ================================================================ .. index:: Ada.Wide_Text_IO.Reset_Standard_Files (a-wrstfi.ads) .. index:: Wide_Text_IO resetting standard files This procedure is used to reset the status of the standard files used by Ada.Wide_Text_IO. This is useful in a situation (such as a restart in an embedded application) where the status of the files may change during execution (for example a standard input file may be redefined to be interactive). .. _`Ada.Wide_Wide_Characters.Unicode_(a-zchuni.ads)`: ``Ada.Wide_Wide_Characters.Unicode`` (:file:`a-zchuni.ads`) =========================================================== .. index:: Ada.Wide_Wide_Characters.Unicode (a-zchuni.ads) .. index:: Unicode categorization, Wide_Wide_Character This package provides subprograms that allow categorization of Wide_Wide_Character values according to Unicode categories. .. _`Ada.Wide_Wide_Text_IO.C_Streams_(a-ztcstr.ads)`: ``Ada.Wide_Wide_Text_IO.C_Streams`` (:file:`a-ztcstr.ads`) ========================================================== .. index:: Ada.Wide_Wide_Text_IO.C_Streams (a-ztcstr.ads) .. index:: C Streams, Interfacing with ``Wide_Wide_Text_IO`` This package provides subprograms that allow interfacing between C streams and ``Wide_Wide_Text_IO``. The stream identifier can be extracted from a file opened on the Ada side, and an Ada file can be constructed from a stream opened on the C side. .. _`Ada.Wide_Wide_Text_IO.Reset_Standard_Files_(a-zrstfi.ads)`: ``Ada.Wide_Wide_Text_IO.Reset_Standard_Files`` (:file:`a-zrstfi.ads`) ===================================================================== .. index:: Ada.Wide_Wide_Text_IO.Reset_Standard_Files (a-zrstfi.ads) .. index:: Wide_Wide_Text_IO resetting standard files This procedure is used to reset the status of the standard files used by Ada.Wide_Wide_Text_IO. This is useful in a situation (such as a restart in an embedded application) where the status of the files may change during execution (for example a standard input file may be redefined to be interactive). .. _`GNAT.Altivec_(g-altive.ads)`: ``GNAT.Altivec`` (:file:`g-altive.ads`) ======================================= .. index:: GNAT.Altivec (g-altive.ads) .. index:: AltiVec This is the root package of the GNAT AltiVec binding. It provides definitions of constants and types common to all the versions of the binding. .. _`GNAT.Altivec.Conversions_(g-altcon.ads)`: ``GNAT.Altivec.Conversions`` (:file:`g-altcon.ads`) =================================================== .. index:: GNAT.Altivec.Conversions (g-altcon.ads) .. index:: AltiVec This package provides the Vector/View conversion routines. .. _`GNAT.Altivec.Vector_Operations_(g-alveop.ads)`: ``GNAT.Altivec.Vector_Operations`` (:file:`g-alveop.ads`) ========================================================= .. index:: GNAT.Altivec.Vector_Operations (g-alveop.ads) .. index:: AltiVec This package exposes the Ada interface to the AltiVec operations on vector objects. A soft emulation is included by default in the GNAT library. The hard binding is provided as a separate package. This unit is common to both bindings. .. _`GNAT.Altivec.Vector_Types_(g-alvety.ads)`: ``GNAT.Altivec.Vector_Types`` (:file:`g-alvety.ads`) ==================================================== .. index:: GNAT.Altivec.Vector_Types (g-alvety.ads) .. index:: AltiVec This package exposes the various vector types part of the Ada binding to AltiVec facilities. .. _`GNAT.Altivec.Vector_Views_(g-alvevi.ads)`: ``GNAT.Altivec.Vector_Views`` (:file:`g-alvevi.ads`) ==================================================== .. index:: GNAT.Altivec.Vector_Views (g-alvevi.ads) .. index:: AltiVec This package provides public 'View' data types from/to which private vector representations can be converted via GNAT.Altivec.Conversions. This allows convenient access to individual vector elements and provides a simple way to initialize vector objects. .. _`GNAT.Array_Split_(g-arrspl.ads)`: ``GNAT.Array_Split`` (:file:`g-arrspl.ads`) =========================================== .. index:: GNAT.Array_Split (g-arrspl.ads) .. index:: Array splitter Useful array-manipulation routines: given a set of separators, split an array wherever the separators appear, and provide direct access to the resulting slices. .. _`GNAT.AWK_(g-awk.ads)`: ``GNAT.AWK`` (:file:`g-awk.ads`) ================================ .. index:: GNAT.AWK (g-awk.ads) .. index:: Parsing .. index:: AWK Provides AWK-like parsing functions, with an easy interface for parsing one or more files containing formatted data. The file is viewed as a database where each record is a line and a field is a data element in this line. .. _`GNAT.Binary_Search_(g-binsea.ads)`: ``GNAT.Binary_Search`` (:file:`g-binsea.ads`) ================================================ .. index:: GNAT.Binary_Search (g-binsea.ads) .. index:: Binary search Allow binary search of a sorted array (or of an array-like container; the generic does not reference the array directly). .. _`GNAT.Bind_Environment_(g-binenv.ads)`: ``GNAT.Bind_Environment`` (:file:`g-binenv.ads`) ================================================ .. index:: GNAT.Bind_Environment (g-binenv.ads) .. index:: Bind environment Provides access to key=value associations captured at bind time. These associations can be specified using the :switch:`-V` binder command line switch. .. _`GNAT.Branch_Prediction_(g-brapre.ads)`: ``GNAT.Branch_Prediction`` (:file:`g-brapre.ads`) ================================================= .. index:: GNAT.Branch_Prediction (g-brapre.ads) .. index:: Branch Prediction Provides routines giving hints to the branch predictor of the code generator. .. _`GNAT.Bounded_Buffers_(g-boubuf.ads)`: ``GNAT.Bounded_Buffers`` (:file:`g-boubuf.ads`) =============================================== .. index:: GNAT.Bounded_Buffers (g-boubuf.ads) .. index:: Parsing .. index:: Bounded Buffers Provides a concurrent generic bounded buffer abstraction. Instances are useful directly or as parts of the implementations of other abstractions, such as mailboxes. .. _`GNAT.Bounded_Mailboxes_(g-boumai.ads)`: ``GNAT.Bounded_Mailboxes`` (:file:`g-boumai.ads`) ================================================= .. index:: GNAT.Bounded_Mailboxes (g-boumai.ads) .. index:: Parsing .. index:: Mailboxes Provides a thread-safe asynchronous intertask mailbox communication facility. .. _`GNAT.Bubble_Sort_(g-bubsor.ads)`: ``GNAT.Bubble_Sort`` (:file:`g-bubsor.ads`) =========================================== .. index:: GNAT.Bubble_Sort (g-bubsor.ads) .. index:: Sorting .. index:: Bubble sort Provides a general implementation of bubble sort usable for sorting arbitrary data items. Exchange and comparison procedures are provided by passing access-to-procedure values. .. _`GNAT.Bubble_Sort_A_(g-busora.ads)`: ``GNAT.Bubble_Sort_A`` (:file:`g-busora.ads`) ============================================= .. index:: GNAT.Bubble_Sort_A (g-busora.ads) .. index:: Sorting .. index:: Bubble sort Provides a general implementation of bubble sort usable for sorting arbitrary data items. Move and comparison procedures are provided by passing access-to-procedure values. This is an older version, retained for compatibility. Usually ``GNAT.Bubble_Sort`` will be preferable. .. _`GNAT.Bubble_Sort_G_(g-busorg.ads)`: ``GNAT.Bubble_Sort_G`` (:file:`g-busorg.ads`) ============================================= .. index:: GNAT.Bubble_Sort_G (g-busorg.ads) .. index:: Sorting .. index:: Bubble sort Similar to ``Bubble_Sort_A`` except that the move and sorting procedures are provided as generic parameters, this improves efficiency, especially if the procedures can be inlined, at the expense of duplicating code for multiple instantiations. .. _`GNAT.Byte_Order_Mark_(g-byorma.ads)`: ``GNAT.Byte_Order_Mark`` (:file:`g-byorma.ads`) =============================================== .. index:: GNAT.Byte_Order_Mark (g-byorma.ads) .. index:: UTF-8 representation .. index:: Wide character representations Provides a routine which given a string, reads the start of the string to see whether it is one of the standard byte order marks (BOM's) which signal the encoding of the string. The routine includes detection of special XML sequences for various UCS input formats. .. _`GNAT.Byte_Swapping_(g-bytswa.ads)`: ``GNAT.Byte_Swapping`` (:file:`g-bytswa.ads`) ============================================= .. index:: GNAT.Byte_Swapping (g-bytswa.ads) .. index:: Byte swapping .. index:: Endianness General routines for swapping the bytes in 2-, 4-, and 8-byte quantities. Machine-specific implementations are available in some cases. .. _`GNAT.C_Time_(g-c_time.ads)`: ``GNAT.C_Time`` (:file:`g-c_time.ads`) ====================================== .. index:: GNAT.C_Time (g-c_time.ads) .. index:: Time Provides the time_t, timeval and timespec types corresponding to the C types defined by the OS, as well as various conversion functions. .. _`GNAT.Calendar_(g-calend.ads)`: ``GNAT.Calendar`` (:file:`g-calend.ads`) ======================================== .. index:: GNAT.Calendar (g-calend.ads) .. index:: Calendar Extends the facilities provided by ``Ada.Calendar`` to include handling of days of the week, an extended ``Split`` and ``Time_Of`` capability. .. _`GNAT.Calendar.Time_IO_(g-catiio.ads)`: ``GNAT.Calendar.Time_IO`` (:file:`g-catiio.ads`) ================================================ .. index:: Calendar .. index:: Time .. index:: GNAT.Calendar.Time_IO (g-catiio.ads) .. _`GNAT.CRC32_(g-crc32.ads)`: ``GNAT.CRC32`` (:file:`g-crc32.ads`) ==================================== .. index:: GNAT.CRC32 (g-crc32.ads) .. index:: CRC32 .. index:: Cyclic Redundancy Check This package implements the CRC-32 algorithm. For a full description of this algorithm see *Computation of Cyclic Redundancy Checks via Table Look-Up*, :title:`Communications of the ACM`, Vol. 31 No. 8, pp. 1008-1013, Aug. 1988. Sarwate, D.V. .. _`GNAT.Case_Util_(g-casuti.ads)`: ``GNAT.Case_Util`` (:file:`g-casuti.ads`) ========================================= .. index:: GNAT.Case_Util (g-casuti.ads) .. index:: Casing utilities .. index:: Character handling (``GNAT.Case_Util``) A set of simple routines for handling upper and lower casing of strings without the overhead of the full casing tables in ``Ada.Characters.Handling``. .. _`GNAT.CGI_(g-cgi.ads)`: ``GNAT.CGI`` (:file:`g-cgi.ads`) ================================ .. index:: GNAT.CGI (g-cgi.ads) .. index:: CGI (Common Gateway Interface) This is a package for interfacing a GNAT program with a Web server via the Common Gateway Interface (CGI). Basically this package parses the CGI parameters, which are a set of key/value pairs sent by the Web server. It builds a table whose index is the key and provides some services to deal with this table. .. _`GNAT.CGI.Cookie_(g-cgicoo.ads)`: ``GNAT.CGI.Cookie`` (:file:`g-cgicoo.ads`) ========================================== .. index:: GNAT.CGI.Cookie (g-cgicoo.ads) .. index:: CGI (Common Gateway Interface) cookie support .. index:: Cookie support in CGI This is a package to interface a GNAT program with a Web server via the Common Gateway Interface (CGI). It exports services to deal with Web cookies (piece of information kept in the Web client software). .. _`GNAT.CGI.Debug_(g-cgideb.ads)`: ``GNAT.CGI.Debug`` (:file:`g-cgideb.ads`) ========================================= .. index:: GNAT.CGI.Debug (g-cgideb.ads) .. index:: CGI (Common Gateway Interface) debugging This is a package to help debugging CGI (Common Gateway Interface) programs written in Ada. .. _`GNAT.Command_Line_(g-comlin.ads)`: ``GNAT.Command_Line`` (:file:`g-comlin.ads`) ============================================ .. index:: GNAT.Command_Line (g-comlin.ads) .. index:: Command line Provides a high level interface to ``Ada.Command_Line`` facilities, including the ability to scan for named switches with optional parameters and expand file names using wildcard notations. .. _`GNAT.Compiler_Version_(g-comver.ads)`: ``GNAT.Compiler_Version`` (:file:`g-comver.ads`) ================================================ .. index:: GNAT.Compiler_Version (g-comver.ads) .. index:: Compiler Version .. index:: Version, of compiler Provides a routine for obtaining the version of the compiler used to compile the program. More accurately this is the version of the binder used to bind the program (this will normally be the same as the version of the compiler if a consistent tool set is used to compile all units of a partition). .. _`GNAT.Ctrl_C_(g-ctrl_c.ads)`: ``GNAT.Ctrl_C`` (:file:`g-ctrl_c.ads`) ====================================== .. index:: GNAT.Ctrl_C (g-ctrl_c.ads) .. index:: Interrupt Provides a simple interface to handle Ctrl-C keyboard events. .. _`GNAT.Current_Exception_(g-curexc.ads)`: ``GNAT.Current_Exception`` (:file:`g-curexc.ads`) ================================================= .. index:: GNAT.Current_Exception (g-curexc.ads) .. index:: Current exception .. index:: Exception retrieval Provides access to information on the current exception that has been raised without the need for using the Ada 95 / Ada 2005 exception choice parameter specification syntax. This is particularly useful in simulating typical facilities for obtaining information about exceptions provided by Ada 83 compilers. .. _`GNAT.Debug_Pools_(g-debpoo.ads)`: ``GNAT.Debug_Pools`` (:file:`g-debpoo.ads`) =========================================== .. index:: GNAT.Debug_Pools (g-debpoo.ads) .. index:: Debugging .. index:: Debug pools .. index:: Memory corruption debugging Provides a debugging storage pools that helps tracking memory corruption problems. See ``The GNAT Debug_Pool Facility`` section in the :title:`GNAT User's Guide`. .. _`GNAT.Debug_Utilities_(g-debuti.ads)`: ``GNAT.Debug_Utilities`` (:file:`g-debuti.ads`) =============================================== .. index:: GNAT.Debug_Utilities (g-debuti.ads) .. index:: Debugging Provides a few useful utilities for debugging purposes, including conversion to and from string images of address values. Supports both C and Ada formats for hexadecimal literals. .. _`GNAT.Decode_String_(g-decstr.ads)`: ``GNAT.Decode_String`` (:file:`g-decstr.ads`) ============================================= .. index:: GNAT.Decode_String (g-decstr.ads) .. index:: Decoding strings .. index:: String decoding .. index:: Wide character encoding .. index:: UTF-8 .. index:: Unicode A generic package providing routines for decoding wide character and wide wide character strings encoded as sequences of 8-bit characters using a specified encoding method. Includes validation routines, and also routines for stepping to next or previous encoded character in an encoded string. Useful in conjunction with Unicode character coding. Note there is a preinstantiation for UTF-8. See next entry. .. _`GNAT.Decode_UTF8_String_(g-deutst.ads)`: ``GNAT.Decode_UTF8_String`` (:file:`g-deutst.ads`) ================================================== .. index:: GNAT.Decode_UTF8_String (g-deutst.ads) .. index:: Decoding strings .. index:: Decoding UTF-8 strings .. index:: UTF-8 string decoding .. index:: Wide character decoding .. index:: UTF-8 .. index:: Unicode A preinstantiation of GNAT.Decode_Strings for UTF-8 encoding. .. _`GNAT.Directory_Operations_(g-dirope.ads)`: ``GNAT.Directory_Operations`` (:file:`g-dirope.ads`) ==================================================== .. index:: GNAT.Directory_Operations (g-dirope.ads) .. index:: Directory operations Provides a set of routines for manipulating directories, including changing the current directory, making new directories, and scanning the files in a directory. .. _`GNAT.Directory_Operations.Iteration_(g-diopit.ads)`: ``GNAT.Directory_Operations.Iteration`` (:file:`g-diopit.ads`) ============================================================== .. index:: GNAT.Directory_Operations.Iteration (g-diopit.ads) .. index:: Directory operations iteration A child unit of GNAT.Directory_Operations providing additional operations for iterating through directories. .. _`GNAT.Dynamic_HTables_(g-dynhta.ads)`: ``GNAT.Dynamic_HTables`` (:file:`g-dynhta.ads`) =============================================== .. index:: GNAT.Dynamic_HTables (g-dynhta.ads) .. index:: Hash tables A generic implementation of hash tables that can be used to hash arbitrary data. Provided in two forms, a simple form with built in hash functions, and a more complex form in which the hash function is supplied. This package provides a facility similar to that of ``GNAT.HTable``, except that this package declares a type that can be used to define dynamic instances of the hash table, while an instantiation of ``GNAT.HTable`` creates a single instance of the hash table. .. _`GNAT.Dynamic_Tables_(g-dyntab.ads)`: ``GNAT.Dynamic_Tables`` (:file:`g-dyntab.ads`) ============================================== .. index:: GNAT.Dynamic_Tables (g-dyntab.ads) .. index:: Table implementation .. index:: Arrays, extendable A generic package providing a single dimension array abstraction where the length of the array can be dynamically modified. This package provides a facility similar to that of ``GNAT.Table``, except that this package declares a type that can be used to define dynamic instances of the table, while an instantiation of ``GNAT.Table`` creates a single instance of the table type. .. _`GNAT.Encode_String_(g-encstr.ads)`: ``GNAT.Encode_String`` (:file:`g-encstr.ads`) ============================================= .. index:: GNAT.Encode_String (g-encstr.ads) .. index:: Encoding strings .. index:: String encoding .. index:: Wide character encoding .. index:: UTF-8 .. index:: Unicode A generic package providing routines for encoding wide character and wide wide character strings as sequences of 8-bit characters using a specified encoding method. Useful in conjunction with Unicode character coding. Note there is a preinstantiation for UTF-8. See next entry. .. _`GNAT.Encode_UTF8_String_(g-enutst.ads)`: ``GNAT.Encode_UTF8_String`` (:file:`g-enutst.ads`) ================================================== .. index:: GNAT.Encode_UTF8_String (g-enutst.ads) .. index:: Encoding strings .. index:: Encoding UTF-8 strings .. index:: UTF-8 string encoding .. index:: Wide character encoding .. index:: UTF-8 .. index:: Unicode A preinstantiation of GNAT.Encode_Strings for UTF-8 encoding. .. _`GNAT.Exception_Actions_(g-excact.ads)`: ``GNAT.Exception_Actions`` (:file:`g-excact.ads`) ================================================= .. index:: GNAT.Exception_Actions (g-excact.ads) .. index:: Exception actions Provides callbacks when an exception is raised. Callbacks can be registered for specific exceptions, or when any exception is raised. This can be used for instance to force a core dump to ease debugging. .. _`GNAT.Exception_Traces_(g-exctra.ads)`: ``GNAT.Exception_Traces`` (:file:`g-exctra.ads`) ================================================ .. index:: GNAT.Exception_Traces (g-exctra.ads) .. index:: Exception traces .. index:: Debugging Provides an interface allowing to control automatic output upon exception occurrences. .. _`GNAT.Exceptions_(g-except.ads)`: ``GNAT.Exceptions`` (:file:`g-except.ads`) ========================================== .. index:: GNAT.Exceptions (g-except.ads) .. index:: Exceptions, Pure .. index:: Pure packages, exceptions Normally it is not possible to raise an exception with a message from a subprogram in a pure package, since the necessary types and subprograms are in ``Ada.Exceptions`` which is not a pure unit. ``GNAT.Exceptions`` provides a facility for getting around this limitation for a few predefined exceptions, and for example allows raising ``Constraint_Error`` with a message from a pure subprogram. .. _`GNAT.Expect_(g-expect.ads)`: ``GNAT.Expect`` (:file:`g-expect.ads`) ====================================== .. index:: GNAT.Expect (g-expect.ads) Provides a set of subprograms similar to what is available with the standard Tcl Expect tool. It allows you to easily spawn and communicate with an external process. You can send commands or inputs to the process, and compare the output with some expected regular expression. Currently ``GNAT.Expect`` is implemented on all native GNAT ports. It is not implemented for cross ports, and in particular is not implemented for VxWorks or LynxOS. .. _`GNAT.Expect.TTY_(g-exptty.ads)`: ``GNAT.Expect.TTY`` (:file:`g-exptty.ads`) ========================================== .. index:: GNAT.Expect.TTY (g-exptty.ads) As GNAT.Expect but using pseudo-terminal. Currently ``GNAT.Expect.TTY`` is implemented on all native GNAT ports. It is not implemented for cross ports, and in particular is not implemented for VxWorks or LynxOS. .. _`GNAT.Float_Control_(g-flocon.ads)`: ``GNAT.Float_Control`` (:file:`g-flocon.ads`) ============================================= .. index:: GNAT.Float_Control (g-flocon.ads) .. index:: Floating-Point Processor Provides an interface for resetting the floating-point processor into the mode required for correct semantic operation in Ada. Some third party library calls may cause this mode to be modified, and the Reset procedure in this package can be used to reestablish the required mode. .. _`GNAT.Formatted_String_(g-forstr.ads)`: ``GNAT.Formatted_String`` (:file:`g-forstr.ads`) ================================================ .. index:: GNAT.Formatted_String (g-forstr.ads) .. index:: Formatted String Provides support for C/C++ printf() formatted strings. The format is copied from the printf() routine and should therefore give identical output. Some generic routines are provided to be able to use types derived from Integer, Float or enumerations as values for the formatted string. .. _`GNAT.Generic_Fast_Math_Functions_(g-gfmafu.ads)`: ``GNAT.Generic_Fast_Math_Functions`` (:file:`g-gfmafu.ads`) =========================================================== .. index:: GNAT.Generic_Fast_Math_Functions (g-gfmafu.ads) .. index:: Mathematical functions Provides direct access to the underlying implementation of the common mathematical functions, generally from the system mathematical library. This differs from ``Ada.Numerics.Generic_Elementary_Functions`` in that the implementation may deviate from the semantics specified for these functions in the Reference Manual, for example ``Numerics.Argument_Error`` is not raised. On selected platforms, some of these functions may also have a vector implementation that can be automatically used by the compiler when auto-vectorization is enabled. .. _`GNAT.Heap_Sort_(g-heasor.ads)`: ``GNAT.Heap_Sort`` (:file:`g-heasor.ads`) ========================================= .. index:: GNAT.Heap_Sort (g-heasor.ads) .. index:: Sorting Provides a general implementation of heap sort usable for sorting arbitrary data items. Exchange and comparison procedures are provided by passing access-to-procedure values. The algorithm used is a modified heap sort that performs approximately N*log(N) comparisons in the worst case. .. _`GNAT.Heap_Sort_A_(g-hesora.ads)`: ``GNAT.Heap_Sort_A`` (:file:`g-hesora.ads`) =========================================== .. index:: GNAT.Heap_Sort_A (g-hesora.ads) .. index:: Sorting Provides a general implementation of heap sort usable for sorting arbitrary data items. Move and comparison procedures are provided by passing access-to-procedure values. The algorithm used is a modified heap sort that performs approximately N*log(N) comparisons in the worst case. This differs from ``GNAT.Heap_Sort`` in having a less convenient interface, but may be slightly more efficient. .. _`GNAT.Heap_Sort_G_(g-hesorg.ads)`: ``GNAT.Heap_Sort_G`` (:file:`g-hesorg.ads`) =========================================== .. index:: GNAT.Heap_Sort_G (g-hesorg.ads) .. index:: Sorting Similar to ``Heap_Sort_A`` except that the move and sorting procedures are provided as generic parameters, this improves efficiency, especially if the procedures can be inlined, at the expense of duplicating code for multiple instantiations. .. _`GNAT.HTable_(g-htable.ads)`: ``GNAT.HTable`` (:file:`g-htable.ads`) ====================================== .. index:: GNAT.HTable (g-htable.ads) .. index:: Hash tables A generic implementation of hash tables that can be used to hash arbitrary data. Provides two approaches, one a simple static approach, and the other allowing arbitrary dynamic hash tables. .. _`GNAT.IO_(g-io.ads)`: ``GNAT.IO`` (:file:`g-io.ads`) ============================== .. index:: GNAT.IO (g-io.ads) .. index:: Simple I/O .. index:: Input/Output facilities A simple preelaborable input-output package that provides a subset of simple Text_IO functions for reading characters and strings from Standard_Input, and writing characters, strings and integers to either Standard_Output or Standard_Error. .. _`GNAT.IO_Aux_(g-io_aux.ads)`: ``GNAT.IO_Aux`` (:file:`g-io_aux.ads`) ====================================== .. index:: GNAT.IO_Aux (g-io_aux.ads) .. index:: Text_IO .. index:: Input/Output facilities Provides some auxiliary functions for use with Text_IO, including a test for whether a file exists, and functions for reading a line of text. .. _`GNAT.Lock_Files_(g-locfil.ads)`: ``GNAT.Lock_Files`` (:file:`g-locfil.ads`) ========================================== .. index:: GNAT.Lock_Files (g-locfil.ads) .. index:: File locking .. index:: Locking using files Provides a general interface for using files as locks. Can be used for providing program level synchronization. .. _`GNAT.MBBS_Discrete_Random_(g-mbdira.ads)`: ``GNAT.MBBS_Discrete_Random`` (:file:`g-mbdira.ads`) ==================================================== .. index:: GNAT.MBBS_Discrete_Random (g-mbdira.ads) .. index:: Random number generation The original implementation of ``Ada.Numerics.Discrete_Random``. Uses a modified version of the Blum-Blum-Shub generator. .. _`GNAT.MBBS_Float_Random_(g-mbflra.ads)`: ``GNAT.MBBS_Float_Random`` (:file:`g-mbflra.ads`) ================================================= .. index:: GNAT.MBBS_Float_Random (g-mbflra.ads) .. index:: Random number generation The original implementation of ``Ada.Numerics.Float_Random``. Uses a modified version of the Blum-Blum-Shub generator. .. _`GNAT.MD5_(g-md5.ads)`: ``GNAT.MD5`` (:file:`g-md5.ads`) ================================ .. index:: GNAT.MD5 (g-md5.ads) .. index:: Message Digest MD5 Implements the MD5 Message-Digest Algorithm as described in RFC 1321, and the HMAC-MD5 message authentication function as described in RFC 2104 and FIPS PUB 198. .. _`GNAT.Memory_Dump_(g-memdum.ads)`: ``GNAT.Memory_Dump`` (:file:`g-memdum.ads`) =========================================== .. index:: GNAT.Memory_Dump (g-memdum.ads) .. index:: Dump Memory Provides a convenient routine for dumping raw memory to either the standard output or standard error files. Uses GNAT.IO for actual output. .. _`GNAT.Most_Recent_Exception_(g-moreex.ads)`: ``GNAT.Most_Recent_Exception`` (:file:`g-moreex.ads`) ===================================================== .. index:: GNAT.Most_Recent_Exception (g-moreex.ads) .. index:: Exception, obtaining most recent Provides access to the most recently raised exception. Can be used for various logging purposes, including duplicating functionality of some Ada 83 implementation dependent extensions. .. _`GNAT.OS_Lib_(g-os_lib.ads)`: ``GNAT.OS_Lib`` (:file:`g-os_lib.ads`) ====================================== .. index:: GNAT.OS_Lib (g-os_lib.ads) .. index:: Operating System interface .. index:: Spawn capability Provides a range of target independent operating system interface functions, including time/date management, file operations, subprocess management, including a portable spawn procedure, and access to environment variables and error return codes. .. _`GNAT.Perfect_Hash_Generators_(g-pehage.ads)`: ``GNAT.Perfect_Hash_Generators`` (:file:`g-pehage.ads`) ======================================================= .. index:: GNAT.Perfect_Hash_Generators (g-pehage.ads) .. index:: Hash functions Provides a generator of static minimal perfect hash functions. No collisions occur and each item can be retrieved from the table in one probe (perfect property). The hash table size corresponds to the exact size of the key set and no larger (minimal property). The key set has to be known in advance (static property). The hash functions are also order preserving. If w2 is inserted after w1 in the generator, their hashcode are in the same order. These hashing functions are very convenient for use with realtime applications. .. _`GNAT.Random_Numbers_(g-rannum.ads)`: ``GNAT.Random_Numbers`` (:file:`g-rannum.ads`) ============================================== .. index:: GNAT.Random_Numbers (g-rannum.ads) .. index:: Random number generation Provides random number capabilities which extend those available in the standard Ada library and are more convenient to use. This package is however NOT suitable for situations requiring cryptographically secure randomness. .. _`GNAT.Regexp_(g-regexp.ads)`: ``GNAT.Regexp`` (:file:`g-regexp.ads`) ====================================== .. index:: GNAT.Regexp (g-regexp.ads) .. index:: Regular expressions .. index:: Pattern matching A simple implementation of regular expressions, using a subset of regular expression syntax copied from familiar Unix style utilities. This is the simplest of the three pattern matching packages provided, and is particularly suitable for 'file globbing' applications. .. _`GNAT.Registry_(g-regist.ads)`: ``GNAT.Registry`` (:file:`g-regist.ads`) ======================================== .. index:: GNAT.Registry (g-regist.ads) .. index:: Windows Registry This is a high level binding to the Windows registry. It is possible to do simple things like reading a key value, creating a new key. For full registry API, but at a lower level of abstraction, refer to the Win32.Winreg package provided with the Win32Ada binding .. _`GNAT.Regpat_(g-regpat.ads)`: ``GNAT.Regpat`` (:file:`g-regpat.ads`) ====================================== .. index:: GNAT.Regpat (g-regpat.ads) .. index:: Regular expressions .. index:: Pattern matching A complete implementation of Unix-style regular expression matching, copied from the original V7 style regular expression library written in C by Henry Spencer (and binary compatible with this C library). .. _`GNAT.Rewrite_Data_(g-rewdat.ads)`: ``GNAT.Rewrite_Data`` (:file:`g-rewdat.ads`) ============================================ .. index:: GNAT.Rewrite_Data (g-rewdat.ads) .. index:: Rewrite data A unit to rewrite on-the-fly string occurrences in a stream of data. The implementation has a very minimal memory footprint as the full content to be processed is not loaded into memory all at once. This makes this interface usable for large files or socket streams. .. _`GNAT.Secondary_Stack_Info_(g-sestin.ads)`: ``GNAT.Secondary_Stack_Info`` (:file:`g-sestin.ads`) ==================================================== .. index:: GNAT.Secondary_Stack_Info (g-sestin.ads) .. index:: Secondary Stack Info Provides the capability to query the high water mark of the current task's secondary stack. .. _`GNAT.Semaphores_(g-semaph.ads)`: ``GNAT.Semaphores`` (:file:`g-semaph.ads`) ========================================== .. index:: GNAT.Semaphores (g-semaph.ads) .. index:: Semaphores Provides classic counting and binary semaphores using protected types. .. _`GNAT.Serial_Communications_(g-sercom.ads)`: ``GNAT.Serial_Communications`` (:file:`g-sercom.ads`) ===================================================== .. index:: GNAT.Serial_Communications (g-sercom.ads) .. index:: Serial_Communications Provides a simple interface to send and receive data over a serial port. This is only supported on GNU/Linux and Windows. .. _`GNAT.SHA1_(g-sha1.ads)`: ``GNAT.SHA1`` (:file:`g-sha1.ads`) ================================== .. index:: GNAT.SHA1 (g-sha1.ads) .. index:: Secure Hash Algorithm SHA-1 Implements the SHA-1 Secure Hash Algorithm as described in FIPS PUB 180-3 and RFC 3174, and the HMAC-SHA1 message authentication function as described in RFC 2104 and FIPS PUB 198. .. _`GNAT.SHA224_(g-sha224.ads)`: ``GNAT.SHA224`` (:file:`g-sha224.ads`) ====================================== .. index:: GNAT.SHA224 (g-sha224.ads) .. index:: Secure Hash Algorithm SHA-224 Implements the SHA-224 Secure Hash Algorithm as described in FIPS PUB 180-3, and the HMAC-SHA224 message authentication function as described in RFC 2104 and FIPS PUB 198. .. _`GNAT.SHA256_(g-sha256.ads)`: ``GNAT.SHA256`` (:file:`g-sha256.ads`) ====================================== .. index:: GNAT.SHA256 (g-sha256.ads) .. index:: Secure Hash Algorithm SHA-256 Implements the SHA-256 Secure Hash Algorithm as described in FIPS PUB 180-3, and the HMAC-SHA256 message authentication function as described in RFC 2104 and FIPS PUB 198. .. _`GNAT.SHA384_(g-sha384.ads)`: ``GNAT.SHA384`` (:file:`g-sha384.ads`) ====================================== .. index:: GNAT.SHA384 (g-sha384.ads) .. index:: Secure Hash Algorithm SHA-384 Implements the SHA-384 Secure Hash Algorithm as described in FIPS PUB 180-3, and the HMAC-SHA384 message authentication function as described in RFC 2104 and FIPS PUB 198. .. _`GNAT.SHA512_(g-sha512.ads)`: ``GNAT.SHA512`` (:file:`g-sha512.ads`) ====================================== .. index:: GNAT.SHA512 (g-sha512.ads) .. index:: Secure Hash Algorithm SHA-512 Implements the SHA-512 Secure Hash Algorithm as described in FIPS PUB 180-3, and the HMAC-SHA512 message authentication function as described in RFC 2104 and FIPS PUB 198. .. _`GNAT.Signals_(g-signal.ads)`: ``GNAT.Signals`` (:file:`g-signal.ads`) ======================================= .. index:: GNAT.Signals (g-signal.ads) .. index:: Signals Provides the ability to manipulate the blocked status of signals on supported targets. .. _`GNAT.Sockets_(g-socket.ads)`: ``GNAT.Sockets`` (:file:`g-socket.ads`) ======================================= .. index:: GNAT.Sockets (g-socket.ads) .. index:: Sockets A high level and portable interface to develop sockets based applications. This package is based on the sockets thin binding found in ``GNAT.Sockets.Thin``. Currently ``GNAT.Sockets`` is implemented on all native GNAT ports and on VxWorks cross ports. It is not implemented for the LynxOS cross port. .. _`GNAT.Source_Info_(g-souinf.ads)`: ``GNAT.Source_Info`` (:file:`g-souinf.ads`) =========================================== .. index:: GNAT.Source_Info (g-souinf.ads) .. index:: Source Information Provides subprograms that give access to source code information known at compile time, such as the current file name and line number. Also provides subprograms yielding the date and time of the current compilation (like the C macros ``__DATE__`` and ``__TIME__``) .. _`GNAT.Spelling_Checker_(g-speche.ads)`: ``GNAT.Spelling_Checker`` (:file:`g-speche.ads`) ================================================ .. index:: GNAT.Spelling_Checker (g-speche.ads) .. index:: Spell checking Provides a function for determining whether one string is a plausible near misspelling of another string. .. _`GNAT.Spelling_Checker_Generic_(g-spchge.ads)`: ``GNAT.Spelling_Checker_Generic`` (:file:`g-spchge.ads`) ======================================================== .. index:: GNAT.Spelling_Checker_Generic (g-spchge.ads) .. index:: Spell checking Provides a generic function that can be instantiated with a string type for determining whether one string is a plausible near misspelling of another string. .. _`GNAT.Spitbol.Patterns_(g-spipat.ads)`: ``GNAT.Spitbol.Patterns`` (:file:`g-spipat.ads`) ================================================ .. index:: GNAT.Spitbol.Patterns (g-spipat.ads) .. index:: SPITBOL pattern matching .. index:: Pattern matching A complete implementation of SNOBOL4 style pattern matching. This is the most elaborate of the pattern matching packages provided. It fully duplicates the SNOBOL4 dynamic pattern construction and matching capabilities, using the efficient algorithm developed by Robert Dewar for the SPITBOL system. .. _`GNAT.Spitbol_(g-spitbo.ads)`: ``GNAT.Spitbol`` (:file:`g-spitbo.ads`) ======================================= .. index:: GNAT.Spitbol (g-spitbo.ads) .. index:: SPITBOL interface The top level package of the collection of SPITBOL-style functionality, this package provides basic SNOBOL4 string manipulation functions, such as Pad, Reverse, Trim, Substr capability, as well as a generic table function useful for constructing arbitrary mappings from strings in the style of the SNOBOL4 TABLE function. .. _`GNAT.Spitbol.Table_Boolean_(g-sptabo.ads)`: ``GNAT.Spitbol.Table_Boolean`` (:file:`g-sptabo.ads`) ===================================================== .. index:: GNAT.Spitbol.Table_Boolean (g-sptabo.ads) .. index:: Sets of strings .. index:: SPITBOL Tables A library level of instantiation of ``GNAT.Spitbol.Patterns.Table`` for type ``Standard.Boolean``, giving an implementation of sets of string values. .. _`GNAT.Spitbol.Table_Integer_(g-sptain.ads)`: ``GNAT.Spitbol.Table_Integer`` (:file:`g-sptain.ads`) ===================================================== .. index:: GNAT.Spitbol.Table_Integer (g-sptain.ads) .. index:: Integer maps .. index:: Maps .. index:: SPITBOL Tables A library level of instantiation of ``GNAT.Spitbol.Patterns.Table`` for type ``Standard.Integer``, giving an implementation of maps from string to integer values. .. _`GNAT.Spitbol.Table_VString_(g-sptavs.ads)`: ``GNAT.Spitbol.Table_VString`` (:file:`g-sptavs.ads`) ===================================================== .. index:: GNAT.Spitbol.Table_VString (g-sptavs.ads) .. index:: String maps .. index:: Maps .. index:: SPITBOL Tables A library level of instantiation of ``GNAT.Spitbol.Patterns.Table`` for a variable length string type, giving an implementation of general maps from strings to strings. .. _`GNAT.SSE_(g-sse.ads)`: ``GNAT.SSE`` (:file:`g-sse.ads`) ================================ .. index:: GNAT.SSE (g-sse.ads) Root of a set of units aimed at offering Ada bindings to a subset of the Intel(r) Streaming SIMD Extensions with GNAT on the x86 family of targets. It exposes vector component types together with a general introduction to the binding contents and use. .. _`GNAT.SSE.Vector_Types_(g-ssvety.ads)`: ``GNAT.SSE.Vector_Types`` (:file:`g-ssvety.ads`) ================================================ .. index:: GNAT.SSE.Vector_Types (g-ssvety.ads) SSE vector types for use with SSE related intrinsics. .. _`GNAT.String_Hash(g-strhas.ads)`: ``GNAT.String_Hash`` (:file:`g-strhas.ads`) =========================================== .. index:: GNAT.String_Hash (g-strhas.ads) .. index:: Hash functions Provides a generic hash function working on arrays of scalars. Both the scalar type and the hash result type are parameters. .. _`GNAT.Strings_(g-string.ads)`: ``GNAT.Strings`` (:file:`g-string.ads`) ======================================= .. index:: GNAT.Strings (g-string.ads) Common String access types and related subprograms. Basically it defines a string access and an array of string access types. .. _`GNAT.String_Split_(g-strspl.ads)`: ``GNAT.String_Split`` (:file:`g-strspl.ads`) ============================================ .. index:: GNAT.String_Split (g-strspl.ads) .. index:: String splitter Useful string manipulation routines: given a set of separators, split a string wherever the separators appear, and provide direct access to the resulting slices. This package is instantiated from ``GNAT.Array_Split``. .. _`GNAT.Table_(g-table.ads)`: ``GNAT.Table`` (:file:`g-table.ads`) ==================================== .. index:: GNAT.Table (g-table.ads) .. index:: Table implementation .. index:: Arrays, extendable A generic package providing a single dimension array abstraction where the length of the array can be dynamically modified. This package provides a facility similar to that of ``GNAT.Dynamic_Tables``, except that this package declares a single instance of the table type, while an instantiation of ``GNAT.Dynamic_Tables`` creates a type that can be used to define dynamic instances of the table. .. _`GNAT.Task_Lock_(g-tasloc.ads)`: ``GNAT.Task_Lock`` (:file:`g-tasloc.ads`) ========================================= .. index:: GNAT.Task_Lock (g-tasloc.ads) .. index:: Task synchronization .. index:: Task locking .. index:: Locking A very simple facility for locking and unlocking sections of code using a single global task lock. Appropriate for use in situations where contention between tasks is very rarely expected. .. _`GNAT.Time_Stamp_(g-timsta.ads)`: ``GNAT.Time_Stamp`` (:file:`g-timsta.ads`) ========================================== .. index:: GNAT.Time_Stamp (g-timsta.ads) .. index:: Time stamp .. index:: Current time Provides a simple function that returns a string YYYY-MM-DD HH:MM:SS.SS that represents the current date and time in ISO 8601 format. This is a very simple routine with minimal code and there are no dependencies on any other unit. .. _`GNAT.Threads_(g-thread.ads)`: ``GNAT.Threads`` (:file:`g-thread.ads`) ======================================= .. index:: GNAT.Threads (g-thread.ads) .. index:: Foreign threads .. index:: Threads, foreign Provides facilities for dealing with foreign threads which need to be known by the GNAT run-time system. Consult the documentation of this package for further details if your program has threads that are created by a non-Ada environment which then accesses Ada code. .. _`GNAT.Traceback_(g-traceb.ads)`: ``GNAT.Traceback`` (:file:`g-traceb.ads`) ========================================= .. index:: GNAT.Traceback (g-traceb.ads) .. index:: Trace back facilities Provides a facility for obtaining non-symbolic traceback information, useful in various debugging situations. .. _`GNAT.Traceback.Symbolic_(g-trasym.ads)`: ``GNAT.Traceback.Symbolic`` (:file:`g-trasym.ads`) ================================================== .. index:: GNAT.Traceback.Symbolic (g-trasym.ads) .. index:: Trace back facilities .. _`GNAT.UTF_32_(g-utf_32.ads)`: ``GNAT.UTF_32`` (:file:`g-utf_32.ads`) ====================================== .. index:: GNAT.UTF_32 (g-utf_32.ads) .. index:: Wide character codes This is a package intended to be used in conjunction with the ``Wide_Character`` type in Ada 95 and the ``Wide_Wide_Character`` type in Ada 2005 (available in ``GNAT`` in Ada 2005 mode). This package contains Unicode categorization routines, as well as lexical categorization routines corresponding to the Ada 2005 lexical rules for identifiers and strings, and also a lower case to upper case fold routine corresponding to the Ada 2005 rules for identifier equivalence. .. _`GNAT.UTF_32_Spelling_Checker_(g-u3spch.ads)`: ``GNAT.UTF_32_Spelling_Checker`` (:file:`g-u3spch.ads`) ======================================================= .. index:: GNAT.UTF_32_Spelling_Checker (g-u3spch.ads) .. index:: Spell checking Provides a function for determining whether one wide wide string is a plausible near misspelling of another wide wide string, where the strings are represented using the UTF_32_String type defined in System.Wch_Cnv. .. _`GNAT.Wide_Spelling_Checker_(g-wispch.ads)`: ``GNAT.Wide_Spelling_Checker`` (:file:`g-wispch.ads`) ===================================================== .. index:: GNAT.Wide_Spelling_Checker (g-wispch.ads) .. index:: Spell checking Provides a function for determining whether one wide string is a plausible near misspelling of another wide string. .. _`GNAT.Wide_String_Split_(g-wistsp.ads)`: ``GNAT.Wide_String_Split`` (:file:`g-wistsp.ads`) ================================================= .. index:: GNAT.Wide_String_Split (g-wistsp.ads) .. index:: Wide_String splitter Useful wide string manipulation routines: given a set of separators, split a wide string wherever the separators appear, and provide direct access to the resulting slices. This package is instantiated from ``GNAT.Array_Split``. .. _`GNAT.Wide_Wide_Spelling_Checker_(g-zspche.ads)`: ``GNAT.Wide_Wide_Spelling_Checker`` (:file:`g-zspche.ads`) ========================================================== .. index:: GNAT.Wide_Wide_Spelling_Checker (g-zspche.ads) .. index:: Spell checking Provides a function for determining whether one wide wide string is a plausible near misspelling of another wide wide string. .. _`GNAT.Wide_Wide_String_Split_(g-zistsp.ads)`: ``GNAT.Wide_Wide_String_Split`` (:file:`g-zistsp.ads`) ====================================================== .. index:: GNAT.Wide_Wide_String_Split (g-zistsp.ads) .. index:: Wide_Wide_String splitter Useful wide wide string manipulation routines: given a set of separators, split a wide wide string wherever the separators appear, and provide direct access to the resulting slices. This package is instantiated from ``GNAT.Array_Split``. .. _`Interfaces.C.Extensions_(i-cexten.ads)`: ``Interfaces.C.Extensions`` (:file:`i-cexten.ads`) ================================================== .. index:: Interfaces.C.Extensions (i-cexten.ads) This package contains additional C-related definitions, intended for use with either manually or automatically generated bindings to C libraries. .. _`Interfaces.C.Streams_(i-cstrea.ads)`: ``Interfaces.C.Streams`` (:file:`i-cstrea.ads`) =============================================== .. index:: Interfaces.C.Streams (i-cstrea.ads) .. index:: C streams, interfacing This package is a binding for the most commonly used operations on C streams. .. _`Interfaces.Packed_Decimal_(i-pacdec.ads)`: ``Interfaces.Packed_Decimal`` (:file:`i-pacdec.ads`) ==================================================== .. index:: Interfaces.Packed_Decimal (i-pacdec.ads) .. index:: IBM Packed Format .. index:: Packed Decimal This package provides a set of routines for conversions to and from a packed decimal format compatible with that used on IBM mainframes. .. _`Interfaces.VxWorks_(i-vxwork.ads)`: ``Interfaces.VxWorks`` (:file:`i-vxwork.ads`) ============================================= .. index:: Interfaces.VxWorks (i-vxwork.ads) .. index:: Interfacing to VxWorks .. index:: VxWorks, interfacing This package provides a limited binding to the VxWorks API. .. _`Interfaces.VxWorks.IO_(i-vxwoio.ads)`: ``Interfaces.VxWorks.IO`` (:file:`i-vxwoio.ads`) ================================================ .. index:: Interfaces.VxWorks.IO (i-vxwoio.ads) .. index:: Interfacing to VxWorks' I/O .. index:: VxWorks, I/O interfacing .. index:: VxWorks, Get_Immediate .. index:: Get_Immediate, VxWorks This package provides a binding to the ioctl (IO/Control) function of VxWorks, defining a set of option values and function codes. A particular use of this package is to enable the use of Get_Immediate under VxWorks. .. _`System.Address_Image_(s-addima.ads)`: ``System.Address_Image`` (:file:`s-addima.ads`) =============================================== .. index:: System.Address_Image (s-addima.ads) .. index:: Address image .. index:: Image, of an address This function provides a useful debugging function that gives an (implementation dependent) string which identifies an address. .. _`System.Assertions_(s-assert.ads)`: ``System.Assertions`` (:file:`s-assert.ads`) ============================================ .. index:: System.Assertions (s-assert.ads) .. index:: Assertions .. index:: Assert_Failure, exception This package provides the declaration of the exception raised by an run-time assertion failure, as well as the routine that is used internally to raise this assertion. .. _`System.Atomic_Counters_(s-atocou.ads)`: ``System.Atomic_Counters`` (:file:`s-atocou.ads`) ================================================= .. index:: System.Atomic_Counters (s-atocou.ads) This package provides the declaration of an atomic counter type, together with efficient routines (using hardware synchronization primitives) for incrementing, decrementing, and testing of these counters. This package is implemented on most targets, including all Alpha, AARCH64, ARM, ia64, PowerPC, SPARC V9, x86, and x86_64 platforms. .. _`System.Memory_(s-memory.ads)`: ``System.Memory`` (:file:`s-memory.ads`) ======================================== .. index:: System.Memory (s-memory.ads) .. index:: Memory allocation This package provides the interface to the low level routines used by the generated code for allocation and freeing storage for the default storage pool (analogous to the C routines malloc and free). It also provides a reallocation interface analogous to the C routine realloc. The body of this unit may be modified to provide alternative allocation mechanisms for the default pool, and in addition, direct calls to this unit may be made for low level allocation uses (for example see the body of ``GNAT.Tables``). .. _`System.Multiprocessors_(s-multip.ads)`: ``System.Multiprocessors`` (:file:`s-multip.ads`) ================================================= .. index:: System.Multiprocessors (s-multip.ads) .. index:: Multiprocessor interface This is an Ada 2012 unit defined in the Ada 2012 Reference Manual, but in GNAT we also make it available in Ada 95 and Ada 2005 (where it is technically an implementation-defined addition). .. _`System.Multiprocessors.Dispatching_Domains_(s-mudido.ads)`: ``System.Multiprocessors.Dispatching_Domains`` (:file:`s-mudido.ads`) ===================================================================== .. index:: System.Multiprocessors.Dispatching_Domains (s-mudido.ads) .. index:: Multiprocessor interface This is an Ada 2012 unit defined in the Ada 2012 Reference Manual, but in GNAT we also make it available in Ada 95 and Ada 2005 (where it is technically an implementation-defined addition). .. _`System.Partition_Interface_(s-parint.ads)`: ``System.Partition_Interface`` (:file:`s-parint.ads`) ===================================================== .. index:: System.Partition_Interface (s-parint.ads) .. index:: Partition interfacing functions This package provides facilities for partition interfacing. It is used primarily in a distribution context when using Annex E with ``PolyORB``. **NB!** See the note in :ref:`polyORB` regarding the lifetime of this product. .. _`System.Pool_Global_(s-pooglo.ads)`: ``System.Pool_Global`` (:file:`s-pooglo.ads`) ============================================= .. index:: System.Pool_Global (s-pooglo.ads) .. index:: Storage pool, global .. index:: Global storage pool This package provides a storage pool that is equivalent to the default storage pool used for access types for which no pool is specifically declared. It uses malloc/free to allocate/free and does not attempt to do any automatic reclamation. .. _`System.Pool_Local_(s-pooloc.ads)`: ``System.Pool_Local`` (:file:`s-pooloc.ads`) ============================================ .. index:: System.Pool_Local (s-pooloc.ads) .. index:: Storage pool, local .. index:: Local storage pool This package provides a storage pool that is intended for use with locally defined access types. It uses malloc/free for allocate/free, and maintains a list of allocated blocks, so that all storage allocated for the pool can be freed automatically when the pool is finalized. .. _`System.Restrictions_(s-restri.ads)`: ``System.Restrictions`` (:file:`s-restri.ads`) ============================================== .. index:: System.Restrictions (s-restri.ads) .. index:: Run-time restrictions access This package provides facilities for accessing at run time the status of restrictions specified at compile time for the partition. Information is available both with regard to actual restrictions specified, and with regard to compiler determined information on which restrictions are violated by one or more packages in the partition. .. _`System.Rident_(s-rident.ads)`: ``System.Rident`` (:file:`s-rident.ads`) ======================================== .. index:: System.Rident (s-rident.ads) .. index:: Restrictions definitions This package provides definitions of the restrictions identifiers supported by GNAT, and also the format of the restrictions provided in package System.Restrictions. It is not normally necessary to ``with`` this generic package since the necessary instantiation is included in package System.Restrictions. .. _`System.Strings.Stream_Ops_(s-ststop.ads)`: ``System.Strings.Stream_Ops`` (:file:`s-ststop.ads`) ==================================================== .. index:: System.Strings.Stream_Ops (s-ststop.ads) .. index:: Stream operations .. index:: String stream operations This package provides a set of stream subprograms for standard string types. It is intended primarily to support implicit use of such subprograms when stream attributes are applied to string types, but the subprograms in this package can be used directly by application programs. .. _`System.Unsigned_Types_(s-unstyp.ads)`: ``System.Unsigned_Types`` (:file:`s-unstyp.ads`) ================================================ .. index:: System.Unsigned_Types (s-unstyp.ads) This package contains definitions of standard unsigned types that correspond in size to the standard signed types declared in Standard, and (unlike the types in Interfaces) have corresponding names. It also contains some related definitions for other specialized types used by the compiler in connection with packed array types. .. _`System.Wch_Cnv_(s-wchcnv.ads)`: ``System.Wch_Cnv`` (:file:`s-wchcnv.ads`) ========================================= .. index:: System.Wch_Cnv (s-wchcnv.ads) .. index:: Wide Character, Representation .. index:: Wide String, Conversion .. index:: Representation of wide characters This package provides routines for converting between wide and wide wide characters and a representation as a value of type ``Standard.String``, using a specified wide character encoding method. It uses definitions in package ``System.Wch_Con``. .. _`System.Wch_Con_(s-wchcon.ads)`: ``System.Wch_Con`` (:file:`s-wchcon.ads`) ========================================= .. index:: System.Wch_Con (s-wchcon.ads) This package provides definitions and descriptions of the various methods used for encoding wide characters in ordinary strings. These definitions are used by the package ``System.Wch_Cnv``. --- ### Gcc Ada Doc Gnat Rm The Implementation Of Standard I O (`gcc/ada/doc/gnat_rm/the_implementation_of_standard_i_o.rst`) .. _The_Implementation_of_Standard_I/O: ********************************** The Implementation of Standard I/O ********************************** GNAT implements all the required input-output facilities described in A.6 through A.14. These sections of the Ada Reference Manual describe the required behavior of these packages from the Ada point of view, and if you are writing a portable Ada program that does not need to know the exact manner in which Ada maps to the outside world when it comes to reading or writing external files, then you do not need to read this chapter. As long as your files are all regular files (not pipes or devices), and as long as you write and read the files only from Ada, the description in the Ada Reference Manual is sufficient. However, if you want to do input-output to pipes or other devices, such as the keyboard or screen, or if the files you are dealing with are either generated by some other language, or to be read by some other language, then you need to know more about the details of how the GNAT implementation of these input-output facilities behaves. In this chapter we give a detailed description of exactly how GNAT interfaces to the file system. As always, the sources of the system are available to you for answering questions at an even more detailed level, but for most purposes the information in this chapter will suffice. Another reason that you may need to know more about how input-output is implemented arises when you have a program written in mixed languages where, for example, files are shared between the C and Ada sections of the same program. GNAT provides some additional facilities, in the form of additional child library packages, that facilitate this sharing, and these additional facilities are also described in this chapter. .. _Standard_I/O_Packages: Standard I/O Packages ===================== The Standard I/O packages described in Annex A for * Ada.Text_IO * Ada.Text_IO.Complex_IO * Ada.Text_IO.Text_Streams * Ada.Wide_Text_IO * Ada.Wide_Text_IO.Complex_IO * Ada.Wide_Text_IO.Text_Streams * Ada.Wide_Wide_Text_IO * Ada.Wide_Wide_Text_IO.Complex_IO * Ada.Wide_Wide_Text_IO.Text_Streams * Ada.Stream_IO * Ada.Sequential_IO * Ada.Direct_IO are implemented using the C library streams facility; where * All files are opened using ``fopen``. * All input/output operations use ``fread``/``fwrite``. There is no internal buffering of any kind at the Ada library level. The only buffering is that provided at the system level in the implementation of the library routines that support streams. This facilitates shared use of these streams by mixed language programs. Note though that system level buffering is explicitly enabled at elaboration of the standard I/O packages and that can have an impact on mixed language programs, in particular those using I/O before calling the Ada elaboration routine (e.g., adainit). It is recommended to call the Ada elaboration routine before performing any I/O or when impractical, flush the common I/O streams and in particular Standard_Output before elaborating the Ada code. .. _FORM_Strings: FORM Strings ============ The format of a FORM string in GNAT is: :: "keyword=value,keyword=value,...,keyword=value" where letters may be in upper or lower case, and there are no spaces between values. The order of the entries is not important. Currently the following keywords defined. :: TEXT_TRANSLATION=[YES|NO|TEXT|BINARY|U8TEXT|WTEXT|U16TEXT] SHARED=[YES|NO] WCEM=[n|h|u|s|e|8|b] ENCODING=[UTF8|8BITS] The use of these parameters is described later in this section. If an unrecognized keyword appears in a form string, it is silently ignored and not considered invalid. .. _Direct_IO: Direct_IO ========= Direct_IO can only be instantiated for definite types. This is a restriction of the Ada language, which means that the records are fixed length (the length being determined by ``type'Size``, rounded up to the next storage unit boundary if necessary). The records of a Direct_IO file are simply written to the file in index sequence, with the first record starting at offset zero, and subsequent records following. There is no control information of any kind. For example, if 32-bit integers are being written, each record takes 4-bytes, so the record at index ``K`` starts at offset ``(K-1)*4``. There is no limit on the size of Direct_IO files, they are expanded as necessary to accommodate whatever records are written to the file. .. _Sequential_IO: Sequential_IO ============= Sequential_IO may be instantiated with either a definite (constrained) or indefinite (unconstrained) type. For the definite type case, the elements written to the file are simply the memory images of the data values with no control information of any kind. The resulting file should be read using the same type, no validity checking is performed on input. For the indefinite type case, the elements written consist of two parts. First is the size of the data item, written as the memory image of a ``Interfaces.C.size_t`` value, followed by the memory image of the data value. The resulting file can only be read using the same (unconstrained) type. Normal assignment checks are performed on these read operations, and if these checks fail, ``Data_Error`` is raised. In particular, in the array case, the lengths must match, and in the variant record case, if the variable for a particular read operation is constrained, the discriminants must match. Note that it is not possible to use Sequential_IO to write variable length array items, and then read the data back into different length arrays. For example, the following will raise ``Data_Error``: .. code-block:: ada package IO is new Sequential_IO (String); F : IO.File_Type; S : String (1..4); ... IO.Create (F) IO.Write (F, "hello!") IO.Reset (F, Mode=>In_File); IO.Read (F, S); Put_Line (S); On some Ada implementations, this will print ``hell``, but the program is clearly incorrect, since there is only one element in the file, and that element is the string ``hello!``. In Ada 95 and Ada 2005, this kind of behavior can be legitimately achieved using Stream_IO, and this is the preferred mechanism. In particular, the above program fragment rewritten to use Stream_IO will work correctly. .. _Text_IO: Text_IO ======= Text_IO files consist of a stream of characters containing the following special control characters: :: LF (line feed, 16#0A#) Line Mark FF (form feed, 16#0C#) Page Mark A canonical Text_IO file is defined as one in which the following conditions are met: * The character ``LF`` is used only as a line mark, i.e., to mark the end of the line. * The character ``FF`` is used only as a page mark, i.e., to mark the end of a page and consequently can appear only immediately following a ``LF`` (line mark) character. * The file ends with either ``LF`` (line mark) or ``LF``-`FF` (line mark, page mark). In the former case, the page mark is implicitly assumed to be present. A file written using Text_IO will be in canonical form provided that no explicit ``LF`` or ``FF`` characters are written using ``Put`` or ``Put_Line``. There will be no ``FF`` character at the end of the file unless an explicit ``New_Page`` operation was performed before closing the file. A canonical Text_IO file that is a regular file (i.e., not a device or a pipe) can be read using any of the routines in Text_IO. The semantics in this case will be exactly as defined in the Ada Reference Manual, and all the routines in Text_IO are fully implemented. A text file that does not meet the requirements for a canonical Text_IO file has one of the following: * The file contains ``FF`` characters not immediately following a ``LF`` character. * The file contains ``LF`` or ``FF`` characters written by ``Put`` or ``Put_Line``, which are not logically considered to be line marks or page marks. * The file ends in a character other than ``LF`` or ``FF``, i.e., there is no explicit line mark or page mark at the end of the file. Text_IO can be used to read such non-standard text files but subprograms to do with line or page numbers do not have defined meanings. In particular, a ``FF`` character that does not follow a ``LF`` character may or may not be treated as a page mark from the point of view of page and line numbering. Every ``LF`` character is considered to end a line, and there is an implied ``LF`` character at the end of the file. .. _Stream_Pointer_Positioning: Stream Pointer Positioning -------------------------- ``Ada.Text_IO`` has a definition of current position for a file that is being read. No internal buffering occurs in Text_IO, and usually the physical position in the stream used to implement the file corresponds to this logical position defined by Text_IO. There are two exceptions: * After a call to ``End_Of_Page`` that returns ``True``, the stream is positioned past the ``LF`` (line mark) that precedes the page mark. Text_IO maintains an internal flag so that subsequent read operations properly handle the logical position which is unchanged by the ``End_Of_Page`` call. * After a call to ``End_Of_File`` that returns ``True``, if the Text_IO file was positioned before the line mark at the end of file before the call, then the logical position is unchanged, but the stream is physically positioned right at the end of file (past the line mark, and past a possible page mark following the line mark. Again Text_IO maintains internal flags so that subsequent read operations properly handle the logical position. These discrepancies have no effect on the observable behavior of Text_IO, but if a single Ada stream is shared between a C program and Ada program, or shared (using ``shared=yes`` in the form string) between two Ada files, then the difference may be observable in some situations. .. _Reading_and_Writing_Non-Regular_Files: Reading and Writing Non-Regular Files ------------------------------------- A non-regular file is a device (such as a keyboard), or a pipe. Text_IO can be used for reading and writing. Writing is not affected and the sequence of characters output is identical to the normal file case, but for reading, the behavior of Text_IO is modified to avoid undesirable look-ahead as follows: An input file that is not a regular file is considered to have no page marks. Any ``Ascii.FF`` characters (the character normally used for a page mark) appearing in the file are considered to be data characters. In particular: * ``Get_Line`` and ``Skip_Line`` do not test for a page mark following a line mark. If a page mark appears, it will be treated as a data character. * This avoids the need to wait for an extra character to be typed or entered from the pipe to complete one of these operations. * ``End_Of_Page`` always returns ``False`` * ``End_Of_File`` will return ``False`` if there is a page mark at the end of the file. Output to non-regular files is the same as for regular files. Page marks may be written to non-regular files using ``New_Page``, but as noted above they will not be treated as page marks on input if the output is piped to another Ada program. Another important discrepancy when reading non-regular files is that the end of file indication is not 'sticky'. If an end of file is entered, e.g., by pressing the :kbd:`EOT` key, then end of file is signaled once (i.e., the test ``End_Of_File`` will yield ``True``, or a read will raise ``End_Error``), but then reading can resume to read data past that end of file indication, until another end of file indication is entered. .. _Get_Immediate: Get_Immediate ------------- .. index:: Get_Immediate Get_Immediate returns the next character (including control characters) from the input file. In particular, Get_Immediate will return LF or FF characters used as line marks or page marks. Such operations leave the file positioned past the control character, and it is thus not treated as having its normal function. This means that page, line and column counts after this kind of Get_Immediate call are set as though the mark did not occur. In the case where a Get_Immediate leaves the file positioned between the line mark and page mark (which is not normally possible), it is undefined whether the FF character will be treated as a page mark. .. _Treating_Text_IO_Files_as_Streams: Treating Text_IO Files as Streams --------------------------------- .. index:: Stream files The package ``Text_IO.Streams`` allows a ``Text_IO`` file to be treated as a stream. Data written to a ``Text_IO`` file in this stream mode is binary data. If this binary data contains bytes 16#0A# (``LF``) or 16#0C# (``FF``), the resulting file may have non-standard format. Similarly if read operations are used to read from a Text_IO file treated as a stream, then ``LF`` and ``FF`` characters may be skipped and the effect is similar to that described above for ``Get_Immediate``. .. _Text_IO_Extensions: Text_IO Extensions ------------------ .. index:: Text_IO extensions A package GNAT.IO_Aux in the GNAT library provides some useful extensions to the standard ``Text_IO`` package: * function File_Exists (Name : String) return Boolean; Determines if a file of the given name exists. * function Get_Line return String; Reads a string from the standard input file. The value returned is exactly the length of the line that was read. * function Get_Line (File : Ada.Text_IO.File_Type) return String; Similar, except that the parameter File specifies the file from which the string is to be read. .. _Text_IO_Facilities_for_Unbounded_Strings: Text_IO Facilities for Unbounded Strings ---------------------------------------- .. index:: Text_IO for unbounded strings .. index:: Unbounded_String, Text_IO operations The package ``Ada.Strings.Unbounded.Text_IO`` in library files :file:`a-suteio.ads/adb` contains some GNAT-specific subprograms useful for Text_IO operations on unbounded strings: * function Get_Line (File : File_Type) return Unbounded_String; Reads a line from the specified file and returns the result as an unbounded string. * procedure Put (File : File_Type; U : Unbounded_String); Writes the value of the given unbounded string to the specified file Similar to the effect of ``Put (To_String (U))`` except that an extra copy is avoided. * procedure Put_Line (File : File_Type; U : Unbounded_String); Writes the value of the given unbounded string to the specified file, followed by a ``New_Line``. Similar to the effect of ``Put_Line (To_String (U))`` except that an extra copy is avoided. In the above procedures, ``File`` is of type ``Ada.Text_IO.File_Type`` and is optional. If the parameter is omitted, then the standard input or output file is referenced as appropriate. The package ``Ada.Strings.Wide_Unbounded.Wide_Text_IO`` in library files :file:`a-swuwti.ads` and :file:`a-swuwti.adb` provides similar extended ``Wide_Text_IO`` functionality for unbounded wide strings. The package ``Ada.Strings.Wide_Wide_Unbounded.Wide_Wide_Text_IO`` in library files :file:`a-szuzti.ads` and :file:`a-szuzti.adb` provides similar extended ``Wide_Wide_Text_IO`` functionality for unbounded wide wide strings. .. _Wide_Text_IO: Wide_Text_IO ============ ``Wide_Text_IO`` is similar in most respects to Text_IO, except that both input and output files may contain special sequences that represent wide character values. The encoding scheme for a given file may be specified using a FORM parameter: :: WCEM=`x` as part of the FORM string (WCEM = wide character encoding method), where ``x`` is one of the following characters ========== ==================== Character Encoding ========== ==================== *h* Hex ESC encoding *u* Upper half encoding *s* Shift-JIS encoding *e* EUC Encoding *8* UTF-8 encoding *b* Brackets encoding ========== ==================== The encoding methods match those that can be used in a source program, but there is no requirement that the encoding method used for the source program be the same as the encoding method used for files, and different files may use different encoding methods. The default encoding method for the standard files, and for opened files for which no WCEM parameter is given in the FORM string matches the wide character encoding specified for the main program (the default being brackets encoding if no coding method was specified with -gnatW). *Hex Coding* In this encoding, a wide character is represented by a five character sequence: :: ESC a b c d .. where ``a``, ``b``, ``c``, ``d`` are the four hexadecimal characters (using upper case letters) of the wide character code. For example, ESC A345 is used to represent the wide character with code 16#A345#. This scheme is compatible with use of the full ``Wide_Character`` set. *Upper Half Coding* The wide character with encoding 16#abcd#, where the upper bit is on (i.e., a is in the range 8-F) is represented as two bytes 16#ab# and 16#cd#. The second byte may never be a format control character, but is not required to be in the upper half. This method can be also used for shift-JIS or EUC where the internal coding matches the external coding. *Shift JIS Coding* A wide character is represented by a two character sequence 16#ab# and 16#cd#, with the restrictions described for upper half encoding as described above. The internal character code is the corresponding JIS character according to the standard algorithm for Shift-JIS conversion. Only characters defined in the JIS code set table can be used with this encoding method. *EUC Coding* A wide character is represented by a two character sequence 16#ab# and 16#cd#, with both characters being in the upper half. The internal character code is the corresponding JIS character according to the EUC encoding algorithm. Only characters defined in the JIS code set table can be used with this encoding method. *UTF-8 Coding* A wide character is represented using UCS Transformation Format 8 (UTF-8) as defined in Annex R of ISO 10646-1/Am.2. Depending on the character value, the representation is a one, two, or three byte sequence: :: 16#0000#-16#007f#: 2#0xxxxxxx# 16#0080#-16#07ff#: 2#110xxxxx# 2#10xxxxxx# 16#0800#-16#ffff#: 2#1110xxxx# 2#10xxxxxx# 2#10xxxxxx# .. where the ``xxx`` bits correspond to the left-padded bits of the 16-bit character value. Note that all lower half ASCII characters are represented as ASCII bytes and all upper half characters and other wide characters are represented as sequences of upper-half (The full UTF-8 scheme allows for encoding 31-bit characters as 6-byte sequences, but in this implementation, all UTF-8 sequences of four or more bytes length will raise a Constraint_Error, as will all invalid UTF-8 sequences.) *Brackets Coding* In this encoding, a wide character is represented by the following eight character sequence: :: [ " a b c d " ] .. where ``a``, ``b``, ``c``, ``d`` are the four hexadecimal characters (using uppercase letters) of the wide character code. For example, ``["A345"]`` is used to represent the wide character with code ``16#A345#``. This scheme is compatible with use of the full Wide_Character set. On input, brackets coding can also be used for upper half characters, e.g., ``["C1"]`` for lower case a. However, on output, brackets notation is only used for wide characters with a code greater than ``16#FF#``. Note that brackets coding is not normally used in the context of Wide_Text_IO or Wide_Wide_Text_IO, since it is really just designed as a portable way of encoding source files. In the context of Wide_Text_IO or Wide_Wide_Text_IO, it can only be used if the file does not contain any instance of the left bracket character other than to encode wide character values using the brackets encoding method. In practice it is expected that some standard wide character encoding method such as UTF-8 will be used for text input output. If brackets notation is used, then any occurrence of a left bracket in the input file which is not the start of a valid wide character sequence will cause Constraint_Error to be raised. It is possible to encode a left bracket as ["5B"] and Wide_Text_IO and Wide_Wide_Text_IO input will interpret this as a left bracket. However, when a left bracket is output, it will be output as a left bracket and not as ["5B"]. We make this decision because for normal use of Wide_Text_IO for outputting messages, it is unpleasant to clobber left brackets. For example, if we write: .. code-block:: ada Put_Line ("Start of output [first run]"); we really do not want to have the left bracket in this message clobbered so that the output reads: :: Start of output ["5B"]first run] .. In practice brackets encoding is reasonably useful for normal Put_Line use since we won't get confused between left brackets and wide character sequences in the output. But for input, or when files are written out and read back in, it really makes better sense to use one of the standard encoding methods such as UTF-8. For the coding schemes other than UTF-8, Hex, or Brackets encoding, not all wide character values can be represented. An attempt to output a character that cannot be represented using the encoding scheme for the file causes Constraint_Error to be raised. An invalid wide character sequence on input also causes Constraint_Error to be raised. .. _Stream_Pointer_Positioning_1: Stream Pointer Positioning -------------------------- ``Ada.Wide_Text_IO`` is similar to ``Ada.Text_IO`` in its handling of stream pointer positioning (:ref:`Text_IO`). There is one additional case: If ``Ada.Wide_Text_IO.Look_Ahead`` reads a character outside the normal lower ASCII set, i.e. a character in the range: .. code-block:: ada Wide_Character'Val (16#0080#) .. Wide_Character'Val (16#FFFF#) then although the logical position of the file pointer is unchanged by the ``Look_Ahead`` call, the stream is physically positioned past the wide character sequence. Again this is to avoid the need for buffering or backup, and all ``Wide_Text_IO`` routines check the internal indication that this situation has occurred so that this is not visible to a normal program using ``Wide_Text_IO``. However, this discrepancy can be observed if the wide text file shares a stream with another file. .. _Reading_and_Writing_Non-Regular_Files_1: Reading and Writing Non-Regular Files ------------------------------------- As in the case of Text_IO, when a non-regular file is read, it is assumed that the file contains no page marks (any form characters are treated as data characters), and ``End_Of_Page`` always returns ``False``. Similarly, the end of file indication is not sticky, so it is possible to read beyond an end of file. .. _Wide_Wide_Text_IO: Wide_Wide_Text_IO ================= ``Wide_Wide_Text_IO`` is similar in most respects to Text_IO, except that both input and output files may contain special sequences that represent wide wide character values. The encoding scheme for a given file may be specified using a FORM parameter: :: WCEM=`x` as part of the FORM string (WCEM = wide character encoding method), where ``x`` is one of the following characters ========== ==================== Character Encoding ========== ==================== *h* Hex ESC encoding *u* Upper half encoding *s* Shift-JIS encoding *e* EUC Encoding *8* UTF-8 encoding *b* Brackets encoding ========== ==================== The encoding methods match those that can be used in a source program, but there is no requirement that the encoding method used for the source program be the same as the encoding method used for files, and different files may use different encoding methods. The default encoding method for the standard files, and for opened files for which no WCEM parameter is given in the FORM string matches the wide character encoding specified for the main program (the default being brackets encoding if no coding method was specified with -gnatW). *UTF-8 Coding* A wide character is represented using UCS Transformation Format 8 (UTF-8) as defined in Annex R of ISO 10646-1/Am.2. Depending on the character value, the representation is a one, two, three, or four byte sequence: :: 16#000000#-16#00007f#: 2#0xxxxxxx# 16#000080#-16#0007ff#: 2#110xxxxx# 2#10xxxxxx# 16#000800#-16#00ffff#: 2#1110xxxx# 2#10xxxxxx# 2#10xxxxxx# 16#010000#-16#10ffff#: 2#11110xxx# 2#10xxxxxx# 2#10xxxxxx# 2#10xxxxxx# .. where the ``xxx`` bits correspond to the left-padded bits of the 21-bit character value. Note that all lower half ASCII characters are represented as ASCII bytes and all upper half characters and other wide characters are represented as sequences of upper-half characters. *Brackets Coding* In this encoding, a wide wide character is represented by the following eight character sequence if is in wide character range :: [ " a b c d " ] .. and by the following ten character sequence if not :: [ " a b c d e f " ] .. where ``a``, ``b``, ``c``, ``d``, ``e``, and ``f`` are the four or six hexadecimal characters (using uppercase letters) of the wide wide character code. For example, ``["01A345"]`` is used to represent the wide wide character with code ``16#01A345#``. This scheme is compatible with use of the full Wide_Wide_Character set. On input, brackets coding can also be used for upper half characters, e.g., ``["C1"]`` for lower case a. However, on output, brackets notation is only used for wide characters with a code greater than ``16#FF#``. If is also possible to use the other Wide_Character encoding methods, such as Shift-JIS, but the other schemes cannot support the full range of wide wide characters. An attempt to output a character that cannot be represented using the encoding scheme for the file causes Constraint_Error to be raised. An invalid wide character sequence on input also causes Constraint_Error to be raised. .. _Stream_Pointer_Positioning_2: Stream Pointer Positioning -------------------------- ``Ada.Wide_Wide_Text_IO`` is similar to ``Ada.Text_IO`` in its handling of stream pointer positioning (:ref:`Text_IO`). There is one additional case: If ``Ada.Wide_Wide_Text_IO.Look_Ahead`` reads a character outside the normal lower ASCII set, i.e. a character in the range: .. code-block:: ada Wide_Wide_Character'Val (16#0080#) .. Wide_Wide_Character'Val (16#10FFFF#) then although the logical position of the file pointer is unchanged by the ``Look_Ahead`` call, the stream is physically positioned past the wide character sequence. Again this is to avoid the need for buffering or backup, and all ``Wide_Wide_Text_IO`` routines check the internal indication that this situation has occurred so that this is not visible to a normal program using ``Wide_Wide_Text_IO``. However, this discrepancy can be observed if the wide text file shares a stream with another file. .. _Reading_and_Writing_Non-Regular_Files_2: Reading and Writing Non-Regular Files ------------------------------------- As in the case of Text_IO, when a non-regular file is read, it is assumed that the file contains no page marks (any form characters are treated as data characters), and ``End_Of_Page`` always returns ``False``. Similarly, the end of file indication is not sticky, so it is possible to read beyond an end of file. .. _Stream_IO: Stream_IO ========= A stream file is a sequence of bytes, where individual elements are written to the file as described in the Ada Reference Manual. The type ``Stream_Element`` is simply a byte. There are two ways to read or write a stream file. * The operations ``Read`` and ``Write`` directly read or write a sequence of stream elements with no control information. * The stream attributes applied to a stream file transfer data in the manner described for stream attributes. .. _Text_Translation: Text Translation ================ ``Text_Translation=xxx`` may be used as the Form parameter passed to Text_IO.Create and Text_IO.Open. ``Text_Translation=xxx`` has no effect on Unix systems. Possible values are: * ``Yes`` or ``Text`` is the default, which means to translate LF to/from CR/LF on Windows systems. ``No`` disables this translation; i.e. it uses binary mode. For output files, ``Text_Translation=No`` may be used to create Unix-style files on Windows. * ``wtext`` translation enabled in Unicode mode. (corresponds to _O_WTEXT). * ``u8text`` translation enabled in Unicode UTF-8 mode. (corresponds to O_U8TEXT). * ``u16text`` translation enabled in Unicode UTF-16 mode. (corresponds to_O_U16TEXT). .. _Shared_Files: Shared Files ============ Section A.14 of the Ada Reference Manual allows implementations to provide a wide variety of behavior if an attempt is made to access the same external file with two or more internal files. To provide a full range of functionality, while at the same time minimizing the problems of portability caused by this implementation dependence, GNAT handles file sharing as follows: * In the absence of a ``shared=xxx`` form parameter, an attempt to open two or more files with the same full name is considered an error and is not supported. The exception ``Use_Error`` will be raised. Note that a file that is not explicitly closed by the program remains open until the program terminates. * If the form parameter ``shared=no`` appears in the form string, the file can be opened or created with its own separate stream identifier, regardless of whether other files sharing the same external file are opened. The exact effect depends on how the C stream routines handle multiple accesses to the same external files using separate streams. * If the form parameter ``shared=yes`` appears in the form string for each of two or more files opened using the same full name, the same stream is shared between these files, and the semantics are as described in Ada Reference Manual, Section A.14. When a program that opens multiple files with the same name is ported from another Ada compiler to GNAT, the effect will be that ``Use_Error`` is raised. The documentation of the original compiler and the documentation of the program should then be examined to determine if file sharing was expected, and ``shared=xxx`` parameters added to ``Open`` and ``Create`` calls as required. When a program is ported from GNAT to some other Ada compiler, no special attention is required unless the ``shared=xxx`` form parameter is used in the program. In this case, you must examine the documentation of the new compiler to see if it supports the required file sharing semantics, and form strings modified appropriately. Of course it may be the case that the program cannot be ported if the target compiler does not support the required functionality. The best approach in writing portable code is to avoid file sharing (and hence the use of the ``shared=xxx`` parameter in the form string) completely. One common use of file sharing in Ada 83 is the use of instantiations of Sequential_IO on the same file with different types, to achieve heterogeneous input-output. Although this approach will work in GNAT if ``shared=yes`` is specified, it is preferable in Ada to use Stream_IO for this purpose (using the stream attributes). .. _Filenames_encoding: Filenames encoding ================== An encoding form parameter can be used to specify the filename encoding ``encoding=xxx``. * If the form parameter ``encoding=utf8`` appears in the form string, the filename must be encoded in UTF-8. * If the form parameter ``encoding=8bits`` appears in the form string, the filename must be a standard 8bits string. In the absence of a ``encoding=xxx`` form parameter, the encoding is controlled by the ``GNAT_CODE_PAGE`` environment variable. And if not set ``utf8`` is assumed. *CP_ACP* The current system Windows ANSI code page. *CP_UTF8* UTF-8 encoding This encoding form parameter is only supported on the Windows platform. On the other Operating Systems the run-time is supporting UTF-8 natively. .. _File_content_encoding: File content encoding ===================== For text files it is possible to specify the encoding to use. This is controlled by the by the ``GNAT_CCS_ENCODING`` environment variable. And if not set ``TEXT`` is assumed. The possible values are those supported on Windows: *TEXT* Translated text mode *WTEXT* Translated unicode encoding *U16TEXT* Unicode 16-bit encoding *U8TEXT* Unicode 8-bit encoding This encoding is only supported on the Windows platform. .. _Open_Modes: Open Modes ========== ``Open`` and ``Create`` calls result in a call to ``fopen`` using the mode shown in the following table: +----------------------------+---------------+------------------+ | ``Open`` and ``Create`` Call Modes | +----------------------------+---------------+------------------+ | | **OPEN** | **CREATE** | +============================+===============+==================+ | Append_File | "r+" | "w+" | +----------------------------+---------------+------------------+ | In_File | "r" | "w+" | +----------------------------+---------------+------------------+ | Out_File (Direct_IO) | "r+" | "w" | +----------------------------+---------------+------------------+ | Out_File (all other cases) | "w" | "w" | +----------------------------+---------------+------------------+ | Inout_File | "r+" | "w+" | +----------------------------+---------------+------------------+ If text file translation is required, then either ``b`` or ``t`` is added to the mode, depending on the setting of Text. Text file translation refers to the mapping of CR/LF sequences in an external file to LF characters internally. This mapping only occurs in DOS and DOS-like systems, and is not relevant to other systems. A special case occurs with Stream_IO. As shown in the above table, the file is initially opened in ``r`` or ``w`` mode for the ``In_File`` and ``Out_File`` cases. If a ``Set_Mode`` operation subsequently requires switching from reading to writing or vice-versa, then the file is reopened in ``r+`` mode to permit the required operation. .. _Operations_on_C_Streams: Operations on C Streams ======================= The package ``Interfaces.C_Streams`` provides an Ada program with direct access to the C library functions for operations on C streams: .. code-block:: ada package Interfaces.C_Streams is -- Note: the reason we do not use the types that are in -- Interfaces.C is that we want to avoid dragging in the -- code in this unit if possible. subtype chars is System.Address; -- Pointer to null-terminated array of characters subtype FILEs is System.Address; -- Corresponds to the C type FILE* subtype voids is System.Address; -- Corresponds to the C type void* subtype int is Integer; subtype long is Long_Integer; -- Note: the above types are subtypes deliberately, and it -- is part of this spec that the above correspondences are -- guaranteed. This means that it is legitimate to, for -- example, use Integer instead of int. We provide these -- synonyms for clarity, but in some cases it may be -- convenient to use the underlying types (for example to -- avoid an unnecessary dependency of a spec on the spec -- of this unit). type size_t is mod 2 ** Standard'Address_Size; NULL_Stream : constant FILEs; -- Value returned (NULL in C) to indicate an -- fdopen/fopen/tmpfile error ---------------------------------- -- Constants Defined in stdio.h -- ---------------------------------- EOF : constant int; -- Used by a number of routines to indicate error or -- end of file IOFBF : constant int; IOLBF : constant int; IONBF : constant int; -- Used to indicate buffering mode for setvbuf call SEEK_CUR : constant int; SEEK_END : constant int; SEEK_SET : constant int; -- Used to indicate origin for fseek call function stdin return FILEs; function stdout return FILEs; function stderr return FILEs; -- Streams associated with standard files -------------------------- -- Standard C functions -- -------------------------- -- The functions selected below are ones that are -- available in UNIX (but not necessarily in ANSI C). -- These are very thin interfaces -- which copy exactly the C headers. For more -- documentation on these functions, see the Microsoft C -- "Run-Time Library Reference" (Microsoft Press, 1990, -- ISBN 1-55615-225-6), which includes useful information -- on system compatibility. procedure clearerr (stream : FILEs); function fclose (stream : FILEs) return int; function fdopen (handle : int; mode : chars) return FILEs; function feof (stream : FILEs) return int; function ferror (stream : FILEs) return int; function fflush (stream : FILEs) return int; function fgetc (stream : FILEs) return int; function fgets (strng : chars; n : int; stream : FILEs) return chars; function fileno (stream : FILEs) return int; function fopen (filename : chars; Mode : chars) return FILEs; -- Note: to maintain target independence, use -- text_translation_required, a boolean variable defined in -- a-sysdep.c to deal with the target dependent text -- translation requirement. If this variable is set, -- then b/t should be appended to the standard mode -- argument to set the text translation mode off or on -- as required. function fputc (C : int; stream : FILEs) return int; function fputs (Strng : chars; Stream : FILEs) return int; function fread (buffer : voids; size : size_t; count : size_t; stream : FILEs) return size_t; function freopen (filename : chars; mode : chars; stream : FILEs) return FILEs; function fseek (stream : FILEs; offset : long; origin : int) return int; function ftell (stream : FILEs) return long; function fwrite (buffer : voids; size : size_t; count : size_t; stream : FILEs) return size_t; function isatty (handle : int) return int; procedure mktemp (template : chars); -- The return value (which is just a pointer to template) -- is discarded procedure rewind (stream : FILEs); function rmtmp return int; function setvbuf (stream : FILEs; buffer : chars; mode : int; size : size_t) return int; function tmpfile return FILEs; function ungetc (c : int; stream : FILEs) return int; function unlink (filename : chars) return int; --------------------- -- Extra functions -- --------------------- -- These functions supply slightly thicker bindings than -- those above. They are derived from functions in the -- C Run-Time Library, but may do a bit more work than -- just directly calling one of the Library functions. function is_regular_file (handle : int) return int; -- Tests if given handle is for a regular file (result 1) -- or for a non-regular file (pipe or device, result 0). --------------------------------- -- Control of Text/Binary Mode -- --------------------------------- -- If text_translation_required is true, then the following -- functions may be used to dynamically switch a file from -- binary to text mode or vice versa. These functions have -- no effect if text_translation_required is false (i.e., in -- normal UNIX mode). Use fileno to get a stream handle. procedure set_binary_mode (handle : int); procedure set_text_mode (handle : int); ---------------------------- -- Full Path Name support -- ---------------------------- procedure full_name (nam : chars; buffer : chars); -- Given a NUL terminated string representing a file -- name, returns in buffer a NUL terminated string -- representing the full path name for the file name. -- On systems where it is relevant the drive is also -- part of the full path name. It is the responsibility -- of the caller to pass an actual parameter for buffer -- that is big enough for any full path name. Use -- max_path_len given below as the size of buffer. max_path_len : integer; -- Maximum length of an allowable full path name on the -- system, including a terminating NUL character. end Interfaces.C_Streams; .. _Interfacing_to_C_Streams: Interfacing to C Streams ======================== The packages in this section permit interfacing Ada files to C Stream operations. .. code-block:: ada with Interfaces.C_Streams; package Ada.Sequential_IO.C_Streams is function C_Stream (F : File_Type) return Interfaces.C_Streams.FILEs; procedure Open (File : in out File_Type; Mode : in File_Mode; C_Stream : in Interfaces.C_Streams.FILEs; Form : in String := ""); end Ada.Sequential_IO.C_Streams; with Interfaces.C_Streams; package Ada.Direct_IO.C_Streams is function C_Stream (F : File_Type) return Interfaces.C_Streams.FILEs; procedure Open (File : in out File_Type; Mode : in File_Mode; C_Stream : in Interfaces.C_Streams.FILEs; Form : in String := ""); end Ada.Direct_IO.C_Streams; with Interfaces.C_Streams; package Ada.Text_IO.C_Streams is function C_Stream (F : File_Type) return Interfaces.C_Streams.FILEs; procedure Open (File : in out File_Type; Mode : in File_Mode; C_Stream : in Interfaces.C_Streams.FILEs; Form : in String := ""); end Ada.Text_IO.C_Streams; with Interfaces.C_Streams; package Ada.Wide_Text_IO.C_Streams is function C_Stream (F : File_Type) return Interfaces.C_Streams.FILEs; procedure Open (File : in out File_Type; Mode : in File_Mode; C_Stream : in Interfaces.C_Streams.FILEs; Form : in String := ""); end Ada.Wide_Text_IO.C_Streams; with Interfaces.C_Streams; package Ada.Wide_Wide_Text_IO.C_Streams is function C_Stream (F : File_Type) return Interfaces.C_Streams.FILEs; procedure Open (File : in out File_Type; Mode : in File_Mode; C_Stream : in Interfaces.C_Streams.FILEs; Form : in String := ""); end Ada.Wide_Wide_Text_IO.C_Streams; with Interfaces.C_Streams; package Ada.Stream_IO.C_Streams is function C_Stream (F : File_Type) return Interfaces.C_Streams.FILEs; procedure Open (File : in out File_Type; Mode : in File_Mode; C_Stream : in Interfaces.C_Streams.FILEs; Form : in String := ""); end Ada.Stream_IO.C_Streams; In each of these six packages, the ``C_Stream`` function obtains the ``FILE`` pointer from a currently opened Ada file. It is then possible to use the ``Interfaces.C_Streams`` package to operate on this stream, or the stream can be passed to a C program which can operate on it directly. Of course the program is responsible for ensuring that only appropriate sequences of operations are executed. One particular use of relevance to an Ada program is that the ``setvbuf`` function can be used to control the buffering of the stream used by an Ada file. In the absence of such a call the standard default buffering is used. The ``Open`` procedures in these packages open a file giving an existing C Stream instead of a file name. Typically this stream is imported from a C program, allowing an Ada file to operate on an existing C file. --- ### Gcc Ada Doc Gnat Ugn About This Guide (`gcc/ada/doc/gnat_ugn/about_this_guide.rst`) .. _About_This_Guide: About This Guide ~~~~~~~~~~~~~~~~ .. only:: PRO For ease of exposition, 'GNAT Pro' will be referred to simply as 'GNAT' in the remainder of this document. This guide describes the use of GNAT, a compiler and software development toolset for the full Ada programming language. It documents the features of the compiler and tools, and explains how to use them to build Ada applications. GNAT implements Ada 95, Ada 2005, Ada 2012, and Ada 2022. You may also invoke it in Ada 83 compatibility mode. By default, GNAT assumes Ada 2012, but you can use a compiler switch (:ref:`Compiling_Different_Versions_of_Ada`) to explicitly specify the language version. Throughout this manual, references to 'Ada' without a year suffix apply to all versions of the Ada language starting with Ada 95. GNAT supports both the GCC and LLVM back end compilation families. Most GNAT versions use the GCC back end, but some are now available using the LLVM back end. In some places in this manual, we distinguish between the two back ends, but in most cases, everything in this manual applies to both back ends. We refer to GNAT with the LLVM back end as 'GNAT LLVM'. See :ref:`GNATLLVM` for limitations of GNAT LLVM. What This Guide Contains ======================== This guide contains the following chapters: * :ref:`Getting_Started_with_GNAT` describes how to get started compiling and running Ada programs with the GNAT Ada programming environment. * :ref:`The_GNAT_Compilation_Model` describes the compilation model used by GNAT. * :ref:`Building_Executable_Programs_With_GNAT` describes how to use the main GNAT tools to build executable programs, and it also gives examples of using the GNU make utility with GNAT. * :ref:`GNAT_Utility_Programs` explains the various utility programs that are included in the GNAT environment. * :ref:`GNAT_and_Program_Execution` covers a number of topics related to running, debugging, and tuning the performance of programs developed with GNAT. Appendices cover several additional topics: * :ref:`Platform_Specific_Information` describes the different run-time library implementations and also presents information on how to use GNAT on several specific platforms. * :ref:`Example_of_Binder_Output_File` shows the source code for the binder output file for a sample program. * :ref:`Elaboration_Order_Handling_in_GNAT` describes how GNAT helps you deal with elaboration order issues. * :ref:`Inline_Assembler` shows how to use the inline assembly facility in an Ada program. What You Should Know before Reading This Guide ============================================== .. index:: Ada 95 Language Reference Manual .. index:: Ada 2005 Language Reference Manual This guide assumes a basic familiarity with the Ada 95 language, as described in the International Standard ANSI/ISO/IEC-8652:1995, January 1995. Reference manuals for Ada 95, Ada 2005, Ada 2012 and Ada 2022 are included in the GNAT documentation package. Related Information =================== For further information about Ada and related tools, please refer to the following documents: * :title:`Ada 95 Reference Manual`, :title:`Ada 2005 Reference Manual`, :title:`Ada 2012 Reference Manual`, and :title:`Ada 2022 Reference Manual`, which contain reference material for the several revisions of the Ada language standard. * :title:`GNAT Reference Manual`, which contains all reference material for the GNAT implementation of Ada. * :title:`Using GNAT Studio`, which describes the GNAT Studio Integrated Development Environment. * :title:`GNAT Studio Tutorial`, which introduces the main GNAT Studio features through examples. * :title:`Debugging with GDB`, for all details on the use of the GNU source-level debugger. Conventions =========== .. index:: Conventions, typographical .. index:: Typographical conventions Following are examples of the typographical and graphic conventions used in this guide: * ``Functions``, ``utility program names``, ``standard names``, and ``classes``. * ``Option flags`` * :file:`File names` * ``Variables`` * *Emphasis* * [optional information or parameters] * Examples are described by text .. code-block:: text and then shown this way. * Commands that you enter are shown as preceded by a prompt string comprising the ``$`` character followed by a space. * Full file names are shown with the '/' character as the directory separator; e.g., :file:`parent-dir/subdir/myfile.adb`. If you are using GNAT on a Windows platform, please note that you should use the '\\' character instead. --- ### Gcc Ada Doc Gnat Ugn Building Executable Programs With Gnat (`gcc/ada/doc/gnat_ugn/building_executable_programs_with_gnat.rst`) .. |with| replace:: *with* .. |withs| replace:: *with*\ s .. |withed| replace:: *with*\ ed .. |withing| replace:: *with*\ ing .. -- Example: A |withing| unit has a |with| clause, it |withs| a |withed| unit .. role:: switch(samp) .. _Building_Executable_Programs_With_GNAT: ************************************** Building Executable Programs with GNAT ************************************** This chapter describes first the gnatmake tool (:ref:`The_GNAT_Make_Program_gnatmake`), which automatically determines the set of sources needed by an Ada compilation unit and executes the necessary (re)compilations, binding and linking. It also explains how to use each tool individually: the compiler (gcc, see :ref:`Compiling_with_gcc`), binder (gnatbind, see :ref:`Binding_with_gnatbind`), and linker (gnatlink, see :ref:`Linking_with_gnatlink`) to build executable programs. Finally, this chapter provides examples of how to make use of the general GNU make mechanism in a GNAT context (see :ref:`Using_the_GNU_make_Utility`). .. only:: PRO or GPL For building large systems with components possibly written in different languages (such as Ada, C, C++ and Fortran) and organized into subsystems and libraries, the GPRbuild tool can be used. This tool, and the Project Manager facility that it is based upon, is described in *GPRbuild and GPR Companion Tools User's Guide*. .. _The_GNAT_Make_Program_gnatmake: Building with ``gnatmake`` ========================== .. index:: gnatmake A typical development cycle when working on an Ada program consists of the following steps: #. Edit some sources to fix bugs; #. Add enhancements; #. Compile all sources affected; #. Rebind and relink; and #. Test. .. index:: Dependency rules (compilation) The third step in particular can be tricky, because not only do the modified files have to be compiled, but any files depending on these files must also be recompiled. The dependency rules in Ada can be quite complex, especially in the presence of overloading, ``use`` clauses, generics and inlined subprograms. ``gnatmake`` automatically takes care of the third and fourth steps of this process. It determines which sources need to be compiled, compiles them, and binds and links the resulting object files. Unlike some other Ada make programs, the dependencies are always accurately recomputed from the new sources. The source based approach of the GNAT compilation model makes this possible. This means that if changes to the source program cause corresponding changes in dependencies, they will always be tracked exactly correctly by ``gnatmake``. Note that for advanced forms of project structure, we recommend creating a project file as explained in the *GNAT_Project_Manager* chapter in the *GPRbuild User's Guide*, and using the ``gprbuild`` tool which supports building with project files and works similarly to ``gnatmake``. .. _Running_gnatmake: Running ``gnatmake`` -------------------- The usual form of the ``gnatmake`` command is .. code-block:: sh $ gnatmake [] [] [] The only required argument is one ``file_name``, which specifies a compilation unit that is a main program. Several ``file_names`` can be specified: this will result in several executables being built. If ``switches`` are present, they can be placed before the first ``file_name``, between ``file_names`` or after the last ``file_name``. If ``mode_switches`` are present, they must always be placed after the last ``file_name`` and all ``switches``. If you are using standard file extensions (:file:`.adb` and :file:`.ads`), then the extension may be omitted from the ``file_name`` arguments. However, if you are using non-standard extensions, then it is required that the extension be given. A relative or absolute directory path can be specified in a ``file_name``, in which case, the input source file will be searched for in the specified directory only. Otherwise, the input source file will first be searched in the directory where ``gnatmake`` was invoked and if it is not found, it will be search on the source path of the compiler as described in :ref:`Search_Paths_and_the_Run-Time_Library_RTL`. All ``gnatmake`` output (except when you specify :switch:`-M`) is sent to :file:`stderr`. The output produced by the :switch:`-M` switch is sent to :file:`stdout`. .. _Switches_for_gnatmake: Switches for ``gnatmake`` ------------------------- You may specify any of the following switches to ``gnatmake``: .. index:: --version (gnatmake) :switch:`--version` Display Copyright and version, then exit disregarding all other options. .. index:: --help (gnatmake) :switch:`--help` If ``--version`` was not used, display usage, then exit disregarding all other options. .. index:: -P (gnatmake) :switch:`-P{project}` Build GNAT project file ``project`` using GPRbuild. When this switch is present, all other command-line switches are treated as GPRbuild switches and not ``gnatmake`` switches. .. -- Comment: :ref:`gnatmake_and_Project_Files`. .. index:: --GCC=compiler_name (gnatmake) :switch:`--GCC={compiler_name}` Program used for compiling. The default is ``gcc``. You need to use quotes around ``compiler_name`` if ``compiler_name`` contains spaces or other separator characters. As an example ``--GCC="foo -x -y"`` will instruct ``gnatmake`` to use ``foo -x -y`` as your compiler. A limitation of this syntax is that the name and path name of the executable itself must not include any embedded spaces. Note that switch :switch:`-c` is always inserted after your command name. Thus in the above example the compiler command that will be used by ``gnatmake`` will be ``foo -c -x -y``. If several ``--GCC=compiler_name`` are used, only the last ``compiler_name`` is taken into account. However, all the additional switches are also taken into account. Thus, ``--GCC="foo -x -y" --GCC="bar -z -t"`` is equivalent to ``--GCC="bar -x -y -z -t"``. .. index:: --GNATBIND=binder_name (gnatmake) :switch:`--GNATBIND={binder_name}` Program used for binding. The default is ``gnatbind``. You need to use quotes around ``binder_name`` if ``binder_name`` contains spaces or other separator characters. As an example ``--GNATBIND="bar -x -y"`` will instruct ``gnatmake`` to use ``bar -x -y`` as your binder. Binder switches that are normally appended by ``gnatmake`` to ``gnatbind`` are now appended to the end of ``bar -x -y``. A limitation of this syntax is that the name and path name of the executable itself must not include any embedded spaces. .. index:: --GNATLINK=linker_name (gnatmake) :switch:`--GNATLINK={linker_name}` Program used for linking. The default is ``gnatlink``. You need to use quotes around ``linker_name`` if ``linker_name`` contains spaces or other separator characters. As an example ``--GNATLINK="lan -x -y"`` will instruct ``gnatmake`` to use ``lan -x -y`` as your linker. Linker switches that are normally appended by ``gnatmake`` to ``gnatlink`` are now appended to the end of ``lan -x -y``. A limitation of this syntax is that the name and path name of the executable itself must not include any embedded spaces. :switch:`--create-map-file` When linking an executable, create a map file. The name of the map file has the same name as the executable with extension ".map". :switch:`--create-map-file={mapfile}` When linking an executable, create a map file with the specified name. .. index:: --create-missing-dirs (gnatmake) :switch:`--create-missing-dirs` When using project files (:switch:`-P{project}`), automatically create missing object directories, library directories and exec directories. :switch:`--single-compile-per-obj-dir` Disallow simultaneous compilations in the same object directory when project files are used. :switch:`--subdirs={subdir}` Actual object directory of each project file is the subdirectory subdir of the object directory specified or defaulted in the project file. :switch:`--unchecked-shared-lib-imports` By default, shared library projects are not allowed to import static library projects. When this switch is used on the command line, this restriction is relaxed. :switch:`--source-info={source info file}` Specify a source info file. This switch is active only when project files are used. If the source info file is specified as a relative path, then it is relative to the object directory of the main project. If the source info file does not exist, then after the Project Manager has successfully parsed and processed the project files and found the sources, it creates the source info file. If the source info file already exists and can be read successfully, then the Project Manager will get all the needed information about the sources from the source info file and will not look for them. This reduces the time to process the project files, especially when looking for sources that take a long time. If the source info file exists but cannot be parsed successfully, the Project Manager will attempt to recreate it. If the Project Manager fails to create the source info file, a message is issued, but gnatmake does not fail. ``gnatmake`` "trusts" the source info file. This means that if the source files have changed (addition, deletion, moving to a different source directory), then the source info file need to be deleted and recreated. .. index:: -a (gnatmake) :switch:`-a` Consider all files in the make process, even the GNAT internal system files (for example, the predefined Ada library files), as well as any locked files. Locked files are files whose ALI file is write-protected. By default, ``gnatmake`` does not check these files, because the assumption is that the GNAT internal files are properly up to date, and also that any write protected ALI files have been properly installed. Note that if there is an installation problem, such that one of these files is not up to date, it will be properly caught by the binder. You may have to specify this switch if you are working on GNAT itself. The switch ``-a`` is also useful in conjunction with ``-f`` if you need to recompile an entire application, including run-time files, using special configuration pragmas, such as a ``Normalize_Scalars`` pragma. By default ``gnatmake -a`` compiles all GNAT internal files with ``gcc -c -gnatpg`` rather than ``gcc -c``. .. index:: -b (gnatmake) :switch:`-b` Bind only. Can be combined with :switch:`-c` to do compilation and binding, but no link. Can be combined with :switch:`-l` to do binding and linking. When not combined with :switch:`-c` all the units in the closure of the main program must have been previously compiled and must be up to date. The root unit specified by ``file_name`` may be given without extension, with the source extension or, if no GNAT Project File is specified, with the ALI file extension. .. index:: -c (gnatmake) :switch:`-c` Compile only. Do not perform binding, except when :switch:`-b` is also specified. Do not perform linking, except if both :switch:`-b` and :switch:`-l` are also specified. If the root unit specified by ``file_name`` is not a main unit, this is the default. Otherwise ``gnatmake`` will attempt binding and linking unless all objects are up to date and the executable is more recent than the objects. .. index:: -C (gnatmake) :switch:`-C` Use a temporary mapping file. A mapping file is a way to communicate to the compiler two mappings: from unit names to file names (without any directory information) and from file names to path names (with full directory information). A mapping file can make the compiler's file searches faster, especially if there are many source directories, or the sources are read over a slow network connection. If :switch:`-P` is used, a mapping file is always used, so :switch:`-C` is unnecessary; in this case the mapping file is initially populated based on the project file. If :switch:`-C` is used without :switch:`-P`, the mapping file is initially empty. Each invocation of the compiler will add any newly accessed sources to the mapping file. .. index:: -C= (gnatmake) :switch:`-C={file}` Use a specific mapping file. The file, specified as a path name (absolute or relative) by this switch, should already exist, otherwise the switch is ineffective. The specified mapping file will be communicated to the compiler. This switch is not compatible with a project file (-P`file`) or with multiple compiling processes (-jnnn, when nnn is greater than 1). .. index:: -d (gnatmake) :switch:`-d` Display progress for each source, up to date or not, as a single line: :: completed x out of y (zz%) If the file needs to be compiled this is displayed after the invocation of the compiler. These lines are displayed even in quiet output mode. .. index:: -D (gnatmake) :switch:`-D {dir}` Put all object files and ALI file in directory ``dir``. If the :switch:`-D` switch is not used, all object files and ALI files go in the current working directory. This switch cannot be used when using a project file. .. index:: -eI (gnatmake) :switch:`-eI{nnn}` Indicates that the main source is a multi-unit source and the rank of the unit in the source file is nnn. nnn needs to be a positive number and a valid index in the source. This switch cannot be used when ``gnatmake`` is invoked for several mains. .. index:: -eL (gnatmake) .. index:: symbolic links :switch:`-eL` Follow all symbolic links when processing project files. This should be used if your project uses symbolic links for files or directories, but is not needed in other cases. .. index:: naming scheme This also assumes that no directory matches the naming scheme for files (for instance that you do not have a directory called "sources.ads" when using the default GNAT naming scheme). When you do not have to use this switch (i.e., by default), gnatmake is able to save a lot of system calls (several per source file and object file), which can result in a significant speed up to load and manipulate a project file, especially when using source files from a remote system. .. index:: -eS (gnatmake) :switch:`-eS` Output the commands for the compiler, the binder and the linker on standard output, instead of standard error. .. index:: -f (gnatmake) :switch:`-f` Force recompilations. Recompile all sources, even though some object files may be up to date, but don't recompile predefined or GNAT internal files or locked files (files with a write-protected ALI file), unless the :switch:`-a` switch is also specified. .. index:: -F (gnatmake) :switch:`-F` When using project files, if some errors or warnings are detected during parsing and verbose mode is not in effect (no use of switch -v), then error lines start with the full path name of the project file, rather than its simple file name. .. index:: -g (gnatmake) :switch:`-g` Enable debugging. This switch is simply passed to the compiler and to the linker. .. index:: -i (gnatmake) :switch:`-i` In normal mode, ``gnatmake`` compiles all object files and ALI files into the current directory. If the :switch:`-i` switch is used, then instead object files and ALI files that already exist are overwritten in place. This means that once a large project is organized into separate directories in the desired manner, then ``gnatmake`` will automatically maintain and update this organization. If no ALI files are found on the Ada object path (see :ref:`Search_Paths_and_the_Run-Time_Library_RTL`), the new object and ALI files are created in the directory containing the source being compiled. If another organization is desired, where objects and sources are kept in different directories, a useful technique is to create dummy ALI files in the desired directories. When detecting such a dummy file, ``gnatmake`` will be forced to recompile the corresponding source file, and it will be put the resulting object and ALI files in the directory where it found the dummy file. .. index:: -j (gnatmake) .. index:: Parallel make :switch:`-j{n}` Use ``n`` processes to carry out the (re)compilations. On a multiprocessor machine compilations will occur in parallel. If ``n`` is 0, then the maximum number of parallel compilations is the number of core processors on the platform. In the event of compilation errors, messages from various compilations might get interspersed (but ``gnatmake`` will give you the full ordered list of failing compiles at the end). If this is problematic, rerun the make process with n set to 1 to get a clean list of messages. .. index:: -k (gnatmake) :switch:`-k` Keep going. Continue as much as possible after a compilation error. To ease the programmer's task in case of compilation errors, the list of sources for which the compile fails is given when ``gnatmake`` terminates. If ``gnatmake`` is invoked with several :file:`file_names` and with this switch, if there are compilation errors when building an executable, ``gnatmake`` will not attempt to build the following executables. .. index:: -l (gnatmake) :switch:`-l` Link only. Can be combined with :switch:`-b` to binding and linking. Linking will not be performed if combined with :switch:`-c` but not with :switch:`-b`. When not combined with :switch:`-b` all the units in the closure of the main program must have been previously compiled and must be up to date, and the main program needs to have been bound. The root unit specified by ``file_name`` may be given without extension, with the source extension or, if no GNAT Project File is specified, with the ALI file extension. .. index:: -m (gnatmake) :switch:`-m` Specify that the minimum necessary amount of recompilations be performed. In this mode ``gnatmake`` ignores time stamp differences when the only modifications to a source file consist in adding/removing comments, empty lines, spaces or tabs. This means that if you have changed the comments in a source file or have simply reformatted it, using this switch will tell ``gnatmake`` not to recompile files that depend on it (provided other sources on which these files depend have undergone no semantic modifications). Note that the debugging information may be out of date with respect to the sources if the :switch:`-m` switch causes a compilation to be switched, so the use of this switch represents a trade-off between compilation time and accurate debugging information. .. index:: Dependencies, producing list .. index:: -M (gnatmake) :switch:`-M` Check if all objects are up to date. If they are, output the object dependences to :file:`stdout` in a form that can be directly exploited in a :file:`Makefile`. By default, each source file is prefixed with its (relative or absolute) directory name. This name is whatever you specified in the various :switch:`-aI` and :switch:`-I` switches. If you use ``gnatmake -M`` :switch:`-q` (see below), only the source file names, without relative paths, are output. If you just specify the :switch:`-M` switch, dependencies of the GNAT internal system files are omitted. This is typically what you want. If you also specify the :switch:`-a` switch, dependencies of the GNAT internal files are also listed. Note that dependencies of the objects in external Ada libraries (see switch :switch:`-aL{dir}` in the following list) are never reported. .. index:: -n (gnatmake) :switch:`-n` Don't compile, bind, or link. Checks if all objects are up to date. If they are not, the full name of the first file that needs to be recompiled is printed. Repeated use of this option, followed by compiling the indicated source file, will eventually result in recompiling all required units. .. index:: -o (gnatmake) :switch:`-o {exec_name}` Output executable name. The name of the final executable program will be ``exec_name``. If the :switch:`-o` switch is omitted the default name for the executable will be the name of the input file in appropriate form for an executable file on the host system. This switch cannot be used when invoking ``gnatmake`` with several :file:`file_names`. .. index:: -p (gnatmake) :switch:`-p` Same as :switch:`--create-missing-dirs` .. index:: -q (gnatmake) :switch:`-q` Quiet. When this flag is not set, the commands carried out by ``gnatmake`` are displayed. .. index:: -s (gnatmake) :switch:`-s` Recompile if compiler switches have changed since last compilation. All compiler switches but -I and -o are taken into account in the following way: orders between different 'first letter' switches are ignored, but orders between same switches are taken into account. For example, :switch:`-O -O2` is different than :switch:`-O2 -O`, but :switch:`-g -O` is equivalent to :switch:`-O -g`. This switch is recommended when Integrated Preprocessing is used. .. index:: -u (gnatmake) :switch:`-u` Unique. Recompile at most the main files. It implies -c. Combined with -f, it is equivalent to calling the compiler directly. Note that using -u with a project file and no main has a special meaning. .. --Comment (See :ref:`Project_Files_and_Main_Subprograms`.) .. index:: -U (gnatmake) :switch:`-U` When used without a project file or with one or several mains on the command line, is equivalent to -u. When used with a project file and no main on the command line, all sources of all project files are checked and compiled if not up to date, and libraries are rebuilt, if necessary. .. index:: -v (gnatmake) :switch:`-v` Verbose. Display the reason for all recompilations ``gnatmake`` decides are necessary, with the highest verbosity level. .. index:: -vl (gnatmake) :switch:`-vl` Verbosity level Low. Display fewer lines than in verbosity Medium. .. index:: -vm (gnatmake) :switch:`-vm` Verbosity level Medium. Potentially display fewer lines than in verbosity High. .. index:: -vm (gnatmake) :switch:`-vh` Verbosity level High. Equivalent to -v. :switch:`-vP{x}` Indicate the verbosity of the parsing of GNAT project files. See :ref:`Switches_Related_to_Project_Files`. .. index:: -x (gnatmake) :switch:`-x` Indicate that sources that are not part of any Project File may be compiled. Normally, when using Project Files, only sources that are part of a Project File may be compile. When this switch is used, a source outside of all Project Files may be compiled. The ALI file and the object file will be put in the object directory of the main Project. The compilation switches used will only be those specified on the command line. Even when :switch:`-x` is used, mains specified on the command line need to be sources of a project file. :switch:`-X{name}={value}` Indicate that external variable ``name`` has the value ``value``. The Project Manager will use this value for occurrences of ``external(name)`` when parsing the project file. :ref:`Switches_Related_to_Project_Files`. .. index:: -z (gnatmake) :switch:`-z` No main subprogram. Bind and link the program even if the unit name given on the command line is a package name. The resulting executable will execute the elaboration routines of the package and its closure, then the finalization routines. .. rubric:: GCC switches Any uppercase or multi-character switch that is not a ``gnatmake`` switch is passed to ``gcc`` (e.g., :switch:`-O`, :switch:`-gnato,` etc.) .. rubric:: Source and library search path switches .. index:: -aI (gnatmake) :switch:`-aI{dir}` When looking for source files also look in directory ``dir``. The order in which source files search is undertaken is described in :ref:`Search_Paths_and_the_Run-Time_Library_RTL`. .. index:: -aL (gnatmake) :switch:`-aL{dir}` Consider ``dir`` as being an externally provided Ada library. Instructs ``gnatmake`` to skip compilation units whose :file:`.ALI` files have been located in directory ``dir``. This allows you to have missing bodies for the units in ``dir`` and to ignore out of date bodies for the same units. You still need to specify the location of the specs for these units by using the switches :switch:`-aI{dir}` or :switch:`-I{dir}`. Note: this switch is provided for compatibility with previous versions of ``gnatmake``. The easier method of causing standard libraries to be excluded from consideration is to write-protect the corresponding ALI files. .. index:: -aO (gnatmake) :switch:`-aO{dir}` When searching for library and object files, look in directory ``dir``. The order in which library files are searched is described in :ref:`Search_Paths_for_gnatbind`. .. index:: Search paths, for gnatmake .. index:: -A (gnatmake) :switch:`-A{dir}` Equivalent to :switch:`-aL{dir}` :switch:`-aI{dir}`. .. index:: -I (gnatmake) :switch:`-I{dir}` Equivalent to :switch:`-aO{dir} -aI{dir}`. .. index:: -I- (gnatmake) .. index:: Source files, suppressing search :switch:`-I-` Do not look for source files in the directory containing the source file named in the command line. Do not look for ALI or object files in the directory where ``gnatmake`` was invoked. .. index:: -L (gnatmake) .. index:: Linker libraries :switch:`-L{dir}` Add directory ``dir`` to the list of directories in which the linker will search for libraries. This is equivalent to :switch:`-largs` :switch:`-L{dir}`. Furthermore, under Windows, the sources pointed to by the libraries path set in the registry are not searched for. .. index:: -nostdinc (gnatmake) :switch:`-nostdinc` Do not look for source files in the system default directory. .. index:: -nostdlib (gnatmake) :switch:`-nostdlib` Do not look for library files in the system default directory. .. index:: --RTS (gnatmake) :switch:`--RTS={rts-path}` Specifies the default location of the run-time library. GNAT looks for the run-time in the following directories, and stops as soon as a valid run-time is found (:file:`adainclude` or :file:`ada_source_path`, and :file:`adalib` or :file:`ada_object_path` present): * */$rts_path* * */$rts_path* * */rts-$rts_path* * The selected path is handled like a normal RTS path. .. _Mode_Switches_for_gnatmake: Mode Switches for ``gnatmake`` ------------------------------ The mode switches (referred to as ``mode_switches``) allow the inclusion of switches that are to be passed to the compiler itself, the binder or the linker. The effect of a mode switch is to cause all subsequent switches up to the end of the switch list, or up to the next mode switch, to be interpreted as switches to be passed on to the designated component of GNAT. .. index:: -cargs (gnatmake) :switch:`-cargs {switches}` Compiler switches. Here ``switches`` is a list of switches that are valid switches for ``gcc``. They will be passed on to all compile steps performed by ``gnatmake``. .. index:: -bargs (gnatmake) :switch:`-bargs {switches}` Binder switches. Here ``switches`` is a list of switches that are valid switches for ``gnatbind``. They will be passed on to all bind steps performed by ``gnatmake``. .. index:: -largs (gnatmake) :switch:`-largs {switches}` Linker switches. Here ``switches`` is a list of switches that are valid switches for ``gnatlink``. They will be passed on to all link steps performed by ``gnatmake``. .. index:: -margs (gnatmake) :switch:`-margs {switches}` Make switches. The switches are directly interpreted by ``gnatmake``, regardless of any previous occurrence of :switch:`-cargs`, :switch:`-bargs` or :switch:`-largs`. .. _Notes_on_the_Command_Line: Notes on the Command Line ------------------------- This section contains some additional useful notes on the operation of the ``gnatmake`` command. .. index:: Recompilation (by gnatmake) * If ``gnatmake`` finds no ALI files, it recompiles the main program and all other units required by the main program. This means that ``gnatmake`` can be used for the initial compile, as well as during subsequent steps of the development cycle. * If you enter ``gnatmake foo.adb``, where ``foo`` is a subunit or body of a generic unit, ``gnatmake`` recompiles :file:`foo.adb` (because it finds no ALI) and stops, issuing a warning. * In ``gnatmake`` the switch :switch:`-I` is used to specify both source and library file paths. Use :switch:`-aI` instead if you just want to specify source paths only and :switch:`-aO` if you want to specify library paths only. * ``gnatmake`` will ignore any files whose ALI file is write-protected. This may conveniently be used to exclude standard libraries from consideration and in particular it means that the use of the :switch:`-f` switch will not recompile these files unless :switch:`-a` is also specified. * ``gnatmake`` has been designed to make the use of Ada libraries particularly convenient. Assume you have an Ada library organized as follows: *obj-dir* contains the objects and ALI files for of your Ada compilation units, whereas *include-dir* contains the specs of these units, but no bodies. Then to compile a unit stored in ``main.adb``, which uses this Ada library you would just type: .. code-block:: sh $ gnatmake -aI`include-dir` -aL`obj-dir` main * Using ``gnatmake`` along with the :switch:`-m (minimal recompilation)` switch provides a mechanism for avoiding unnecessary recompilations. Using this switch, you can update the comments/format of your source files without having to recompile everything. Note, however, that adding or deleting lines in a source files may render its debugging info obsolete. If the file in question is a spec, the impact is rather limited, as that debugging info will only be useful during the elaboration phase of your program. For bodies the impact can be more significant. In all events, your debugger will warn you if a source file is more recent than the corresponding object, and alert you to the fact that the debugging information may be out of date. .. _How_gnatmake_Works: How ``gnatmake`` Works ---------------------- Generally ``gnatmake`` automatically performs all necessary recompilations and you don't need to worry about how it works. However, it may be useful to have some basic understanding of the ``gnatmake`` approach and in particular to understand how it uses the results of previous compilations without incorrectly depending on them. First a definition: an object file is considered *up to date* if the corresponding ALI file exists and if all the source files listed in the dependency section of this ALI file have time stamps matching those in the ALI file. This means that neither the source file itself nor any files that it depends on have been modified, and hence there is no need to recompile this file. ``gnatmake`` works by first checking if the specified main unit is up to date. If so, no compilations are required for the main unit. If not, ``gnatmake`` compiles the main program to build a new ALI file that reflects the latest sources. Then the ALI file of the main unit is examined to find all the source files on which the main program depends, and ``gnatmake`` recursively applies the above procedure on all these files. This process ensures that ``gnatmake`` only trusts the dependencies in an existing ALI file if they are known to be correct. Otherwise it always recompiles to determine a new, guaranteed accurate set of dependencies. As a result the program is compiled 'upside down' from what may be more familiar as the required order of compilation in some other Ada systems. In particular, clients are compiled before the units on which they depend. The ability of GNAT to compile in any order is critical in allowing an order of compilation to be chosen that guarantees that ``gnatmake`` will recompute a correct set of new dependencies if necessary. When invoking ``gnatmake`` with several ``file_names``, if a unit is imported by several of the executables, it will be recompiled at most once. Note: when using non-standard naming conventions (:ref:`Using_Other_File_Names`), changing through a configuration pragmas file the version of a source and invoking ``gnatmake`` to recompile may have no effect, if the previous version of the source is still accessible by ``gnatmake``. It may be necessary to use the switch -f. .. _Examples_of_gnatmake_Usage: Examples of ``gnatmake`` Usage ------------------------------ ``gnatmake hello.adb`` Compile all files necessary to bind and link the main program :file:`hello.adb` (containing unit ``Hello``) and bind and link the resulting object files to generate an executable file :file:`hello`. ``gnatmake main1 main2 main3`` Compile all files necessary to bind and link the main programs :file:`main1.adb` (containing unit ``Main1``), :file:`main2.adb` (containing unit ``Main2``) and :file:`main3.adb` (containing unit ``Main3``) and bind and link the resulting object files to generate three executable files :file:`main1`, :file:`main2` and :file:`main3`. ``gnatmake -q Main_Unit -cargs -O2 -bargs -l`` Compile all files necessary to bind and link the main program unit ``Main_Unit`` (from file :file:`main_unit.adb`). All compilations will be done with optimization level 2 and the order of elaboration will be listed by the binder. ``gnatmake`` will operate in quiet mode, not displaying commands it is executing. .. _Compiling_with_gcc: Compiling with ``gcc`` ====================== This section discusses how to compile Ada programs using the ``gcc`` command. It also describes the set of switches that can be used to control the behavior of the compiler. .. _Compiling_Programs: Compiling Programs ------------------ The first step in creating an executable program is to compile the units of the program using the ``gcc`` command. You must compile the following files: * the body file (:file:`.adb`) for a library level subprogram or generic subprogram * the spec file (:file:`.ads`) for a library level package or generic package that has no body * the body file (:file:`.adb`) for a library level package or generic package that has a body You need *not* compile the following files * the spec of a library unit which has a body * subunits because they are compiled as part of compiling related units. GNAT compiles package specs when the corresponding body is compiled, and subunits when the parent is compiled. .. index:: cannot generate code If you attempt to compile any of these files, you will get one of the following error messages (where :file:`fff` is the name of the file you compiled): .. code-block:: text cannot generate code for file fff (package spec) cannot generate code for file fff (missing subunits) cannot generate code for file fff (subprogram spec) cannot generate code for file fff (subunit) As indicated by the above error messages, if you want to submit one of these files to the compiler to check for correct semantics without generating code, then use the :switch:`-gnatc` switch. The basic command for compiling a file containing an Ada unit is: .. code-block:: sh $ gcc -c [switches] where ``file name`` is the name of the Ada file (usually having an extension :file:`.ads` for a spec or :file:`.adb` for a body). You specify the :switch:`-c` switch to tell ``gcc`` to compile, but not link, the file. The result of a successful compilation is an object file, which has the same name as the source file but an extension of :file:`.o` and an Ada Library Information (ALI) file, which also has the same name as the source file, but with :file:`.ali` as the extension. GNAT creates these two output files in the current directory, but you may specify a source file in any directory using an absolute or relative path specification containing the directory information. .. index:: gnat1 ``gcc`` is actually a driver program that looks at the extensions of the file arguments and loads the appropriate compiler. For example, the GNU C compiler is :file:`cc1`, and the Ada compiler is :file:`gnat1`. These programs are in directories known to the driver program (in some configurations via environment variables you set), but need not be in your path. The ``gcc`` driver also calls the assembler and any other utilities needed to complete the generation of the required object files. It is possible to supply several file names on the same ``gcc`` command. This causes ``gcc`` to call the appropriate compiler for each file. For example, the following command lists two separate files to be compiled: .. code-block:: sh $ gcc -c x.adb y.adb calls ``gnat1`` (the Ada compiler) twice to compile :file:`x.adb` and :file:`y.adb`. The compiler generates two object files :file:`x.o` and :file:`y.o` and the two ALI files :file:`x.ali` and :file:`y.ali`. Any switches apply to all the files listed, see :ref:`Switches_for_gcc` for a list of available ``gcc`` switches. .. _Search_Paths_and_the_Run-Time_Library_RTL: Search Paths and the Run-Time Library (RTL) ------------------------------------------- With the GNAT source-based library system, the compiler must be able to find source files for units that are needed by the unit being compiled. Search paths are used to guide this process. The compiler compiles one source file whose name must be given explicitly on the command line. In other words, no searching is done for this file. To find all other source files that are needed (the most common being the specs of units), the compiler examines the following directories, in the following order: * The directory containing the source file of the main unit being compiled (the file name on the command line). * Each directory named by an :switch:`-I` switch given on the ``gcc`` command line, in the order given. .. index:: ADA_PRJ_INCLUDE_FILE * Each of the directories listed in the text file whose name is given by the :envvar:`ADA_PRJ_INCLUDE_FILE` environment variable. :envvar:`ADA_PRJ_INCLUDE_FILE` is normally set by gnatmake or by the gnat driver when project files are used. It should not normally be set by other means. .. index:: ADA_INCLUDE_PATH * Each of the directories listed in the value of the :envvar:`ADA_INCLUDE_PATH` environment variable. Construct this value exactly as the :envvar:`PATH` environment variable: a list of directory names separated by colons (semicolons when working with the NT version). * The content of the :file:`ada_source_path` file which is part of the GNAT installation tree and is used to store standard libraries such as the GNAT Run Time Library (RTL) source files. See also :ref:`Installing_a_library`. Specifying the switch :switch:`-I-` inhibits the use of the directory containing the source file named in the command line. You can still have this directory on your search path, but in this case it must be explicitly requested with a :switch:`-I` switch. Specifying the switch :switch:`-nostdinc` inhibits the search of the default location for the GNAT Run Time Library (RTL) source files. The compiler outputs its object files and ALI files in the current working directory. Caution: The object file can be redirected with the :switch:`-o` switch; however, ``gcc`` and ``gnat1`` have not been coordinated on this so the :file:`ALI` file will not go to the right place. Therefore, you should avoid using the :switch:`-o` switch. .. index:: System.IO The packages ``Ada``, ``System``, and ``Interfaces`` and their children make up the GNAT RTL, together with the simple ``System.IO`` package used in the ``"Hello World"`` example. The sources for these units are needed by the compiler and are kept together in one directory. Not all of the bodies are needed, but all of the sources are kept together anyway. In a normal installation, you need not specify these directory names when compiling or binding. Either the environment variables or the built-in defaults cause these files to be found. In addition to the language-defined hierarchies (``System``, ``Ada`` and ``Interfaces``), the GNAT distribution provides a fourth hierarchy, consisting of child units of ``GNAT``. This is a collection of generally useful types, subprograms, etc. See the :title:`GNAT_Reference_Manual` for further details. Besides simplifying access to the RTL, a major use of search paths is in compiling sources from multiple directories. This can make development environments much more flexible. .. _Order_of_Compilation_Issues: Order of Compilation Issues --------------------------- If, in our earlier example, there was a spec for the ``hello`` procedure, it would be contained in the file :file:`hello.ads`; yet this file would not have to be explicitly compiled. This is the result of the model we chose to implement library management. Some of the consequences of this model are as follows: * There is no point in compiling specs (except for package specs with no bodies) because these are compiled as needed by clients. If you attempt a useless compilation, you will receive an error message. It is also useless to compile subunits because they are compiled as needed by the parent. * There are no order of compilation requirements: performing a compilation never obsoletes anything. The only way you can obsolete something and require recompilations is to modify one of the source files on which it depends. * There is no library as such, apart from the ALI files (:ref:`The_Ada_Library_Information_Files`, for information on the format of these files). For now we find it convenient to create separate ALI files, but eventually the information therein may be incorporated into the object file directly. * When you compile a unit, the source files for the specs of all units that it |withs|, all its subunits, and the bodies of any generics it instantiates must be available (reachable by the search-paths mechanism described above), or you will receive a fatal error message. .. _Examples: Examples -------- The following are some typical Ada compilation command line examples: .. code-block:: sh $ gcc -c xyz.adb Compile body in file :file:`xyz.adb` with all default options. .. code-block:: sh $ gcc -c -O2 -gnata xyz-def.adb Compile the child unit package in file :file:`xyz-def.adb` with extensive optimizations, and pragma ``Assert``/``Debug`` statements enabled. .. code-block:: sh $ gcc -c -gnatc abc-def.adb Compile the subunit in file :file:`abc-def.adb` in semantic-checking-only mode. .. _Switches_for_gcc: Compiler Switches ================= The ``gcc`` command accepts switches that control the compilation process. These switches are fully described in this section: first an alphabetical listing of all switches with a brief description, and then functionally grouped sets of switches with more detailed information. More switches exist for GCC than those documented here, especially for specific targets. However, their use is not recommended as they may change code generation in ways that are incompatible with the Ada run-time library, or can cause inconsistencies between compilation units. .. _Alphabetical_List_of_All_Switches: Alphabetical List of All Switches --------------------------------- .. index:: -b (gcc) :switch:`-b {target}` Compile your program to run on ``target``, which is the name of a system configuration. You must have a GNAT cross-compiler built if ``target`` is not the same as your host system. .. index:: -B (gcc) :switch:`-B{dir}` Load compiler executables (for example, ``gnat1``, the Ada compiler) from ``dir`` instead of the default location. Only use this switch when multiple versions of the GNAT compiler are available. See the "Options for Directory Search" section in the :title:`Using the GNU Compiler Collection (GCC)` manual for further details. You would normally use the :switch:`-b` or :switch:`-V` switch instead. .. index:: -c (gcc) :switch:`-c` Compile. Always use this switch when compiling Ada programs. Note: for some other languages when using ``gcc``, notably in the case of C and C++, it is possible to use use ``gcc`` without a :switch:`-c` switch to compile and link in one step. In the case of GNAT, you cannot use this approach, because the binder must be run and ``gcc`` cannot be used to run the GNAT binder. .. index:: -fcallgraph-info (gcc) :switch:`-fcallgraph-info[=su,da]` Makes the compiler output callgraph information for the program, on a per-file basis. The information is generated in the VCG format. It can be decorated with additional, per-node and/or per-edge information, if a list of comma-separated markers is additionally specified. When the ``su`` marker is specified, the callgraph is decorated with stack usage information; it is equivalent to :switch:`-fstack-usage`. When the ``da`` marker is specified, the callgraph is decorated with information about dynamically allocated objects. .. index:: -fdiagnostics-format (gcc) :switch:`-fdiagnostics-format=json` Makes GNAT emit warning and error messages as JSON. Inhibits printing of text warning and errors messages except if :switch:`-gnatv` or :switch:`-gnatl` are present. Uses absolute file paths when used along :switch:`-gnatef`. .. index:: -fdump-scos (gcc) :switch:`-fdump-scos` Generates SCO (Source Coverage Obligation) information in the ALI file. This information is used by advanced coverage tools. See unit :file:`SCOs` in the compiler sources for details in files :file:`scos.ads` and :file:`scos.adb`. .. index:: -fgnat-encodings (gcc) :switch:`-fgnat-encodings=[all|gdb|minimal]` This switch controls the balance between GNAT encodings and standard DWARF emitted in the debug information. .. index:: -flto (gcc) :switch:`-flto[={n}]` Enables Link Time Optimization. This switch must be used in conjunction with the :switch:`-Ox` switches (but not with the :switch:`-gnatn` switch since it is a full replacement for the latter) and instructs the compiler to defer most optimizations until the link stage. The advantage of this approach is that the compiler can do a whole-program analysis and choose the best interprocedural optimization strategy based on a complete view of the program, instead of a fragmentary view with the usual approach. This can also speed up the compilation of big programs and reduce the size of the executable, compared with a traditional per-unit compilation with inlining across units enabled by the :switch:`-gnatn` switch. The drawback of this approach is that it may require more memory and that the debugging information generated by ``-g`` with it might be hardly usable. The switch, as well as the accompanying :switch:`-Ox` switches, must be specified both for the compilation and the link phases. If the ``n`` parameter is specified, the optimization and final code generation at link time are executed using ``n`` parallel jobs by means of an installed ``make`` program. .. index:: -fno-inline (gcc) :switch:`-fno-inline` Suppresses all inlining, unless requested with pragma ``Inline_Always``. The effect is enforced regardless of other optimization or inlining switches. Note that inlining can also be suppressed on a finer-grained basis with pragma ``No_Inline``. .. index:: -fno-inline-functions (gcc) :switch:`-fno-inline-functions` Suppresses automatic inlining of subprograms, which is enabled if :switch:`-O3` is used. .. index:: -fno-inline-small-functions (gcc) :switch:`-fno-inline-small-functions` Suppresses automatic inlining of small subprograms, which is enabled if :switch:`-O2` is used. .. index:: -fno-inline-functions-called-once (gcc) :switch:`-fno-inline-functions-called-once` Suppresses inlining of subprograms local to the unit and called once from within it, which is enabled if :switch:`-O1` is used. .. index:: -fno-ivopts (gcc) :switch:`-fno-ivopts` Suppresses high-level loop induction variable optimizations, which are enabled if :switch:`-O1` is used. These optimizations are generally profitable but, for some specific cases of loops with numerous uses of the iteration variable that follow a common pattern, they may end up destroying the regularity that could be exploited at a lower level and thus producing inferior code. .. index:: -fno-strict-aliasing (gcc) :switch:`-fno-strict-aliasing` Causes the compiler to avoid assumptions regarding non-aliasing of objects of different types. See :ref:`Optimization_and_Strict_Aliasing` for details. .. index:: -fno-strict-overflow (gcc) :switch:`-fno-strict-overflow` Causes the compiler to avoid assumptions regarding the rules of signed integer overflow. These rules specify that signed integer overflow will result in a Constraint_Error exception at run time and are enforced in default mode by the compiler, so this switch should not be necessary in normal operating mode. It might be useful in conjunction with :switch:`-gnato0` for very peculiar cases of low-level programming. .. index:: -fstack-check (gcc) :switch:`-fstack-check` Activates stack checking. See :ref:`Stack_Overflow_Checking` for details. .. index:: -fstack-usage (gcc) :switch:`-fstack-usage` Makes the compiler output stack usage information for the program, on a per-subprogram basis. See :ref:`Static_Stack_Usage_Analysis` for details. .. index:: -g (gcc) :switch:`-g` Generate debugging information. This information is stored in the object file and copied from there to the final executable file by the linker, where it can be read by the debugger. You must use the :switch:`-g` switch if you plan on using the debugger. .. index:: -gnat05 (gcc) :switch:`-gnat05` Allow full Ada 2005 features. .. index:: -gnat12 (gcc) :switch:`-gnat12` Allow full Ada 2012 features. .. index:: -gnat83 (gcc) .. index:: -gnat2005 (gcc) :switch:`-gnat2005` Allow full Ada 2005 features (same as :switch:`-gnat05`) .. index:: -gnat2012 (gcc) :switch:`-gnat2012` Allow full Ada 2012 features (same as :switch:`-gnat12`) .. index:: -gnat2022 (gcc) :switch:`-gnat2022` Allow full Ada 2022 features :switch:`-gnat83` Enforce Ada 83 restrictions. .. index:: -gnat95 (gcc) :switch:`-gnat95` Enforce Ada 95 restrictions. Note: for compatibility with some Ada 95 compilers which support only the ``overriding`` keyword of Ada 2005, the :switch:`-gnatd.D` switch can be used along with :switch:`-gnat95` to achieve a similar effect with GNAT. :switch:`-gnatd.D` instructs GNAT to consider ``overriding`` as a keyword and handle its associated semantic checks, even in Ada 95 mode. .. index:: -gnata (gcc) :switch:`-gnata` Assertions enabled. ``Pragma Assert`` and ``pragma Debug`` to be activated. Note that these pragmas can also be controlled using the configuration pragmas ``Assertion_Policy`` and ``Debug_Policy``. It also activates pragmas ``Check``, ``Precondition``, and ``Postcondition``. Note that these pragmas can also be controlled using the configuration pragma ``Check_Policy``. In Ada 2012, it also activates all assertions defined in the RM as aspects: preconditions, postconditions, type invariants and (sub)type predicates. In all Ada modes, corresponding pragmas for type invariants and (sub)type predicates are also activated. The default is that all these assertions are disabled, and have no effect, other than being checked for syntactic validity, and in the case of subtype predicates, constructions such as membership tests still test predicates even if assertions are turned off. .. index:: -gnatA (gcc) :switch:`-gnatA` Avoid processing :file:`gnat.adc`. If a :file:`gnat.adc` file is present, it will be ignored. .. index:: -gnatb (gcc) :switch:`-gnatb` Generate brief messages to :file:`stderr` even if verbose mode set. .. index:: -gnatB (gcc) :switch:`-gnatB` Assume no invalid (bad) values except for 'Valid attribute use (:ref:`Validity_Checking`). .. index:: -gnatc (gcc) :switch:`-gnatc` Check syntax and semantics only (no code generation attempted). When the compiler is invoked by ``gnatmake``, if the switch :switch:`-gnatc` is only given to the compiler (after :switch:`-cargs` or in package Compiler of the project file), ``gnatmake`` will fail because it will not find the object file after compilation. If ``gnatmake`` is called with :switch:`-gnatc` as a builder switch (before :switch:`-cargs` or in package Builder of the project file) then ``gnatmake`` will not fail because it will not look for the object files after compilation, and it will not try to build and link. .. index:: -gnatC (gcc) :switch:`-gnatC` Generate CodePeer intermediate format (no code generation attempted). This switch will generate an intermediate representation suitable for use by CodePeer (:file:`.scil` files). This switch is not compatible with code generation (it will, among other things, disable some switches such as ``-gnatn``, and enable others such as ``-gnata``). .. index:: -gnatd (gcc) :switch:`-gnatd` Specify debug options for the compiler. The string of characters after the :switch:`-gnatd` specifies the specific debug options. The possible characters are 0-9, a-z, A-Z, optionally preceded by a dot or underscore. See compiler source file :file:`debug.adb` for details of the implemented debug options. Certain debug options are relevant to application programmers, and these are documented at appropriate points in this user's guide. .. index:: -gnatD[nn] (gcc) :switch:`-gnatD` Create expanded source files for source level debugging. This switch also suppresses generation of cross-reference information (see :switch:`-gnatx`). Note that this switch is not allowed if a previous ``-gnatR`` switch has been given, since these two switches are not compatible. .. index:: -gnateA (gcc) :switch:`-gnateA` Check that the actual parameters of a subprogram call are not aliases of one another. To qualify as aliasing, their memory locations must be identical or overlapping, at least one of the corresponding formal parameters must be of mode OUT or IN OUT, and at least one of the corresponding formal parameters must have its parameter passing mechanism not specified. .. code-block:: ada type Rec_Typ is record Data : Integer := 0; end record; function Self (Val : Rec_Typ) return Rec_Typ is begin return Val; end Self; procedure Detect_Aliasing (Val_1 : in out Rec_Typ; Val_2 : Rec_Typ) is begin null; end Detect_Aliasing; Obj : Rec_Typ; Detect_Aliasing (Obj, Obj); Detect_Aliasing (Obj, Self (Obj)); In the example above, the first call to ``Detect_Aliasing`` fails with a ``Program_Error`` at run time because the actuals for ``Val_1`` and ``Val_2`` denote the same object. The second call executes without raising an exception because ``Self(Obj)`` produces an anonymous object which does not share the memory location of ``Obj``. .. index:: -gnateb (gcc) :switch:`-gnateb` Store configuration files by their basename in ALI files. This switch is used for instance by gprbuild for distributed builds in order to prevent issues where machine-specific absolute paths could end up being stored in ALI files. .. index:: -gnatec (gcc) :switch:`-gnatec={path}` Specify a configuration pragma file (the equal sign is optional) (:ref:`The_Configuration_Pragmas_Files`). .. index:: -gnateC (gcc) :switch:`-gnateC` Generate CodePeer messages in a compiler-like format. This switch is only effective if :switch:`-gnatcC` is also specified and requires an installation of CodePeer. .. index:: -gnated (gcc) :switch:`-gnated` Disable atomic synchronization .. index:: -gnateD (gcc) :switch:`-gnateDsymbol[={value}]` Defines a symbol, associated with ``value``, for preprocessing. (:ref:`Integrated_Preprocessing`). .. index:: -gnateE (gcc) :switch:`-gnateE` Generate extra information in exception messages. In particular, display extra column information and the value and range associated with index and range check failures, and extra column information for access checks. In cases where the compiler is able to determine at compile time that a check will fail, it gives a warning, and the extra information is not produced at run time. .. index:: -gnatef (gcc) :switch:`-gnatef` Display full source path name in brief error messages and absolute paths in :switch:`-fdiagnostics-format=json`'s output. .. index:: -gnateF (gcc) :switch:`-gnateF` Check for overflow on all floating-point operations, including those for unconstrained predefined types. See description of pragma ``Check_Float_Overflow`` in GNAT RM. .. index:: -gnateg (gcc) :switch:`-gnateg` :switch:`-gnatceg` The :switch:`-gnatc` switch must always be specified before this switch, e.g. :switch:`-gnatceg`. Generate a C header from the Ada input file. See :ref:`Generating_C_Headers_for_Ada_Specifications` for more information. .. index:: -gnateG (gcc) :switch:`-gnateG[bce]` Save result of preprocessing in a text file. An optional character (b, c, or e) can be appended to indicate that filtered lines are to be replaced by blank lines, comment lines that include the filtered line, or empty comment lines ("--!"), respectively. The default is to replace filtered lines with blank lines. .. index:: -gnateH (gcc) :switch:`-gnateH` Set the threshold from which the RM 13.5.1(13.3/2) clause applies to 64. This is useful only on 64-bit platforms where this threshold is 128, but used to be 64 in earlier versions of the compiler. .. index:: -gnatei (gcc) :switch:`-gnatei{nnn}` Set maximum number of instantiations during compilation of a single unit to ``nnn``. This may be useful in increasing the default maximum of 8000 for the rare case when a single unit legitimately exceeds this limit. .. index:: -gnateI (gcc) :switch:`-gnateI{nnn}` Indicates that the source is a multi-unit source and that the index of the unit to compile is ``nnn``. ``nnn`` needs to be a positive number and need to be a valid index in the multi-unit source. .. index:: -gnatel (gcc) :switch:`-gnatel` This switch can be used with the static elaboration model to issue info messages showing where implicit ``pragma Elaborate`` and ``pragma Elaborate_All`` are generated. This is useful in diagnosing elaboration circularities caused by these implicit pragmas when using the static elaboration model. See the section in this guide on elaboration checking for further details. These messages are not generated by default, and are intended only for temporary use when debugging circularity problems. .. index:: -gnatel (gcc) :switch:`-gnateL` This switch turns off the info messages about implicit elaboration pragmas. .. index:: -gnatem (gcc) :switch:`-gnatem={path}` Specify a mapping file (the equal sign is optional) (:ref:`Units_to_Sources_Mapping_Files`). .. index:: -gnatep (gcc) :switch:`-gnatep={file}` Specify a preprocessing data file (the equal sign is optional) (:ref:`Integrated_Preprocessing`). .. index:: -gnateP (gcc) :switch:`-gnateP` Turn categorization dependency errors into warnings. Ada requires that units that WITH one another have compatible categories, for example a Pure unit cannot WITH a Preelaborate unit. If this switch is used, these errors become warnings (which can be ignored, or suppressed in the usual manner). This can be useful in some specialized circumstances such as the temporary use of special test software. .. index:: -gnateS (gcc) :switch:`-gnateS` Synonym of :switch:`-fdump-scos`, kept for backwards compatibility. .. index:: -gnatet=file (gcc) :switch:`-gnatet={path}` Generate target dependent information. The format of the output file is described in the section about switch :switch:`-gnateT`. .. index:: -gnateT (gcc) :switch:`-gnateT={path}` Read target dependent information, such as endianness or sizes and alignments of base type. If this switch is passed, the default target dependent information of the compiler is replaced by the one read from the input file. This is used by tools other than the compiler, e.g. to do semantic analysis of programs that will run on some other target than the machine on which the tool is run. The following target dependent values should be defined, where ``Nat`` denotes a natural integer value, ``Pos`` denotes a positive integer value, and fields marked with a question mark are boolean fields, where a value of 0 is False, and a value of 1 is True: :: Bits_BE : Nat; -- Bits stored big-endian? Bits_Per_Unit : Pos; -- Bits in a storage unit Bits_Per_Word : Pos; -- Bits in a word Bytes_BE : Nat; -- Bytes stored big-endian? Char_Size : Pos; -- Standard.Character'Size Double_Float_Alignment : Nat; -- Alignment of double float Double_Scalar_Alignment : Nat; -- Alignment of double length scalar Double_Size : Pos; -- Standard.Long_Float'Size Float_Size : Pos; -- Standard.Float'Size Float_Words_BE : Nat; -- Float words stored big-endian? Int_Size : Pos; -- Standard.Integer'Size Long_Double_Size : Pos; -- Standard.Long_Long_Float'Size Long_Long_Long_Size : Pos; -- Standard.Long_Long_Long_Integer'Size Long_Long_Size : Pos; -- Standard.Long_Long_Integer'Size Long_Size : Pos; -- Standard.Long_Integer'Size Maximum_Alignment : Pos; -- Maximum permitted alignment Max_Unaligned_Field : Pos; -- Maximum size for unaligned bit field Pointer_Size : Pos; -- System.Address'Size Short_Enums : Nat; -- Foreign enums use short size? Short_Size : Pos; -- Standard.Short_Integer'Size Strict_Alignment : Nat; -- Strict alignment? System_Allocator_Alignment : Nat; -- Alignment for malloc calls Wchar_T_Size : Pos; -- Interfaces.C.wchar_t'Size Words_BE : Nat; -- Words stored big-endian? ``Bits_Per_Unit`` is the number of bits in a storage unit, the equivalent of GCC macro ``BITS_PER_UNIT`` documented as follows: `Define this macro to be the number of bits in an addressable storage unit (byte); normally 8.` ``Bits_Per_Word`` is the number of bits in a machine word, the equivalent of GCC macro ``BITS_PER_WORD`` documented as follows: `Number of bits in a word; normally 32.` ``Double_Float_Alignment``, if not zero, is the maximum alignment that the compiler can choose by default for a 64-bit floating-point type or object. ``Double_Scalar_Alignment``, if not zero, is the maximum alignment that the compiler can choose by default for a 64-bit or larger scalar type or object. ``Maximum_Alignment`` is the maximum alignment that the compiler can choose by default for a type or object, which is also the maximum alignment that can be specified in GNAT. It is computed for GCC back ends as ``BIGGEST_ALIGNMENT / BITS_PER_UNIT`` where GCC macro ``BIGGEST_ALIGNMENT`` is documented as follows: `Biggest alignment that any data type can require on this machine, in bits.` ``Max_Unaligned_Field`` is the maximum size for unaligned bit field, which is 64 for the majority of GCC targets (but can be different on some targets). ``Strict_Alignment`` is the equivalent of GCC macro ``STRICT_ALIGNMENT`` documented as follows: `Define this macro to be the value 1 if instructions will fail to work if given data not on the nominal alignment. If instructions will merely go slower in that case, define this macro as 0.` ``System_Allocator_Alignment`` is the guaranteed alignment of data returned by calls to ``malloc``. The format of the input file is as follows. First come the values of the variables defined above, with one line per value: :: name value where ``name`` is the name of the parameter, spelled out in full, and cased as in the above list, and ``value`` is an unsigned decimal integer. Two or more blanks separates the name from the value. All the variables must be present, in alphabetical order (i.e. the same order as the list above). Then there is a blank line to separate the two parts of the file. Then come the lines showing the floating-point types to be registered, with one line per registered mode: :: name digs float_rep size alignment where ``name`` is the string name of the type (which can have single spaces embedded in the name, e.g. long double), ``digs`` is the number of digits for the floating-point type, ``float_rep`` is the float representation (I for IEEE-754-Binary, which is the only one supported at this time), ``size`` is the size in bits, ``alignment`` is the alignment in bits. The name is followed by at least two blanks, fields are separated by at least one blank, and a LF character immediately follows the alignment field. Here is an example of a target parameterization file: :: Bits_BE 0 Bits_Per_Unit 8 Bits_Per_Word 64 Bytes_BE 0 Char_Size 8 Double_Float_Alignment 0 Double_Scalar_Alignment 0 Double_Size 64 Float_Size 32 Float_Words_BE 0 Int_Size 64 Long_Double_Size 128 Long_Long_Long_Size 128 Long_Long_Size 64 Long_Size 64 Maximum_Alignment 16 Max_Unaligned_Field 64 Pointer_Size 64 Short_Size 16 Strict_Alignment 0 System_Allocator_Alignment 16 Wchar_T_Size 32 Words_BE 0 float 15 I 64 64 double 15 I 64 64 long double 18 I 80 128 TF 33 I 128 128 .. index:: -gnateu (gcc) :switch:`-gnateu` Ignore unrecognized validity, warning, and style switches that appear after this switch is given. This may be useful when compiling sources developed on a later version of the compiler with an earlier version. Of course the earlier version must support this switch. .. index:: -gnateV (gcc) :switch:`-gnateV` Check that all actual parameters of a subprogram call are valid according to the rules of validity checking (:ref:`Validity_Checking`). .. index:: -gnateY (gcc) :switch:`-gnateY` Ignore all STYLE_CHECKS pragmas. Full legality checks are still carried out, but the pragmas have no effect on what style checks are active. This allows all style checking options to be controlled from the command line. .. index:: -gnatE (gcc) :switch:`-gnatE` Dynamic elaboration checking mode enabled. For further details see :ref:`Elaboration_Order_Handling_in_GNAT`. .. index:: -gnatf (gcc) :switch:`-gnatf` Full errors. Multiple errors per line, all undefined references, do not attempt to suppress cascaded errors. .. index:: -gnatF (gcc) :switch:`-gnatF` Externals names are folded to all uppercase. .. index:: -gnatg (gcc) :switch:`-gnatg` Internal GNAT implementation mode. This should not be used for applications programs, it is intended only for use by the compiler and its run-time library. For documentation, see the GNAT sources. Note that :switch:`-gnatg` implies :switch:`-gnatw.ge` and :switch:`-gnatyg` so that all standard warnings and all standard style options are turned on. All warnings and style messages are treated as errors. .. index:: -gnatG[nn] (gcc) :switch:`-gnatG=nn` List generated expanded code in source form. .. index:: -gnath (gcc) :switch:`-gnath` Output usage information. The output is written to :file:`stdout`. .. index:: -gnatH (gcc) :switch:`-gnatH` Legacy elaboration-checking mode enabled. When this switch is in effect, the pre-18.x access-before-elaboration model becomes the de facto model. For further details see :ref:`Elaboration_Order_Handling_in_GNAT`. .. index:: -gnati (gcc) :switch:`-gnati{c}` Identifier character set (``c`` = 1/2/3/4/5/9/p/8/f/n/w). For details of the possible selections for ``c``, see :ref:`Character_Set_Control`. .. index:: -gnatI (gcc) :switch:`-gnatI` Ignore representation clauses. When this switch is used, representation clauses are treated as comments. This is useful when initially porting code where you want to ignore rep clause problems, and also for compiling foreign code. The representation clauses that are ignored are: enumeration_representation_clause, record_representation_clause, and attribute_definition_clause for the following attributes: Address, Alignment, Bit_Order, Component_Size, Machine_Radix, Object_Size, Scalar_Storage_Order, Size, Small, Stream_Size, and Value_Size. Pragma Default_Scalar_Storage_Order is also ignored. Note that this option should be used only for compiling -- the code is likely to malfunction at run time. .. index:: -gnatjnn (gcc) :switch:`-gnatj{nn}` Reformat error messages to fit on ``nn`` character lines .. index:: -gnatJ (gcc) :switch:`-gnatJ` Permissive elaboration-checking mode enabled. When this switch is in effect, the post-18.x access-before-elaboration model ignores potential issues with: - Accept statements - Activations of tasks defined in instances - Assertion pragmas - Calls from within an instance to its enclosing context - Calls through generic formal parameters - Calls to subprograms defined in instances - Entry calls - Indirect calls using 'Access - Requeue statements - Select statements - Synchronous task suspension and does not emit compile-time diagnostics or run-time checks. For further details see :ref:`Elaboration_Order_Handling_in_GNAT`. .. index:: -gnatk (gcc) :switch:`-gnatk={n}` Limit file names to ``n`` (1-999) characters (``k`` = krunch). .. index:: -gnatl (gcc) :switch:`-gnatl` Output full source listing with embedded error messages. .. index:: -gnatL (gcc) :switch:`-gnatL` Used in conjunction with -gnatG or -gnatD to intersperse original source lines (as comment lines with line numbers) in the expanded source output. .. index:: -gnatm (gcc) :switch:`-gnatm={n}` Limit number of detected error or warning messages to ``n`` where ``n`` is in the range 1..999999. The default setting if no switch is given is 9999. If the number of warnings reaches this limit, then a message is output and further warnings are suppressed, but the compilation is continued. If the number of error messages reaches this limit, then a message is output and the compilation is abandoned. The equal sign here is optional. A value of zero means that no limit applies. .. index:: -gnatn (gcc) :switch:`-gnatn[12]` Activate inlining across units for subprograms for which pragma ``Inline`` is specified. This inlining is performed by the GCC back end. An optional digit sets the inlining level: 1 for moderate inlining across units or 2 for full inlining across units. If no inlining level is specified, the compiler will pick it based on the optimization level. .. index:: -gnatN (gcc) :switch:`-gnatN` Activate front end inlining for subprograms for which pragma ``Inline`` is specified. This inlining is performed by the front end and will be visible in the :switch:`-gnatG` output. When using a gcc-based back end, then the use of :switch:`-gnatN` is deprecated, and the use of :switch:`-gnatn` is preferred. Historically front end inlining was more extensive than the gcc back end inlining, but that is no longer the case. .. index:: -gnato0 (gcc) :switch:`-gnato0` Suppresses overflow checking. This causes the behavior of the compiler to match the default for older versions where overflow checking was suppressed by default. This is equivalent to having ``pragma Suppress (Overflow_Check)`` in a configuration pragma file. .. index:: -gnato?? (gcc) :switch:`-gnato??` Set default mode for handling generation of code to avoid intermediate arithmetic overflow. Here ``??`` is two digits, a single digit, or nothing. Each digit is one of the digits ``1`` through ``3``: ===== =============================================================== Digit Interpretation ----- --------------------------------------------------------------- *1* All intermediate overflows checked against base type (``STRICT``) *2* Minimize intermediate overflows (``MINIMIZED``) *3* Eliminate intermediate overflows (``ELIMINATED``) ===== =============================================================== If only one digit appears, then it applies to all cases; if two digits are given, then the first applies outside assertions, pre/postconditions, and type invariants, and the second applies within assertions, pre/postconditions, and type invariants. If no digits follow the :switch:`-gnato`, then it is equivalent to :switch:`-gnato11`, causing all intermediate overflows to be handled in strict mode. This switch also causes arithmetic overflow checking to be performed (as though ``pragma Unsuppress (Overflow_Check)`` had been specified). The default if no option :switch:`-gnato` is given is that overflow handling is in ``STRICT`` mode (computations done using the base type), and that overflow checking is enabled. Note that division by zero is a separate check that is not controlled by this switch (divide-by-zero checking is on by default). See also :ref:`Specifying_the_Desired_Mode`. .. index:: -gnatp (gcc) :switch:`-gnatp` Suppress all checks. See :ref:`Run-Time_Checks` for details. This switch has no effect if cancelled by a subsequent :switch:`-gnat-p` switch. .. index:: -gnat-p (gcc) :switch:`-gnat-p` Cancel effect of previous :switch:`-gnatp` switch. .. index:: -gnatq (gcc) :switch:`-gnatq` Don't quit. Try semantics, even if parse errors. .. index:: -gnatQ (gcc) :switch:`-gnatQ` Don't quit. Generate :file:`ALI` and tree files even if illegalities. Note that code generation is still suppressed in the presence of any errors, so even with :switch:`-gnatQ` no object file is generated. .. index:: -gnatr (gcc) :switch:`-gnatr` Treat pragma Restrictions as Restriction_Warnings. .. index:: -gnatR (gcc) :switch:`-gnatR[0|1|2|3|4][e][h][m][j][s]` Output representation information for declared types, objects and subprograms. Note that this switch is not allowed if a previous :switch:`-gnatD` switch has been given, since these two switches are not compatible. .. index:: -gnats (gcc) :switch:`-gnats` Syntax check only. .. index:: -gnatS (gcc) :switch:`-gnatS` Print package Standard. .. index:: -gnatT (gcc) :switch:`-gnatT{nnn}` All compiler tables start at ``nnn`` times usual starting size. .. index:: -gnatu (gcc) :switch:`-gnatu` List units for this compilation. .. index:: -gnatU (gcc) :switch:`-gnatU` Tag all error messages with the unique string 'error:' .. index:: -gnatv (gcc) :switch:`-gnatv` Verbose mode. Full error output with source lines to :file:`stdout`. .. index:: -gnatV (gcc) :switch:`-gnatV` Control level of validity checking (:ref:`Validity_Checking`). .. index:: -gnatw (gcc) :switch:`-gnatw{xxx}` Warning mode where ``xxx`` is a string of option letters that denotes the exact warnings that are enabled or disabled (:ref:`Warning_Message_Control`). .. index:: -gnatW (gcc) :switch:`-gnatW{e}` Wide character encoding method (``e``\ =n/h/u/s/e/8). .. index:: -gnatx (gcc) :switch:`-gnatx` Suppress generation of cross-reference information. .. index:: -gnatX (gcc) :switch:`-gnatX` Enable core GNAT implementation extensions and latest Ada version. .. index:: -gnatX0 (gcc) :switch:`-gnatX0` Enable all GNAT implementation extensions and latest Ada version. .. index:: -gnaty (gcc) :switch:`-gnaty` Enable built-in style checks (:ref:`Style_Checking`). .. index:: -gnatz (gcc) :switch:`-gnatz{m}` Distribution stub generation and compilation (``m``\ =r/c for receiver/caller stubs). .. index:: -I (gcc) :switch:`-I{dir}` .. index:: RTL Direct GNAT to search the ``dir`` directory for source files needed by the current compilation (see :ref:`Search_Paths_and_the_Run-Time_Library_RTL`). .. index:: -I- (gcc) :switch:`-I-` .. index:: RTL Except for the source file named in the command line, do not look for source files in the directory containing the source file named in the command line (see :ref:`Search_Paths_and_the_Run-Time_Library_RTL`). .. index:: -o (gcc) :switch:`-o {file}` This switch is used in ``gcc`` to redirect the generated object file and its associated ALI file. Beware of this switch with GNAT, because it may cause the object file and ALI file to have different names which in turn may confuse the binder and the linker. .. index:: -nostdinc (gcc) :switch:`-nostdinc` Inhibit the search of the default location for the GNAT Run Time Library (RTL) source files. .. index:: -nostdlib (gcc) :switch:`-nostdlib` Inhibit the search of the default location for the GNAT Run Time Library (RTL) ALI files. .. index:: -O (gcc) :switch:`-O[{n}]` ``n`` controls the optimization level: ======= ================================================================== *n* Effect ------- ------------------------------------------------------------------ *0* No optimization, the default setting if no :switch:`-O` appears. *1* Moderate optimization, same as :switch:`-O` without an operand. A good compromise between code quality and compilation time. *2* Extensive optimization, should improve execution time, possibly at the cost of substantially increased compilation time. *3* Full optimization, may further improve execution time, possibly at the cost of substantially larger generated code. *s* Optimize for size (code and data) rather than speed. *z* Optimize aggressively for size (code and data) rather than speed. *g* Optimize for debugging experience rather than speed. ======= ================================================================== See also :ref:`Optimization_Levels`. .. index:: -pass-exit-codes (gcc) :switch:`-pass-exit-codes` Catch exit codes from the compiler and use the most meaningful as exit status. .. index:: --RTS (gcc) :switch:`--RTS={rts-path}` Specifies the default location of the run-time library. Same meaning as the equivalent ``gnatmake`` flag (:ref:`Switches_for_gnatmake`). .. index:: -S (gcc) :switch:`-S` Used in place of :switch:`-c` to cause the assembler source file to be generated, using :file:`.s` as the extension, instead of the object file. This may be useful if you need to examine the generated assembly code. .. index:: -fverbose-asm (gcc) :switch:`-fverbose-asm` Used in conjunction with :switch:`-S` to cause the generated assembly code file to be annotated with variable names, making it significantly easier to follow. .. index:: -v (gcc) :switch:`-v` Show commands generated by the ``gcc`` driver. Normally used only for debugging purposes or if you need to be sure what version of the compiler you are executing. .. index:: -V (gcc) :switch:`-V {ver}` Execute ``ver`` version of the compiler. This is the ``gcc`` version, not the GNAT version. .. index:: -w (gcc) :switch:`-w` Turn off warnings generated by the back end of the compiler. Use of this switch also causes the default for front end warnings to be set to suppress (as though :switch:`-gnatws` had appeared at the start of the options). .. index:: Combining GNAT switches You may combine a sequence of GNAT switches into a single switch. For example, the combined switch :: -gnatofi3 is equivalent to specifying the following sequence of switches: :: -gnato -gnatf -gnati3 The following restrictions apply to the combination of switches in this manner: * The switch :switch:`-gnatc` if combined with other switches must come first in the string. * The switch :switch:`-gnats` if combined with other switches must come first in the string. * The switches :switch:`-gnatzc` and :switch:`-gnatzr` may not be combined with any other switches, and only one of them may appear in the command line. * The switch :switch:`-gnat-p` may not be combined with any other switch. * Once a 'y' appears in the string (that is a use of the :switch:`-gnaty` switch), then all further characters in the switch are interpreted as style modifiers (see description of :switch:`-gnaty`). * Once a 'd' appears in the string (that is a use of the :switch:`-gnatd` switch), then all further characters in the switch are interpreted as debug flags (see description of :switch:`-gnatd`). * Once a 'w' appears in the string (that is a use of the :switch:`-gnatw` switch), then all further characters in the switch are interpreted as warning mode modifiers (see description of :switch:`-gnatw`). * Once a 'V' appears in the string (that is a use of the :switch:`-gnatV` switch), then all further characters in the switch are interpreted as validity checking options (:ref:`Validity_Checking`). * Option 'em', 'ec', 'ep', 'l=' and 'R' must be the last options in a combined list of options. .. _Output_and_Error_Message_Control: Output and Error Message Control -------------------------------- .. index:: stderr The standard default format for error messages is called 'brief format'. Brief format messages are written to :file:`stderr` (the standard error file) and have the following form: :: e.adb:3:04: Incorrect spelling of keyword "function" e.adb:4:20: ";" should be "is" The first integer after the file name is the line number in the file, and the second integer is the column number within the line. ``GNAT Studio`` can parse the error messages and point to the referenced character. The following switches provide control over the error message format: .. index:: -gnatv (gcc) :switch:`-gnatv` The ``v`` stands for verbose. The effect of this setting is to write long-format error messages to :file:`stdout` (the standard output file). The same program compiled with the :switch:`-gnatv` switch would generate: :: 3. funcion X (Q : Integer) | >>> Incorrect spelling of keyword "function" 4. return Integer; | >>> ";" should be "is" The vertical bar indicates the location of the error, and the ``>>>`` prefix can be used to search for error messages. When this switch is used the only source lines output are those with errors. .. index:: -gnatl (gcc) :switch:`-gnatl` The ``l`` stands for list. This switch causes a full listing of the file to be generated. In the case where a body is compiled, the corresponding spec is also listed, along with any subunits. Typical output from compiling a package body :file:`p.adb` might look like:: Compiling: p.adb 1. package body p is 2. procedure a; 3. procedure a is separate; 4. begin 5. null | >>> missing ";" 6. end; Compiling: p.ads 1. package p is 2. pragma Elaborate_Body | >>> missing ";" 3. end p; Compiling: p-a.adb 1. separate p | >>> missing "(" 2. procedure a is 3. begin 4. null | >>> missing ";" 5. end; When you specify the :switch:`-gnatv` or :switch:`-gnatl` switches and standard output is redirected, a brief summary is written to :file:`stderr` (standard error) giving the number of error messages and warning messages generated. .. index:: -gnatl=fname (gcc) :switch:`-gnatl={fname}` This has the same effect as :switch:`-gnatl` except that the output is written to a file instead of to standard output. If the given name :file:`fname` does not start with a period, then it is the full name of the file to be written. If :file:`fname` is an extension, it is appended to the name of the file being compiled. For example, if file :file:`xyz.adb` is compiled with :switch:`-gnatl=.lst`, then the output is written to file xyz.adb.lst. .. index:: -gnatU (gcc) :switch:`-gnatU` This switch forces all error messages to be preceded by the unique string 'error:'. This means that error messages take a few more characters in space, but allows easy searching for and identification of error messages. .. index:: -gnatb (gcc) :switch:`-gnatb` The ``b`` stands for brief. This switch causes GNAT to generate the brief format error messages to :file:`stderr` (the standard error file) as well as the verbose format message or full listing (which as usual is written to :file:`stdout`, the standard output file). .. index:: -gnatm (gcc) :switch:`-gnatm={n}` The ``m`` stands for maximum. ``n`` is a decimal integer in the range of 1 to 999999 and limits the number of error or warning messages to be generated. For example, using :switch:`-gnatm2` might yield :: e.adb:3:04: Incorrect spelling of keyword "function" e.adb:5:35: missing ".." fatal error: maximum number of errors detected compilation abandoned The default setting if no switch is given is 9999. If the number of warnings reaches this limit, then a message is output and further warnings are suppressed, but the compilation is continued. If the number of error messages reaches this limit, then a message is output and the compilation is abandoned. A value of zero means that no limit applies. Note that the equal sign is optional, so the switches :switch:`-gnatm2` and :switch:`-gnatm=2` are equivalent. .. index:: -gnatf (gcc) :switch:`-gnatf` .. index:: Error messages, suppressing The ``f`` stands for full. Normally, the compiler suppresses error messages that are likely to be redundant. This switch causes all error messages to be generated. In particular, in the case of references to undefined variables. If a given variable is referenced several times, the normal format of messages is :: e.adb:7:07: "V" is undefined (more references follow) where the parenthetical comment warns that there are additional references to the variable ``V``. Compiling the same program with the :switch:`-gnatf` switch yields :: e.adb:7:07: "V" is undefined e.adb:8:07: "V" is undefined e.adb:8:12: "V" is undefined e.adb:8:16: "V" is undefined e.adb:9:07: "V" is undefined e.adb:9:12: "V" is undefined The :switch:`-gnatf` switch also generates additional information for some error messages. Some examples are: * Details on possibly non-portable unchecked conversion * List possible interpretations for ambiguous calls * Additional details on incorrect parameters .. index:: -gnatjnn (gcc) :switch:`-gnatjnn` In normal operation mode (or if :switch:`-gnatj0` is used), then error messages with continuation lines are treated as though the continuation lines were separate messages (and so a warning with two continuation lines counts as three warnings, and is listed as three separate messages). If the :switch:`-gnatjnn` switch is used with a positive value for nn, then messages are output in a different manner. A message and all its continuation lines are treated as a unit, and count as only one warning or message in the statistics totals. Furthermore, the message is reformatted so that no line is longer than nn characters. .. index:: -gnatq (gcc) :switch:`-gnatq` The ``q`` stands for quit (really 'don't quit'). In normal operation mode, the compiler first parses the program and determines if there are any syntax errors. If there are, appropriate error messages are generated and compilation is immediately terminated. This switch tells GNAT to continue with semantic analysis even if syntax errors have been found. This may enable the detection of more errors in a single run. On the other hand, the semantic analyzer is more likely to encounter some internal fatal error when given a syntactically invalid tree. .. index:: -gnatQ (gcc) :switch:`-gnatQ` In normal operation mode, the :file:`ALI` file is not generated if any illegalities are detected in the program. The use of :switch:`-gnatQ` forces generation of the :file:`ALI` file. This file is marked as being in error, so it cannot be used for binding purposes, but it does contain reasonably complete cross-reference information, and thus may be useful for use by tools (e.g., semantic browsing tools or integrated development environments) that are driven from the :file:`ALI` file. This switch implies :switch:`-gnatq`, since the semantic phase must be run to get a meaningful ALI file. When :switch:`-gnatQ` is used and the generated :file:`ALI` file is marked as being in error, ``gnatmake`` will attempt to recompile the source when it finds such an :file:`ALI` file, including with switch :switch:`-gnatc`. Note that :switch:`-gnatQ` has no effect if :switch:`-gnats` is specified, since ALI files are never generated if :switch:`-gnats` is set. .. _Warning_Message_Control: Warning Message Control ----------------------- .. index:: Warning messages In addition to error messages, which correspond to illegalities as defined in the Ada Reference Manual, the compiler detects two kinds of warning situations. First, the compiler considers some constructs suspicious and generates a warning message to alert you to a possible error. Second, if the compiler detects a situation that is sure to raise an exception at run time, it generates a warning message. The following shows an example of warning messages: :: e.adb:4:24: warning: creation of object may raise Storage_Error e.adb:10:17: warning: static value out of range e.adb:10:17: warning: "Constraint_Error" will be raised at run time GNAT considers a large number of situations as appropriate for the generation of warning messages. As always, warnings are not definite indications of errors. For example, if you do an out-of-range assignment with the deliberate intention of raising a ``Constraint_Error`` exception, then the warning that may be issued does not indicate an error. Some of the situations for which GNAT issues warnings (at least some of the time) are given in the following list. This list is not complete, and new warnings are often added to subsequent versions of GNAT. The list is intended to give a general idea of the kinds of warnings that are generated. * Possible infinitely recursive calls * Out-of-range values being assigned * Possible order of elaboration problems * Size not a multiple of alignment for a record type * Assertions (pragma Assert) that are sure to fail * Unreachable code * Address clauses with possibly unaligned values, or where an attempt is made to overlay a smaller variable with a larger one. * Fixed-point type declarations with a null range * Direct_IO or Sequential_IO instantiated with a type that has access values * Variables that are never assigned a value * Variables that are referenced before being initialized * Task entries with no corresponding ``accept`` statement * Duplicate accepts for the same task entry in a ``select`` * Objects that take too much storage * Unchecked conversion between types of differing sizes * Missing ``return`` statement along some execution path in a function * Incorrect (unrecognized) pragmas * Incorrect external names * Allocation from empty storage pool * Potentially blocking operation in protected type * Suspicious parenthesization of expressions * Mismatching bounds in an aggregate * Attempt to return local value by reference * Premature instantiation of a generic body * Attempt to pack aliased components * Out of bounds array subscripts * Wrong length on string assignment * Violations of style rules if style checking is enabled * Unused |with| clauses * ``Bit_Order`` usage that does not have any effect * ``Standard.Duration`` used to resolve universal fixed expression * Dereference of possibly null value * Declaration that is likely to cause storage error * Internal GNAT unit |withed| by application unit * Values known to be out of range at compile time * Unreferenced or unmodified variables. Note that a special exemption applies to variables which contain any of the substrings ``DISCARD, DUMMY, IGNORE, JUNK, UNUSE, TMP, TEMP`` in any casing. Such variables are considered likely to be intentionally used in a situation where otherwise a warning would be given, so warnings of this kind are always suppressed for such variables. * Address overlays that could clobber memory * Unexpected initialization when address clause present * Bad alignment for address clause * Useless type conversions * Redundant assignment statements and other redundant constructs * Useless exception handlers * Accidental hiding of name by child unit * Access before elaboration detected at compile time * A range in a ``for`` loop that is known to be null or might be null The following section lists compiler switches that are available to control the handling of warning messages. It is also possible to exercise much finer control over what warnings are issued and suppressed using the GNAT pragma Warnings (see the description of the pragma in the :title:`GNAT_Reference_manual`). .. index:: -gnatwa (gcc) :switch:`-gnatwa` *Activate most optional warnings.* This switch activates most optional warning messages. See the remaining list in this section for details on optional warning messages that can be individually controlled. The warnings that are not turned on by this switch are: * :switch:`-gnatwd` (implicit dereferencing) * :switch:`-gnatw.d` (tag warnings with -gnatw switch) * :switch:`-gnatwh` (hiding) * :switch:`-gnatw.h` (holes in record layouts) * :switch:`-gnatw.j` (late primitives of tagged types) * :switch:`-gnatw.k` (redefinition of names in standard) * :switch:`-gnatwl` (elaboration warnings) * :switch:`-gnatw.l` (inherited aspects) * :switch:`-gnatw.n` (atomic synchronization) * :switch:`-gnatwo` (address clause overlay) * :switch:`-gnatw.o` (values set by out parameters ignored) * :switch:`-gnatw.q` (questionable layout of record types) * :switch:`-gnatw_q` (ignored equality) * :switch:`-gnatw_r` (out-of-order record representation clauses) * :switch:`-gnatw.s` (overridden size clause) * :switch:`-gnatw_s` (ineffective predicate test) * :switch:`-gnatwt` (tracking of deleted conditional code) * :switch:`-gnatw.u` (unordered enumeration) * :switch:`-gnatw.w` (use of Warnings Off) * :switch:`-gnatw.y` (reasons for package needing body) All other optional warnings are turned on. .. index:: -gnatwA (gcc) :switch:`-gnatwA` *Suppress all optional errors.* This switch suppresses all optional warning messages, see remaining list in this section for details on optional warning messages that can be individually controlled. Note that unlike switch :switch:`-gnatws`, the use of switch :switch:`-gnatwA` does not suppress warnings that are normally given unconditionally and cannot be individually controlled (for example, the warning about a missing exit path in a function). Also, again unlike switch :switch:`-gnatws`, warnings suppressed by the use of switch :switch:`-gnatwA` can be individually turned back on. For example the use of switch :switch:`-gnatwA` followed by switch :switch:`-gnatwd` will suppress all optional warnings except the warnings for implicit dereferencing. .. index:: -gnatw.a (gcc) :switch:`-gnatw.a` *Activate warnings on failing assertions.* .. index:: Assert failures This switch activates warnings for assertions where the compiler can tell at compile time that the assertion will fail. Note that this warning is given even if assertions are disabled. The default is that such warnings are generated. .. index:: -gnatw.A (gcc) :switch:`-gnatw.A` *Suppress warnings on failing assertions.* .. index:: Assert failures This switch suppresses warnings for assertions where the compiler can tell at compile time that the assertion will fail. .. index:: -gnatw_a :switch:`-gnatw_a` *Activate warnings on anonymous allocators.* .. index:: Anonymous allocators This switch activates warnings for allocators of anonymous access types, which can involve run-time accessibility checks and lead to unexpected accessibility violations. For more details on the rules involved, see RM 3.10.2 (14). .. index:: -gnatw_A :switch:`-gnatw_A` *Suppress warnings on anonymous allocators.* .. index:: Anonymous allocators This switch suppresses warnings for anonymous access type allocators. .. index:: -gnatwb (gcc) :switch:`-gnatwb` *Activate warnings on bad fixed values.* .. index:: Bad fixed values .. index:: Fixed-point Small value .. index:: Small value This switch activates warnings for static fixed-point expressions whose value is not an exact multiple of Small. Such values are implementation dependent, since an implementation is free to choose either of the multiples that surround the value. GNAT always chooses the closer one, but this is not required behavior, and it is better to specify a value that is an exact multiple, ensuring predictable execution. The default is that such warnings are not generated. .. index:: -gnatwB (gcc) :switch:`-gnatwB` *Suppress warnings on bad fixed values.* This switch suppresses warnings for static fixed-point expressions whose value is not an exact multiple of Small. .. index:: -gnatw.b (gcc) :switch:`-gnatw.b` *Activate warnings on biased representation.* .. index:: Biased representation This switch activates warnings when a size clause, value size clause, component clause, or component size clause forces the use of biased representation for an integer type (e.g. representing a range of 10..11 in a single bit by using 0/1 to represent 10/11). The default is that such warnings are generated. .. index:: -gnatwB (gcc) :switch:`-gnatw.B` *Suppress warnings on biased representation.* This switch suppresses warnings for representation clauses that force the use of biased representation. .. index:: -gnatwc (gcc) :switch:`-gnatwc` *Activate warnings on conditionals.* .. index:: Conditionals, constant This switch activates warnings for boolean expressions that are known to be True or False at compile time. The default is that such warnings are not generated. Note that this warning does not get issued for the use of boolean constants whose values are known at compile time, since this is a standard technique for conditional compilation in Ada, and this would generate too many false positive warnings. This warning option also activates a special test for comparisons using the operators '>=' and' <='. If the compiler can tell that only the equality condition is possible, then it will warn that the '>' or '<' part of the test is useless and that the operator could be replaced by '='. An example would be comparing a ``Natural`` variable <= 0. This warning option also generates warnings if one or both tests is optimized away in a membership test for integer values if the result can be determined at compile time. Range tests on enumeration types are not included, since it is common for such tests to include an end point. This warning can also be turned on using :switch:`-gnatwa`. .. index:: -gnatwC (gcc) :switch:`-gnatwC` *Suppress warnings on conditionals.* This switch suppresses warnings for conditional expressions used in tests that are known to be True or False at compile time. .. index:: -gnatw.c (gcc) :switch:`-gnatw.c` *Activate warnings on missing component clauses.* .. index:: Component clause, missing This switch activates warnings for record components where a record representation clause is present and has component clauses for the majority, but not all, of the components. A warning is given for each component for which no component clause is present. .. index:: -gnatw.C (gcc) :switch:`-gnatw.C` *Suppress warnings on missing component clauses.* This switch suppresses warnings for record components that are missing a component clause in the situation described above. .. index:: -gnatw_c (gcc) :switch:`-gnatw_c` *Activate warnings on unknown condition in Compile_Time_Warning.* .. index:: Compile_Time_Warning .. index:: Compile_Time_Error This switch activates warnings on a pragma Compile_Time_Warning or Compile_Time_Error whose condition has a value that is not known at compile time. The default is that such warnings are generated. .. index:: -gnatw_C (gcc) :switch:`-gnatw_C` *Suppress warnings on unknown condition in Compile_Time_Warning.* This switch suppresses warnings on a pragma Compile_Time_Warning or Compile_Time_Error whose condition has a value that is not known at compile time. .. index:: -gnatwd (gcc) :switch:`-gnatwd` *Activate warnings on implicit dereferencing.* If this switch is set, then the use of a prefix of an access type in an indexed component, slice, or selected component without an explicit ``.all`` will generate a warning. With this warning enabled, access checks occur only at points where an explicit ``.all`` appears in the source code (assuming no warnings are generated as a result of this switch). The default is that such warnings are not generated. .. index:: -gnatwD (gcc) :switch:`-gnatwD` *Suppress warnings on implicit dereferencing.* .. index:: Implicit dereferencing .. index:: Dereferencing, implicit This switch suppresses warnings for implicit dereferences in indexed components, slices, and selected components. .. index:: -gnatw.d (gcc) :switch:`-gnatw.d` *Activate tagging of warning and info messages.* If this switch is set, then warning messages are tagged, with one of the following strings: - *[-gnatw?]* Used to tag warnings controlled by the switch :switch:`-gnatwx` where x is a letter a-z. - *[-gnatw.?]* Used to tag warnings controlled by the switch :switch:`-gnatw.x` where x is a letter a-z. - *[-gnatel]* Used to tag elaboration information (info) messages generated when the static model of elaboration is used and the :switch:`-gnatel` switch is set. - *[restriction warning]* Used to tag warning messages for restriction violations, activated by use of the pragma ``Restriction_Warnings``. - *[warning-as-error]* Used to tag warning messages that have been converted to error messages by use of the pragma Warning_As_Error. Note that such warnings are prefixed by the string "error: " rather than "warning: ". - *[enabled by default]* Used to tag all other warnings that are always given by default, unless warnings are completely suppressed using pragma *Warnings(Off)* or the switch :switch:`-gnatws`. .. index:: -gnatw.d (gcc) :switch:`-gnatw.D` *Deactivate tagging of warning and info messages messages.* If this switch is set, then warning messages return to the default mode in which warnings and info messages are not tagged as described above for :switch:`-gnatw.d`. .. index:: -gnatwe (gcc) .. index:: Warnings, treat as error :switch:`-gnatwe` *Treat warnings and style checks as errors.* This switch causes warning messages and style check messages to be treated as errors. The warning string still appears, but the warning messages are counted as errors, and prevent the generation of an object file. Note that this is the only -gnatw switch that affects the handling of style check messages. Note also that this switch has no effect on info (information) messages, which are not treated as errors if this switch is present. .. index:: -gnatw.e (gcc) :switch:`-gnatw.e` *Activate every optional warning.* .. index:: Warnings, activate every optional warning This switch activates all optional warnings, including those which are not activated by :switch:`-gnatwa`. The use of this switch is not recommended for normal use. If you turn this switch on, it is almost certain that you will get large numbers of useless warnings. The warnings that are excluded from :switch:`-gnatwa` are typically highly specialized warnings that are suitable for use only in code that has been specifically designed according to specialized coding rules. .. index:: -gnatwE (gcc) .. index:: Warnings, treat as error :switch:`-gnatwE` *Treat all run-time exception warnings as errors.* This switch causes warning messages regarding errors that will be raised during run-time execution to be treated as errors. .. index:: -gnatwf (gcc) :switch:`-gnatwf` *Activate warnings on unreferenced formals.* .. index:: Formals, unreferenced This switch causes a warning to be generated if a formal parameter is not referenced in the body of the subprogram. This warning can also be turned on using :switch:`-gnatwu`. The default is that these warnings are not generated. .. index:: -gnatwF (gcc) :switch:`-gnatwF` *Suppress warnings on unreferenced formals.* This switch suppresses warnings for unreferenced formal parameters. Note that the combination :switch:`-gnatwu` followed by :switch:`-gnatwF` has the effect of warning on unreferenced entities other than subprogram formals. .. index:: -gnatwg (gcc) :switch:`-gnatwg` *Activate warnings on unrecognized pragmas.* .. index:: Pragmas, unrecognized This switch causes a warning to be generated if an unrecognized pragma is encountered. Apart from issuing this warning, the pragma is ignored and has no effect. The default is that such warnings are issued (satisfying the Ada Reference Manual requirement that such warnings appear). .. index:: -gnatwG (gcc) :switch:`-gnatwG` *Suppress warnings on unrecognized pragmas.* This switch suppresses warnings for unrecognized pragmas. .. index:: -gnatw.g (gcc) :switch:`-gnatw.g` *Warnings used for GNAT sources.* This switch sets the warning categories that are used by the standard GNAT style. Currently this is equivalent to :switch:`-gnatwAao.q.s.CI.V.X.Z` but more warnings may be added in the future without advanced notice. .. index:: -gnatwh (gcc) :switch:`-gnatwh` *Activate warnings on hiding.* .. index:: Hiding of Declarations This switch activates warnings on hiding declarations that are considered potentially confusing. Not all cases of hiding cause warnings; for example an overriding declaration hides an implicit declaration, which is just normal code. The default is that warnings on hiding are not generated. .. index:: -gnatwH (gcc) :switch:`-gnatwH` *Suppress warnings on hiding.* This switch suppresses warnings on hiding declarations. .. index:: -gnatw.h (gcc) :switch:`-gnatw.h` *Activate warnings on holes/gaps in records.* .. index:: Record Representation (gaps) This switch activates warnings on component clauses in record representation clauses that leave holes (gaps) in the record layout. If a record representation clause does not specify a location for every component of the record type, then the warnings generated (or not generated) are unspecified. For example, there may be gaps for which either no warning is generated or a warning is generated that incorrectly describes the location of the gap. This undesirable situation can sometimes be avoided by adding (and specifying the location for) unused fill fields. .. index:: -gnatw.H (gcc) :switch:`-gnatw.H` *Suppress warnings on holes/gaps in records.* This switch suppresses warnings on component clauses in record representation clauses that leave holes (haps) in the record layout. .. index:: -gnatwi (gcc) :switch:`-gnatwi` *Activate warnings on implementation units.* This switch activates warnings for a |with| of an internal GNAT implementation unit, defined as any unit from the ``Ada``, ``Interfaces``, ``GNAT``, or ``System`` hierarchies that is not documented in either the Ada Reference Manual or the GNAT Programmer's Reference Manual. Such units are intended only for internal implementation purposes and should not be |withed| by user programs. The default is that such warnings are generated .. index:: -gnatwI (gcc) :switch:`-gnatwI` *Disable warnings on implementation units.* This switch disables warnings for a |with| of an internal GNAT implementation unit. .. index:: -gnatw.i (gcc) :switch:`-gnatw.i` *Activate warnings on overlapping actuals.* This switch enables a warning on statically detectable overlapping actuals in a subprogram call, when one of the actuals is an in-out parameter, and the types of the actuals are not by-copy types. This warning is off by default. .. index:: -gnatw.I (gcc) :switch:`-gnatw.I` *Disable warnings on overlapping actuals.* This switch disables warnings on overlapping actuals in a call. .. index:: -gnatwj (gcc) :switch:`-gnatwj` *Activate warnings on obsolescent features (Annex J).* .. index:: Features, obsolescent .. index:: Obsolescent features If this warning option is activated, then warnings are generated for calls to subprograms marked with ``pragma Obsolescent`` and for use of features in Annex J of the Ada Reference Manual. In the case of Annex J, not all features are flagged. In particular, uses of package ``ASCII`` are not flagged, since these are very common and would generate many annoying positive warnings. The default is that such warnings are not generated. In addition to the above cases, warnings are also generated for GNAT features that have been provided in past versions but which have been superseded (typically by features in the new Ada standard). For example, ``pragma Ravenscar`` will be flagged since its function is replaced by ``pragma Profile(Ravenscar)``, and ``pragma Interface_Name`` will be flagged since its function is replaced by ``pragma Import``. Note that this warning option functions differently from the restriction ``No_Obsolescent_Features`` in two respects. First, the restriction applies only to annex J features. Second, the restriction does flag uses of package ``ASCII``. .. index:: -gnatwJ (gcc) :switch:`-gnatwJ` *Suppress warnings on obsolescent features (Annex J).* This switch disables warnings on use of obsolescent features. .. index:: -gnatw.j (gcc) :switch:`-gnatw.j` *Activate warnings on late declarations of tagged type primitives.* This switch activates warnings on visible primitives added to a tagged type after deriving a private extension from it. .. index:: -gnatw.J (gcc) :switch:`-gnatw.J` *Suppress warnings on late declarations of tagged type primitives.* This switch suppresses warnings on visible primitives added to a tagged type after deriving a private extension from it. .. index:: -gnatwk (gcc) :switch:`-gnatwk` *Activate warnings on variables that could be constants.* This switch activates warnings for variables that are initialized but never modified, and then could be declared constants. The default is that such warnings are not given. .. index:: -gnatwK (gcc) :switch:`-gnatwK` *Suppress warnings on variables that could be constants.* This switch disables warnings on variables that could be declared constants. .. index:: -gnatw.k (gcc) :switch:`-gnatw.k` *Activate warnings on redefinition of names in standard.* This switch activates warnings for declarations that declare a name that is defined in package Standard. Such declarations can be confusing, especially since the names in package Standard continue to be directly visible, meaning that use visibility on such redeclared names does not work as expected. Names of discriminants and components in records are not included in this check. .. index:: -gnatwK (gcc) :switch:`-gnatw.K` *Suppress warnings on redefinition of names in standard.* This switch disables warnings for declarations that declare a name that is defined in package Standard. .. index:: -gnatwl (gcc) :switch:`-gnatwl` *Activate warnings for elaboration pragmas.* .. index:: Elaboration, warnings This switch activates warnings for possible elaboration problems, including suspicious use of ``Elaborate`` pragmas, when using the static elaboration model, and possible situations that may raise ``Program_Error`` when using the dynamic elaboration model. See the section in this guide on elaboration checking for further details. The default is that such warnings are not generated. .. index:: -gnatwL (gcc) :switch:`-gnatwL` *Suppress warnings for elaboration pragmas.* This switch suppresses warnings for possible elaboration problems. .. index:: -gnatw.l (gcc) :switch:`-gnatw.l` *List inherited aspects as info messages.* This switch causes the compiler to list inherited invariants, preconditions, and postconditions from Type_Invariant'Class, Invariant'Class, Pre'Class, and Post'Class aspects. Also list inherited subtype predicates. .. index:: -gnatw.L (gcc) :switch:`-gnatw.L` *Suppress listing of inherited aspects as info messages.* This switch suppresses listing of inherited aspects. .. index:: -gnatw_l (gcc) :switch:`-gnatw_l` *Activate warnings on implicitly limited types.* This switch causes the compiler trigger warnings on record types that do not have a limited keyword but contain a component that is a limited type. .. index:: -gnatw_L (gcc) :switch:`-gnatw_L` *Suppress warnings on implicitly limited types.* This switch suppresses warnings on implicitly limited types. .. index:: -gnatwm (gcc) :switch:`-gnatwm` *Activate warnings on modified but unreferenced variables.* This switch activates warnings for variables that are assigned (using an initialization value or with one or more assignment statements) but whose value is never read. The warning is suppressed for volatile variables and also for variables that are renamings of other variables or for which an address clause is given. The default is that these warnings are not given. .. index:: -gnatwM (gcc) :switch:`-gnatwM` *Disable warnings on modified but unreferenced variables.* This switch disables warnings for variables that are assigned or initialized, but never read. .. index:: -gnatw.m (gcc) :switch:`-gnatw.m` *Activate warnings on suspicious modulus values.* This switch activates warnings for modulus values that seem suspicious. The cases caught are where the size is the same as the modulus (e.g. a modulus of 7 with a size of 7 bits), and modulus values of 32 or 64 with no size clause. The guess in both cases is that 2**x was intended rather than x. In addition expressions of the form 2*x for small x generate a warning (the almost certainly accurate guess being that 2**x was intended). This switch also activates warnings for negative literal values of a modular type, which are interpreted as large positive integers after wrap-around. The default is that these warnings are given. .. index:: -gnatw.M (gcc) :switch:`-gnatw.M` *Disable warnings on suspicious modulus values.* This switch disables warnings for suspicious modulus values. .. index:: -gnatwn (gcc) :switch:`-gnatwn` *Set normal warnings mode.* This switch sets normal warning mode, in which enabled warnings are issued and treated as warnings rather than errors. This is the default mode. the switch :switch:`-gnatwn` can be used to cancel the effect of an explicit :switch:`-gnatws` or :switch:`-gnatwe`. It also cancels the effect of the implicit :switch:`-gnatwe` that is activated by the use of :switch:`-gnatg`. .. index:: -gnatw.n (gcc) .. index:: Atomic Synchronization, warnings :switch:`-gnatw.n` *Activate warnings on atomic synchronization.* This switch activates warnings when an access to an atomic variable requires the generation of atomic synchronization code. These warnings are off by default. .. index:: -gnatw.N (gcc) :switch:`-gnatw.N` *Suppress warnings on atomic synchronization.* .. index:: Atomic Synchronization, warnings This switch suppresses warnings when an access to an atomic variable requires the generation of atomic synchronization code. .. index:: -gnatwo (gcc) .. index:: Address Clauses, warnings :switch:`-gnatwo` *Activate warnings on address clause overlays.* This switch activates warnings for possibly unintended initialization effects of defining address clauses that cause one variable to overlap another. The default is that such warnings are generated. .. index:: -gnatwO (gcc) :switch:`-gnatwO` *Suppress warnings on address clause overlays.* This switch suppresses warnings on possibly unintended initialization effects of defining address clauses that cause one variable to overlap another. .. index:: -gnatw.o (gcc) :switch:`-gnatw.o` *Activate warnings on modified but unreferenced out parameters.* This switch activates warnings for variables that are modified by using them as actuals for a call to a procedure with an out mode formal, where the resulting assigned value is never read. It is applicable in the case where there is more than one out mode formal. If there is only one out mode formal, the warning is issued by default (controlled by -gnatwu). The warning is suppressed for volatile variables and also for variables that are renamings of other variables or for which an address clause is given. The default is that these warnings are not given. .. index:: -gnatw.O (gcc) :switch:`-gnatw.O` *Disable warnings on modified but unreferenced out parameters.* This switch suppresses warnings for variables that are modified by using them as actuals for a call to a procedure with an out mode formal, where the resulting assigned value is never read. .. index:: -gnatwp (gcc) .. index:: Inlining, warnings :switch:`-gnatwp` *Activate warnings on ineffective pragma Inlines.* This switch activates warnings for failure of cross-unit inlining (activated by :switch:`-gnatn`) to inline calls to a subprogram. There are many reasons for not being able to inline these calls, including most commonly that the subprogram body is too complex to inline. The default is that such warnings are not given. Warnings on ineffective inlining (within units) by the back end can be activated separately, using the -Winline switch. .. index:: -gnatwP (gcc) :switch:`-gnatwP` *Suppress warnings on ineffective pragma Inlines.* This switch suppresses warnings on ineffective pragma Inlines. If the inlining mechanism cannot inline a call, it will simply ignore the request silently. .. index:: -gnatw.p (gcc) .. index:: Parameter order, warnings :switch:`-gnatw.p` *Activate warnings on parameter ordering.* This switch activates warnings for cases of suspicious parameter ordering when the list of arguments are all simple identifiers that match the names of the formals, but are in a different order. The warning is suppressed if any use of named parameter notation is used, so this is the appropriate way to suppress a false positive (and serves to emphasize that the "misordering" is deliberate). The default is that such warnings are not given. .. index:: -gnatw.P (gcc) :switch:`-gnatw.P` *Suppress warnings on parameter ordering.* This switch suppresses warnings on cases of suspicious parameter ordering. .. index:: -gnatw_p (gcc) :switch:`-gnatw_p` *Activate warnings for pedantic checks.* This switch activates warnings for the failure of certain pedantic checks. The only case currently supported is a check that the subtype_marks given for corresponding formal parameter and function results in a subprogram declaration and its body denote the same subtype declaration. The default is that such warnings are not given. .. index:: -gnatw_P (gcc) :switch:`-gnatw_P` *Suppress warnings for pedantic checks.* This switch suppresses warnings on violations of pedantic checks. .. index:: -gnatwq (gcc) .. index:: Parentheses, warnings :switch:`-gnatwq` *Activate warnings on questionable missing parentheses.* This switch activates warnings for cases where parentheses are not used and the result is potential ambiguity from a readers point of view. For example (not a > b) when a and b are modular means ((not a) > b) and very likely the programmer intended (not (a > b)). Similarly (-x mod 5) means (-(x mod 5)) and quite likely ((-x) mod 5) was intended. In such situations it seems best to follow the rule of always parenthesizing to make the association clear, and this warning switch warns if such parentheses are not present. The default is that these warnings are given. .. index:: -gnatwQ (gcc) :switch:`-gnatwQ` *Suppress warnings on questionable missing parentheses.* This switch suppresses warnings for cases where the association is not clear and the use of parentheses is preferred. .. index:: -gnatw.q (gcc) .. index:: Layout, warnings :switch:`-gnatw.q` *Activate warnings on questionable layout of record types.* This switch activates warnings for cases where the default layout of a record type, that is to say the layout of its components in textual order of the source code, would very likely cause inefficiencies in the code generated by the compiler, both in terms of space and speed during execution. One warning is issued for each problematic component without representation clause in the nonvariant part and then in each variant recursively, if any. The purpose of these warnings is neither to prescribe an optimal layout nor to force the use of representation clauses, but rather to get rid of the most blatant inefficiencies in the layout. Therefore, the default layout is matched against the following synthetic ordered layout and the deviations are flagged on a component-by-component basis: * first all components or groups of components whose length is fixed and a multiple of the storage unit, * then the remaining components whose length is fixed and not a multiple of the storage unit, * then the remaining components whose length doesn't depend on discriminants (that is to say, with variable but uniform length for all objects), * then all components whose length depends on discriminants, * finally the variant part (if any), for the nonvariant part and for each variant recursively, if any. The exact wording of the warning depends on whether the compiler is allowed to reorder the components in the record type or precluded from doing it by means of pragma ``No_Component_Reordering``. The default is that these warnings are not given. .. index:: -gnatw.Q (gcc) :switch:`-gnatw.Q` *Suppress warnings on questionable layout of record types.* This switch suppresses warnings for cases where the default layout of a record type would very likely cause inefficiencies. .. index:: -gnatw_q (gcc) :switch:`-gnatw_q` *Activate warnings for ignored equality operators.* This switch activates warnings for a user-defined "=" function that does not compose (i.e. is ignored for a predefined "=" for a composite type containing a component whose type has the user-defined "=" as primitive). Note that the user-defined "=" must be a primitive operator in order to trigger the warning. See RM-4.5.2(14/3-15/5, 21, 24/3, 32.1/1) for the exact Ada rules on composability of "=". The default is that these warnings are not given. .. index:: -gnatw_Q (gcc) :switch:`-gnatw_Q` *Suppress warnings for ignored equality operators.* .. index:: -gnatwr (gcc) :switch:`-gnatwr` *Activate warnings on redundant constructs.* This switch activates warnings for redundant constructs. The following is the current list of constructs regarded as redundant: * Assignment of an item to itself. * Type conversion that converts an expression to its own subtype. * Use of the attribute ``Base`` where ``typ'Base`` is the same as ``typ``. * Use of pragma ``Pack`` when all components are placed by a record representation clause. * Exception handler containing only a reraise statement (raise with no operand) which has no effect. * Use of the operator abs on an operand that is known at compile time to be non-negative * Comparison of an object or (unary or binary) operation of boolean type to an explicit True value. * Import of parent package. The default is that warnings for redundant constructs are not given. .. index:: -gnatwR (gcc) :switch:`-gnatwR` *Suppress warnings on redundant constructs.* This switch suppresses warnings for redundant constructs. .. index:: -gnatw.r (gcc) :switch:`-gnatw.r` *Activate warnings for object renaming function.* This switch activates warnings for an object renaming that renames a function call, which is equivalent to a constant declaration (as opposed to renaming the function itself). The default is that these warnings are given. .. index:: -gnatw.R (gcc) :switch:`-gnatw.R` *Suppress warnings for object renaming function.* This switch suppresses warnings for object renaming function. .. index:: -gnatw_r (gcc) :switch:`-gnatw_r` *Activate warnings for out-of-order record representation clauses.* This switch activates warnings for record representation clauses, if the order of component declarations, component clauses, and bit-level layout do not all agree. The default is that these warnings are not given. .. index:: -gnatw_R (gcc) :switch:`-gnatw_R` *Suppress warnings for out-of-order record representation clauses.* .. index:: -gnatws (gcc) :switch:`-gnatws` *Suppress all warnings.* This switch completely suppresses the output of all warning messages from the GNAT front end, including both warnings that can be controlled by switches described in this section, and those that are normally given unconditionally. The effect of this suppress action can only be cancelled by a subsequent use of the switch :switch:`-gnatwn`. Note that switch :switch:`-gnatws` does not suppress warnings from the ``gcc`` back end. To suppress these back end warnings as well, use the switch :switch:`-w` in addition to :switch:`-gnatws`. Also this switch has no effect on the handling of style check messages. .. index:: -gnatw.s (gcc) .. index:: Record Representation (component sizes) :switch:`-gnatw.s` *Activate warnings on overridden size clauses.* This switch activates warnings on component clauses in record representation clauses where the length given overrides that specified by an explicit size clause for the component type. A warning is similarly given in the array case if a specified component size overrides an explicit size clause for the array component type. .. index:: -gnatw.S (gcc) :switch:`-gnatw.S` *Suppress warnings on overridden size clauses.* This switch suppresses warnings on component clauses in record representation clauses that override size clauses, and similar warnings when an array component size overrides a size clause. .. index:: -gnatw_s (gcc) .. index:: Warnings :switch:`-gnatw_s` *Activate warnings on ineffective predicate tests.* This switch activates warnings on Static_Predicate aspect specifications that test for values that do not belong to the parent subtype. Not all such ineffective tests are detected. .. index:: -gnatw_S (gcc) :switch:`-gnatw_S` *Suppress warnings on ineffective predicate tests.* This switch suppresses warnings on Static_Predicate aspect specifications that test for values that do not belong to the parent subtype. .. index:: -gnatwt (gcc) .. index:: Deactivated code, warnings .. index:: Deleted code, warnings :switch:`-gnatwt` *Activate warnings for tracking of deleted conditional code.* This switch activates warnings for tracking of code in conditionals (IF and CASE statements) that is detected to be dead code which cannot be executed, and which is removed by the front end. This warning is off by default. This may be useful for detecting deactivated code in certified applications. .. index:: -gnatwT (gcc) :switch:`-gnatwT` *Suppress warnings for tracking of deleted conditional code.* This switch suppresses warnings for tracking of deleted conditional code. .. index:: -gnatw.t (gcc) :switch:`-gnatw.t` *Activate warnings on suspicious contracts.* This switch activates warnings on suspicious contracts. This includes warnings on suspicious postconditions (whether a pragma ``Postcondition`` or a ``Post`` aspect in Ada 2012) and suspicious contract cases (pragma or aspect ``Contract_Cases``). A function postcondition or contract case is suspicious when no postcondition or contract case for this function mentions the result of the function. A procedure postcondition or contract case is suspicious when it only refers to the pre-state of the procedure, because in that case it should rather be expressed as a precondition. This switch also controls warnings on suspicious cases of expressions typically found in contracts like quantified expressions and uses of Update attribute. The default is that such warnings are generated. .. index:: -gnatw.T (gcc) :switch:`-gnatw.T` *Suppress warnings on suspicious contracts.* This switch suppresses warnings on suspicious contracts. .. index:: -gnatwu (gcc) :switch:`-gnatwu` *Activate warnings on unused entities.* This switch activates warnings to be generated for entities that are declared but not referenced, and for units that are |withed| and not referenced. In the case of packages, a warning is also generated if no entities in the package are referenced. This means that if a with'ed package is referenced but the only references are in ``use`` clauses or ``renames`` declarations, a warning is still generated. A warning is also generated for a generic package that is |withed| but never instantiated. In the case where a package or subprogram body is compiled, and there is a |with| on the corresponding spec that is only referenced in the body, a warning is also generated, noting that the |with| can be moved to the body. The default is that such warnings are not generated. This switch also activates warnings on unreferenced formals (it includes the effect of :switch:`-gnatwf`). .. index:: -gnatwU (gcc) :switch:`-gnatwU` *Suppress warnings on unused entities.* This switch suppresses warnings for unused entities and packages. It also turns off warnings on unreferenced formals (and thus includes the effect of :switch:`-gnatwF`). .. index:: -gnatw.u (gcc) :switch:`-gnatw.u` *Activate warnings on unordered enumeration types.* This switch causes enumeration types to be considered as conceptually unordered, unless an explicit pragma ``Ordered`` is given for the type. The effect is to generate warnings in clients that use explicit comparisons or subranges, since these constructs both treat objects of the type as ordered. (A *client* is defined as a unit that is other than the unit in which the type is declared, or its body or subunits.) Please refer to the description of pragma ``Ordered`` in the :title:`GNAT Reference Manual` for further details. The default is that such warnings are not generated. .. index:: -gnatw.U (gcc) :switch:`-gnatw.U` *Deactivate warnings on unordered enumeration types.* This switch causes all enumeration types to be considered as ordered, so that no warnings are given for comparisons or subranges for any type. .. index:: -gnatwv (gcc) .. index:: Unassigned variable warnings :switch:`-gnatwv` *Activate warnings on unassigned variables.* This switch activates warnings for access to variables which may not be properly initialized. The default is that such warnings are generated. This switch will also be emitted when initializing an array or record object via the following aggregate: .. code-block:: ada Array_Or_Record : XXX := (others => <>); unless the relevant type fully initializes all components. .. index:: -gnatwV (gcc) :switch:`-gnatwV` *Suppress warnings on unassigned variables.* This switch suppresses warnings for access to variables which may not be properly initialized. .. index:: -gnatw.v (gcc) .. index:: bit order warnings :switch:`-gnatw.v` *Activate warnings for non-default bit order.* This switch activates warning messages about the effects of non-default bit-order on records to which a component clause is applied. The effect of specifying non-default bit ordering is a bit subtle (and changed with Ada 2005), so these messages, which are given by default, are useful in understanding the exact consequences of using this feature. .. index:: -gnatw.V (gcc) :switch:`-gnatw.V` *Suppress warnings for non-default bit order.* This switch suppresses warnings for the effects of specifying non-default bit order on record components with component clauses. .. index:: -gnatww (gcc) .. index:: String indexing warnings :switch:`-gnatww` *Activate warnings on wrong low bound assumption.* This switch activates warnings for indexing an unconstrained string parameter with a literal or S'Length. This is a case where the code is assuming that the low bound is one, which is in general not true (for example when a slice is passed). The default is that such warnings are generated. .. index:: -gnatwW (gcc) :switch:`-gnatwW` *Suppress warnings on wrong low bound assumption.* This switch suppresses warnings for indexing an unconstrained string parameter with a literal or S'Length. Note that this warning can also be suppressed in a particular case by adding an assertion that the lower bound is 1, as shown in the following example: .. code-block:: ada procedure K (S : String) is pragma Assert (S'First = 1); ... .. index:: -gnatw.w (gcc) .. index:: Warnings Off control :switch:`-gnatw.w` *Activate warnings on Warnings Off pragmas.* This switch activates warnings for use of ``pragma Warnings (Off, entity)`` where either the pragma is entirely useless (because it suppresses no warnings), or it could be replaced by ``pragma Unreferenced`` or ``pragma Unmodified``. Also activates warnings for the case of Warnings (Off, String), where either there is no matching Warnings (On, String), or the Warnings (Off) did not suppress any warning. The default is that these warnings are not given. .. index:: -gnatw.W (gcc) :switch:`-gnatw.W` *Suppress warnings on unnecessary Warnings Off pragmas.* This switch suppresses warnings for use of ``pragma Warnings (Off, ...)``. .. index:: -gnatwx (gcc) .. index:: Export/Import pragma warnings :switch:`-gnatwx` *Activate warnings on Export/Import pragmas.* This switch activates warnings on Export/Import pragmas when the compiler detects a possible conflict between the Ada and foreign language calling sequences. For example, the use of default parameters in a convention C procedure is dubious because the C compiler cannot supply the proper default, so a warning is issued. The default is that such warnings are generated. .. index:: -gnatwX (gcc) :switch:`-gnatwX` *Suppress warnings on Export/Import pragmas.* This switch suppresses warnings on Export/Import pragmas. The sense of this is that you are telling the compiler that you know what you are doing in writing the pragma, and it should not complain at you. .. index:: -gnatw.x (gcc) :switch:`-gnatw.x` *Activate warnings for No_Exception_Propagation mode.* This switch activates warnings for exception usage when pragma Restrictions (No_Exception_Propagation) is in effect. Warnings are given for implicit or explicit exception raises which are not covered by a local handler, and for exception handlers which do not cover a local raise. The default is that these warnings are given for units that contain exception handlers. :switch:`-gnatw.X` *Disable warnings for No_Exception_Propagation mode.* This switch disables warnings for exception usage when pragma Restrictions (No_Exception_Propagation) is in effect. .. index:: -gnatwy (gcc) .. index:: Ada compatibility issues warnings :switch:`-gnatwy` *Activate warnings for Ada compatibility issues.* For the most part, newer versions of Ada are upwards compatible with older versions. For example, Ada 2005 programs will almost always work when compiled as Ada 2012. However there are some exceptions (for example the fact that ``some`` is now a reserved word in Ada 2012). This switch activates several warnings to help in identifying and correcting such incompatibilities. The default is that these warnings are generated. Note that at one point Ada 2005 was called Ada 0Y, hence the choice of character. .. index:: -gnatwY (gcc) .. index:: Ada compatibility issues warnings :switch:`-gnatwY` *Disable warnings for Ada compatibility issues.* This switch suppresses the warnings intended to help in identifying incompatibilities between Ada language versions. .. index:: -gnatw.y (gcc) .. index:: Package spec needing body :switch:`-gnatw.y` *Activate information messages for why package spec needs body.* There are a number of cases in which a package spec needs a body. For example, the use of pragma Elaborate_Body, or the declaration of a procedure specification requiring a completion. This switch causes information messages to be output showing why a package specification requires a body. This can be useful in the case of a large package specification which is unexpectedly requiring a body. The default is that such information messages are not output. .. index:: -gnatw.Y (gcc) .. index:: No information messages for why package spec needs body :switch:`-gnatw.Y` *Disable information messages for why package spec needs body.* This switch suppresses the output of information messages showing why a package specification needs a body. .. index:: -gnatwz (gcc) .. index:: Unchecked_Conversion warnings :switch:`-gnatwz` *Activate warnings on unchecked conversions.* This switch activates warnings for unchecked conversions where the types are known at compile time to have different sizes. The default is that such warnings are generated. Warnings are also generated for subprogram pointers with different conventions. .. index:: -gnatwZ (gcc) :switch:`-gnatwZ` *Suppress warnings on unchecked conversions.* This switch suppresses warnings for unchecked conversions where the types are known at compile time to have different sizes or conventions. .. index:: -gnatw.z (gcc) .. index:: Size/Alignment warnings :switch:`-gnatw.z` *Activate warnings for size not a multiple of alignment.* This switch activates warnings for cases of array and record types with specified ``Size`` and ``Alignment`` attributes where the size is not a multiple of the alignment, resulting in an object size that is greater than the specified size. The default is that such warnings are generated. .. index:: -gnatw.Z (gcc) .. index:: Size/Alignment warnings :switch:`-gnatw.Z` *Suppress warnings for size not a multiple of alignment.* This switch suppresses warnings for cases of array and record types with specified ``Size`` and ``Alignment`` attributes where the size is not a multiple of the alignment, resulting in an object size that is greater than the specified size. The warning can also be suppressed by giving an explicit ``Object_Size`` value. .. index:: -Wunused (gcc) :switch:`-Wunused` The warnings controlled by the :switch:`-gnatw` switch are generated by the front end of the compiler. The GCC back end can provide additional warnings and they are controlled by the :switch:`-W` switch. For example, :switch:`-Wunused` activates back end warnings for entities that are declared but not referenced. .. index:: -Wuninitialized (gcc) :switch:`-Wuninitialized` Similarly, :switch:`-Wuninitialized` activates the back end warning for uninitialized variables. This switch must be used in conjunction with an optimization level greater than zero. .. index:: -Wstack-usage (gcc) :switch:`-Wstack-usage={len}` Warn if the stack usage of a subprogram might be larger than ``len`` bytes. See :ref:`Static_Stack_Usage_Analysis` for details. .. index:: -Wall (gcc) :switch:`-Wall` This switch enables most warnings from the GCC back end. The code generator detects a number of warning situations that are missed by the GNAT front end, and this switch can be used to activate them. The use of this switch also sets the default front-end warning mode to :switch:`-gnatwa`, that is, most front-end warnings are activated as well. .. index:: -w (gcc) :switch:`-w` Conversely, this switch suppresses warnings from the GCC back end. The use of this switch also sets the default front-end warning mode to :switch:`-gnatws`, that is, front-end warnings are suppressed as well. .. index:: -Werror (gcc) :switch:`-Werror` This switch causes warnings from the GCC back end to be treated as errors. The warning string still appears, but the warning messages are counted as errors, and prevent the generation of an object file. The use of this switch also sets the default front-end warning mode to :switch:`-gnatwe`, that is, front-end warning messages and style check messages are treated as errors as well. A string of warning parameters can be used in the same parameter. For example:: -gnatwaGe will turn on all optional warnings except for unrecognized pragma warnings, and also specify that warnings should be treated as errors. When no switch :switch:`-gnatw` is used, this is equivalent to: * :switch:`-gnatw.a` * :switch:`-gnatwB` * :switch:`-gnatw.b` * :switch:`-gnatwC` * :switch:`-gnatw.C` * :switch:`-gnatwD` * :switch:`-gnatw.D` * :switch:`-gnatwF` * :switch:`-gnatw.F` * :switch:`-gnatwg` * :switch:`-gnatwH` * :switch:`-gnatw.H` * :switch:`-gnatwi` * :switch:`-gnatwJ` * :switch:`-gnatw.J` * :switch:`-gnatwK` * :switch:`-gnatw.K` * :switch:`-gnatwL` * :switch:`-gnatw.L` * :switch:`-gnatwM` * :switch:`-gnatw.m` * :switch:`-gnatwn` * :switch:`-gnatw.N` * :switch:`-gnatwo` * :switch:`-gnatw.O` * :switch:`-gnatwP` * :switch:`-gnatw.P` * :switch:`-gnatwq` * :switch:`-gnatw.Q` * :switch:`-gnatwR` * :switch:`-gnatw.R` * :switch:`-gnatw.S` * :switch:`-gnatwT` * :switch:`-gnatw.t` * :switch:`-gnatwU` * :switch:`-gnatw.U` * :switch:`-gnatwv` * :switch:`-gnatw.v` * :switch:`-gnatww` * :switch:`-gnatw.W` * :switch:`-gnatwx` * :switch:`-gnatw.X` * :switch:`-gnatwy` * :switch:`-gnatw.Y` * :switch:`-gnatwz` * :switch:`-gnatw.z` .. _Debugging_and_Assertion_Control: Info message Control -------------------- In addition to the warning messages, the compiler can also generate info messages. In order to control the generation of these messages, the following switch is provided: :switch:`-gnatis` *Suppress all info messages.* This switch completely suppresses the output of all info messages from the GNAT front end. Debugging and Assertion Control ------------------------------- .. index:: -gnata (gcc) :switch:`-gnata` .. index:: Assert .. index:: Debug .. index:: Assertions .. index:: Precondition .. index:: Postcondition .. index:: Type invariants .. index:: Subtype predicates The :switch:`-gnata` option is equivalent to the following ``Assertion_Policy`` pragma:: pragma Assertion_Policy (Check); Which is a shorthand for:: pragma Assertion_Policy -- Ada RM assertion pragmas (Assert => Check, Static_Predicate => Check, Dynamic_Predicate => Check, Pre => Check, Pre'Class => Check, Post => Check, Post'Class => Check, Type_Invariant => Check, Type_Invariant'Class => Check, Default_Initial_Condition => Check, -- GNAT specific assertion pragmas Assert_And_Cut => Check, Assume => Check, Contract_Cases => Check, Debug => Check, Ghost => Check, Initial_Condition => Check, Loop_Invariant => Check, Loop_Variant => Check, Postcondition => Check, Precondition => Check, Predicate => Check, Refined_Post => Check, Subprogram_Variant => Check); The pragmas ``Assert`` and ``Debug`` normally have no effect and are ignored. This switch, where ``a`` stands for 'assert', causes pragmas ``Assert`` and ``Debug`` to be activated. This switch also causes preconditions, postconditions, subtype predicates, and type invariants to be activated. The pragmas have the form:: pragma Assert ( [, ]) pragma Debug () pragma Type_Invariant (, ) pragma Predicate (, ) pragma Precondition (, ) pragma Postcondition (, ) The aspects have the form:: with [Pre|Post|Type_Invariant|Dynamic_Predicate|Static_Predicate] => ; The ``Assert`` pragma causes ``Boolean-expression`` to be tested. If the result is ``True``, the pragma has no effect (other than possible side effects from evaluating the expression). If the result is ``False``, the exception ``Assert_Failure`` declared in the package ``System.Assertions`` is raised (passing ``static-string-expression``, if present, as the message associated with the exception). If no string expression is given, the default is a string containing the file name and line number of the pragma. The ``Debug`` pragma causes ``procedure`` to be called. Note that ``pragma Debug`` may appear within a declaration sequence, allowing debugging procedures to be called between declarations. For the aspect specification, the ``Boolean-expression`` is evaluated. If the result is ``True``, the aspect has no effect. If the result is ``False``, the exception ``Assert_Failure`` is raised. .. _Validity_Checking: Validity Checking ----------------- .. index:: Validity Checking The Ada Reference Manual defines the concept of invalid values (see RM 13.9.1). The primary source of invalid values is uninitialized variables. A scalar variable that is left uninitialized may contain an invalid value; the concept of invalid does not apply to access or composite types. It is an error to read an invalid value, but the RM does not require run-time checks to detect such errors, except for some minimal checking to prevent erroneous execution (i.e. unpredictable behavior). This corresponds to the :switch:`-gnatVd` switch below, which is the default. For example, by default, if the expression of a case statement is invalid, it will raise Constraint_Error rather than causing a wild jump, and if an array index on the left-hand side of an assignment is invalid, it will raise Constraint_Error rather than overwriting an arbitrary memory location. The :switch:`-gnatVa` may be used to enable additional validity checks, which are not required by the RM. These checks are often very expensive (which is why the RM does not require them). These checks are useful in tracking down uninitialized variables, but they are not usually recommended for production builds, and in particular we do not recommend using these extra validity checking options in combination with optimization, since this can confuse the optimizer. If performance is a consideration, leading to the need to optimize, then the validity checking options should not be used. The other :switch:`-gnatV{x}` switches below allow finer-grained control; you can enable whichever validity checks you desire. However, for most debugging purposes, :switch:`-gnatVa` is sufficient, and the default :switch:`-gnatVd` (i.e. standard Ada behavior) is usually sufficient for non-debugging use. The :switch:`-gnatB` switch tells the compiler to assume that all values are valid (that is, within their declared subtype range) except in the context of a use of the Valid attribute. This means the compiler can generate more efficient code, since the range of values is better known at compile time. However, an uninitialized variable can cause wild jumps and memory corruption in this mode. The :switch:`-gnatV{x}` switch allows control over the validity checking mode as described below. The ``x`` argument is a string of letters that indicate validity checks that are performed or not performed in addition to the default checks required by Ada as described above. .. index:: -gnatVa (gcc) :switch:`-gnatVa` *All validity checks.* All validity checks are turned on. That is, :switch:`-gnatVa` is equivalent to ``gnatVcdefimoprst``. .. index:: -gnatVc (gcc) :switch:`-gnatVc` *Validity checks for copies.* The right-hand side of assignments, and the (explicit) initializing values of object declarations are validity checked. .. index:: -gnatVd (gcc) :switch:`-gnatVd` *Default (RM) validity checks.* Some validity checks are required by Ada (see RM 13.9.1 (9-11)); these (and only these) validity checks are enabled by default. For case statements (and case expressions) that lack a "when others =>" choice, a check is made that the value of the selector expression belongs to its nominal subtype. If it does not, Constraint_Error is raised. For assignments to array components (and for indexed components in some other contexts), a check is made that each index expression belongs to the corresponding index subtype. If it does not, Constraint_Error is raised. Both these validity checks may be turned off using switch :switch:`-gnatVD`. They are turned on by default. If :switch:`-gnatVD` is specified, a subsequent switch :switch:`-gnatVd` will leave the checks turned on. Switch :switch:`-gnatVD` should be used only if you are sure that all such expressions have valid values. If you use this switch and invalid values are present, then the program is erroneous, and wild jumps or memory overwriting may occur. .. index:: -gnatVe (gcc) :switch:`-gnatVe` *Validity checks for scalar components.* In the absence of this switch, assignments to scalar components of enclosing record or array objects are not validity checked, even if validity checks for assignments generally (:switch:`-gnatVc`) are turned on. Specifying this switch enables such checks. This switch has no effect if the :switch:`-gnatVc` switch is not specified. .. index:: -gnatVf (gcc) :switch:`-gnatVf` *Validity checks for floating-point values.* Specifying this switch enables validity checking for floating-point values in the same contexts where validity checking is enabled for other scalar values. In the absence of this switch, validity checking is not performed for floating-point values. This takes precedence over other statements about performing validity checking for scalar objects in various scenarios. One way to look at it is that if this switch is not set, then whenever any of the other rules in this section use the word "scalar" they really mean "scalar and not floating-point". If :switch:`-gnatVf` is specified, then validity checking also applies for floating-point values, and NaNs and infinities are considered invalid, as well as out-of-range values for constrained types. The exact contexts in which floating-point values are checked depends on the setting of other options. For example, :switch:`-gnatVif` or :switch:`-gnatVfi` (the order does not matter) specifies that floating-point parameters of mode ``in`` should be validity checked. .. index:: -gnatVi (gcc) :switch:`-gnatVi` *Validity checks for ``in`` mode parameters.* Arguments for parameters of mode ``in`` are validity checked in function and procedure calls at the point of call. .. index:: -gnatVm (gcc) :switch:`-gnatVm` *Validity checks for ``in out`` mode parameters.* Arguments for parameters of mode ``in out`` are validity checked in procedure calls at the point of call. The ``'m'`` here stands for modify, since this concerns parameters that can be modified by the call. Note that there is no specific option to test ``out`` parameters, but any reference within the subprogram will be tested in the usual manner, and if an invalid value is copied back, any reference to it will be subject to validity checking. .. index:: -gnatVn (gcc) :switch:`-gnatVn` *No validity checks.* This switch turns off all validity checking, including the default checking for case statements and left hand side subscripts. Note that the use of the switch :switch:`-gnatp` suppresses all run-time checks, including validity checks, and thus implies :switch:`-gnatVn`. When this switch is used, it cancels any other :switch:`-gnatV` previously issued. .. index:: -gnatVo (gcc) :switch:`-gnatVo` *Validity checks for operator and attribute operands.* Scalar arguments for predefined operators and for attributes are validity checked. This includes all operators in package ``Standard``, the shift operators defined as intrinsic in package ``Interfaces`` and operands for attributes such as ``Pos``. Checks are also made on individual component values for composite comparisons, and on the expressions in type conversions and qualified expressions. Checks are also made on explicit ranges using :samp:`..` (e.g., slices, loops etc). .. index:: -gnatVp (gcc) :switch:`-gnatVp` *Validity checks for parameters.* This controls the treatment of formal parameters within a subprogram (as opposed to :switch:`-gnatVi` and :switch:`-gnatVm`, which control validity testing of actual parameters of a call). If either of these call options is specified, then normally an assumption is made within a subprogram that the validity of any incoming formal parameters of the corresponding mode(s) has already been checked at the point of call and does not need rechecking. If :switch:`-gnatVp` is set, then this assumption is not made and so their validity may be checked (or rechecked) within the subprogram. If neither of the two call-related options is specified, then this switch has no effect. .. index:: -gnatVr (gcc) :switch:`-gnatVr` *Validity checks for function returns.* The expression in simple ``return`` statements in functions is validity checked. .. index:: -gnatVs (gcc) :switch:`-gnatVs` *Validity checks for subscripts.* All subscript expressions are checked for validity, whatever context they occur in (in default mode some subscripts are not validity checked; for example, validity checking may be omitted in some cases involving a read of a component of an array). .. index:: -gnatVt (gcc) :switch:`-gnatVt` *Validity checks for tests.* Expressions used as conditions in ``if``, ``while`` or ``exit`` statements are checked, as well as guard expressions in entry calls. The :switch:`-gnatV` switch may be followed by a string of letters to turn on a series of validity checking options. For example, :switch:`-gnatVcr` specifies that in addition to the default validity checking, copies and function return expressions are to be validity checked. In order to make it easier to specify the desired combination of effects, the upper case letters ``CDFIMORST`` may be used to turn off the corresponding lower case option. Thus :switch:`-gnatVaM` turns on all validity checking options except for checking of ``in out`` parameters. The specification of additional validity checking generates extra code (and in the case of :switch:`-gnatVa` the code expansion can be substantial). However, these additional checks can be very useful in detecting uninitialized variables, incorrect use of unchecked conversion, and other errors leading to invalid values. The use of pragma ``Initialize_Scalars`` is useful in conjunction with the extra validity checking, since this ensures that wherever possible uninitialized variables have invalid values. See also the pragma ``Validity_Checks`` which allows modification of the validity checking mode at the program source level, and also allows for temporary disabling of validity checks. .. _Style_Checking: Style Checking -------------- .. index:: Style checking .. index:: -gnaty (gcc) The :switch:`-gnaty` switch causes the compiler to enforce specified style rules. A limited set of style rules has been used in writing the GNAT sources themselves. This switch allows user programs to activate all or some of these checks. If the source program fails a specified style check, an appropriate message is given, preceded by the character sequence '(style)'. This message does not prevent successful compilation (unless the :switch:`-gnatwe` switch is used). Note that this is by no means intended to be a general facility for checking arbitrary coding standards. It is simply an embedding of the style rules we have chosen for the GNAT sources. If you are starting a project which does not have established style standards, you may find it useful to adopt the entire set of GNAT coding standards, or some subset of them. .. only:: PRO or GPL If you already have an established set of coding standards, then the selected style checking options may indeed correspond to choices you have made, but for general checking of an existing set of coding rules, you should look to the gnatcheck tool, which is designed for that purpose. The string ``x`` is a sequence of letters or digits indicating the particular style checks to be performed. The following checks are defined: .. index:: -gnaty[0-9] (gcc) :switch:`-gnaty0` *Specify indentation level.* If a digit from 1-9 appears in the string after :switch:`-gnaty` then proper indentation is checked, with the digit indicating the indentation level required. A value of zero turns off this style check. The rule checks that the following constructs start on a column that is one plus a multiple of the alignment level: * beginnings of declarations (except record component declarations) and statements; * beginnings of the structural components of compound statements; * ``end`` keyword that completes the declaration of a program unit declaration or body or that completes a compound statement. Full line comments must be aligned with the ``--`` starting on a column that is one plus a multiple of the alignment level, or they may be aligned the same way as the following non-blank line (this is useful when full line comments appear in the middle of a statement), or they may be aligned with the source line on the previous non-blank line. .. index:: -gnatya (gcc) :switch:`-gnatya` *Check attribute casing.* Attribute names, including the case of keywords such as ``digits`` used as attributes names, must be written in mixed case, that is, the initial letter and any letter following an underscore must be uppercase. All other letters must be lowercase. .. index:: -gnatyA (gcc) :switch:`-gnatyA` *Use of array index numbers in array attributes.* When using the array attributes First, Last, Range, or Length, the index number must be omitted for one-dimensional arrays and is required for multi-dimensional arrays. .. index:: -gnatyb (gcc) :switch:`-gnatyb` *Blanks not allowed at statement end.* Trailing blanks are not allowed at the end of statements. The purpose of this rule, together with h (no horizontal tabs), is to enforce a canonical format for the use of blanks to separate source tokens. .. index:: -gnatyB (gcc) :switch:`-gnatyB` *Check Boolean operators.* The use of AND/OR operators is not permitted except in the cases of modular operands, array operands, and simple stand-alone boolean variables or boolean constants. In all other cases ``and then``/`or else` are required. .. index:: -gnatyc (gcc) :switch:`-gnatyc` *Check comments, double space.* Comments must meet the following set of rules: * The ``--`` that starts the column must either start in column one, or else at least one blank must precede this sequence. * Comments that follow other tokens on a line must have at least one blank following the ``--`` at the start of the comment. * Full line comments must have at least two blanks following the ``--`` that starts the comment, with the following exceptions. * A line consisting only of the ``--`` characters, possibly preceded by blanks is permitted. * A comment starting with ``--x`` where ``x`` is a special character is permitted. This allows proper processing of the output from specialized tools such as ``gnatprep`` (where ``--!`` is used) and in earlier versions of the SPARK annotation language (where ``--#`` is used). For the purposes of this rule, a special character is defined as being in one of the ASCII ranges ``16#21#...16#2F#`` or ``16#3A#...16#3F#``. Note that this usage is not permitted in GNAT implementation units (i.e., when :switch:`-gnatg` is used). * A line consisting entirely of minus signs, possibly preceded by blanks, is permitted. This allows the construction of box comments where lines of minus signs are used to form the top and bottom of the box. * A comment that starts and ends with ``--`` is permitted as long as at least one blank follows the initial ``--``. Together with the preceding rule, this allows the construction of box comments, as shown in the following example: .. code-block:: ada --------------------------- -- This is a box comment -- -- with two text lines. -- --------------------------- .. index:: -gnatyC (gcc) :switch:`-gnatyC` *Check comments, single space.* This is identical to ``c`` except that only one space is required following the ``--`` of a comment instead of two. .. index:: -gnatyd (gcc) :switch:`-gnatyd` *Check no DOS line terminators present.* All lines must be terminated by a single ASCII.LF character (in particular the DOS line terminator sequence CR/LF is not allowed). .. index:: -gnatyD (gcc) :switch:`-gnatyD` *Check declared identifiers in mixed case.* Declared identifiers must be in mixed case, as in This_Is_An_Identifier. Use -gnatyr in addition to ensure that references match declarations. .. index:: -gnatye (gcc) :switch:`-gnatye` *Check end/exit labels.* Optional labels on ``end`` statements ending subprograms and on ``exit`` statements exiting named loops, are required to be present. .. index:: -gnatyf (gcc) :switch:`-gnatyf` *No form feeds or vertical tabs.* Neither form feeds nor vertical tab characters are permitted in the source text. .. index:: -gnatyg (gcc) :switch:`-gnatyg` *GNAT style mode.* The set of style check switches is set to match that used by the GNAT sources. This may be useful when developing code that is eventually intended to be incorporated into GNAT. Currently this is equivalent to :switch:`-gnatyydISuxz`) but additional style switches may be added to this set in the future without advance notice. .. index:: -gnatyh (gcc) :switch:`-gnatyh` *No horizontal tabs.* Horizontal tab characters are not permitted in the source text. Together with the b (no blanks at end of line) check, this enforces a canonical form for the use of blanks to separate source tokens. .. index:: -gnatyi (gcc) :switch:`-gnatyi` *Check if-then layout.* The keyword ``then`` must appear either on the same line as corresponding ``if``, or on a line on its own, lined up under the ``if``. .. index:: -gnatyI (gcc) :switch:`-gnatyI` *check mode IN keywords.* Mode ``in`` (the default mode) is not allowed to be given explicitly. ``in out`` is fine, but not ``in`` on its own. .. index:: -gnatyk (gcc) :switch:`-gnatyk` *Check keyword casing.* All keywords must be in lower case (with the exception of keywords such as ``digits`` used as attribute names to which this check does not apply). A single error is reported for each line breaking this rule even if multiple casing issues exist on a same line. .. index:: -gnatyl (gcc) :switch:`-gnatyl` *Check layout.* Layout of statement and declaration constructs must follow the recommendations in the Ada Reference Manual, as indicated by the form of the syntax rules. For example an ``else`` keyword must be lined up with the corresponding ``if`` keyword. There are two respects in which the style rule enforced by this check option are more liberal than those in the Ada Reference Manual. First in the case of record declarations, it is permissible to put the ``record`` keyword on the same line as the ``type`` keyword, and then the ``end`` in ``end record`` must line up under ``type``. This is also permitted when the type declaration is split on two lines. For example, any of the following three layouts is acceptable: .. code-block:: ada type q is record a : integer; b : integer; end record; type q is record a : integer; b : integer; end record; type q is record a : integer; b : integer; end record; Second, in the case of a block statement, a permitted alternative is to put the block label on the same line as the ``declare`` or ``begin`` keyword, and then line the ``end`` keyword up under the block label. For example both the following are permitted: .. code-block:: ada Block : declare A : Integer := 3; begin Proc (A, A); end Block; Block : declare A : Integer := 3; begin Proc (A, A); end Block; The same alternative format is allowed for loops. For example, both of the following are permitted: .. code-block:: ada Clear : while J < 10 loop A (J) := 0; end loop Clear; Clear : while J < 10 loop A (J) := 0; end loop Clear; .. index:: -gnatyLnnn (gcc) :switch:`-gnatyL` *Set maximum nesting level.* The maximum level of nesting of constructs (including subprograms, loops, blocks, packages, and conditionals) may not exceed the given value *nnn*. A value of zero disconnects this style check. .. index:: -gnatym (gcc) :switch:`-gnatym` *Check maximum line length.* The length of source lines must not exceed 79 characters, including any trailing blanks. The value of 79 allows convenient display on an 80 character wide device or window, allowing for possible special treatment of 80 character lines. Note that this count is of characters in the source text. This means that a tab character counts as one character in this count and a wide character sequence counts as a single character (however many bytes are needed in the encoding). .. index:: -gnatyMnnn (gcc) :switch:`-gnatyM` *Set maximum line length.* The length of lines must not exceed the given value *nnn*. The maximum value that can be specified is 32767. If neither style option for setting the line length is used, then the default is 255. This also controls the maximum length of lexical elements, where the only restriction is that they must fit on a single line. .. index:: -gnatyn (gcc) :switch:`-gnatyn` *Check casing of entities in Standard.* Any identifier from Standard must be cased to match the presentation in the Ada Reference Manual (for example, ``Integer`` and ``ASCII.NUL``). .. index:: -gnatyN (gcc) :switch:`-gnatyN` *Turn off all style checks.* All style check options are turned off. .. index:: -gnatyo (gcc) :switch:`-gnatyo` *Check order of subprogram bodies.* All subprogram bodies in a given scope (e.g., a package body) must be in alphabetical order. The ordering rule uses normal Ada rules for comparing strings, ignoring casing of letters, except that if there is a trailing numeric suffix, then the value of this suffix is used in the ordering (e.g., Junk2 comes before Junk10). .. index:: -gnatyO (gcc) :switch:`-gnatyO` *Check that overriding subprograms are explicitly marked as such.* This applies to all subprograms of a derived type that override a primitive operation of the type, for both tagged and untagged types. In particular, the declaration of a primitive operation of a type extension that overrides an inherited operation must carry an overriding indicator. Another case is the declaration of a function that overrides a predefined operator (such as an equality operator). .. index:: -gnatyp (gcc) :switch:`-gnatyp` *Check pragma casing.* Pragma names must be written in mixed case, that is, the initial letter and any letter following an underscore must be uppercase. All other letters must be lowercase. An exception is that SPARK_Mode is allowed as an alternative for Spark_Mode. .. index:: -gnatyr (gcc) :switch:`-gnatyr` *Check references.* All identifier references must be cased in the same way as the corresponding declaration. No specific casing style is imposed on identifiers. The only requirement is for consistency of references with declarations. .. index:: -gnatys (gcc) :switch:`-gnatys` *Check separate specs.* Separate declarations ('specs') are required for subprograms (a body is not allowed to serve as its own declaration). The only exception is that parameterless library level procedures are not required to have a separate declaration. This exception covers the most frequent form of main program procedures. .. index:: -gnatyS (gcc) :switch:`-gnatyS` *Check no statements after then/else.* No statements are allowed on the same line as a ``then`` or ``else`` keyword following the keyword in an ``if`` statement. ``or else`` and ``and then`` are not affected, and a special exception allows a pragma to appear after ``else``. .. index:: -gnatyt (gcc) :switch:`-gnatyt` *Check token spacing.* The following token spacing rules are enforced: * The keywords ``abs`` and ``not`` must be followed by a space. * The token ``=>`` must be surrounded by spaces. * The token ``<>`` must be preceded by a space or a left parenthesis. * Binary operators other than ``**`` must be surrounded by spaces. There is no restriction on the layout of the ``**`` binary operator. * Colon must be surrounded by spaces. * Colon-equal (assignment, initialization) must be surrounded by spaces. * Comma must be the first non-blank character on the line, or be immediately preceded by a non-blank character, and must be followed by a space. * If the token preceding a left parenthesis ends with a letter or digit, then a space must separate the two tokens. * If the token following a right parenthesis starts with a letter or digit, then a space must separate the two tokens. * A right parenthesis must either be the first non-blank character on a line, or it must be preceded by a non-blank character. * A semicolon must not be preceded by a space, and must not be followed by a non-blank character. * A unary plus or minus may not be followed by a space. * A vertical bar must be surrounded by spaces. Exactly one blank (and no other white space) must appear between a ``not`` token and a following ``in`` token. .. index:: -gnatyu (gcc) :switch:`-gnatyu` *Check unnecessary blank lines.* Unnecessary blank lines are not allowed. A blank line is considered unnecessary if it appears at the end of the file, or if more than one blank line occurs in sequence. .. index:: -gnatyx (gcc) :switch:`-gnatyx` *Check extra parentheses.* Unnecessary extra levels of parentheses (C-style) are not allowed around conditions (or selection expressions) in ``if``, ``while``, ``case``, and ``exit`` statements, as well as part of ranges. .. index:: -gnatyy (gcc) :switch:`-gnatyy` *Set all standard style check options.* This is equivalent to ``gnaty3aAbcefhiklmnprst``, that is all checking options enabled with the exception of :switch:`-gnatyB`, :switch:`-gnatyd`, :switch:`-gnatyI`, :switch:`-gnatyLnnn`, :switch:`-gnatyo`, :switch:`-gnatyO`, :switch:`-gnatyS`, :switch:`-gnatyu`, and :switch:`-gnatyx`. .. index:: -gnatyz (gcc) :switch:`-gnatyz` *Check extra parentheses (operator precedence).* Extra levels of parentheses that are not required by operator precedence rules are flagged. See also ``-gnatyx``. .. index:: -gnaty- (gcc) :switch:`-gnaty-` *Remove style check options.* This causes any subsequent options in the string to act as canceling the corresponding style check option. To cancel maximum nesting level control, use the ``L`` parameter without any integer value after that, because any digit following *-* in the parameter string of the :switch:`-gnaty` option will be treated as canceling the indentation check. The same is true for the ``M`` parameter. ``y`` and ``N`` parameters are not allowed after *-*. .. index:: -gnaty+ (gcc) :switch:`-gnaty+` *Enable style check options.* This causes any subsequent options in the string to enable the corresponding style check option. That is, it cancels the effect of a previous -, if any. .. end of switch description (leave this comment to ease automatic parsing for .. GNAT Studio) In the above rules, appearing in column one is always permitted, that is, counts as meeting either a requirement for a required preceding space, or as meeting a requirement for no preceding space. Appearing at the end of a line is also always permitted, that is, counts as meeting either a requirement for a following space, or as meeting a requirement for no following space. If any of these style rules is violated, a message is generated giving details on the violation. The initial characters of such messages are always '`(style)`'. Note that these messages are treated as warning messages, so they normally do not prevent the generation of an object file. The :switch:`-gnatwe` switch can be used to treat warning messages, including style messages, as fatal errors. The switch :switch:`-gnaty` on its own (that is not followed by any letters or digits) is equivalent to the use of :switch:`-gnatyy` as described above, that is all built-in standard style check options are enabled. The switch :switch:`-gnatyN` clears any previously set style checks. .. _Run-Time_Checks: Run-Time Checks --------------- .. index:: Division by zero .. index:: Access before elaboration .. index:: Checks, division by zero .. index:: Checks, access before elaboration .. index:: Checks, stack overflow checking By default, the following checks are suppressed: stack overflow checks, and checks for access before elaboration on subprogram calls. All other checks, including overflow checks, range checks and array bounds checks, are turned on by default. The following ``gcc`` switches refine this default behavior. .. index:: -gnatp (gcc) :switch:`-gnatp` .. index:: Suppressing checks .. index:: Checks, suppressing This switch causes the unit to be compiled as though ``pragma Suppress (All_checks)`` had been present in the source. Validity checks are also eliminated (in other words :switch:`-gnatp` also implies :switch:`-gnatVn`. Use this switch to improve the performance of the code at the expense of safety in the presence of invalid data or program bugs. Note that when checks are suppressed, the compiler is allowed, but not required, to omit the checking code. If the run-time cost of the checking code is zero or near-zero, the compiler will generate it even if checks are suppressed. In particular, if the compiler can prove that a certain check will necessarily fail, it will generate code to do an unconditional 'raise', even if checks are suppressed. The compiler warns in this case. Another case in which checks may not be eliminated is when they are embedded in certain run-time routines such as math library routines. Of course, run-time checks are omitted whenever the compiler can prove that they will not fail, whether or not checks are suppressed. Note that if you suppress a check that would have failed, program execution is erroneous, which means the behavior is totally unpredictable. The program might crash, or print wrong answers, or do anything else. It might even do exactly what you wanted it to do (and then it might start failing mysteriously next week or next year). The compiler will generate code based on the assumption that the condition being checked is true, which can result in erroneous execution if that assumption is wrong. The checks subject to suppression include all the checks defined by the Ada standard, as well as all implementation-defined checks, including any checks introduced using ``pragma Check_Name``. If the code depends on certain checks being active, you can use pragma ``Unsuppress`` either as a configuration pragma or as a local pragma to make sure that a specified check is performed even if ``gnatp`` is specified. The :switch:`-gnatp` switch has no effect if a subsequent :switch:`-gnat-p` switch appears. .. index:: -gnat-p (gcc) .. index:: Suppressing checks .. index:: Checks, suppressing .. index:: Suppress :switch:`-gnat-p` This switch cancels the effect of a previous ``gnatp`` switch. .. index:: -gnato?? (gcc) .. index:: Overflow checks .. index:: Overflow mode .. index:: Check, overflow :switch:`-gnato??` This switch controls the mode used for computing intermediate arithmetic integer operations, and also enables overflow checking. For a full description of overflow mode and checking control, see the :ref:`Overflow_Check_Handling_in_GNAT` appendix in this User's Guide. .. index:: -gnatE (gcc) .. index:: Elaboration checks .. index:: Check, elaboration :switch:`-gnatE` Enables dynamic checks for access-before-elaboration on subprogram calls and generic instantiations. Note that :switch:`-gnatE` is not necessary for safety, because in the default mode, GNAT ensures statically that the checks would not fail. For full details of the effect and use of this switch, :ref:`Compiling_with_gcc`. .. index:: -fstack-check (gcc) .. index:: Stack Overflow Checking .. index:: Checks, stack overflow checking :switch:`-fstack-check` Activates stack overflow checking. For full details of the effect and use of this switch see :ref:`Stack_Overflow_Checking`. .. index:: Unsuppress The setting of these switches only controls the default setting of the checks. You may modify them using either ``Suppress`` (to remove checks) or ``Unsuppress`` (to add back suppressed checks) pragmas in the program source. .. _Using_gcc_for_Syntax_Checking: Using ``gcc`` for Syntax Checking --------------------------------- .. index:: -gnats (gcc) :switch:`-gnats` The ``s`` stands for 'syntax'. Run GNAT in syntax checking only mode. For example, the command :: $ gcc -c -gnats x.adb compiles file :file:`x.adb` in syntax-check-only mode. You can check a series of files in a single command , and can use wildcards to specify such a group of files. Note that you must specify the :switch:`-c` (compile only) flag in addition to the :switch:`-gnats` flag. You may use other switches in conjunction with :switch:`-gnats`. In particular, :switch:`-gnatl` and :switch:`-gnatv` are useful to control the format of any generated error messages. When the source file is empty or contains only empty lines and/or comments, the output is a warning: :: $ gcc -c -gnats -x ada toto.txt toto.txt:1:01: warning: empty file, contains no compilation units $ Otherwise, the output is simply the error messages, if any. No object file or ALI file is generated by a syntax-only compilation. Also, no units other than the one specified are accessed. For example, if a unit ``X`` |withs| a unit ``Y``, compiling unit ``X`` in syntax check only mode does not access the source file containing unit ``Y``. .. index:: Multiple units, syntax checking Normally, GNAT allows only a single unit in a source file. However, this restriction does not apply in syntax-check-only mode, and it is possible to check a file containing multiple compilation units concatenated together. This is primarily used by the ``gnatchop`` utility (:ref:`Renaming_Files_with_gnatchop`). .. _Using_gcc_for_Semantic_Checking: Using ``gcc`` for Semantic Checking ----------------------------------- .. index:: -gnatc (gcc) :switch:`-gnatc` The ``c`` stands for 'check'. Causes the compiler to operate in semantic check mode, with full checking for all illegalities specified in the Ada Reference Manual, but without generation of any object code (no object file is generated). Because dependent files must be accessed, you must follow the GNAT semantic restrictions on file structuring to operate in this mode: * The needed source files must be accessible (see :ref:`Search_Paths_and_the_Run-Time_Library_RTL`). * Each file must contain only one compilation unit. * The file name and unit name must match (:ref:`File_Naming_Rules`). The output consists of error messages as appropriate. No object file is generated. An :file:`ALI` file is generated for use in the context of cross-reference tools, but this file is marked as not being suitable for binding (since no object file is generated). The checking corresponds exactly to the notion of legality in the Ada Reference Manual. Any unit can be compiled in semantics-checking-only mode, including units that would not normally be compiled (subunits, and specifications where a separate body is present). .. _Compiling_Different_Versions_of_Ada: Compiling Different Versions of Ada ----------------------------------- The switches described in this section allow you to explicitly specify the version of the Ada language that your programs are written in. The default mode is Ada 2012, but you can also specify Ada 95, Ada 2005 mode, or indicate Ada 83 compatibility mode. .. index:: Compatibility with Ada 83 .. index:: -gnat83 (gcc) .. index:: ACVC, Ada 83 tests .. index:: Ada 83 mode :switch:`-gnat83` (Ada 83 Compatibility Mode) Although GNAT is primarily an Ada 95 / Ada 2005 compiler, this switch specifies that the program is to be compiled in Ada 83 mode. With :switch:`-gnat83`, GNAT rejects most post-Ada 83 extensions and applies Ada 83 semantics where this can be done easily. It is not possible to guarantee this switch does a perfect job; some subtle tests, such as are found in earlier ACVC tests (and that have been removed from the ACATS suite for Ada 95), might not compile correctly. Nevertheless, this switch may be useful in some circumstances, for example where, due to contractual reasons, existing code needs to be maintained using only Ada 83 features. With few exceptions (most notably the need to use ``<>`` on unconstrained :index:`generic formal parameters `, the use of the new Ada 95 / Ada 2005 reserved words, and the use of packages with optional bodies), it is not necessary to specify the :switch:`-gnat83` switch when compiling Ada 83 programs, because, with rare exceptions, Ada 95 and Ada 2005 are upwardly compatible with Ada 83. Thus a correct Ada 83 program is usually also a correct program in these later versions of the language standard. For further information please refer to the *Compatibility and Porting Guide* chapter in the :title:`GNAT Reference Manual`. .. index:: -gnat95 (gcc) .. index:: Ada 95 mode :switch:`-gnat95` (Ada 95 mode) This switch directs the compiler to implement the Ada 95 version of the language. Since Ada 95 is almost completely upwards compatible with Ada 83, Ada 83 programs may generally be compiled using this switch (see the description of the :switch:`-gnat83` switch for further information about Ada 83 mode). If an Ada 2005 program is compiled in Ada 95 mode, uses of the new Ada 2005 features will cause error messages or warnings. This switch also can be used to cancel the effect of a previous :switch:`-gnat83`, :switch:`-gnat05/2005`, or :switch:`-gnat12/2012` switch earlier in the command line. .. index:: -gnat05 (gcc) .. index:: -gnat2005 (gcc) .. index:: Ada 2005 mode :switch:`-gnat05` or :switch:`-gnat2005` (Ada 2005 mode) This switch directs the compiler to implement the Ada 2005 version of the language, as documented in the official Ada standards document. Since Ada 2005 is almost completely upwards compatible with Ada 95 (and thus also with Ada 83), Ada 83 and Ada 95 programs may generally be compiled using this switch (see the description of the :switch:`-gnat83` and :switch:`-gnat95` switches for further information). .. index:: -gnat12 (gcc) .. index:: -gnat2012 (gcc) .. index:: Ada 2012 mode :switch:`-gnat12` or :switch:`-gnat2012` (Ada 2012 mode) This switch directs the compiler to implement the Ada 2012 version of the language (also the default). Since Ada 2012 is almost completely upwards compatible with Ada 2005 (and thus also with Ada 83, and Ada 95), Ada 83 and Ada 95 programs may generally be compiled using this switch (see the description of the :switch:`-gnat83`, :switch:`-gnat95`, and :switch:`-gnat05/2005` switches for further information). .. index:: -gnat2022 (gcc) .. index:: Ada 2022 mode :switch:`-gnat2022` (Ada 2022 mode) This switch directs the compiler to implement the Ada 2022 version of the language. .. index:: -gnatX0 (gcc) .. index:: Ada language extensions .. index:: GNAT extensions :switch:`-gnatX0` (Enable GNAT Extensions) This switch directs the compiler to implement the latest version of the language (currently Ada 2022) and also to enable certain GNAT implementation extensions that are not part of any Ada standard. For a full list of these extensions, see the GNAT reference manual, ``Pragma Extensions_Allowed``. .. index:: -gnatX (gcc) .. index:: Ada language extensions .. index:: GNAT extensions :switch:`-gnatX` (Enable core GNAT Extensions) This switch is similar to -gnatX0 except that only some, not all, of the GNAT-defined language extensions are enabled. For a list of the extensions enabled by this switch, see the GNAT reference manual ``Pragma Extensions_Allowed`` and the description of that pragma's "On" (as opposed to "All") argument. .. _Character_Set_Control: Character Set Control --------------------- .. index:: -gnati (gcc) :switch:`-gnati{c}` Normally GNAT recognizes the Latin-1 character set in source program identifiers, as described in the Ada Reference Manual. This switch causes GNAT to recognize alternate character sets in identifiers. ``c`` is a single character indicating the character set, as follows: ========== ====================================================== *1* ISO 8859-1 (Latin-1) identifiers *2* ISO 8859-2 (Latin-2) letters allowed in identifiers *3* ISO 8859-3 (Latin-3) letters allowed in identifiers *4* ISO 8859-4 (Latin-4) letters allowed in identifiers *5* ISO 8859-5 (Cyrillic) letters allowed in identifiers *9* ISO 8859-15 (Latin-9) letters allowed in identifiers *p* IBM PC letters (code page 437) allowed in identifiers *8* IBM PC letters (code page 850) allowed in identifiers *f* Full upper-half codes allowed in identifiers *n* No upper-half codes allowed in identifiers *w* Wide-character codes (that is, codes greater than 255) allowed in identifiers ========== ====================================================== See :ref:`Foreign_Language_Representation` for full details on the implementation of these character sets. .. index:: -gnatW (gcc) :switch:`-gnatW{e}` Specify the method of encoding for wide characters. ``e`` is one of the following: ========== ====================================================== *h* Hex encoding (brackets coding also recognized) *u* Upper half encoding (brackets encoding also recognized) *s* Shift/JIS encoding (brackets encoding also recognized) *e* EUC encoding (brackets encoding also recognized) *8* UTF-8 encoding (brackets encoding also recognized) *b* Brackets encoding only (default value) ========== ====================================================== For full details on these encoding methods see :ref:`Wide_Character_Encodings`. Note that brackets coding is always accepted, even if one of the other options is specified, so for example :switch:`-gnatW8` specifies that both brackets and UTF-8 encodings will be recognized. The units that are with'ed directly or indirectly will be scanned using the specified representation scheme, and so if one of the non-brackets scheme is used, it must be used consistently throughout the program. However, since brackets encoding is always recognized, it may be conveniently used in standard libraries, allowing these libraries to be used with any of the available coding schemes. Note that brackets encoding only applies to program text. Within comments, brackets are considered to be normal graphic characters, and bracket sequences are never recognized as wide characters. If no :switch:`-gnatW?` parameter is present, then the default representation is normally Brackets encoding only. However, if the first three characters of the file are 16#EF# 16#BB# 16#BF# (the standard byte order mark or BOM for UTF-8), then these three characters are skipped and the default representation for the file is set to UTF-8. Note that the wide character representation that is specified (explicitly or by default) for the main program also acts as the default encoding used for Wide_Text_IO files if not specifically overridden by a WCEM form parameter. When no :switch:`-gnatW?` is specified, then characters (other than wide characters represented using brackets notation) are treated as 8-bit Latin-1 codes. The codes recognized are the Latin-1 graphic characters, and ASCII format effectors (CR, LF, HT, VT). Other lower half control characters in the range 16#00#..16#1F# are not accepted in program text or in comments. Upper half control characters (16#80#..16#9F#) are rejected in program text, but allowed and ignored in comments. Note in particular that the Next Line (NEL) character whose encoding is 16#85# is not recognized as an end of line in this default mode. If your source program contains instances of the NEL character used as a line terminator, you must use UTF-8 encoding for the whole source program. In default mode, all lines must be ended by a standard end of line sequence (CR, CR/LF, or LF). Note that the convention of simply accepting all upper half characters in comments means that programs that use standard ASCII for program text, but UTF-8 encoding for comments are accepted in default mode, providing that the comments are ended by an appropriate (CR, or CR/LF, or LF) line terminator. This is a common mode for many programs with foreign language comments. .. _File_Naming_Control: File Naming Control ------------------- .. index:: -gnatk (gcc) :switch:`-gnatk{n}` Activates file name 'krunching'. ``n``, a decimal integer in the range 1-999, indicates the maximum allowable length of a file name (not including the :file:`.ads` or :file:`.adb` extension). The default is not to enable file name krunching. For the source file naming rules, :ref:`File_Naming_Rules`. .. _Subprogram_Inlining_Control: Subprogram Inlining Control --------------------------- .. index:: -gnatn (gcc) :switch:`-gnatn[12]` The ``n`` here is intended to suggest the first syllable of the word 'inline'. GNAT recognizes and processes ``Inline`` pragmas. However, for inlining to actually occur, optimization must be enabled and, by default, inlining of subprograms across units is not performed. If you want to additionally enable inlining of subprograms specified by pragma ``Inline`` across units, you must also specify this switch. In the absence of this switch, GNAT does not attempt inlining across units and does not access the bodies of subprograms for which ``pragma Inline`` is specified if they are not in the current unit. You can optionally specify the inlining level: 1 for moderate inlining across units, which is a good compromise between compilation times and performances at run time, or 2 for full inlining across units, which may bring about longer compilation times. If no inlining level is specified, the compiler will pick it based on the optimization level: 1 for :switch:`-O1`, :switch:`-O2` or :switch:`-Os` and 2 for :switch:`-O3`. If you specify this switch the compiler will access these bodies, creating an extra source dependency for the resulting object file, and where possible, the call will be inlined. For further details on when inlining is possible see :ref:`Inlining_of_Subprograms`. .. index:: -gnatN (gcc) :switch:`-gnatN` This switch activates front-end inlining which also generates additional dependencies. When using a gcc-based back end, then the use of :switch:`-gnatN` is deprecated, and the use of :switch:`-gnatn` is preferred. Historically front end inlining was more extensive than the gcc back end inlining, but that is no longer the case. .. _Auxiliary_Output_Control: Auxiliary Output Control ------------------------ .. index:: -gnatu (gcc) :switch:`-gnatu` Print a list of units required by this compilation on :file:`stdout`. The listing includes all units on which the unit being compiled depends either directly or indirectly. .. index:: -pass-exit-codes (gcc) :switch:`-pass-exit-codes` If this switch is not used, the exit code returned by ``gcc`` when compiling multiple files indicates whether all source files have been successfully used to generate object files or not. When :switch:`-pass-exit-codes` is used, ``gcc`` exits with an extended exit status and allows an integrated development environment to better react to a compilation failure. Those exit status are: ========== ====================================================== *5* There was an error in at least one source file. *3* At least one source file did not generate an object file. *2* The compiler died unexpectedly (internal error for example). *0* An object file has been generated for every source file. ========== ====================================================== .. _Debugging_Control: Debugging Control ----------------- .. index:: Debugging options .. index:: -gnatd (gcc) :switch:`-gnatd{x}` Activate internal debugging switches. ``x`` is a letter or digit, or string of letters or digits, which specifies the type of debugging outputs desired. Normally these are used only for internal development or system debugging purposes. You can find full documentation for these switches in the body of the ``Debug`` unit in the compiler source file :file:`debug.adb`. .. index:: -gnatG (gcc) :switch:`-gnatG[={nn}]` This switch causes the compiler to generate auxiliary output containing a pseudo-source listing of the generated expanded code. Like most Ada compilers, GNAT works by first transforming the high level Ada code into lower level constructs. For example, tasking operations are transformed into calls to the tasking run-time routines. A unique capability of GNAT is to list this expanded code in a form very close to normal Ada source. This is very useful in understanding the implications of various Ada usage on the efficiency of the generated code. There are many cases in Ada (e.g., the use of controlled types), where simple Ada statements can generate a lot of run-time code. By using :switch:`-gnatG` you can identify these cases, and consider whether it may be desirable to modify the coding approach to improve efficiency. The optional parameter ``nn`` if present after -gnatG specifies an alternative maximum line length that overrides the normal default of 72. This value is in the range 40-999999, values less than 40 being silently reset to 40. The equal sign is optional. The format of the output is very similar to standard Ada source, and is easily understood by an Ada programmer. The following special syntactic additions correspond to low level features used in the generated code that do not have any exact analogies in pure Ada source form. The following is a partial list of these special constructions. See the spec of package ``Sprint`` in file :file:`sprint.ads` for a full list. .. index:: -gnatL (gcc) If the switch :switch:`-gnatL` is used in conjunction with :switch:`-gnatG`, then the original source lines are interspersed in the expanded source (as comment lines with the original line number). :samp:`new {xxx} [storage_pool = {yyy}]` Shows the storage pool being used for an allocator. :samp:`at end {procedure-name};` Shows the finalization (cleanup) procedure for a scope. :samp:`(if {expr} then {expr} else {expr})` Conditional expression equivalent to the ``x?y:z`` construction in C. :samp:`{target}^({source})` A conversion with floating-point truncation instead of rounding. :samp:`{target}?({source})` A conversion that bypasses normal Ada semantic checking. In particular enumeration types and fixed-point types are treated simply as integers. :samp:`{target}?^({source})` Combines the above two cases. :samp:`{x} #/ {y}` :samp:`{x} #mod {y}` :samp:`{x} # {y}` :samp:`{x} #rem {y}` A division or multiplication of fixed-point values which are treated as integers without any kind of scaling. :samp:`free {expr} [storage_pool = {xxx}]` Shows the storage pool associated with a ``free`` statement. :samp:`[subtype or type declaration]` Used to list an equivalent declaration for an internally generated type that is referenced elsewhere in the listing. :samp:`freeze {type-name} [{actions}]` Shows the point at which ``type-name`` is frozen, with possible associated actions to be performed at the freeze point. :samp:`reference {itype}` Reference (and hence definition) to internal type ``itype``. :samp:`{function-name}! ({arg}, {arg}, {arg})` Intrinsic function call. :samp:`{label-name} : label` Declaration of label ``labelname``. :samp:`#$ {subprogram-name}` An implicit call to a run-time support routine (to meet the requirement of H.3.1(9) in a convenient manner). :samp:`{expr} && {expr} && {expr} ... && {expr}` A multiple concatenation (same effect as ``expr`` & ``expr`` & ``expr``, but handled more efficiently). :samp:`[constraint_error]` Raise the ``Constraint_Error`` exception. :samp:`{expression}'reference` A pointer to the result of evaluating {expression}. :samp:`{target-type}!({source-expression})` An unchecked conversion of ``source-expression`` to ``target-type``. :samp:`[{numerator}/{denominator}]` Used to represent internal real literals (that) have no exact representation in base 2-16 (for example, the result of compile time evaluation of the expression 1.0/27.0). .. index:: -gnatD (gcc) :switch:`-gnatD[=nn]` When used in conjunction with :switch:`-gnatG`, this switch causes the expanded source, as described above for :switch:`-gnatG` to be written to files with names :file:`xxx.dg`, where :file:`xxx` is the normal file name, instead of to the standard output file. For example, if the source file name is :file:`hello.adb`, then a file :file:`hello.adb.dg` will be written. The debugging information generated by the ``gcc`` :switch:`-g` switch will refer to the generated :file:`xxx.dg` file. This allows you to do source level debugging using the generated code which is sometimes useful for complex code, for example to find out exactly which part of a complex construction raised an exception. This switch also suppresses generation of cross-reference information (see :switch:`-gnatx`) since otherwise the cross-reference information would refer to the :file:`.dg` file, which would cause confusion since this is not the original source file. Note that :switch:`-gnatD` actually implies :switch:`-gnatG` automatically, so it is not necessary to give both options. In other words :switch:`-gnatD` is equivalent to :switch:`-gnatDG`). .. index:: -gnatL (gcc) If the switch :switch:`-gnatL` is used in conjunction with :switch:`-gnatDG`, then the original source lines are interspersed in the expanded source (as comment lines with the original line number). The optional parameter ``nn`` if present after -gnatD specifies an alternative maximum line length that overrides the normal default of 72. This value is in the range 40-999999, values less than 40 being silently reset to 40. The equal sign is optional. .. index:: -gnatr (gcc) .. index:: pragma Restrictions :switch:`-gnatr` This switch causes pragma Restrictions to be treated as Restriction_Warnings so that violation of restrictions causes warnings rather than illegalities. This is useful during the development process when new restrictions are added or investigated. The switch also causes pragma Profile to be treated as Profile_Warnings, and pragma Restricted_Run_Time and pragma Ravenscar set restriction warnings rather than restrictions. .. index:: -gnatR (gcc) :switch:`-gnatR[0|1|2|3|4][e][h][m][j][s]` This switch controls output from the compiler of a listing showing representation information for declared types, objects and subprograms. For :switch:`-gnatR0`, no information is output (equivalent to omitting the :switch:`-gnatR` switch). For :switch:`-gnatR1` (which is the default, so :switch:`-gnatR` with no parameter has the same effect), size and alignment information is listed for declared array and record types. For :switch:`-gnatR2`, size and alignment information is listed for all declared types and objects. The ``Linker_Section`` is also listed for any entity for which the ``Linker_Section`` is set explicitly or implicitly (the latter case occurs for objects of a type for which a ``Linker_Section`` is set). For :switch:`-gnatR3`, symbolic expressions for values that are computed at run time for records are included. These symbolic expressions have a mostly obvious format with #n being used to represent the value of the n'th discriminant. See source files :file:`repinfo.ads/adb` in the GNAT sources for full details on the format of :switch:`-gnatR3` output. For :switch:`-gnatR4`, information for relevant compiler-generated types is also listed, i.e. when they are structurally part of other declared types and objects. If the switch is followed by an ``e`` (e.g. :switch:`-gnatR2e`), then extended representation information for record sub-components of records is included. If the switch is followed by a ``h`` (e.g. :switch:`-gnatR3h`), then the components of records are sorted by increasing offsets and holes between consecutive components are flagged. If the switch is followed by an ``m`` (e.g. :switch:`-gnatRm`), then subprogram conventions and parameter passing mechanisms for all the subprograms are included. If the switch is followed by a ``j`` (e.g. :switch:`-gnatRj`), then the output is in the JSON data interchange format specified by the ECMA-404 standard. The semantic description of this JSON output is available in the specification of the Repinfo unit present in the compiler sources. If the switch is followed by an ``s`` (e.g. :switch:`-gnatR3s`), then the output is to a file with the name :file:`file.rep` where ``file`` is the name of the corresponding source file, except if ``j`` is also specified, in which case the file name is :file:`file.json`. Note that it is possible for record components to have zero size. In this case, the component clause uses an obvious extension of permitted Ada syntax, for example ``at 0 range 0 .. -1``. .. index:: -gnatS (gcc) :switch:`-gnatS` The use of the switch :switch:`-gnatS` for an Ada compilation will cause the compiler to output a representation of package Standard in a form very close to standard Ada. It is not quite possible to do this entirely in standard Ada (since new numeric base types cannot be created in standard Ada), but the output is easily readable to any Ada programmer, and is useful to determine the characteristics of target dependent types in package Standard. .. index:: -gnatx (gcc) :switch:`-gnatx` Normally the compiler generates full cross-referencing information in the :file:`ALI` file. This information is used by a number of tools. The :switch:`-gnatx` switch suppresses this information. This saves some space and may slightly speed up compilation, but means that tools depending on this information cannot be used. .. index:: -fgnat-encodings (gcc) :switch:`-fgnat-encodings=[all|gdb|minimal]` This switch controls the balance between GNAT encodings and standard DWARF emitted in the debug information. Historically, old debug formats like stabs were not powerful enough to express some Ada types (for instance, variant records or fixed-point types). To work around this, GNAT introduced proprietary encodings that embed the missing information ("GNAT encodings"). Recent versions of the DWARF debug information format are now able to correctly describe most of these Ada constructs ("standard DWARF"). As third-party tools started to use this format, GNAT has been enhanced to generate it. However, most tools (including GDB) are still relying on GNAT encodings. To support all tools, GNAT needs to be versatile about the balance between generation of GNAT encodings and standard DWARF. This is what :switch:`-fgnat-encodings` is about. * ``=all``: Emit all GNAT encodings, and then emit as much standard DWARF as possible so it does not conflict with GNAT encodings. * ``=gdb``: Emit as much standard DWARF as possible as long as the current GDB handles it. Emit GNAT encodings for the rest. * ``=minimal``: Emit as much standard DWARF as possible and emit GNAT encodings for the rest. .. _Exception_Handling_Control: Exception Handling Control -------------------------- GNAT uses two methods for handling exceptions at run time. The ``setjmp/longjmp`` method saves the context when entering a frame with an exception handler. Then when an exception is raised, the context can be restored immediately, without the need for tracing stack frames. This method provides very fast exception propagation, but introduces significant overhead for the use of exception handlers, even if no exception is raised. The other approach is called 'zero cost' exception handling. With this method, the compiler builds static tables to describe the exception ranges. No dynamic code is required when entering a frame containing an exception handler. When an exception is raised, the tables are used to control a back trace of the subprogram invocation stack to locate the required exception handler. This method has considerably poorer performance for the propagation of exceptions, but there is no overhead for exception handlers if no exception is raised. Note that in this mode and in the context of mixed Ada and C/C++ programming, to propagate an exception through a C/C++ code, the C/C++ code must be compiled with the :switch:`-funwind-tables` GCC's option. The following switches may be used to control which of the two exception handling methods is used. .. index:: --RTS=sjlj (gnatmake) :switch:`--RTS=sjlj` This switch causes the setjmp/longjmp run-time (when available) to be used for exception handling. If the default mechanism for the target is zero cost exceptions, then this switch can be used to modify this default, and must be used for all units in the partition. This option is rarely used. One case in which it may be advantageous is if you have an application where exception raising is common and the overall performance of the application is improved by favoring exception propagation. .. index:: --RTS=zcx (gnatmake) .. index:: Zero Cost Exceptions :switch:`--RTS=zcx` This switch causes the zero cost approach to be used for exception handling. If this is the default mechanism for the target (see below), then this switch is unneeded. If the default mechanism for the target is setjmp/longjmp exceptions, then this switch can be used to modify this default, and must be used for all units in the partition. This option can only be used if the zero cost approach is available for the target in use, otherwise it will generate an error. The same option :switch:`--RTS` must be used both for ``gcc`` and ``gnatbind``. Passing this option to ``gnatmake`` (:ref:`Switches_for_gnatmake`) will ensure the required consistency through the compilation and binding steps. .. _Units_to_Sources_Mapping_Files: Units to Sources Mapping Files ------------------------------ .. index:: -gnatem (gcc) :switch:`-gnatem={path}` A mapping file is a way to communicate to the compiler two mappings: from unit names to file names (without any directory information) and from file names to path names (with full directory information). These mappings are used by the compiler to short-circuit the path search. The use of mapping files is not required for correct operation of the compiler, but mapping files can improve efficiency, particularly when sources are read over a slow network connection. In normal operation, you need not be concerned with the format or use of mapping files, and the :switch:`-gnatem` switch is not a switch that you would use explicitly. It is intended primarily for use by automatic tools such as ``gnatmake`` running under the project file facility. The description here of the format of mapping files is provided for completeness and for possible use by other tools. A mapping file is a sequence of sets of three lines. In each set, the first line is the unit name, in lower case, with ``%s`` appended for specs and ``%b`` appended for bodies; the second line is the file name; and the third line is the path name. Example:: main%b main.2.ada /gnat/project1/sources/main.2.ada When the switch :switch:`-gnatem` is specified, the compiler will create in memory the two mappings from the specified file. If there is any problem (nonexistent file, truncated file or duplicate entries), no mapping will be created. Several :switch:`-gnatem` switches may be specified; however, only the last one on the command line will be taken into account. When using a project file, ``gnatmake`` creates a temporary mapping file and communicates it to the compiler using this switch. .. _Code_Generation_Control: Code Generation Control ----------------------- The GCC technology provides a wide range of target dependent :switch:`-m` switches for controlling details of code generation with respect to different versions of architectures. This includes variations in instruction sets (e.g., different members of the power pc family), and different requirements for optimal arrangement of instructions (e.g., different members of the x86 family). The list of available :switch:`-m` switches may be found in the GCC documentation. Use of these :switch:`-m` switches may in some cases result in improved code performance. The GNAT technology is tested and qualified without any :switch:`-m` switches, so generally the most reliable approach is to avoid the use of these switches. However, we generally expect most of these switches to work successfully with GNAT, and many customers have reported successful use of these options. Our general advice is to avoid the use of :switch:`-m` switches unless special needs lead to requirements in this area. In particular, there is no point in using :switch:`-m` switches to improve performance unless you actually see a performance improvement. .. _Linker_Switches: Linker Switches =============== Linker switches can be specified after :switch:`-largs` builder switch. .. index:: -fuse-ld=name :switch:`-fuse-ld={name}` Linker to be used. The default is ``bfd`` for :file:`ld.bfd`; ``gold`` (for :file:`ld.gold`) and ``mold`` (for :file:`ld.mold`) are more recent and faster alternatives, but only available on GNU/Linux platforms. .. _Binding_with_gnatbind: Binding with ``gnatbind`` ========================= .. index:: ! gnatbind This chapter describes the GNAT binder, ``gnatbind``, which is used to bind compiled GNAT objects. The ``gnatbind`` program performs four separate functions: * Checks that a program is consistent, in accordance with the rules in Chapter 10 of the Ada Reference Manual. In particular, error messages are generated if a program uses inconsistent versions of a given unit. * Checks that an acceptable order of elaboration exists for the program and issues an error message if it cannot find an order of elaboration that satisfies the rules in Chapter 10 of the Ada Language Manual. * Generates a main program incorporating the given elaboration order. This program is a small Ada package (body and spec) that must be subsequently compiled using the GNAT compiler. The necessary compilation step is usually performed automatically by ``gnatlink``. The two most important functions of this program are to call the elaboration routines of units in an appropriate order and to call the main program. * Determines the set of object files required by the given main program. This information is output in the forms of comments in the generated program, to be read by the ``gnatlink`` utility used to link the Ada application. .. _Running_gnatbind: Running ``gnatbind`` -------------------- The form of the ``gnatbind`` command is .. code-block:: sh $ gnatbind [ switches ] mainprog[.ali] [ switches ] where :file:`mainprog.adb` is the Ada file containing the main program unit body. ``gnatbind`` constructs an Ada package in two files whose names are :file:`b~mainprog.ads`, and :file:`b~mainprog.adb`. For example, if given the parameter :file:`hello.ali`, for a main program contained in file :file:`hello.adb`, the binder output files would be :file:`b~hello.ads` and :file:`b~hello.adb`. When doing consistency checking, the binder takes into consideration any source files it can locate. For example, if the binder determines that the given main program requires the package ``Pack``, whose :file:`.ALI` file is :file:`pack.ali` and whose corresponding source spec file is :file:`pack.ads`, it attempts to locate the source file :file:`pack.ads` (using the same search path conventions as previously described for the ``gcc`` command). If it can locate this source file, it checks that the time stamps or source checksums of the source and its references to in :file:`ALI` files match. In other words, any :file:`ALI` files that mentions this spec must have resulted from compiling this version of the source file (or in the case where the source checksums match, a version close enough that the difference does not matter). .. index:: Source files, use by binder The effect of this consistency checking, which includes source files, is that the binder ensures that the program is consistent with the latest version of the source files that can be located at bind time. Editing a source file without compiling files that depend on the source file cause error messages to be generated by the binder. For example, suppose you have a main program :file:`hello.adb` and a package ``P``, from file :file:`p.ads` and you perform the following steps: * Enter ``gcc -c hello.adb`` to compile the main program. * Enter ``gcc -c p.ads`` to compile package ``P``. * Edit file :file:`p.ads`. * Enter ``gnatbind hello``. At this point, the file :file:`p.ali` contains an out-of-date time stamp because the file :file:`p.ads` has been edited. The attempt at binding fails, and the binder generates the following error messages: :: error: "hello.adb" must be recompiled ("p.ads" has been modified) error: "p.ads" has been modified and must be recompiled Now both files must be recompiled as indicated, and then the bind can succeed, generating a main program. You need not normally be concerned with the contents of this file, but for reference purposes a sample binder output file is given in :ref:`Example_of_Binder_Output_File`. In most normal usage, the default mode of ``gnatbind`` which is to generate the main package in Ada, as described in the previous section. In particular, this means that any Ada programmer can read and understand the generated main program. It can also be debugged just like any other Ada code provided the :switch:`-g` switch is used for ``gnatbind`` and ``gnatlink``. .. _Switches_for_gnatbind: Switches for ``gnatbind`` ------------------------- The following switches are available with ``gnatbind``; details will be presented in subsequent sections. .. index:: --version (gnatbind) :switch:`--version` Display Copyright and version, then exit disregarding all other options. .. index:: --help (gnatbind) :switch:`--help` If :switch:`--version` was not used, display usage, then exit disregarding all other options. .. index:: -a (gnatbind) :switch:`-a` Indicates that, if supported by the platform, the adainit procedure should be treated as an initialisation routine by the linker (a constructor). This is intended to be used by the Project Manager to automatically initialize shared Stand-Alone Libraries. .. index:: -aO (gnatbind) :switch:`-aO` Specify directory to be searched for ALI files. .. index:: -aI (gnatbind) :switch:`-aI` Specify directory to be searched for source file. .. index:: -A (gnatbind) :switch:`-A[={filename}]` Output ALI list (to standard output or to the named file). .. index:: -b (gnatbind) :switch:`-b` Generate brief messages to :file:`stderr` even if verbose mode set. .. index:: -c (gnatbind) :switch:`-c` Check only, no generation of binder output file. .. index:: -dnn[k|m] (gnatbind) :switch:`-d{nn}[k|m]` This switch can be used to change the default task stack size value to a specified size ``nn``, which is expressed in bytes by default, or in kilobytes when suffixed with ``k`` or in megabytes when suffixed with ``m``. In the absence of a :samp:`[k|m]` suffix, this switch is equivalent, in effect, to completing all task specs with .. code-block:: ada pragma Storage_Size (nn); When they do not already have such a pragma. .. index:: -D (gnatbind) :switch:`-D{nn}[k|m]` Set the default secondary stack size to ``nn``. The suffix indicates whether the size is in bytes (no suffix), kilobytes (``k`` suffix) or megabytes (``m`` suffix). The secondary stack holds objects of unconstrained types that are returned by functions, for example unconstrained Strings. The size of the secondary stack can be dynamic or fixed depending on the target. For most targets, the secondary stack grows on demand and is implemented as a chain of blocks in the heap. In this case, the default secondary stack size determines the initial size of the secondary stack for each task and the smallest amount the secondary stack can grow by. For Light, Light-Tasking, and Embedded run-times the size of the secondary stack is fixed. This switch can be used to change the default size of these stacks. The default secondary stack size can be overridden on a per-task basis if individual tasks have different secondary stack requirements. This is achieved through the Secondary_Stack_Size aspect, which takes the size of the secondary stack in bytes. .. index:: -e (gnatbind) :switch:`-e` Output complete list of elaboration-order dependencies. .. index:: -Ea (gnatbind) :switch:`-Ea` Store tracebacks in exception occurrences when the target supports it. The "a" is for "address"; tracebacks will contain hexadecimal addresses, unless symbolic tracebacks are enabled. See also the packages ``GNAT.Traceback`` and ``GNAT.Traceback.Symbolic`` for more information. Note that on x86 ports, you must not use :switch:`-fomit-frame-pointer` ``gcc`` option. .. index:: -Es (gnatbind) :switch:`-Es` Store tracebacks in exception occurrences when the target supports it. The "s" is for "symbolic"; symbolic tracebacks are enabled. .. index:: -E (gnatbind) :switch:`-E` Currently the same as ``-Ea``. .. index:: -f (gnatbind) :switch:`-f{elab-order}` Force elaboration order. For further details see :ref:`Elaboration_Control` and :ref:`Elaboration_Order_Handling_in_GNAT`. .. index:: -F (gnatbind) :switch:`-F` Force the checks of elaboration flags. ``gnatbind`` does not normally generate checks of elaboration flags for the main executable, except when a Stand-Alone Library is used. However, there are cases when this cannot be detected by gnatbind. An example is importing an interface of a Stand-Alone Library through a pragma Import and only specifying through a linker switch this Stand-Alone Library. This switch is used to guarantee that elaboration flag checks are generated. .. index:: -h (gnatbind) :switch:`-h` Output usage (help) information. .. index:: -H (gnatbind) :switch:`-H` Legacy elaboration order model enabled. For further details see :ref:`Elaboration_Order_Handling_in_GNAT`. .. index:: -H32 (gnatbind) :switch:`-H32` Use 32-bit allocations for ``__gnat_malloc`` (and thus for access types). For further details see :ref:`Dynamic_Allocation_Control`. .. index:: -H64 (gnatbind) .. index:: __gnat_malloc :switch:`-H64` Use 64-bit allocations for ``__gnat_malloc`` (and thus for access types). For further details see :ref:`Dynamic_Allocation_Control`. .. index:: -I (gnatbind) :switch:`-I` Specify directory to be searched for source and ALI files. .. index:: -I- (gnatbind) :switch:`-I-` Do not look for sources in the current directory where ``gnatbind`` was invoked, and do not look for ALI files in the directory containing the ALI file named in the ``gnatbind`` command line. .. index:: -k (gnatbind) :switch:`-k` Disable checking of elaboration flags. When using :switch:`-n` either explicitly or implicitly, :switch:`-F` is also implied, unless :switch:`-k` is used. This switch should be used with care and you should ensure manually that elaboration routines are not called twice unintentionally. .. index:: -K (gnatbind) :switch:`-K` Give list of linker options specified for link. .. index:: -l (gnatbind) :switch:`-l` Output chosen elaboration order. .. index:: -L (gnatbind) :switch:`-L{xxx}` Bind the units for library building. In this case the ``adainit`` and ``adafinal`` procedures (:ref:`Binding_with_Non-Ada_Main_Programs`) are renamed to :samp:`{xxx}init` and :samp:`{xxx}final`. Implies -n. (:ref:`GNAT_and_Libraries`, for more details.) .. index:: -M (gnatbind) :switch:`-M{xyz}` Rename generated main program from main to xyz. This option is supported on cross environments only. .. index:: -m (gnatbind) :switch:`-m{n}` Limit number of detected errors or warnings to ``n``, where ``n`` is in the range 1..999999. The default value if no switch is given is 9999. If the number of warnings reaches this limit, then a message is output and further warnings are suppressed, the bind continues in this case. If the number of errors reaches this limit, then a message is output and the bind is abandoned. A value of zero means that no limit is enforced. The equal sign is optional. .. index:: -minimal (gnatbind) :switch:`-minimal` Generate a binder file suitable for space-constrained applications. When active, binder-generated objects not required for program operation are no longer generated. **Warning:** this option comes with the following limitations: * Starting the program's execution in the debugger will cause it to stop at the start of the ``main`` function instead of the main subprogram. This can be worked around by manually inserting a breakpoint on that subprogram and resuming the program's execution until reaching that breakpoint. * Programs using GNAT.Compiler_Version will not link. .. index:: -n (gnatbind) :switch:`-n` No main program. .. index:: -nostdinc (gnatbind) :switch:`-nostdinc` Do not look for sources in the system default directory. .. index:: -nostdlib (gnatbind) :switch:`-nostdlib` Do not look for library files in the system default directory. .. index:: --RTS (gnatbind) :switch:`--RTS={rts-path}` Specifies the default location of the run-time library. Same meaning as the equivalent ``gnatmake`` flag (:ref:`Switches_for_gnatmake`). .. index:: -o (gnatbind) :switch:`-o {file}` Name the output file ``file`` (default is :file:`b~{xxx}.adb`). Note that if this option is used, then linking must be done manually, gnatlink cannot be used. .. index:: -O (gnatbind) :switch:`-O[={filename}]` Output object list (to standard output or to the named file). .. index:: -p (gnatbind) :switch:`-p` Pessimistic (worst-case) elaboration order. .. index:: -P (gnatbind) :switch:`-P` Generate binder file suitable for CodePeer. .. index:: -Q (gnatbind) :switch:`-Q{nnn}` Generate ``nnn`` additional default-sized secondary stacks. Tasks declared at the library level that use default-size secondary stacks have their secondary stacks allocated from a pool of stacks generated by gnatbind. This allows the default secondary stack size to be quickly changed by rebinding the application. While the binder sizes this pool to match the number of such tasks defined in the application, the pool size may need to be increased with the :switch:`-Q` switch to accommodate foreign threads registered with the Light run-time. For more information, please see the *The Primary and Secondary Stack* chapter in the *GNAT User’s Guide Supplement for Cross Platforms*. .. index:: -R (gnatbind) :switch:`-R` Output closure source list, which includes all non-run-time units that are included in the bind. .. index:: -Ra (gnatbind) :switch:`-Ra` Like :switch:`-R` but the list includes run-time units. .. index:: -s (gnatbind) :switch:`-s` Require all source files to be present. .. index:: -S (gnatbind) :switch:`-S{xxx}` Specifies the value to be used when detecting uninitialized scalar objects with pragma Initialize_Scalars. The ``xxx`` string specified with the switch is one of: * ``in`` for an invalid value. If zero is invalid for the discrete type in question, then the scalar value is set to all zero bits. For signed discrete types, the largest possible negative value of the underlying scalar is set (i.e. a one bit followed by all zero bits). For unsigned discrete types, the underlying scalar value is set to all one bits. For floating-point types, a NaN value is set (see body of package System.Scalar_Values for exact values). * ``lo`` for low value. If zero is invalid for the discrete type in question, then the scalar value is set to all zero bits. For signed discrete types, the largest possible negative value of the underlying scalar is set (i.e. a one bit followed by all zero bits). For unsigned discrete types, the underlying scalar value is set to all zero bits. For floating-point, a small value is set (see body of package System.Scalar_Values for exact values). * ``hi`` for high value. If zero is invalid for the discrete type in question, then the scalar value is set to all one bits. For signed discrete types, the largest possible positive value of the underlying scalar is set (i.e. a zero bit followed by all one bits). For unsigned discrete types, the underlying scalar value is set to all one bits. For floating-point, a large value is set (see body of package System.Scalar_Values for exact values). * ``xx`` for hex value (two hex digits). The underlying scalar is set to a value consisting of repeated bytes, whose value corresponds to the given value. For example if ``BF`` is given, then a 32-bit scalar value will be set to the bit pattern ``16#BFBFBFBF#``. .. index:: GNAT_INIT_SCALARS In addition, you can specify :switch:`-Sev` to indicate that the value is to be set at run time. In this case, the program will look for an environment variable of the form :samp:`GNAT_INIT_SCALARS={yy}`, where ``yy`` is one of :samp:`in/lo/hi/{xx}` with the same meanings as above. If no environment variable is found, or if it does not have a valid value, then the default is ``in`` (invalid values). .. index:: -static (gnatbind) :switch:`-static` Link against a static GNAT run-time. .. index:: -shared (gnatbind) :switch:`-shared` Link against a shared GNAT run-time when available. .. index:: -t (gnatbind) :switch:`-t` Tolerate time stamp and other consistency errors. .. index:: -T (gnatbind) :switch:`-T{n}` Set the time slice value to ``n`` milliseconds. If the system supports the specification of a specific time slice value, then the indicated value is used. If the system does not support specific time slice values, but does support some general notion of round-robin scheduling, then any nonzero value will activate round-robin scheduling. A value of zero is treated specially. It turns off time slicing, and in addition, indicates to the tasking run-time that the semantics should match as closely as possible the Annex D requirements of the Ada RM, and in particular sets the default scheduling policy to ``FIFO_Within_Priorities``. .. index:: -u (gnatbind) :switch:`-u{n}` Enable dynamic stack usage, with ``n`` results stored and displayed at program termination. A result is generated when a task terminates. Results that can't be stored are displayed on the fly, at task termination. This option is currently not supported on Itanium platforms. (See :ref:`Dynamic_Stack_Usage_Analysis` for details.) .. index:: -v (gnatbind) :switch:`-v` Verbose mode. Write error messages, header, summary output to :file:`stdout`. .. index:: -V (gnatbind) :switch:`-V{key}={value}` Store the given association of ``key`` to ``value`` in the bind environment. Values stored this way can be retrieved at run time using ``GNAT.Bind_Environment``. .. index:: -w (gnatbind) :switch:`-w{x}` Warning mode; ``x`` = s/e for suppress/treat as error. .. index:: -Wx (gnatbind) :switch:`-Wx{e}` Override default wide character encoding for standard Text_IO files. .. index:: -x (gnatbind) :switch:`-x` Exclude source files (check object consistency only). .. index:: -xdr (gnatbind) :switch:`-xdr` Use the target-independent XDR protocol for stream oriented attributes instead of the default implementation which is based on direct binary representations and is therefore target-and endianness-dependent. However it does not support 128-bit integer types and the exception ``Ada.IO_Exceptions.Device_Error`` is raised if any attempt is made at streaming 128-bit integer types with it. .. index:: -Xnnn (gnatbind) :switch:`-X{nnn}` Set default exit status value, normally 0 for POSIX compliance. .. index:: -y (gnatbind) :switch:`-y` Enable leap seconds support in ``Ada.Calendar`` and its children. .. index:: -z (gnatbind) :switch:`-z` No main subprogram. You may obtain this listing of switches by running ``gnatbind`` with no arguments. .. _Consistency-Checking_Modes: Consistency-Checking Modes ^^^^^^^^^^^^^^^^^^^^^^^^^^ As described earlier, by default ``gnatbind`` checks that object files are consistent with one another and are consistent with any source files it can locate. The following switches control binder access to sources. .. index:: -s (gnatbind) :switch:`-s` Require source files to be present. In this mode, the binder must be able to locate all source files that are referenced, in order to check their consistency. In normal mode, if a source file cannot be located it is simply ignored. If you specify this switch, a missing source file is an error. .. index:: -Wx (gnatbind) :switch:`-Wx{e}` Override default wide character encoding for standard Text_IO files. Normally the default wide character encoding method used for standard [Wide\_[Wide\_]]Text_IO files is taken from the encoding specified for the main source input (see description of switch :switch:`-gnatWx` for the compiler). The use of this switch for the binder (which has the same set of possible arguments) overrides this default as specified. .. index:: -x (gnatbind) :switch:`-x` Exclude source files. In this mode, the binder only checks that ALI files are consistent with one another. Source files are not accessed. The binder runs faster in this mode, and there is still a guarantee that the resulting program is self-consistent. If a source file has been edited since it was last compiled, and you specify this switch, the binder will not detect that the object file is out of date with respect to the source file. Note that this is the mode that is automatically used by ``gnatmake`` because in this case the checking against sources has already been performed by ``gnatmake`` in the course of compilation (i.e., before binding). .. _Binder_Error_Message_Control: Binder Error Message Control ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The following switches provide control over the generation of error messages from the binder: .. index:: -v (gnatbind) :switch:`-v` Verbose mode. In the normal mode, brief error messages are generated to :file:`stderr`. If this switch is present, a header is written to :file:`stdout` and any error messages are directed to :file:`stdout`. All that is written to :file:`stderr` is a brief summary message. .. index:: -b (gnatbind) :switch:`-b` Generate brief error messages to :file:`stderr` even if verbose mode is specified. This is relevant only when used with the :switch:`-v` switch. .. index:: -m (gnatbind) :switch:`-m{n}` Limits the number of error messages to ``n``, a decimal integer in the range 1-999. The binder terminates immediately if this limit is reached. .. index:: -M (gnatbind) :switch:`-M{xxx}` Renames the generated main program from ``main`` to ``xxx``. This is useful in the case of some cross-building environments, where the actual main program is separate from the one generated by ``gnatbind``. .. index:: -ws (gnatbind) .. index:: Warnings :switch:`-ws` Suppress all warning messages. .. index:: -we (gnatbind) :switch:`-we` Treat any warning messages as fatal errors. .. index:: -t (gnatbind) .. index:: Time stamp checks, in binder .. index:: Binder consistency checks .. index:: Consistency checks, in binder :switch:`-t` The binder performs a number of consistency checks including: * Check that time stamps of a given source unit are consistent * Check that checksums of a given source unit are consistent * Check that consistent versions of ``GNAT`` were used for compilation * Check consistency of configuration pragmas as required Normally failure of such checks, in accordance with the consistency requirements of the Ada Reference Manual, causes error messages to be generated which abort the binder and prevent the output of a binder file and subsequent link to obtain an executable. The :switch:`-t` switch converts these error messages into warnings, so that binding and linking can continue to completion even in the presence of such errors. The result may be a failed link (due to missing symbols), or a non-functional executable which has undefined semantics. .. note:: This means that :switch:`-t` should be used only in unusual situations, with extreme care. .. _Elaboration_Control: Elaboration Control ^^^^^^^^^^^^^^^^^^^ The following switches provide additional control over the elaboration order. For further details see :ref:`Elaboration_Order_Handling_in_GNAT`. .. index:: -f (gnatbind) :switch:`-f{elab-order}` Force elaboration order. ``elab-order`` should be the name of a "forced elaboration order file", that is, a text file containing library item names, one per line. A name of the form "some.unit%s" or "some.unit (spec)" denotes the spec of Some.Unit. A name of the form "some.unit%b" or "some.unit (body)" denotes the body of Some.Unit. Each pair of lines is taken to mean that there is an elaboration dependence of the second line on the first. For example, if the file contains: .. code-block:: ada this (spec) this (body) that (spec) that (body) then the spec of This will be elaborated before the body of This, and the body of This will be elaborated before the spec of That, and the spec of That will be elaborated before the body of That. The first and last of these three dependences are already required by Ada rules, so this file is really just forcing the body of This to be elaborated before the spec of That. The given order must be consistent with Ada rules, or else ``gnatbind`` will give elaboration cycle errors. For example, if you say x (body) should be elaborated before x (spec), there will be a cycle, because Ada rules require x (spec) to be elaborated before x (body); you can't have the spec and body both elaborated before each other. If you later add "with That;" to the body of This, there will be a cycle, in which case you should erase either "this (body)" or "that (spec)" from the above forced elaboration order file. Blank lines and Ada-style comments are ignored. Unit names that do not exist in the program are ignored. Units in the GNAT predefined library are also ignored. .. index:: -p (gnatbind) :switch:`-p` Pessimistic elaboration order This switch is only applicable to the pre-20.x legacy elaboration models. The post-20.x elaboration model uses a more informed approach of ordering the units. Normally the binder attempts to choose an elaboration order that is likely to minimize the likelihood of an elaboration order error resulting in raising a ``Program_Error`` exception. This switch reverses the action of the binder, and requests that it deliberately choose an order that is likely to maximize the likelihood of an elaboration error. This is useful in ensuring portability and avoiding dependence on accidental fortuitous elaboration ordering. Normally it only makes sense to use the :switch:`-p` switch if dynamic elaboration checking is used (:switch:`-gnatE` switch used for compilation). This is because in the default static elaboration mode, all necessary ``Elaborate`` and ``Elaborate_All`` pragmas are implicitly inserted. These implicit pragmas are still respected by the binder in :switch:`-p` mode, so a safe elaboration order is assured. Note that :switch:`-p` is not intended for production use; it is more for debugging/experimental use. .. _Output_Control: Output Control ^^^^^^^^^^^^^^ The following switches allow additional control over the output generated by the binder. .. index:: -c (gnatbind) :switch:`-c` Check only. Do not generate the binder output file. In this mode the binder performs all error checks but does not generate an output file. .. index:: -e (gnatbind) :switch:`-e` Output complete list of elaboration-order dependencies, showing the reason for each dependency. This output can be rather extensive but may be useful in diagnosing problems with elaboration order. The output is written to :file:`stdout`. .. index:: -h (gnatbind) :switch:`-h` Output usage information. The output is written to :file:`stdout`. .. index:: -K (gnatbind) :switch:`-K` Output linker options to :file:`stdout`. Includes library search paths, contents of pragmas Ident and Linker_Options, and libraries added by ``gnatbind``. .. index:: -l (gnatbind) :switch:`-l` Output chosen elaboration order. The output is written to :file:`stdout`. .. index:: -O (gnatbind) :switch:`-O` Output full names of all the object files that must be linked to provide the Ada component of the program. The output is written to :file:`stdout`. This list includes the files explicitly supplied and referenced by the user as well as implicitly referenced run-time unit files. The latter are omitted if the corresponding units reside in shared libraries. The directory names for the run-time units depend on the system configuration. .. index:: -o (gnatbind) :switch:`-o {file}` Set name of output file to ``file`` instead of the normal :file:`b~`mainprog`.adb` default. Note that ``file`` denote the Ada binder generated body filename. Note that if this option is used, then linking must be done manually. It is not possible to use gnatlink in this case, since it cannot locate the binder file. .. index:: -r (gnatbind) :switch:`-r` Generate list of ``pragma Restrictions`` that could be applied to the current unit. This is useful for code audit purposes, and also may be used to improve code generation in some cases. .. _Dynamic_Allocation_Control: Dynamic Allocation Control ^^^^^^^^^^^^^^^^^^^^^^^^^^ The heap control switches -- :switch:`-H32` and :switch:`-H64` -- determine whether dynamic allocation uses 32-bit or 64-bit memory. They only affect compiler-generated allocations via ``__gnat_malloc``; explicit calls to ``malloc`` and related functions from the C run-time library are unaffected. :switch:`-H32` Allocate memory on 32-bit heap :switch:`-H64` Allocate memory on 64-bit heap. This is the default unless explicitly overridden by a ``'Size`` clause on the access type. These switches are only effective on VMS platforms. .. _Binding_with_Non-Ada_Main_Programs: Binding with Non-Ada Main Programs ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The description so far has assumed that the main program is in Ada, and that the task of the binder is to generate a corresponding function ``main`` that invokes this Ada main program. GNAT also supports the building of executable programs where the main program is not in Ada, but some of the called routines are written in Ada and compiled using GNAT (:ref:`Mixed_Language_Programming`). The following switch is used in this situation: .. index:: -n (gnatbind) :switch:`-n` No main program. The main program is not in Ada. In this case, most of the functions of the binder are still required, but instead of generating a main program, the binder generates a file containing the following callable routines: .. index:: adainit ``adainit`` You must call this routine to initialize the Ada part of the program by calling the necessary elaboration routines. A call to ``adainit`` is required before the first call to an Ada subprogram. Note that it is assumed that the basic execution environment must be setup to be appropriate for Ada execution at the point where the first Ada subprogram is called. In particular, if the Ada code will do any floating-point operations, then the FPU must be setup in an appropriate manner. For the case of the x86, for example, full precision mode is required. The procedure GNAT.Float_Control.Reset may be used to ensure that the FPU is in the right state. .. index:: adafinal ``adafinal`` You must call this routine to perform any library-level finalization required by the Ada subprograms. A call to ``adafinal`` is required after the last call to an Ada subprogram, and before the program terminates. .. index:: -n (gnatbind) .. index:: Binder, multiple input files If the :switch:`-n` switch is given, more than one ALI file may appear on the command line for ``gnatbind``. The normal ``closure`` calculation is performed for each of the specified units. Calculating the closure means finding out the set of units involved by tracing |with| references. The reason it is necessary to be able to specify more than one ALI file is that a given program may invoke two or more quite separate groups of Ada units. The binder takes the name of its output file from the last specified ALI file, unless overridden by the use of the :switch:`-o file`. .. index:: -o (gnatbind) The output is an Ada unit in source form that can be compiled with GNAT. This compilation occurs automatically as part of the ``gnatlink`` processing. Currently the GNAT run-time requires a FPU using 80 bits mode precision. Under targets where this is not the default it is required to call GNAT.Float_Control.Reset before using floating point numbers (this include float computation, float input and output) in the Ada code. A side effect is that this could be the wrong mode for the foreign code where floating point computation could be broken after this call. .. _Binding_Programs_with_No_Main_Subprogram: Binding Programs with No Main Subprogram ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ It is possible to have an Ada program which does not have a main subprogram. This program will call the elaboration routines of all the packages, then the finalization routines. The following switch is used to bind programs organized in this manner: .. index:: -z (gnatbind) :switch:`-z` Normally the binder checks that the unit name given on the command line corresponds to a suitable main subprogram. When this switch is used, a list of ALI files can be given, and the execution of the program consists of elaboration of these units in an appropriate order. Note that the default wide character encoding method for standard Text_IO files is always set to Brackets if this switch is set (you can use the binder switch :switch:`-Wx` to override this default). .. _Command-Line_Access: Command-Line Access ------------------- The package ``Ada.Command_Line`` provides access to the command-line arguments and program name. In order for this interface to operate correctly, the two variables .. code-block:: c int gnat_argc; char **gnat_argv; .. index:: gnat_argv .. index:: gnat_argc are declared in one of the GNAT library routines. These variables must be set from the actual ``argc`` and ``argv`` values passed to the main program. With no :switch:`-n` present, ``gnatbind`` generates the C main program to automatically set these variables. If the :switch:`-n` switch is used, there is no automatic way to set these variables. If they are not set, the procedures in ``Ada.Command_Line`` will not be available, and any attempt to use them will raise ``Constraint_Error``. If command line access is required, your main program must set ``gnat_argc`` and ``gnat_argv`` from the ``argc`` and ``argv`` values passed to it. .. _Search_Paths_for_gnatbind: Search Paths for ``gnatbind`` ----------------------------- The binder takes the name of an ALI file as its argument and needs to locate source files as well as other ALI files to verify object consistency. For source files, it follows exactly the same search rules as ``gcc`` (see :ref:`Search_Paths_and_the_Run-Time_Library_RTL`). For ALI files the directories searched are: * The directory containing the ALI file named in the command line, unless the switch :switch:`-I-` is specified. * All directories specified by :switch:`-I` switches on the ``gnatbind`` command line, in the order given. .. index:: ADA_PRJ_OBJECTS_FILE * Each of the directories listed in the text file whose name is given by the :envvar:`ADA_PRJ_OBJECTS_FILE` environment variable. :envvar:`ADA_PRJ_OBJECTS_FILE` is normally set by gnatmake or by the gnat driver when project files are used. It should not normally be set by other means. .. index:: ADA_OBJECTS_PATH * Each of the directories listed in the value of the :envvar:`ADA_OBJECTS_PATH` environment variable. Construct this value exactly as the :envvar:`PATH` environment variable: a list of directory names separated by colons (semicolons when working with the NT version of GNAT). * The content of the :file:`ada_object_path` file which is part of the GNAT installation tree and is used to store standard libraries such as the GNAT Run-Time Library (RTL) unless the switch :switch:`-nostdlib` is specified. See :ref:`Installing_a_library` .. index:: -I (gnatbind) .. index:: -aI (gnatbind) .. index:: -aO (gnatbind) In the binder the switch :switch:`-I` is used to specify both source and library file paths. Use :switch:`-aI` instead if you want to specify source paths only, and :switch:`-aO` if you want to specify library paths only. This means that for the binder :switch:`-I{dir}` is equivalent to :switch:`-aI{dir}` :switch:`-aO`{dir}`. The binder generates the bind file (a C language source file) in the current working directory. .. index:: Ada .. index:: System .. index:: Interfaces .. index:: GNAT The packages ``Ada``, ``System``, and ``Interfaces`` and their children make up the GNAT Run-Time Library, together with the package GNAT and its children, which contain a set of useful additional library functions provided by GNAT. The sources for these units are needed by the compiler and are kept together in one directory. The ALI files and object files generated by compiling the RTL are needed by the binder and the linker and are kept together in one directory, typically different from the directory containing the sources. In a normal installation, you need not specify these directory names when compiling or binding. Either the environment variables or the built-in defaults cause these files to be found. Besides simplifying access to the RTL, a major use of search paths is in compiling sources from multiple directories. This can make development environments much more flexible. .. _Examples_of_gnatbind_Usage: Examples of ``gnatbind`` Usage ------------------------------ Here are some examples of ``gnatbind`` invocations: :: gnatbind hello The main program ``Hello`` (source program in :file:`hello.adb`) is bound using the standard switch settings. The generated main program is :file:`b~hello.adb`. This is the normal, default use of the binder. :: gnatbind hello -o mainprog.adb The main program ``Hello`` (source program in :file:`hello.adb`) is bound using the standard switch settings. The generated main program is :file:`mainprog.adb` with the associated spec in :file:`mainprog.ads`. Note that you must specify the body here not the spec. Note that if this option is used, then linking must be done manually, since gnatlink will not be able to find the generated file. .. _Linking_with_gnatlink: Linking with ``gnatlink`` ========================= .. index:: ! gnatlink This chapter discusses ``gnatlink``, a tool that links an Ada program and builds an executable file. This utility invokes the system linker (via the ``gcc`` command) with a correct list of object files and library references. ``gnatlink`` automatically determines the list of files and references for the Ada part of a program. It uses the binder file generated by the ``gnatbind`` to determine this list. .. _Running_gnatlink: Running ``gnatlink`` -------------------- The form of the ``gnatlink`` command is .. code-block:: sh $ gnatlink [ switches ] mainprog [.ali] [ non-Ada objects ] [ linker options ] The arguments of ``gnatlink`` (switches, main ``ALI`` file, non-Ada objects or linker options) may be in any order, provided that no non-Ada object may be mistaken for a main :file:`ALI` file. Any file name :file:`F` without the :file:`.ali` extension will be taken as the main :file:`ALI` file if a file exists whose name is the concatenation of :file:`F` and :file:`.ali`. :file:`mainprog.ali` references the ALI file of the main program. The :file:`.ali` extension of this file can be omitted. From this reference, ``gnatlink`` locates the corresponding binder file :file:`b~mainprog.adb` and, using the information in this file along with the list of non-Ada objects and linker options, constructs a linker command file to create the executable. The arguments other than the ``gnatlink`` switches and the main :file:`ALI` file are passed to the linker uninterpreted. They typically include the names of object files for units written in other languages than Ada and any library references required to resolve references in any of these foreign language units, or in ``Import`` pragmas in any Ada units. ``linker options`` is an optional list of linker specific switches. The default linker called by gnatlink is ``gcc`` which in turn calls the appropriate system linker. One useful option for the linker is :switch:`-s`: it reduces the size of the executable by removing all symbol table and relocation information from the executable. Standard options for the linker such as :switch:`-lmy_lib` or :switch:`-Ldir` can be added as is. For options that are not recognized by ``gcc`` as linker options, use the ``gcc`` switches :switch:`-Xlinker` or :switch:`-Wl,`. Refer to the GCC documentation for details. Here is an example showing how to generate a linker map: .. code-block:: sh $ gnatlink my_prog -Wl,-Map,MAPFILE Using ``linker options`` it is possible to set the program stack and heap size. See :ref:`Setting_Stack_Size_from_gnatlink` and :ref:`Setting_Heap_Size_from_gnatlink`. ``gnatlink`` determines the list of objects required by the Ada program and prepends them to the list of objects passed to the linker. ``gnatlink`` also gathers any arguments set by the use of ``pragma Linker_Options`` and adds them to the list of arguments presented to the linker. .. _Switches_for_gnatlink: Switches for ``gnatlink`` ------------------------- The following switches are available with the ``gnatlink`` utility: .. index:: --version (gnatlink) :switch:`--version` Display Copyright and version, then exit disregarding all other options. .. index:: --help (gnatlink) :switch:`--help` If :switch:`--version` was not used, display usage, then exit disregarding all other options. .. index:: Command line length .. index:: -f (gnatlink) :switch:`-f` On some targets, the command line length is limited, and ``gnatlink`` will generate a separate file for the linker if the list of object files is too long. The :switch:`-f` switch forces this file to be generated even if the limit is not exceeded. This is useful in some cases to deal with special situations where the command line length is exceeded. .. index:: Debugging information, including .. index:: -g (gnatlink) :switch:`-g` The option to include debugging information causes the Ada bind file (in other words, :file:`b~mainprog.adb`) to be compiled with :switch:`-g`. In addition, the binder does not delete the :file:`b~mainprog.adb`, :file:`b~mainprog.o` and :file:`b~mainprog.ali` files. Without :switch:`-g`, the binder removes these files by default. .. index:: -n (gnatlink) :switch:`-n` Do not compile the file generated by the binder. This may be used when a link is rerun with different options, but there is no need to recompile the binder file. .. index:: -v (gnatlink) :switch:`-v` Verbose mode. Causes additional information to be output, including a full list of the included object files. This switch option is most useful when you want to see what set of object files are being used in the link step. .. index:: -v -v (gnatlink) :switch:`-v -v` Very verbose mode. Requests that the compiler operate in verbose mode when it compiles the binder file, and that the system linker run in verbose mode. .. index:: -o (gnatlink) :switch:`-o {exec-name}` ``exec-name`` specifies an alternate name for the generated executable program. If this switch is omitted, the executable has the same name as the main unit. For example, ``gnatlink try.ali`` creates an executable called :file:`try`. .. index:: -B (gnatlink) :switch:`-B{dir}` Load compiler executables (for example, ``gnat1``, the Ada compiler) from ``dir`` instead of the default location. Only use this switch when multiple versions of the GNAT compiler are available. See the ``Directory Options`` section in :title:`The_GNU_Compiler_Collection` for further details. You would normally use the :switch:`-b` or :switch:`-V` switch instead. .. index:: -M (gnatlink) :switch:`-M` When linking an executable, create a map file. The name of the map file has the same name as the executable with extension ".map". .. index:: -M= (gnatlink) :switch:`-M={mapfile}` When linking an executable, create a map file. The name of the map file is ``mapfile``. .. index:: --GCC=compiler_name (gnatlink) :switch:`--GCC={compiler_name}` Program used for compiling the binder file. The default is ``gcc``. You need to use quotes around ``compiler_name`` if ``compiler_name`` contains spaces or other separator characters. As an example ``--GCC="foo -x -y"`` will instruct ``gnatlink`` to use ``foo -x -y`` as your compiler. Note that switch ``-c`` is always inserted after your command name. Thus in the above example the compiler command that will be used by ``gnatlink`` will be ``foo -c -x -y``. A limitation of this syntax is that the name and path name of the executable itself must not include any embedded spaces. If the compiler executable is different from the default one (gcc or -gcc), then the back end switches in the ALI file are not used to compile the binder generated source. For example, this is the case with ``--GCC="foo -x -y"``. But the back end switches will be used for ``--GCC="gcc -gnatv"``. If several ``--GCC=compiler_name`` are used, only the last ``compiler_name`` is taken into account. However, all the additional switches are also taken into account. Thus, ``--GCC="foo -x -y" --GCC="bar -z -t"`` is equivalent to ``--GCC="bar -x -y -z -t"``. .. index:: --LINK= (gnatlink) :switch:`--LINK={name}` ``name`` is the name of the linker to be invoked. This is especially useful in mixed language programs since languages such as C++ require their own linker to be used. When this switch is omitted, the default name for the linker is ``gcc``. When this switch is used, the specified linker is called instead of ``gcc`` with exactly the same parameters that would have been passed to ``gcc`` so if the desired linker requires different parameters it is necessary to use a wrapper script that massages the parameters before invoking the real linker. It may be useful to control the exact invocation by using the verbose switch. .. _Using_the_GNU_make_Utility: Using the GNU ``make`` Utility ============================== .. index:: make (GNU), GNU make This chapter offers some examples of makefiles that solve specific problems. It does not explain how to write a makefile, nor does it try to replace the ``gnatmake`` utility (:ref:`The_GNAT_Make_Program_gnatmake`). All the examples in this section are specific to the GNU version of make. Although ``make`` is a standard utility, and the basic language is the same, these examples use some advanced features found only in ``GNU make``. .. _Using_gnatmake_in_a_Makefile: Using gnatmake in a Makefile ---------------------------- .. index makefile (GNU make) Complex project organizations can be handled in a very powerful way by using GNU make combined with gnatmake. For instance, here is a Makefile which allows you to build each subsystem of a big project into a separate shared library. Such a makefile allows you to significantly reduce the link time of very big applications while maintaining full coherence at each step of the build process. The list of dependencies are handled automatically by ``gnatmake``. The Makefile is simply used to call gnatmake in each of the appropriate directories. Note that you should also read the example on how to automatically create the list of directories (:ref:`Automatically_Creating_a_List_of_Directories`) which might help you in case your project has a lot of subdirectories. .. code-block:: makefile ## This Makefile is intended to be used with the following directory ## configuration: ## - The sources are split into a series of csc (computer software components) ## Each of these csc is put in its own directory. ## Their name are referenced by the directory names. ## They will be compiled into shared library (although this would also work ## with static libraries) ## - The main program (and possibly other packages that do not belong to any ## csc) is put in the top level directory (where the Makefile is). ## toplevel_dir __ first_csc (sources) __ lib (will contain the library) ## \\_ second_csc (sources) __ lib (will contain the library) ## \\_ ... ## Although this Makefile is build for shared library, it is easy to modify ## to build partial link objects instead (modify the lines with -shared and ## gnatlink below) ## ## With this makefile, you can change any file in the system or add any new ## file, and everything will be recompiled correctly (only the relevant shared ## objects will be recompiled, and the main program will be re-linked). # The list of computer software component for your project. This might be # generated automatically. CSC_LIST=aa bb cc # Name of the main program (no extension) MAIN=main # If we need to build objects with -fPIC, uncomment the following line #NEED_FPIC=-fPIC # The following variable should give the directory containing libgnat.so # You can get this directory through 'gnatls -v'. This is usually the last # directory in the Object_Path. GLIB=... # The directories for the libraries # (This macro expands the list of CSC to the list of shared libraries, you # could simply use the expanded form: # LIB_DIR=aa/lib/libaa.so bb/lib/libbb.so cc/lib/libcc.so LIB_DIR=${foreach dir,${CSC_LIST},${dir}/lib/lib${dir}.so} ${MAIN}: objects ${LIB_DIR} gnatbind ${MAIN} ${CSC_LIST:%=-aO%/lib} -shared gnatlink ${MAIN} ${CSC_LIST:%=-l%} objects:: # recompile the sources gnatmake -c -i ${MAIN}.adb ${NEED_FPIC} ${CSC_LIST:%=-I%} # Note: In a future version of GNAT, the following commands will be simplified # by a new tool, gnatmlib ${LIB_DIR}: mkdir -p ${dir $@ } cd ${dir $@ } && gcc -shared -o ${notdir $@ } ../*.o -L${GLIB} -lgnat cd ${dir $@ } && cp -f ../*.ali . # The dependencies for the modules # Note that we have to force the expansion of *.o, since in some cases # make won't be able to do it itself. aa/lib/libaa.so: ${wildcard aa/*.o} bb/lib/libbb.so: ${wildcard bb/*.o} cc/lib/libcc.so: ${wildcard cc/*.o} # Make sure all of the shared libraries are in the path before starting the # program run:: LD_LIBRARY_PATH=`pwd`/aa/lib:`pwd`/bb/lib:`pwd`/cc/lib ./${MAIN} clean:: ${RM} -rf ${CSC_LIST:%=%/lib} ${RM} ${CSC_LIST:%=%/*.ali} ${RM} ${CSC_LIST:%=%/*.o} ${RM} *.o *.ali ${MAIN} .. _Automatically_Creating_a_List_of_Directories: Automatically Creating a List of Directories -------------------------------------------- In most makefiles, you will have to specify a list of directories, and store it in a variable. For small projects, it is often easier to specify each of them by hand, since you then have full control over what is the proper order for these directories, which ones should be included. However, in larger projects, which might involve hundreds of subdirectories, it might be more convenient to generate this list automatically. The example below presents two methods. The first one, although less general, gives you more control over the list. It involves wildcard characters, that are automatically expanded by ``make``. Its shortcoming is that you need to explicitly specify some of the organization of your project, such as for instance the directory tree depth, whether some directories are found in a separate tree, etc. The second method is the most general one. It requires an external program, called ``find``, which is standard on all Unix systems. All the directories found under a given root directory will be added to the list. .. code-block:: makefile # The examples below are based on the following directory hierarchy: # All the directories can contain any number of files # ROOT_DIRECTORY -> a -> aa -> aaa # -> ab # -> ac # -> b -> ba -> baa # -> bb # -> bc # This Makefile creates a variable called DIRS, that can be reused any time # you need this list (see the other examples in this section) # The root of your project's directory hierarchy ROOT_DIRECTORY=. #### # First method: specify explicitly the list of directories # This allows you to specify any subset of all the directories you need. #### DIRS := a/aa/ a/ab/ b/ba/ #### # Second method: use wildcards # Note that the argument(s) to wildcard below should end with a '/'. # Since wildcards also return file names, we have to filter them out # to avoid duplicate directory names. # We thus use make's ``dir`` and ``sort`` functions. # It sets DIRs to the following value (note that the directories aaa and baa # are not given, unless you change the arguments to wildcard). # DIRS= ./a/a/ ./b/ ./a/aa/ ./a/ab/ ./a/ac/ ./b/ba/ ./b/bb/ ./b/bc/ #### DIRS := ${sort ${dir ${wildcard ${ROOT_DIRECTORY}/*/ ${ROOT_DIRECTORY}/*/*/}}} #### # Third method: use an external program # This command is much faster if run on local disks, avoiding NFS slowdowns. # This is the most complete command: it sets DIRs to the following value: # DIRS= ./a ./a/aa ./a/aa/aaa ./a/ab ./a/ac ./b ./b/ba ./b/ba/baa ./b/bb ./b/bc #### DIRS := ${shell find ${ROOT_DIRECTORY} -type d -print} .. _Generating_the_Command_Line_Switches: Generating the Command Line Switches ------------------------------------ Once you have created the list of directories as explained in the previous section (:ref:`Automatically_Creating_a_List_of_Directories`), you can easily generate the command line arguments to pass to gnatmake. For the sake of completeness, this example assumes that the source path is not the same as the object path, and that you have two separate lists of directories. .. code-block:: makefile # see "Automatically creating a list of directories" to create # these variables SOURCE_DIRS= OBJECT_DIRS= GNATMAKE_SWITCHES := ${patsubst %,-aI%,${SOURCE_DIRS}} GNATMAKE_SWITCHES += ${patsubst %,-aO%,${OBJECT_DIRS}} all: gnatmake ${GNATMAKE_SWITCHES} main_unit .. _Overcoming_Command_Line_Length_Limits: Overcoming Command Line Length Limits ------------------------------------- One problem that might be encountered on big projects is that many operating systems limit the length of the command line. It is thus hard to give gnatmake the list of source and object directories. This example shows how you can set up environment variables, which will make ``gnatmake`` behave exactly as if the directories had been specified on the command line, but have a much higher length limit (or even none on most systems). It assumes that you have created a list of directories in your Makefile, using one of the methods presented in :ref:`Automatically_Creating_a_List_of_Directories`. For the sake of completeness, we assume that the object path (where the ALI files are found) is different from the sources patch. Note a small trick in the Makefile below: for efficiency reasons, we create two temporary variables (SOURCE_LIST and OBJECT_LIST), that are expanded immediately by ``make``. This way we overcome the standard make behavior which is to expand the variables only when they are actually used. On Windows, if you are using the standard Windows command shell, you must replace colons with semicolons in the assignments to these variables. .. code-block:: makefile # In this example, we create both ADA_INCLUDE_PATH and ADA_OBJECTS_PATH. # This is the same thing as putting the -I arguments on the command line. # (the equivalent of using -aI on the command line would be to define # only ADA_INCLUDE_PATH, the equivalent of -aO is ADA_OBJECTS_PATH). # You can of course have different values for these variables. # # Note also that we need to keep the previous values of these variables, since # they might have been set before running 'make' to specify where the GNAT # library is installed. # see "Automatically creating a list of directories" to create these # variables SOURCE_DIRS= OBJECT_DIRS= empty:= space:=${empty} ${empty} SOURCE_LIST := ${subst ${space},:,${SOURCE_DIRS}} OBJECT_LIST := ${subst ${space},:,${OBJECT_DIRS}} ADA_INCLUDE_PATH += ${SOURCE_LIST} ADA_OBJECTS_PATH += ${OBJECT_LIST} export ADA_INCLUDE_PATH export ADA_OBJECTS_PATH all: gnatmake main_unit .. _GNATLLVM: GNAT with the LLVM Back End =========================== This section outlines the usage of the GNAT compiler with the LLVM back end and highlights its key limitations. Certain GNAT versions, referred to as GNAT LLVM, include an alternative LLVM back end alongside the GCC back end, providing access to utilities that operate at the LLVM Intermediate Representation (IR) level. This also enhances safety by facilitating dissimilar redundancy through diverse code generation techniques, allowing for the creation of two distinct binaries from the same source code. Although both GNAT LLVM and the GCC-based GNAT follow most ABI rules, there are some cases where there you may encounter an incompatibility between the two compilers. One such case for the 64-bit Intel X86 is a difference in parameter passing when a structure that consists of 64 bits is passed. The native LLVM handling (and hence that of GNAT LLVM) and ``clang`` disagree in this case. GCC follows ``clang``. The formal ABI agrees with LLVM. In any case, we don't recommend you link code compiled with GNAT LLVM to code compiled by the GCC version of GNAT. This is a specific case of the general rule that you should compile all your Ada code with the same version of GNAT. Both ``gnatmake`` and ``gprbuild`` ensure this is done. You may, however, run into this incompatibility if you pass such a record between C and Ada. In general, we recommend keeping the data passed between C and Ada as simple as practical. GNAT LLVM currently provides limited support for debugging data. It provides full line number information for declarations and statements, and complete debugging data for all types. In some situations, a name in the source may not be available. For example, renamings are not currently emitted into the debugging data. In addition, debugging information may be confusing if you have ``out`` parameters to subprograms. If you have a procedure with only one ``out`` parameter, GNAT LLVM converts that to a function returning an object of that type. If you have multiple ``out`` parameters or have a function that also has an ``out`` parameter, GNAT LLVM converts that subprogram into a function that returns a record where each field is either an ``out`` parameter or the function return value, if any. The debug information reflects these transformations and not the original Ada source code. GNAT LLVM doesn't fully implement the :switch:`-fstack-check` switch. When you specify it, the code generated by GNAT LLVM tests for allocating overly-large items on the stack, but not all cases of stack overflow. For example, if you have a very deep recursion where each call only uses a small amount of stack and the total stack depth exceeds the amount of available stack, the program will be terminated by a signal instead of raising an Ada exception. GNAT LLVM doesn't support the ``Scalar_Storage_Order`` attribute except when it's used to confirm the chosen storage order. This is because this facility is provided by GCC but not by LLVM. GNAT LLVM doesn't support Convention C++, which provides so-called 'name mangling' by encoding parameter and return datatypes into a function name. The inline assembler in GNAT LLVM is not yet feature complete. For example, it can't handle more than one output variable, and it doesn't support all constraints. .. only:: PRO Unlike the GCC versions of GNAT, GNAT LLVM doesn't include any Ada libraries. You can obtain source packages from GNATtracker and build them with the compiler configuration appropriate for your project. AdaCore internally builds and tests the GNAT Components Collection, XML/Ada, and AUnit libraries to ensure compatibility with GNAT LLVM, but not all options of those libraries are currently supported with all targets of GNAT LLVM. If you need to use any of these libraries, please open a support ticket. We provide two options that you can use to build code with GNAT LLVM: * GNAT LLVM includes a version of ``gnatmake`` called ``llvm-gnatmake``, which is equivalent to ``gnatmake`` and has the same switches, except that it uses GNAT LLVM instead of the GCC version of GNAT. * ``gprbuild`` can detect and use GNAT LLVM when it is installed. ``gprbuild`` uses the first applicable compiler on the executable search path, including GNAT LLVM. An easy way to build with GNAT LLVM is to make it available on the operating system's search path before any other Ada compiler (such as the GCC version of GNAT). To avoid accidentally using a different compiler than the one you want to use, we recommend explicitly selecting GNAT LLVM in your existing GPR project file by adding :code:`for Toolchain_Name ("Ada") use "GNAT_LLVM";`. You can also generate an explicit toolchain configuration file with ``gprconfig`` and use it with ``gprbuild``; see the *GPRbuild and GPR Companion Tools User's Guide* for details. You can determine from the first line of the :file:`.ali` file which version of GNAT built that file because it contains either :code:`GNAT` or :code:`GNAT-LLVM`. .. only:: PRO If your project uses one of the libraries packaged with the GCC version of GNAT, it is important that you use the GPR tools from the GNAT LLVM package when building with GNAT LLVM. Otherwise, ``gprbuild`` would try to use the precompiled libraries shipped with the GCC version of GNAT, mixing those objects with code generated by GNAT LLVM, which is not supported. GNAT LLVM understands the same target triplets as the GCC version of GNAT. .. only:: PRO It provides the same runtimes. --- ### Gcc Ada Doc Gnat Ugn Elaboration Order Handling In Gnat (`gcc/ada/doc/gnat_ugn/elaboration_order_handling_in_gnat.rst`) .. role:: switch(samp) .. |with| replace:: *with* .. |withs| replace:: *with*\ s .. |withed| replace:: *with*\ ed .. |withing| replace:: *with*\ ing .. -- Example: A |withing| unit has a |with| clause, it |withs| a |withed| unit .. _Elaboration_Order_Handling_in_GNAT: ********************************** Elaboration Order Handling in GNAT ********************************** .. index:: Order of elaboration .. index:: Elaboration control This appendix describes the handling of elaboration code in Ada and GNAT and discusses how you can control the order of elaboration of program units in GNAT, either automatically or with explicit programming features. .. _Elaboration_Code: Elaboration Code ================ Ada defines the term *execution* as the process by which a construct achieves its run-time effect. This process is also referred to as **elaboration** for declarations and *evaluation* for expressions. The execution model in Ada allows for certain sections of an Ada program to be executed prior to execution of the program itself, primarily with the intent of initializing data. These sections are referred to as **elaboration code**. Elaboration code is executed as follows: * All partitions of an Ada program are executed in parallel with one another, possibly in a separate address space and possibly on a separate computer. * The execution of a partition involves running the environment task for that partition. * The environment task executes all elaboration code (if available) for all units within that partition. This code is said to be executed at **elaboration time**. * The environment task executes the Ada program (if available) for that partition. In addition to the Ada terminology, this appendix defines the following terms: * *Invocation* The act of calling a subprogram, instantiating a generic, or activating a task. * *Scenario* A construct that is elaborated or invoked by elaboration code is referred to as an *elaboration scenario* or simply a **scenario**. GNAT recognizes the following scenarios: - ``'Access`` of entries, operators, and subprograms - Activation of tasks - Calls to entries, operators, and subprograms - Instantiations of generic templates * *Target* A construct elaborated by a scenario is referred to as an *elaboration target* or simply a **target**. GNAT recognizes the following targets: - For ``'Access`` of entries, operators, and subprograms, the target is the entry, operator, or subprogram being aliased. - For activation of tasks, the target is the task body - For calls to entries, operators, and subprograms, the target is the entry, operator, or subprogram being invoked. - For instantiations of generic templates, the target is the generic template being instantiated. Elaboration code may appear in two distinct contexts: * *Library level* A scenario appears at the library level when it is encapsulated by a package [body] compilation unit, ignoring any other package [body] declarations in between. .. code-block:: ada with Server; package Client is procedure Proc; package Nested is Val : ... := Server.Func; end Nested; end Client; In the example above, the call to ``Server.Func`` is an elaboration scenario because it appears at the library level of package ``Client``. Note that the declaration of package ``Nested`` is ignored according to the definition given above. As a result, the call to ``Server.Func`` will be invoked when the spec of unit ``Client`` is elaborated. * *Package body statements* A scenario appears within the statement sequence of a package body when it is bounded by the region starting from the ``begin`` keyword of the package body and ending at the ``end`` keyword of the package body. .. code-block:: ada package body Client is procedure Proc is begin ... end Proc; begin Proc; end Client; In the example above, the call to ``Proc`` is an elaboration scenario because it appears within the statement sequence of package body ``Client``. As a result, the call to ``Proc`` will be invoked when the body of ``Client`` is elaborated. .. _Elaboration_Order: Elaboration Order ================= The sequence by which the elaboration code of all units within a partition is executed is referred to as **elaboration order**. Within a single unit, elaboration code is executed in sequential order. .. code-block:: ada package body Client is Result : ... := Server.Func; procedure Proc is package Inst is new Server.Gen; begin Inst.Eval (Result); end Proc; begin Proc; end Client; In the example above, the elaboration order within package body ``Client`` is as follows: 1. The object declaration of ``Result`` is elaborated. * Function ``Server.Func`` is invoked. 2. The subprogram body of ``Proc`` is elaborated. 3. Procedure ``Proc`` is invoked. * Generic unit ``Server.Gen`` is instantiated as ``Inst``. * Instance ``Inst`` is elaborated. * Procedure ``Inst.Eval`` is invoked. The elaboration order of all units within a partition depends on the following factors: * |withed| units * parent units * purity of units * preelaborability of units * presence of elaboration-control pragmas * invocations performed in elaboration code A program may have several possible elaboration orders depending on its structure: .. code-block:: ada package Server is function Func (Index : Integer) return Integer; end Server; .. code-block:: ada package body Server is Results : array (1 .. 5) of Integer := (1, 2, 3, 4, 5); function Func (Index : Integer) return Integer is begin return Results (Index); end Func; end Server; .. code-block:: ada with Server; package Client is Val : constant Integer := Server.Func (3); end Client; .. code-block:: ada with Client; procedure Main is begin null; end Main; The following elaboration order exhibits a fundamental problem referred to as *access-before-elaboration* or simply **ABE**. :: spec of Server spec of Client body of Server body of Main The elaboration of ``Server``'s spec materializes function ``Func``, making it callable. The elaboration of ``Client``'s spec elaborates the declaration of ``Val``. This invokes function ``Server.Func``, however the body of ``Server.Func`` has not been elaborated yet because ``Server``'s body comes after ``Client``'s spec in the elaboration order. As a result, the value of constant ``Val`` is now undefined. Without any guarantees from the language, an undetected ABE problem may hinder proper initialization of data, which in turn may lead to undefined behavior at run time. To prevent such ABE problems, Ada employs dynamic checks in the same vein as index or null exclusion checks. A failed ABE check raises exception ``Program_Error``. The following elaboration order avoids the ABE problem and the program can be successfully elaborated. :: spec of Server body of Server spec of Client body of Main Ada states that a total elaboration order must exist, but it does not define what this order is. A compiler is thus tasked with choosing a suitable elaboration order which satisfies the dependencies imposed by |with| clauses, unit categorization, elaboration-control pragmas, and invocations performed in elaboration code. Ideally, an order that avoids ABE problems should be chosen, however a compiler may not always find such an order due to complications with respect to control and data flow. .. _Checking_the_Elaboration_Order: Checking the Elaboration Order ============================== To avoid placing the entire elaboration-order burden on the programmer, Ada provides three lines of defense: * *Static semantics* Static semantic rules restrict the possible choice of elaboration order. For instance, if unit Client |withs| unit Server, then the spec of Server is always elaborated prior to Client. The same principle applies to child units - the spec of a parent unit is always elaborated prior to the child unit. * *Dynamic semantics* Dynamic checks are performed at run time to ensure that a target is elaborated prior to a scenario that invokes it, thus avoiding ABE problems. A failed run-time check raises exception ``Program_Error``. The following restrictions apply: - *Restrictions on calls* An entry, operator, or subprogram can be called from elaboration code only when the corresponding body has been elaborated. - *Restrictions on instantiations* A generic unit can be instantiated by elaboration code only when the corresponding body has been elaborated. - *Restrictions on task activation* A task can be activated by elaboration code only when the body of the associated task type has been elaborated. The restrictions above can be summarized by the following rule: *If a target has a body, then this body must be elaborated prior to the scenario that invokes the target.* * *Elaboration control* Ada provides pragmas for you to specify the desired elaboration order. .. _Controlling_the_Elaboration_Order_in_Ada: Controlling the Elaboration Order in Ada ======================================== Ada provides several idioms and pragmas to aid you in specifying your desired elaboration order and avoiding ABE problems. * *Packages without a body* A library package that does not require a completing body does not suffer from ABE problems. .. code-block:: ada package Pack is generic type Element is private; package Containers is type Element_Array is array (1 .. 10) of Element; end Containers; end Pack; In the example above, package ``Pack`` does not require a body because it does not contain any constructs which require completion in a body. As a result, generic ``Pack.Containers`` can be instantiated without encountering any ABE problems. .. index:: pragma Pure * *pragma Pure* Pragma ``Pure`` places sufficient restrictions on a unit to guarantee that no scenario within the unit can result in an ABE problem. .. index:: pragma Preelaborate * *pragma Preelaborate* Pragma ``Preelaborate`` is slightly less restrictive than pragma ``Pure``, but still strong enough to prevent ABE problems within a unit. .. index:: pragma Elaborate_Body * *pragma Elaborate_Body* Pragma ``Elaborate_Body`` requires that the body of a unit is elaborated immediately after its spec. This restriction guarantees that no client scenario can invoke a server target before the target body has been elaborated because the spec and body are effectively "glued" together. .. code-block:: ada package Server is pragma Elaborate_Body; function Func return Integer; end Server; .. code-block:: ada package body Server is function Func return Integer is begin ... end Func; end Server; .. code-block:: ada with Server; package Client is Val : constant Integer := Server.Func; end Client; In the example above, pragma ``Elaborate_Body`` guarantees the following elaboration order: :: spec of Server body of Server spec of Client because the spec of ``Server`` must be elaborated prior to ``Client`` by virtue of the |with| clause and the body of ``Server`` must be elaborated immediately after the spec of ``Server``. Removing pragma ``Elaborate_Body`` could result in the following incorrect elaboration order: :: spec of Server spec of Client body of Server where ``Client`` invokes ``Server.Func``, but the body of ``Server.Func`` has not been elaborated yet. The pragmas outlined above allow a server unit to guarantee safe elaboration use by client units. Thus it is a good rule to mark units as ``Pure`` or ``Preelaborate``, and if this is not possible, mark them as ``Elaborate_Body``. There are however situations where ``Pure``, ``Preelaborate``, and ``Elaborate_Body`` are not applicable. Ada provides another set of pragmas for use by client units to help ensure the elaboration safety of server units they depend on. .. index:: pragma Elaborate (Unit) * *pragma Elaborate (Unit)* You can place pragma ``Elaborate`` in the context clauses of a unit, after a |with| clause. It guarantees that both the spec and body of its argument will be elaborated prior to the unit with the pragma. Note that other unrelated units may be elaborated in between the spec and the body. .. code-block:: ada package Server is function Func return Integer; end Server; .. code-block:: ada package body Server is function Func return Integer is begin ... end Func; end Server; .. code-block:: ada with Server; pragma Elaborate (Server); package Client is Val : constant Integer := Server.Func; end Client; In the example above, pragma ``Elaborate`` guarantees the following elaboration order: :: spec of Server body of Server spec of Client Removing pragma ``Elaborate`` could result in the following incorrect elaboration order: :: spec of Server spec of Client body of Server where ``Client`` invokes ``Server.Func``, but the body of ``Server.Func`` has not been elaborated yet. .. index:: pragma Elaborate_All (Unit) * *pragma Elaborate_All (Unit)* You can place pragma ``Elaborate_All`` in the context clauses of a unit, after a |with| clause. It guarantees that both the spec and body of its argument will be elaborated prior to the unit with the pragma as well as all units |withed| by the spec and body of the argument, recursively. Note that other unrelated units may be elaborated in between the spec and the body. .. code-block:: ada package Math is function Factorial (Val : Natural) return Natural; end Math; .. code-block:: ada package body Math is function Factorial (Val : Natural) return Natural is begin ...; end Factorial; end Math; .. code-block:: ada package Computer is type Operation_Kind is (None, Op_Factorial); function Compute (Val : Natural; Op : Operation_Kind) return Natural; end Computer; .. code-block:: ada with Math; package body Computer is function Compute (Val : Natural; Op : Operation_Kind) return Natural is if Op = Op_Factorial then return Math.Factorial (Val); end if; return 0; end Compute; end Computer; .. code-block:: ada with Computer; pragma Elaborate_All (Computer); package Client is Val : constant Natural := Computer.Compute (123, Computer.Op_Factorial); end Client; In the example above, pragma ``Elaborate_All`` can result in the following elaboration order: :: spec of Math body of Math spec of Computer body of Computer spec of Client Note that there are several allowable suborders for the specs and bodies of ``Math`` and ``Computer``, but the point is that these specs and bodies will be elaborated prior to ``Client``. Removing pragma ``Elaborate_All`` could result in the following incorrect elaboration order: :: spec of Math spec of Computer body of Computer spec of Client body of Math where ``Client`` invokes ``Computer.Compute``, which in turn invokes ``Math.Factorial``, but the body of ``Math.Factorial`` has not been elaborated yet. All pragmas shown above can be summarized by the following rule: *If a client unit elaborates a server target directly or indirectly, then if the server unit requires a body and does not have pragma Pure, Preelaborate, or Elaborate_Body, then the client unit should have pragma Elaborate or Elaborate_All for the server unit.* If you do not follow the rule outlined above, a program may fall in one of the following ways: * *No elaboration order exists* In this case a compiler must diagnose the situation and refuse to build an executable program. * *One or more incorrect elaboration orders exist* In this case a compiler can build an executable program, but ``Program_Error`` will be raised when the program is run. * *Several elaboration orders exist, some correct, some incorrect* In this case, you have not controlled the elaboration order. As a result, a compiler may or may not pick one of the correct orders and the program may or may not raise ``Program_Error`` when it is run. This is the worst possible state because the program may fail on another compiler or even a different version of the same compiler. * *One or more correct orders exist* In this case a compiler can build an executable program and the program is run successfully. This state may be guaranteed by following the outlined rules or may be the result of good program architecture. Note that one additional advantage of using ``Elaborate`` and ``Elaborate_All`` is that the program continues to stay in the last state (one or more correct orders exist) even if maintenance changes the bodies of targets. .. _Controlling_the_Elaboration_Order_in_GNAT: Controlling the Elaboration Order in GNAT ========================================= In addition to Ada semantics and rules synthesized from them, GNAT offers three elaboration models to aid you in specifying the correct elaboration order and in diagnosing elaboration problems. .. index:: Dynamic elaboration model * *Dynamic elaboration model* This is the most permissive of the three elaboration models and emulates the behavior specified by the Ada Reference Manual. When the dynamic model is in effect, GNAT makes the following assumptions: - All code within all units in a partition is considered to be elaboration code. - Some of the invocations in elaboration code may not take place at run time due to conditional execution. GNAT performs extensive diagnostics on a unit-by-unit basis for all scenarios that invoke internal targets. In addition, GNAT generates run-time checks for all external targets and for all scenarios that may exhibit ABE problems. The elaboration order is obtained by honoring all |with| clauses, purity and preelaborability of units, and elaboration-control pragmas. The dynamic model attempts to take all invocations in elaboration code into account. If an invocation leads to a circularity, GNAT ignores the invocation based on the assumptions stated above. An order obtained using the dynamic model may fail an ABE check at run time when GNAT ignored an invocation. You enable the dynamic model with the compiler switch :switch:`-gnatE`. .. index:: Static elaboration model * *Static elaboration model* This is the middle ground of the three models. When the static model is in effect, GNAT makes the following assumptions: - Only code at the library level and in package body statements within all units in a partition is considered to be elaboration code. - All invocations in elaboration will take place at run time, regardless of conditional execution. GNAT performs extensive diagnostics on a unit-by-unit basis for all scenarios that invoke internal targets. In addition, GNAT generates run-time checks for all external targets and for all scenarios that may exhibit ABE problems. The elaboration order is obtained by honoring all |with| clauses, purity and preelaborability of units, presence of elaboration-control pragmas, and all invocations in elaboration code. An order obtained using the static model is guaranteed to be ABE problem-free, excluding dispatching calls and access-to-subprogram types. The static model is the default model in GNAT. .. index:: SPARK elaboration model * *SPARK elaboration model* This is the most conservative of the three models and enforces the SPARK rules of elaboration as defined in the SPARK Reference Manual, section 7.7. The SPARK model is in effect only when a scenario and a target reside in a region subject to ``SPARK_Mode On``, otherwise the dynamic or static model is in effect. The SPARK model is enabled with compiler switch :switch:`-gnatd.v`. .. index:: Legacy elaboration models * *Legacy elaboration models* In addition to the three elaboration models outlined above, GNAT provides the following legacy models: - `Legacy elaboration-checking model` available in pre-18.x versions of GNAT. You can enable this model with compiler switch :switch:`-gnatH`. - `Legacy elaboration-order model` available in pre-20.x versions of GNAT. You can enable this model with binder switch :switch:`-H`. .. index:: Relaxed elaboration mode You can relax the dynamic, legacy, and static models by specifying compiler switch :switch:`-gnatJ`, which makes them more permissive. Note that in this mode, GNAT may not diagnose certain elaboration issues or install run-time checks. .. _Mixing_Elaboration_Models: Mixing Elaboration Models ========================= You can mix units compiled with different elaboration models. However you must observe the following rules: * A client unit compiled with the dynamic model can only |with| a server unit that meets at least one of the following criteria: - The server unit is compiled with the dynamic model. - The server unit is a GNAT implementation unit from the ``Ada``, ``GNAT``, ``Interfaces``, or ``System`` hierarchies. - The server unit has pragma ``Pure`` or ``Preelaborate``. - The client unit has an explicit ``Elaborate_All`` pragma for the server unit. These rules ensure that elaboration checks are not omitted. If the rules are violated, the binder emits a warning: :: warning: "x.ads" has dynamic elaboration checks and with's warning: "y.ads" which has static elaboration checks You can suppress these warnings by specifying binder switch :switch:`-ws`. .. _ABE_Diagnostics: ABE Diagnostics =============== GNAT performs extensive diagnostics on a unit-by-unit basis for all scenarios that invoke internal targets, regardless of whether the dynamic, SPARK, or static model is in effect. Note that GNAT emits warnings rather than errors whenever it encounters an elaboration problem. This is because the elaboration model in effect may be too conservative or a particular scenario may not be invoked due to conditional execution. You can selectively suppress the warnings with ``pragma Warnings (Off)`` or globally with compiler switch :switch:`-gnatwL`. A *guaranteed ABE* arises when the body of a target is not elaborated early enough and causes *all* scenarios that directly invoke the target to fail. .. code-block:: ada package body Guaranteed_ABE is function ABE return Integer; Val : constant Integer := ABE; function ABE return Integer is begin ... end ABE; end Guaranteed_ABE; In the example above, the elaboration of ``Guaranteed_ABE``'s body elaborates the declaration of ``Val``. This invokes function ``ABE``, however the body of ``ABE`` has not been elaborated yet. GNAT emits the following diagnostic: :: 4. Val : constant Integer := ABE; | >>> warning: cannot call "ABE" before body seen >>> warning: Program_Error will be raised at run time A *conditional ABE* arises when the body of a target is not elaborated early enough and causes *some* scenarios that directly invoke the target to fail. .. code-block:: ada 1. package body Conditional_ABE is 2. procedure Force_Body is null; 3. 4. generic 5. with function Func return Integer; 6. package Gen is 7. Val : constant Integer := Func; 8. end Gen; 9. 10. function ABE return Integer; 11. 12. function Cause_ABE return Boolean is 13. package Inst is new Gen (ABE); 14. begin 15. ... 16. end Cause_ABE; 17. 18. Val : constant Boolean := Cause_ABE; 19. 20. function ABE return Integer is 21. begin 22. ... 23. end ABE; 24. 25. Safe : constant Boolean := Cause_ABE; 26. end Conditional_ABE; In the example above, the elaboration of package body ``Conditional_ABE`` elaborates the declaration of ``Val``. This invokes function ``Cause_ABE``, which instantiates generic unit ``Gen`` as ``Inst``. The elaboration of ``Inst`` invokes function ``ABE``, however the body of ``ABE`` has not been elaborated yet. GNAT emits the following diagnostic: :: 13. package Inst is new Gen (ABE); | >>> warning: in instantiation at line 7 >>> warning: cannot call "ABE" before body seen >>> warning: Program_Error may be raised at run time >>> warning: body of unit "Conditional_ABE" elaborated >>> warning: function "Cause_ABE" called at line 18 >>> warning: function "ABE" called at line 7, instance at line 13 Note that the same ABE problem does not occur with the elaboration of declaration ``Safe`` because the body of function ``ABE`` has already been elaborated at that point. .. _SPARK_Diagnostics: SPARK Diagnostics ================= GNAT enforces the SPARK rules of elaboration as defined in the SPARK Reference Manual section 7.7 when you specify compiler switch :switch:`-gnatd.v`. Note that GNAT emits errors whenever it encounters a violation of the SPARK rules. :: 1. with Server; 2. package body SPARK_Diagnostics with SPARK_Mode is 3. Val : constant Integer := Server.Func; | >>> call to "Func" during elaboration in SPARK >>> unit "SPARK_Diagnostics" requires pragma "Elaborate_All" for "Server" >>> body of unit "SPARK_Model" elaborated >>> function "Func" called at line 3 4. end SPARK_Diagnostics; .. _Elaboration_Circularities: Elaboration Circularities ========================= An **elaboration circularity** occurs whenever the elaboration of a set of units enters a deadlocked state, where each unit is waiting for another unit to be elaborated. This situation may be the result of improper use of |with| clauses, elaboration-control pragmas, or invocations in elaboration code. The following example exhibits an elaboration circularity. .. code-block:: ada with B; pragma Elaborate (B); package A is end A; .. code-block:: ada package B is procedure Force_Body; end B; .. code-block:: ada with C; package body B is procedure Force_Body is null; Elab : constant Integer := C.Func; end B; .. code-block:: ada package C is function Func return Integer; end C; .. code-block:: ada with A; package body C is function Func return Integer is begin ... end Func; end C; The binder emits the following diagnostic: :: error: Elaboration circularity detected info: info: Reason: info: info: unit "a (spec)" depends on its own elaboration info: info: Circularity: info: info: unit "a (spec)" has with clause and pragma Elaborate for unit "b (spec)" info: unit "b (body)" is in the closure of pragma Elaborate info: unit "b (body)" invokes a construct of unit "c (body)" at elaboration time info: unit "c (body)" has with clause for unit "a (spec)" info: info: Suggestions: info: info: remove pragma Elaborate for unit "b (body)" in unit "a (spec)" info: use the dynamic elaboration model (compiler switch -gnatE) The diagnostic consist of the following sections: * Reason This section provides a short explanation describing why the set of units could not be ordered. * Circularity This section enumerates the units comprising the deadlocked set, along with their interdependencies. * Suggestions This section enumerates various tactics for eliminating the circularity. .. _Resolving_Elaboration_Circularities: Resolving Elaboration Circularities =================================== The most desirable option from the point of view of long-term maintenance is to rearrange the program so that the elaboration problems are avoided. One useful technique is to place the elaboration code into separate child packages. Another is to move some of the initialization code to explicitly invoked subprograms, where the program controls the order of initialization explicitly. Although this is the most desirable option, it may be impractical and involve too much modification, especially in the case of complex legacy code. When faced with an elaboration circularity, you should also consider the tactics given in the suggestions section of the circularity diagnostic. Depending on the units involved in the circularity, their |with| clauses, purity, preelaborability, presence of elaboration-control pragmas and invocations at elaboration time, the binder may suggest one or more of the following tactics to eliminate the circularity: * Pragma Elaborate elimination :: remove pragma Elaborate for unit "..." in unit "..." The binder suggests this tactic when it has determined that: - pragma ``Elaborate`` prevents a set of units from being elaborated. - The removal of the pragma will not eliminate the semantic effects of the pragma. In other words, the argument of the pragma will still be elaborated prior to the unit containing the pragma. - The removal of the pragma will enable the successful ordering of the units. You should remove the pragma as advised and rebuild the program. * Pragma Elaborate_All elimination :: remove pragma Elaborate_All for unit "..." in unit "..." The binder suggests this tactic when it has determined that: - pragma ``Elaborate_All`` prevents a set of units from being elaborated. - The removal of the pragma will not eliminate the semantic effects of the pragma. In other words, the argument of the pragma along with its |with| closure will still be elaborated prior to the unit containing the pragma. - The removal of the pragma will enable the successful ordering of the units. You should remove the pragma as advised and rebuild the program. * Pragma Elaborate_All downgrade :: change pragma Elaborate_All for unit "..." to Elaborate in unit "..." The binder always suggests this tactic when it suggests the pragma ``Elaborate_All`` elimination tactic. It offers a different alternative of guaranteeing that the argument of the pragma will still be elaborated prior to the unit containing the pragma. You should update the pragma as advised and rebuild the program. * Pragma Elaborate_Body elimination :: remove pragma Elaborate_Body in unit "..." The binder suggests this tactic when it has determined that: - pragma ``Elaborate_Body`` prevents a set of units from being elaborated. - The removal of the pragma will enable the successful ordering of the units. Note that the binder cannot determine whether the pragma is required for other purposes, such as guaranteeing the initialization of a variable declared in the spec by elaboration code in the body. If the pragma is not required for another purpose, you should remove the pragma as advised and rebuild the program. * Use of pragma Restrictions :: use pragma Restrictions (No_Entry_Calls_In_Elaboration_Code) The binder suggests this tactic when it has determined that a task activation at elaboration time: - Prevents a set of units from being elaborated. Note that the binder cannot determine with certainty whether the task will block at elaboration time. The programmer should create a configuration file, place the pragma within, update the general compilation arguments, and rebuild the program. * Use of dynamic elaboration model :: use the dynamic elaboration model (compiler switch -gnatE) This tactic is suggested when the binder has determined that an invocation at elaboration time: - Prevents a set of units from being elaborated. - The use of the dynamic model will enable the successful ordering of the units. You have two options: - Determine the units involved in the invocation using the detailed invocation information and add compiler switch :switch:`-gnatE` to the compilation arguments of those units only. This approach will yield safer elaboration orders compared to the other option because it will minimize the opportunities presented to the dynamic model for ignoring invocations. - Add compiler switch :switch:`-gnatE` to the global compilation arguments. * Use of detailed invocation information :: use detailed invocation information (compiler switch -gnatd_F) The binder always suggests this tactic when it suggests use of the dynamic model tactic. It causes the circularity section of the circularity diagnostic to describe the flow of elaboration code from a unit to a unit, enumerating all such paths in the process. You should analyze this information to determine which units should be compiled with the dynamic model. * Forced-dependency elimination :: remove the dependency of unit "..." on unit "..." from the argument of switch -f The binder suggests this tactic when it has determined that a dependency present in the forced-elaboration-order file indicated by binder switch :switch:`-f`: - Prevents a set of units from being elaborated. - The removal of the dependency will enable the successful ordering of the units. You should edit the forced-elaboration-order file, remove the dependency, and rebind the program. * All forced-dependency elimination :: remove switch -f The binder suggests this tactic when editing the forced-elaboration-order file is not an option. You should remove binder switch :switch:`-f` from the binder arguments and rebind. * Multiple-circularities diagnostic :: diagnose all circularities (binder switch -d_C) By default, the binder only diagnoses the highest-precedence circularity. If the program contains multiple circularities, the binder will suggest the use of binder switch :switch:`-d_C` in order to obtain the diagnostics for each circularity. You should add binder switch :switch:`-d_C` to the binder arguments and rebind. If none of the tactics suggested by the binder eliminate the elaboration circularity, you should consider using one of the legacy elaboration models, in the following order: * Use the pre-20.x legacy elaboration-order model, with binder switch :switch:`-H`. * Use both pre-18.x and pre-20.x legacy elaboration models, with compiler switch :switch:`-gnatH` and binder switch :switch:`-H`. * Use the relaxed static-elaboration model, with compiler switches :switch:`-gnatH` :switch:`-gnatJ` and binder switch :switch:`-H`. * Use the relaxed dynamic-elaboration model, with compiler switches :switch:`-gnatH` :switch:`-gnatJ` :switch:`-gnatE` and binder switch :switch:`-H`. .. _Elaboration_Related_Compiler_Switches: Elaboration-related Compiler Switches ===================================== GNAT has several switches that affect the elaboration model and consequently the elaboration order chosen by the binder. .. index:: -gnatE (gnat) :switch:`-gnatE` Dynamic elaboration checking mode enabled When this switch is in effect, GNAT activates the dynamic model. .. index:: -gnatel (gnat) :switch:`-gnatel` Turn on informational messages on generated Elaborate[_All] pragmas This switch is only applicable to the pre-20.x legacy elaboration models. The post-20.x elaboration model no longer relies on implicitly generated ``Elaborate`` and ``Elaborate_All`` pragmas to order units. When this switch is in effect, GNAT will emit the following supplementary information depending on the elaboration model in effect. - *Dynamic model* GNAT will indicate missing ``Elaborate`` and ``Elaborate_All`` pragmas for all library-level scenarios within the partition. - *Static model* GNAT will indicate all scenarios invoked during elaboration. In addition, it will provide detailed traceback when an implicit ``Elaborate`` or ``Elaborate_All`` pragma is generated. - *SPARK model* GNAT will indicate how an elaboration requirement is met by the context of a unit. This diagnostic requires compiler switch :switch:`-gnatd.v`. :: 1. with Server; pragma Elaborate_All (Server); 2. package Client with SPARK_Mode is 3. Val : constant Integer := Server.Func; | >>> info: call to "Func" during elaboration in SPARK >>> info: "Elaborate_All" requirement for unit "Server" met by pragma at line 1 4. end Client; .. index:: -gnatH (gnat) :switch:`-gnatH` Legacy elaboration checking mode enabled When this switch is in effect, GNAT will utilize the pre-18.x elaboration model. .. index:: -gnatJ (gnat) :switch:`-gnatJ` Relaxed elaboration checking mode enabled When this switch is in effect, GNAT will not process certain scenarios, resulting in a more permissive elaboration model. Note that this may eliminate some diagnostics and run-time checks. .. index:: -gnatw.f (gnat) :switch:`-gnatw.f` Turn on warnings for suspicious Subp'Access When this switch is in effect, GNAT will treat ``'Access`` of an entry, operator, or subprogram as a potential call to the target and issue warnings: :: 1. package body Attribute_Call is 2. function Func return Integer; 3. type Func_Ptr is access function return Integer; 4. 5. Ptr : constant Func_Ptr := Func'Access; | >>> warning: "Access" attribute of "Func" before body seen >>> warning: possible Program_Error on later references >>> warning: body of unit "Attribute_Call" elaborated >>> warning: "Access" of "Func" taken at line 5 6. 7. function Func return Integer is 8. begin 9. ... 10. end Func; 11. end Attribute_Call; In the example above, the elaboration of declaration ``Ptr`` is assigned ``Func'Access`` before the body of ``Func`` has been elaborated. .. index:: -gnatwl (gnat) :switch:`-gnatwl` Turn on warnings for elaboration problems When this switch is in effect, GNAT emits diagnostics in the form of warnings concerning various elaboration problems. The warnings are enabled by default. The switch is provided in case all warnings are suppressed, but elaboration warnings are still desired. :switch:`-gnatwL` Turn off warnings for elaboration problems When this switch is in effect, GNAT no longer emits any diagnostics in the form of warnings. Selective suppression of elaboration problems is possible using ``pragma Warnings (Off)``. :: 1. package body Selective_Suppression is 2. function ABE return Integer; 3. 4. Val_1 : constant Integer := ABE; | >>> warning: cannot call "ABE" before body seen >>> warning: Program_Error will be raised at run time 5. 6. pragma Warnings (Off); 7. Val_2 : constant Integer := ABE; 8. pragma Warnings (On); 9. 10. function ABE return Integer is 11. begin 12. ... 13. end ABE; 14. end Selective_Suppression; Note that suppressing elaboration warnings does not eliminate run-time checks. The example above will still fail at run time with an ABE. .. _Summary_of_Procedures_for_Elaboration_Control: Summary of Procedures for Elaboration Control ============================================= You should first compile the program with the default options, using none of the binder or compiler switches. If the binder succeeds in finding an elaboration order, then apart from possible cases involving dispatching calls and access-to-subprogram types, the program is free of elaboration errors. If it is important for the program to be portable to compilers other than GNAT, you should use compiler switch :switch:`-gnatel` and consider the messages about missing or implicitly created ``Elaborate`` and ``Elaborate_All`` pragmas. If the binder reports an elaboration circularity, you have several options: * Ensure that elaboration warnings are enabled. This allows the static model to output trace information of elaboration issues. The trace information could shed light on previously unforeseen dependencies as well as their origins. You enable elaboration warnings with compiler switch :switch:`-gnatwl`. * Cosider the tactics given in the suggestions section of the circularity diagnostic. * If none of the steps outlined above resolve the circularity, use a more permissive elaboration model, in the following order: - Use the pre-20.x legacy elaboration-order model, with binder switch :switch:`-H`. - Use both pre-18.x and pre-20.x legacy elaboration models, with compiler switch :switch:`-gnatH` and binder switch :switch:`-H`. - Use the relaxed static elaboration model, with compiler switches :switch:`-gnatH` :switch:`-gnatJ` and binder switch :switch:`-H`. - Use the relaxed dynamic elaboration model, with compiler switches :switch:`-gnatH` :switch:`-gnatJ` :switch:`-gnatE` and binder switch :switch:`-H`. .. _Inspecting_the_Chosen_Elaboration_Order: Inspecting the Chosen Elaboration Order ======================================= To see the elaboration order chosen by the binder, inspect the contents of file `b~xxx.adb`. On certain targets, this file appears as `b_xxx.adb`. The elaboration order appears as a sequence of calls to ``Elab_Body`` and ``Elab_Spec``, interspersed with assignments to `Exxx` which indicates that a particular unit is elaborated. For example: :: System.Soft_Links'Elab_Body; E14 := True; System.Secondary_Stack'Elab_Body; E18 := True; System.Exception_Table'Elab_Body; E24 := True; Ada.Io_Exceptions'Elab_Spec; E67 := True; Ada.Tags'Elab_Spec; Ada.Streams'Elab_Spec; E43 := True; Interfaces.C'Elab_Spec; E69 := True; System.Finalization_Root'Elab_Spec; E60 := True; System.Os_Lib'Elab_Body; E71 := True; System.Finalization_Implementation'Elab_Spec; System.Finalization_Implementation'Elab_Body; E62 := True; Ada.Finalization'Elab_Spec; E58 := True; Ada.Finalization.List_Controller'Elab_Spec; E76 := True; System.File_Control_Block'Elab_Spec; E74 := True; System.File_Io'Elab_Body; E56 := True; Ada.Tags'Elab_Body; E45 := True; Ada.Text_Io'Elab_Spec; Ada.Text_Io'Elab_Body; E07 := True; Note also binder switch :switch:`-l`, which outputs the chosen elaboration order and provides a more readable form of the above: :: ada (spec) interfaces (spec) system (spec) system.case_util (spec) system.case_util (body) system.concat_2 (spec) system.concat_2 (body) system.concat_3 (spec) system.concat_3 (body) system.htable (spec) system.parameters (spec) system.parameters (body) system.crtl (spec) interfaces.c_streams (spec) interfaces.c_streams (body) system.restrictions (spec) system.restrictions (body) system.standard_library (spec) system.exceptions (spec) system.exceptions (body) system.storage_elements (spec) system.storage_elements (body) system.secondary_stack (spec) system.stack_checking (spec) system.stack_checking (body) system.string_hash (spec) system.string_hash (body) system.htable (body) system.strings (spec) system.strings (body) system.traceback (spec) system.traceback (body) system.traceback_entries (spec) system.traceback_entries (body) ada.exceptions (spec) ada.exceptions.last_chance_handler (spec) system.soft_links (spec) system.soft_links (body) ada.exceptions.last_chance_handler (body) system.secondary_stack (body) system.exception_table (spec) system.exception_table (body) ada.io_exceptions (spec) ada.tags (spec) ada.streams (spec) interfaces.c (spec) interfaces.c (body) system.finalization_root (spec) system.finalization_root (body) system.memory (spec) system.memory (body) system.standard_library (body) system.os_lib (spec) system.os_lib (body) system.unsigned_types (spec) system.stream_attributes (spec) system.stream_attributes (body) system.finalization_implementation (spec) system.finalization_implementation (body) ada.finalization (spec) ada.finalization (body) ada.finalization.list_controller (spec) ada.finalization.list_controller (body) system.file_control_block (spec) system.file_io (spec) system.file_io (body) system.val_uns (spec) system.val_util (spec) system.val_util (body) system.val_uns (body) system.wch_con (spec) system.wch_con (body) system.wch_cnv (spec) system.wch_jis (spec) system.wch_jis (body) system.wch_cnv (body) system.wch_stw (spec) system.wch_stw (body) ada.tags (body) ada.exceptions (body) ada.text_io (spec) ada.text_io (body) text_io (spec) gdbstr (body) --- ### Gcc Ada Doc Gnat Ugn Example Of Binder Output (`gcc/ada/doc/gnat_ugn/example_of_binder_output.rst`) .. _Example_of_Binder_Output_File: ***************************** Example of Binder Output File ***************************** .. index:: Binder output (example) This Appendix displays the source code for the output file generated by ``gnatbind`` for a simple 'Hello World' program. Comments have been added for clarification purposes. .. code-block:: ada -- The package is called Ada_Main unless this name is actually used -- as a unit name in the partition, in which case some other unique -- name is used. pragma Ada_95; with System; package ada_main is pragma Warnings (Off); -- The main program saves the parameters (argument count, -- argument values, environment pointer) in global variables -- for later access by other units including Ada.Command_Line. gnat_argc : Integer; gnat_argv : System.Address; gnat_envp : System.Address; -- The actual variables are stored in a library routine. This -- is useful for some shared library situations, where there -- are problems if variables are not in the library. pragma Import (C, gnat_argc); pragma Import (C, gnat_argv); pragma Import (C, gnat_envp); -- The exit status is similarly an external location gnat_exit_status : Integer; pragma Import (C, gnat_exit_status); GNAT_Version : constant String := "GNAT Version: Pro 7.4.0w (20141119-49)" & ASCII.NUL; pragma Export (C, GNAT_Version, "__gnat_version"); Ada_Main_Program_Name : constant String := "_ada_hello" & ASCII.NUL; pragma Export (C, Ada_Main_Program_Name, "__gnat_ada_main_program_name"); -- This is the generated adainit routine that performs -- initialization at the start of execution. In the case -- where Ada is the main program, this main program makes -- a call to adainit at program startup. procedure adainit; pragma Export (C, adainit, "adainit"); -- This is the generated adafinal routine that performs -- finalization at the end of execution. In the case where -- Ada is the main program, this main program makes a call -- to adafinal at program termination. procedure adafinal; pragma Export (C, adafinal, "adafinal"); -- This routine is called at the start of execution. It is -- a dummy routine that is used by the debugger to breakpoint -- at the start of execution. -- This is the actual generated main program (it would be -- suppressed if the no main program switch were used). As -- required by standard system conventions, this program has -- the external name main. function main (argc : Integer; argv : System.Address; envp : System.Address) return Integer; pragma Export (C, main, "main"); -- The following set of constants give the version -- identification values for every unit in the bound -- partition. This identification is computed from all -- dependent semantic units, and corresponds to the -- string that would be returned by use of the -- Body_Version or Version attributes. -- The following Export pragmas export the version numbers -- with symbolic names ending in B (for body) or S -- (for spec) so that they can be located in a link. The -- information provided here is sufficient to track down -- the exact versions of units used in a given build. type Version_32 is mod 2 ** 32; u00001 : constant Version_32 := 16#8ad6e54a#; pragma Export (C, u00001, "helloB"); u00002 : constant Version_32 := 16#fbff4c67#; pragma Export (C, u00002, "system__standard_libraryB"); u00003 : constant Version_32 := 16#1ec6fd90#; pragma Export (C, u00003, "system__standard_libraryS"); u00004 : constant Version_32 := 16#3ffc8e18#; pragma Export (C, u00004, "adaS"); u00005 : constant Version_32 := 16#28f088c2#; pragma Export (C, u00005, "ada__text_ioB"); u00006 : constant Version_32 := 16#f372c8ac#; pragma Export (C, u00006, "ada__text_ioS"); u00007 : constant Version_32 := 16#2c143749#; pragma Export (C, u00007, "ada__exceptionsB"); u00008 : constant Version_32 := 16#f4f0cce8#; pragma Export (C, u00008, "ada__exceptionsS"); u00009 : constant Version_32 := 16#a46739c0#; pragma Export (C, u00009, "ada__exceptions__last_chance_handlerB"); u00010 : constant Version_32 := 16#3aac8c92#; pragma Export (C, u00010, "ada__exceptions__last_chance_handlerS"); u00011 : constant Version_32 := 16#1d274481#; pragma Export (C, u00011, "systemS"); u00012 : constant Version_32 := 16#a207fefe#; pragma Export (C, u00012, "system__soft_linksB"); u00013 : constant Version_32 := 16#467d9556#; pragma Export (C, u00013, "system__soft_linksS"); u00014 : constant Version_32 := 16#b01dad17#; pragma Export (C, u00014, "system__parametersB"); u00015 : constant Version_32 := 16#630d49fe#; pragma Export (C, u00015, "system__parametersS"); u00016 : constant Version_32 := 16#b19b6653#; pragma Export (C, u00016, "system__secondary_stackB"); u00017 : constant Version_32 := 16#b6468be8#; pragma Export (C, u00017, "system__secondary_stackS"); u00018 : constant Version_32 := 16#39a03df9#; pragma Export (C, u00018, "system__storage_elementsB"); u00019 : constant Version_32 := 16#30e40e85#; pragma Export (C, u00019, "system__storage_elementsS"); u00020 : constant Version_32 := 16#41837d1e#; pragma Export (C, u00020, "system__stack_checkingB"); u00021 : constant Version_32 := 16#93982f69#; pragma Export (C, u00021, "system__stack_checkingS"); u00022 : constant Version_32 := 16#393398c1#; pragma Export (C, u00022, "system__exception_tableB"); u00023 : constant Version_32 := 16#b33e2294#; pragma Export (C, u00023, "system__exception_tableS"); u00024 : constant Version_32 := 16#ce4af020#; pragma Export (C, u00024, "system__exceptionsB"); u00025 : constant Version_32 := 16#75442977#; pragma Export (C, u00025, "system__exceptionsS"); u00026 : constant Version_32 := 16#37d758f1#; pragma Export (C, u00026, "system__exceptions__machineS"); u00027 : constant Version_32 := 16#b895431d#; pragma Export (C, u00027, "system__exceptions_debugB"); u00028 : constant Version_32 := 16#aec55d3f#; pragma Export (C, u00028, "system__exceptions_debugS"); u00029 : constant Version_32 := 16#570325c8#; pragma Export (C, u00029, "system__img_intB"); u00030 : constant Version_32 := 16#1ffca443#; pragma Export (C, u00030, "system__img_intS"); u00031 : constant Version_32 := 16#b98c3e16#; pragma Export (C, u00031, "system__tracebackB"); u00032 : constant Version_32 := 16#831a9d5a#; pragma Export (C, u00032, "system__tracebackS"); u00033 : constant Version_32 := 16#9ed49525#; pragma Export (C, u00033, "system__traceback_entriesB"); u00034 : constant Version_32 := 16#1d7cb2f1#; pragma Export (C, u00034, "system__traceback_entriesS"); u00035 : constant Version_32 := 16#8c33a517#; pragma Export (C, u00035, "system__wch_conB"); u00036 : constant Version_32 := 16#065a6653#; pragma Export (C, u00036, "system__wch_conS"); u00037 : constant Version_32 := 16#9721e840#; pragma Export (C, u00037, "system__wch_stwB"); u00038 : constant Version_32 := 16#2b4b4a52#; pragma Export (C, u00038, "system__wch_stwS"); u00039 : constant Version_32 := 16#92b797cb#; pragma Export (C, u00039, "system__wch_cnvB"); u00040 : constant Version_32 := 16#09eddca0#; pragma Export (C, u00040, "system__wch_cnvS"); u00041 : constant Version_32 := 16#6033a23f#; pragma Export (C, u00041, "interfacesS"); u00042 : constant Version_32 := 16#ece6fdb6#; pragma Export (C, u00042, "system__wch_jisB"); u00043 : constant Version_32 := 16#899dc581#; pragma Export (C, u00043, "system__wch_jisS"); u00044 : constant Version_32 := 16#10558b11#; pragma Export (C, u00044, "ada__streamsB"); u00045 : constant Version_32 := 16#2e6701ab#; pragma Export (C, u00045, "ada__streamsS"); u00046 : constant Version_32 := 16#db5c917c#; pragma Export (C, u00046, "ada__io_exceptionsS"); u00047 : constant Version_32 := 16#12c8cd7d#; pragma Export (C, u00047, "ada__tagsB"); u00048 : constant Version_32 := 16#ce72c228#; pragma Export (C, u00048, "ada__tagsS"); u00049 : constant Version_32 := 16#c3335bfd#; pragma Export (C, u00049, "system__htableB"); u00050 : constant Version_32 := 16#99e5f76b#; pragma Export (C, u00050, "system__htableS"); u00051 : constant Version_32 := 16#089f5cd0#; pragma Export (C, u00051, "system__string_hashB"); u00052 : constant Version_32 := 16#3bbb9c15#; pragma Export (C, u00052, "system__string_hashS"); u00053 : constant Version_32 := 16#807fe041#; pragma Export (C, u00053, "system__unsigned_typesS"); u00054 : constant Version_32 := 16#d27be59e#; pragma Export (C, u00054, "system__val_lluB"); u00055 : constant Version_32 := 16#fa8db733#; pragma Export (C, u00055, "system__val_lluS"); u00056 : constant Version_32 := 16#27b600b2#; pragma Export (C, u00056, "system__val_utilB"); u00057 : constant Version_32 := 16#b187f27f#; pragma Export (C, u00057, "system__val_utilS"); u00058 : constant Version_32 := 16#d1060688#; pragma Export (C, u00058, "system__case_utilB"); u00059 : constant Version_32 := 16#392e2d56#; pragma Export (C, u00059, "system__case_utilS"); u00060 : constant Version_32 := 16#84a27f0d#; pragma Export (C, u00060, "interfaces__c_streamsB"); u00061 : constant Version_32 := 16#8bb5f2c0#; pragma Export (C, u00061, "interfaces__c_streamsS"); u00062 : constant Version_32 := 16#6db6928f#; pragma Export (C, u00062, "system__crtlS"); u00063 : constant Version_32 := 16#4e6a342b#; pragma Export (C, u00063, "system__file_ioB"); u00064 : constant Version_32 := 16#ba56a5e4#; pragma Export (C, u00064, "system__file_ioS"); u00065 : constant Version_32 := 16#b7ab275c#; pragma Export (C, u00065, "ada__finalizationB"); u00066 : constant Version_32 := 16#19f764ca#; pragma Export (C, u00066, "ada__finalizationS"); u00067 : constant Version_32 := 16#95817ed8#; pragma Export (C, u00067, "system__finalization_rootB"); u00068 : constant Version_32 := 16#52d53711#; pragma Export (C, u00068, "system__finalization_rootS"); u00069 : constant Version_32 := 16#769e25e6#; pragma Export (C, u00069, "interfaces__cB"); u00070 : constant Version_32 := 16#4a38bedb#; pragma Export (C, u00070, "interfaces__cS"); u00071 : constant Version_32 := 16#07e6ee66#; pragma Export (C, u00071, "system__os_libB"); u00072 : constant Version_32 := 16#d7b69782#; pragma Export (C, u00072, "system__os_libS"); u00073 : constant Version_32 := 16#1a817b8e#; pragma Export (C, u00073, "system__stringsB"); u00074 : constant Version_32 := 16#639855e7#; pragma Export (C, u00074, "system__stringsS"); u00075 : constant Version_32 := 16#e0b8de29#; pragma Export (C, u00075, "system__file_control_blockS"); u00076 : constant Version_32 := 16#b5b2aca1#; pragma Export (C, u00076, "system__finalization_mastersB"); u00077 : constant Version_32 := 16#69316dc1#; pragma Export (C, u00077, "system__finalization_mastersS"); u00078 : constant Version_32 := 16#57a37a42#; pragma Export (C, u00078, "system__address_imageB"); u00079 : constant Version_32 := 16#bccbd9bb#; pragma Export (C, u00079, "system__address_imageS"); u00080 : constant Version_32 := 16#7268f812#; pragma Export (C, u00080, "system__img_boolB"); u00081 : constant Version_32 := 16#e8fe356a#; pragma Export (C, u00081, "system__img_boolS"); u00082 : constant Version_32 := 16#d7aac20c#; pragma Export (C, u00082, "system__ioB"); u00083 : constant Version_32 := 16#8365b3ce#; pragma Export (C, u00083, "system__ioS"); u00084 : constant Version_32 := 16#6d4d969a#; pragma Export (C, u00084, "system__storage_poolsB"); u00085 : constant Version_32 := 16#e87cc305#; pragma Export (C, u00085, "system__storage_poolsS"); u00086 : constant Version_32 := 16#e34550ca#; pragma Export (C, u00086, "system__pool_globalB"); u00087 : constant Version_32 := 16#c88d2d16#; pragma Export (C, u00087, "system__pool_globalS"); u00088 : constant Version_32 := 16#9d39c675#; pragma Export (C, u00088, "system__memoryB"); u00089 : constant Version_32 := 16#445a22b5#; pragma Export (C, u00089, "system__memoryS"); u00090 : constant Version_32 := 16#6a859064#; pragma Export (C, u00090, "system__storage_pools__subpoolsB"); u00091 : constant Version_32 := 16#e3b008dc#; pragma Export (C, u00091, "system__storage_pools__subpoolsS"); u00092 : constant Version_32 := 16#63f11652#; pragma Export (C, u00092, "system__storage_pools__subpools__finalizationB"); u00093 : constant Version_32 := 16#fe2f4b3a#; pragma Export (C, u00093, "system__storage_pools__subpools__finalizationS"); -- BEGIN ELABORATION ORDER -- ada%s -- interfaces%s -- system%s -- system.case_util%s -- system.case_util%b -- system.htable%s -- system.img_bool%s -- system.img_bool%b -- system.img_int%s -- system.img_int%b -- system.io%s -- system.io%b -- system.parameters%s -- system.parameters%b -- system.crtl%s -- interfaces.c_streams%s -- interfaces.c_streams%b -- system.standard_library%s -- system.exceptions_debug%s -- system.exceptions_debug%b -- system.storage_elements%s -- system.storage_elements%b -- system.stack_checking%s -- system.stack_checking%b -- system.string_hash%s -- system.string_hash%b -- system.htable%b -- system.strings%s -- system.strings%b -- system.os_lib%s -- system.traceback_entries%s -- system.traceback_entries%b -- ada.exceptions%s -- system.soft_links%s -- system.unsigned_types%s -- system.val_llu%s -- system.val_util%s -- system.val_util%b -- system.val_llu%b -- system.wch_con%s -- system.wch_con%b -- system.wch_cnv%s -- system.wch_jis%s -- system.wch_jis%b -- system.wch_cnv%b -- system.wch_stw%s -- system.wch_stw%b -- ada.exceptions.last_chance_handler%s -- ada.exceptions.last_chance_handler%b -- system.address_image%s -- system.exception_table%s -- system.exception_table%b -- ada.io_exceptions%s -- ada.tags%s -- ada.streams%s -- ada.streams%b -- interfaces.c%s -- system.exceptions%s -- system.exceptions%b -- system.exceptions.machine%s -- system.finalization_root%s -- system.finalization_root%b -- ada.finalization%s -- ada.finalization%b -- system.storage_pools%s -- system.storage_pools%b -- system.finalization_masters%s -- system.storage_pools.subpools%s -- system.storage_pools.subpools.finalization%s -- system.storage_pools.subpools.finalization%b -- system.memory%s -- system.memory%b -- system.standard_library%b -- system.pool_global%s -- system.pool_global%b -- system.file_control_block%s -- system.file_io%s -- system.secondary_stack%s -- system.file_io%b -- system.storage_pools.subpools%b -- system.finalization_masters%b -- interfaces.c%b -- ada.tags%b -- system.soft_links%b -- system.os_lib%b -- system.secondary_stack%b -- system.address_image%b -- system.traceback%s -- ada.exceptions%b -- system.traceback%b -- ada.text_io%s -- ada.text_io%b -- hello%b -- END ELABORATION ORDER end ada_main; .. code-block:: ada pragma Ada_95; -- The following source file name pragmas allow the generated file -- names to be unique for different main programs. They are needed -- since the package name will always be Ada_Main. pragma Source_File_Name (ada_main, Spec_File_Name => "b~hello.ads"); pragma Source_File_Name (ada_main, Body_File_Name => "b~hello.adb"); pragma Suppress (Overflow_Check); with Ada.Exceptions; -- Generated package body for Ada_Main starts here package body ada_main is pragma Warnings (Off); -- These values are reference counters associated with units that have -- been elaborated. They are used to avoid elaborating the -- same unit twice. E72 : Short_Integer; pragma Import (Ada, E72, "system__os_lib_E"); E13 : Short_Integer; pragma Import (Ada, E13, "system__soft_links_E"); E23 : Short_Integer; pragma Import (Ada, E23, "system__exception_table_E"); E46 : Short_Integer; pragma Import (Ada, E46, "ada__io_exceptions_E"); E48 : Short_Integer; pragma Import (Ada, E48, "ada__tags_E"); E45 : Short_Integer; pragma Import (Ada, E45, "ada__streams_E"); E70 : Short_Integer; pragma Import (Ada, E70, "interfaces__c_E"); E25 : Short_Integer; pragma Import (Ada, E25, "system__exceptions_E"); E68 : Short_Integer; pragma Import (Ada, E68, "system__finalization_root_E"); E66 : Short_Integer; pragma Import (Ada, E66, "ada__finalization_E"); E85 : Short_Integer; pragma Import (Ada, E85, "system__storage_pools_E"); E77 : Short_Integer; pragma Import (Ada, E77, "system__finalization_masters_E"); E91 : Short_Integer; pragma Import (Ada, E91, "system__storage_pools__subpools_E"); E87 : Short_Integer; pragma Import (Ada, E87, "system__pool_global_E"); E75 : Short_Integer; pragma Import (Ada, E75, "system__file_control_block_E"); E64 : Short_Integer; pragma Import (Ada, E64, "system__file_io_E"); E17 : Short_Integer; pragma Import (Ada, E17, "system__secondary_stack_E"); E06 : Short_Integer; pragma Import (Ada, E06, "ada__text_io_E"); Local_Priority_Specific_Dispatching : constant String := ""; Local_Interrupt_States : constant String := ""; Is_Elaborated : Boolean := False; procedure finalize_library is begin E06 := E06 - 1; declare procedure F1; pragma Import (Ada, F1, "ada__text_io__finalize_spec"); begin F1; end; E77 := E77 - 1; E91 := E91 - 1; declare procedure F2; pragma Import (Ada, F2, "system__file_io__finalize_body"); begin E64 := E64 - 1; F2; end; declare procedure F3; pragma Import (Ada, F3, "system__file_control_block__finalize_spec"); begin E75 := E75 - 1; F3; end; E87 := E87 - 1; declare procedure F4; pragma Import (Ada, F4, "system__pool_global__finalize_spec"); begin F4; end; declare procedure F5; pragma Import (Ada, F5, "system__storage_pools__subpools__finalize_spec"); begin F5; end; declare procedure F6; pragma Import (Ada, F6, "system__finalization_masters__finalize_spec"); begin F6; end; declare procedure Reraise_Library_Exception_If_Any; pragma Import (Ada, Reraise_Library_Exception_If_Any, "__gnat_reraise_library_exception_if_any"); begin Reraise_Library_Exception_If_Any; end; end finalize_library; ------------- -- adainit -- ------------- procedure adainit is Main_Priority : Integer; pragma Import (C, Main_Priority, "__gl_main_priority"); Time_Slice_Value : Integer; pragma Import (C, Time_Slice_Value, "__gl_time_slice_val"); WC_Encoding : Character; pragma Import (C, WC_Encoding, "__gl_wc_encoding"); Locking_Policy : Character; pragma Import (C, Locking_Policy, "__gl_locking_policy"); Queuing_Policy : Character; pragma Import (C, Queuing_Policy, "__gl_queuing_policy"); Task_Dispatching_Policy : Character; pragma Import (C, Task_Dispatching_Policy, "__gl_task_dispatching_policy"); Priority_Specific_Dispatching : System.Address; pragma Import (C, Priority_Specific_Dispatching, "__gl_priority_specific_dispatching"); Num_Specific_Dispatching : Integer; pragma Import (C, Num_Specific_Dispatching, "__gl_num_specific_dispatching"); Main_CPU : Integer; pragma Import (C, Main_CPU, "__gl_main_cpu"); Interrupt_States : System.Address; pragma Import (C, Interrupt_States, "__gl_interrupt_states"); Num_Interrupt_States : Integer; pragma Import (C, Num_Interrupt_States, "__gl_num_interrupt_states"); Unreserve_All_Interrupts : Integer; pragma Import (C, Unreserve_All_Interrupts, "__gl_unreserve_all_interrupts"); Detect_Blocking : Integer; pragma Import (C, Detect_Blocking, "__gl_detect_blocking"); Default_Stack_Size : Integer; pragma Import (C, Default_Stack_Size, "__gl_default_stack_size"); Leap_Seconds_Support : Integer; pragma Import (C, Leap_Seconds_Support, "__gl_leap_seconds_support"); procedure Runtime_Initialize; pragma Import (C, Runtime_Initialize, "__gnat_runtime_initialize"); Finalize_Library_Objects : No_Param_Proc; pragma Import (C, Finalize_Library_Objects, "__gnat_finalize_library_objects"); -- Start of processing for adainit begin -- Record various information for this partition. The values -- are derived by the binder from information stored in the ali -- files by the compiler. if Is_Elaborated then return; end if; Is_Elaborated := True; Main_Priority := -1; Time_Slice_Value := -1; WC_Encoding := 'b'; Locking_Policy := ' '; Queuing_Policy := ' '; Task_Dispatching_Policy := ' '; Priority_Specific_Dispatching := Local_Priority_Specific_Dispatching'Address; Num_Specific_Dispatching := 0; Main_CPU := -1; Interrupt_States := Local_Interrupt_States'Address; Num_Interrupt_States := 0; Unreserve_All_Interrupts := 0; Detect_Blocking := 0; Default_Stack_Size := -1; Leap_Seconds_Support := 0; Runtime_Initialize; Finalize_Library_Objects := finalize_library'access; -- Now we have the elaboration calls for all units in the partition. -- The Elab_Spec and Elab_Body attributes generate references to the -- implicit elaboration procedures generated by the compiler for -- each unit that requires elaboration. Also increment a reference -- counter for each unit. System.Soft_Links'Elab_Spec; System.Exception_Table'Elab_Body; E23 := E23 + 1; Ada.Io_Exceptions'Elab_Spec; E46 := E46 + 1; Ada.Tags'Elab_Spec; Ada.Streams'Elab_Spec; E45 := E45 + 1; Interfaces.C'Elab_Spec; System.Exceptions'Elab_Spec; E25 := E25 + 1; System.Finalization_Root'Elab_Spec; E68 := E68 + 1; Ada.Finalization'Elab_Spec; E66 := E66 + 1; System.Storage_Pools'Elab_Spec; E85 := E85 + 1; System.Finalization_Masters'Elab_Spec; System.Storage_Pools.Subpools'Elab_Spec; System.Pool_Global'Elab_Spec; E87 := E87 + 1; System.File_Control_Block'Elab_Spec; E75 := E75 + 1; System.File_Io'Elab_Body; E64 := E64 + 1; E91 := E91 + 1; System.Finalization_Masters'Elab_Body; E77 := E77 + 1; E70 := E70 + 1; Ada.Tags'Elab_Body; E48 := E48 + 1; System.Soft_Links'Elab_Body; E13 := E13 + 1; System.Os_Lib'Elab_Body; E72 := E72 + 1; System.Secondary_Stack'Elab_Body; E17 := E17 + 1; Ada.Text_Io'Elab_Spec; Ada.Text_Io'Elab_Body; E06 := E06 + 1; end adainit; -------------- -- adafinal -- -------------- procedure adafinal is procedure s_stalib_adafinal; pragma Import (C, s_stalib_adafinal, "system__standard_library__adafinal"); procedure Runtime_Finalize; pragma Import (C, Runtime_Finalize, "__gnat_runtime_finalize"); begin if not Is_Elaborated then return; end if; Is_Elaborated := False; Runtime_Finalize; s_stalib_adafinal; end adafinal; -- We get to the main program of the partition by using -- pragma Import because if we try to 'with' the unit and -- call it in Ada style, not only do we waste time recompiling it, -- but we don't know the right switches (e.g.@: identifier -- character set) to be used to compile it. procedure Ada_Main_Program; pragma Import (Ada, Ada_Main_Program, "_ada_hello"); ---------- -- main -- ---------- -- main is actually a function, as in the ANSI C standard, -- defined to return the exit status. The three parameters -- are the argument count, argument values and environment -- pointer. function main (argc : Integer; argv : System.Address; envp : System.Address) return Integer is -- The initialize routine performs low level system -- initialization using a standard library routine which -- sets up signal handling and performs any other -- required setup. The routine can be found in file -- a-init.c. procedure initialize; pragma Import (C, initialize, "__gnat_initialize"); -- The finalize routine performs low level system -- finalization using a standard library routine. The -- routine is found in file a-final.c and in the standard -- distribution is a dummy routine that does nothing, so -- really this is a hook for special user finalization. procedure finalize; pragma Import (C, finalize, "__gnat_finalize"); -- The following is to initialize the SEH exceptions SEH : aliased array (1 .. 2) of Integer; Ensure_Reference : aliased System.Address := Ada_Main_Program_Name'Address; pragma Volatile (Ensure_Reference); -- Start of processing for main begin -- Save global variables gnat_argc := argc; gnat_argv := argv; gnat_envp := envp; -- Call low level system initialization Initialize (SEH'Address); -- Call our generated Ada initialization routine adainit; -- Now we call the main program of the partition Ada_Main_Program; -- Perform Ada finalization adafinal; -- Perform low level system finalization Finalize; -- Return the proper exit status return (gnat_exit_status); end; -- This section is entirely comments, so it has no effect on the -- compilation of the Ada_Main package. It provides the list of -- object files and linker options, as well as some standard -- libraries needed for the link. The gnatlink utility parses -- this b~hello.adb file to read these comment lines to generate -- the appropriate command line arguments for the call to the -- system linker. The BEGIN/END lines are used for sentinels for -- this parsing operation. -- The exact file names will of course depend on the environment, -- host/target and location of files on the host system. -- BEGIN Object file/option list -- ./hello.o -- -L./ -- -L/usr/local/gnat/lib/gcc-lib/i686-pc-linux-gnu/2.8.1/adalib/ -- /usr/local/gnat/lib/gcc-lib/i686-pc-linux-gnu/2.8.1/adalib/libgnat.a -- END Object file/option list end ada_main; The Ada code in the above example is exactly what is generated by the binder. We have added comments to more clearly indicate the function of each part of the generated ``Ada_Main`` package. The code is standard Ada in all respects, and can be processed by any tools that handle Ada. In particular, you can use the debugger in Ada mode to debug the generated ``Ada_Main`` package. For example, suppose that for reasons you don't understand, your program is crashing during elaboration of the body of ``Ada.Text_IO``. To locate this bug, you can place a breakpoint on the call: .. code-block:: ada Ada.Text_Io'Elab_Body; and trace the elaboration routine for this package to find out where the problem might be (more usually, of course, you would be debugging elaboration code in your own application). --- ### Gcc Ada Doc Gnat Ugn Getting Started With Gnat (`gcc/ada/doc/gnat_ugn/getting_started_with_gnat.rst`) .. role:: switch(samp) .. _Getting_Started_with_GNAT: ************************* Getting Started with GNAT ************************* This chapter describes how to use GNAT's command line interface to build executable Ada programs. On most platforms a visually oriented Integrated Development Environment is also available: GNAT Studio. GNAT Studio offers a graphical "look and feel", support for development in other programming languages, comprehensive browsing features, and many other capabilities. For information on GNAT Studio please refer to the :title:`GNAT Studio documentation`. .. _System_Requirements: System Requirements =================== Even though any machine can run the GNAT toolset and GNAT Studio IDE, to get the best experience we recommend using a machine with as many cores as possible, allowing individual compilations to run in parallel. A comfortable setup for a compiler server is a machine with 24 physical cores or more, with at least 48 GB of memory (2 GB per core). For a desktop machine, we recommend a minimum of 4 cores (8 is preferred), with at least 2GB per core (so 8 to 16GB). In addition, for running and smoothly navigating sources in GNAT Studio, we recommend at least 1.5 GB, plus 3 GB of RAM per million source lines of code. So we recommend at least 3 GB for 500K lines of code and 7.5 GB for 2 million lines of code. Using fast, local drives can make a significant difference in build and link times. You should avoid network drives such as NFS, SMB, or worse, configuration management filesystems (such as ClearCase dynamic views) as much as possible since these will produce very degraded performance (typically 2 to 3 times slower than on fast, local drives). If you cannot avoid using such slow drives for accessing source code, you should at least configure your project file so the result of the compilation is stored on a drive local to the machine performing the compilation. You can do this by setting the ``Object_Dir`` project file attribute. .. _Running_GNAT: Running GNAT ============ You need to take three steps to create an executable file from an Ada source file: * You must compile the source file(s). * You must bind the file(s) using the GNAT binder. * You must link all appropriate object files to produce an executable. You most commonly perform all three steps by using the ``gnatmake`` utility program. You pass it the name of the main program and it automatically performs the necessary compilation, binding, and linking steps. .. _Running_a_Simple_Ada_Program: Running a Simple Ada Program ============================ You may use any text editor to prepare an Ada program. (If you use Emacs, an optional Ada mode may be helpful in laying out the program.) The program text is a normal text file. We will assume in our initial example that you have used your editor to prepare the following standard format text file named :file:`hello.adb`: .. code-block:: ada with Ada.Text_IO; use Ada.Text_IO; procedure Hello is begin Put_Line ("Hello WORLD!"); end Hello; With the normal default file naming conventions, GNAT requires that each file contain a single compilation unit whose file name is the unit name with periods replaced by hyphens; the extension is :file:`ads` for a spec and :file:`adb` for a body. You can override this default file naming convention by use of the special pragma ``Source_File_Name`` (see :ref:`Using_Other_File_Names`). Alternatively, if you want to rename your files according to this default convention, which is probably more convenient if you will be using GNAT for all your compilations, then you use can use the ``gnatchop`` utility to generate correctly-named source files (see :ref:`Renaming_Files_with_gnatchop`). You can compile the program using the following command (``$`` is used as the command prompt in the examples in this document): .. code-block:: sh $ gcc -c hello.adb ``gcc`` is the command used to run the compiler. It is capable of compiling programs in several languages, including Ada and C. It assumes you have given it an Ada program if the file extension is either :file:`.ads` or :file:`.adb`, in which case it will call the GNAT compiler to compile the specified file. The :switch:`-c` switch is required. It tells ``gcc`` to only do a compilation. (For C programs, ``gcc`` can also do linking, but this capability is not used directly for Ada programs, so you must always specify the :switch:`-c`.) This compile command generates a file :file:`hello.o`, which is the object file corresponding to your Ada program. It also generates an 'Ada Library Information' file :file:`hello.ali`, which contains additional information used to check that an Ada program is consistent. To build an executable file, use either ``gnatmake`` or ``gprbuild`` with the name of the main file: these tools are builders that perform all the necessary build steps in the correct order. In particular, these builders automatically recompile any sources that have been modified since they were last compiled, as well as sources that depend on such modified sources, so that 'version skew' is avoided. .. index:: Version skew (avoided by ``gnatmake``) .. code-block:: sh $ gnatmake hello.adb The result is an executable program called :file:`hello`, which you can run by entering: .. code-block:: sh $ hello assuming that the current directory is on the search path for executable programs. and, if all has gone well, you will see:: Hello WORLD! appear in response to this command. .. _Running_a_Program_with_Multiple_Units: Running a Program with Multiple Units ===================================== Consider a slightly more complicated example with three files: a main program and the spec and body of a package: .. code-block:: ada package Greetings is procedure Hello; procedure Goodbye; end Greetings; with Ada.Text_IO; use Ada.Text_IO; package body Greetings is procedure Hello is begin Put_Line ("Hello WORLD!"); end Hello; procedure Goodbye is begin Put_Line ("Goodbye WORLD!"); end Goodbye; end Greetings; with Greetings; procedure Gmain is begin Greetings.Hello; Greetings.Goodbye; end Gmain; Following the one-unit-per-file rule, place this program in the following three separate files: *greetings.ads* spec of package ``Greetings`` *greetings.adb* body of package ``Greetings`` *gmain.adb* body of main program Note that there is no required order of compilation when using GNAT. In particular it is perfectly fine to compile the main program first. Also, it is not necessary to compile package specs in the case where there is an accompanying body; you only need compile the body. If you want to submit these files to the compiler for semantic checking and not code generation, use the :switch:`-gnatc` switch: .. code-block:: sh $ gcc -c greetings.ads -gnatc Although you can do the compilation in separate steps, in practice it's almost always more convenient to use the ``gnatmake`` or ``gprbuild`` tools: .. code-block:: sh $ gnatmake gmain.adb --- ### Gcc Ada Doc Gnat Ugn Gnat And Program Execution (`gcc/ada/doc/gnat_ugn/gnat_and_program_execution.rst`) .. role:: switch(samp) .. |with| replace:: *with* .. |withs| replace:: *with*\ s .. |withed| replace:: *with*\ ed .. |withing| replace:: *with*\ ing .. -- Example: A |withing| unit has a |with| clause, it |withs| a |withed| unit .. _GNAT_and_Program_Execution: ************************** GNAT and Program Execution ************************** This chapter covers several topics: * `Running and Debugging Ada Programs`_ * `Profiling`_ * `Improving Performance`_ * `Overflow Check Handling in GNAT`_ * `Performing Dimensionality Analysis in GNAT`_ * `Stack Related Facilities`_ * `Memory Management Issues`_ * `Sanitizers for Ada`_ .. _Running_and_Debugging_Ada_Programs: Running and Debugging Ada Programs ================================== .. index:: Debugging This section discusses how to debug Ada programs. The GNAT compiler handles an incorrect Ada program in three ways: * The illegality may be a violation of the static semantics of Ada. In that case, GNAT diagnoses the constructs in the program that are illegal. It's then a straightforward matter for you to modify those parts of the program. * The illegality may be a violation of the dynamic semantics of Ada. In that case the program compiles and executes, but may generate incorrect results or may terminate abnormally with some exception. * When presented with a program that contains convoluted errors, GNAT itself may terminate abnormally without providing full diagnostics on the incorrect user program. .. index:: Debugger .. index:: ! gdb .. _The_GNAT_Debugger_GDB: The GNAT Debugger GDB --------------------- ``GDB`` is a general purpose, platform-independent debugger that you can use to debug mixed-language programs, including compiled with ``gcc``, and in particular is capable of debugging Ada programs compiled with GNAT. The latest versions of ``GDB`` are Ada-aware and can handle complex Ada data structures. See :title:`Debugging with GDB`, for full details on the usage of ``GDB``, including a section on its usage on programs. That manual should be consulted for full details. The section that follows is a brief introduction to the philosophy and use of ``GDB``. When programs are compiled, the compiler optionally writes debugging information into the generated object file, including information on line numbers and on declared types and variables. This information is separate from the generated code. It makes the object files considerably larger, but it does not add to the size of the actual executable that is loaded into memory and has no impact on run-time performance. The generation of debug information is triggered by the use of the :switch:`-g` switch in the ``gcc`` or ``gnatmake`` command you used to perform the compilations. It is important to emphasize that it's a goal of GCC, and hence GNAT, that the use of this switch does not change the generated code. The compiler writes the debugging information in standard system formats that are used by many tools, including debuggers and profilers. The format of the information is typically designed to describe C types and semantics, but GNAT implements a translation scheme which allows full details about Ada types and variables to be encoded into these standard C formats. Details of this encoding scheme may be found in the file :file:`exp_dbug.ads` in the GNAT source distribution. However, the details of this encoding are, in most cases, of no interest to a user, since ``GDB`` automatically performs the necessary decoding. When a program is bound and linked, the debugging information is collected from the object files and stored in the executable image of the program. Again, this process significantly increases the size of the generated executable file, but does not increase the size of the executable program in memory. Furthermore, if this program is run in the normal manner, it runs exactly as if the debug information were not present and takes no more actual memory. However, if the program is run under control of ``GDB``, the debugger is activated. The image of the program is loaded, at which point it is ready to run. If you give a run command, the program runs exactly as it would have if ``GDB`` were not present. This is a crucial part of the ``GDB`` design philosophy: ``GDB`` is entirely non-intrusive until a breakpoint is encountered. If no breakpoint is ever hit, the program runs exactly as it would if no debugger were present. When a breakpoint is hit, ``GDB`` accesses the debugging information and can respond to user commands to inspect variables and more generally to report on the state of execution. .. _Running_GDB: Running GDB ----------- This section describes how to initiate the debugger. You can launch the debugger from a ``GNAT Studio`` menu or directly from the command line. The description below covers the latter use. You can use all the commands shown in the ``GNAT Studio`` debug console window, but there are usually more GUI-based ways to achieve the same effect. The command to run ``GDB`` is :: $ gdb program where ``program`` is the name of the executable file. This activates the debugger and results in a prompt for debugger commands. The simplest command is simply ``run``, which causes the program to run exactly as if the debugger were not present. The following section describes some of the additional commands that you can give to ``GDB``. .. _Introduction_to_GDB_Commands: Introduction to GDB Commands ---------------------------- ``GDB`` contains a large repertoire of commands. See :title:`Debugging with GDB` for extensive documentation on the use of these commands, together with examples of their use. Furthermore, the command *help* invoked from within GDB activates a simple help facility which summarizes the available commands and their options. In this section, we summarize a few of the most commonly used commands to give an idea of what ``GDB`` is about. You should create a simple program with debugging information and experiment with the use of these ``GDB`` commands on that program as you read through the following section. * :samp:`set args {arguments}` *arguments* is a list of arguments to be passed to the program on a subsequent run command, just as though the arguments had been entered on a normal invocation of the program. You do not need the ``set args`` command if the program does not require arguments. * :samp:`run` The ``run`` command causes execution of the program to start from the beginning. If the program is already running, that is to say if you are currently positioned at a breakpoint, then a prompt will ask for confirmation that you want to abandon the current execution and restart. You can also specify program arguments on this command and if you specify ``run`` with no arguments, the arguments used on the previous command will be used again. * :samp:`breakpoint {location}` This command sets a breakpoint, that is to say a point at which execution will halt and ``GDB`` will await further commands. *location* is either a line number within a file, which you specify in the format ``file:linenumber``, or the name of a subprogram. If you request a breakpoint be set on a subprogram that is overloaded, either a prompt will ask you to specify on which of those subprograms you want to breakpoint or a breakpoint will be set on all of them. If the program is run and execution encounters the breakpoint, the program stops and ``GDB`` signals that the breakpoint was encountered by printing the line of code before which the program is halted. * :samp:`catch exception {name}` This command causes the program execution to stop whenever exception ``name`` is raised. If you omit ``name``, execution is suspended when any exception is raised. * :samp:`print {expression}` This prints the value of the given expression. Most Ada expression formats are properly handled by ``GDB``, so the expression can contain function calls, variables, operators, and attribute references. * :samp:`continue` Continues execution following a breakpoint until the next breakpoint or the termination of the program. * :samp:`step` Executes a single line after a breakpoint. If the next statement is a subprogram call, execution continues into (the first statement of) the called subprogram. * :samp:`next` Executes a single line. If this line is a subprogram call, the program executes that call and returns. * :samp:`list` Lists a few lines around the current source location. In practice, it is usually more convenient to have a separate edit window open with the relevant source file displayed. ``emacs`` has debugging modes that display both the relevant source and ``GDB`` commands and output. Successive applications of this command print subsequent lines. You can give this command an argument which is a line number, in which case it displays a few lines around the specified line. * :samp:`backtrace` Displays a backtrace of the call chain. This command is typically used after a breakpoint has occurred to examine the sequence of calls that leads to the current breakpoint. The display includes one line for each activation record (frame) corresponding to an active subprogram. * :samp:`up` At a breakpoint, ``GDB`` can display the values of variables local to the current frame. You can use the command ``up`` to examine the contents of other active frames by moving the focus up the stack, that is to say from callee to caller, one frame at a time. * :samp:`down` Moves the focus of ``GDB`` down from the frame currently being examined to the frame of its callee (the reverse of the previous command), * :samp:`frame {n}` Inspect the frame with the given number. The value 0 denotes the frame of the current breakpoint, that is to say the top of the call stack. * :samp:`kill` Kills the child process in which the program is running under GDB. You may find this useful for several purposes: * It allows you to recompile and relink your program, since on many systems you cannot regenerate an executable file while it is running in a process. * You can run your program outside the debugger on systems that do not permit executing a program outside GDB while breakpoints are set within GDB. * It allows you to debug a core dump rather than a running process. The above is a very short introduction to the commands that ``GDB`` provides. Important additional capabilities, including conditional breakpoints, the ability to execute command sequences on a breakpoint, the ability to debug at the machine instruction level and many other features are described in detail in :title:`Debugging with GDB`. Note that most commands can be abbreviated (for example, "c" for ``continue`` and "bt" for ``backtrace``) and only enough characters need be typed to disambiguate the command (e.g., "br" for ``breakpoint``). .. _Using_Ada_Expressions: Using Ada Expressions --------------------- .. index:: Ada expressions (in gdb) ``GDB`` supports a very large subset of Ada expression syntax, with some extensions. The philosophy behind the design of this subset is * ``GDB`` should provide basic literals and access to operations for arithmetic, dereferencing, field selection, indexing, and subprogram calls, leaving more sophisticated computations to subprograms written into the program (which therefore may be called from ``GDB``). * Type safety and strict adherence to Ada language restrictions are not particularly relevant in a debugging context. * Brevity is important to the ``GDB`` user. Thus, for brevity, the debugger acts as if there were implicit ``with`` and ``use`` clauses in effect for all user-written packages, thus making it unnecessary to fully qualify most names with their packages, regardless of context. Where this causes ambiguity, ``GDB`` asks the user's intent. For details on the supported Ada syntax, see :title:`Debugging with GDB`. .. _Calling_User-Defined_Subprograms: Calling User-Defined Subprograms -------------------------------- An important capability of ``GDB`` is the ability to call user-defined subprograms while debugging. You do this by simply entering a subprogram call statement in the form: :: call subprogram-name (parameters) You can omit the keyword ``call`` in the normal case where the ``subprogram-name`` does not coincide with any of the predefined ``GDB`` commands. The effect is to invoke the given subprogram, passing it the list of parameters that is supplied. The parameters you specify can be expressions and can include variables from the program being debugged. The subprogram must be defined at the library level within your program and ``GDB`` will call the subprogram within the environment of your program execution (which means that the subprogram is free to access or even modify variables within your program). The most important use of this facility that you can include debugging routines that are tailored to particular data structures in your program. You can write such debugging routines to provide a suitably high-level description of an abstract type, rather than a low-level dump of its physical layout. After all, the standard ``GDB print`` command only knows the physical layout of your types, not their abstract meaning. Debugging routines can provide information at the desired semantic level and are thus enormously useful. For example, when debugging GNAT itself, it is crucial to have access to the contents of the tree nodes used to represent the program internally. But tree nodes are represented simply by an integer value (which in turn is an index into a table of nodes). Using the ``print`` command on a tree node would simply print this integer value, which is not very useful. But the ``PN`` routine (defined in file :file:`treepr.adb` in the GNAT sources) takes a tree node as input and displays a useful high level representation of the tree node, which includes the syntactic category of the node, its position in the source, the descendant nodes and parent node, as well as lots of semantic information. To study this example in more detail, you might want to look at the body of the ``PN`` procedure in the above file. Another useful application of this capability is to deal with situations where complex data which are not handled suitably by GDB. For example, if you specify Convention Fortran for a multi-dimensional array, GDB does not know that the ordering of array elements has been switched and will not properly address the array elements. In such a case, instead of trying to print the elements directly from GDB, you can write a callable procedure that prints the elements in the format you desire. .. _Using_the_Next_Command_in_a_Function: Using the *next* Command in a Function -------------------------------------- When you use the ``next`` command in a function, the current source location will advance to the next statement as usual. A special case arises in the case of a ``return`` statement. Part of the code for a return statement is the 'epilogue' of the function. This is the code that returns to the caller. There is only one copy of this epilogue code and it is typically associated with the last return statement in the function if there is more than one return. In some implementations, this epilogue is associated with the first statement of the function. The result is that if you use the ``next`` command from a return statement that is not the last return statement of the function you may see a strange apparent jump to the last return statement or to the start of the function. You should simply ignore this odd jump. The value returned is always that from the first return statement that was stepped through. .. _Stopping_When_Ada_Exceptions_Are_Raised: Stopping When Ada Exceptions Are Raised --------------------------------------- .. index:: Exceptions (in gdb) You can set catchpoints that stop the program execution when your program raises selected exceptions. * :samp:`catch exception` Set a catchpoint that stops execution whenever (any task in the) program raises any exception. * :samp:`catch exception {name}` Set a catchpoint that stops execution whenever (any task in the) program raises the exception *name*. * :samp:`catch exception unhandled` Set a catchpoint that stops executing whenever (any task in the) program raises an exception for which there is no handler. * :samp:`info exceptions`, :samp:`info exceptions {regexp}` The ``info exceptions`` command permits the user to examine all defined exceptions within Ada programs. With a regular expression, *regexp*, as argument, prints out only those exceptions whose name matches *regexp*. .. index:: Tasks (in gdb) .. _Ada_Tasks: Ada Tasks --------- ``GDB`` allows the following task-related commands: * :samp:`info tasks` This command shows a list of current Ada tasks, as in the following example: :: (gdb) info tasks ID TID P-ID Thread Pri State Name 1 8088000 0 807e000 15 Child Activation Wait main_task 2 80a4000 1 80ae000 15 Accept/Select Wait b 3 809a800 1 80a4800 15 Child Activation Wait a * 4 80ae800 3 80b8000 15 Running c In this listing, the asterisk before the first task indicates it's currently running task. The first column lists the task ID used to refer to tasks in the following commands. .. index:: Breakpoints and tasks * ``break`` *linespec* ``task`` *taskid*, ``break`` *linespec* ``task`` *taskid* ``if`` ... These commands are like the ``break ... thread ...``. *linespec* specifies source lines. Use the qualifier :samp:`task {taskid}` with a breakpoint command to specify that you only want ``GDB`` to stop the program when that particular Ada task reaches this breakpoint. *taskid* is one of the numeric task identifiers assigned by ``GDB``, shown in the first column of the ``info tasks`` display. If you don't specify :samp:`task {taskid}` when you set a breakpoint, the breakpoint applies to *all* tasks of your program. You can use the ``task`` qualifier on conditional breakpoints as well; in this case, place :samp:`task {taskid}` before the breakpoint condition (before the ``if``). .. index:: Task switching (in gdb) * :samp:`task {taskno}` This command allows switching to the task referred by *taskno*. In particular, it allows browsing the backtrace of the specified task. You should switch back to the original task before continuing execution; otherwise the scheduling of the program may be disturbed. For more detailed information on tasking support, see :title:`Debugging with GDB`. .. index:: Debugging Generic Units .. index:: Generics .. _Debugging_Generic_Units: Debugging Generic Units ----------------------- GNAT always uses the code expansion mechanism for generic instantiation. This means that each time an instantiation occurs, the compiler makes a complete copy of the original code, with appropriate substitutions of formals by actuals. You can't refer to the original generic entities in ``GDB``, but you can debug a particular instance of a generic by using the appropriate expanded names. For example, if we have .. code-block:: ada procedure g is generic package k is procedure kp (v1 : in out integer); end k; package body k is procedure kp (v1 : in out integer) is begin v1 := v1 + 1; end kp; end k; package k1 is new k; package k2 is new k; var : integer := 1; begin k1.kp (var); k2.kp (var); k1.kp (var); k2.kp (var); end; Then to break on a call to procedure kp in the k2 instance, simply use the command: :: (gdb) break g.k2.kp When the breakpoint occurs, you can step through the code of the instance in the normal manner and examine the values of local variables, as you do for other units. .. index:: Remote Debugging with gdbserver .. _Remote_Debugging_with_gdbserver: Remote Debugging with gdbserver ------------------------------- On platforms that support ``gdbserver``, you can use this tool to debug your application remotely. This can be useful in situations where the program needs to be run on a target host that is different from the host used for development, particularly when the target has a limited amount of resources (either CPU and/or memory). To do so, start your program using ``gdbserver`` on the target machine. ``gdbserver`` automatically suspends the execution of your program at its entry point, waiting for a debugger to connect to it. You use the following commands to start an application and tell ``gdbserver`` to wait for a connection with the debugger on ``localhost`` port 4444. :: $ gdbserver localhost:4444 program Process program created; pid = 5685 Listening on port 4444 Once ``gdbserver`` has started listening, you can tell the debugger to establish a connection with this ``gdbserver``, and then start a debugging session as if the program was being debugged on the same host, directly under the control of ``GDB``. :: $ gdb program (gdb) target remote targethost:4444 Remote debugging using targethost:4444 0x00007f29936d0af0 in ?? () from /lib64/ld-linux-x86-64.so. (gdb) b foo.adb:3 Breakpoint 1 at 0x401f0c: file foo.adb, line 3. (gdb) continue Continuing. Breakpoint 1, foo () at foo.adb:4 4 end foo; You can also use ``gdbserver`` to attach to an already running program, in which case the execution of that program is suspended until you have established the connection between the debugger and ``gdbserver``. For more information on how to use ``gdbserver``, see the *Using the gdbserver Program* section in :title:`Debugging with GDB`. GNAT provides support for ``gdbserver`` on x86-linux, x86-windows and x86_64-linux. .. index:: Abnormal Termination or Failure to Terminate .. _GNAT_Abnormal_Termination_or_Failure_to_Terminate: GNAT Abnormal Termination or Failure to Terminate ------------------------------------------------- When presented with programs that contain serious errors in syntax or semantics, GNAT may, on rare occasions, experience problems such as aborting with a segmentation fault or illegal memory access, raising an internal exception, terminating abnormally, or failing to terminate at all. In such cases, you can activate various features of GNAT that can help you pinpoint the construct in your program that is the likely source of the problem. The following strategies for you to use in such cases are presented in increasing order of difficulty, corresponding to your experience in using GNAT and your familiarity with compiler internals. * Run ``gcc`` with the :switch:`-gnatf`. This switch causes all errors on a given line to be reported. In its absence, GNAT only displays the first error on a line. The :switch:`-gnatdO` switch causes errors to be displayed as soon as they are encountered, rather than after compilation is terminated. If GNAT terminates prematurely or goes into an infinite loop, the last error message displayed may help to pinpoint the culprit. * Run ``gcc`` with the :switch:`-v` (verbose) switch. In this mode, ``gcc`` produces ongoing information about the progress of the compilation and provides the name of each procedure as code is generated. This switch allows you to find which Ada procedure was being compiled when it encountered a problem. .. index:: -gnatdc switch * Run ``gcc`` with the :switch:`-gnatdc` switch. This is a GNAT specific switch that does for the front-end what :switch:`-v` does for the back end. The system prints the name of each unit, either a compilation unit or nested unit, as it is being analyzed. * Finally, you can start ``gdb`` directly on the ``gnat1`` executable. ``gnat1`` is the front-end of GNAT and can be run independently (normally it is just called from ``gcc``). You can use ``gdb`` on ``gnat1`` as you would on a C program (but :ref:`The_GNAT_Debugger_GDB` for caveats). The ``where`` command is the first line of attack; the variable ``lineno`` (seen by ``print lineno``), used by the second phase of ``gnat1`` and by the ``gcc`` back end, indicates the source line at which the execution stopped, and ``input_file name`` indicates the name of the source file. .. _Naming_Conventions_for_GNAT_Source_Files: Naming Conventions for GNAT Source Files ---------------------------------------- In order to better understand the workings of the GNAT system, the following brief description of its organization may be helpful: * Files with prefix :file:`sc` contain the lexical scanner. * All files prefixed with :file:`par` are components of the parser. The numbers correspond to chapters of the Ada Reference Manual. For example, parsing of select statements can be found in :file:`par-ch9.adb`. * All files prefixed with :file:`sem` perform semantic analysis. The numbers correspond to chapters of the Ada standard. For example, all issues involving context clauses can be found in :file:`sem_ch10.adb`. In addition, some features of the language require sufficient special processing to justify their own semantic files, such as :file:`sem_aggr.adb` for aggregates and :file:`sem_disp.adb` for dynamic dispatching. * All files prefixed with :file:`exp` perform normalization and expansion of the intermediate representation (abstract syntax tree, or AST). The expansion has the effect of lowering the semantic level of the AST to a level closer to what the back end can handle. For example, it converts tasking operations into calls to the appropriate runtime routines. These files use the same numbering scheme as the parser and semantics files. For example, the construction of record initialization procedures is done in :file:`exp_ch3.adb`. * The files prefixed with :file:`bind` implement the binder, which verifies the consistency of the compilation, determines an order of elaboration, and generates the bind file. * The files :file:`atree.ads` and :file:`atree.adb` detail the low-level data structures used by the front-end. * The files :file:`sinfo.ads` and :file:`sinfo.adb` detail the structure of the abstract syntax tree as produced by the parser. * The files :file:`einfo.ads` and :file:`einfo.adb` detail the attributes of all entities, computed during semantic analysis. * The files prefixed with :file:`gen_il` generate most of the functions defined in :file:`sinfo.ads` and :file:`einfo.ads`, which set and get various fields and flags of the AST. * Library management issues are dealt with in files with prefix :file:`lib`. .. index:: Annex A (in Ada Reference Manual) * Ada files with the prefix :file:`a-` are children of ``Ada``, as defined in Annex A. .. index:: Annex B (in Ada reference Manual) * Files with prefix :file:`i-` are children of ``Interfaces``, as defined in Annex B. .. index:: System (package in Ada Reference Manual) * Files with prefix :file:`s-` are children of ``System``. This includes both language-defined children and GNAT run-time routines. .. index:: GNAT (package) * Files with prefix :file:`g-` are children of ``GNAT``. These are useful general-purpose packages, fully documented in their specs. All the other :file:`.c` files are modifications of common ``gcc`` files. .. _Getting_Internal_Debugging_Information: Getting Internal Debugging Information -------------------------------------- Most compilers have internal debugging switches and modes. GNAT does too, except GNAT internal debugging switches and modes are not secret. A summary and full description of all the compiler and binder debug flags are in the file :file:`debug.adb`. You must obtain the sources of the compiler to see the full detailed effects of these flags. The switches that print the source of the program (reconstructed from the internal tree) are of general interest for user programs, as are the options to print the full internal tree and the entity table (the symbol table information). The reconstructed source provides a readable version of the program after the front-end has completed analysis and expansion and is useful when studying the performance of specific constructs. For example, constraint checks are shown explicitly, complex aggregates are replaced with loops and assignments, and tasking primitives are replaced with run-time calls. .. index:: traceback .. index:: stack traceback .. index:: stack unwinding .. _Stack_Traceback: Stack Traceback --------------- Traceback is a mechanism to display the sequence of subprogram calls that leads to a specified execution point in a program. Often (but not always) the execution point is an instruction at which an exception has been raised. This mechanism is also known as *stack unwinding* because it obtains its information by scanning the run-time stack and recovering the activation records of all active subprograms. Stack unwinding is one of the most important tools for program debugging. The first entry stored in traceback corresponds to the deepest calling level, that is to say the subprogram currently executing the instruction from which we want to obtain the traceback. Note that there is no runtime performance penalty when stack traceback is enabled and no exception is raised during program execution. .. index:: traceback, non-symbolic .. _Non-Symbolic_Traceback: Non-Symbolic Traceback ^^^^^^^^^^^^^^^^^^^^^^ Note: this feature is not supported on all platforms. See :samp:`GNAT.Traceback` spec in :file:`g-traceb.ads` for a complete list of supported platforms. .. rubric:: Tracebacks From an Unhandled Exception A runtime non-symbolic traceback is a list of addresses of call instructions. To enable this feature you must use the :switch:`-E` ``gnatbind`` switch. With this switch, a stack traceback is stored at runtime as part of exception information. You can translate this information using the ``addr2line`` tool, provided that the program is compiled with debugging options (see :ref:`Switches_for_gcc`) and linked at a fixed position with :switch:`-no-pie`. Here's a simple example with ``gnatmake``: .. code-block:: ada procedure STB is procedure P1 is begin raise Constraint_Error; end P1; procedure P2 is begin P1; end P2; begin P2; end STB; :: $ gnatmake stb -g -bargs -E -largs -no-pie $ stb Execution of stb terminated by unhandled exception raised CONSTRAINT_ERROR : stb.adb:5 explicit raise Load address: 0x400000 Call stack traceback locations: 0x401373 0x40138b 0x40139c 0x401335 0x4011c4 0x4011f1 0x77e892a4 As we can see, the traceback lists a sequence of addresses for the unhandled exception ``CONSTRAINT_ERROR`` raised in procedure P1. It's easy to see that this exception come from procedure P1. To translate these addresses into the source lines where the calls appear, you need to invoke the ``addr2line`` tool like this: :: $ addr2line -e stb 0x401373 0x40138b 0x40139c 0x401335 0x4011c4 0x4011f1 0x77e892a4 d:/stb/stb.adb:5 d:/stb/stb.adb:10 d:/stb/stb.adb:14 d:/stb/b~stb.adb:197 crtexe.c:? crtexe.c:? ??:0 The ``addr2line`` tool has several other useful options: ========================= ==================================================== :samp:`-a --addresses` to show the addresses alongside the line numbers :samp:`-f --functions` to get the function name corresponding to a location :samp:`-p --pretty-print` to print all the information on a single line :samp:`--demangle=gnat` to use the GNAT decoding mode for the function names ========================= ==================================================== :: $ addr2line -e stb -a -f -p --demangle=gnat 0x401373 0x40138b 0x40139c 0x401335 0x4011c4 0x4011f1 0x77e892a4 0x00401373: stb.p1 at d:/stb/stb.adb:5 0x0040138B: stb.p2 at d:/stb/stb.adb:10 0x0040139C: stb at d:/stb/stb.adb:14 0x00401335: main at d:/stb/b~stb.adb:197 0x004011c4: ?? at crtexe.c:? 0x004011f1: ?? at crtexe.c:? 0x77e892a4: ?? ??:0 From this traceback, we can see that the exception was raised in :file:`stb.adb` at line 5, which was reached from a procedure call in :file:`stb.adb` at line 10, and so on. :file:`b~std.adb` is the binder file, which contains the call to the main program; :ref:`Running_gnatbind`. The remaining entries are assorted runtime routines. The output will vary from platform to platform. You can also use ``GDB`` with these traceback addresses to debug the program. For example, we can break at a given code location, as reported in the stack traceback:: $ gdb -nw stb (gdb) break *0x401373 Breakpoint 1 at 0x401373: file stb.adb, line 5. It is important to note that the stack traceback addresses do not change when debug information is included. This is particularly useful because it makes it possible to release software without debug information (to minimize object size), get a field report that includes a stack traceback whenever an internal bug occurs, and then be able to retrieve the sequence of calls with the same program compiled with debug information. However the ``addr2line`` tool does not work with Position-Independent Code (PIC), the historical example being Linux dynamic libraries and Windows DLLs, which nowadays encompasse Position-Independent Executables (PIE) on recent Linux and Windows versions. In order to translate addresses the source lines with Position-Independent Executables on recent Linux and Windows versions, in other words without using the switch :switch:`-no-pie` during linking, you need to use the ``gnatsymbolize`` tool with :switch:`--load` instead of the ``addr2line`` tool. The main difference is that you need to copy the Load Address output in the traceback ahead of the sequence of addresses. The default mode of ``gnatsymbolize`` is equivalent to that of ``addr2line`` with the above switches, so none of them are needed:: $ gnatmake stb -g -bargs -E $ stb Execution of stb terminated by unhandled exception raised CONSTRAINT_ERROR : stb.adb:5 explicit raise Load address: 0x400000 Call stack traceback locations: 0x401373 0x40138b 0x40139c 0x401335 0x4011c4 0x4011f1 0x77e892a4 $ gnatsymbolize --load stb 0x400000 0x401373 0x40138b 0x40139c 0x401335 \ 0x4011c4 0x4011f1 0x77e892a4 0x00401373 Stb.P1 at stb.adb:5 0x0040138B Stb.P2 at stb.adb:10 0x0040139C Stb at stb.adb:14 0x00401335 Main at b~stb.adb:197 0x004011c4 __tmainCRTStartup at ??? 0x004011f1 mainCRTStartup at ??? 0x77e892a4 ??? at ??? .. rubric:: Tracebacks From Exception Occurrences Non-symbolic tracebacks are obtained by using the :switch:`-E` binder switch. The stack traceback is attached to the exception information string and you can retrieve it in an exception handler within the Ada program by means of the Ada facilities defined in ``Ada.Exceptions``. Here's a simple example: .. code-block:: ada with Ada.Text_IO; with Ada.Exceptions; procedure STB is use Ada; use Ada.Exceptions; procedure P1 is K : Positive := 1; begin K := K - 1; exception when E : others => Text_IO.Put_Line (Exception_Information (E)); end P1; procedure P2 is begin P1; end P2; begin P2; end STB; :: $ gnatmake stb -g -bargs -E -largs -no-pie $ stb raised CONSTRAINT_ERROR : stb.adb:12 range check failed Load address: 0x400000 Call stack traceback locations: 0x4015e4 0x401633 0x401644 0x401461 0x4011c4 0x4011f1 0x77e892a4 .. rubric:: Tracebacks From Anywhere in a Program You can also retrieve a stack traceback from anywhere in a program. For this, you need to use the ``GNAT.Traceback`` API. This package includes a procedure called ``Call_Chain`` that computes a complete stack traceback as well as useful display procedures described below. You don't have to use the :switch:`-E` ``gnatbind`` switch in this case because the stack traceback mechanism is invoked explicitly. In the following example, we compute a traceback at a specific location in the program and display it using ``GNAT.Debug_Utilities.Image`` to convert addresses to strings: .. code-block:: ada with Ada.Text_IO; with GNAT.Traceback; with GNAT.Debug_Utilities; with System; procedure STB is use Ada; use Ada.Text_IO; use GNAT; use GNAT.Traceback; use System; LA : constant Address := Executable_Load_Address; procedure P1 is TB : Tracebacks_Array (1 .. 10); -- We are asking for a maximum of 10 stack frames. Len : Natural; -- Len will receive the actual number of stack frames returned. begin Call_Chain (TB, Len); Put ("In STB.P1 : "); for K in 1 .. Len loop Put (Debug_Utilities.Image_C (TB (K))); Put (' '); end loop; New_Line; end P1; procedure P2 is begin P1; end P2; begin if LA /= Null_Address then Put_Line ("Load address: " & Debug_Utilities.Image_C (LA)); end if; P2; end STB; :: $ gnatmake stb -g $ stb Load address: 0x400000 In STB.P1 : 0x40F1E4 0x4014F2 0x40170B 0x40171C 0x401461 0x4011C4 \ 0x4011F1 0x77E892A4 You can get even more information by invoking the ``addr2line`` tool or the ``gnatsymbolize`` tool as described earlier (note that the hexadecimal addresses need to be specified in C format, with a leading '0x'). .. index:: traceback, symbolic .. _Symbolic_Traceback: Symbolic Traceback ^^^^^^^^^^^^^^^^^^ A symbolic traceback is a stack traceback in which procedure names are associated with each code location. Note that this feature is not supported on all platforms. See :samp:`GNAT.Traceback.Symbolic` spec in :file:`g-trasym.ads` for a complete list of currently supported platforms. Note that the symbolic traceback requires that the program be compiled with debug information. If you do not compile it with debug information, only the non-symbolic information will be valid. .. rubric:: Tracebacks From Exception Occurrences Here is an example: .. code-block:: ada with Ada.Text_IO; with GNAT.Traceback.Symbolic; procedure STB is procedure P1 is begin raise Constraint_Error; end P1; procedure P2 is begin P1; end P2; procedure P3 is begin P2; end P3; begin P3; exception when E : others => Ada.Text_IO.Put_Line (GNAT.Traceback.Symbolic.Symbolic_Traceback (E)); end STB; :: $ gnatmake -g stb -bargs -E $ stb 0040149F in stb.p1 at stb.adb:8 004014B7 in stb.p2 at stb.adb:13 004014CF in stb.p3 at stb.adb:18 004015DD in ada.stb at stb.adb:22 00401461 in main at b~stb.adb:168 004011C4 in __mingw_CRTStartup at crt1.c:200 004011F1 in mainCRTStartup at crt1.c:222 77E892A4 in ?? at ??:0 .. rubric:: Tracebacks From Anywhere in a Program You can get a symbolic stack traceback from anywhere in a program, just as you can for non-symbolic tracebacks. The first step is to obtain a non-symbolic traceback. Then call ``Symbolic_Traceback`` to compute the symbolic information. Here is an example: .. code-block:: ada with Ada.Text_IO; with GNAT.Traceback; with GNAT.Traceback.Symbolic; procedure STB is use Ada; use GNAT.Traceback; use GNAT.Traceback.Symbolic; procedure P1 is TB : Tracebacks_Array (1 .. 10); -- We are asking for a maximum of 10 stack frames. Len : Natural; -- Len will receive the actual number of stack frames returned. begin Call_Chain (TB, Len); Text_IO.Put_Line (Symbolic_Traceback (TB (1 .. Len))); end P1; procedure P2 is begin P1; end P2; begin P2; end STB; .. rubric:: Automatic Symbolic Tracebacks You may also enable symbolic tracebacks by using the :switch:`-Es` switch to gnatbind (as in ``gprbuild -g ... -bargs -Es``). This causes the Exception_Information to contain a symbolic traceback, which will also be printed if an unhandled exception terminates the program. .. _Pretty-Printers_For_The_GNAT_Runtime: Pretty-Printers for the GNAT runtime ------------------------------------ As discussed in :title:`Calling User-Defined Subprograms`, GDB's ``print`` command only knows about the physical layout of program data structures and therefore normally displays only low-level dumps, which are often hard to understand. An example of this is when trying to display the contents of an Ada standard container, such as ``Ada.Containers.Ordered_Maps.Map``: .. code-block:: ada with Ada.Containers.Ordered_Maps; procedure PP is package Int_To_Nat is new Ada.Containers.Ordered_Maps (Integer, Natural); Map : Int_To_Nat.Map; begin Map.Insert (1, 10); Map.Insert (2, 20); Map.Insert (3, 30); Map.Clear; -- BREAK HERE end PP; When this program is built with debugging information and run under ``GDB`` up to the ``Map.Clear`` statement, trying to print ``Map`` will yield information that is only relevant to the developers of the standard containers: :: (gdb) print map $1 = ( tree => ( first => 0x64e010, last => 0x64e070, root => 0x64e040, length => 3, tc => ( busy => 0, lock => 0 ) ) ) Fortunately, ``GDB ``has a feature called `pretty-printers `_, which allows customizing how ``GDB`` displays data structures. The ``GDB`` shipped with GNAT embeds such pretty-printers for the most common containers in the standard library. To enable them, either run the following command manually under ``GDB`` or add it to your :file:`.gdbinit` file: :: python import gnatdbg; gnatdbg.setup() Once you've done this, ``GDB``'s ``print`` command will automatically use these pretty-printers when appropriate. Using the previous example: :: (gdb) print map $1 = pp.int_to_nat.map of length 3 = { [1] = 10, [2] = 20, [3] = 30 } Pretty-printers are invoked each time GDB tries to display a value, including when displaying the arguments of a called subprogram (in GDB's ``backtrace`` command) or when printing the value returned by a function (in GDB's ``finish`` command). To display a value without involving pretty-printers, you can invoke ``print`` with its ``/r`` option: :: (gdb) print/r map $1 = ( tree => (... You can also obtain finer control of pretty-printers: see `GDB's online documentation `_ for more information. .. index:: Profiling .. _Profiling: Profiling ========= This section describes how to use the ``gprof`` profiler tool on Ada programs. .. index:: ! gprof .. index:: Profiling .. _Profiling_an_Ada_Program_with_gprof: Profiling an Ada Program with gprof ----------------------------------- This section is not meant to be an exhaustive documentation of ``gprof``. You can find full documentation for it in the :title:`GNU Profiler User's Guide` documentation that is part of this GNAT distribution. Profiling a program helps determine the parts of a program that are executed most often and are therefore the most time-consuming. ``gprof`` is the standard GNU profiling tool; it has been enhanced to better handle Ada programs and multitasking. It's currently supported on the following platforms * Linux x86/x86_64 * Windows x86/x86_64 (without PIE support) In order to profile a program using ``gprof``, you need to perform the following steps: #. Instrument the code, which requires a full recompilation of the project with the proper switches. #. Execute the program under the analysis conditions, i.e. with the desired input. #. Analyze the results using the ``gprof`` tool. The following sections detail the different steps and indicate how to interpret the results. .. _Compilation_for_profiling: Compilation for profiling ^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: -pg (gcc), for profiling .. index:: -pg (gnatlink), for profiling In order to profile a program, you must first to tell the compiler to generate the necessary profiling information. You do this using the compiler switch :switch:`-pg`, which you must add to other compilation switches. You need to specify this switch during compilation and link stages, and you can specified it only once when using ``gnatmake``: :: $ gnatmake -f -pg -P my_project Note that only the objects that were compiled with the :switch:`-pg` switch will be profiled; if you need to profile your whole project, use the :switch:`-f` ``gnatmake`` switch to force full recompilation. Note that on Windows, ``gprof`` does not support PIE. You should add the :switch:`-no-pie` switch to the linker flags to disable PIE. .. _Program_execution: Program execution ^^^^^^^^^^^^^^^^^ Once the program has been compiled for profiling, you can run it as usual. The only constraint imposed by profiling is that the program must terminate normally. An interrupted program (via a Ctrl-C, kill, etc.) will not be properly analyzed. Once the program completes execution, a data file called :file:`gmon.out` is generated in the directory where the program was launched from. If this file already exists, it will be overwritten by running the program. .. _Running_gprof: Running gprof ^^^^^^^^^^^^^ You can call the ``gprof`` tool as follows: :: $ gprof my_prog gmon.out or simply: :: $ gprof my_prog The complete form of the ``gprof`` command line is the following: :: $ gprof [switches] [executable [data-file]] ``gprof`` supports numerous switches, whose order does not matter. You can find the full list of switches in the :title:`GNU Profiler User's Guide`. The following are the most relevant of those switches: .. index:: --demangle (gprof) :switch:`--demangle[={style}]`, :switch:`--no-demangle` These switches control whether symbol names should be demangled when printing output. The default is to demangle C++ symbols. You can use :switch:`--no-demangle` to turn off demangling. Different compilers have different mangling styles. The optional demangling style argument can be used to choose an appropriate demangling style for your compiler, in particular Ada symbols generated by GNAT can be demangled using :switch:`--demangle=gnat`. .. index:: -e (gprof) :switch:`-e {function_name}` The :switch:`-e {function}` option tells ``gprof`` not to print information about the function ``function_name`` and its children in the call graph. The function will still be listed as a child of any functions that call it, but its index number will be shown as ``[not printed]``. You may specify more than one :switch:`-e` switch, but you may only include one ``function_name`` with each :switch:`-e` switch. .. index:: -E (gprof) :switch:`-E {function_name}` The :switch:`-E {function}` switch works like the :switch:`-e` switch, but execution time spent in the function (and children who were not called from anywhere else) will not be used to compute the percentages-of-time for the call graph. You may specify more than one :switch:`-E` switch, but you may only include one ``function_name`` with each :switch:`-E` switch. .. index:: -f (gprof) :switch:`-f {function_name}` The :switch:`-f {function}` switch causes ``gprof`` to limit the call graph to the function ``function_name`` and its children and their children. You may specify more than one :switch:`-f` switch, but you may only include one ``function_name`` with each :switch:`-f` switch. .. index:: -F (gprof) :switch:`-F {function_name}` The :switch:`-F {function}` switch works like the :switch:`-f` switch, but only time spent in the function and its children and their children will be used to determine total-time and percentages-of-time for the call graph. You may specify more than one :switch:`-F` switch, but you may include only one ``function_name`` with each :switch:`-F` switch. The :switch:`-F` switch overrides the :switch:`-E` switch. .. _Interpretation_of_profiling_results: Interpretation of profiling results ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The results of the profiling analysis are represented by two arrays: the 'flat profile' and the 'call graph'. You can find full documentation of those outputs in the :title:`GNU Profiler User's Guide`. The flat profile shows the time spent in each function of the program and how many time it has been called. This allows you to easily locate the most time-consuming functions. The call graph shows, for each subprogram, the subprograms that call it, and the subprograms that it calls. It also provides an estimate of the time spent in each of those callers and called subprograms. .. _Improving_Performance: Improving Performance ===================== .. index:: Improving performance This section presents several topics related to program performance. It first describes some of the tradeoffs that you need to consider and some of the techniques for making your program run faster. It then documents the unused subprogram/data elimination feature, which can reduce the size of program executables. .. _Performance_Considerations: Performance Considerations -------------------------- The GNAT system provides a number of options that allow a trade-off between: * performance of the generated code * speed of compilation * minimization of dependences and recompilation * the degree of run-time checking. The default (if you don't select any switches) aims at improving the speed of compilation and minimizing dependences, at the expense of performance of the generated code and consists of: * no optimization * no inlining of subprogram calls * all run-time checks enabled except overflow and elaboration checks These options are suitable for most program development purposes. This section describes how you can modify these choices and also provides some guidelines on debugging optimized code. .. _Controlling_Run-Time_Checks: Controlling Run-Time Checks ^^^^^^^^^^^^^^^^^^^^^^^^^^^ By default, GNAT generates all run-time checks, except stack overflow checks and checks for access before elaboration on subprogram calls. The latter are not required in default mode because all necessary checking is done at compile time. .. index:: -gnatp (gcc) .. index:: -gnato (gcc) The GNAT switch, :switch:`-gnatp` allows you to modify this default; see :ref:`Run-Time_Checks`. Our experience is that the default is suitable for most development purposes. Elaboration checks are off by default and also not needed by default since GNAT uses a static elaboration analysis approach that avoids the need for run-time checking. This manual contains a full chapter discussing the issue of elaboration checks and you should read this chapter if the default is not satisfactory for your use, For validity checks, the minimal checks required by the Ada Reference Manual (for case statements and assignments to array elements) are enabled by default. You can suppress these by using the :switch:`-gnatVn` switch. Note that in Ada 83, there were no validity checks, so if the Ada 83 mode is acceptable (or when comparing GNAT performance with an Ada 83 compiler), it may be reasonable to routinely use :switch:`-gnatVn`. Validity checks are also suppressed entirely if you use :switch:`-gnatp`. .. index:: Overflow checks .. index:: Checks, overflow .. index:: Suppress .. index:: Unsuppress .. index:: pragma Suppress .. index:: pragma Unsuppress Note that the setting of the switches controls the default setting of the checks. You may modify them using either ``pragma Suppress`` (to remove checks) or ``pragma Unsuppress`` (to add back suppressed checks) in your program source. .. _Use_of_Restrictions: Use of Restrictions ^^^^^^^^^^^^^^^^^^^ You can use pragma Restrictions to control which features are permitted in your program. In most cases, the use of this pragma itself does not affect the generated code (but, of course, if you avoid relatively expensive features like finalization, you'll have more efficient programs and that's enforceable by the use of pragma Restrictions (No_Finalization). One notable exception to this rule is that the possibility of task abort results in some distributed overhead, particularly if finalization or exception handlers are used. This is because certain sections of code must be marked as non-abortable. If you use neither the ``abort`` statement nor asynchronous transfer of control (``select ... then abort``), this distributed overhead can be removed, which may have a general positive effect in improving overall performance, especially in code involving frequent use of tasking constructs and controlled types, which will show much improved performance. The relevant restrictions pragmas are .. code-block:: ada pragma Restrictions (No_Abort_Statements); pragma Restrictions (Max_Asynchronous_Select_Nesting => 0); We recommend that you use these restriction pragmas if possible. If you do this, it also means you can write code without worrying about the possibility of an immediate abort at any point. .. _Optimization_Levels: Optimization Levels ^^^^^^^^^^^^^^^^^^^ .. index:: -O (gcc) Without any optimization switch, the compiler's goal is to reduce the cost of compilation and to make debugging produce the expected results. This means that statements are independent: if you stop the program with a breakpoint between statements, you can then assign a new value to any variable or change the program counter to any other statement in the subprogram and get exactly the results you would expect from the source code. However, the generated programs are considerably larger and slower than when optimization is enabled. Turning on optimization makes the compiler attempt to improve the performance and/or code size at the expense of compilation time and possibly the ability to debug the program. You can pass the :switch:`-O` switch, with or without an operand (the permitted forms with an operand are :switch:`-O0`, :switch:`-O1`, :switch:`-O2`, :switch:`-O3`, :switch:`-Os`, :switch:`-Oz`, and :switch:`-Og`) to ``gcc`` to control the optimization level. If you pass multiple :switch:`-O` switches, with or without an operand, the last such switch is the one that's used: * :switch:`-O0` No optimization (the default); generates unoptimized code but has the fastest compilation time. Debugging is easiest with this switch. Note that many other compilers do substantial optimization even if 'no optimization' is specified. With GCC, it is very unusual to use :switch:`-O0` for production if execution time is of any concern, since :switch:`-O0` means (almost) no optimization. You should keep this difference between GCC and other compilers in mind when doing performance comparisons. * :switch:`-O1` Moderate optimization (same as :switch:`-O` without an operand); optimizes reasonably well but does not degrade compilation time significantly. You may not be able to see some variables in the debugger, and changing the value of some variables in the debugger may not have the effect you desire. * :switch:`-O2` Extensive optimization; generates highly optimized code but has an increased compilation time. You may see significant impacts on your ability to display and modify variables in the debugger. * :switch:`-O3` Full optimization; attempts more sophisticated transformations, in particular on loops, possibly at the cost of larger generated code. You may be hardly able to use the debugger at this optimization level. * :switch:`-Os` Optimize for size (code and data) of resulting binary rather than speed; based on the :switch:`-O2` optimization level, but disables some of its transformations that often increase code size, as well as performs further optimizations designed to reduce code size. * :switch:`-Oz` Optimize aggressively for size (code and data) of resulting binary rather than speed; may increase the number of instructions executed if these instructions require fewer bytes to be encoded. * :switch:`-Og` Optimize for debugging experience rather than speed; based on the :switch:`-O1` optimization level, but attempts to eliminate all the negative effects of optimization on debugging. Higher optimization levels perform more global transformations on the program and apply more expensive analysis algorithms in order to generate faster and more compact code. The price in compilation time, and the resulting improvement in execution time, both depend on the particular application and the hardware environment. You should experiment to find the best level for your application. Since the precise set of optimizations done at each level will vary from release to release (and sometime from target to target), it is best to think of the optimization settings in general terms. See the *Options That Control Optimization* section in :title:`Using the GNU Compiler Collection (GCC)` for details about the :switch:`-O` settings and a number of :switch:`-f` switches that individually enable or disable specific optimizations. Unlike some other compilation systems, GCC has been tested extensively at all optimization levels. There are some bugs which appear only with optimization turned on, but there have also been bugs which show up only in *unoptimized* code. Selecting a lower level of optimization does not improve the reliability of the code generator, which in practice is highly reliable at all optimization levels. A note regarding the use of :switch:`-O3`: The use of this optimization level ought not to be automatically preferred over that of level :switch:`-O2`, since it often results in larger executables which may run more slowly. See further discussion of this point in :ref:`Inlining_of_Subprograms`. .. _Debugging_Optimized_Code: Debugging Optimized Code ^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: Debugging optimized code .. index:: Optimization and debugging Although it is possible to do a reasonable amount of debugging at nonzero optimization levels, the higher the level the more likely that source-level constructs will have been eliminated by optimization. For example, if a loop is strength-reduced, the loop control variable may be completely eliminated and thus cannot be displayed in the debugger. This can only happen at :switch:`-O2` or :switch:`-O3`. Explicit temporary variables that you code might be eliminated at level :switch:`-O1` or higher. .. index:: -g (gcc) The use of the :switch:`-g` switch, which is needed for source-level debugging, affects the size of the program executable on disk, and indeed the debugging information can be quite large. However, it has no effect on the generated code (and thus does not degrade performance) Since the compiler generates debugging tables for a compilation unit before it performs optimizations, the optimizing transformations may invalidate some of the debugging data. You therefore need to anticipate certain anomalous situations that may arise while debugging optimized code. These are the most common cases: * *The 'hopping Program Counter':* Repeated ``step`` or ``next`` commands show the PC bouncing back and forth in the code. This may result from any of the following optimizations: - *Common subexpression elimination:* using a single instance of code for a quantity that the source computes several times. As a result you may not be able to stop on what looks like a statement. - *Invariant code motion:* moving an expression that does not change within a loop to the beginning of the loop. - *Instruction scheduling:* moving instructions so as to overlap loads and stores (typically) with other code or in general to move computations of values closer to their uses. Often this causes you to pass an assignment statement without the assignment happening and then later bounce back to the statement when the value is actually needed. Placing a breakpoint on a line of code and then stepping over it may, therefore, not always cause all the expected side-effects. * *The 'big leap':* More commonly known as *cross-jumping*, in which two identical pieces of code are merged and the program counter suddenly jumps to a statement that is not supposed to be executed, simply because it (and the code following) translates to the same thing as the code that *was* supposed to be executed. This effect is typically seen in sequences that end in a jump, such as a ``goto``, a ``return``, or a ``break`` in a C ``switch`` statement. * *The 'roving variable':* The symptom is an unexpected value in a variable. There are various reasons for this effect: - In a subprogram prologue, a parameter may not yet have been moved to its 'home'. - A variable may be dead and its register re-used. This is probably the most common cause. - As mentioned above, the assignment of a value to a variable may have been moved. - A variable may be eliminated entirely by value propagation or other means. In this case, GCC may incorrectly generate debugging information for the variable In general, when an unexpected value appears for a local variable or parameter you should first ascertain if that value was actually computed by your program as opposed to being incorrectly reported by the debugger. Record fields or array elements in an object designated by an access value are generally less of a problem once you have verified that the access value is sensible. Typically, this means checking variables in the preceding code and in the calling subprogram to verify that the value observed is explainable from other values (you must apply the procedure recursively to those other values); or re-running the code and stopping a little earlier (perhaps before the call) and stepping to better see how the variable obtained the value in question; or continuing to step *from* the point of the strange value to see if code motion had simply moved the variable's assignments later. In light of such anomalies, a recommended technique is to use :switch:`-O0` early in the software development cycle, when extensive debugging capabilities are most needed, and then move to :switch:`-O1` and later :switch:`-O2` as the debugger becomes less critical. Whether to use the :switch:`-g` switch in the release version is a release management issue. Note that if you use :switch:`-g` you can then use the ``strip`` program on the resulting executable, which removes both debugging information and global symbols. .. _Inlining_of_Subprograms: Inlining of Subprograms ^^^^^^^^^^^^^^^^^^^^^^^ A call to a subprogram in the current unit is inlined if all the following conditions are met: * The optimization level is at least :switch:`-O1`. * The called subprogram is suitable for inlining: it must be small enough and not contain something that the back end cannot support in inlined subprograms. .. index:: pragma Inline .. index:: Inline * Any one of the following applies: ``pragma Inline`` is applied to the subprogram; the subprogram is local to the unit and called once from within it; the subprogram is small and optimization level :switch:`-O2` is specified; optimization level :switch:`-O3` is specified; or the subprogram is an expression function. Calls to subprograms in |withed| units are normally not inlined. To achieve inlining in those case (that is, replacement of the call by the code in the body of the subprogram), the following conditions must all be true: * The optimization level is at least :switch:`-O1`. * The called subprogram is suitable for inlining: It must be small enough and not contain something that the back end cannot support in inlined subprograms. * There is a ``pragma Inline`` for the subprogram. * The :switch:`-gnatn` switch is used on the command line. Even if all these conditions are met, it may not be possible for the compiler to inline the call due to the length of the body, or features in the body that make it impossible for the compiler to do the inlining. Note that specifying the :switch:`-gnatn` switch causes additional compilation dependencies. Consider the following: .. code-block:: ada package R is procedure Q; pragma Inline (Q); end R; package body R is ... end R; with R; procedure Main is begin ... R.Q; end Main; With the default behavior (no :switch:`-gnatn` switch specified), the compilation of the ``Main`` procedure depends only on its own source, :file:`main.adb`, and the spec of the package in file :file:`r.ads`. This means that editing the body of ``R`` does not require recompiling ``Main``. On the other hand, the call ``R.Q`` is not inlined under these circumstances. If the :switch:`-gnatn` switch is present when ``Main`` is compiled, the call will be inlined if the body of ``Q`` is small enough, but now ``Main`` depends on the body of ``R`` in :file:`r.adb` as well as on the spec. This means that if this body is edited, the main program must be recompiled. Note that this extra dependency occurs whether or not the call is in fact inlined by the back end. The use of front end inlining with :switch:`-gnatN` generates similar additional dependencies. .. index:: -fno-inline (gcc) Note: The :switch:`-fno-inline` switch overrides all other conditions and ensures that no inlining occurs, unless requested with pragma Inline_Always for most back ends. The extra dependences resulting from :switch:`-gnatn` will still be active, even if this switch is used to suppress the resulting inlining actions. .. index:: -fno-inline-functions (gcc) For the GCC back end, you can use the :switch:`-fno-inline-functions` switch to prevent automatic inlining of subprograms if you use :switch:`-O3`. .. index:: -fno-inline-small-functions (gcc) For the GCC back end, you can use the :switch:`-fno-inline-small-functions` switch to prevent automatic inlining of small subprograms if you use :switch:`-O2`. .. index:: -fno-inline-functions-called-once (gcc) For the GC back end, you can use the :switch:`-fno-inline-functions-called-once` switch to prevent inlining of subprograms local to the unit and called once from within it if you use :switch:`-O1`. A note regarding the use of :switch:`-O3`: :switch:`-gnatn` is made up of two sub-switches :switch:`-gnatn1` and :switch:`-gnatn2` that you can directly specify. :switch:`-gnatn` is translated into one of them based on the optimization level. With :switch:`-O2` or below, :switch:`-gnatn` is equivalent to :switch:`-gnatn1` which activates pragma ``Inline`` with moderate inlining across modules. With :switch:`-O3`, :switch:`-gnatn` is equivalent to :switch:`-gnatn2` which activates pragma ``Inline`` with full inlining across modules. If you have used pragma ``Inline`` in appropriate cases, it's usually much better to use :switch:`-O2` and :switch:`-gnatn` and avoid the use of :switch:`-O3` which has the additional effect of inlining subprograms you did not think should be inlined. We have found that the use of :switch:`-O3` may slow down the compilation and increase the code size by performing excessive inlining, leading to increased instruction cache pressure from the increased code size and thus minor performance degradations. So the bottom line here is that you should not automatically assume that :switch:`-O3` is better than :switch:`-O2` and indeed you should use :switch:`-O3` only if tests show that it actually improves performance for your program. .. _Floating_Point_Operations: Floating Point Operations ^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: Floating-Point Operations On almost all targets, GNAT maps ``Float`` and ``Long_Float`` to the 32-bit and 64-bit standard IEEE floating-point representations and operations will use standard IEEE arithmetic as provided by the processor. On most, but not all, architectures, the attribute ``Machine_Overflows`` is ``False`` for these types, meaning that the semantics of overflow is implementation-defined. In the case of GNAT, these semantics correspond to the normal IEEE treatment of infinities and NaN (not a number) values. For example, 1.0 / 0.0 yields plus infinitiy and 0.0 / 0.0 yields a NaN. By avoiding explicit overflow checks, the performance is greatly improved on many targets. However, if required, you can enable floating-point overflow by using the pragma ``Check_Float_Overflow``. Another consideration that applies specifically to x86 32-bit architectures is which form of floating-point arithmetic is used. By default, the operations use the old style x86 floating-point, which implements an 80-bit extended precision form (on these architectures the type ``Long_Long_Float`` corresponds to that form). In addition, generation of efficient code in this mode means that the extended precision form is used for intermediate results. This may be helpful in improving the final precision of a complex expression, but it means that the results obtained on the x86 may be different from those on other architectures and, for some algorithms, the extra intermediate precision can be detrimental. In addition to this old-style floating-point, all modern x86 chips implement an alternative floating-point operation model referred to as SSE2. In this model, there is no extended form and execution performance is significantly enhanced. To force GNAT to use this more modern form, use both of the switches: -msse2 -mfpmath=sse A unit compiled with these switches will automatically use the more efficient SSE2 instruction set for ``Float`` and ``Long_Float`` operations. Note that the ABI has the same form for both floating-point models, so you can mix units compiled with and without these switches. .. _Vectorization_of_loops: Vectorization of loops ^^^^^^^^^^^^^^^^^^^^^^ .. index:: Optimization Switches The GCC and LLVM back ends have an auto-vectorizer that's enabled by default at some optimization levels. For the GCC back end, it's enabled by default at :switch:`-O3` and you can request it at other levels with :switch:`-ftree-vectorize`. For the LLVM back end, it's enabled by default at lower levels, but you can explicitly enable or disable it with the :switch:`-fno-vectorize`, :switch:`-fvectorize`, :switch:`-fno-slp-vectorize`, and :switch:`-fslp-vectorize` switches. To get auto-vectorization, you also need to make sure that the target architecture features a supported SIMD instruction set. For example, for the x86 architecture, you should at least specify :switch:`-msse2` to get significant vectorization (but you don't need to specify it for x86-64 as it is part of the base 64-bit architecture). Similarly, for the PowerPC architecture, you should specify :switch:`-maltivec`. The preferred loop form for vectorization is the ``for`` iteration scheme. Loops with a ``while`` iteration scheme can also be vectorized if they are very simple, but the vectorizer will quickly give up otherwise. With either iteration scheme, the flow of control must be straight, in particular no ``exit`` statement may appear in the loop body. The loop may however contain a single nested loop, if it can be vectorized when considered alone: .. code-block:: ada A : array (1..4, 1..4) of Long_Float; S : array (1..4) of Long_Float; procedure Sum is begin for I in A'Range(1) loop for J in A'Range(2) loop S (I) := S (I) + A (I, J); end loop; end loop; end Sum; The vectorizable operations depend on the targeted SIMD instruction set, but addition and some multiplication operators are generally supported, as well as the logical operators for modular types. Note that compiling with :switch:`-gnatp` might well reveal cases where some checks do thwart vectorization. Type conversions may also prevent vectorization if they involve semantics that are not directly supported by the code generator or the SIMD instruction set. A typical example is direct conversion from floating-point to integer types. The solution in this case is to use the following idiom: .. code-block:: ada Integer (S'Truncation (F)) if ``S`` is the subtype of floating-point object ``F``. In most cases, the vectorizable loops are loops that iterate over arrays. All kinds of array types are supported, i.e. constrained array types with static bounds: .. code-block:: ada type Array_Type is array (1 .. 4) of Long_Float; constrained array types with dynamic bounds: .. code-block:: ada type Array_Type is array (1 .. Q.N) of Long_Float; type Array_Type is array (Q.K .. 4) of Long_Float; type Array_Type is array (Q.K .. Q.N) of Long_Float; or unconstrained array types: .. code-block:: ada type Array_Type is array (Positive range <>) of Long_Float; The quality of the generated code decreases when the dynamic aspect of the array type increases, the worst code being generated for unconstrained array types. This is because the less information the compiler has about the bounds of the array, the more fallback code it needs to generate in order to fix things up at run time. You can specify that a given loop should be subject to vectorization preferably to other optimizations by means of pragma ``Loop_Optimize``: .. code-block:: ada pragma Loop_Optimize (Vector); placed immediately within the loop will convey the appropriate hint to the compiler for this loop. This is currently only supported for the GCC back end. You can also help the compiler generate better vectorized code for a given loop by asserting that there are no loop-carried dependencies in the loop. Consider for example the procedure: .. code-block:: ada type Arr is array (1 .. 4) of Long_Float; procedure Add (X, Y : not null access Arr; R : not null access Arr) is begin for I in Arr'Range loop R(I) := X(I) + Y(I); end loop; end; By default, the compiler cannot unconditionally vectorize the loop because assigning to a component of the array designated by R in one iteration could change the value read from the components of the array designated by X or Y in a later iteration. As a result, the compiler will generate two versions of the loop in the object code, one vectorized and the other not vectorized, as well as a test to select the appropriate version at run time. This can be overcome by another hint: .. code-block:: ada pragma Loop_Optimize (Ivdep); placed immediately within the loop will tell the compiler that it can safely omit the non-vectorized version of the loop as well as the run-time test. This is also currently only supported by the GCC back end. .. _Other_Optimization_Switches: Other Optimization Switches ^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: Optimization Switches You can also use any specialized optimization switches supported by the back end being used. These switches have not been extensively tested with GNAT but can generally be expected to work. Examples of switches in this category for the GCC back end are :switch:`-funroll-loops` and the various target-specific :switch:`-m` options (in particular, it has been observed that :switch:`-march=xxx` can significantly improve performance on appropriate machines). For full details of these switches, see the *Submodel Options* section in the *Hardware Models and Configurations* chapter of :title:`Using the GNU Compiler Collection (GCC)`. .. _Optimization_and_Strict_Aliasing: Optimization and Strict Aliasing ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: Aliasing .. index:: Strict Aliasing .. index:: No_Strict_Aliasing The strong typing capabilities of Ada allow an optimizer to generate efficient code in situations where other languages would be forced to make worst case assumptions preventing such optimizations. Consider the following example: .. code-block:: ada procedure M is type Int1 is new Integer; I1 : Int1; type Int2 is new Integer; type A2 is access Int2; V2 : A2; ... begin ... for J in Data'Range loop if Data (J) = I1 then V2.all := V2.all + 1; end if; end loop; ... end; Here, since ``V2`` can only access objects of type ``Int2`` and ``I1`` is not one of them, there is no possibility that the assignment to ``V2.all`` affects the value of ``I1``. This means that the compiler optimizer can infer that the value ``I1`` is constant for all iterations of the loop and load it from memory only once, before entering the loop, instead of in every iteration (this is called load hoisting). This kind of optimizations, based on strict type-based aliasing, is triggered by specifying an optimization level of :switch:`-O2` or higher (or :switch:`-Os`) for the GCC back end and :switch:`-O1` or higher for the LLVM back end and allows the compiler to generate more efficient code. However, although this optimization is always correct in terms of the formal semantics of the Ada Reference Manual, you can run into difficulties arise if you use features like ``Unchecked_Conversion`` to break the typing system. Consider the following complete program example: .. code-block:: ada package P1 is type Int1 is new Integer; type A1 is access Int1; type Int2 is new Integer; type A2 is access Int2; end P1; with P1; use P1; package P2 is function To_A2 (Input : A1) return A2; end p2; with Ada.Unchecked_Conversion; package body P2 is function To_A2 (Input : A1) return A2 is function Conv is new Ada.Unchecked_Conversion (A1, A2); begin return Conv (Input); end To_A2; end P2; with P1; use P1; with P2; use P2; with Text_IO; use Text_IO; procedure M is V1 : A1 := new Int1; V2 : A2 := To_A2 (V1); begin V1.all := 1; V2.all := 0; Put_Line (Int1'Image (V1.all)); end; This program prints out ``0`` in :switch:`-O0` mode, but it prints out ``1`` in :switch:`-O2` mode. That's because in strict aliasing mode, the compiler may and does assume that the assignment to ``V2.all`` could not affect the value of ``V1.all``, since different types are involved. This behavior is not a case of non-conformance with the standard, since the Ada RM specifies that an unchecked conversion where the resulting bit pattern is not a correct value of the target type can result in an abnormal value and attempting to reference an abnormal value makes the execution of a program erroneous. That's the case here since the result does not point to an object of type ``Int2``. This means that the effect is entirely unpredictable. However, although that explanation may satisfy a language lawyer, in practice, you probably expect an unchecked conversion involving pointers to create true aliases and the behavior of printing ``1`` is questionable. In this case, the strict type-based aliasing optimizations are clearly unwelcome. Indeed, the compiler recognizes this possibility and the instantiation of Unchecked_Conversion generates a warning: :: p2.adb:5:07: warning: possible aliasing problem with type "A2" p2.adb:5:07: warning: use -fno-strict-aliasing switch for references p2.adb:5:07: warning: or use "pragma No_Strict_Aliasing (A2);" Unfortunately the problem is only recognized when compiling the body of package ``P2``, but the actual problematic code is generated while compiling the body of ``M`` and this latter compilation does not see the suspicious instance of ``Unchecked_Conversion``. As implied by the warning message, there are approaches you can use to avoid the unwanted strict aliasing optimizations in a case like this. One possibility is to simply avoid the use of higher levels of optimization, but that is quite drastic, since it throws away a number of useful optimizations that don't involve strict aliasing assumptions. A less drastic approach is for you to compile the program using the :switch:`-fno-strict-aliasing` switch. Actually, it is only the unit containing the dereferencing of the suspicious pointer that you need to compile with that switch. So, in this case, if you compile unit ``M`` with this switch, you get the expected value of ``0`` printed. Analyzing which units might need the switch can be painful, so you may find it a more reasonable approach is to compile the entire program with options :switch:`-O2` and :switch:`-fno-strict-aliasing`. If you obtain satisfactory performance with this combination of options, then the advantage is that you have avoided the entire issue of possible problematic optimizations due to strict aliasing. To avoid the use of compiler switches, you may use the configuration pragma ``No_Strict_Aliasing`` with no parameters to specify that for all access types, the strict aliasing optimizations should be suppressed. However, these approaches are still overkill, in that they cause all manipulations of all access values to be deoptimized. A more refined approach is to concentrate attention on the specific access type identified as problematic. The first possibility is to move the instantiation of unchecked conversion to the unit in which the type is declared. In this example, you would move the instantiation of ``Unchecked_Conversion`` from the body of package ``P2`` to the spec of package ``P1``. Now, the warning disappears because any use of the access type knows there is a suspicious unchecked conversion and the strict aliasing optimizations are automatically suppressed for it. If it's not practical to move the unchecked conversion to the same unit in which the destination access type is declared (perhaps because the source type is not visible in that unit), the second possibility is for you to use pragma ``No_Strict_Aliasing`` for the type. You must place this pragma in the same declarative part as the declaration of the access type: .. code-block:: ada type A2 is access Int2; pragma No_Strict_Aliasing (A2); Here again, the compiler now knows that strict aliasing optimizations should be suppressed for any dereference made through type ``A2`` and the expected behavior is obtained. The third possibility is to declare that one of the designated types involved, namely ``Int1`` or ``Int2``, is allowed to alias any other type in the universe, by using pragma ``Universal_Aliasing``: .. code-block:: ada type Int2 is new Integer; pragma Universal_Aliasing (Int2); The effect is equivalent to applying pragma ``No_Strict_Aliasing`` to every access type designating ``Int2``, in particular ``A2``, and, more generally, to every reference made to an object of declared type ``Int2``, so it's very powerful and effectively takes ``Int2`` out of the alias analysis performed by the compiler in all circumstances. You can also use this pragma used to deal with aliasing issues that arise from the use of ``Unchecked_Conversion`` in the source code but without the presence of access types. The typical example is code that streams data by means of arrays of storage units (bytes): .. code-block:: ada type Byte is mod 2**System.Storage_Unit; for Byte'Size use System.Storage_Unit; type Chunk_Of_Bytes is array (1 .. 64) of Byte; procedure Send (S : Chunk_Of_Bytes); type Rec is record ... end record; procedure Dump (R : Rec) is function To_Stream is new Ada.Unchecked_Conversion (Rec, Chunk_Of_Bytes); begin Send (To_Stream (R)); end; This generates the following warning for the call to ``Send``: :: dump.adb:8:25: warning: unchecked conversion implemented by copy dump.adb:8:25: warning: use pragma Universal_Aliasing on either type dump.adb:8:25: warning: to enable RM 13.9(12) implementation permission This occurs because the formal parameter ``S`` of ``Send`` is passed by reference by the compiler and it's not possible to pass a reference to ``R`` directly in the call without violating strict type-based aliasing. That's why the compiler generates a temporary of type ``Chunk_Of_Bytes`` just before the call and passes a reference to this temporary instead. As implied by the warning message, you can avoid the temporary (and the warning) by means of pragma ``Universal_Aliasing``: .. code-block:: ada type Chunk_Of_Bytes is array (1 .. 64) of Byte; pragma Universal_Aliasing (Chunk_Of_Bytes); You can also apply this pragma to the component type instead: .. code-block:: ada type Byte is mod 2**System.Storage_Unit; for Byte'Size use System.Storage_Unit; pragma Universal_Aliasing (Byte); and every array type whose component is ``Byte`` will inherit the pragma. To summarize, the alias analysis performed in strict aliasing mode by the compiler can have significant benefits. We've seen cases of large scale application code where the execution time is increased by up to 5% when these optimizations are turned off. However, if you have code that make significant use of unchecked conversion, you might want to just stick with :switch:`-O1` (with the GCC back end) and avoid the entire issue. If you get adequate performance at this level of optimization, that's probably the safest approach. If tests show that you really need higher levels of optimization, then you can experiment with :switch:`-O2` and :switch:`-O2 -fno-strict-aliasing` to see how much effect this has on size and speed of the code. If you really need to use :switch:`-O2` with strict aliasing in effect, then you should review any uses of unchecked conversion, particularly if you are getting the warnings described above. .. _Aliased_Variables_and_Optimization: Aliased Variables and Optimization ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: Aliasing There are scenarios in which your programs may use low level techniques to modify variables that otherwise might be considered to be unassigned. For example, you can pass a variable to a procedure by reference by taking the address of the parameter and using that address to modify the variable's value, even though the address is passed as an ``in`` parameter. Consider the following example: .. code-block:: ada procedure P is Max_Length : constant Natural := 16; type Char_Ptr is access all Character; procedure Get_String(Buffer: Char_Ptr; Size : Integer); pragma Import (C, Get_String, "get_string"); Name : aliased String (1 .. Max_Length) := (others => ' '); Temp : Char_Ptr; function Addr (S : String) return Char_Ptr is function To_Char_Ptr is new Ada.Unchecked_Conversion (System.Address, Char_Ptr); begin return To_Char_Ptr (S (S'First)'Address); end; begin Temp := Addr (Name); Get_String (Temp, Max_Length); end; where Get_String is a C function that uses the address in ``Temp`` to modify the variable ``Name``. This code is dubious, and arguably erroneous, and the compiler is entitled to assume that ``Name`` is never modified, and generate code accordingly. However, in practice, this could cause some existing code that seems to work with no optimization to start failing at higher levels of optimization. What the compiler does for such cases, is to assume that marking a variable as aliased indicates that some "funny business" may be going on. The optimizer recognizes the ``aliased`` keyword and inhibits any optimizations that assume the variable cannot be assigned to. This means that the above example will in fact "work" reliably, that is, it will produce the expected results. However, you should nevertheless avoid code such as this if possible because it's not portable and may not function as you expect with all compilers. .. _Atomic_Variables_and_Optimization: Atomic Variables and Optimization ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: Atomic You need to take two things into consideration with regard to performance when you use atomic variables. First, the RM only guarantees that access to atomic variables be atomic, but has nothing to say about how this is achieved, though there is a strong implication that this should not be achieved by explicit locking code. Indeed, GNAT never generates any locking code for atomic variable access; it will simply reject any attempt to make a variable or type atomic if the atomic access cannot be achieved without such locking code. That being said, it's important to understand that you cannot assume the the program will always access the entire variable. Consider this example: .. code-block:: ada type R is record A,B,C,D : Character; end record; for R'Size use 32; for R'Alignment use 4; RV : R; pragma Atomic (RV); X : Character; ... X := RV.B; You cannot assume that the reference to ``RV.B`` will read the entire 32-bit variable with a single load instruction. It is perfectly legitimate, if the hardware allows it, to do a byte read of just the ``B`` field. This read is still atomic, which is all the RM requires. GNAT can and does take advantage of this, depending on the architecture and optimization level. Any assumption to the contrary is non-portable and risky. Even if you examine the assembly language and see a full 32-bit load, this might change in a future version of the compiler. If your application requires that all accesses to ``RV`` in this example be full 32-bit loads, you need to make a copy for the access as in: .. code-block:: ada declare RV_Copy : constant R := RV; begin X := RV_Copy.B; end; Now the reference to ``RV`` must read the whole variable. Actually, one can imagine some compiler which figures out that the whole copy is not required (because only the ``B`` field is actually accessed), but GNAT certainly won't do that, and we don't know of any compiler that would not handle this right, and the above code will in practice work portably across all architectures (that permit the Atomic declaration). The second issue with atomic variables has to do with the possible requirement of generating synchronization code. For more details on this, consult the sections on the pragmas Enable/Disable_Atomic_Synchronization in the :title:``GNAT Reference Manual``. If performance is critical, and such synchronization code is not required, you may find it useful to disable it. .. _Passive_Task_Optimization: Passive Task Optimization ^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: Passive Task A passive task is one which is sufficiently simple that, in theory, a compiler could recognize it and implement it efficiently without creating a new thread. The original design of Ada 83 had in mind this kind of passive task optimization, but only a few Ada 83 compilers attempted it. The reason was that it was difficult to determine the exact conditions under which the optimization was possible. The result is a very fragile optimization where a very minor change in the program can suddenly silently make a task non-optimizable. With the revisiting of this issue in Ada 95, there was general agreement that this approach was fundamentally flawed and the notion of protected types was introduced. When using protected types, the restrictions are well defined, you KNOW that the operations will be optimized, and furthermore this optimized performance is fully portable. Although it would theoretically be possible for GNAT to attempt to do this optimization, it really doesn't make sense in the context of Ada 95 and none of the Ada 95 compilers implement this optimization as far as we know. GNAT never attempts to perform this optimization. In any new Ada 95 code that you write, you should always use protected types in place of tasks that might be able to be optimized in this manner. Of course, this does not help if you have legacy Ada 83 code that depends on this optimization, but it is unusual to encounter a case where the performance gains from this optimization are significant. Your program should work correctly without this optimization. If you have performance problems, the most practical approach is to figure out exactly where these performance problems arise and update those particular tasks to be protected types. Note that typically clients of the tasks who call entries will not have to be modified, only the task definitions themselves. .. _Text_IO_Suggestions: ``Text_IO`` Suggestions ----------------------- .. index:: Text_IO and performance The ``Ada.Text_IO`` package has fairly high overhead due in part to the requirement of maintaining page and line counts. If performance is critical, one recommendation is to use ``Stream_IO`` instead of ``Text_IO`` for large-volume output, since it has less overhead. If you must use ``Text_IO``, note that output to the standard output and standard error files is unbuffered by default (this provides better behavior when output statements are used for debugging or if the progress of a program is observed by tracking the output, e.g. by using the Unix *tail -f* command to watch redirected output). If you're generating large volumes of output with ``Text_IO`` and performance is an important factor, use a designated file instead of the standard output file or change the standard output file to be buffered using ``Interfaces.C_Streams.setvbuf``. .. _Reducing_Size_of_Executables_with_Unused_Subprogram/Data_Elimination: Reducing Size of Executables with Unused Subprogram/Data Elimination -------------------------------------------------------------------- .. index:: Uunused subprogram/data elimination This section describes how you can eliminate unused subprograms and data from your executable just by setting options at compilation time. .. _About_unused_subprogram/data_elimination: About unused subprogram/data elimination ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ By default, an executable contains all code and data of its objects (directly linked or coming from statically linked libraries), even data or code never used by this executable. This feature eliminates such unused code from your executable, thus making it smaller (in disk and in memory). You can use this functionality on all Linux platforms except for the IA-64 architecture and on all cross platforms using the ELF binary file format. In both cases, GNU binutils version 2.16 or later are required to enable it. .. _Compilation_options: Compilation options ^^^^^^^^^^^^^^^^^^^ The operation of eliminating the unused code and data from the final executable is directly performed by the linker. .. index:: -ffunction-sections (gcc) .. index:: -fdata-sections (gcc) In order to do this, it has to work with objects compiled with the following switches passed to the GCC back end: :switch:`-ffunction-sections` :switch:`-fdata-sections`. These options are usable with C and Ada files. They cause the compiler to place each function or data in a separate section in the resulting object file. Once you've created the objects and static libraries with these switches, the linker can perform the dead code elimination. You can do this by specifying the :switch:`-Wl,--gc-sections` switch to your ``gcc`` command or in the :switch:`-largs` section of your invocation of ``gnatmake``. This causes the linker to perform a garbage collection and remove code and data that are never referenced. If the linker performs a partial link (:switch:`-r` linker switch), then you need to provide the entry point using the :switch:`-e` / :switch:`--entry` linker switch. Note that objects compiled without the :switch:`-ffunction-sections` and :switch:`-fdata-sections` options can still be linked with the executable. However, no dead code elimination can be performed on those objects (they will be linked as is). The GNAT static library is compiled with :switch:`-ffunction-sections` and :switch:`-fdata-sections` on some platforms. This allows you to eliminate the unused code and data of the GNAT library from your executable. .. _Example_of_unused_subprogram/data_elimination: Example of unused subprogram/data elimination ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Here's a simple example: .. code-block:: ada with Aux; procedure Test is begin Aux.Used (10); end Test; package Aux is Used_Data : Integer; Unused_Data : Integer; procedure Used (Data : Integer); procedure Unused (Data : Integer); end Aux; package body Aux is procedure Used (Data : Integer) is begin Used_Data := Data; end Used; procedure Unused (Data : Integer) is begin Unused_Data := Data; end Unused; end Aux; ``Unused`` and ``Unused_Data`` are never referenced in this code excerpt and hence may be safely removed from the final executable. :: $ gnatmake test $ nm test | grep used 020015f0 T aux__unused 02005d88 B aux__unused_data 020015cc T aux__used 02005d84 B aux__used_data $ gnatmake test -cargs -fdata-sections -ffunction-sections \\ -largs -Wl,--gc-sections $ nm test | grep used 02005350 T aux__used 0201ffe0 B aux__used_data You can see that the procedure ``Unused`` and the object ``Unused_Data`` are removed by the linker when you've used the appropriate switches. .. index:: Overflow checks .. index:: Checks (overflow) .. _Overflow_Check_Handling_in_GNAT: Overflow Check Handling in GNAT =============================== This section explains how to control the handling of overflow checks. .. _Background: Background ---------- Overflow checks are checks that the compiler may make to ensure that intermediate results are not out of range. For example: .. code-block:: ada A : Integer; ... A := A + 1; If ``A`` has the value ``Integer'Last``, the addition will cause overflow since the result is out of range of the type ``Integer``. In this case, execution will raise ``Constraint_Error`` if checks are enabled. A trickier situation arises in cases like the following: .. code-block:: ada A, C : Integer; ... A := (A + 1) + C; where ``A`` is ``Integer'Last`` and ``C`` is ``-1``. Here, the final result of the expression on the right hand side is ``Integer'Last`` which is in range, but the question arises whether the intermediate addition of ``(A + 1)`` raises an overflow error. The (perhaps surprising) answer is that the Ada language definition does not answer this question. Instead, it leaves it up to the implementation to do one of two things if overflow checks are enabled. * raise an exception (``Constraint_Error``), or * yield the correct mathematical result which is then used in subsequent operations. If the compiler chooses the first approach, the execution of this example will indeed raise ``Constraint_Error`` if overflow checking is enabled or result in erroneous execution if overflow checks are suppressed. But if the compiler chooses the second approach, it can perform both additions yielding the correct mathematical result, which is in range, so no exception is raised and the right result is obtained, regardless of whether overflow checks are suppressed. Note that in the first example, an exception will be raised in either case, since if the compiler gives the correct mathematical result for the addition, it will be out of range of the target type of the assignment and thus fails the range check. This lack of specified behavior in the handling of overflow for intermediate results is a source of non-portability and can thus be problematic when you port programs. Most typically, this arises in a situation where the original compiler did not raise an exception and you move the application to a compiler where the check is performed on the intermediate result and an unexpected exception is raised. Furthermore, when using Ada 2012's preconditions and other assertion forms, another issue arises. Consider: .. code-block:: ada procedure P (A, B : Integer) with Pre => A + B <= Integer'Last; We often want to regard arithmetic in a context such as this from a purely mathematical point of view. So, for example, if the two actual parameters for a call to ``P`` are both ``Integer'Last`` then the precondition should be evaluated as ``False``. If we're executing in a mode with run-time checks enabled for preconditions, then we would like this precondition to fail, rather than raising an exception because of the intermediate overflow. However, the language definition leaves the specification of whether the above condition fails (raising ``Assert_Error``) or causes an intermediate overflow (raising ``Constraint_Error``) up to the implementation. The situation is worse in a case such as the following: .. code-block:: ada procedure Q (A, B, C : Integer) with Pre => A + B + C <= Integer'Last; Consider the call .. code-block:: ada Q (A => Integer'Last, B => 1, C => -1); From a mathematical point of view the precondition is ``True``, but at run time we may (but are not guaranteed to) get an exception raised because of the intermediate overflow (and we really would prefer this precondition to be considered ``True`` at run time). .. _Management_of_Overflows_in_GNAT: Management of Overflows in GNAT ------------------------------- To deal with the portability issue and with the problem of mathematical versus run-time interpretation of the expressions in assertions, GNAT provides comprehensive control over the handling of intermediate overflows. It can operate in three modes, and in addition, permits separate selection of operating modes for the expressions within assertions (here the term 'assertions' is used in the technical sense, which includes preconditions and so forth) and for expressions appearing outside assertions. The three modes are: * *Use base type for intermediate operations* (``STRICT``) In this mode, all intermediate results for predefined arithmetic operators are computed using the base type, and the result must be in range of the base type. If this is not the case, then either an exception is raised (if overflow checks are enabled) or the execution is erroneous (if overflow checks are suppressed). This is the normal default mode. * *Most intermediate overflows avoided* (``MINIMIZED``) In this mode, the compiler attempts to avoid intermediate overflows by using a larger integer type, typically ``Long_Long_Integer``, as the type in which arithmetic is performed for predefined arithmetic operators. This may be slightly more expensive at run time (compared to suppressing intermediate overflow checks), though the cost is negligible on modern 64-bit machines. For the examples given earlier, no intermediate overflows would have resulted in exceptions, since the intermediate results are all in the range of ``Long_Long_Integer`` (typically 64-bits on nearly all implementations of GNAT). In addition, if checks are enabled, this reduces the number of checks that must be made, so this choice may actually result in an improvement in space and time behavior. However, there are cases where ``Long_Long_Integer`` is not large enough. Consider the following example: .. code-block:: ada procedure R (A, B, C, D : Integer) with Pre => (A**2 * B**2) / (C**2 * D**2) <= 10; where ``A`` = ``B`` = ``C`` = ``D`` = ``Integer'Last``. Now the intermediate results are out of the range of ``Long_Long_Integer`` even though the final result is in range and the precondition is ``True`` from a mathematical point of view. In such a case, operating in this mode, an overflow occurs for the intermediate computation (which is why this mode says *most* intermediate overflows are avoided). In this case, an exception is raised if overflow checks are enabled, and the execution is erroneous if overflow checks are suppressed. * *All intermediate overflows avoided* (``ELIMINATED``) In this mode, the compiler avoids all intermediate overflows by using arbitrary precision arithmetic as required. In this mode, the above example with ``A**2 * B**2`` would not cause intermediate overflow, because the intermediate result would be evaluated using sufficient precision, and the result of evaluating the precondition would be ``True``. This mode has the advantage of avoiding any intermediate overflows, but at the expense of significant run-time overhead, including the use of a library (included automatically in this mode) for multiple-precision arithmetic. This mode provides cleaner semantics for assertions, since now the run-time behavior emulates true arithmetic behavior for the predefined arithmetic operators, meaning that there is never a conflict between the mathematical view of the assertion and its run-time behavior. Note that in this mode, the behavior is unaffected by whether or not overflow checks are suppressed, since overflow does not occur. Gigantic intermediate expressions can still raise ``Storage_Error`` as a result of attempting to compute the results of such expressions (e.g. ``Integer'Last ** Integer'Last``) but overflow is impossible. Note that these modes apply only to the evaluation of predefined arithmetic, membership, and comparison operators for signed integer arithmetic. For fixed-point arithmetic, you suppress checks. But if checks are enabled, fixed-point values are always checked for overflow against the base type for intermediate expressions (i.e., such checks always operate in the equivalent of ``STRICT`` mode). For floating-point, on nearly all architectures, ``Machine_Overflows`` is ``False``, and IEEE infinities are generated, so overflow exceptions are never raised. If you want to avoid infinities and check that final results of expressions are in range, you can declare a constrained floating-point type and range checks are carried out in the normal manner (with infinite values always failing all range checks). .. _Specifying_the_Desired_Mode: Specifying the Desired Mode --------------------------- .. index:: pragma Overflow_Mode You can specify the desired mode of for handling intermediate overflow using either the ``Overflow_Mode`` pragma or an equivalent compiler switch. The pragma has the form: .. code-block:: ada pragma Overflow_Mode ([General =>] MODE [, [Assertions =>] MODE]); where ``MODE`` is one of * ``STRICT``: intermediate overflows checked (using base type) * ``MINIMIZED``: minimize intermediate overflows * ``ELIMINATED``: eliminate intermediate overflows The case is ignored, so ``MINIMIZED``, ``Minimized`` and ``minimized`` all have the same effect. If you only specify the ``General`` parameter, the given ``MODE`` applies to expressions both within and outside assertions. If you specify both arguments, the value of ``General`` applies to expressions outside assertions, and ``Assertions`` applies to expressions within assertions. For example: .. code-block:: ada pragma Overflow_Mode (General => Minimized, Assertions => Eliminated); specifies that expressions outside assertions be evaluated in 'minimize intermediate overflows' mode and expressions within assertions be evaluated in 'eliminate intermediate overflows' mode. This is often a reasonable choice, avoiding excessive overhead outside assertions, but assuring a high degree of portability when importing code from another compiler while incurring the extra overhead for assertion expressions to ensure that the behavior at run time matches the expected mathematical behavior. The ``Overflow_Mode`` pragma has the same scoping and placement rules as pragma ``Suppress``, so you can use it either as a configuration pragma, specifying a default for the whole program, or in a declarative scope, where it applies to the remaining declarations and statements in that scope. Note that pragma ``Overflow_Mode`` does not affect whether overflow checks are enabled or suppressed. It only controls the method used to compute intermediate values. To control whether overflow checking is enabled or suppressed, use pragma ``Suppress`` or ``Unsuppress`` in the usual manner. .. index:: -gnato? (gcc) .. index:: -gnato?? (gcc) Additionally, you can use the compiler switch :switch:`-gnato?` or :switch:`-gnato??` to control the checking mode default (which you can subsequently override using the above pragmas). Here ``?`` is one of the digits ``1`` through ``3``: ====== ====================================================== ``1`` use base type for intermediate operations (``STRICT``) ``2`` minimize intermediate overflows (``MINIMIZED``) ``3`` eliminate intermediate overflows (``ELIMINATED``) ====== ====================================================== As with the pragma, if only one digit appears, it applies to all cases; if two digits are given, the first applies to expressions outside assertions and the second within assertions. Thus the equivalent of the example pragma above would be :switch:`-gnato23`. If you don't provide any digits following the :switch:`-gnato`, it's equivalent to :switch:`-gnato11`, causing all intermediate operations to be computed using the base type (``STRICT`` mode). .. _Default_Settings: Default Settings ---------------- The default mode for overflow checks is :: General => Strict which causes all computations both inside and outside assertions to use the base type, and is equivalent to :switch:`-gnato` (with no digits following). The pragma ``Suppress (Overflow_Check)`` disables overflow checking but has no effect on the method used for computing intermediate results. The pragma ``Unsuppress (Overflow_Check)`` enables overflow checking but has no effect on the method used for computing intermediate results. .. _Implementation_Notes: Implementation Notes -------------------- In practice, on typical 64-bit machines, the ``MINIMIZED`` mode is reasonably efficient and you can generally use it. It also helps to ensure compatibility with code imported from other compilers to GNAT. Setting all intermediate overflows checking (``STRICT`` mode) makes sense if you want to make sure your code is compatible with any other Ada implementations. You may find this useful in ensuring portability for code that is to be exported to some other compiler than GNAT. The Ada standard allows the reassociation of expressions at the same precedence level if no parentheses are present. For example, ``A+B+C`` parses as though it were ``(A+B)+C``, but the compiler can reinterpret this as ``A+(B+C)``, possibly introducing or eliminating an overflow exception. The GNAT compiler never takes advantage of this freedom, and the expression ``A+B+C`` will be evaluated as ``(A+B)+C``. If you need the other order, you can write the parentheses explicitly ``A+(B+C)`` and GNAT will respect this order. The use of ``ELIMINATED`` mode will cause the compiler to automatically include an appropriate arbitrary precision integer arithmetic package. The compiler will make calls to this package, though only in cases where it cannot be sure that ``Long_Long_Integer`` is sufficient to guard against intermediate overflows. This package does not use dynamic allocation, but it does use the secondary stack, so an appropriate secondary stack package must be present (this is always true for standard full Ada, but may require specific steps for restricted run times such as ZFP). Although ``ELIMINATED`` mode causes expressions to use arbitrary precision arithmetic, avoiding overflow, the final result must be in an appropriate range. This is true even if the final result is of type ``[Long_[Long_]]Integer'Base``, which still has the same bounds as its associated constrained type at run-time. Currently, the ``ELIMINATED`` mode is only available on target platforms for which ``Long_Long_Integer`` is at least 64-bits (nearly all GNAT platforms). .. _Performing_Dimensionality_Analysis_in_GNAT: Performing Dimensionality Analysis in GNAT ========================================== .. index:: Dimensionality analysis The GNAT compiler supports dimensionality checking. You can specify physical units for objects and the compiler verifies that uses of these objects are compatible with their dimension, in a fashion that is familiar to engineering practice. The dimensions of algebraic expressions (including powers with static exponents) are computed from their constituents. .. index:: Dimension_System aspect .. index:: Dimension aspect This feature depends on Ada 2012 aspect specifications and is available for versions 7.0.1 and later of GNAT. The GNAT-specific aspect ``Dimension_System`` allows you to define a system of units; the aspect ``Dimension`` allows you to declare dimensioned quantities within a given system. (These aspects are described in the *Implementation Defined Aspects* chapter of the :title:`GNAT Reference Manual`). The major advantage of this model is that it does not require the declaration of multiple operators for all possible combinations of types: you is only need to use the proper subtypes in object declarations. .. index:: System.Dim.Mks package (GNAT library) .. index:: MKS_Type type The simplest way to impose dimensionality checking on a computation is to make use of one of the instantiations of the package ``System.Dim.Generic_Mks``, which is part of the GNAT library. This generic package defines a floating-point type ``MKS_Type``, for which a sequence of dimension names are specified, together with their conventional abbreviations. You should read the following together with the full specification of the package, in file :file:`s-digemk.ads`. .. index:: s-digemk.ads file .. code-block:: ada type Mks_Type is new Float_Type with Dimension_System => ( (Unit_Name => Meter, Unit_Symbol => 'm', Dim_Symbol => 'L'), (Unit_Name => Kilogram, Unit_Symbol => "kg", Dim_Symbol => 'M'), (Unit_Name => Second, Unit_Symbol => 's', Dim_Symbol => 'T'), (Unit_Name => Ampere, Unit_Symbol => 'A', Dim_Symbol => 'I'), (Unit_Name => Kelvin, Unit_Symbol => 'K', Dim_Symbol => "Theta"), (Unit_Name => Mole, Unit_Symbol => "mol", Dim_Symbol => 'N'), (Unit_Name => Candela, Unit_Symbol => "cd", Dim_Symbol => 'J')); The package then defines a series of subtypes that correspond to these conventional units. For example: .. code-block:: ada subtype Length is Mks_Type with Dimension => (Symbol => 'm', Meter => 1, others => 0); and similarly for ``Mass``, ``Time``, ``Electric_Current``, ``Thermodynamic_Temperature``, ``Amount_Of_Substance``, and ``Luminous_Intensity`` (the standard set of units of the SI system). The package also defines conventional names for values of each unit, for example: .. code-block:: ada m : constant Length := 1.0; kg : constant Mass := 1.0; s : constant Time := 1.0; A : constant Electric_Current := 1.0; as well as useful multiples of these units: .. code-block:: ada cm : constant Length := 1.0E-02; g : constant Mass := 1.0E-03; min : constant Time := 60.0; day : constant Time := 60.0 * 24.0 * min; ... There are three instantiations of ``System.Dim.Generic_Mks`` defined in the GNAT library: * ``System.Dim.Float_Mks`` based on ``Float`` defined in :file:`s-diflmk.ads`. * ``System.Dim.Long_Mks`` based on ``Long_Float`` defined in :file:`s-dilomk.ads`. * ``System.Dim.Mks`` based on ``Long_Long_Float`` defined in :file:`s-dimmks.ads`. Using one of these packages, you can then define a derived unit by providing the aspect that specifies its dimensions within the MKS system as well as the string to be used for output of a value of that unit: .. code-block:: ada subtype Acceleration is Mks_Type with Dimension => ("m/sec^2", Meter => 1, Second => -2, others => 0); Here's a complete example: .. code-block:: ada with System.Dim.MKS; use System.Dim.Mks; with System.Dim.Mks_IO; use System.Dim.Mks_IO; with Text_IO; use Text_IO; procedure Free_Fall is subtype Acceleration is Mks_Type with Dimension => ("m/sec^2", 1, 0, -2, others => 0); G : constant acceleration := 9.81 * m / (s ** 2); T : Time := 10.0*s; Distance : Length; begin Put ("Gravitational constant: "); Put (G, Aft => 2, Exp => 0); Put_Line (""); Distance := 0.5 * G * T ** 2; Put ("distance travelled in 10 seconds of free fall "); Put (Distance, Aft => 2, Exp => 0); Put_Line (""); end Free_Fall; Execution of this program yields: :: Gravitational constant: 9.81 m/sec^2 distance travelled in 10 seconds of free fall 490.50 m However, incorrect assignments such as: .. code-block:: ada Distance := 5.0; Distance := 5.0 * kg; are rejected with the following diagnoses: :: Distance := 5.0; >>> dimensions mismatch in assignment >>> left-hand side has dimension [L] >>> right-hand side is dimensionless Distance := 5.0 * kg: >>> dimensions mismatch in assignment >>> left-hand side has dimension [L] >>> right-hand side has dimension [M] The dimensions of an expression are properly displayed even if there is no explicit subtype for it. If we add to the program: .. code-block:: ada Put ("Final velocity: "); Put (G * T, Aft =>2, Exp =>0); Put_Line (""); the output includes: :: Final velocity: 98.10 m.s**(-1) .. index:: Dimensionable type .. index:: Dimensioned subtype The type ``Mks_Type`` is said to be a *dimensionable type* since it has a ``Dimension_System`` aspect, and the subtypes ``Length``, ``Mass``, etc., are said to be *dimensioned subtypes* since each one has a ``Dimension`` aspect. .. index:: Dimension Vector (for a dimensioned subtype) .. index:: Dimension aspect .. index:: Dimension_System aspect The ``Dimension`` aspect of a dimensioned subtype ``S`` defines a mapping from the base type's Unit_Names to integer (or, more generally, rational) values. This mapping is the *dimension vector* (also referred to as the *dimensionality*) for that subtype, denoted by ``DV(S)``, and thus for each object of that subtype. Intuitively, the value specified for each ``Unit_Name`` is the exponent associated with that unit; a zero value means that the unit is not used. For example: .. code-block:: ada declare Acc : Acceleration; ... begin ... end; Here ``DV(Acc)`` = ``DV(Acceleration)`` = ``(Meter=>1, Kilogram=>0, Second=>-2, Ampere=>0, Kelvin=>0, Mole=>0, Candela=>0)``. Symbolically, we can express this as ``Meter / Second**2``. The dimension vector of an arithmetic expression is synthesized from the dimension vectors of its components, with compile-time dimensionality checks that help prevent mismatches such as using an ``Acceleration`` where a ``Length`` is required. The dimension vector of the result of an arithmetic expression *expr*, or :samp:`DV({expr})`, is defined as follows, assuming conventional mathematical definitions for the vector operations that are used: * If *expr* is of the type *universal_real*, or is not of a dimensioned subtype, then *expr* is dimensionless; :samp:`DV({expr})` is the empty vector. * :samp:`DV({op expr})`, where *op* is a unary operator, is :samp:`DV({expr})` * :samp:`DV({expr1 op expr2})`, where *op* is "+" or "-", is :samp:`DV({expr1})` provided that :samp:`DV({expr1})` = :samp:`DV({expr2})`. If this condition is not met then the construct is illegal. * :samp:`DV({expr1} * {expr2})` is :samp:`DV({expr1})` + :samp:`DV({expr2})`, and :samp:`DV({expr1} / {expr2})` = :samp:`DV({expr1})` - :samp:`DV({expr2})`. In this context if one of the *expr*\ s is dimensionless then its empty dimension vector is treated as ``(others => 0)``. * :samp:`DV({expr} ** {power})` is *power* * :samp:`DV({expr})`, provided that *power* is a static rational value. If this condition is not met then the construct is illegal. Note that, by the above rules, it is illegal to use binary "+" or "-" to combine a dimensioned and dimensionless value. Thus an expression such as ``acc-10.0`` is illegal, where ``acc`` is an object of subtype ``Acceleration``. The dimensionality checks for relationals use the same rules as for "+" and "-" except when comparing to a literal; thus .. code-block:: ada acc > len is equivalent to .. code-block:: ada acc-len > 0.0 and is thus illegal, but .. code-block:: ada acc > 10.0 is accepted with a warning. Analogously, a conditional expression requires the same dimension vector for each branch (with no exception for literals). The dimension vector of a type conversion :samp:`T({expr})` is defined as follows, based on the nature of ``T``: * If ``T`` is a dimensioned subtype, then :samp:`DV(T({expr}))` is ``DV(T)`` provided that either *expr* is dimensionless or :samp:`DV(T)` = :samp:`DV({expr})`. The conversion is illegal if *expr* is dimensioned and :samp:`DV({expr})` /= ``DV(T)``. Note that vector equality does not require that the corresponding Unit_Names be the same. As a consequence of the above rule, you can convert between different dimension systems that follow the same international system of units, with the seven physical components given in the standard order (length, mass, time, etc.). Thus, you can convert a length in meters to a length in inches (with a suitable conversion factor) but not, for example, to a mass in pounds. * If ``T`` is the base type for *expr* (and the dimensionless root type of the dimension system), then :samp:`DV(T({expr}))` is ``DV(expr)``. Thus, if *expr* is of a dimensioned subtype of ``T``, the conversion may be regarded as a "view conversion" that preserves dimensionality. This rule means you can write generic code that can be instantiated with compatible dimensioned subtypes. You include in the generic unit conversions that will consequently be present in instantiations, but conversions to the base type will preserve dimensionality and make it possible to write generic code that is correct with respect to dimensionality. * Otherwise (i.e., ``T`` is neither a dimensioned subtype nor a dimensionable base type), :samp:`DV(T({expr}))` is the empty vector. Thus, a dimensioned value can be explicitly converted to a non-dimensioned subtype, which of course then escapes dimensionality analysis. The dimension vector for a type qualification :samp:`T'({expr})` is the same as for the type conversion :samp:`T({expr})`. An assignment statement .. code-block:: ada Source := Target; requires ``DV(Source)`` = ``DV(Target)`` and analogously for parameter passing (the dimension vector for the actual parameter must be equal to the dimension vector for the formal parameter). When using dimensioned types with elementary functions, you need not instantiate the ``Ada.Numerics.Generic_Elementary_Functions`` package using the ``Mks_Type`` nor for any of the derived subtypes such as ``Distance``. For functions such as ``Sqrt``, the dimensional analysis will fail when using the subtypes because both the parameter and return are of the same type. An example instantiation .. code-block:: ada package Mks_Numerics is new Ada.Numerics.Generic_Elementary_Functions (System.Dim.Mks.Mks_Type); .. _Stack_Related_Facilities: Stack Related Facilities ======================== This section describes some useful tools associated with stack checking and analysis. In particular, it deals with dynamic and static stack usage measurements. .. _Stack_Overflow_Checking: Stack Overflow Checking ----------------------- .. index:: Stack Overflow Checking .. index:: -fstack-check (gcc) For most operating systems, ``gcc`` does not perform stack overflow checking by default. This means that if the main environment task or some other task exceeds the available stack space, unpredictable behavior will occur. Most native systems offer some level of protection by adding a guard page at the end of each task stack. This mechanism is usually not enough for dealing properly with stack overflow situations because a large local variable could "jump" above the guard page. Furthermore, when the guard page is hit, there may not be any space left on the stack for executing the exception propagation code. Enabling stack checking avoids such situations. To activate stack checking, compile all units with the ``gcc`` switch :switch:`-fstack-check`. For example: :: $ gcc -c -fstack-check package1.adb Units compiled with this option will generate extra instructions to check that any use of the stack (for procedure calls or for declaring local variables in declare blocks) does not exceed the available stack space. If the space is exceeded, a ``Storage_Error`` exception is raised. For declared tasks, the default stack size is defined by the GNAT runtime, whose size may be modified at bind time through the ``-d`` bind switch (:ref:`Switches_for_gnatbind`). You can set task specific stack sizes using the ``Storage_Size`` pragma. For the environment task, the stack size is determined by the operating system. Consequently, to modify the size of the environment task please refer to your operating system documentation. When using the LLVM back end, this switch doesn't perform full stack overflow checking, but just checks for very large local dynamic allocations. .. _Static_Stack_Usage_Analysis: Static Stack Usage Analysis --------------------------- .. index:: Static Stack Usage Analysis .. index:: -fstack-usage A unit compiled with the :switch:`-fstack-usage` switch generate an extra file that specifies the maximum amount of stack used on a per-function basis. The file has the same basename as the target object file with a :file:`.su` extension. Each line of this file is made up of three fields: * The name of the function. * A number of bytes. * One or more qualifiers: ``static``, ``dynamic``, ``bounded``. The second field corresponds to the size of the known part of the function frame. The qualifier ``static`` means that the function frame size is purely static. It usually means that all local variables have a static size. In this case, the second field is a reliable measure of the function stack utilization. The qualifier ``dynamic`` means that the function frame size is not static. It happens mainly when some local variables have a dynamic size. When this qualifier appears alone, the second field is not a reliable measure of the function stack analysis. When it is qualified with ``bounded``, it means that the second field is a reliable maximum of the function stack utilization. Compilation of a unit with the :switch:`-Wstack-usage` switch will issue a warning for each subprogram whose stack usage might be larger than the specified amount of bytes. The wording of that warning is consistent with that in the file documented above. This is not supported by the LLVM back end. .. _Dynamic_Stack_Usage_Analysis: Dynamic Stack Usage Analysis ---------------------------- You can measure the maximum amount of stack used by a task by adding a switch to ``gnatbind``, as: :: $ gnatbind -u0 file With this option, at each task termination, its stack usage is output on :file:`stderr`. Note that this switch is not compatible with tools like Valgrind and DrMemory; they will report errors. It is not always convenient to output the stack usage when the program is still running. Hence, you can delay this output until the termination of the number of tasks specified as the argument of the :switch:`-u` switch. For example: :: $ gnatbind -u100 file buffers the stack usage information of the first 100 tasks to terminate and outputs it when the program terminates. Results are displayed in four columns: :: Index | Task Name | Stack Size | Stack Usage where: * *Index* is a number associated with each task. * *Task Name* is the name of the task analyzed. * *Stack Size* is the maximum size for the stack. * *Stack Usage* is the measure done by the stack analyzer. In order to prevent overflow, the stack is not entirely analyzed, and it's not possible to know exactly how much has actually been used. By default, ``gnatbind`` does not process the environment task stack, the stack that contains the main unit. To enable processing of the environment task stack, set the environment variable GNAT_STACK_LIMIT to the maximum size of the environment task stack. This amount is given in kilobytes. For example: :: $ set GNAT_STACK_LIMIT 1600 would specify to the analyzer that the environment task stack has a limit of 1.6 megabytes. Any stack usage beyond this will be ignored by the analysis. This is not supported by the LLVM back end. The package ``GNAT.Task_Stack_Usage`` provides facilities to get stack-usage reports at run time. See its body for the details. .. _Memory_Management_Issues: Memory Management Issues ======================== This section describes some useful memory pools provided in the GNAT library, and in particular the GNAT Debug Pool facility, which can be used to detect incorrect uses of access values (including 'dangling references'). .. only:: PRO or GPL It also describes the ``gnatmem`` tool, which can be used to track down "memory leaks". .. _Some_Useful_Memory_Pools: Some Useful Memory Pools ------------------------ .. index:: Memory Pool .. index:: storage, pool The ``System.Pool_Global`` package provides the ``Unbounded_No_Reclaim_Pool`` storage pool. Allocations use the standard system call ``malloc`` while deallocations use the standard system call ``free``. No reclamation is performed when the pool goes out of scope. For performance reasons, the standard default Ada allocators/deallocators do not use any explicit storage pools but if they did, they could use this storage pool without any change in behavior. That is why this storage pool is used when the user makes the default implicit allocator explicit as in this example: .. code-block:: ada type T1 is access Something; -- no Storage pool is defined for T2 type T2 is access Something_Else; for T2'Storage_Pool use T1'Storage_Pool; -- the above is equivalent to for T2'Storage_Pool use System.Pool_Global.Global_Pool_Object; The ``System.Pool_Local`` package provides the ``Unbounded_Reclaim_Pool`` storage pool. Its allocation strategy is similar to ``Pool_Local`` except that the all storage allocated with this pool is reclaimed when the pool object goes out of scope. This pool provides a explicit mechanism similar to the implicit one provided by several Ada 83 compilers for allocations performed through a local access type and whose purpose was to reclaim memory when exiting the scope of a given local access. As an example, the following program does not leak memory even though it does not perform explicit deallocation: .. code-block:: ada with System.Pool_Local; procedure Pooloc1 is procedure Internal is type A is access Integer; X : System.Pool_Local.Unbounded_Reclaim_Pool; for A'Storage_Pool use X; v : A; begin for I in 1 .. 50 loop v := new Integer; end loop; end Internal; begin for I in 1 .. 100 loop Internal; end loop; end Pooloc1; The ``System.Pool_Size`` package implements the ``Stack_Bounded_Pool`` used when ``Storage_Size`` is specified for an access type. The whole storage for the pool is allocated at once, usually on the stack at the point where the access type is elaborated. It is automatically reclaimed when exiting the scope where the access type is defined. This package is not intended to be used directly by the user; it is implicitly used for each declaration with a specified ``Storage_Size``: .. code-block:: ada type T1 is access Something; for T1'Storage_Size use 10_000; .. _The_GNAT_Debug_Pool_Facility: The GNAT Debug Pool Facility ---------------------------- .. index:: Debug Pool .. index:: storage, pool, memory corruption Using unchecked deallocation and unchecked conversion can easily lead to incorrect memory references. The problems generated by such references are usually difficult to find because the symptoms can be very remote from the origin of the problem. In such cases, it is very helpful to detect the problem as early as possible. This is the purpose of the Storage Pool provided by ``GNAT.Debug_Pools``. In order to use the GNAT specific debugging pool, you must associate a debug pool object with each of the access types that may be related to suspected memory problems. See Ada Reference Manual 13.11. .. code-block:: ada type Ptr is access Some_Type; Pool : GNAT.Debug_Pools.Debug_Pool; for Ptr'Storage_Pool use Pool; ``GNAT.Debug_Pools`` is derived from a GNAT-specific kind of pool: the ``Checked_Pool``. Such pools, like standard Ada storage pools, allow you to redefine allocation and deallocation strategies. They also provide a checkpoint for each dereference through the use of the primitive operation ``Dereference`` which is implicitly called at each dereference of an access value. Once you have associated an access type with a debug pool, operations on values of the type may raise four distinct exceptions, which correspond to four potential kinds of memory corruption: * ``GNAT.Debug_Pools.Accessing_Not_Allocated_Storage`` * ``GNAT.Debug_Pools.Accessing_Deallocated_Storage`` * ``GNAT.Debug_Pools.Freeing_Not_Allocated_Storage`` * ``GNAT.Debug_Pools.Freeing_Deallocated_Storage`` For types associated with a Debug_Pool, dynamic allocation is performed using the standard GNAT allocation routine. References to all allocated chunks of memory are kept in an internal dictionary. Several deallocation strategies are provided, allowing you to choose to release the memory to the system, keep it allocated for further invalid access checks, or fill it with an easily recognizable pattern for debug sessions. The memory pattern is the old IBM hexadecimal convention: ``16#DEADBEEF#``. See the documentation in the file :file:`g-debpoo.ads` for more information on the various strategies. Upon each dereference, a check is made that the access value denotes a properly allocated memory location. Here's a complete example of use of ``Debug_Pools``, which includes typical instances of memory corruption: .. code-block:: ada with GNAT.IO; use GNAT.IO; with Ada.Unchecked_Deallocation; with Ada.Unchecked_Conversion; with GNAT.Debug_Pools; with System.Storage_Elements; with Ada.Exceptions; use Ada.Exceptions; procedure Debug_Pool_Test is type T is access Integer; type U is access all T; P : GNAT.Debug_Pools.Debug_Pool; for T'Storage_Pool use P; procedure Free is new Ada.Unchecked_Deallocation (Integer, T); function UC is new Ada.Unchecked_Conversion (U, T); A, B : aliased T; procedure Info is new GNAT.Debug_Pools.Print_Info(Put_Line); begin Info (P); A := new Integer; B := new Integer; B := A; Info (P); Free (A); begin Put_Line (Integer'Image(B.all)); exception when E : others => Put_Line ("raised: " & Exception_Name (E)); end; begin Free (B); exception when E : others => Put_Line ("raised: " & Exception_Name (E)); end; B := UC(A'Access); begin Put_Line (Integer'Image(B.all)); exception when E : others => Put_Line ("raised: " & Exception_Name (E)); end; begin Free (B); exception when E : others => Put_Line ("raised: " & Exception_Name (E)); end; Info (P); end Debug_Pool_Test; The debug pool mechanism provides the following precise diagnostics on the execution of this erroneous program: :: Debug Pool info: Total allocated bytes : 0 Total deallocated bytes : 0 Current Water Mark: 0 High Water Mark: 0 Debug Pool info: Total allocated bytes : 8 Total deallocated bytes : 0 Current Water Mark: 8 High Water Mark: 8 raised: GNAT.DEBUG_POOLS.ACCESSING_DEALLOCATED_STORAGE raised: GNAT.DEBUG_POOLS.FREEING_DEALLOCATED_STORAGE raised: GNAT.DEBUG_POOLS.ACCESSING_NOT_ALLOCATED_STORAGE raised: GNAT.DEBUG_POOLS.FREEING_NOT_ALLOCATED_STORAGE Debug Pool info: Total allocated bytes : 8 Total deallocated bytes : 4 Current Water Mark: 4 High Water Mark: 8 .. only:: PRO or GPL .. _The_gnatmem_Tool: The ``gnatmem`` Tool -------------------- .. index:: ! gnatmem The ``gnatmem`` utility monitors dynamic allocation and deallocation activity in a program, and displays information about incorrect deallocations and possible sources of memory leaks. It is designed to work for fixed-position executables that use a static runtime library and, in this context, provides three types of information: * General information concerning memory management, such as the total number of allocations and deallocations, the amount of allocated memory and the high water mark, i.e., the largest amount of allocated memory in the course of program execution. * Backtraces for all incorrect deallocations, which are deallocations that do not correspond to a valid allocation. * Information on each allocation that is potentially the origin of a memory leak. .. _Running_gnatmem: Running ``gnatmem`` ^^^^^^^^^^^^^^^^^^^ ``gnatmem`` makes use of the output created by the special version of allocation and deallocation routines that record call information. This allows it to obtain accurate dynamic memory usage history at a minimal cost to the execution speed. Note however, that ``gnatmem`` is only supported on GNU/Linux and Windows. The ``gnatmem`` command has the form :: $ gnatmem [ switches ] [ DEPTH ] user_program You must link your program with the instrumented version of the allocation and deallocation routines. You do this by linking with the :file:`libgmem.a` library. For correct symbolic backtrace information, you should also compile your program with debugging options (see :ref:`Switches_for_gcc`) and be linked at a fixed position (with :switch:`-no-pie`). For example to build :file:`my_program` with ``gnatmake``: :: $ gnatmake my_program -g -largs -lgmem -no-pie Because library :file:`libgmem.a` contains an alternate body for package ``System.Memory``, you should not compile and link :file:`s-memory.adb` when you link an executable with library :file:`libgmem.a`. In that case, we don't recommended specifying switch :switch:`-a` to ``gnatmake``. When :file:`my_program` is executed, the file :file:`gmem.out` is produced. This file contains information about all allocations and deallocations performed by the program. It is produced by the instrumented allocations and deallocations routines and will be used by ``gnatmem``. To produce symbolic backtrace information for allocations and deallocations performed by the GNAT run-time library, you need to use a version of that library that has been compiled with the :switch:`-g` switch (see :ref:`Rebuilding_the_GNAT_Run-Time_Library`). You must supply ``gnatmem`` with the :file:`gmem.out` file and the executable to examine. If the location of :file:`gmem.out` file was not explicitly supplied by :switch:`-i` switch, ``gnatmem`` assumes that this file can be found in the current directory. For example, after you have executed :file:`my_program`, :file:`gmem.out` can be analyzed by ``gnatmem`` using the command: :: $ gnatmem my_program This will produce the output with the following format: :: $ gnatmem my_program Global information ------------------ Total number of allocations : 45 Total number of deallocations : 6 Final Water Mark (non freed mem) : 11.29 Kilobytes High Water Mark : 11.40 Kilobytes . . . Allocation Root # 2 ------------------- Number of non freed allocations : 11 Final Water Mark (non freed mem) : 1.16 Kilobytes High Water Mark : 1.27 Kilobytes Backtrace : my_program.adb:23 my_program.alloc . . . The first block of output gives general information. In this case, the Ada construct ``new`` was executed 45 times and only 6 calls to an ``Unchecked_Deallocation`` routine occurred. Subsequent paragraphs display information on all allocation roots. An *allocation root* is a specific point in the execution of the program that generates some dynamic allocation, such as a ``new`` construct. This root is represented by an execution backtrace (or subprogram call stack). By default, the backtrace depth for allocations roots is 1, so that a root corresponds exactly to a source location. The backtrace can be made deeper, to make the root more specific. .. _Switches_for_gnatmem: Switches for ``gnatmem`` ^^^^^^^^^^^^^^^^^^^^^^^^ ``gnatmem`` recognizes the following switches: .. index:: -q (gnatmem) :switch:`-q` Quiet. Gives the minimum output needed to identify the origin of the memory leaks. Omits statistical information. .. index:: DEPTH switch (gnatmem) :switch:`{DEPTH}` ``DEPTH`` is an integer literal (usually between 1 and 10) which controls the depth of the backtraces defining allocation root. The default value for DEPTH is 1. The deeper the backtrace, the more precise the localization of the root. Note that the total number of roots can depend on this parameter; in other words there may be more roots when the requested backtrace depth is higher. You must specify this parameter *before* the name of the executable to be analyzed, to avoid ambiguity. .. index:: -b (gnatmem) :switch:`-b {N}` This switch has the same effect as just a depth parameter ``N``. .. index:: -i (gnatmem) :switch:`-i {file}` Do the ``gnatmem`` processing starting from :file:`file`, rather than :file:`gmem.out` in the current directory. .. index:: -m (gnatmem) :switch:`-m {n}` This switch causes ``gnatmem`` to mask the allocation roots that have less than ``n`` leaks. The default value is 1. Specifying the value of 0 will allow examination of even the roots that did not result in leaks. .. index:: -s (gnatmem) :switch:`-s {order}` This switch causes ``gnatmem`` to sort the allocation roots according to the specified sort criteria, each identified by a single letter. The currently supported criteria are ``n``, ``h``, and ``w`` representing, respectively, the number of unfreed allocations, the high watermark, and the final watermark corresponding to a specific root. The default order is ``nwh``. .. index:: -t (gnatmem) :switch:`-t` This switch causes memory allocated size to be always output in bytes. The default ``gnatmem`` behavior is to show memory sizes less then 1 kilobyte in bytes, from 1 kilobyte till 1 megabyte in kilobytes and the rest in megabytes. .. _Example_of_gnatmem_Usage: Example of ``gnatmem`` Usage ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The following example shows the use of ``gnatmem`` on a simple memory-leaking program. Suppose that we have the following Ada program: .. code-block:: ada with Ada.Unchecked_Deallocation; procedure Test_Gm is type T is array (1..1000) of Integer; type Ptr is access T; procedure Free is new Ada.Unchecked_Deallocation (T, Ptr); A : Ptr; procedure My_Alloc is begin A := new T; end My_Alloc; procedure My_DeAlloc is B : Ptr := A; begin Free (B); end My_DeAlloc; begin My_Alloc; for I in 1 .. 5 loop for J in I .. 5 loop My_Alloc; end loop; My_Dealloc; end loop; end; The program needs to be compiled with the debugging option and linked with the ``gmem`` library: :: $ gnatmake -g test_gm -largs -lgmem We execute the program as usual: :: $ test_gm ``gnatmem`` is invoked simply with :: $ gnatmem test_gm which produces the following output (the details may vary on different platforms): :: Global information ------------------ Total number of allocations : 18 Total number of deallocations : 5 Final Water Mark (non freed mem) : 53.00 Kilobytes High Water Mark : 56.90 Kilobytes Allocation Root # 1 ------------------- Number of non freed allocations : 11 Final Water Mark (non freed mem) : 42.97 Kilobytes High Water Mark : 46.88 Kilobytes Backtrace : test_gm.adb:11 test_gm.my_alloc Allocation Root # 2 ------------------- Number of non freed allocations : 1 Final Water Mark (non freed mem) : 10.02 Kilobytes High Water Mark : 10.02 Kilobytes Backtrace : s-secsta.adb:81 system.secondary_stack.ss_init Allocation Root # 3 ------------------- Number of non freed allocations : 1 Final Water Mark (non freed mem) : 12 Bytes High Water Mark : 12 Bytes Backtrace : s-secsta.adb:181 system.secondary_stack.ss_init Note that the GNAT runtime itself contains a certain number of allocations that have no corresponding deallocations, as shown here for root #2 and root #3. This is a normal behavior when the number of non-freed allocations is one: it allocates dynamic data structures that the run time needs for the complete lifetime of the program. Note also that there is only one allocation root in the user program, with a single line back trace: ``test_gm.adb:11 test_gm.my_alloc``, whereas a careful analysis of the program shows that ``My_Alloc`` is called at 2 different points in the source (line 21 and line 24). If those two allocation roots need to be distinguished, you can use the backtrace depth parameter: :: $ gnatmem 3 test_gm which produces the following output: :: Global information ------------------ Total number of allocations : 18 Total number of deallocations : 5 Final Water Mark (non freed mem) : 53.00 Kilobytes High Water Mark : 56.90 Kilobytes Allocation Root # 1 ------------------- Number of non freed allocations : 10 Final Water Mark (non freed mem) : 39.06 Kilobytes High Water Mark : 42.97 Kilobytes Backtrace : test_gm.adb:11 test_gm.my_alloc test_gm.adb:24 test_gm b_test_gm.c:52 main Allocation Root # 2 ------------------- Number of non freed allocations : 1 Final Water Mark (non freed mem) : 10.02 Kilobytes High Water Mark : 10.02 Kilobytes Backtrace : s-secsta.adb:81 system.secondary_stack.ss_init s-secsta.adb:283 b_test_gm.c:33 adainit Allocation Root # 3 ------------------- Number of non freed allocations : 1 Final Water Mark (non freed mem) : 3.91 Kilobytes High Water Mark : 3.91 Kilobytes Backtrace : test_gm.adb:11 test_gm.my_alloc test_gm.adb:21 test_gm b_test_gm.c:52 main Allocation Root # 4 ------------------- Number of non freed allocations : 1 Final Water Mark (non freed mem) : 12 Bytes High Water Mark : 12 Bytes Backtrace : s-secsta.adb:181 system.secondary_stack.ss_init s-secsta.adb:283 b_test_gm.c:33 adainit The allocation root #1 of the first example has been split in 2 roots #1 and #3, thanks to the more precise associated backtrace. .. _Sanitizers_for_Ada: Sanitizers for Ada ================== .. index:: Sanitizers This section explains how to use sanitizers with Ada code. Sanitizers offer code instrumentation and run-time libraries that detect certain memory issues and undefined behaviors during execution. They provide dynamic analysis capabilities useful for debugging and testing. While many sanitizer capabilities overlap with Ada's built-in runtime checks, they are particularly valuable for identifying issues that arise from unchecked features or low-level operations. .. _AddressSanitizer: AddressSanitizer ---------------- .. index:: AddressSanitizer .. index:: ASan .. index:: -fsanitize=address AddressSanitizer (aka ASan) is a memory error detector activated with the :switch:`-fsanitize=address` switch. Note that many of the typical memory errors, such as use after free or buffer overflow, are detected by Ada’s ``Access_Check`` and ``Index_Check``. It can detect the following types of problems: * Wrong memory overlay A memory overlay is a situation in which an object of one type is placed at the same memory location as a distinct object of a different type, thus overlaying one object over the other in memory. When there is an overflow because the objects do not overlap (like in the following example), the sanitizer can signal it. .. code-block:: ada procedure Wrong_Size_Overlay is type Block is array (Natural range <>) of Integer; Block4 : aliased Block := (1 .. 4 => 4); Block5 : Block (1 .. 5) with Address => Block4'Address; begin Block5 (Block5'Last) := 5; -- Outside the object end Wrong_Size_Overlay; If the code is built with the :switch:`-fsanitize=address` and :switch:`-g` options, the following error is shown at execution time: :: ... SUMMARY: AddressSanitizer: stack-buffer-overflow wrong_size_overlay.adb:7 in _ada_wrong_size_overlay ... * Buffer overflow Ada’s ``Index_Check`` detects buffer overflows caused by out-of-bounds array access. If run-time checks are disabled, the sanitizer can still detect such overflows at execution time the same way as it signalled the previous wrong memory overlay. Note that if both the Ada run-time checks and the sanitizer are enabled, the Ada run-time exception takes precedence. .. code-block:: ada procedure Buffer_Overrun is Size : constant := 100; Buffer : array (1 .. Size) of Integer := (others => 0); Wrong_Index : Integer := Size + 1 with Export; begin -- Access outside the boundaries Put_Line ("Value: " & Integer'Image (Buffer (Wrong_Index))); end Buffer_Overrun; * Use after lifetime Ada’s ``Accessibility_Check`` helps prevent use-after-return and use-after-scope errors by enforcing lifetime rules. When these checks are bypassed using ``Unchecked_Access``, sanitizers can still detect such violations during execution. .. code-block:: ada with Ada.Text_IO; use Ada.Text_IO; procedure Use_After_Return is type Integer_Access is access all Integer; Ptr : Integer_Access; procedure Inner; procedure Inner is Local : aliased Integer := 42; begin Ptr := Local'Unchecked_Access; end Inner; begin Inner; -- Accessing Local after it has gone out of scope Put_Line ("Value: " & Integer'Image (Ptr.all)); end Use_After_Return; If the code is built with the :switch:`-fsanitize=address` and :switch:`-g` options, the following error is shown at execution time: :: ... ==1793927==ERROR: AddressSanitizer: stack-use-after-return on address 0xf6fa1a409060 at pc 0xb20b6cb6cac0 bp 0xffffcc89c8b0 sp 0xffffcc89c8c8 READ of size 4 at 0xf6fa1a409060 thread T0 #0 0xb20b6cb6cabc in _ada_use_after_return use_after_return.adb:18 ... Address 0xf6fa1a409060 is located in stack of thread T0 at offset 32 in frame #0 0xb20b6cb6c794 in use_after_return__inner use_after_return.adb:9 This frame has 1 object(s): [32, 36) 'local' (line 10) <== Memory access at offset 32 is inside this variable SUMMARY: AddressSanitizer: stack-use-after-return use_after_return.adb:18 in _ada_use_after_return ... * Memory leak A memory leak happens when a program allocates memory from the heap but fails to release it after it is no longer needed and loses all references to it like in the following example. .. code-block:: ada procedure Memory_Leak is type Integer_Access is access Integer; procedure Allocate is Ptr : Integer_Access := new Integer'(42); begin null; end Allocate; begin -- Memory leak occurs in the following procedure Allocate; end Memory_Leak; If the code is built with the :switch:`-fsanitize=address` and :switch:`-g` options, the following error is emitted at execution time showing the location of the offending allocation. :: ==1810634==ERROR: LeakSanitizer: detected memory leaks Direct leak of 4 byte(s) in 1 object(s) allocated from: #0 0xe3cbee4bb4a8 in __interceptor_malloc asan_malloc_linux.cpp:69 #1 0xc15bb25d0af8 in __gnat_malloc (memory_leak+0x10af8) (BuildId: f5914a6eac10824f81d512de50b514e7d5f733be) #2 0xc15bb25c9060 in memory_leak__allocate memory_leak.adb:5 ... SUMMARY: AddressSanitizer: 4 byte(s) leaked in 1 allocation(s). .. _UndefinedBehaviorSanitizer: UndefinedBehaviorSanitizer -------------------------- .. index:: UndefinedBehaviorSanitizer .. index:: UBSan .. index:: -fsanitize=undefined UndefinedBehaviorSanitizer (aka UBSan) modifies the program at compile-time to catch various kinds of undefined behavior during program execution. Different sanitize options (:switch:`-fsanitize=alignment,float-cast-overflow,signed-integer-overflow`) detect the following types of problems: * Wrong alignment The :switch:`-fsanitize=alignment` flag (included also in :switch:`-fsanitize=undefined`) enables run-time checks for misaligned memory accesses, ensuring that objects are accessed at addresses that conform to the alignment constraints of their declared types. Violations may lead to crashes or performance penalties on certain architectures. In the following example: .. code-block:: ada with Ada.Text_IO; use Ada.Text_IO; with System.Storage_Elements; use System.Storage_Elements; procedure Misaligned_Address is type Aligned_Integer is new Integer with Alignment => 4; -- Ensure 4-byte alignment Reference : Aligned_Integer := 42; -- Properly aligned object -- Create a misaligned object by modifying the address manually Misaligned : Aligned_Integer with Address => Reference'Address + 1; begin -- This causes undefined behavior or an alignment exception on strict architectures Put_Line ("Misaligned Value: " & Aligned_Integer'Image (Misaligned)); end Misaligned_Address; If the code is built with the :switch:`-fsanitize=alignment` and :switch:`-g` options, the following error is shown at execution time. :: misaligned_address.adb:15:51: runtime error: load of misaligned address 0xffffd836dd45 for type 'volatile misaligned_address__aligned_integer', which requires 4 byte alignment * Signed integer overflow Ada performs range checks at runtime in arithmetic operation on signed integers to ensure the value is within the target type's bounds. If this check is removed, the :switch:`-fsanitize=signed-integer-overflow` flag (included also in :switch:`-fsanitize=undefined`) enables run-time checks for signed integer overflows. In the following example: .. code-block:: ada procedure Signed_Integer_Overflow is type Small_Int is range -128 .. 127; X, Y, Z : Small_Int with Export; begin X := 100; Y := 50; -- This addition will exceed 127, causing an overflow Z := X + Y; end Signed_Integer_Overflow; If the code is built with the :switch:`-fsanitize=signed-integer-overflow` and :switch:`-g` options, the following error is shown at execution time. :: signed_integer_overflow.adb:8:11: runtime error: signed integer overflow: 100 + 50 cannot be represented in type 'signed_integer_overflow__small_int' * Float to integer overflow When converting a floating-point value to an integer type, Ada performs a range check at runtime to ensure the value is within the target type's bounds. If this check is removed, the sanitizer can detect overflows in conversions from floating point to integer types. In the following code: .. code-block:: ada procedure Float_Cast_Overflow is Flt : Float := Float'Last with Export; Int : Integer; begin Int := Integer (Flt); -- Overflow end Float_Cast_Overflow; If the code is built with the :switch:`-fsanitize=float-cast-overflow` and :switch:`-g` options, the following error is shown at execution time. :: float_cast_overflow.adb:5:20: runtime error: 3.40282e+38 is outside the range of representable values of type 'integer' Sanitizers in mixed-language applications ----------------------------------------- Most of the checks performed by sanitizers operate at a global level, which means they can detect issues even when they span across language boundaries. This applies notably to: * All checks performed by the AddressSanitizer: wrong memory overlays, buffer overflows, uses after lifetime, memory leaks. These checks apply globally, regardless of where the objects are allocated or defined, or where they are destroyed * Wrong alignment checks performed by the UndefinedBehaviorSanitizer. It will check whether an object created in a given language is accessed in another with an incompatible alignment An interesting case that highlights the benefit of global sanitization is a buffer overflow caused by a mismatch in language bindings. Consider the following C function, which allocates an array of 4 characters: .. code-block:: c char *get_str (void) { char *str = malloc (4 * sizeof (char)); } This function is then bound to Ada code, which incorrectly assumes the buffer is of size 5: .. code-block:: ada type Buffer is array (1 .. 5) of Character; function Get_Str return access Buffer with Import => True, Convention => C, External_Name => "get_str"; Str : access Buffer := Get_Str; Ch : Character := S (S'Last); -- Detected by AddressSanitizer as erroneous On the Ada side, accessing ``Str (5)`` appears valid because the array type declares five elements. However, the actual memory allocated in C only holds four. This mismatch is not detectable by Ada run-time checks, because Ada has no visibility into how the memory was allocated. However, the AddressSanitizer will detect the heap buffer overflow at runtime, halting execution and providing a clear diagnostic: :: ... SUMMARY: AddressSanitizer: heap-buffer-overflow buffer_overflow.adb:20 in _ada_buffer_overflow ... --- ### Gcc Ada Doc Gnat Ugn Gnat Utility Programs (`gcc/ada/doc/gnat_ugn/gnat_utility_programs.rst`) .. role:: switch(samp) .. |rightarrow| unicode:: 0x2192 .. _GNAT_Utility_Programs: ********************* GNAT Utility Programs ********************* This chapter describes a number of utility programs: .. only:: PRO or GPL * :ref:`The_File_Cleanup_Utility_gnatclean` * :ref:`The_GNAT_Library_Browser_gnatls` * :ref:`The_GNAT_Pretty_Printer_gnatpp` * :ref:`The_Body_Stub_Generator_gnatstub` * :ref:`The_Backtrace_Symbolizer_gnatsymbolize` It also describes how you can use several of these tools in conjunction with project files: :ref:`Using_Project_Files_with_GNAT_Tools` .. only:: FSF * :ref:`The_File_Cleanup_Utility_gnatclean` * :ref:`The_GNAT_Library_Browser_gnatls` Other GNAT utilities are described elsewhere in this manual: * :ref:`Handling_Arbitrary_File_Naming_Conventions_with_gnatname` * :ref:`File_Name_Krunching_with_gnatkr` * :ref:`Renaming_Files_with_gnatchop` * :ref:`Preprocessing_with_gnatprep` .. _The_File_Cleanup_Utility_gnatclean: The File Cleanup Utility ``gnatclean`` ====================================== .. index:: File cleanup tool .. index:: gnatclean ``gnatclean`` is a tool that deletes some files produced by the compiler, binder and linker, including ALI files, object files, tree files, expanded source files, library files, interface copy source files, binder generated files and executable files. .. _Running_gnatclean: Running ``gnatclean`` --------------------- You run the ``gnatclean`` command as follow: :: $ gnatclean switches names where ``names`` is a list of source file names. You may omit suffixes :file:`.ads` and :file:`adb`. If a project file is specified using switch :switch:`-P`, then you may completely omit ``names``. In normal mode, ``gnatclean`` deletes the files produced by the compiler and, if switch :switch:`-c` is not specified, produced by the binder and linker. In information-only mode, specified by switch :switch:`-n`, ``gnatclean`` lists the files that would have been deleted in normal mode, but doesn't actually delete any files. .. _Switches_for_gnatclean: Switches for ``gnatclean`` -------------------------- ``gnatclean`` recognizes the following switches: .. index:: --version (gnatclean) :switch:`--version` Display copyright and version, then exit, disregarding all other options. .. index:: --help (gnatclean) :switch:`--help` If :switch:`--version` was not specified, display usage, then exit disregarding all other options. :switch:`--subdirs={subdir}` Actual object directory of each project file, which is the subdirectory ``subdir`` of the object directory specified or defaulted in the project file. :switch:`--unchecked-shared-lib-imports` By default, shared library projects are not allowed to import static library projects. When this switch is specified, this restriction is lifted. .. index:: -c (gnatclean) :switch:`-c` Only attempt to delete the files produced by the compiler, not those produced by the binder or the linker. The files that are not to be deleted are library files, interface copy files, binder generated files and executable files. .. index:: -D (gnatclean) :switch:`-D {dir}` Indicate that ALI and object files should normally be found in directory ``dir``. .. index:: -F (gnatclean) :switch:`-F` When using project files, if some errors or warnings are detected during parsing and verbose mode is not in effect (the switch :switch:`-v` is not specified), error lines start with the full path name of the project file, rather than its simple file name. .. index:: -h (gnatclean) :switch:`-h` Output a message explaining the usage of ``gnatclean``. .. index:: -n (gnatclean) :switch:`-n` Informative-only mode. Do not delete any files. Output the list of the files that would have been deleted if this switch was not specified. .. index:: -P (gnatclean) :switch:`-P{project}` Use project file ``project``. You can specify only one such switch. When cleaning a project file, ``gnatclean`` deletes the files produced by the compilation of the immediate sources or inherited sources of the project files. This does not depend on whether or not you include executable names on the command line. .. index:: -q (gnatclean) :switch:`-q` Quiet output. If there are no errors, do not output anything, except in verbose mode (:switch:`-v`) or in information-only mode (:switch:`-n`). .. index:: -r (gnatclean) :switch:`-r` When a project file is specified (using :switch:`-P`), clean all imported and extended project files, recursively. If you don't specify this switch, ``gnatclean`` only deletes the files related to the main project file. This switch has no effect if you don't specify a project file. .. index:: -v (gnatclean) :switch:`-v` Verbose mode. .. index:: -vP (gnatclean) :switch:`-vP{x}` Indicates the verbosity of the parsing of GNAT project files. :ref:`Switches_Related_to_Project_Files`. .. index:: -X (gnatclean) :switch:`-X{name}={value}` Indicates that external variable ``name`` has the value ``value``. The Project Manager will use this value for occurrences of ``external(name)`` when parsing the project file. See :ref:`Switches_Related_to_Project_Files`. .. index:: -aO (gnatclean) :switch:`-aO{dir}` When searching for ALI and object files, look in directory ``dir``. .. index:: -I (gnatclean) :switch:`-I{dir}` Equivalent to :switch:`-aO{dir}`. .. index:: -I- (gnatclean) .. index:: Source files, suppressing search :switch:`-I-` Do not look for ALI or object files in the directory where ``gnatclean`` was invoked. .. _The_GNAT_Library_Browser_gnatls: The GNAT Library Browser ``gnatls`` =================================== .. index:: Library browser .. index:: ! gnatls ``gnatls`` is a tool that outputs information about compiled units. It gives the relationship between objects, unit names and source files. You can also use it to check the source dependencies of a unit as well as various characteristics. .. _Running_gnatls: Running ``gnatls`` ------------------ You run the ``gnatls`` command as follows: :: $ gnatls switches object_or_ali_file The main argument is the list of object or :file:`ali` files (see :ref:`The_Ada_Library_Information_Files`) for which you are requesting information. In the default mode, without additional options, ``gnatls`` produces a four-column listing. Each line contains information for a specific object. The first column gives the full path of the object, the second column gives the name of the principal unit in the object, the third column gives the status of the source and the fourth column gives the full path of the source representing this unit. Here's a simple example: :: $ gnatls *.o ./demo1.o demo1 DIF demo1.adb ./demo2.o demo2 OK demo2.adb ./hello.o h1 OK hello.adb ./instr-child.o instr.child MOK instr-child.adb ./instr.o instr OK instr.adb ./tef.o tef DIF tef.adb ./text_io_example.o text_io_example OK text_io_example.adb ./tgef.o tgef DIF tgef.adb You should interpret the first line as follows: the main unit, which is contained in object file :file:`demo1.o`, is demo1, whose main source is in :file:`demo1.adb`. Furthermore, the version of the source used for the compilation of demo1 has been modified (DIF). Each source file has a status qualifier which can be: *OK (unchanged)* The version of the source file used for the compilation of the specified unit corresponds exactly to the actual source file. *MOK (slightly modified)* The version of the source file used for the compilation of the specified unit differs from the actual source file but not enough to require recompilation (e.g., only comments have been changed). If you run ``gnatmake`` with the option :switch:`-m` (minimal recompilation), it will not recompile a file marked MOK. *DIF (modified)* No version of the source found on the path corresponds to the source used to build this object. *??? (file not found)* No source file was found for this unit. *HID (hidden, unchanged version not first on PATH)* The version of the source that corresponds exactly to the source used for compilation has been found on the path but it is hidden by another version of the same source that has been modified. .. _Switches_for_gnatls: Switches for ``gnatls`` ----------------------- You can specify the following switches to ``gnatls``: .. index:: --version (gnatls) :switch:`--version` Display copyright and version, then exit, disregarding all other options. .. index:: --help (gnatls) :switch:`--help` If :switch:`--version` was not specified, display usage, then exit, disregarding all other options. .. index:: -a (gnatls) :switch:`-a` Consider all units, including those of the predefined Ada library. Especially useful with :switch:`-d`. .. index:: -d (gnatls) :switch:`-d` List sources that specified units depend on. .. index:: -h (gnatls) :switch:`-h` Output the list of options. .. index:: -o (gnatls) :switch:`-o` Only output information about object files. .. index:: -s (gnatls) :switch:`-s` Only output information about source files. .. index:: -u (gnatls) :switch:`-u` Only output information about compilation units. .. index:: -files (gnatls) :switch:`-files={file}` Take as arguments the files listed in text file ``file``, which may contain empty lines that are ignored. Each nonempty line should contain the name of an existing file. Several such switches may be specified on the same command. .. index:: -aO (gnatls) .. index:: -aI (gnatls) .. index:: -I (gnatls) .. index:: -I- (gnatls) :switch:`-aO{dir}`, :switch:`-aI{dir}`, :switch:`-I{dir}`, :switch:`-I-`, :switch:`-nostdinc` Source path manipulation. It has the same meaning as the equivalent ``gnatmake`` switches (:ref:`Switches_for_gnatmake`). .. index:: -aP (gnatls) :switch:`-aP{dir}` Add ``dir`` at the beginning of the project search dir. .. index:: --RTS (gnatls) :switch:`--RTS={rts-path}` Specifies the default location of the runtime library. It has the same meaning as the equivalent ``gnatmake`` switch (:ref:`Switches_for_gnatmake`). .. index:: -v (gnatls) :switch:`-v` Verbose mode. Output the complete source, object and project paths. Don't use the default column layout but instead use long format giving as much as information possible on each requested units, including special characteristics such as: * *Preelaborable*: The unit is preelaborable in the Ada sense. * *No_Elab_Code*: No elaboration code has been produced by the compiler for this unit. * *Pure*: The unit is pure in the Ada sense. * *Elaborate_Body*: The unit contains a pragma Elaborate_Body. * *Remote_Types*: The unit contains a pragma Remote_Types. * *Shared_Passive*: The unit contains a pragma Shared_Passive. * *Predefined*: This unit is part of the predefined environment and cannot be modified by the user. * *Remote_Call_Interface*: The unit contains a pragma Remote_Call_Interface. .. _Example_of_gnatls_Usage: Example of ``gnatls`` Usage --------------------------- Here's an example of using the verbose switch. Note how the source and object paths are affected by the :switch:`-I` switch. :: $ gnatls -v -I.. demo1.o GNATLS 5.03w (20041123-34) Copyright 1997-2004 Free Software Foundation, Inc. Source Search Path: ../ /home/comar/local/adainclude/ Object Search Path: ../ /home/comar/local/lib/gcc-lib/x86-linux/3.4.3/adalib/ Project Search Path: /home/comar/local/lib/gnat/ ./demo1.o Unit => Name => demo1 Kind => subprogram body Flags => No_Elab_Code Source => demo1.adb modified Here's an example of use of the dependency list. Note the use of the :switch:`-s` switch, which gives a simple list of source files. You may find this useful for building specialized scripts. :: $ gnatls -d demo2.o ./demo2.o demo2 OK demo2.adb OK gen_list.ads OK gen_list.adb OK instr.ads OK instr-child.ads $ gnatls -d -s -a demo1.o demo1.adb /home/comar/local/adainclude/ada.ads /home/comar/local/adainclude/a-finali.ads /home/comar/local/adainclude/a-filico.ads /home/comar/local/adainclude/a-stream.ads /home/comar/local/adainclude/a-tags.ads gen_list.ads gen_list.adb /home/comar/local/adainclude/gnat.ads /home/comar/local/adainclude/g-io.ads instr.ads /home/comar/local/adainclude/system.ads /home/comar/local/adainclude/s-exctab.ads /home/comar/local/adainclude/s-finimp.ads /home/comar/local/adainclude/s-finroo.ads /home/comar/local/adainclude/s-secsta.ads /home/comar/local/adainclude/s-stalib.ads /home/comar/local/adainclude/s-stoele.ads /home/comar/local/adainclude/s-stratt.ads /home/comar/local/adainclude/s-tasoli.ads /home/comar/local/adainclude/s-unstyp.ads /home/comar/local/adainclude/unchconv.ads .. only:: PRO or GPL .. _The_GNAT_Pretty_Printer_gnatpp: The GNAT Pretty Printer ``gnatpp`` ================================== .. index:: ! gnatpp .. index:: pretty printer The ``gnatpp`` tool is a utility for source reformatting / pretty printing. It takes an Ada source file as input and generates a reformatted version as output. You can specify various style directives via switches; e.g., identifier case conventions, rules of indentation, and comment layout. ``gnatpp`` is a project-aware tool (see :ref:`Using_Project_Files_with_GNAT_Tools` for a description of the project-related switches). The project file package that can specify ``gnatpp`` switches is named ``Pretty_Printer``. ``gnatpp`` cannot process sources that contain preprocessing directives. The ``gnatpp`` command has the form :: $ gnatpp [ switches ] filename where * ``switches`` is an optional sequence of switches defining such properties as the formatting rules, the source search path, and the destination for the output source file * ``filename`` is the name of the source file to reformat; wildcards or several file names on the same gnatpp command are allowed. The file name may contain path information; it does not have to follow the GNAT file naming rules Note that you no longer need to specify the Ada language version; ``gnatpp`` can process Ada source code written in any version from Ada 83 onward without specifying any language version switch. .. _Switches_for_gnatpp: Switches for ``gnatpp`` ----------------------- The following subsections describe the various switches accepted by ``gnatpp``, organized by category. You specify a switch by supplying a name and usually also a value. In many cases the values for a switch with a given name are incompatible with each other (for example the switch that controls the casing of a reserved word may have exactly one value: upper case, lower case, or mixed case) and thus exactly one such switch can be in effect for an invocation of ``gnatpp``. If more than one is supplied, the last one is used. However, some values for the same switch are mutually compatible. You may supply several such switches to ``gnatpp``, but each must be specified in full, with both the name and the value. Abbreviated forms (the name appearing once, followed by each value) are not permitted. .. _Layout_Control: Layout Control ^^^^^^^^^^^^^^ .. index:: Layout control in gnatpp ``gnatpp`` provides a layout switch which controls the general formatting style: .. index:: layout(gnatpp) :switch:`--layout=default|minimal|compact|tall` :switch:`default` The default layout is a compact style, but ``gnatpp`` adds alignment and puts some keywords on a separate line. ``gnatpp`` adds alignment in the the following constructs: * ``:`` in declarations, * ``:=`` in initializations in declarations, * ``:=`` in assignment statements, * ``=>`` in associations, and * ``at`` keywords in the component clauses in record representation clauses. In addition, ``gnatpp`` also lines up ``in`` and ``out`` keywords in parameter specifications. ``gnatpp`` places the keyword ``is`` on a separate line in a subprogram body in case the spec occupies more than one line. ``gnatpp`` places the keyword ``return`` on a separate line if a subprogram spec does not fit on one line. :switch:`minimal` The minimal layout will avoid changing the source layout by keeping all line breaks from the original source (it will not insert or delete any). It will add indentation where appropriate as long as it does not exceed the line length limit. :switch:`compact` The compact layout avoids adding line breaks and alignment by packing as many subexpressions on the same line as possible. Whole-line comments that form a paragraph are filled in typical word processor style (that is, moving words between lines to make them similar in length, except the last one which may be shorter). For each whole-line comment that does not end with two hyphens, inserts spaces if necessary after the starting two hyphens to ensure that there are at least two spaces preceding the first non-blank character of the comment. :switch:`tall` The tall layout favors adding lines breaks and alignment. It adds all the alignment and line breaks defined in the ``default`` option, and in addition: * Places the keyword ``loop`` in FOR and WHILE loop statements on a separate line * Places the keyword ``then`` in IF statements on a separate line * Places each keyword ``use`` in USE clauses on a separate line * Splits the line just before the keyword ``record`` in a RECORD type declaration * Indents named blocks and loop statements with respect to the name * When necessary, splits binary operators always before the operator * Inserts an extra blank before various occurrences of ``(`` and ``:`` * When it is necessary to split a line between two subexpressions (because otherwise the construct would exceed --max-line-length), then all such subexpressions are placed on separate lines * Formats enumeration type declarations “vertically”, e.g. each enumeration literal goes on a separate line * Formats array type declarations “vertically”, e.g. for multidimensional arrays, each index_subtype_definition or discrete_subtype_definition goes on a separate line * Format aggregates “vertically” if named notation is used for all component_associations, e.g. each component_association goes on a separate line * Formats case statements, case expressions, and variant parts with additional line breaks * Inserts blank lines where appropriate (between bodies and other large constructs) * Similarly to the :switch:`compact` layout, two spaces are added in the beginning of a whole-line comment when needed .. _Casing_Control: Casing Control ^^^^^^^^^^^^^^ .. index:: Casing control in gnatpp ``gnatpp`` allows you to specify the casing for reserved words, pragma names, attribute designators, and identifiers. For identifiers, you may define a general rule for name casing but also override this rule via a set of dictionary files. Three types of casing are supported: 'Lower Case', 'Upper Case', and 'Mixed Case'. 'Mixed case' means that the first letter and each letter immediately following an underscore are converted to their uppercase forms and all other letters are converted to their lowercase forms. .. index:: --name-case-as-declared (gnatpp) :switch:`--name-case-as-declared, -nD` Name casing for defining occurrences are as they appear in the source file (this is the default). .. index:: --name-upper-case (gnatpp) :switch:`--name-upper-case, -nU` Names are in upper case. .. index:: --name-lower-case (gnatpp) :switch:`--name-lower-case, -nL` Names are in lower case. .. index:: --name-mixed-case (gnatpp) :switch:`--name-mixed-case, -nM` Names are in mixed case. .. index:: --attribute-lower-case (gnatpp) :switch:`--attribute-lower-case, -aL` Attribute designators are lower case. .. index:: --attribute-upper-case (gnatpp) :switch:`--attribute-upper-case, -aU` Attribute designators are upper case. .. index:: --attribute-mixed-case (gnatpp) :switch:`--attribute-mixed-case, -aM` Attribute designators are mixed case (this is the default). .. index:: --keyword-lower-case (gnatpp) :switch:`--keyword-lower-case, -kL` Keywords (technically, these are known in Ada as *reserved words*) are lower case (this is the default). .. index:: --keyword-upper-case (gnatpp) :switch:`--keyword-upper-case, -kU` Keywords are upper case. .. index:: --enum-case-as-declared (gnatpp) :switch:`--enum-case-as-declared, -neD` Enumeration literal casing for defining occurrences are as they appear in the source file. Overrides the :switch:`-n` casing setting. .. index:: --enum-upper-case (gnatpp) :switch:`--enum-upper-case, -neU` Enumeration literals are in upper case. Overrides the :switch:`-n` casing setting. .. index:: --enum-lower-case (gnatpp) :switch:`--enum-lower-case, -neL` Enumeration literals are in lower case. Overrides the :switch:`-n` casing setting. .. index:: --enum-mixed-case (gnatpp) :switch:`--enum-mixed-case, -neM` Enumeration literals are in mixed case. Overrides the :switch:`-n` casing setting. .. index:: --type-case-as-declared (gnatpp) :switch:`--type-case-as-declared, -ntD` Names introduced by type and subtype declarations are always cased as they appear in the declaration in the source file. Overrides the :switch:`-n` casing setting. .. index:: --type-upper-case (gnatpp) :switch:`--type-upper-case, -ntU` Names introduced by type and subtype declarations are always in upper case. Overrides the :switch:`-n` casing setting. .. index:: --type-lower-case (gnatpp) :switch:`--type-lower-case, -ntL` Names introduced by type and subtype declarations are always in lower case. Overrides the :switch:`-n` casing setting. .. index:: --type-mixed-case (gnatpp) :switch:`--type-mixed-case, -ntM` Names introduced by type and subtype declarations are always in mixed case. Overrides the :switch:`-n` casing setting. .. index:: --number-upper-case (gnatpp) :switch:`--number-upper-case, -nnU` Names introduced by number declarations are always in upper case. Overrides the :switch:`-n` casing setting. .. index:: --number-lower-case (gnatpp) :switch:`--number-lower-case, -nnL` Names introduced by number declarations are always in lower case. Overrides the :switch:`-n` casing setting. .. index:: --number-mixed-case (gnatpp) :switch:`--number-mixed-case, -nnM` Names introduced by number declarations are always in mixed case. Overrides -n casing setting. .. index:: --pragma-lower-case (gnatpp) :switch:`--pragma-lower-case, -pL` Pragma names are lower case. .. index:: --pragma-upper-case (gnatpp) :switch:`--pragma-upper-case, -pU` Pragma names are upper case. .. index:: --pragma-mixed-case (gnatpp) :switch:`--pragma-mixed-case, -pM` Pragma names are mixed case (this is the default). .. index:: --constant-case-as-non-constant (gnatpp) :switch:`--constant-case-as-non-constant, -cN` Constant object declaration names have the same case as the name casing (this is the default). .. index:: --constant-case-as-declared (gnatpp) :switch:`--constant-case-as-declared, -cD` Constant object declaration names are as they appear in the source file. .. index:: --constant-lower-case (gnatpp) :switch:`--constant-lower-case, -cL` Constant object declaration names are lower case. .. index:: --constant-upper-case (gnatpp) :switch:`--constan-upper-case, -cU` Constant object declaration names are upper case. .. index:: --constant-mixed-case (gnatpp) :switch:`--constant-mixed-case, -cM` Constant object declaration names are mixed case. .. index:: --syntax-only (gnatpp) :switch:`--syntax-only` Disable the semantic analysis (name resolution) done by libadalang. This means ``gnatpp`` is not able to support any of the "as-declared" switches. .. index:: --dictionary (gnatpp) :switch:`--dictionary={file}, -D={file}` Use ``file`` as a *dictionary file* that defines the casing for a set of specified names, thereby overriding the effect on these names by any explicit or implicit -n switch. To supply more than one dictionary file, use several ``--dictionary`` switches. ``gnatpp`` implicitly uses a *default dictionary file* to define the casing for the Ada predefined names and the names declared in the GNAT libraries. .. index:: --dictionary=- (gnatpp) :switch:`--dictionary=-, -D=-` Do not use the default dictionary file; instead, use the casing defined by a :switch:`-n` switch and any explicit dictionary file(s) The structure of a dictionary file, and details on the conventions used in the default dictionary file, are defined in :ref:`Name_Casing`. The :switch:`--dictionary=-` and :switch:`--dictionary={file}` switches are mutually compatible. This group of ``gnatpp`` switches controls the layout of comments and complex syntactic constructs. See :ref:`Formatting_Comments` for details on their effect. .. _General_Text_Layout_Control: General Text Layout Control ^^^^^^^^^^^^^^^^^^^^^^^^^^^ These switches allow control over line length and indentation. .. index:: --max-line-length (gnatpp) :switch:`--max-line-length={nnn}, -M={nnn}` Maximum line length, ``nnn`` from 32...256. The default value is 79 .. index:: --indentation (gnatpp) :switch:`--indentation={nnn}, -i={nnn}` Indentation level, ``nnn`` from 1...9. The default value is 3 .. index:: --indent-continuation (gnatpp) :switch:`--indent-continuation={nnn}, -cl={nnn}` Indentation level for continuation lines (relative to the line being continued), ``nnn`` from 1...9. The default value is one less than the (normal) indentation level, unless the indentation is set to 1 (in which case the default value for continuation line indentation is also 1) .. _Other_Formatting_Options: Other Formatting Options ^^^^^^^^^^^^^^^^^^^^^^^^ These switches control other formatting not listed above. .. index:: --decimal-grouping (gnatpp) :switch:`--decimal-grouping={n}` Put underscores in decimal literals (numeric literals without a base) every ``n`` characters. If a literal already has one or more underscores, ``gnatpp`` will not modify it. For example, with ``--decimal-grouping=3``, ``1000000`` is changed to ``1_000_000``. .. index:: --based-grouping (gnatpp) :switch:`--based-grouping={n}` Same as ``--decimal-grouping``, but for based literals. For example, with ``--based-grouping=4``, ``16#0001FFFE#`` is changed to ``16#0001_FFFE#``. .. index:: --call-threshold (gnatpp) :switch:`--call-threshold={nnn}` If the number of parameter associations is greater than ``nnn`` and if at least one association uses named notation, start each association from a new line. If ``nnn`` is 0, ``gnatpp`` does not check for the number of associations; this is the default. .. index:: --par-threshold (gnatpp) :switch:`--par-threshold={nnn}` If the number of parameter specifications is greater than ``nnn`` (or equal to ``nnn`` in case of a function), start each specification from a new line. If ``nnn`` is 0, and :switch:`--no-separate-is` was not specified, then the ``is`` is placed on a separate line. This option is disabled by default. .. _Setting_the_Source_Search_Path: Setting the Source Search Path ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ To define the search path for the input source file, pass ``gnatpp`` the same switches as the GNAT compiler, with the same effects: .. index:: -I (gnatpp) :switch:`-I{dir}` .. index:: -I- (gnatpp) :switch:`-I-` .. index:: -gnatec (gnatpp) :switch:`-gnatec={path}` .. _Output_File_Control-gnatpp: Output File Control ^^^^^^^^^^^^^^^^^^^ By default, the output overwrites the input file. You can specify the location of the output with the following switches: .. index:: --replace (gnatpp) :switch:`--replace, -rnb` This is the default. Replace the input source file with the reformatted output without creating any backup copy of the input source. .. index:: --output-dir (gnatpp) :switch:`--output-dir={dir}` Generate the output file in directory :file:`dir` with the same name as the input file. If :file:`dir` is the same as the directory containing the input file, ``gnatpp`` does not read or process the input file; use ``--replace`` if you want to update the input file in place. .. index:: --pipe (gnatpp) :switch:`--pipe, -pipe` Send the output to ``Standard_Output`` .. index:: --output (gnatpp) :switch:`--output={output_file}, -o={output_file}` Write the output into ``output_file``. If ``output_file`` already exists, ``gnatpp`` terminates without reading or processing the input file. .. index:: --output-force (gnatpp) :switch:`--output-force={output_file}, -of={output_file}` Write the output into ``output_file``, overwriting the existing file (if one is present). .. index:: --replace-backup (gnatpp) :switch:`--replace-backup, -r` Replace the input source file with the reformatted output and copy the original input source into the file whose name is obtained by appending the :file:`.npp` suffix to the name of the input file. If a file with this name already exists, ``gnatpp`` terminates without reading or processing the input file. .. index:: --replace-force-backup (gnatpp) :switch:`--replace-force-backup, -rf` Like ``--replace-backup`` except that if the file with the specified name already exists, it is overwritten. .. index:: --eol (gnatpp) :switch:`--eol={xxx}` Specifies the line-ending style of the reformatted output file. The ``xxx`` string specified with the switch may be: * *dos* - MS DOS style, lines end with CR LF characters* * *crlf* - the same as *dos* * *unix* - UNIX style, lines end with LF character* * *lf* - the same as *unix* The default is to use the same end-of-line convention as the input. .. index:: --wide-character-encoding (gnatpp) :switch:`--wide-character-encoding={e}, -W={e}` Specify the wide character encoding method for the input and output files. ``e`` is one of the following: * *8* - UTF-8 encoding * *b* - Brackets encoding (default value) You may only specify options :switch:`--output-file` and :switch:`--output-force` if the call to ``gnatpp`` contains only one file to reformat. You may not use the option :switch:`--eol` or :switch:`--wide-character-encoding` together with the :switch:`--pipe` option. .. _Other_gnatpp_Switches: Other ``gnatpp`` Switches ^^^^^^^^^^^^^^^^^^^^^^^^^ You can also use the additional ``gnatpp`` switches defined in this subsection. .. index:: --version (gnatpp) :switch:`--version` Display copyright and version, then exit, disregarding all other options. .. index:: --help (gnatpp) :switch:`--help` Display usage, then exit, disregarding all other options. .. index:: -P (gnatpp) :switch:`-P {file}` Specifies the name of the project file that describes the set of sources to be processed. The exact set of argument sources depends on other options specified; see below. .. index:: -U (gnatpp) :switch:`-U` If you specify a project file but don't specify a source file, either directly or by means of a :switch:`--files` option, ``gnatpp`` processes all the units of the closure of the specified project. Otherwise this option has no effect. :switch:`-U {main_unit}` If you specify a project file but don't specify a source file, either directly or by means of :switch:`--files` option, ``gnatpp`` will process the closure of units rooted at ``main_unit``. Otherwise this option has no effect. .. index:: -X (gnatpp) :switch:`-X{name}={value}` Indicates that external variable ``name`` in the specified project has the value ``value``. Has no effect if you don't specify a project. .. index:: --RTS (gnatpp) :switch:`--RTS={rts-path}` Specifies the default location of the runtime library. It has the same meaning as the equivalent ``gnatmake`` switch (:ref:`Switches_for_gnatmake`). .. index:: --incremental (gnatpp) :switch:`--incremental` ``gnatpp`` will perform incremental processing on a per-file basis. It will only process a source file if it has been modified, or if files it depends on have been modified. This is similar to the way ``gnatmake``/``gprbuild`` only compiles files that need to be recompiled. You must specify a project file in this mode, and the gnat driver (as in *gnat pretty*) is not supported. (Note: this switch is not yet supported in the libadalang-based version of ``gnatpp``.) .. index:: --pp-off (gnatpp) :switch:`--pp-off={xxx}` Use :switch:`--xxx` as the command to turn off pretty printing, instead of the default ``--!pp off``. .. index:: --pp-on (gnatpp) :switch:`--pp-on={xxx}` Use :switch:`--xxx` as the command to turn pretty printing back on, instead of the default ``--!pp on``. .. index:: --files (gnatpp) :switch:`--files={filename}, -files={filename}` Take as arguments the files listed in text file ``file``, which may contain empty lines that are ignored. Each nonempty line should contain the name of an existing file. You may specify several such switches on the same command line. .. index:: --ignore (gnatpp) :switch:`--ignore={filename}` ``gnatpp`` will not process the sources listed in the specified file. You can't specify this option in incremental mode. .. index:: --jobs (gnatpp) :switch:`--jobs={n}, -j={n}` With :switch:`--incremental`, use *n* ``gnatpp`` processes to perform pretty printing in parallel. If *n* is 0, the maximum number processes is the number of core processors on the host. .. index:: --verbose (gnatpp) :switch:`--verbose, -v` Verbose mode .. index:: --quiet (gnatpp) :switch:`--quiet, -q` Quiet mode If you specify a project file, but no source files (either directly or by means of a :switch:`--files` option), and you specify the :switch:`-U`, then the set of processed sources is all the immediate units of the argument project. .. _Formatting_Rules: Formatting Rules ---------------- The following subsections show how ``gnatpp`` treats comments, program layout, and name casing as well as how to disable ``gnatpp`` in source code regions. They provide more details of the switches shown above. .. _Disabling_Pretty_Printing: Disabling Pretty Printing ^^^^^^^^^^^^^^^^^^^^^^^^^ Pretty printing is highly heuristic in nature and sometimes doesn't do exactly what you want. If you want to format a certain region of code by hand, you can turn off pretty printing in that region by surrounding it with special comments that start with ``--!pp off`` and ``--!pp on``. The text in that region is reproduced verbatim in the output with no formatting. To disable pretty printing for an entire file, put ``--!pp off`` at the top, with no following ``--!pp on``. You must place each of these comments on a line by themselves, with nothing preceding except spaces. You must have the initial text of the comment be exactly ``--!pp off`` or ``--!pp on`` (case sensitive), but you may follow it by arbitrary additional text. For example: .. code-block:: ada package Interrupts is --!pp off -- turn off pretty printing so "Interrupt_Kind" lines up type Interrupt_Kind is (Asynchronous_Interrupt_Kind, Synchronous_Interrupt_Kind, Green_Interrupt_Kind); --!pp on -- reenable pretty printing ... You can specify different comment strings using the :switch:`--pp-off` and :switch:`--pp-on` switches. For example, if you say: :: $ gnatpp --pp-off=' pp-' *.ad? ``gnatpp`` will recognize comments of the form ``-- pp-`` instead of ``--!pp off`` for disabling pretty printing. Note that you do not include the leading ``--`` of the comment in the argument to these switches. .. _Formatting_Comments: Formatting Comments ^^^^^^^^^^^^^^^^^^^ Only :switch:`--layout=compact` and :switch:`--layout=tall` format comments. Comments in Ada code are of two kinds: * *whole-line comments*, which appear by themselves (possibly preceded by white space) on a line * *end-of-line comments*, which follows some other Ada code on the same line. ``gnatpp`` indents whole-line comment according to the surrounding code, with some exceptions. Comments that start in column 1 are kept there. If possible, comments are not moved so far to the right that the maximum line length is exceeded. Special-form comments such as SPARK-style ``--#...`` are left alone. For an end-of-line comment, ``gnatpp`` tries to leave the same number of spaces between the end of the preceding Ada code and the beginning of the comment, as it appears in the original source. For each whole-line comment that does not end with two hyphens or that are not special-form comments, ``gnatpp`` inserts spaces if necessary after the starting two hyphens to ensure that there are at least two spaces between these hyphens and the first non-blank character of the comment. With :switch:`--layout=compact`, ``gnatpp`` fills in whole-line comments that form a paragraph in typical word processor style (that is, moving words between lines to make the lines other than the last similar in length). .. _Name_Casing: Name Casing ^^^^^^^^^^^ ``gnatpp`` always converts the usage occurrence of a (simple) name to the same casing as the corresponding defining identifier. You control the casing for defining occurrences via the :switch:`--name...` switches. With :switch:`--name-case-as-declared`, which is the default, defining occurrences appear exactly as in the source file where they are declared. The other values for this switch -- :switch:`--name-upper-case`, :switch:`--name-lower-case`, and :switch:`--name-mixed-case` -- result in upper, lower, or mixed case, respectively. If ``gnatpp`` changes the casing of a defining occurrence, it similarly changes the casing of all the usage occurrences of this name. If the defining occurrence of a name is not in the source compilation unit currently being processed by ``gnatpp``, ``gnatpp`` changes the casing of each reference to this name according to the switch (subject to the dictionary file mechanism described below). Thus ``gnatpp`` acts as though the switch had affected the casing for the defining occurrence of the name. The options :switch:`--attribute...`, :switch:`--keyword...`, :switch:`--enum...`, :switch:`--type...`, :switch:`--number...`, and :switch:`--pragma...` allow finer-grained control over casing for attributes, keywords, enumeration literals, types, named numbers and pragmas, respectively. :switch:`--type...` cover subtypes as well. You may need some names to be spelled with casing conventions that are not covered by the upper-, lower-, and mixed-case transformations. You can have ``gnatpp`` produce the correct casing by placing such names in a *dictionary file*, and specifying a :switch:`--dictionary` switch. Specifying any dictionary files overrides any :switch:`--name...` switch. ``gnatpp`` uses a default dictionary file to choose the casing of Ada predefined names and the names from GNAT libraries, Each predefined entity is converted to the same casing as the entity in the :title:`Ada Reference Manual` (usually mixed case) and each entity in the GNAT libraries is cased the same as its declaration in the library. You can specify the :switch:`--dictionary=-` switch to suppress the use of the default dictionary file. Instead, the casing for predefined and GNAT-defined names is given by the :switch:`-n` switch or explicit dictionary files. For example, by default the names ``Ada.Text_IO`` and ``GNAT.OS_Lib`` appear as just shown, even in the presence of a :switch:`--name-upper-case` switch. To ensure that even such names are rendered in uppercase, you must specify the :switch:`--dictionary=-` switch or place these names in upper case in a dictionary file. A dictionary file is a plain text file; each line in this file is either a blank line (containing only space characters), an Ada comment line, or the specification of exactly one *casing schema*. A casing schema is a string with the following syntax: :: casing_schema ::= identifier | simple_identifier simple_identifier ::= letter{letter_or_digit} (See :title:`Ada Reference Manual`, Section 2.3) for the definition of the ``identifier`` lexical element and the ``letter_or_digit`` category.) You can follow a casing schema string by white space and/or an Ada-style comment. You can also have any amount of white space before the string. If you pass a dictionary file as the value of a :switch:`--dictionary={file}` switch, ``gnatpp`` checks every simple name and identifier to see if the dictionary defines the casing for the name or for some of its parts (the term 'subword' is used below to denote the part of a name that is delimited by '_' or by the beginning or end of the word that doesn't contain any '_' characters): * if the complete name is in the dictionary, ``gnatpp`` uses the casing defined by the dictionary for this name; it does not check any subwords * for every subword, ``gnatpp`` checks if the dictionary contains the corresponding string of the form ``simple_identifier``, and if it does, the casing of this ``simple_identifier`` is used for this subword * if the complete name does not contain any '_' characters and if for this name the dictionary contains two entries -- one of the form ``identifier``, and another of the form ``simple_identifier`` -- ``gnatpp`` uses the first one to obtain the casing of this name * if you pass more than one dictionary file as ``gnatpp`` switches, each dictionary adds new casing exceptions and overrides all the existing casing exceptions set by the previous dictionaries * when ``gnatpp`` checks if the word or subword is in the dictionary, it uses a check that's not case sensitive For example, suppose we have the following source to reformat: .. code-block:: ada procedure test is name1 : integer := 1; name4_name3_name2 : integer := 2; name2_name3_name4 : Boolean; name1_var : Float; begin name2_name3_name4 := name4_name3_name2 > name1; end; And suppose we have two dictionaries: :: *dict1:* NAME1 *NaMe3* *Name1* *dict2:* *NAME3* If ``gnatpp`` is called with the following switches: :: $ gnatpp --name-mixed-case --dictionary=dict1 --dictionary=dict2 test.adb then we get the following name casing in the ``gnatpp`` output: .. code-block:: ada procedure Test is NAME1 : Integer := 1; Name4_NAME3_Name2 : Integer := 2; Name2_NAME3_Name4 : Boolean; Name1_Var : Float; begin Name2_NAME3_Name4 := Name4_NAME3_Name2 > NAME1; end Test; .. _Preprocessor_directives: Preprocessor Directives ^^^^^^^^^^^^^^^^^^^^^^^ ``gnatpp`` has some support for preprocessor directives. You can use preprocessor symbols such as ``$symbol``. In addition, you can use conditional compilation as long as the program text is syntactically legal Ada code after removing all the preprocessor directives (lines starting with ``#``). For example, ``gnatpp`` can format the following: .. code-block:: ada package P is #IF SOMETHING X : constant Integer := 123; #ELSE X : constant Integer := 456; #END IF; end P; which will be formatted as if it were: .. code-block:: ada package P is X : constant Integer := 123; X : constant Integer := 456; end P; except that the ``#`` lines will be preserved. However, ``gnatpp`` cannot format the following: .. code-block:: ada procedure P is begin #IF SOMETHING if X = 0 then #ELSE if X = 1 then #END IF; null; end if; end P; because removing the ``#`` lines gives: .. code-block:: ada procedure P is begin if X = 0 then if X = 1 then null; end if; end P; which is not syntactically legal. Legacy Switches ^^^^^^^^^^^^^^^ .. index:: --alignment (gnatpp) :switch:`--alignment, --no-alignment` Programs can be easier to read if certain constructs are vertically aligned. * ``:`` in declarations, * ``:=`` in initializations in declarations, * ``:=`` in assignment statements, * ``=>`` in associations, and * ``at`` keywords in the component clauses in record representation clauses. .. index:: --align-modes (gnatpp) :switch:`--align-modes, --no-align-modes` Line up ``in`` and ``out`` in parameter specifications. .. index:: --compact (gnatpp) :switch:`--compact, --no-compact` In calls and similar constructs, :switch:`--compact` packs as many subexpressions into the same line as is possible. Example: .. code-block:: ada Some_Procedure (Short_One, Another_Short_One, A_Very_Very_Very_Very_Very_Very_Very_Very_Long_One); On the other hand, if you specify :switch:`--no-compact`, ``gnatpp`` places all such subexpressions on separate lines if it's necessary to split a line between two subexpressions to avoid the construct exceeding :switch:`--max-line-length`. For example: .. code-block:: ada Some_Procedure (Short_One, Another_Short_One, A_Very_Very_Very_Very_Very_Very_Very_Very_Long_One); .. index:: --end-id (gnatpp) :switch:`--end-id, --no-end-id` Insert the name of a unit after ``end``. Otherwise, leave whatever comes after ``end``, if anything, alone. .. index:: --separate-is (gnatpp) :switch:`--separate-is, --no-separate-is` Place the keyword ``is`` on a separate line in a subprogram body in case if the spec occupies more than one line. .. index:: --separate-return (gnatpp) :switch:`--separate-return, --no-separate-return` If a subprogram spec does not fit on one line, place the ``return`` on a separate line. Otherwise try to place it in the same line as the last parameter specification. .. index:: --separate-loop (gnatpp) :switch:`--separate-loop, --no-separate-loop` Place the keyword ``loop`` in FOR and WHILE loop statements on a separate line. .. index:: --separate-then (gnatpp) :switch:`--separate-then, --no-separate-then` Place the keyword ``then`` in IF statements on a separate line. .. index:: --separate-loop-then (gnatpp) :switch:`--separate-loop-then, --no-separate-loop-then` Equivalent to :switch:`--separate-loop` :switch:`--separate-then`. .. index:: --split-line-before-record (gnatpp) :switch:`--split-line-before-record, --no-split-line-before-record` Split the line just before ``record`` in a record type declaration. .. index:: --indent-named-statements (gnatpp) :switch:`--indent-named-statements, --no-indent-named-statements` Indent block and loop statements with respect to the name. .. index:: --split-line-before-op (gnatpp) :switch:`--split-line-before-op, --no-split-line-before-op` If it is necessary to split a line at a binary operator, by default the line is split after the operator. When you specify this switch, it is split before the operator. .. index:: --use-on-new-line (gnatpp) :switch:`--use-on-new-line, --no-use-one-new-line` Start each USE clause in a context clause from a separate line. .. index:: --vertical-enum-types (gnatpp) :switch:`--vertical-enum-types, --no-vertical-enum-types` Format enumeration type declarations "vertically", e.g. each enumeration literal goes on a separate line. .. index:: --vertical-array-types (gnatpp) :switch:`--vertical-array-types, --no-vertical-array-types` Format array type declarations "vertically", e.g. for multidimensional arrays, each index_subtype_definition or discrete_subtype_definition goes on a separate line. .. index:: --vertical-named-aggregates (gnatpp) :switch:`--vertical-named-aggregates, --no-vertical-named-aggregates` Format aggregates "vertically" if named notation is used for all component_associations, e.g. each component_association goes on a separate line. .. index:: --vertical-case-alternatives (gnatpp) :switch:`--vertical-case-alternatives, --no-vertical-case-alternatives` Format case statements, case expressions, and variant parts with additional line breaks. .. index:: --RM-style-spacing (gnatpp) :switch:`--RM-style-spacing` Do not insert an extra blank before various occurrences of '(' and ':'. Alignment is off by default in this mode; use :switch:`--alignment` to turn it on. .. index:: --insert-blank-lines (gnatpp) :switch:`--insert-blank-lines, --no-insert-blank-lines` Insert blank lines where appropriate (between bodies and other large constructs). .. index:: --preserve-blank-lines (gnatpp) :switch:`--preserve-blank-lines, --no-preserve-blank-lines` Preserve blank lines in the input. By default, gnatpp will squeeze multiple blank lines down to one. .. index:: --preserve-line-breaks (gnatpp) :switch:`--preserve-line-breaks, --no-preserve-line-breaks` Preserve line breaks in the input, to the extent possible. By default, line breaks are also inserted at appropriate places. .. index:: --source-line-breaks (gnatpp) :switch:`--source-line-breaks, --no-source-line-breaks` Keep the line breaks from the source; do not insert or delete any line breaks. .. index:: --spaces-only (gnatpp) :switch:`--spaces-only, --no-spaces-only` Disable all formatting except for inserting and removing spaces. This implies --source-line-breaks. .. index:: -c (gnatpp) :switch:`--comments-unchanged, -c0` All comments remain unchanged. :switch:`--comments-gnat-indentation, -c1` GNAT-style comment line indentation. :switch:`--comments-gnat-beginning, -c3` GNAT-style comment beginning. :switch:`--comments-fill, -c4` Fill comment blocks. :switch:`--comments-special, -c5` Keep unchanged special form comments. :switch:`--comments-only` Format just the comments. .. only:: PRO or GPL .. _The_Body_Stub_Generator_gnatstub: The Body Stub Generator ``gnatstub`` ==================================== .. index:: ! gnatstub ``gnatstub`` creates empty but compilable bodies for library unit declarations and empty but compilable subunits for body stubs. ``gnatstub`` is a project-aware tool. (See :ref:`Using_Project_Files_with_GNAT_Tools` for a description of the project-related switches but note that ``gnatstub`` does not support the :switch:`-U`, :switch:`-U {main_unit}`, :switch:`--subdirs={dir}`, or :switch:`--no-objects-dir` switches.) The project file package that can specify ``gnatstub`` switches is named ``gnatstub``. By default, all program unit bodies generated by ``gnatstub`` raise ``Program_Error``, which will catch accidental calls of generated stubs. You can change this behavior with switch :switch:`--no-exception` (see below). .. _Running_gnatstub: Running ``gnatstub`` -------------------- You invoke ``gnatstub`` like this: :: $ gnatstub [ switches ] {filename} where * *filename* is the name of the source file that contains a library unit declaration for which you want a body to be created or a library unit body for which you want subunits to be created for the body stubs declared in this body. The file name may contain path information. If the name does not follow GNAT file naming conventions and the set of switches does not contain a project file that defines naming conventions, you must explicitly provide the name of the body file as the value of the :switch:`--output={body-name}` switch. If the file name follows the GNAT file naming conventions and you do not provide the name of the body file ``gnatstub`` uses the naming conventions for the generated source from the project file provided as a parameter of a :switch:`-P` switch if any, or creates the name file using the standard GNAT naming conventions. Note that you no longer need to specify the Ada language version; ``gnatstub`` can process Ada source code written in any version from Ada 83 onward without specifying any language version switch. * *switches* is an optional sequence of switches as described in the next section .. _Switches_for_gnatstub: Switches for ``gnatstub`` ------------------------- .. index:: --version (gnatstub) :switch:`--version` Display copyright and version, then exit, disregarding all other options. .. index:: --help (gnatstub) :switch:`--help` Display usage, then exit, disregarding all other options. .. index:: -P (gnatstub) :switch:`-P {file}` Indicates the name of the project file that describes the set of sources to be processed. You can specify an aggregate project as the file parameter only if it has exactly one non-aggregate project being aggregated. .. index:: -X (gnatstub) :switch:`-X{name}={value}` Indicates that external variable ``name`` in the argument project has the value ``value``. Has no effect if no project is specified. .. index:: --RTS (gnatstub) :switch:`--RTS={rts-path}` Specifies the default location of the runtime library. It has the same meaning as the equivalent ``gnatmake`` flag (:ref:`Switches_for_gnatmake`). .. index:: --subunits (gnatstub) :switch:`--subunits` Generate subunits for body stubs. If you specify this switch, ``gnatstub`` expects a library unit body as an argument file; otherwise a library unit declaration is expected. If a body stub already has a corresponding subunit, ``gnatstub`` does not generate anything for it. .. index:: --force (gnatstub) :switch:`--force` If the destination directory already contains a file with the name of the body file for the argument spec file, replace it with the generated body stub. This switch cannot be used together with :switch:`--subunits`. .. index:: --comment-header-spec (gnatstub) :switch:`--comment-header-spec` Put the comment header (i.e., all the comments preceding the compilation unit) from the source of the library unit declaration into the body stub. .. index:: --comment-header-sample (gnatstub) :switch:`--comment-header-sample` Put a sample comment header into the body stub. .. index:: --header-file (gnatstub) :switch:`--header-file={filename}` Use the content of the file as the comment header for a generated body stub. .. index:: --max-line-length (gnatstub) :switch:`--max-line-length={n}` (``n`` is a non-negative integer). Set the maximum line length for the output files. The default is 79. The maximum value that you can specify is 32767. .. index:: --indentation (gnatstub) :switch:`--indentation={n}` (``n`` is an integer from 1 to 9). Set the indentation level in the generated files to ``n``. The default indentation is 3. .. index:: --alphabetical-order (gnatstub) :switch:`--alphabetical-order` Order local bodies alphabetically. By default local bodies are ordered in the same way as the corresponding local specs in the argument spec file. .. index:: --no-exception (gnatstub) :switch:`--no-exception` Avoid raising Program_Error in the generated bodies of program unit stubs, except in the case of functions, where there will be value to return. .. index:: --no-local-header (gnatstub) :switch:`--no-local-header` Do not place a local comment header with unit name before body stub for a unit. .. index:: --files (gnatstub) :switch:`--files={filename}` Take as arguments the files listed in text file ``file``, which may contain empty lines that are ignored. You should specify the name of an existing file in each non-empty line. You may specify multiple :switch:`--files=` switches. .. index:: --output (gnatstub) :switch:`--output={body-name}` Body file name. You should set this if the argument file name does not follow the default GNAT file naming conventions and the naming conventions are not specified by a project file. If you omit both this switch and :switch:`-P`, ``gnatpp`` will choose the name for the body according to the default GNAT file naming conventions. .. index:: --output-dir (gnatstub) :switch:`--output-dir={dir-name}` The directory in which to place the output files. If you do not specify this switch, ``gnatpp`` places the generated library unit body in the current directory and generated sununits in the directory where the argument body is located. .. index:: --wide-character-encoding (gnatstub) :switch:`--wide-character-encoding={e}` Specify the wide character encoding method for the input and output files. ``e`` is one of the following: * *8* - UTF-8 encoding * *b* - Brackets encoding (default value) .. index:: --quiet (gnatstub) .. index:: -q (gnatstub) :switch:`--quiet` / :switch:`-q` Quiet mode. .. index:: --verbose (gnatstub) .. index:: -v (gnatstub) :switch:`--verbose` / :switch:`-v` Verbose mode. Legacy Switches ^^^^^^^^^^^^^^^ Some switches have a short form, mostly for legacy reasons, as shown below. .. index:: -M (gnatstub) :switch:`-gnatyM{nnn}` :switch:`--max-line-length={nnn}` .. index:: -i (gnatstub) :switch:`-i{nnn}` :switch:`--indentation={nnn}` .. index:: -gnaty (gnatstub) :switch:`-gnaty{nnn}` :switch:`--indentation={nnn}` .. index:: -f (gnatstub) :switch:`-f` :switch:`--force` .. index:: -gnatyo (gnatstub) :switch:`-gnatyo` :switch:`--alphabetical-order` .. index:: -hg (gnatstub) :switch:`-hg` :switch:`--comment-header-sample` .. index:: -hs (gnatstub) :switch:`-hs` :switch:`--comment-header-spec` .. index:: -o (gnatstub) :switch:`-o {output-file}` :switch:`--output={output-file}` .. index:: -dir (gnatstub) :switch:`-dir {dir-name}` :switch:`--output-dir={dir-name}` .. index:: -W (gnatstub) :switch:`-W{e}` :switch:`--wide-character-encoding={e}` .. index:: -files (gnatstub) :switch:`-files {filename}` :switch:`--files={filename}` .. only:: PRO or GPL .. _The_Backtrace_Symbolizer_gnatsymbolize: Translating Code Addresses into Source Locations with ``gnatsymbolize`` ======================================================================= .. index:: ! gnatsymbolize ``gnatsymbolize`` is a program which translates addresses into their corresponding filename, line number, and function names. Running ``gnatsymbolize`` ------------------------- :: $ gnatsymbolize [ switches ] filename [ addresses ] For example, consider the following Ada program: .. code-block:: ada package Pck is Global_Val : Integer := 0; procedure Call_Me_First; end Pck; with GNAT.IO; use GNAT.IO; with GNAT.Traceback; use GNAT.Traceback; with GNAT.Debug_Utilities; package body Pck is procedure Call_Me_Third is TB : Tracebacks_Array (1 .. 5); TB_len : Natural; begin Global_Val := Global_Val + 1; Call_Chain (TB, TB_Len); for K in 1 .. TB_Len loop Put_Line (GNAT.Debug_Utilities.Image_C (TB (K))); end loop; end Call_Me_Third; procedure Call_Me_Second is begin Call_Me_Third; end Call_Me_Second; procedure Call_Me_First is begin Call_Me_Second; end Call_Me_First; end Pck; with GNAT.IO; use GNAT.IO; with GNAT.Debug_Utilities; with GNAT.Traceback; with System; with Pck; use Pck; procedure Foo is LA : constant System.Address := \ GNAT.Traceback.Executable_Load_Address; use type System.Address; begin if LA /= System.Null_Address then Put_Line ("Load address: " & GNAT.Debug_Utilities.Image_C (LA)); end if; Global_Val := 123; Call_Me_First; end Foo; This program, when built and run, prints a list of addresses which correspond to the traceback when inside function ``Call_Me_Third``. For example, on x86-64 GNU/Linux: :: $ gnatmake -g -q foo.adb $ ./foo Load address: 0x00005586C9D7D000 0x00005586C9D81105 0x00005586C9D8119B 0x00005586C9D811A7 0x00005586C9D8128C 0x00005586C9D81069 You can use ``gnatsymbolize`` to translate those addresses into code locations as follows: :: $ gnatsymbolize --load foo 0x00005586C9D7D000 0x00005586C9D81105 \ 0x00005586C9D8119B 0x00005586C9D811A7 0x00005586C9D8128C \ 0x00005586C9D81069 0x5586c9d81105 Pck.Call_Me_Third at pck.adb:12 0x5586c9d8119b Pck.Call_Me_Second at pck.adb:20 0x5586c9d811a7 Pck.Call_Me_First at pck.adb:25 0x5586c9d8128c Foo at foo.adb:6 0x5586c9d81069 Main at b~foo.adb:199 Switches for ``gnatsymbolize`` ------------------------------ You can specify the following switches for ``gnatsymbolize``: .. index:: --help (gnatsymbolize) :switch:`--help` Display the program's usage, and then exit, disregarding all other options. :switch:`--cache` Read the symbolic information from the executable and cache them in memory to accelerate the translation of each address into a symbolic location. Depending on the size of the executable and the number of addresses to translate, this may not always make ``gnatsymbolize`` faster overall. :switch:`--dump` If you have specified :switch:`--cache`, dump the contents of the cache to Standard Output. Has no effect otherwise. :switch:`--count={N}` Compute the symbolic traceback ``N`` times in a row. You use this switch mostly for measuring the performance of ``gnatsymbolize``, particularly in the case where you have specified the cache to be used. :switch:`--load` Interpret the first address as the load address of the executable. This is needed for position-independent executables on Linux and Windows. Requirements for Correct Operation ---------------------------------- The translation is performed by reading the DWARF debugging information produced by the compiler for each unit. You must therefore compile all units for which the translation is to be done in a way that DWARF debugging information is produced. In most cases, you do this by simply compiling with :switch:`-g`. This program provides a functionality similar to ``addr2line``. It has fewer options to tailor its output, but has been designed to require fewer DWARF sections to be present in the executable. In particular, it works for code compiled with ``-g1``. .. only:: PRO or GPL .. _Using_Project_Files_with_GNAT_Tools: Using Project Files with GNAT Tools =================================== This section describes how you can use project files in conjunction with a number of GNAT tools. For a comprehensive description of project files and the overall GNAT Project Manager facility, please refer to the *GNAT Project Manager* chapter in the *GPRbuild and GPR Companion Tools User's Guide*. .. index:: Project-aware tool If a tool can take a project file as an option and extract the needed information, such a tool is called a *project-aware* tool. .. _Switches_Related_to_Project_Files: Switches Related to Project Files --------------------------------- The following switches are used by the project-aware GNAT tools: :switch:`-P{project_file}` Indicates the name of the project file whose source files are to be processed. The exact set of sources depends on other options specified, see below. :switch:`-U` If you supply a project file, say for project ``proj``, but you don't specify any sources for ``proj`` (either by a project attribute or through a tool option that provides a list of the files to be used), process all the source files from projects imported either directly or indirectly by ``proj``. Otherwise this option has no effect. :switch:`-U {source_file}` Similar to :switch:`-U`, but if you don't specify any sources, process only those source files for units in the closure of the Ada source contained in ``source_file``. Note that this option expects the source file name, not the Ada unit name, as its parameter. :switch:`-X{name}={val}` Indicates that the external variable ``name`` in the project has the value ``val``. Has no effect if no project has been specified. :switch:`--subdirs={dir}` Use the ``dir`` subdirectory of the project's object directory (or the ``dir`` subdirectory of the project file directory if the project does not specify an object directory) for tool output files. Has no effect if you haven't specified a project and if you haven't specified the if :switch:`--no-objects-dir` switch. :switch:`--no-objects-dir` Place all the result files into the current directory (i.e., the directory from which the tool invocation command is issued) instead of the project's object directory. Has no effect if you haven't specified a project. :switch:`-eL` Follow all symbolic links when processing project files. If you specify a project file but neither a :switch:`-U` option, nor a :switch:`-U {main_unit}` option, nor any other explicit option to specify the source files, then the sources to be processed are the immediate sources of the specified project (i.e., the source files directly defined by that project, either implicitly by residing in the project source directories or explicitly through any of the source-related attributes). .. _Tool-specific_packages_in_project files: Tool-specific packages in project files --------------------------------------- Each project-aware tool may have a corresponding package in a project file; the package names are given elsewhere in this manual, in the sections that describe the respective tools. A tool-specific package in a project file may define the ``Default_Switches`` attribute indexed by "ada" (as language name). You set the value of this attribute to a list of switches that you want the tool to use when it's invoked. You cannot specify project-specific switches through this attribute. --- ### Gcc Ada Doc Gnat Ugn Inline Assembler (`gcc/ada/doc/gnat_ugn/inline_assembler.rst`) .. role:: switch(samp) .. _Inline_Assembler: **************** Inline Assembler **************** .. index:: Inline Assembler If you need to write low-level software that interacts directly with the hardware, Ada provides two ways for you to incorporate assembly language code into your program. First, you can import and invoke external routines written in assembly language, an Ada feature fully supported by GNAT. However, for small sections of code, it may be simpler or more efficient to include assembly language statements directly in your Ada source program, using the facilities of the implementation-defined package ``System.Machine_Code``, which incorporates the gcc Inline Assembler. The Inline Assembler approach offers a number of advantages, including the following: * No need to use non-Ada tools * Consistent interface over different targets * Automatic usage of the proper calling conventions * Access to Ada constants and variables * Definition of intrinsic routines * Possibility of inlining a subprogram consisting of assembler code * Code optimizer can take Inline Assembler code into account This appendix presents a series of examples to show you how to use the Inline Assembler. Although it focuses on the Intel x86, the general approach applies also to other processors. It is assumed you are familiar with both Ada and assembly language programming. .. _Basic_Assembler_Syntax: Basic Assembler Syntax ====================== The assembler used by GNAT and gcc is based not on the Intel assembly language, but rather on a language that descends from the AT&T Unix assembler ``as`` (and which is often referred to as 'AT&T syntax'). The following table summarizes the main features of ``as`` syntax and points out the differences from the Intel conventions. See the gcc ``as`` and ``gas`` (an ``as`` macro pre-processor) documentation for further information. | *Register names* | gcc / ``as``: Prefix with '%'; for example ``%eax`` | Intel: No extra punctuation; for example ``eax`` | *Immediate operand* | gcc / ``as``: Prefix with '$'; for example ``$4`` | Intel: No extra punctuation; for example ``4`` | *Address* | gcc / ``as``: Prefix with '$'; for example ``$loc`` | Intel: No extra punctuation; for example ``loc`` | *Memory contents* | gcc / ``as``: No extra punctuation; for example ``loc`` | Intel: Square brackets; for example ``[loc]`` | *Register contents* | gcc / ``as``: Parentheses; for example ``(%eax)`` | Intel: Square brackets; for example ``[eax]`` | *Hexadecimal numbers* | gcc / ``as``: Leading '0x' (C language syntax); for example ``0xA0`` | Intel: Trailing 'h'; for example ``A0h`` | *Operand size* | gcc / ``as``: Explicit in op code; for example ``movw`` to move a 16-bit word | Intel: Implicit, deduced by assembler; for example ``mov`` | *Instruction repetition* | gcc / ``as``: Split into two lines; for example | ``rep`` | ``stosl`` | Intel: Keep on one line; for example ``rep stosl`` | *Order of operands* | gcc / ``as``: Source first; for example ``movw $4, %eax`` | Intel: Destination first; for example ``mov eax, 4`` .. _A_Simple_Example_of_Inline_Assembler: A Simple Example of Inline Assembler ==================================== The following example generate a single assembly language statement, ``nop``, which does nothing. Despite its lack of run-time effect, the example is useful in illustrating the basics of the Inline Assembler facility. .. code-block:: ada with System.Machine_Code; use System.Machine_Code; procedure Nothing is begin Asm ("nop"); end Nothing; ``Asm`` is a procedure declared in package ``System.Machine_Code``; here it takes one parameter, a *template string* that must be a static expression that produces the generated instruction. ``Asm`` may be regarded as a compile-time procedure that parses the template string and any additional parameters (none, in this case) and generates one or more assembly language instructions. The examples in this chapter will illustrate several of the forms for invoking ``Asm``; a complete specification of the syntax is found in the ``Machine_Code_Insertions`` section of the :title:`GNAT Reference Manual`. Under the standard GNAT conventions, you should put the ``Nothing`` procedure in a file named :file:`nothing.adb`. You can build the executable in the usual way: :: $ gnatmake nothing However, the interesting aspect of this example is not its run-time behavior but rather the generated assembly code. To see this output, invoke the compiler as follows: :: $ gcc -c -S -fomit-frame-pointer -gnatp nothing.adb where the options are: * :switch:`-c` compile only (no bind or link) * :switch:`-S` generate assembler listing * :switch:`-fomit-frame-pointer` do not set up separate stack frames * :switch:`-gnatp` do not add runtime checks This gives a human-readable assembler version of the code. The resulting file has the same name as the Ada source file but with a ``.s`` extension. In our example, the file :file:`nothing.s` has the following contents: :: .file "nothing.adb" gcc2_compiled.: ___gnu_compiled_ada: .text .align 4 .globl __ada_nothing __ada_nothing: #APP nop #NO_APP jmp L1 .align 2,0x90 L1: ret The assembly code you included is clearly indicated by the compiler, between the ``#APP`` and ``#NO_APP`` delimiters. The character before the 'APP' and 'NOAPP' can differ on different targets. For example, GNU/Linux uses '#APP' while on NT you will see '/APP'. If you make a mistake in your assembler code (such as using the wrong size modifier or using a wrong operand for the instruction) GNAT will report this error in a temporary file, which is deleted when the compilation is finished. Generating an assembler file will help in such cases, since you can assemble this file separately using the ``as`` assembler that comes with gcc. Assembling the file using the command :: $ as nothing.s will give you error messages whose lines correspond to the assembler input file, so you can easily find and correct any mistakes you made. If there are no errors, ``as`` generates an object file called :file:`nothing.out`. .. _Output_Variables_in_Inline_Assembler: Output Variables in Inline Assembler ==================================== The examples in this section, showing how to access the processor flags, illustrate how to specify the destination operands for assembly language statements. .. code-block:: ada with Interfaces; use Interfaces; with Ada.Text_IO; use Ada.Text_IO; with System.Machine_Code; use System.Machine_Code; procedure Get_Flags is Flags : Unsigned_32; use ASCII; begin Asm ("pushfl" & LF & HT & -- push flags on stack "popl %%eax" & LF & HT & -- load eax with flags "movl %%eax, %0", -- store flags in variable Outputs => Unsigned_32'Asm_Output ("=g", Flags)); Put_Line ("Flags register:" & Flags'Img); end Get_Flags; We have separated multiple assembler statements in the Asm template string with linefeed (ASCII.LF) and horizontal tab (ASCII.HT) characters in order to have a nicely aligned assembly listing. The resulting section of the assembly output file is: :: #APP pushfl popl %eax movl %eax, -40(%ebp) #NO_APP It would have been legal to write the Asm invocation as: .. code-block:: ada Asm ("pushfl popl %%eax movl %%eax, %0") but in the generated assembler file, this would come out as: :: #APP pushfl popl %eax movl %eax, -40(%ebp) #NO_APP which is not so convenient for the human reader. We use Ada comments at the end of each line to explain what the assembler instructions actually do. This is a useful convention. When writing Inline Assembler instructions, you need to precede each register and variable name with a percent sign. Since the assembler already requires a percent sign at the beginning of a register name, you need two consecutive percent signs for such names in the Asm template string, thus ``%%eax``. In the generated assembly code, one of the percent signs will be stripped off. Names such as ``%0``, ``%1``, ``%2``, etc., denote input or output variables: operands you later define using ``Input`` or ``Output`` parameters to ``Asm``. An output variable is shown in the third section of the Asm template string: :: movl %%eax, %0 The intent of this section is to store the contents of the ``eax`` register in a variable that can be accessed in Ada. Simply writing ``movl %%eax, Flags`` would not necessarily work, since the compiler might optimize by using a register to hold ``Flags``, and the expansion of the ``movl`` instruction would not be aware of this optimization. The solution is not to store the result directly but rather to advise the compiler to choose the correct operand form; that is the purpose of the ``%0`` output variable. Information about the output variable is supplied in the ``Outputs`` parameter to ``Asm``: .. code-block:: ada Outputs => Unsigned_32'Asm_Output ("=g", Flags)); The output is defined by the ``Asm_Output`` attribute of the target type; the general format is: .. code-block:: ada Type'Asm_Output (constraint_string, variable_name) The constraint string directs the compiler how to store/access the associated variable. In the example: .. code-block:: ada Unsigned_32'Asm_Output ("=m", Flags); the ``"m"`` (memory) constraint tells the compiler that the variable ``Flags`` should be stored in a memory variable, thus preventing the optimizer from keeping it in a register. In contrast, .. code-block:: ada Unsigned_32'Asm_Output ("=r", Flags); uses the ``"r"`` (register) constraint, telling the compiler to store the variable in a register. If you precede the constraint with the equal character ('='), it tells the compiler that the variable will have data stored into it. In the ``Get_Flags`` example, we used the ``"g"`` (global) constraint, allowing the optimizer to choose whatever operand it deems best. There are a fairly large number of constraints, but the ones that are most useful for the Intel x86 processor are the following: ====== ========================================== *=* output constraint *g* global (i.e., can be stored anywhere) *m* in memory *I* a constant *a* use eax *b* use ebx *c* use ecx *d* use edx *S* use esi *D* use edi *r* use one of eax, ebx, ecx or edx *q* use one of eax, ebx, ecx, edx, esi or edi ====== ========================================== The full set of constraints is described in the ``gcc`` and ``as`` documentation; note that you can combine certain constraints into one constraint string. You specify the association of an output variable with an assembler operand through the :samp:`%{n}` notation, where *n* is a non-negative integer. Thus in .. code-block:: ada Asm ("pushfl" & LF & HT & -- push flags on stack "popl %%eax" & LF & HT & -- load eax with flags "movl %%eax, %0", -- store flags in variable Outputs => Unsigned_32'Asm_Output ("=g", Flags)); ``%0`` is replaced in the expanded code by the appropriate operand, whatever the compiler chose for the ``Flags`` variable. In general, you may have any number of output variables: * Count the operands starting at 0; thus ``%0``, ``%1``, etc. * Specify the ``Outputs`` parameter as a parenthesized comma-separated list of ``Asm_Output`` attributes For example: .. code-block:: ada Asm ("movl %%eax, %0" & LF & HT & "movl %%ebx, %1" & LF & HT & "movl %%ecx, %2", Outputs => (Unsigned_32'Asm_Output ("=g", Var_A), -- %0 = Var_A Unsigned_32'Asm_Output ("=g", Var_B), -- %1 = Var_B Unsigned_32'Asm_Output ("=g", Var_C))); -- %2 = Var_C where ``Var_A``, ``Var_B``, and ``Var_C`` are variables in the Ada program. As a variation on the ``Get_Flags`` example, we can use the constraint string to direct the compiler to store the ``eax`` register into the ``Flags`` variable, instead of including the store instruction explicitly in the ``Asm`` template string: .. code-block:: ada with Interfaces; use Interfaces; with Ada.Text_IO; use Ada.Text_IO; with System.Machine_Code; use System.Machine_Code; procedure Get_Flags_2 is Flags : Unsigned_32; use ASCII; begin Asm ("pushfl" & LF & HT & -- push flags on stack "popl %%eax", -- save flags in eax Outputs => Unsigned_32'Asm_Output ("=a", Flags)); Put_Line ("Flags register:" & Flags'Img); end Get_Flags_2; The ``"a"`` constraint tells the compiler that the ``Flags`` variable will come from the ``eax`` register. Here is the resulting code: :: #APP pushfl popl %eax #NO_APP movl %eax,-40(%ebp) The compiler generated the store of eax into Flags after expanding the assembler code. In fact, there was no need to pop the flags into the ``eax`` register; more simply, we could just pop the flags directly into the program variable: .. code-block:: ada with Interfaces; use Interfaces; with Ada.Text_IO; use Ada.Text_IO; with System.Machine_Code; use System.Machine_Code; procedure Get_Flags_3 is Flags : Unsigned_32; use ASCII; begin Asm ("pushfl" & LF & HT & -- push flags on stack "pop %0", -- save flags in Flags Outputs => Unsigned_32'Asm_Output ("=g", Flags)); Put_Line ("Flags register:" & Flags'Img); end Get_Flags_3; .. _Input_Variables_in_Inline_Assembler: Input Variables in Inline Assembler =================================== The example in this section illustrates how to specify the source operands for assembly language statements. The procedure simply increments its input value by 1: .. code-block:: ada with Interfaces; use Interfaces; with Ada.Text_IO; use Ada.Text_IO; with System.Machine_Code; use System.Machine_Code; procedure Increment is function Incr (Value : Unsigned_32) return Unsigned_32 is Result : Unsigned_32; begin Asm ("incl %0", Outputs => Unsigned_32'Asm_Output ("=a", Result), Inputs => Unsigned_32'Asm_Input ("a", Value)); return Result; end Incr; Value : Unsigned_32; begin Value := 5; Put_Line ("Value before is" & Value'Img); Value := Incr (Value); Put_Line ("Value after is" & Value'Img); end Increment; The ``Outputs`` parameter to ``Asm`` specifies that the result is in the ``eax`` register and that it is to be stored in the ``Result`` variable. The ``Inputs`` parameter looks much like the ``Outputs`` parameter, but with an ``Asm_Input`` attribute. The ``"="`` constraint, indicating an output value, is not present. You can have multiple input variables in the same way you can have more than one output variable. The parameter count (%0, %1) etc, still starts at the first output statement, and continues with the input statements. Just as the ``Outputs`` parameter causes the register to be stored into the target variable after execution of the assembler statements, the ``Inputs`` parameter causes its variable to be loaded into the register before execution of the assembler statements. Thus the effect of the ``Asm`` invocation is: * load the 32-bit value of ``Value`` into ``eax`` * execute the ``incl %eax`` instruction * store the contents of eax into the ``Result`` variable The resulting assembler file (with :switch:`-O2` optimization) contains: :: _increment__incr.1: subl $4,%esp movl 8(%esp),%eax #APP incl %eax #NO_APP movl %eax,%edx movl %ecx,(%esp) addl $4,%esp ret .. _Inlining_Inline_Assembler_Code: Inlining Inline Assembler Code ============================== For a short subprogram such as the ``Incr`` function in the previous section, the overhead of the call and return (creating / deleting the stack frame) can be significant, compared to the amount of code in the subprogram body. A solution is to apply Ada's ``Inline`` pragma to the subprogram, which directs the compiler to expand invocations of the subprogram at the point(s) of call, instead of setting up a stack frame for out-of-line calls. Here's the resulting program: .. code-block:: ada with Interfaces; use Interfaces; with Ada.Text_IO; use Ada.Text_IO; with System.Machine_Code; use System.Machine_Code; procedure Increment_2 is function Incr (Value : Unsigned_32) return Unsigned_32 is Result : Unsigned_32; begin Asm ("incl %0", Outputs => Unsigned_32'Asm_Output ("=a", Result), Inputs => Unsigned_32'Asm_Input ("a", Value)); return Result; end Incr; pragma Inline (Increment); Value : Unsigned_32; begin Value := 5; Put_Line ("Value before is" & Value'Img); Value := Increment (Value); Put_Line ("Value after is" & Value'Img); end Increment_2; Compile the program with both optimization (:switch:`-O2`) and inlining (:switch:`-gnatn`) enabled. The ``Incr`` function is still compiled as usual, but at the point in ``Increment`` where our function used to be called: :: pushl %edi call _increment__incr.1 the code for the function body directly appears: :: movl %esi,%eax #APP incl %eax #NO_APP movl %eax,%edx thus saving the overhead of stack frame setup and an out-of-line call. .. _Other_Asm_Functionality: Other ``Asm`` Functionality =========================== This section describes two important parameters to the ``Asm`` procedure: ``Clobber``, which identifies register usage; and ``Volatile``, which inhibits unwanted optimizations. .. _The_Clobber_Parameter: The ``Clobber`` Parameter ------------------------- One of the dangers of intermixing assembly language and a compiled language such as Ada is that the compiler needs to be aware of which registers are being used by the assembly code. In some cases, such as the earlier examples, the constraint string is sufficient to indicate register usage (e.g., ``"a"`` for the ``eax`` register). But, more generally, the compiler needs an explicit identification of the registers that are used by the Inline Assembly statements. Using a register that the compiler doesn't know about could be a side effect of an instruction (like ``mull``, which stores its result into both ``eax`` and ``edx``). It can also arise from explicit register usage within your assembly code; for example: .. code-block:: ada Asm ("movl %0, %%ebx" & LF & HT & "movl %%ebx, %1", Outputs => Unsigned_32'Asm_Output ("=g", Var_Out), Inputs => Unsigned_32'Asm_Input ("g", Var_In)); where the compiler (since it does not analyze the ``Asm`` template string) does not know you are using the ``ebx`` register. In such cases you need to supply the ``Clobber`` parameter to ``Asm``, to identify the registers used by your assembly code: .. code-block:: ada Asm ("movl %0, %%ebx" & LF & HT & "movl %%ebx, %1", Outputs => Unsigned_32'Asm_Output ("=g", Var_Out), Inputs => Unsigned_32'Asm_Input ("g", Var_In), Clobber => "ebx"); The Clobber parameter is a static string expression specifying the register(s) you are using. Note that register names are *not* prefixed by a percent sign. Also, if more than one register is used, you separate their names by commas; e.g., ``"eax, ebx"`` The ``Clobber`` parameter has several additional uses: * Use 'register' name ``cc`` to indicate that flags might have changed * Use 'register' name ``memory`` if you changed a memory location .. _The_Volatile_Parameter: The ``Volatile`` Parameter -------------------------- .. index:: Volatile parameter Compiler optimizations in the presence of Inline Assembler may sometimes have unwanted effects. For example, when an ``Asm`` invocation with an input variable is inside a loop, the compiler might move the loading of the input variable outside the loop, regarding it as a one-time initialization. If you don't want this to happen, you can disable such optimizations by setting the ``Volatile`` parameter to ``True``; for example: .. code-block:: ada Asm ("movl %0, %%ebx" & LF & HT & "movl %%ebx, %1", Outputs => Unsigned_32'Asm_Output ("=g", Var_Out), Inputs => Unsigned_32'Asm_Input ("g", Var_In), Clobber => "ebx", Volatile => True); By default, ``Volatile`` is set to ``False`` unless there is no ``Outputs`` parameter. Although setting ``Volatile`` to ``True`` prevents unwanted optimizations, it also disables other optimizations that might be important for efficiency. In general, you should set ``Volatile`` to ``True`` only if the compiler's optimizations have created problems. --- ### Gcc Ada Doc Gnat Ugn Platform Specific Information (`gcc/ada/doc/gnat_ugn/platform_specific_information.rst`) .. role:: switch(samp) .. -- Non-breaking space in running text -- E.g. Ada |nbsp| 95 .. |nbsp| unicode:: 0xA0 :trim: .. _Platform_Specific_Information: ***************************** Platform-Specific Information ***************************** This appendix contains information relating to the implementation of run-time libraries on various platforms. It also covers topics related to the GNAT implementation on specific Operating Systems. .. _`Run_Time_Libraries`: Run-Time Libraries ================== .. index:: Tasking and threads libraries .. index:: Threads libraries and tasking .. index:: Run-time libraries (platform-specific information) The GNAT run-time implementation may vary with respect to both the underlying threads library and the exception-handling scheme. For threads support, the default run-time will bind to the thread package of the underlying operating system. For exception handling, either or both of two models are supplied: .. index:: Zero-Cost Exceptions .. index:: ZCX (Zero-Cost Exceptions) * **Zero-Cost Exceptions** ("ZCX"), which uses binder-generated tables that are interrogated at run time to locate a handler. .. index:: setjmp/longjmp Exception Model .. index:: SJLJ (setjmp/longjmp Exception Model) * **setjmp / longjmp** ('SJLJ'), which uses dynamically-set data to establish the set of handlers Most programs should experience a substantial speed improvement by being compiled with a ZCX run-time. This is especially true for tasking applications or applications with many exception handlers. Note however that the ZCX run-time does not support asynchronous abort of tasks (``abort`` and ``select-then-abort`` constructs) and will instead implement abort by polling points in the runtime. You can also add additional polling points explicitly if needed in your application via ``pragma Abort_Defer``. This section summarizes which combinations of threads and exception support are supplied on various GNAT platforms. .. _Summary_of_Run-Time_Configurations: Summary of Run-Time Configurations ---------------------------------- +-----------------+--------------+-------------------------+------------+ | Platform | Run-Time | Tasking | Exceptions | +=================+==============+=========================+============+ | GNU/Linux | rts-native | pthread library | ZCX | | | (default) | | | | +--------------+-------------------------+------------+ | | rts-sjlj | pthread library | SJLJ | +-----------------+--------------+-------------------------+------------+ | Windows | rts-native | native Win32 threads | ZCX | | | (default) | | | | +--------------+-------------------------+------------+ | | rts-sjlj | native Win32 threads | SJLJ | +-----------------+--------------+-------------------------+------------+ | Mac OS | rts-native | pthread library | ZCX | +-----------------+--------------+-------------------------+------------+ .. _Specifying_a_Run-Time_Library: Specifying a Run-Time Library ============================= The :file:`adainclude` subdirectory containing the sources of the GNAT run-time library and the :file:`adalib` subdirectory containing the :file:`ALI` files and the static and/or shared GNAT library are located in the gcc target-dependent area: :: target=$prefix/lib/gcc/gcc-*dumpmachine*/gcc-*dumpversion*/ As indicated above, on some platforms, several run-time libraries are supplied. These libraries are installed in the target dependent area and contain a complete source and binary subdirectory. The detailed description below explains the differences between the different libraries in terms of their thread support. The default run-time library (when GNAT is installed) is *rts-native*. This default run-time is selected by the means of soft links. For example on x86-linux: .. -- -- $(target-dir) -- | -- +--- adainclude----------+ -- | | -- +--- adalib-----------+ | -- | | | -- +--- rts-native | | -- | | | | -- | +--- adainclude <---+ -- | | | -- | +--- adalib <----+ -- | -- +--- rts-sjlj -- | -- +--- adainclude -- | -- +--- adalib .. only:: html or latex .. image:: rtlibrary-structure.png .. only:: not (html or latex) :: $(target-dir) __/ / \ \___ _______/ / \ \_________________ / / \ \ / / \ \ ADAINCLUDE ADALIB rts-native rts-sjlj : : / \ / \ : : / \ / \ : : / \ / \ : : / \ / \ +-------------> adainclude adalib adainclude adalib : ^ : : +---------------------+ Run-Time Library Directory Structure (Upper-case names and dotted/dashed arrows represent soft links) If you want to select the *rts-sjlj* library on a permanent basis, you can modify these soft links with the following commands: :: $ cd $target $ rm -f adainclude adalib $ ln -s rts-sjlj/adainclude adainclude $ ln -s rts-sjlj/adalib adalib Alternatively, you can specify :file:`rts-sjlj/adainclude` in the file :file:`$target/ada_source_path` and :file:`rts-sjlj/adalib` in :file:`$target/ada_object_path`. .. index:: --RTS switch You can select another run-time library temporarily by using the :switch:`--RTS` switch, e.g., :switch:`--RTS=sjlj` .. index:: Linux .. index:: GNU/Linux .. _GNU_Linux_Topics: GNU/Linux Topics ================ This section describes topics that are specific to GNU/Linux platforms. .. _Required_packages_on_GNU_Linux: Required Packages on GNU/Linux ------------------------------ GNAT requires the C library developer's package to be installed. The name of of that package depends on your GNU/Linux distribution: * RedHat, SUSE: ``glibc-devel``; * Debian, Ubuntu: ``libc6-dev`` (normally installed by default). If you're using the 32-bit version of GNAT on a 64-bit version of GNU/Linux, you'll need the 32-bit version of the following packages: * RedHat, SUSE: ``glibc.i686``, ``glibc-devel.i686``, ``ncurses-libs.i686`` * SUSE: ``glibc-locale-base-32bit`` * Debian, Ubuntu: ``libc6:i386``, ``libc6-dev:i386``, ``lib32ncursesw5`` Other GNU/Linux distributions might choose different name for those packages. .. _PIE_Enabled_By_Default_On_Linux: Position Independent Executable (PIE) Enabled by Default on Linux ----------------------------------------------------------------- GNAT generates Position Independent Executable (PIE) code by default. PIE binaries are loaded into random memory locations, introducing an additional layer of protection against attacks. Building PIE binaries requires that all of their dependencies also be built as Position Independent. If the link of your project fails with an error like:: /[...]/ld: /path/to/object/file: relocation R_X86_64_32S against symbol `symbol name' can not be used when making a PIE object; recompile with -fPIE it means the identified object file has not been built as Position Independent. If you are not interested in building PIE binaries, you can simply turn this feature off by first compiling your code with :samp:`-fno-pie` and then by linking with :samp:`-no-pie` (note the subtle but important difference in the names of the switches -- the linker switch does **not** have an `f` after the dash!). When using gprbuild, you do this by updating the *Required_Switches* attribute in package `Compiler` and, depending on your type of project, either attribute *Switches* or attribute *Library_Options* in package `Linker`. On the other hand, if you would like to build PIE binaries and you are getting the error above, a quick and easy workaround to allow linking to succeed again is to disable PIE during the link, thus temporarily lifting the requirement that all dependencies also be Position Independent code. To do so, you simply need to add :samp:`-no-pie` to the list of switches passed to the linker. As part of this workaround, there is no need to adjust the compiler switches. From there, to be able to link your binaries with PIE and therefore drop the :samp:`-no-pie` workaround, you'll need to get the identified dependencies rebuilt with PIE enabled (compiled with :samp:`-fPIE` and linked with :samp:`-pie`). .. _Choosing_the_Scheduling_Policy_With_GNU_Linux: .. index:: SCHED_FIFO scheduling policy .. index:: SCHED_RR scheduling policy .. index:: SCHED_OTHER scheduling policy Choosing the Scheduling Policy with GNU/Linux --------------------------------------------- When using a POSIX threads implementation, you have a choice of several scheduling policies: ``SCHED_FIFO``, ``SCHED_RR`` and ``SCHED_OTHER``. Typically, the default is ``SCHED_OTHER``, while using ``SCHED_FIFO`` or ``SCHED_RR`` requires special (e.g., root) privileges. .. index:: pragma Time_Slice .. index:: -T0 switch .. index:: pragma Task_Dispatching_Policy By default, GNAT uses the ``SCHED_OTHER`` policy. To specify ``SCHED_FIFO``, you can use one of the following: * ``pragma Time_Slice (0.0)`` * the corresponding binder switch :switch:`-T0` * ``pragma Task_Dispatching_Policy (FIFO_Within_Priorities)`` To specify ``SCHED_RR``, you should use ``pragma Time_Slice`` with a value greater than 0.0, or else use the corresponding :switch:`-T` binder switch. To make sure a program is running as root, you can put something like this in a library package body in your application: .. code-block:: ada function geteuid return Integer; pragma Import (C, geteuid, "geteuid"); Ignore : constant Boolean := (if geteuid = 0 then True else raise Program_Error with "must be root"); This gets the effective user id and if it's not 0 (i.e. root), it raises Program_Error. Note that if you're running the code in a container, this may not be sufficient as you may have sufficient privilege on the container, but not on the host machine running the container, so check that you also have sufficient privilege for running the container image. .. _A_GNU_Linux_debug_quirk: A GNU/Linux Debug Quirk ----------------------- On SuSE 15, some kernels have a defect causing issues when debugging programs using threads or Ada tasks. Due to the lack of documentation found regarding this kernel issue, we can only provide limited information about which kernels are impacted. Kernel version 5.3.18 is known to be impacted and kernels in the 5.14 range or newer are believed to fix this problem. The bug affects the debugging of 32-bit processes on a 64-bit system. Symptoms can vary: Unexpected ``SIGABRT`` signals being received by the program, "The futex facility returned an unexpected error code" error message, and inferior programs hanging indefinitely range among the symptoms most commonly observed. .. index:: Windows .. _Microsoft_Windows_Topics: Microsoft Windows Topics ======================== This section describes topics that are specific to the Microsoft Windows platforms. .. only:: PRO .. rubric:: Installing from the Command Line By default the GNAT installers display a GUI that prompts you to enter the installation path and similar information and then guides you through the installation process. You can also perform silent installations using the command-line interface. To install one of the GNAT installers from the command line, you should pass parameter :switch:`/S` (and, optionally, :switch:`/D=`) as command-line arguments. For example, for an unattended installation of GNAT 19.2 into the default directory :file:`C:\\GNATPRO\\19.2` you would run:: gnatpro-19.2-x86-windows-bin /S To install into a custom directory, say, :file:`C:\\TOOLS\\GNATPRO\\19.2`:: gnatpro-19.2-x86-windows-bin /S /D=C:\TOOLS\GNATPRO\19.2 You can use the same syntax for all installers. Note that unattended installations don't modify system path, nor create file associations, so you need to do such activities by hand. .. _Using_GNAT_on_Windows: Using GNAT on Windows --------------------- One of the strengths of the GNAT technology is that its tool set (``gcc``, ``gnatbind``, ``gnatlink``, ``gnatmake``, the ``gdb`` debugger, etc.) is used in the same way regardless of the platform. On Windows, this tool set is complemented by a number of Microsoft-specific tools that have been provided to facilitate interoperability with Windows when this is required. With these tools: * You can build applications using the ``CONSOLE`` or ``WINDOWS`` subsystems. * You can use any Dynamically Linked Library (DLL) in your Ada code (both relocatable and non-relocatable DLLs are supported). * You can build Ada DLLs for use in other applications. You can write these applications in a language other than Ada (e.g., C, C++, etc). Again, both relocatable and non-relocatable Ada DLLs are supported. * You can include Windows resources in your Ada application. * You can use or create COM/DCOM objects. Listed immediately below are all known general GNAT-for-Windows restrictions. We list other restrictions about specific features such as Windows Resources and DLLs in separate sections below. * You cannot use ``GetLastError`` and ``SetLastError`` when tasking, protected records, or exceptions are used. In these cases, in order to implement Ada semantics, the GNAT run-time system calls certain Win32 routines that set the last error variable to 0 upon success. You may be able to use ``GetLastError`` and ``SetLastError`` when tasking, protected record, and exception features are not used, but it is not guaranteed to work. * You cannot link against Microsoft C++ libraries except for import libraries. You must do interfacing by means of DLLs. * You can link against Microsoft C libraries. However, the preferred solution is to use C/C++ compiler that comes with GNAT, since it doesn't require having two different development environments and makes the inter-language debugging experience smoother. * When the compilation environment is located on FAT32 drives, you may experience recompilations of source files that have not changed if Daylight Saving Time (DST) state has changed since the last time files were compiled. NTFS drives do not have this problem. * No components of the GNAT toolset use any entries in the Windows registry. The only entries installation of GNAT may create are file associations and PATH settings, provided you chose to create them at installation time, as well as some minimal bookkeeping information needed to correctly uninstall or integrate different GNAT products. .. _Using_a_network_installation_of_GNAT: Using a network installation of GNAT ------------------------------------ Make sure the system on which GNAT is installed is accessible from the current machine, i.e., the install location is shared over the network. Shared resources are accessed on Windows by means of UNC paths, which have the format ``\\\\server\\sharename\\path`` In order to use such a network installation, simply add the UNC path of the :file:`bin` directory of your GNAT installation in front of your PATH. For example, if GNAT is installed in :file:`\\GNAT` directory of a share location called :file:`c-drive` on a machine :file:`LOKI`, the following command will make it available: :: $ path \\loki\c-drive\gnat\bin;%path%` Be aware that every compilation using the network installation results in the transfer of large amounts of data across the network and will likely cause a serious performance penalty. .. _CONSOLE_and_WINDOWS_subsystems: CONSOLE and WINDOWS subsystems ------------------------------ .. index:: CONSOLE Subsystem .. index:: WINDOWS Subsystem .. index:: -mwindows There are two main subsystems under Windows. The ``CONSOLE`` subsystem (which is the default subsystem) always creates a console when launching the application. This is not something desirable when the application has a Windows GUI. To remove this console, your application must use the ``WINDOWS`` subsystem. To do so, you must specify the :switch:`-mwindows` linker switch. :: $ gnatmake winprog -largs -mwindows .. _Temporary_Files: Temporary Files --------------- .. index:: Temporary files You can control where temporary files get created by setting the :envvar:`TMP` environment variable. The file will be created: * Under the directory pointed to by the :envvar:`TMP` environment variable if this directory exists. * Under :file:`c:\\temp`, if the :envvar:`TMP` environment variable is not set (or not pointing to a directory) and if this directory exists. * Under the current working directory otherwise. This allows you to determine exactly where the temporary file will be created. This is particularly useful in networked environments where you may not have write access to some directories. Disabling Command Line Argument Expansion ----------------------------------------- .. index:: Command Line Argument Expansion By default, an executable compiled for the Windows platform will do the following postprocessing on the arguments passed on the command line: * If the argument contains the characters ``*`` and/or ``?``, file expansion will be attempted. For example, if the current directory contains :file:`a.txt` and :file:`b.txt`, then when calling:: $ my_ada_program *.txt The following arguments will effectively be passed to the main program (for example when using ``Ada.Command_Line.Argument``):: Ada.Command_Line.Argument (1) -> "a.txt" Ada.Command_Line.Argument (2) -> "b.txt" * You can disable filename expansion for a given argument by using single quotes. Thus, calling:: $ my_ada_program '*.txt' will result in:: Ada.Command_Line.Argument (1) -> "*.txt" Note that if the program is launched from a shell such as Cygwin Bash, quote removal might be performed by that shell. In some contexts, it might be useful to disable this feature (for example if the program performs its own argument expansion). In order to do this, a C symbol needs to be defined and set to ``0``. You can do this by adding the following code fragment in one of your Ada units: .. code-block:: ada Do_Argv_Expansion : Integer := 0; pragma Export (C, Do_Argv_Expansion, "__gnat_do_argv_expansion"); The results of previous examples will be respectively:: Ada.Command_Line.Argument (1) -> "*.txt" and:: Ada.Command_Line.Argument (1) -> "'*.txt'" .. _Choosing_the_Scheduling_Policy_With_Windows: Choosing the Scheduling Policy with Windows ------------------------------------------- Under Windows, the standard 31 priorities of the Ada model are mapped onto Window's seven standard priority levels by default: Idle, Lowest, Below Normal, Normal, Above Normal, When using the ``FIFO_Within_Priorities`` task dispatching policy, GNAT assigns the ``REALTIME_PRIORITY_CLASS`` priority class to the application and maps the Ada priority range to the sixteen priorities made available under ``REALTIME_PRIORITY_CLASS``. For details on the values of the different priority mappings, see declarations in :file:`system.ads`. For more information about Windows priorities, please refer to Microsoft documentation. Windows Socket Timeouts ----------------------- Microsoft Windows desktops older than ``8.0`` and Microsoft Windows Servers older than ``2019`` set a socket timeout 500 milliseconds longer than the value set by setsockopt with ``SO_RCVTIMEO`` and ``SO_SNDTIMEO`` options. The GNAT runtime makes a correction for the difference in the corresponding Windows versions. For Windows Server starting with version ``2019``, you must provide a manifest file for the GNAT runtime to be able to recognize that the Windows version does not need the timeout correction. The manifest file should be located in the same directory as the executable file and its file name must match the executable name suffixed by ``.manifest``. For example, if the executable name is :file:`sock_wto.exe`, the manifest file name must be :file:`sock_wto.exe.manifest`. The manifest file must contain at least the following data:: Without the manifest file, the socket timeout will be overcorrected on these Windows Server versions and the actual time will be 500 milliseconds shorter than what was set with ``GNAT.Sockets.Set_Socket_Option``. Note that on Microsoft Windows versions where correction is necessary, there is no way to set a socket timeout shorter than 500 ms. If a socket timeout shorter than 500 ms is needed on these Windows versions, you should add a call to ``Check_Selector`` before any socket read or write operations. .. _Mixed-Language_Programming_on_Windows: Mixed-Language Programming on Windows ------------------------------------- Developing pure Ada applications on Windows is no different than on other GNAT-supported platforms. However, when developing or porting an application that contains a mix of Ada and C/C++, the choice of your Windows C/C++ development environment conditions your overall interoperability strategy. If you use ``gcc`` or Microsoft C to compile the non-Ada part of your application, there are no Windows-specific restrictions that affect the overall interoperability with your Ada code. If you do want to use the Microsoft tools for your C++ code, you have two choices: * You can encapsulate your C++ code in a DLL to be linked with your Ada application. In this case, use the Microsoft or other environment to build the DLL and use GNAT to build your executable (:ref:`Using_DLLs_with_GNAT`). * You can encapsulate your Ada code in a DLL to be linked with the other part of your application. In this case, use GNAT to build the DLL (:ref:`Building_DLLs_with_GNAT_Project_files`) and use the Microsoft or other environment to build your executable. In addition to the description about C ``main`` in :ref:`Mixed_Language_Programming` section, if the C ``main`` uses a stand-alone library, it is required on x86-windows to setup the SEH context. For this, the C ``main`` must looks like this: .. code-block:: c /* main.c */ extern void adainit (void); extern void adafinal (void); extern void __gnat_initialize(void*); extern void call_to_ada (void); int main (int argc, char *argv[]) { int SEH [2]; /* Initialize the SEH context */ __gnat_initialize (&SEH); adainit(); /* Then call Ada services in the stand-alone library */ call_to_ada(); adafinal(); } Note that you need not do this on x86_64-windows where the Windows native SEH support is used. .. _Windows_Calling_Conventions: Windows Calling Conventions ^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: Stdcall .. index:: APIENTRY This section pertain only to Win32. On Win64, there is a single native calling convention. All convention specifiers are ignored on this platform. When a subprogram ``F`` (caller) calls a subprogram ``G`` (callee), there are several ways to push ``G``\ 's parameters on the stack and there are several possible scenarios to clean up the stack upon ``G``\ 's return. A calling convention is an agreed upon software protocol whereby the responsibilities between the caller (``F``) and the callee (``G``) are clearly defined. Several calling conventions are available for Windows: * ``C`` (Microsoft defined) * ``Stdcall`` (Microsoft defined) * ``Win32`` (GNAT specific) * ``DLL`` (GNAT specific) .. _C_Calling_Convention: ``C`` Calling Convention """""""""""""""""""""""" This is the default calling convention used when interfacing to C/C++ routines compiled with either ``gcc`` or Microsoft Visual C++. In the ``C`` calling convention, subprogram parameters are pushed on the stack by the caller from right to left. The caller itself is in charge of cleaning up the stack after the call. In addition, the name of a routine with ``C`` calling convention is mangled by adding a leading underscore. The name to use on the Ada side when importing (or exporting) a routine with ``C`` calling convention is the name of the routine. For example you should import the C function: :: int get_val (long); from Ada as follows: .. code-block:: ada function Get_Val (V : Interfaces.C.long) return Interfaces.C.int; pragma Import (C, Get_Val, External_Name => "get_val"); Note that in this particular case, you could have omitted the ``External_Name`` parameter since, when missing, this parameter is set to the name of the Ada entity in lower case. When the ``Link_Name`` parameter is missing, as in the above example, this parameter is set the ``External_Name`` with a leading underscore. When importing a variable defined in C, you should always use the ``C`` calling convention unless the object containing the variable is part of a DLL (in which case you should use the ``Stdcall`` calling convention, :ref:`Stdcall_Calling_Convention`). .. _Stdcall_Calling_Convention: ``Stdcall`` Calling Convention """""""""""""""""""""""""""""" This convention, which was the calling convention used for Pascal programs, is used by Microsoft for all the routines in the Win32 API for efficiency reasons. You must use it to import any routine for which this convention was specified. In the ``Stdcall`` calling convention, subprogram parameters are also pushed on the stack by the caller from right to left. However, the callee, not the caller, is in charge of cleaning up the stack on routine exit. In addition, the name of a routine with ``Stdcall`` calling convention is mangled by adding a leading underscore (as for the ``C`` calling convention) and a trailing :samp:`@{nn}`, where ``nn`` is the overall size (in bytes) of the parameters passed to the routine. The name to use on the Ada side when importing a C routine with a ``Stdcall`` calling convention is the name of the C routine. The leading underscore and trailing :samp:`@{nn}` are added automatically by the compiler. For example, you could import the Win32 function: :: APIENTRY int get_val (long); from Ada as follows: .. code-block:: ada function Get_Val (V : Interfaces.C.long) return Interfaces.C.int; pragma Import (Stdcall, Get_Val); -- On the x86 a long is 4 bytes, so the Link_Name is "_get_val@4" Like the case for the ``C`` calling convention, when the ``External_Name`` parameter is missing, it is the name of the Ada entity in lower case. If instead of writing the above import pragma you write: .. code-block:: ada function Get_Val (V : Interfaces.C.long) return Interfaces.C.int; pragma Import (Stdcall, Get_Val, External_Name => "retrieve_val"); the imported routine is ``_retrieve_val@4``. However, if instead of specifying the ``External_Name`` parameter, you specify the ``Link_Name`` as in the following example: .. code-block:: ada function Get_Val (V : Interfaces.C.long) return Interfaces.C.int; pragma Import (Stdcall, Get_Val, Link_Name => "retrieve_val"); the imported routine is ``retrieve_val``. There is no decoration at all; no leading underscore and no Stdcall suffix :samp:`@{nn}`. This is especially important as in some special cases a DLL's entry point name lacks a trailing :samp:`@{nn}` while the exported name generated for a call has it. You can also import variables defined in a DLL by using an import pragma for a variable. As an example, if a DLL contains a variable defined as: .. code-block:: c int my_var; then, to access this variable from Ada you should write: .. code-block:: ada My_Var : Interfaces.C.int; pragma Import (Stdcall, My_Var); Note that to ease building cross-platform bindings, this convention will be handled as a ``C`` calling convention on non-Windows platforms. .. _Win32_Calling_Convention: ``Win32`` Calling Convention """""""""""""""""""""""""""" This convention, which is GNAT-specific, is fully equivalent to the ``Stdcall`` calling convention described above. .. _DLL_Calling_Convention: ``DLL`` Calling Convention """""""""""""""""""""""""" This convention, which is GNAT-specific, is fully equivalent to the ``Stdcall`` calling convention described above. .. _Introduction_to_Dynamic_Link_Libraries_DLLs: Introduction to Dynamic Link Libraries (DLLs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: DLL A Dynamically Linked Library (DLL) is a library that can be shared by several applications running under Windows. A DLL can contain any number of routines and variables. One advantage of DLLs is that you can change and enhance them without forcing all the applications that depend on them to be relinked or recompiled. However, you should be aware than all calls to DLL routines are slower since, as you will understand below, such calls are indirect. To illustrate the remainder of this section, suppose that an application wants to use the services of a DLL :file:`API.dll`. To use the services provided by :file:`API.dll`, you must statically link against the DLL or an import library which contains a jump table with an entry for each routine and variable exported by the DLL. In the Microsoft world, this import library is called :file:`API.lib`. When using GNAT, this import library is called either :file:`libAPI.dll.a`, :file:`libapi.dll.a`, :file:`libAPI.a` or :file:`libapi.a` (names are case insensitive). After you have linked your application with the DLL or the import library and you run your application, here is what happens: * Your application is loaded into memory. * The DLL :file:`API.dll` is mapped into the address space of your application. This means that: - The DLL uses the stack of the calling thread. - The DLL uses the virtual address space of the calling process. - The DLL allocates memory from the virtual address space of the calling process. - Handles (pointers) can be safely exchanged between routines in the DLL routines and routines in the application using the DLL. * The entries in the jump table (from the import library :file:`libAPI.dll.a` or :file:`API.lib` or automatically created when linking against a DLL) which is part of your application are initialized with the addresses of the routines and variables in :file:`API.dll`. * If present in :file:`API.dll`, routines ``DllMain`` or ``DllMainCRTStartup`` are invoked. These routines typically contain the initialization code needed for the well-being of the routines and variables exported by the DLL. There is an additional point which is worth mentioning. In the Windows world, there are two kind of DLLs: relocatable and non-relocatable DLLs. Non-relocatable DLLs can only be loaded at a specific address in the target application address space. If the addresses of two non-relocatable DLLs overlap and these happen to be used by the same application, a conflict occurs and the application will run incorrectly. Hence, when possible, you should always use and build relocatable DLLs. Both relocatable and non-relocatable DLLs are supported by GNAT. Note that the :switch:`-s` linker switch (see GNU Linker User's Guide) removes the debugging symbols from the DLL, but the DLL can still be relocated. As a side note, an interesting difference between Microsoft DLLs and Unix shared libraries is the fact that on most Unix systems all public routines are exported by default in a Unix shared library, while under Windows it is possible (but not required) to list exported routines in a definition file (see :ref:`The Definition File `). .. _Using_DLLs_with_GNAT: Using DLLs with GNAT ^^^^^^^^^^^^^^^^^^^^ To use the services of a DLL, say :file:`API.dll`, in your Ada application you must have: * The Ada spec for the routines and/or variables you want to access in :file:`API.dll`. If not available, you must build this Ada spec from the C/C++ header files provided with the DLL. * The import library (:file:`libAPI.dll.a` or :file:`API.lib`). As previously mentioned, an import library is a statically linked library containing the import table, which is filled at load time to point to the actual :file:`API.dll` routines. Sometimes you don't have an import library for the DLL you want to use. The following sections will explain how to build one. Note that this is optional. * The actual DLL, :file:`API.dll`. Once you have all the above, to compile an Ada application that uses the services of :file:`API.dll` and whose main subprogram is ``My_Ada_App``, you simply issue the command :: $ gnatmake my_ada_app -largs -lAPI The argument :switch:`-largs -lAPI` at the end of the ``gnatmake`` command tells the GNAT linker to look for an import library. The linker will look for a library name in this specific order: * :file:`libAPI.dll.a` * :file:`API.dll.a` * :file:`libAPI.a` * :file:`API.lib` * :file:`libAPI.dll` * :file:`API.dll` The first three are the GNU-style import libraries. The third is the Microsoft-style import libraries. The last two are the actual DLL names. Note that if the Ada package spec for :file:`API.dll` contains the following pragma .. code-block:: ada pragma Linker_Options ("-lAPI"); you do not have to add :switch:`-largs -lAPI` at the end of the ``gnatmake`` command. If any one of the items above is missing, you will have to create it yourself. The following sections explain how to do so using as an example a fictitious DLL called :file:`API.dll`. .. _Creating_an_Ada_Spec_for_the_DLL_Services: Creating an Ada Spec for the DLL Services """"""""""""""""""""""""""""""""""""""""" A DLL typically comes with a C/C++ header file which provides the definitions of the routines and variables exported by the DLL. The Ada equivalent of this header file is a package spec that contains definitions for the imported entities. If the DLL you intend to use does not come with an Ada spec, you have to generate such a spec yourself. For example, if the header file of :file:`API.dll` is a file :file:`api.h` containing the following two definitions: .. code-block:: c int some_var; int get (char *); then the equivalent Ada spec could be: .. code-block:: ada with Interfaces.C.Strings; package API is use Interfaces; Some_Var : C.int; function Get (Str : C.Strings.Chars_Ptr) return C.int; private pragma Import (C, Get); pragma Import (DLL, Some_Var); end API; .. _Creating_an_Import_Library: Creating an Import Library """""""""""""""""""""""""" .. index:: Import library If a Microsoft-style import library :file:`API.lib` or a GNAT-style import library :file:`libAPI.dll.a` or :file:`libAPI.a` is available with :file:`API.dll` you can skip this section. You can also skip this section if :file:`API.dll` or :file:`libAPI.dll` is built with GNU tools as in this case it is possible to link directly against the DLL. Otherwise read on. .. index:: Definition file .. _The_Definition_File: .. rubric:: The Definition File As previously mentioned, and unlike Unix systems, the list of symbols that are exported from a DLL must be provided explicitly in Windows. The main goal of a definition file is precisely that: list the symbols exported by a DLL. A definition file (usually a file with a ``.def`` suffix) has the following structure: :: [LIBRARY ``name``] [DESCRIPTION ``string``] EXPORTS ``symbol1`` ``symbol2`` ... *LIBRARY name* This section, which is optional, gives the name of the DLL. *DESCRIPTION string* This section, which is optional, gives a description string that will be embedded in the import library. *EXPORTS* This section gives the list of exported symbols (procedures, functions or variables). For example, in the case of :file:`API.dll` the ``EXPORTS`` section of :file:`API.def` looks like: :: EXPORTS some_var get Note that you must specify the correct suffix (:samp:`@{nn}`) (see :ref:`Windows_Calling_Conventions`) for a Stdcall calling convention function in the exported symbols list. There can actually be other sections in a definition file, but these sections are not relevant to the discussion at hand. .. _Create_Def_File_Automatically: .. rubric:: Creating a Definition File Automatically You can automatically create the definition file :file:`API.def` (see :ref:`The Definition File `) from a DLL. To do that, use the ``dlltool`` program as follows: :: $ dlltool API.dll -z API.def --export-all-symbols Note that if some routines in the DLL have the ``Stdcall`` convention (:ref:`Windows_Calling_Conventions`) with stripped :samp:`@{nn}` suffix then you'll have to edit :file:`api.def` to add it and specify :switch:`-k` to ``gnatdll`` when creating the import library. Here are some hints to find the right :samp:`@{nn}` suffix. - If you have the Microsoft import library (.lib), you may be able to find the right symbols by using the Microsoft ``dumpbin`` tool (see the corresponding Microsoft documentation for further details). :: $ dumpbin /exports api.lib - If you get a message about a missing symbol at link time, the compiler tells you what symbol is expected. You then can go back to the definition file and add the right suffix. .. _GNAT-Style_Import_Library: .. rubric:: GNAT-Style Import Library To create a static import library from :file:`API.dll` with the GNAT tools, you should create the :file:`.def` file and use the ``gnatdll`` tool (see :ref:`Using_gnatdll`) as follows: :: $ gnatdll -e API.def -d API.dll ``gnatdll`` takes as input a definition file :file:`API.def` and the name of the DLL containing the services listed in the definition file :file:`API.dll`. The name of the static import library generated is computed from the name of the definition file as follows: if the definition file name is :file:`xyz.def`, the import library name will be :file:`libxyz.a`. Note that in the previous example, the switch :switch:`-e` could have been removed because the name of the definition file (before the ``.def`` suffix) is the same as the name of the DLL (:ref:`Using_gnatdll` for more information about ``gnatdll``). .. _MSVS-Style_Import_Library: .. rubric:: Microsoft-Style Import Library A Microsoft import library is needed only if you plan to make an Ada DLL available to applications developed with Microsoft tools (:ref:`Mixed-Language_Programming_on_Windows`). To create a Microsoft-style import library for :file:`API.dll` you should create the :file:`.def` file, then build the actual import library using Microsoft's ``lib`` utility: :: $ lib -machine:IX86 -def:API.def -out:API.lib If you use the above command, the definition file :file:`API.def` must contain a line giving the name of the DLL: :: LIBRARY "API" See the Microsoft documentation for further details about the usage of ``lib``. .. _Building_DLLs_with_GNAT_Project_files: Building DLLs with GNAT Project files ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: DLLs, building There is nothing specific to Windows in the build process. See the *Library Projects* section in the *GNAT Project Manager* chapter of the *GPRbuild User's Guide*. Due to a system limitation, you cannot create threads under Windows when inside the ``DllMain`` routine which is used for auto-initialization of shared libraries, so you can't have library level tasks in SALs. .. _Building_DLLs_with_GNAT: Building DLLs with GNAT ^^^^^^^^^^^^^^^^^^^^^^^ .. index:: DLLs, building This section explains how to build DLLs using the GNAT built-in DLL support. With the following procedure, it is straightforward to build and use DLLs with GNAT. * Building object files. The first step is to build all objects files that are to be included into the DLL. This is done using the standard ``gnatmake`` tool. * Building the DLL. To build the DLL, you must use the ``gcc`` :switch:`-shared` and :switch:`-shared-libgcc` switches. It's quite simple to use this method: :: $ gcc -shared -shared-libgcc -o api.dll obj1.o obj2.o ... It's important to note that in this case all symbols found in the object files are automatically exported. You can restrict the set of symbols to export by passing to ``gcc`` a definition file (see :ref:`The Definition File `). For example: :: $ gcc -shared -shared-libgcc -o api.dll api.def obj1.o obj2.o ... If you use a definition file, you must export the elaboration procedures for every package that requires one. Elaboration procedures are named using the package name followed by "_E". * Preparing DLL to be used. For the DLL to be used by client programs, the bodies must be hidden from it and the :file:`.ali` set with read-only attribute. This is very important because otherwise GNAT will recompile all packages and will not actually use the code in the DLL. For example: :: $ mkdir apilib $ copy *.ads *.ali api.dll apilib $ attrib +R apilib\\*.ali At this point, you can use the DLL by directly linking against it. Note that you must use the GNAT shared runtime when using GNAT shared libraries. You do this with the :switch:`-shared` binder switch. :: $ gnatmake main -Iapilib -bargs -shared -largs -Lapilib -lAPI .. _Building_DLLs_with_gnatdll: Building DLLs with gnatdll ^^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: DLLs, building Note that it is preferred to use GNAT Project files (:ref:`Building_DLLs_with_GNAT_Project_files`) or the built-in GNAT DLL support (:ref:`Building_DLLs_with_GNAT`) to build DLLs. This section explains how to build DLLs containing Ada code using ``gnatdll``. These DLLs will be referred to as Ada DLLs in the remainder of this section. The steps required to build an Ada DLL that is to be used by Ada as well as non-Ada applications are as follows: * You need to mark each Ada entity exported by the DLL with a ``C`` or ``Stdcall`` calling convention to avoid any Ada name mangling for the entities exported by the DLL (see :ref:`Exporting Ada Entities `). You can skip this step if you plan to use the Ada DLL only from Ada applications. * Your Ada code must export an initialization routine which calls the routine ``adainit`` (generated by ``gnatbind``) to perform the elaboration of the Ada code in the DLL (:ref:`Ada_DLLs_and_Elaboration`). The initialization routine exported by the Ada DLL must be invoked by the clients of the DLL to initialize the DLL. * When useful, the DLL should also export a finalization routine which calls routine ``adafinal`` (also generated by ``gnatbind``) to perform the finalization of the Ada code in the DLL (:ref:`Ada_DLLs_and_Finalization`). The finalization routine exported by the Ada DLL must be invoked by the clients of the DLL when the DLL services are no further needed. * You must provide a spec for the services exported by the Ada DLL in each of the programming languages to which you plan to make the DLL available. * You must provide a definition file listing the exported entities (:ref:`The Definition File `). * Finally, you must use ``gnatdll`` to produce the DLL and the import library (:ref:`Using_gnatdll`). Note that a relocatable DLL stripped using the ``strip`` binutils tool is no longer relocatable. To build a DLL without debug information, pass :switch:`-largs -s` to ``gnatdll``. This restriction does not apply to a DLL built using a Library Project. See the *Library Projects* section in the *GNAT Project Manager* chapter of the *GPRbuild User's Guide*. .. Limitations_When_Using_Ada_DLLs_from Ada: Limitations When Using Ada DLLs from Ada """""""""""""""""""""""""""""""""""""""" When using Ada DLLs from Ada applications there is a limitation you should be aware of. On Windows, the GNAT run-time is not in a DLL of its own, so each Ada DLL includes a part of the GNAT run-time. Specifically, each Ada DLL includes the services of the GNAT run-time that are necessary for the Ada code inside the DLL. As a result, when an Ada program uses an Ada DLL there are two independent GNAT run-times: one in the Ada DLL and one in the main program. It is therefore not possible to exchange GNAT run-time objects between the Ada DLL and the main Ada program. Example of GNAT run-time objects are file handles (e.g., ``Text_IO.File_Type``), tasks types, protected objects types, etc. It is completely safe to exchange plain elementary, array or record types, Windows object handles, etc. .. _Exporting_Ada_Entities: Exporting Ada Entities """""""""""""""""""""" .. index:: Export table Building a DLL is a way to encapsulate a set of services usable from any application. As a result, the Ada entities exported by a DLL should be exported with the ``C`` or ``Stdcall`` calling conventions to avoid any Ada name mangling. As an example here is an Ada package ``API``, spec and body, exporting two procedures, a function, and a variable: .. code-block:: ada with Interfaces.C; use Interfaces; package API is Count : C.int := 0; function Factorial (Val : C.int) return C.int; procedure Initialize_API; procedure Finalize_API; -- Initialization & Finalization routines. More in the next section. private pragma Export (C, Initialize_API); pragma Export (C, Finalize_API); pragma Export (C, Count); pragma Export (C, Factorial); end API; .. code-block:: ada package body API is function Factorial (Val : C.int) return C.int is Fact : C.int := 1; begin Count := Count + 1; for K in 1 .. Val loop Fact := Fact * K; end loop; return Fact; end Factorial; procedure Initialize_API is procedure Adainit; pragma Import (C, Adainit); begin Adainit; end Initialize_API; procedure Finalize_API is procedure Adafinal; pragma Import (C, Adafinal); begin Adafinal; end Finalize_API; end API; If the Ada DLL you are building will only be used by Ada applications, you do not have to export Ada entities with a ``C`` or ``Stdcall`` convention. As an example, the previous package could be written as follows: .. code-block:: ada package API is Count : Integer := 0; function Factorial (Val : Integer) return Integer; procedure Initialize_API; procedure Finalize_API; -- Initialization and Finalization routines. end API; .. code-block:: ada package body API is function Factorial (Val : Integer) return Integer is Fact : Integer := 1; begin Count := Count + 1; for K in 1 .. Val loop Fact := Fact * K; end loop; return Fact; end Factorial; ... -- The remainder of this package body is unchanged. end API; Note that if you do not export the Ada entities with a ``C`` or ``Stdcall`` convention, you will have to provide the mangled Ada names in the definition file of the Ada DLL (:ref:`Creating_the_Definition_File`). .. _Ada_DLLs_and_Elaboration: Ada DLLs and Elaboration """""""""""""""""""""""" .. index:: DLLs and elaboration The DLL that you are building contains your Ada code as well as all the routines in the Ada library that are needed by it. The first thing a user of your DLL must do is elaborate the Ada code (:ref:`Elaboration_Order_Handling_in_GNAT`). To allow this, you must export an initialization routine (``Initialize_API`` in the previous example), which must be invoked before using any of the DLL services. This elaboration routine must call the Ada elaboration routine ``adainit`` generated by the GNAT binder (:ref:`Binding_with_Non-Ada_Main_Programs`). See the body of ``Initialize_Api`` for an example. Note that the GNAT binder is automatically invoked during the DLL build process by the ``gnatdll`` tool (:ref:`Using_gnatdll`). When a DLL is loaded, Windows systematically invokes a routine called ``DllMain``. It should therefore be possible to call ``adainit`` directly from ``DllMain`` without having to provide an explicit initialization routine. Unfortunately, you can't call ``adainit`` from the ``DllMain`` if your program has library level tasks because access to the ``DllMain`` entry point is serialized by the system (that is, only a single thread can execute 'through' it at a time), which means that the GNAT run-time will deadlock waiting for a newly created task to complete its initialization. .. _Ada_DLLs_and_Finalization: Ada DLLs and Finalization ^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: DLLs and finalization When the services of an Ada DLL are no longer needed, the client code should invoke the DLL finalization routine, if available. The DLL finalization routine is in charge of releasing all resources acquired by the DLL. In the case of the Ada code contained in the DLL, this is achieved by calling routine ``adafinal`` generated by the GNAT binder (:ref:`Binding_with_Non-Ada_Main_Programs`). See the body of ``Finalize_Api`` for an example. As already pointed out the GNAT binder is automatically invoked during the DLL build process by the ``gnatdll`` tool (:ref:`Using_gnatdll`). .. _Creating_a_Spec_for_Ada_DLLs: Creating a Spec for Ada DLLs ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ To use the services exported by the Ada DLL from another programming language (e.g., C), you have to translate the specs of the exported Ada entities in that language. For instance in the case of ``API.dll``, the corresponding C header file could look like: .. code-block:: c extern int *_imp__count; #define count (*_imp__count) int factorial (int); It is important to understand that when building an Ada DLL to be used by other Ada applications, you need two different specs for the packages contained in the DLL: one for building the DLL and the other for using the DLL. This is because the ``DLL`` calling convention is needed to use a variable defined in a DLL, but when building the DLL, the variable must have either the ``Ada`` or ``C`` calling convention. As an example consider a DLL consisting of the following package ``API``: .. code-block:: ada package API is Count : Integer := 0; ... -- Remainder of the package omitted. end API; After producing a DLL containing package ``API``, the spec that must be used to import ``API.Count`` from Ada code outside of the DLL is: .. code-block:: ada package API is Count : Integer; pragma Import (DLL, Count); end API; .. _Creating_the_Definition_File: Creating the Definition File """""""""""""""""""""""""""" The definition file is the last file you need to build the DLL. It lists the exported symbols. As an example, the definition file for a DLL containing only package ``API`` above (where all the entities are exported with a ``C`` calling convention) is: :: EXPORTS count factorial finalize_api initialize_api If the ``C`` calling convention is missing from package ``API``, the definition file contains the mangled Ada names of the above entities, which in this case are: :: EXPORTS api__count api__factorial api__finalize_api api__initialize_api .. _Using_gnatdll: Using ``gnatdll`` """"""""""""""""" .. index:: gnatdll ``gnatdll`` is a tool to automate the DLL build process once all the Ada and non-Ada sources that make up your DLL have been compiled. ``gnatdll`` is actually in charge of two distinct tasks: building both the static import library for the DLL and the actual DLL. You invoke the ``gnatdll`` command as :: $ gnatdll [ switches ] list-of-files [ -largs opts ] where ``list-of-files`` is a list of ALI and object files. The object file list must be the exact list of objects corresponding to the non-Ada sources whose services are to be included in the DLL. The ALI file list must be the exact list of ALI files for the corresponding Ada sources whose services are to be included in the DLL. If ``list-of-files`` is missing, only the static import library is generated. You may specify any of the following switches to ``gnatdll``: .. index:: -a (gnatdll) :switch:`-a[{address}]` Build a non-relocatable DLL at ``address``. If you don't specify ``address``, ``gnatdll`` uses the default address of ``0x11000000``. By default, when this switch is missing, ``gnatdll`` builds a relocatable DLL. We advise you to build relocatable DLL. .. index:: -bargs (gnatdll) :switch:`-bargs {opts}` Binder switches. Pass ``opts`` to the binder. .. index:: -d (gnatdll) :switch:`-d {dllfile}` ``dllfile`` is the name of the DLL. You must specify this switch for ``gnatdll`` to do anything. ``gnatdll`` names the generated import library algorithmically from ``dllfile`` as shown in the following example: if ``dllfile`` is :file:`xyz.dll`, the import library name is :file:`libxyz.dll.a`. ``gnatdll`` obtains the name of the definition file (if not specified by switch :switch:`-e`) algorithmically from ``dllfile`` as shown in the following example: if ``dllfile`` is :file:`xyz.dll`, the definition file used is :file:`xyz.def`. .. index:: -e (gnatdll) :switch:`-e {deffile}` ``deffile`` is the name of the definition file. .. index:: -g (gnatdll) :switch:`-g` Generate debugging information. This information is stored in the object file and copied from there to the final DLL file by the linker, where it can be read by the debugger. You must use the :switch:`-g` switch if you plan on using the debugger or the symbolic stack traceback. .. index:: -h (gnatdll) :switch:`-h` Help mode. Displays ``gnatdll`` switch usage information. .. index:: -I (gnatdll) :switch:`-I{dir}` Direct ``gnatdll`` to search the ``dir`` directory for source and object files needed to build the DLL. (:ref:`Search_Paths_and_the_Run-Time_Library_RTL`). .. index:: -k (gnatdll) :switch:`-k` Removes the :samp:`@{nn}` suffix from the import library's exported names, but keeps them for the link names. You must specify this switch if you want to use a ``Stdcall`` function in a DLL for which the :samp:`@{nn}` suffix has been removed. This is the case for most of the Windows NT DLL for example. This switch has no effect if you specify the :switch:`-n` switch. .. index:: -l (gnatdll) :switch:`-l {file}` The list of ALI and object files used to build the DLL are listed in ``file``, instead of being given in the command line. Each line in ``file`` contains the name of an ALI or object file. .. index:: -m (gnatdll) :switch:`-m` Generate map file .. index:: -n (gnatdll) :switch:`-n` No Import. Do not create the import library. .. index:: -q (gnatdll) :switch:`-q` Quiet mode. Do not display unnecessary messages. .. index:: -v (gnatdll) :switch:`-v` Verbose mode. Display extra information. .. index:: -largs (gnatdll) :switch:`-largs {opts}` Linker switches. Pass ``opts`` to the linker. .. rubric:: ``gnatdll`` Example As an example, the command to build a relocatable DLL from :file:`api.adb` once :file:`api.adb` has been compiled and :file:`api.def` created is :: $ gnatdll -d api.dll api.ali The above command creates two files: :file:`libapi.dll.a` (the import library) and :file:`api.dll` (the actual DLL). If you want to create only the DLL, just type: :: $ gnatdll -d api.dll -n api.ali Alternatively, if you want to create just the import library, type: :: $ gnatdll -d api.dll .. rubric:: ``gnatdll`` behind the Scenes This section details the steps involved in creating a DLL. ``gnatdll`` does these steps for you. Unless you are interested in understanding what goes on behind the scenes, you should skip this section. We use the previous example of a DLL containing the Ada package ``API``, to illustrate the steps necessary to build a DLL. The starting point is a set of objects that make up the DLL and the corresponding ALI files. In the case of this example, this means :file:`api.o` and :file:`api.ali`. To build a relocatable DLL, ``gnatdll`` does the following: * builds the base file (:file:`api.base`). A base file gives the information necessary to generate relocation information for the DLL. :: $ gnatbind -n api $ gnatlink api -o api.jnk -mdll -Wl,--base-file,api.base In addition to the base file, the ``gnatlink`` command generates an output file :file:`api.jnk`, which can be discarded. The :switch:`-mdll` switch asks ``gnatlink`` to generate the routines ``DllMain`` and ``DllMainCRTStartup`` that are called by the Windows loader when the DLL is loaded into memory. * uses ``dlltool`` (see :ref:`Using dlltool `) to build the export table (:file:`api.exp`). The export table contains the relocation information in a form which can be used during the final link to ensure that the Windows loader is able to place the DLL anywhere in memory. :: $ dlltool --dllname api.dll --def api.def --base-file api.base \\ --output-exp api.exp * builds the base file using the new export table. Note that ``gnatbind`` must be called once again since the binder generated file has been deleted during the previous call to ``gnatlink``. :: $ gnatbind -n api $ gnatlink api -o api.jnk api.exp -mdll -Wl,--base-file,api.base * builds the new export table using the new base file and generates the DLL import library :file:`libAPI.dll.a`. :: $ dlltool --dllname api.dll --def api.def --base-file api.base \\ --output-exp api.exp --output-lib libAPI.a * Finally, builds the relocatable DLL using the final export table. :: $ gnatbind -n api $ gnatlink api api.exp -o api.dll -mdll .. _Using_dlltool: .. rubric:: Using ``dlltool`` ``dlltool`` is the low-level tool used by ``gnatdll`` to build DLLs and static import libraries. This section summarizes the most common ``dlltool`` switches. You run ``dlltool`` as follows: :: $ dlltool [`switches`] ``dlltool`` switches include: .. index:: --base-file (dlltool) :switch:`--base-file {basefile}` Read the base file ``basefile`` generated by the linker. You use this switch to create a relocatable DLL. .. index:: --def (dlltool) :switch:`--def {deffile}` Read the definition file. .. index:: --dllname (dlltool) :switch:`--dllname {name}` Gives the name of the DLL. You use this switch to embed the name of the DLL in the static import library generated by ``dlltool`` with switch :switch:`--output-lib`. .. index:: -k (dlltool) :switch:`-k` Kill :samp:`@{nn}` from exported names (:ref:`Windows_Calling_Conventions` for a discussion about ``Stdcall``-style symbols). .. index:: --help (dlltool) :switch:`--help` Prints the ``dlltool`` switches with a concise description. .. index:: --output-exp (dlltool) :switch:`--output-exp {exportfile}` Generate an export file ``exportfile``. The export file contains the export table (list of symbols in the DLL) and is used to create the DLL. .. index:: --output-lib (dlltool) :switch:`--output-lib {libfile}` Generate a static import library ``libfile``. .. index:: -v (dlltool) :switch:`-v` Verbose mode. .. index:: --as (dlltool) :switch:`--as {assembler-name}` Use ``assembler-name`` as the assembler. The default is ``as``. .. _GNAT_and_Windows_Resources: GNAT and Windows Resources ^^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: Resources, windows Resources are an easy way to add Windows-specific objects to your application. The objects that you can add as resources include: * menus * accelerators * dialog boxes * string tables * bitmaps * cursors * icons * fonts * version information You can use a version information resource to embed information into an executable or a DLL. This information can be viewed using the file properties from the Windows Explorer. Here's an example of a version information resource: :: 1 VERSIONINFO FILEVERSION 1,0,0,0 PRODUCTVERSION 1,0,0,0 BEGIN BLOCK "StringFileInfo" BEGIN BLOCK "080904E4" BEGIN VALUE "CompanyName", "My Company Name" VALUE "FileDescription", "My application" VALUE "FileVersion", "1.0" VALUE "InternalName", "my_app" VALUE "LegalCopyright", "My Name" VALUE "OriginalFilename", "my_app.exe" VALUE "ProductName", "My App" VALUE "ProductVersion", "1.0" END END BLOCK "VarFileInfo" BEGIN VALUE "Translation", 0x809, 1252 END END The value ``0809`` (langID) is for the U.K English language and ``04E4`` (charsetID), which is equal to ``1252`` decimal, for multilingual. This section explains how to build, compile and use resources. Note that this section does not cover all resource objects; for a complete description see the corresponding Microsoft documentation. .. _Building_Resources: Building Resources """""""""""""""""" .. index:: Resources, building A resource file is an ASCII file. By convention, resource files have an :file:`.rc` extension. The easiest way to build a resource file is to use Microsoft tools such as ``imagedit.exe`` to build bitmaps, icons and cursors and ``dlgedit.exe`` to build dialogs. You can always build an :file:`.rc` file yourself by writing a resource script. It's not our objective to explain how to write a resource file. A complete description of the resource script language can be found in the Microsoft documentation. .. _Compiling_Resources: Compiling Resources """"""""""""""""""" .. index:: rc .. index:: windres .. index:: Resources, compiling This section describes how you can build a GNAT-compatible (COFF) object file containing the resources. You do this using the Resource Compiler ``windres`` as follows: :: $ windres -i myres.rc -o myres.o By default ``windres`` runs ``gcc`` to preprocess the :file:`.rc` file. You can specify an alternate preprocessor (usually named :file:`cpp.exe`) using the ``windres`` :switch:`--preprocessor` parameter. You can obtain a list of all possible switches by entering the command ``windres`` :switch:`--help`. You can also use the Microsoft resource compiler ``rc.exe`` to produce a :file:`.res` file (binary resource file). See the corresponding Microsoft documentation for further details. In this case, you need to use ``windres`` to translate the :file:`.res` file to a GNAT-compatible object file as follows: :: $ windres -i myres.res -o myres.o .. _Using_Resources: Using Resources """"""""""""""" .. index:: Resources, using To include the resource file in your program just add the GNAT-compatible object file for the resource(s) to the linker arguments. With ``gnatmake`` you do this using the :switch:`-largs` switch: :: $ gnatmake myprog -largs myres.o .. _Using_GNAT_DLL_from_MSVS: Using GNAT DLLs from Microsoft Visual Studio Applications ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: Microsoft Visual Studio, use with GNAT DLLs This section describes a common case of mixed GNAT/Microsoft Visual Studio application development, where the main program is developed using MSVS and is linked with a DLL developed using GNAT. You should develop such a mixed application following the general guidelines outlined above; below is the cookbook-style sequence of steps to follow: 1. First develop and build the GNAT shared library using a library project (let's assume the project is :file:`mylib.gpr`, producing the library :file:`libmylib.dll`): :: $ gprbuild -p mylib.gpr 2. Produce a :file:`.def` file for the symbols you need to interface with, either by hand or automatically with possibly some manual adjustments (see :ref:`Creating Definition File Automatically `): :: $ dlltool libmylib.dll -z libmylib.def --export-all-symbols 3. Make sure that MSVS command-line tools are accessible on the path. 4. Create the Microsoft-style import library (see :ref:`MSVS-Style Import Library `): :: $ lib -machine:IX86 -def:libmylib.def -out:libmylib.lib If you are using a 64-bit toolchain, the above becomes... :: $ lib -machine:X64 -def:libmylib.def -out:libmylib.lib 5. Build the C ``main``: :: $ cl /O2 /MD main.c libmylib.lib 6. Before running the executable, make sure you have set the PATH to include the DLL or copy the DLL into into the directory containing the :file:`.exe`. .. _Debugging_a_DLL: Debugging a DLL ^^^^^^^^^^^^^^^ .. index:: DLL debugging Debugging a DLL is similar to debugging a standard program, but you have to deal with two different executable parts: the DLL and the program that uses it. There are the following four possibilities: * The program and DLL are built with GCC/GNAT. * The program is built with foreign tools and the DLL is built with GCC/GNAT. * The program is built with GCC/GNAT and the DLL is built with foreign tools. In this section we address only cases one and two above. Note that there is no point in trying to debug a DLL with GNU/GDB if there is no GDB-compatible debugging information in it. To do so, you must use a debugger compatible with the tools suite used to build the DLL. .. _Program_and_DLL_Both_Built_with_GCC/GNAT: Program and DLL Both Built with GCC/GNAT """""""""""""""""""""""""""""""""""""""" This is the simplest case. Both the DLL and the program have ``GDB`` compatible debugging information. You can then break anywhere in the process. Let's suppose the main procedure is named ``ada_main`` and in the DLL there's an entry point named ``ada_dll``. The DLL (:ref:`Introduction_to_Dynamic_Link_Libraries_DLLs`) and program must have been built with the debugging information (see the GNAT :switch:`-g` switch). Here are the step-by-step instructions for debugging it: * Launch ``GDB`` on the main program. :: $ gdb -nw ada_main * Start the program and stop at the beginning of the main procedure :: (gdb) start This step is required to be able to set a breakpoint inside the DLL. Until the program is run, the DLL is not loaded. This has the consequence that the DLL debugging information is also not loaded, so it is not possible to set a breakpoint in the DLL. * Set a breakpoint inside the DLL :: (gdb) break ada_dll (gdb) cont At this stage, a breakpoint is set inside the DLL. From there on you can use standard ``GDB`` commands to debug the whole program (:ref:`Running_and_Debugging_Ada_Programs`). .. _Program_Built_with_Foreign_Tools_and_DLL_Built_with_GCC/GNAT: Program Built with Foreign Tools and DLL Built with GCC/GNAT """""""""""""""""""""""""""""""""""""""""""""""""""""""""""" In this case, things are slightly more complex because you can't start the main program and then break at the beginning to load the DLL and the associated DLL debugging information. It's not possible to break at the beginning of the program because there's no ``GDB`` debugging information, and therefore there's no direct way of getting initial control. This section addresses this issue by describing some methods that you can use to break somewhere in the DLL to debug it. First, suppose that the main procedure is named ``main`` (this is the case, for example, for some C code built with Microsoft Visual C) and that there's a DLL named ``test.dll`` containing an Ada entry point named ``ada_dll``. The DLL (see :ref:`Introduction_to_Dynamic_Link_Libraries_DLLs`) must have been built with debugging information (see the GNAT :switch:`-g` switch). .. rubric:: Debugging the DLL Directly * Determine the executable's starting address :: $ objdump --file-header main.exe The starting address is reported on the last line. For example: :: main.exe: file format pei-i386 architecture: i386, flags 0x0000010a: EXEC_P, HAS_DEBUG, D_PAGED start address 0x00401010 * Launch the debugger on the executable. :: $ gdb main.exe * Set a breakpoint at the starting address and launch the program. :: $ (gdb) break *0x00401010 $ (gdb) run The program will stop at the specified address. * Set a breakpoint on a DLL subroutine. :: (gdb) break ada_dll.adb:45 Or if you want to break using a symbol on the DLL, you need first to select the Ada language (language used by the DLL). :: (gdb) set language ada (gdb) break ada_dll * Continue the program. :: (gdb) cont This runs the program until it reaches the breakpoint that you've set. From that point, you can use standard ``GDB`` commands to debug a program as described in (:ref:`Running_and_Debugging_Ada_Programs`). You can also debug the DLL by attaching ``GDB`` to a running process. .. rubric:: Attaching to a Running Process .. index:: DLL debugging, attach to process With ``GDB``, you can always debug a running process by attaching to it. It's possible to debug a DLL this way. The limitation of this approach is that the DLL must run long enough to perform the attach operation. To ensure this, you may want, for example, to insert a time-wasting loop in the code of the DLL to allow this to happen. * Launch the main program :file:`main.exe`. :: $ main * Use the Windows *Task Manager* to find the process ID. Let's say that the process PID for :file:`main.exe` is 208. * Launch gdb. :: $ gdb * Attach to the running process to be debugged. :: (gdb) attach 208 * Load the process debugging information. :: (gdb) symbol-file main.exe * Break somewhere in the DLL. :: (gdb) break ada_dll * Continue process execution. :: (gdb) cont This last step will resume the process execution and stop at the breakpoint we have set. From there you can use standard ``GDB`` commands to debug a program, as described in :ref:`Running_and_Debugging_Ada_Programs`. .. _Setting_Stack_Size_from_gnatlink: Setting Stack Size from ``gnatlink`` ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ You can specify the program stack size at link time. On most versions of Windows, starting with XP, this is mostly useful to set the size of the main stack (environment task). The other task stacks are set with pragma Storage_Size or with the *gnatbind -d* command. The specified size will become the reserved memory size of the underlying thread. Since very old versions of Windows (2000, NT4, etc.) don't allow setting the reserve size of individual tasks, for those versions the link-time stack size applies to all tasks, and pragma Storage_Size has no effect. In particular, Stack Overflow checks are made against this link-time specified size. You can set this with ``gnatlink`` using either of the following: * :switch:`-Xlinker` linker switch :: $ gnatlink hello -Xlinker --stack=0x10000,0x1000 This sets the stack reserve size to 0x10000 bytes and the stack commit size to 0x1000 bytes. * :switch:`-Wl` linker switch :: $ gnatlink hello -Wl,--stack=0x1000000 This sets the stack reserve size to 0x1000000 bytes. Note that with :switch:`-Wl` switch, you can't also set the stack commit size because the comma is a separator for this switch. .. _Setting_Heap_Size_from_gnatlink: Setting Heap Size from ``gnatlink`` ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Under Windows systems, it is possible to specify the program heap size from ``gnatlink`` using either of the following: * :switch:`-Xlinker` linker switch :: $ gnatlink hello -Xlinker --heap=0x10000,0x1000 This sets the heap reserve size to 0x10000 bytes and the heap commit size to 0x1000 bytes. * :switch:`-Wl` linker switch :: $ gnatlink hello -Wl,--heap=0x1000000 This sets the heap reserve size to 0x1000000 bytes. Note that with :switch:`-Wl` switch, you can't also set the heap commit size because the comma is a separator for this switch. .. _Win32_Specific_Addons: Windows Specific Add-Ons ------------------------- This section describes the Windows specific add-ons. .. _Win32Ada: Win32Ada ^^^^^^^^ ``Win32Ada`` is a binding for the Microsoft Win32 API, which you can easily install using the provided installer. To use it, you need to use a project file and add a single with_clause to give you full access to the ``Win32Ada`` binding sources and ensure that the proper libraries are passed to the linker. .. code-block:: gpr with "win32ada"; project P is for Sources use ...; end P; To build the application, you just need to call ``gprbuild`` for the application's project, here :file:`p.gpr`: .. code-block:: sh gprbuild p.gpr .. _wPOSIX: wPOSIX ^^^^^^ ``wPOSIX`` is a minimal POSIX binding whose goal is to help with building cross-platforms applications. This binding is not complete though, as the Win32 API does not provide the necessary support for all POSIX APIs. To use the ``wPOSIX`` binding, you need to use a project file and add a single *with* clause to give you full access to the ``wPOSIX`` binding sources and ensure that the proper libraries are passed to the linker. .. code-block:: gpr with "wposix"; project P is for Sources use ...; end P; To build the application, you just need to call ``gprbuild`` for the application's project, here :file:`p.gpr`: .. code-block:: sh gprbuild p.gpr .. _Mac_OS_Topics: Mac OS Topics ============= .. index:: OS X This section describes topics that are specific to Apple's OS X platform. Codesigning the Debugger ------------------------ The Darwin Kernel, used by Apple's OS X, requires the debugger to have special permissions before it's allowed to control other processes. These permissions are granted by codesigning the GDB executable. Without these permissions, the debugger will report error messages such as:: Starting program: /x/y/foo Unable to find Mach task port for process-id 28885: (os/kern) failure (0x5). (please check gdb is codesigned - see taskgated(8)) Codesigning requires a certificate. The following procedure explains how to create one: * Start the Keychain Access application (in /Applications/Utilities/Keychain Access.app) * Select the Keychain Access -> Certificate Assistant -> Create a Certificate... menu * Then: * Choose a name for the new certificate (this procedure will use "gdb-cert" as an example) * Set "Identity Type" to "Self Signed Root" * Set "Certificate Type" to "Code Signing" * Activate the "Let me override defaults" option * Click several times on "Continue" until the "Specify a Location For The Certificate" screen appears, then set "Keychain" to "System" * Click on "Continue" until the certificate is created * Finally, in the view, double-click on the new certificate, and set "When using this certificate" to "Always Trust" * Exit the Keychain Access application and restart the computer (this is unfortunately required) Once you've created a certificate as above, you can codesign the debugger by running the following command in a Terminal: :: $ codesign -f -s "gdb-cert" /bin/gdb with ``gdb-cert`` replaced by the actual certificate name chosen above, and ``gnat_install_prefix`` replaced by the location where you installed GNAT. Also, be sure that users of ``GDB`` are in the Unix group ``_developer``. --- ### Gcc Ada Doc Gnat Ugn The Gnat Compilation Model (`gcc/ada/doc/gnat_ugn/the_gnat_compilation_model.rst`) .. role:: switch(samp) .. |with| replace:: *with* .. |withs| replace:: *with*\ s .. |withed| replace:: *with*\ ed .. |withing| replace:: *with*\ ing .. -- Example: A |withing| unit has a |with| clause, it |withs| a |withed| unit .. _The_GNAT_Compilation_Model: ************************** The GNAT Compilation Model ************************** .. index:: ! GNAT compilation model .. index:: Compilation model This chapter describes the compilation model used by GNAT. Although similar to that used by other languages such as C and C++, this model is substantially different from the traditional Ada compilation models, which are based on a centralized program library. The chapter covers the following material: * Topics related to source file makeup and naming * :ref:`Source_Representation` * :ref:`Foreign_Language_Representation` * :ref:`File_Naming_Topics_and_Utilities` * :ref:`Configuration_Pragmas` * :ref:`Generating_Object_Files` * :ref:`Source_Dependencies` * :ref:`The_Ada_Library_Information_Files` * :ref:`Binding_an_Ada_Program` * :ref:`GNAT_and_Libraries` * :ref:`Conditional_Compilation` * :ref:`Mixed_Language_Programming` * :ref:`GNAT_and_Other_Compilation_Models` * :ref:`Using_GNAT_Files_with_External_Tools` .. _Source_Representation: Source Representation ===================== .. index:: Latin-1 .. index:: VT, HT, CR, LF, FF Ada source programs are represented in standard text files, using Latin-1 coding. Latin-1 is an 8-bit code that includes the familiar 7-bit ASCII set plus additional characters used for representing foreign languages (see :ref:`Foreign_Language_Representation` for support of non-USA character sets). The format effector characters are represented using their standard ASCII encodings, as follows: =========== ======================= =========== Character Effect Code ----------- ----------------------- ----------- :kbd:`VT` Vertical tab ``16#0B#`` :kbd:`HT` Horizontal tab ``16#09#`` :kbd:`CR` Carriage return ``16#0D#`` :kbd:`LF` Line feed ``16#0A#`` :kbd:`FF` Form feed ``16#0C#`` =========== ======================= =========== Source files are in standard text file format. In addition, GNAT recognizes a wide variety of stream formats, in which the end of physical lines is marked by any of the following sequences: ``LF``, ``CR``, ``CR-LF``, or ``LF-CR``. This is useful in accommodating files imported from other operating systems. .. index:: pair: End of source file; Source file, end .. index:: SUB (control character) The end of a source file is normally represented by the physical end of file. However, the control character ``16#1A#`` (:kbd:`SUB`) is also recognized as signalling the end of the source file. Again, this is provided for compatibility with other, legacy, operating systems where this code is used to represent the end of file. .. index:: spec (definition), compilation (definition) Each file contains a single Ada compilation unit, including any pragmas associated with the unit. For example, this means you must place a package declaration (a package *spec*) and the corresponding body in separate files. An Ada *compilation* (which is a sequence of compilation units) is represented using a sequence of files. Similarly, you place each subunit or child unit in a separate file. .. _Foreign_Language_Representation: Foreign Language Representation =============================== GNAT supports the standard character sets defined in Ada as well as several other non-standard character sets for use in localized versions of the compiler (:ref:`Character_Set_Control`). .. _Latin-1: Latin-1 ------- .. index:: Latin-1 The basic character set is Latin-1. This character set is defined by ISO standard 8859, part 1. The lower half (character codes ``16#00#`` ... ``16#7F#)`` is identical to standard ASCII coding but the upper half is used to represent additional characters. These include extended letters used by European languages, such as French accents, the vowels with umlauts used in German, and the extra letter A-ring used in Swedish. .. index:: Ada.Characters.Latin_1 For a complete list of Latin-1 codes and their encodings, see the source file of library unit ``Ada.Characters.Latin_1`` in file :file:`a-chlat1.ads`. You may use any of these extended characters freely in character or string literals. In addition, the extended characters that represent letters can be used in identifiers. .. _Other_8-Bit_Codes: Other 8-Bit Codes ----------------- GNAT also supports several other 8-bit coding schemes: .. index:: Latin-2 .. index:: ISO 8859-2 *ISO 8859-2 (Latin-2)* Latin-2 letters allowed in identifiers, with uppercase and lowercase equivalence. .. index:: Latin-3 .. index:: ISO 8859-3 *ISO 8859-3 (Latin-3)* Latin-3 letters allowed in identifiers, with uppercase and lowercase equivalence. .. index:: Latin-4 .. index:: ISO 8859-4 *ISO 8859-4 (Latin-4)* Latin-4 letters allowed in identifiers, with uppercase and lowercase equivalence. .. index:: ISO 8859-5 .. index:: Cyrillic *ISO 8859-5 (Cyrillic)* ISO 8859-5 letters (Cyrillic) allowed in identifiers, with uppercase and lowercase equivalence. .. index:: ISO 8859-15 .. index:: Latin-9 *ISO 8859-15 (Latin-9)* ISO 8859-15 (Latin-9) letters allowed in identifiers, with uppercase and lowercase equivalence. .. index:: code page 437 (IBM PC) *IBM PC (code page 437)* This code page is the normal default for PCs in the US. It corresponds to the original IBM PC character set. This set has some, but not all, of the extended Latin-1 letters, but these letters do not have the same encoding as Latin-1. In this mode, these letters are allowed in identifiers with uppercase and lowercase equivalence. .. index:: code page 850 (IBM PC) *IBM PC (code page 850)* This code page is a modification of 437 extended to include all the Latin-1 letters, but still not with the usual Latin-1 encoding. In this mode, all these letters are allowed in identifiers with uppercase and lowercase equivalence. *Full Upper 8-bit* Any character in the range 80-FF is allowed in identifiers and all are considered distinct. In other words, there are no uppercase and lowercase equivalences in this range. This is useful in conjunction with certain encoding schemes used for some foreign character sets (e.g., the typical method of representing Chinese characters on the PC). *No Upper-Half* No upper-half characters in the range 80-FF are allowed in identifiers. This gives Ada 83 compatibility for identifier names. For precise data on the encodings permitted, and the uppercase and lowercase equivalences that are recognized, see the file :file:`csets.adb` in the GNAT compiler sources. You will need to obtain a full source release of GNAT to obtain this file. .. _Wide_Character_Encodings: Wide_Character Encodings ------------------------ GNAT allows wide character codes to appear in character and string literals, and also optionally in identifiers, by means of the following possible encoding schemes: *Hex Coding* In this encoding, a wide character is represented by the following five character sequence:: ESC a b c d where ``a``, ``b``, ``c``, ``d`` are the four hexadecimal characters (using uppercase letters) of the wide character code. For example, ESC A345 is used to represent the wide character with code ``16#A345#``. This scheme is compatible with use of the full Wide_Character set. .. index:: Upper-Half Coding *Upper-Half Coding* The wide character with encoding ``16#abcd#`` where the upper bit is on (in other words, 'a' is in the range 8-F) is represented as two bytes, ``16#ab#`` and ``16#cd#``. The second byte cannot be a format control character, but is not required to be in the upper half. This method can be also used for shift-JIS or EUC, where the internal coding matches the external coding. .. index:: Shift JIS Coding *Shift JIS Coding* A wide character is represented by a two-character sequence, ``16#ab#`` and ``16#cd#``, with the restrictions described for upper-half encoding as described above. The internal character code is the corresponding JIS character according to the standard algorithm for Shift-JIS conversion. You can only use characters defined in the JIS code set table with this encoding method. .. index:: EUC Coding *EUC Coding* A wide character is represented by a two-character sequence ``16#ab#`` and ``16#cd#``, with both characters being in the upper half. The internal character code is the corresponding JIS character according to the EUC encoding algorithm. You can only use characters defined in the JIS code set table with this encoding method. *UTF-8 Coding* A wide character is represented using UCS Transformation Format 8 (UTF-8) as defined in Annex R of ISO 10646-1/Am.2. Depending on the character value, the representation is a one, two, or three byte sequence:: 16#0000#-16#007f#: 2#0xxxxxxx# 16#0080#-16#07ff#: 2#110xxxxx# 2#10xxxxxx# 16#0800#-16#ffff#: 2#1110xxxx# 2#10xxxxxx# 2#10xxxxxx# where the ``xxx`` bits correspond to the left-padded bits of the 16-bit character value. Note that all lower half ASCII characters are represented as ASCII bytes and all upper half characters and other wide characters are represented as sequences of upper-half (The full UTF-8 scheme allows for encoding 31-bit characters as 6-byte sequences the use of these sequences is documented in the following section on wide wide characters.) *Brackets Coding* In this encoding, a wide character is represented by the following eight character sequence:: [ " a b c d " ] where ``a``, ``b``, ``c``, ``d`` are the four hexadecimal characters (using uppercase letters) of the wide character code. For example, ['A345'] is used to represent the wide character with code ``16#A345#``. You can also (though you are not required to) use the Brackets coding for upper half characters. For example, you can represent the code ``16#A3#`` as ``['A3']``. This scheme is compatible with use of the full ``Wide_Character`` set, and is also the method used for wide character encoding in some standard ACATS (Ada Conformity Assessment Test Suite) test suite distributions. .. note:: Some of these coding schemes do not permit the full use of the Ada character set. For example, neither Shift JIS nor EUC allow the use of the upper half of the Latin-1 set. .. _Wide_Wide_Character_Encodings: Wide_Wide_Character Encodings ----------------------------- GNAT allows wide wide character codes to appear in character and string literals, and also optionally in identifiers, by means of the following possible encoding schemes: *UTF-8 Coding* A wide character is represented using UCS Transformation Format 8 (UTF-8) as defined in Annex R of ISO 10646-1/Am.2. Depending on the character value, the representation of character codes with values greater than 16#FFFF# is a is a four, five, or six byte sequence:: 16#01_0000#-16#10_FFFF#: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx 16#0020_0000#-16#03FF_FFFF#: 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 16#0400_0000#-16#7FFF_FFFF#: 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx where the ``xxx`` bits correspond to the left-padded bits of the 32-bit character value. *Brackets Coding* In this encoding, a wide wide character is represented by the following ten or twelve byte character sequence:: [ " a b c d e f " ] [ " a b c d e f g h " ] where ``a-h`` are the six or eight hexadecimal characters (using uppercase letters) of the wide wide character code. For example, ["1F4567"] is used to represent the wide wide character with code ``16#001F_4567#``. This scheme is compatible with use of the full ``Wide_Wide_Character`` set, and is also the method used for wide wide character encoding in some standard ACATS (Ada Conformity Assessment Test Suite) test suite distributions. .. _File_Naming_Topics_and_Utilities: File Naming Topics and Utilities ================================ GNAT has a default file naming scheme, but and also provides you with a high degree of control over how the names and extensions of your source files correspond to the Ada compilation units that they contain. .. _File_Naming_Rules: File Naming Rules ----------------- GNAT determines the default file name by the name of the unit that the file contains. The name is formed by taking the full expanded name of the unit, replacing the separating dots with hyphens, and using lowercase for all letters. An exception occurs if the file name generated by the above rules starts with one of the characters ``a``, ``g``, ``i``, or ``s`` and the second character is a hyphen. In this case, the character tilde is used in place of the hyphen. This special rule avoids clashes with the standard names for child units of the packages ``System``, ``Ada``, ``Interfaces``, and ``GNAT``, which use the prefixes ``s-``, ``a-``, ``i-``, and ``g-``, respectively. The file extension is :file:`.ads` for a spec and :file:`.adb` for a body. The following table shows some examples of these rules. ============================ =============================== Source File Ada Compilation Unit ---------------------------- ------------------------------- :file:`main.ads` Main (spec) :file:`main.adb` Main (body) :file:`arith_functions.ads` Arith_Functions (package spec) :file:`arith_functions.adb` Arith_Functions (package body) :file:`func-spec.ads` Func.Spec (child package spec) :file:`func-spec.adb` Func.Spec (child package body) :file:`main-sub.adb` Sub (subunit of Main) :file:`a~bad.adb` A.Bad (child package body) ============================ =============================== Following these rules can result in excessively long file names if corresponding unit names are long (for example, if child units or subunits are heavily nested). An option is available to shorten such long file names (called file name 'krunching'). You may find this particularly useful when programs being developed with GNAT are to be used on operating systems with limited file name lengths. :ref:`Using_gnatkr`. Of course, no file shortening algorithm can guarantee uniqueness over all possible unit names; if file name krunching is used, it is your responsibility to ensure no name clashes occur. Alternatively, you can specify the exact file names that you want used, as described in the next section. Finally, if your Ada programs are migrating from a compiler with a different naming convention, you can use the ``gnatchop`` utility to produce source files that follow the GNAT naming conventions. (For details see :ref:`Renaming_Files_with_gnatchop`.) Note: in the case of Windows or Mac OS operating systems, case is not significant. So, for example, on Windows if the canonical name is :file:`main-sub.adb`, you can use the file name :file:`Main-Sub.adb` instead. However, case is significant for other operating systems, so, for example, if you want to use other than canonically cased file names on a Unix system, you need to follow the procedures described in the next section. .. _Using_Other_File_Names: Using Other File Names ---------------------- .. index:: File names The previous section described the default rules used by GNAT to determine the file name in which a given unit resides. It is usually convenient to follow these default rules, and if you follow them, the compiler knows without being explicitly told where to find all the files it needs. .. index:: Source_File_Name pragma However, in some cases, particularly when a program is imported from another Ada compiler environment, it may be more convenient for you to specify which file names contain which units. GNAT allows arbitrary file names to be used by means of the ``Source_File_Name`` pragma. The form of this pragma is as shown in the following examples: .. code-block:: ada pragma Source_File_Name (My_Utilities.Stacks, Spec_File_Name => "myutilst_a.ada"); pragma Source_File_name (My_Utilities.Stacks, Body_File_Name => "myutilst.ada"); As shown in this example, the first argument for the pragma is the unit name (in this example a child unit). The second argument has the form of a named association. The identifier indicates whether the file name is for a spec or a body; the file name itself is given by a string literal. The source file name pragma is a configuration pragma, which means that normally you will place it in the :file:`gnat.adc` file used to hold configuration pragmas that apply to a complete compilation environment. For more details on how the :file:`gnat.adc` file is created and used see :ref:`Handling_of_Configuration_Pragmas`. .. index:: gnat.adc GNAT allows you to specify completely arbitrary file names using the source file name pragma. However, if the file name specified has an extension other than :file:`.ads` or :file:`.adb` you must use a special syntax when compiling the file. The name on the command line in this case must be preceded by the special sequence ``-x`` followed by a space and the name of the language, here ``ada``, as in: .. code-block:: sh $ gcc -c -x ada peculiar_file_name.sim ``gnatmake`` handles non-standard file names in the usual manner (the non-standard file name for the main program is simply used as the argument to ``gnatmake``). Note that if the extension is also non-standard, you must include it in the ``gnatmake`` command; it may not be omitted. .. _Alternative_File_Naming_Schemes: Alternative File Naming Schemes ------------------------------- .. index:: File naming schemes, alternative .. index:: File names The previous section described the use of the ``Source_File_Name`` pragma to allow arbitrary names to be assigned to individual source files. However, this approach requires one pragma for each file and, especially in large systems, can result in very long :file:`gnat.adc` files, which can create a maintenance problem. .. index:: Source_File_Name pragma GNAT also provides a facility for specifying systematic file naming schemes other than the standard default naming scheme previously described. An alternative scheme for naming is specified by the use of ``Source_File_Name`` pragmas having the following format: .. code-block:: ada pragma Source_File_Name ( Spec_File_Name => FILE_NAME_PATTERN [ , Casing => CASING_SPEC] [ , Dot_Replacement => STRING_LITERAL ] ); pragma Source_File_Name ( Body_File_Name => FILE_NAME_PATTERN [ , Casing => CASING_SPEC ] [ , Dot_Replacement => STRING_LITERAL ] ) ; pragma Source_File_Name ( Subunit_File_Name => FILE_NAME_PATTERN [ , Casing => CASING_SPEC ] [ , Dot_Replacement => STRING_LITERAL ] ) ; FILE_NAME_PATTERN ::= STRING_LITERAL CASING_SPEC ::= Lowercase | Uppercase | Mixedcase The ``FILE_NAME_PATTERN`` string shows how the file name is constructed. It contains a single asterisk character, and the unit name is substituted systematically for this asterisk. The optional parameter ``Casing`` indicates whether the unit name is to be all upper-case letters, all lower-case letters, or mixed-case. If no ``Casing`` parameter is used, the default is all lower-case. You use the optional ``Dot_Replacement`` string to replace any periods that occur in subunit or child unit names. If you don't specify a ``Dot_Replacement`` argument, separating dots appear unchanged in the resulting file name. The above syntax indicates that the ``Casing`` argument must appear before the ``Dot_Replacement`` argument, but you can write these arguments in any order. As indicated, you can specify different naming schemes for bodies, specs, and subunits. Quite often, the rule for subunits is the same as the rule for bodies, in which case, you need not provide a separate ``Subunit_File_Name`` rule; in this case the ``Body_File_name`` rule is used for subunits as well. You can also use the separate rule for subunits to implement the rather unusual case of a compilation environment (e.g., a single directory) which contains a subunit and a child unit with the same unit name. Although both units cannot appear in the same partition, the Ada Reference Manual allows (but does not require) the possibility of the two units coexisting in the same environment. File name translation consists of the following steps: * If there is a specific ``Source_File_Name`` pragma for the given unit, this is always used and any general pattern rules are ignored. * If there is a pattern type ``Source_File_Name`` pragma that applies to the unit, the resulting file name is used if the file exists. If more than one pattern matches, the latest one is tried first and the first attempt that results in a reference to a file that exists is used. * If no pattern type ``Source_File_Name`` pragma that applies to the unit for which the corresponding file exists, the standard GNAT default naming rules are used. As an example of the use of this mechanism, consider a commonly used scheme in which file names are all lower case, with separating periods copied unchanged to the resulting file name, specs end with :file:`.1.ada`, and bodies end with :file:`.2.ada`. GNAT will follow this scheme if the following two pragmas appear: .. code-block:: ada pragma Source_File_Name (Spec_File_Name => ".1.ada"); pragma Source_File_Name (Body_File_Name => ".2.ada"); The default GNAT scheme is equivalent to providing the following default pragmas: .. code-block:: ada pragma Source_File_Name (Spec_File_Name => ".ads", Dot_Replacement => "-"); pragma Source_File_Name (Body_File_Name => ".adb", Dot_Replacement => "-"); Our final example implements a scheme typically used with one of the legacy Ada 83 compilers, where the separator character for subunits was '__' (two underscores), specs were identified by adding :file:`_.ADA`, bodies by adding :file:`.ADA`, and subunits by adding :file:`.SEP`. All file names were upper case. Child units were not present, of course, since this was an Ada 83 compiler, but it seems reasonable to extend this scheme to use the same double underscore separator for child units. .. code-block:: ada pragma Source_File_Name (Spec_File_Name => "_.ADA", Dot_Replacement => "__", Casing = Uppercase); pragma Source_File_Name (Body_File_Name => ".ADA", Dot_Replacement => "__", Casing = Uppercase); pragma Source_File_Name (Subunit_File_Name => ".SEP", Dot_Replacement => "__", Casing = Uppercase); .. index:: ! gnatname .. _Handling_Arbitrary_File_Naming_Conventions_with_gnatname: Handling Arbitrary File Naming Conventions with ``gnatname`` ------------------------------------------------------------ .. index:: File Naming Conventions .. _Arbitrary_File_Naming_Conventions: Arbitrary File Naming Conventions ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The GNAT compiler must know the source file name of a compilation unit in order to compile it. When using the standard GNAT default file naming conventions (``.ads`` for specs, ``.adb`` for bodies), it does not need additional information. When the source file names do not follow the standard GNAT default file naming conventions, you must give the GNAT compiler additional information through a configuration pragmas file (:ref:`Configuration_Pragmas`) or a project file. When the non-standard file naming conventions are well-defined, a small number of pragmas ``Source_File_Name`` specifying a naming pattern (:ref:`Alternative_File_Naming_Schemes`) may be sufficient. However, if the file naming conventions are irregular or arbitrary, you must define a number of pragma ``Source_File_Name`` for individual compilation units. To help maintain the correspondence between compilation unit names and source file names within the compiler, GNAT provides a tool ``gnatname`` to generate the required pragmas for a set of files. .. _Running_gnatname: Running ``gnatname`` ^^^^^^^^^^^^^^^^^^^^ The usual form of the ``gnatname`` command is: .. code-block:: sh $ gnatname [ switches ] naming_pattern [ naming_patterns ] [--and [ switches ] naming_pattern [ naming_patterns ]] All of the arguments are optional. If invoked without any arguments, ``gnatname`` will display its usage. When used with at least one naming pattern, ``gnatname`` attempts to find all the compilation units in files that follow at least one of the naming patterns. To find these compilation units, ``gnatname`` uses the GNAT compiler in syntax-check-only mode on all regular files. One or several *Naming Patterns* may be given as arguments to ``gnatname``. Each Naming Pattern is enclosed between double quotes (or single quotes on Windows). A Naming Pattern is a regular expression similar to the wildcard patterns used in file names by the Unix shells or the DOS prompt. You may call ``gnatname`` with several sections of directories/patterns. Sections are separated by the switch :switch:`--and`. In each section, you must include at least one pattern. If you don't specify a directory a section, the current directory (or the project directory if :switch:`-P` is used) is used. The options other that the directory switches and the patterns apply globally even if they are in different sections. Examples of Naming Patterns are:: "*.[12].ada" "*.ad[sb]*" "body_*" "spec_*" For a more complete description of the syntax of Naming Patterns, see the second kind of regular expressions described in :file:`g-regexp.ads` (the 'Glob' regular expressions). When invoked without the switch :switch:`-P`, ``gnatname`` will create a configuration pragmas file :file:`gnat.adc` in the current working directory, with pragmas ``Source_File_Name`` for each file that contains a valid Ada unit. .. _Switches_for_gnatname: Switches for ``gnatname`` ^^^^^^^^^^^^^^^^^^^^^^^^^ Switches for ``gnatname`` must precede any specified Naming Pattern. You may specify any of the following switches to ``gnatname``: .. index:: --version (gnatname) :switch:`--version` Display Copyright and version, then exit disregarding, all other options. .. index:: --help (gnatname) :switch:`--help` If :switch:`--version` was not used, display usage, then exit, disregarding all other options. :switch:`--subdirs={dir}` Actual object, library or exec directories are subdirectories of the specified ones. :switch:`--no-backup` Do not create a backup copy of an existing project file. :switch:`--and` Start another section of directories/patterns. .. index:: -c (gnatname) :switch:`-c{filename}` Create a configuration pragmas file :file:`filename` (instead of the default :file:`gnat.adc`). There may be zero, one, or more space between :switch:`-c` and :file:`filename`. :file:`filename` may include directory information. :file:`filename` must be writable. You can specify only one switch :switch:`-c`. When a switch :switch:`-c` is specified, you may not specify switch :switch:`-P` (see below). .. index:: -d (gnatname) :switch:`-d{dir}` Look for source files in directory :file:`dir`. You may put zero, one or more spaces between :switch:`-d` and :file:`dir`. :file:`dir` may end with ``/**``, i.e., you may write it the form ``root_dir/**``. In this case, the directory ``root_dir`` and all of its subdirectories, recursively, have to be searched for sources. When you specify a :switch:`-d` switch, the current working directory will is not searched for source files unless you explicitly specify it with a :switch:`-d` or :switch:`-D` switch. You may specify several switches :switch:`-d`. If :file:`dir` is a relative path, it is relative to the directory of the configuration pragmas file specified with switch :switch:`-c`, or to the directory of the project file specified with switch :switch:`-P` or, if you don't specify either switch :switch:`-c` or switch :switch:`-P`, it's relative to the current working directory. The directory you specified with switch :switch:`-d` must exist and be readable. .. index:: -D (gnatname) :switch:`-D{filename}` Look for source files in all directories listed in text file :file:`filename`. You may place zero, one or more spaces between :switch:`-D` and :file:`filename`. :file:`filename` must be an existing, readable text file. Each nonempty line in :file:`filename` must be a directory. Specifying switch :switch:`-D` is equivalent to specifying as many switches :switch:`-d` as there are nonempty lines in :file:`file`. :switch:`-eL` Follow symbolic links when processing project files. .. index:: -f (gnatname) :switch:`-f{pattern}` Foreign patterns. Using this switch, you can add sources of languages other than Ada to the list of sources of a project file, but it's only useful if you also specify a :switch:`-P` switch. For example, .. code-block:: sh gnatname -Pprj -f"*.c" "*.ada" looks for Ada units in all files with the :file:`.ada` extension, and adds the C files with extension :file:`.c` to the list of file for project :file:`prj.gpr` . .. index:: -h (gnatname) :switch:`-h` Output usage (help) information. The output is written to :file:`stdout`. .. index:: -P (gnatname) :switch:`-P{proj}` Create or update project file :file:`proj`. You may place zero, one or more space between :switch:`-P` and :file:`proj`. :file:`proj` may include directory information. :file:`proj` must be writable. There may be only one switch :switch:`-P`. When you specify switch :switch:`-P`, you may not also include switch :switch:`-c`. On all platforms except VMS when ``gnatname`` is invoked for an existing project file :file:`.gpr``, ``gnatname`` creates a backup copy of the project file in the project directory with file name :file:`.gpr.saved_x` where ``x`` is the first non negative number that creates a unique filename. .. index:: -v (gnatname) :switch:`-v` Verbose mode. Output detailed explanation of what it's doing to :file:`stdout`. This includes name of the file written, the name of the directories searched, and, for each file in those directories whose name matches at least one of the Naming Patterns, an indication of whether the file contains a unit, and, if so, the name of the unit. .. index:: -v -v (gnatname) :switch:`-v -v` Very verbose mode. In addition to the output produced in verbose mode (a single :switch:`-v` switch), for each file in the searched directories whose name matches none of the Naming Patterns, ``gnatname`` indicates that there is no match. .. index:: -x (gnatname) :switch:`-x{pattern}` Excluded patterns. Using this switch, you can exclude some files that otherwise would match the name patterns. For example, .. code-block:: sh gnatname -x "*_nt.ada" "*.ada" looks for Ada units in all files with the :file:`.ada` extension, except those whose names end with :file:`_nt.ada`. .. _Examples_of_gnatname_Usage: Examples of ``gnatname`` Usage ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. code-block:: sh $ gnatname -c /home/me/names.adc -d sources "[a-z]*.ada*" In this example, the directory :file:`/home/me` must already exist and be writable. In addition, the directory :file:`/home/me/sources` (specified by :switch:`-d sources`) must exist and be readable. Note the optional spaces after :switch:`-c` and :switch:`-d`. .. code-block:: sh $ gnatname -P/home/me/proj -x "*_nt_body.ada" -dsources -dsources/plus -Dcommon_dirs.txt "body_*" "spec_*" Note that you may use several :switch:`-d` switches, even in conjunction with one or several :switch:`-D` switches. This example illustrates multiple Naming Patterns and one excluded pattern. .. _File_Name_Krunching_with_gnatkr: File Name Krunching with ``gnatkr`` ----------------------------------- .. index:: ! gnatkr Here we discuss the method used by the compiler to shorten the default file names chosen for Ada units so that they do not exceed the maximum length permitted. We also describe the ``gnatkr`` utility, which you can use to determine the result of applying this shortening. .. _About_gnatkr: About ``gnatkr`` ^^^^^^^^^^^^^^^^ GNAT requires that the file name must be derived from the unit name. The default rule is as follows: * Take the unit name and replace all dots by hyphens. * If such a replacement occurs in the second character position of a name, and the first character is :samp:`a`, :samp:`g`, :samp:`s`, or :samp:`i`, then replace the dot by the character :samp:`~` (tilde) instead of a hyphen. This exception avoids clashes with the standard names for children of System, Ada, Interfaces, and GNAT, which use the prefixes :samp:`s-`, :samp:`a-`, :samp:`i-`, and :samp:`g-`, respectively. The :switch:`-gnatk{nn}` switch of the compiler activates a 'krunching' circuit that limits file names to ``nn`` characters (where ``nn`` is a decimal integer). You can use the ``gnatkr`` utility to determine the krunched name for a given file when krunched to a specified maximum length. .. _Using_gnatkr: Using ``gnatkr`` ^^^^^^^^^^^^^^^^ You invoke the ``gnatkr`` command as follows: .. code-block:: sh $ gnatkr name [ length ] ``name`` is the uncrunched file name, derived from the name of the unit in the default manner described in the previous section (i.e., in particular all dots are replaced by hyphens). You may or may not include an extension (defined as a suffix of the form period followed by arbitrary characters other than period) in the filename. If you do, ``gnatkr`` will preserve it in the output. For example, when krunching :file:`hellofile.ads` to eight characters, the result will be :file:`hellofil.ads`. Note: for compatibility with previous versions of ``gnatkr``, you can use dots in the name instead of hyphens, but ``gnatkr`` always interprets the last dot as the start of an extension. So if you pass ``gnatkr`` an argument such as :file:`Hello.World.adb`, it treats it exactly as if the first period had been a hyphen, so, for example, krunching to eight characters gives the result :file:`hellworl.adb`. Note that the result is always all lower case. Other characters are folded as required. ``length`` represents the length of the krunched name. The default if you don't specify it, is 8 characters. A length of zero means unlimited, in other words don't chop except for system files where the implied crunching length is always eight characters. The output is the krunched name. The output has an extension only if the original argument was a file name with an extension. .. _Krunching_Method: Krunching Method ^^^^^^^^^^^^^^^^ The initial file name is determined by the name of the unit that the file contains. The name is formed by taking the full expanded name of the unit and replacing the separating dots with hyphens and using lowercase for all letters, except that a hyphen in the second character position is replaced by a tilde if the first character is :samp:`a`, :samp:`i`, :samp:`g`, or :samp:`s`. The extension is ``.ads`` for a spec and ``.adb`` for a body. Krunching does not affect the extension, but the file name is shortened to the specified length by following these rules: * The name is divided into segments separated by hyphens, tildes, or underscores and all hyphens, tildes, and underscores are eliminated. If this leaves the name short enough, we are done. * If the name is too long, the longest segment is located (left-most if there are two of equal length) and shortened by dropping its last character. This is repeated until the name is short enough. As an example, consider the krunching of :file:`our-strings-wide_fixed.adb` to fit the name into 8 characters, as required by some operating systems:: our-strings-wide_fixed 22 our strings wide fixed 19 our string wide fixed 18 our strin wide fixed 17 our stri wide fixed 16 our stri wide fixe 15 our str wide fixe 14 our str wid fixe 13 our str wid fix 12 ou str wid fix 11 ou st wid fix 10 ou st wi fix 9 ou st wi fi 8 Final file name: oustwifi.adb * The file names for all predefined units are always krunched to eight characters. The krunching of these predefined units uses the following special prefix replacements: ===================== ============== Prefix Replacement --------------------- -------------- :file:`ada-` :file:`a-` :file:`gnat-` :file:`g-` :file:`interfac es-` :file:`i-` :file:`system-` :file:`s-` ===================== ============== These system files have a hyphen in the second character position. That's is why normal user files replace such a character with a tilde. As an example of this special rule, consider :file:`ada-strings-wide_fixed.adb`, which gets krunched as follows:: ada-strings-wide_fixed 22 a- strings wide fixed 18 a- string wide fixed 17 a- strin wide fixed 16 a- stri wide fixed 15 a- stri wide fixe 14 a- str wide fixe 13 a- str wid fixe 12 a- str wid fix 11 a- st wid fix 10 a- st wi fix 9 a- st wi fi 8 Final file name: a-stwifi.adb Of course, no file shortening algorithm can guarantee uniqueness over all possible unit names. If file name krunching is used, it'is your responsibility to ensure that no name clashes occur. The utility program ``gnatkr`` is supplied so that you can conveniently determine the krunched name of a file. .. _Examples_of_gnatkr_Usage: Examples of ``gnatkr`` Usage ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ :: $ gnatkr very_long_unit_name.ads --> velounna.ads $ gnatkr grandparent-parent-child.ads --> grparchi.ads $ gnatkr Grandparent.Parent.Child.ads --> grparchi.ads $ gnatkr grandparent-parent-child --> grparchi $ gnatkr very_long_unit_name.ads/count=6 --> vlunna.ads $ gnatkr very_long_unit_name.ads/count=0 --> very_long_unit_name.ads .. _Renaming_Files_with_gnatchop: Renaming Files with ``gnatchop`` -------------------------------- .. index:: ! gnatchop This section discusses how to handle files with multiple units by using the ``gnatchop`` utility. You will also find this utility useful in renaming files to meet the standard GNAT default file naming conventions. .. _Handling_Files_with_Multiple_Units: Handling Files with Multiple Units ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ GNAT's fundamental compilation model requires that a file submitted to the compiler contain only one unit and there be a strict correspondence between the file name and the unit name. If you want to have your files contain multiple units, perhaps to maintain compatibility with some other Ada compilation system, you can use ``gnatname`` to generate or update your project files, which can be processed by GNAT. See :ref:`Handling_Arbitrary_File_Naming_Conventions_with_gnatname` for more details on how to use `gnatname`. Alternatively, if you want to permanently restructure a set of 'foreign' files so that they match the GNAT rules, and do the remaining development using the GNAT structure, you can simply use ``gnatchop`` once, generate the new set of files containing only one unit per file, and work with them from that point on. Note that if your file containing multiple units starts with a byte order mark (BOM) specifying UTF-8 encoding, each file generated by gnatchop will start with a copy of this BOM, meaning that they can be compiled automatically in UTF-8 mode without you needing to specify an explicit encoding. .. _Operating_gnatchop_in_Compilation_Mode: Operating gnatchop in Compilation Mode ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The basic function of ``gnatchop`` is to take a file with multiple units and split it into separate files. The boundary between units is reasonably clear, except for the issue of comments and pragmas. In default mode, the rule is that any pragmas between units belong to the previous unit, except that configuration pragmas always belong to the following unit. Any comments belong to the following unit. These rules almost always result in the right choice of the split point without you needing to mark it explicitly and you'll likely find this default to be what you want. In this default mode, you may not submit a file containing only configuration pragmas, or one that ends in configuration pragmas, to ``gnatchop``. However, using a special switch to activate 'compilation mode', ``gnatchop`` can perform another function, which is to provide exactly the semantics required by the RM for the handling of configuration pragmas in a compilation. In the absence of configuration pragmas at the main file level, this switch has no effect, but it causes such configuration pragmas to be handled in a very different manner. First, in compilation mode, if you give ``gnatchop`` a file that consists of only configuration pragmas, it appends this file to the :file:`gnat.adc` file in the current directory. This behavior provides the required behavior described in the RM for the actions to be taken on submitting such a file to the compiler, namely that these pragmas should apply to all subsequent compilations in the same compilation environment. Using GNAT, the current directory, possibly containing a :file:`gnat.adc` file is the representation of a compilation environment. For more information on the :file:`gnat.adc` file, see :ref:`Handling_of_Configuration_Pragmas`. Second, in compilation mode, if you give ``gnatchop`` a file that starts with configuration pragmas and contains one or more units, then configuration pragmas are prepended to each of the chopped files. This behavior provides the required behavior described in the RM for the actions to be taken on compiling such a file, namely that the pragmas apply to all units in the compilation, but not to subsequently compiled units. Finally, if configuration pragmas appear between units, they are appended to the previous unit. This results in the previous unit being illegal, since the compiler does not accept configuration pragmas that follow a unit. This provides the required RM behavior that forbids configuration pragmas other than those preceding the first compilation unit of a compilation. For most purposes, you will use ``gnatchop`` in default mode. You only use the compilation mode described above if you need precisely accurate behavior with respect to compilations and you have files that contain multiple units and configuration pragmas. In this circumstance, the use of ``gnatchop`` with the compilation mode switch provides the required behavior. This is the mode in which GNAT processes the ACVC tests. .. _Command_Line_for_gnatchop: Command Line for ``gnatchop`` ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ You call ``gnatchop`` as follows: .. code-block:: sh $ gnatchop switches file_name [file_name ...] [directory] The only required argument is the file name of the file to be chopped. There are no restrictions on the form of this file name. The file itself contains one or more Ada units, in normal GNAT format, concatenated together. As shown, more than one file may be presented to be chopped. When run in default mode, ``gnatchop`` generates one output file in the current directory for each unit in each of the files. ``directory``, if specified, gives the name of the directory to which the output files will be written. If you don't specify it, all files are written to the current directory. For example, given a file called :file:`hellofiles` containing .. code-block:: ada procedure Hello; with Ada.Text_IO; use Ada.Text_IO; procedure Hello is begin Put_Line ("Hello"); end Hello; the command .. code-block:: sh $ gnatchop hellofiles generates two files in the current directory, one called :file:`hello.ads` containing the single line that is the procedure spec, and the other called :file:`hello.adb` containing the remaining text. The original file is not affected. You can compile these generated files in the normal manner. When you invoke ``gnatchop`` on a file that is empty or contains only empty lines and/or comments, ``gnatchop`` will complete normally, but won't produce any new file. For example, given a file called :file:`toto.txt` containing .. code-block:: ada -- Just a comment the command .. code-block:: sh $ gnatchop toto.txt will not produce any new file and will result in the following warnings:: toto.txt:1:01: warning: empty file, contains no compilation units no compilation units found no source files written .. _Switches_for_gnatchop: Switches for ``gnatchop`` ^^^^^^^^^^^^^^^^^^^^^^^^^ ``gnatchop`` recognizes the following switches: .. index:: --version (gnatchop) :switch:`--version` Display copyright and version, then exit, disregarding all other options. .. index:: --help (gnatchop) :switch:`--help` If :switch:`--version` is not present, display usage, then exit, disregarding all other options. .. index:: -c (gnatchop) :switch:`-c` Causes ``gnatchop`` to operate in compilation mode, in which configuration pragmas are handled according to strict RM rules. See the previous section for a full description of this mode. :switch:`-gnat{xxx}` This passes the given :switch:`-gnat{xxx}` switch to ``gcc`` which is used to parse the given file. Not all *xxx* options make sense, but, for example, the use of :switch:`-gnati2` allows ``gnatchop`` to process a source file that uses Latin-2 coding for identifiers. :switch:`-h` Causes ``gnatchop`` to generate a brief help summary to the standard output file showing usage information. .. index:: -k (gnatchop) :switch:`-k{mm}` Limit generated file names to the specified number ``mm`` of characters. This is useful if the resulting set of files is required to be interoperable with systems which limit the length of file names. You may not place any space between the :switch:`-k` and the numeric value. You can omit the numeric value, in which case ``gnatchop`` will use a default of :switch:`-k8`, suitable for use with DOS-like file systems. If you don't specify a :switch:`-k` switch, there is no limit on the length of file names. .. index:: -p (gnatchop) :switch:`-p` Causes the file modification time stamp of the input file to be preserved and used for the time stamp of the output file(s). You may find this useful for preserving coherency of time stamps in an environment where ``gnatchop`` is used as part of a standard build process. .. index:: -q (gnatchop) :switch:`-q` Causes output of informational messages indicating the set of generated files to be suppressed. Warnings and error messages are unaffected. .. index:: -r (gnatchop) .. index:: Source_Reference pragmas :switch:`-r` Generate ``Source_Reference`` pragmas. Use this switch if the output files are regarded as temporary and development is to be done from of the original unchopped file. This switch causes ``Source_Reference`` pragmas to be inserted into each of the generated files to refer back to the original file name and line number. The result is that all error messages refer back to the original unchopped file. In addition, the debugging information placed into the object file (when the :switch:`-g` switch of ``gcc`` or ``gnatmake`` is specified) also refers back to this original file so that tools like profilers and debuggers will give information in terms of the original unchopped file. If the original file to be chopped itself contains a ``Source_Reference`` pragma referencing a third file, ``gnatchop`` respects these pragmas and the generated ``Source_Reference`` pragmas in the chopped file refer to the original file, with appropriate line numbers. This is particularly useful when ``gnatchop`` is used in conjunction with ``gnatprep`` to compile files that contain preprocessing statements and multiple units. .. index:: -v (gnatchop) :switch:`-v` Causes ``gnatchop`` to operate in verbose mode. It outputs the version number and copyright notice as well as exact copies of the commands spawned to obtain the information needed to control chopping. .. index:: -w (gnatchop) :switch:`-w` Overwrite existing file names. Normally, ``gnatchop`` treats it as a fatal error if there's already a file with the same name as a file it would otherwise output. This can happen either if you've previously chopped that file or if the files to be chopped contain duplicated units. This switch bypasses this check and causes all but the last instance of such duplicated units to be skipped. .. index:: --GCC= (gnatchop) :switch:`--GCC={xxxx}` Specify the path of the GNAT parser to be used. When this switch is used, ``gnatchop`` makes no attempt to add a prefix to the GNAT parser executable, so it must include the full pathname. .. _Examples_of_gnatchop_Usage: Examples of ``gnatchop`` Usage ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. code-block:: sh $ gnatchop -w hello_s.ada prerelease/files Chops the source file :file:`hello_s.ada`. The output files are placed in the directory :file:`prerelease/files`, overwriting any files with matching names in that directory (no files in the current directory are modified). .. code-block:: sh $ gnatchop archive Chops the source file :file:`archive` into the current directory. One useful application of ``gnatchop`` is in sending sets of sources around, for example in email messages. The required sources are simply concatenated (for example, using a Unix ``cat`` command) and ``gnatchop`` is used at the other end to reconstitute the original files. .. code-block:: sh $ gnatchop file1 file2 file3 direc Chops all units in files :file:`file1`, :file:`file2`, :file:`file3`, placing the resulting files in the directory :file:`direc`. Note that if any units occur more than once anywhere within this set of files, ``gnatchop`` generates an error message, and doesn't write any files. To override this check, use the :switch:`-w` switch, in which case the last occurrence in the last file will be the one that is output and ``gnatchop`` will skip earlier duplicate occurrences for the same unit. .. _Configuration_Pragmas: Configuration Pragmas ===================== .. index:: Configuration pragmas .. index:: Pragmas, configuration Configuration pragmas supported by GNAT consist of those pragmas described as such in the Ada Reference Manual and the implementation-dependent pragmas that are configuration pragmas. See the ``Implementation_Defined_Pragmas`` chapter in the :title:`GNAT_Reference_Manual` for details on these additional GNAT-specific configuration pragmas. Most notably, the pragma ``Source_File_Name``, which allows specifying non-default names for source files, is a configuration pragma. The following is a complete list of configuration pragmas recognized by GNAT:: Ada_83 Ada_95 Ada_05 Ada_2005 Ada_12 Ada_2012 Ada_2022 Aggregate_Individually_Assign Allow_Integer_Address Annotate Assertion_Policy Assume_No_Invalid_Values C_Pass_By_Copy Check_Float_Overflow Check_Name Check_Policy Component_Alignment Convention_Identifier Debug_Policy Default_Scalar_Storage_Order Default_Storage_Pool Detect_Blocking Disable_Atomic_Synchronization Discard_Names Elaboration_Checks Eliminate Enable_Atomic_Synchronization Extend_System Extensions_Allowed External_Name_Casing Fast_Math Favor_Top_Level Ignore_Pragma Implicit_Packing Initialize_Scalars Interrupt_State Interrupts_System_By_Default License Locking_Policy No_Component_Reordering No_Heap_Finalization No_Strict_Aliasing Normalize_Scalars Optimize_Alignment Overflow_Mode Overriding_Renamings Partition_Elaboration_Policy Persistent_BSS Prefix_Exception_Messages Priority_Specific_Dispatching Profile Profile_Warnings Queuing_Policy Rename_Pragma Restrictions Restriction_Warnings Reviewable Short_Circuit_And_Or Source_File_Name Source_File_Name_Project SPARK_Mode Style_Checks Suppress Suppress_Exception_Locations Task_Dispatching_Policy Unevaluated_Use_Of_Old Unsuppress Use_VADS_Size User_Aspect_Definition Validity_Checks Warning_As_Error Warnings Wide_Character_Encoding .. _Handling_of_Configuration_Pragmas: Handling of Configuration Pragmas --------------------------------- You can place configuration pragmas either appear at the start of a compilation unit or in a configuration pragma file that applies to all compilations performed in a given compilation environment. Configuration pragmas placed before a library level package specification are not propagated to the corresponding package body (see RM 10.1.5(8)); they must be added explicitly to the package body. GNAT includes the ``gnatchop`` utility to provide an automatic way to handle configuration pragmas that follows the semantics for compilations (that is, files with multiple units) described in the RM. See :ref:`Operating_gnatchop_in_Compilation_Mode` for details. However, for most purposes, you will find it more convenient to edit the :file:`gnat.adc` file that contains configuration pragmas directly, as described in the following section. In the case of ``Restrictions`` pragmas appearing as configuration pragmas in individual compilation units, the exact handling depends on the type of restriction. Restrictions that require partition-wide consistency (like ``No_Tasking``) are recognized wherever they appear and can be freely inherited, e.g. from a |withed| unit to the |withing| unit. This makes sense since the binder will always insist on seeing consistent us, so any unit not conforming to any restrictions anywhere in the partition will be rejected and it's better for you to find that out at compile time rather than bind time. For restrictions that do not require partition-wide consistency, e.g. ``SPARK`` or ``No_Implementation_Attributes``, the restriction normally applies only to the unit in which the pragma appears, and not to any other units. The exception is ``No_Elaboration_Code``, which always applies to the entire object file from a compilation, i.e. to the body, spec, and all subunits. You can apply this restriction in a configuration pragma file or you can ace it in the body and/or the spec (in either case it applies to all the relevant units). You can place it on a subunit only if you have previously placed it in the body of spec. .. _The_Configuration_Pragmas_Files: The Configuration Pragmas Files ------------------------------- .. index:: gnat.adc In GNAT, a compilation environment is defined by the current directory at the time that a compile command is given. This current directory is searched for a file whose name is :file:`gnat.adc`. If this file is present, it is expected to contain one or more configuration pragmas that will be applied to the current compilation. However, if you specify the switch :switch:`-gnatA`, GNAT ignores :file:`gnat.adc`. When used, GNAT adds :file:`gnat.adc` to the dependencies so that if :file:`gnat.adc` is modified later, the source will be recompiled on a future invocation of ``gnatmake``. You can add configuration pragmas into the :file:`gnat.adc` file either by running ``gnatchop`` on a source file consisting only of configuration pragmas or, more conveniently, by directly editing the :file:`gnat.adc` file, which is a standard format source file. Besides :file:`gnat.adc`, you may apply additional files containing configuration pragmas to the current compilation using the :switch:`-gnatec={path}` switch, where ``path`` must designate an existing file that contains only configuration pragmas. These configuration pragmas are in addition to those found in :file:`gnat.adc` (provided :file:`gnat.adc` is present and you do not use switch :switch:`-gnatA`). You can specify multiple :switch:`-gnatec=` switches. GNAT will add files containing configuration pragmas specified with switches :switch:`-gnatec=` to the dependencies, unless they are temporary files. A file is considered temporary if its name ends in :file:`.tmp` or :file:`.TMP`. Certain tools follow this naming convention because they pass information to ``gcc`` via temporary files that are immediately deleted; it doesn't make sense to depend on a file that no longer exists. Such tools include ``gprbuild``, ``gnatmake``, and ``gnatcheck``. By default, configuration pragma files are stored by their absolute paths in ALI files. You can use the :switch:`-gnateb` switch to request they be stored instead by just their basename. If you are using project file, they provide a separate mechanism using project attributes. .. --Comment See :ref:`Specifying_Configuration_Pragmas` for more details. .. _Generating_Object_Files: Generating Object Files ======================= An Ada program consists of a set of source files and the first step in compiling the program is generating the corresponding object files. You generate these by compiling a subset of these source files. The files you need to compile are the following: * If a package spec has no body, compile the package spec to produce the object file for the package. * If a package has both a spec and a body, compile the body to produce the object file for the package. You need not compile the source file for the package spec in this case because there's only one object file, which contains the code for both the spec and body of the package. * For a subprogram, compile the subprogram body to produce the object file for the subprogram. You need not compile the spec, if such a file is present. .. index:: Subunits * In the case of subunits, only compile the parent unit. GNAT generates a single object file for the entire subunit tree, which includes all the subunits. * Compile child units independently of their parent units (though, of course, the spec of all the ancestor unit must be present in order to compile a child unit). .. index:: Generics * Compile generic units in the same manner as any other units. The object files in this case are small dummy files that contain, at most, the flag used for elaboration checking. This is because GNAT always handles generic instantiation by means of macro expansion. However, you still must compile generic units for dependency checking and elaboration purposes. The preceding rules describe the set of files that must be compiled to generate all the object files for a program. See the following section on dependencies for more details on computing that set of files. Each object file has the same name as the corresponding source file, except that the extension is :file:`.o`, as usual. You may wish to compile other files for the purpose of checking their syntactic and semantic correctness. For example, in the case where a package has a separate spec and body, you would not normally compile the spec. However, it is convenient in practice to compile the spec to make sure it is error-free before compiling clients of this spec because such compilations will fail if there is an error in the spec. GNAT provides an option for compiling such files purely for the purposes of checking correctness; such compilations are not required as part of the process of building a program. To compile a file in this checking mode, use the :switch:`-gnatc` switch. .. _Source_Dependencies: Source Dependencies =================== Each object file obviously depends on at least the source file which is compiled to produce it. Here we are using "depends" in the sense of a typical ``make`` utility; in other words, an object file depends on a source file if changes to the source file require the object file to be recompiled. In addition to this basic dependency, a given object may depend on additional source files as follows: * If a file being compiled |withs| a unit ``X``, the object file depends on the file containing the spec of unit ``X``. This includes files that are |withed| implicitly either because they are parents of |withed| child units or are run-time units required by the language constructs used in a particular unit. * If a file being compiled instantiates a library level generic unit, the object file depends on both the spec and body files for this generic unit. * If a file being compiled instantiates a generic unit defined within a package, the object file depends on the body file for the package as well as the spec file. .. index:: Inline .. index:: -gnatn switch * If a file being compiled contains a call to a subprogram for which pragma ``Inline`` applies and you have activated inlining with the :switch:`-gnatn` switch, the object file depends on the file containing the body of this subprogram as well as on the file containing the spec. Note that for inlining to actually occur as a result of the use of this switch, you must compile in optimizing mode. .. index:: -gnatN switch The use of :switch:`-gnatN` activates inlining optimization that is performed by the front end of the compiler. This inlining does not require that the code generation be optimized. Like :switch:`-gnatn`, the use of this switch generates additional dependencies. When using a ``gcc`` or LLVM based back end, the use of :switch:`-gnatN` is deprecated and the use of :switch:`-gnatn` is preferred. Historically front end inlining was more extensive than back end inlining, but that is no longer the case. * If an object file :file:`O` depends on the proper body of a subunit through inlining or instantiation, it depends on the parent unit of the subunit. This means that any modification of the parent unit or one of its subunits affects the compilation of :file:`O`. * The object file for a parent unit depends on all its subunit body files. * The previous two rules means that, for purposes of computing dependencies and recompilation, a body and all its subunits are treated as an indivisible whole. These rules are applied transitively: if unit ``A`` |withs| unit ``B``, whose elaboration calls an inlined procedure in package ``C``, the object file for unit ``A`` depends on the body of ``C``, in file :file:`c.adb`. The set of dependent files described by these rules includes all the files on which the unit is semantically dependent, as dictated by the Ada language standard. However, it is a superset of what the standard describes, because it includes generic, inline, and subunit dependencies. An object file must be recreated by recompiling the corresponding source file if any of the source files on which it depends are modified. For example, if the ``make`` utility is used to control compilation, the rule for an Ada object file must mention all the source files on which the object file depends, according to the above definition. Invoking ``gnatmake`` will cause it to determine the necessary recompilations. .. _The_Ada_Library_Information_Files: The Ada Library Information Files ================================= .. index:: Ada Library Information files .. index:: ALI files Each compilation actually generates two output files. The first of these is the actual object file that has a :file:`.o` extension. The second is a text file containing full dependency information. It has the same name as the source file, but an :file:`.ali` extension. This file is known as the Ada Library Information (:file:`ALI`) file. The following information is contained in that file: * Version information (indicates which version of GNAT was used to compile the unit(s) in question) * Main program information (including priority and time slice settings, as well as the wide character encoding used during compilation). * List of arguments used in the compilation command * Attributes of the unit, including the configuration pragmas used, an indication of whether the compilation was successful, and the exception model used. * A list of relevant restrictions applying to the unit (used for consistency checking). * Categorization information (e.g., use of pragma ``Pure``). * Information on all |withed| units, including presence of ``Elaborate`` or ``Elaborate_All`` pragmas. * Information from any ``Linker_Options`` pragmas used in the unit * Information on the use of ``Body_Version`` or ``Version`` attributes in the unit. * Dependency information. This is a list of files, together with time stamp and checksum information. These are files on which the unit depends in the sense that the modification of any of these units requires the recompilation of the unit in question. * Cross-reference data. Contains information on all entities referenced in the unit. Used by some tools to provide cross-reference information. For a full detailed description of the format of the :file:`ALI` file, see the source of the spec of unit ``Lib.Writ``, contained in file :file:`lib-writ.ads` in the GNAT compiler sources. .. _Binding_an_Ada_Program: Binding an Ada Program ====================== When using languages such as C and C++, once the source files have been compiled the only remaining step in building an executable program is linking the object modules together. This means that you can link an inconsistent version of a program, in which two units have included different versions of the same header. The rules of Ada do not permit such an inconsistent program to be built. For example, if two clients have different versions of the same package, it is illegal to build a program containing these two clients. These rules are enforced by the GNAT binder, which also determines an elaboration order consistent with the Ada rules. The GNAT binder is run after all the object files for a program have been created. It is given the name of the main program unit and from this determines the set of units required by the program by reading the corresponding ALI files. It generates error messages if the program is inconsistent or if no valid order of elaboration exists. If no errors are detected, the binder produces a main program in Ada that contains calls to the elaboration procedures of those compilation unit that require them, followed by a call to the main program. This Ada program is compiled to generate the object file for the main program. The name of the Ada file is :file:`b~xxx.adb` (with the corresponding spec :file:`b~xxx.ads`) where ``xxx`` is the name of the main program unit. Finally, the linker is used to build the resulting executable program, using the object from the main program from the bind step as well as the object files for the Ada units of the program. .. _GNAT_and_Libraries: GNAT and Libraries ================== .. index:: Library building and referencing This section describes how to build and use libraries with GNAT and how to recompile the GNAT run-time library. You should be familiar with the Project Manager facility (see the *GNAT_Project_Manager* chapter of the :title:`GPRbuild User's Guide`) before reading this chapter. .. _Introduction_to_Libraries_in_GNAT: Introduction to Libraries in GNAT --------------------------------- A library is, conceptually, a collection of objects which does not have its own main thread of execution but instead provides certain services to the applications that use it. A library can be either statically linked with the application, in which case its code is directly included in the application, or, on platforms that support it, be dynamically linked, in which case its code is shared by all applications making use of this library. GNAT supports both types of libraries. In the static case, you can provide the compiled code in different ways. The simplest approach is to directly provide the set of objects resulting from compilation of the library source files. Alternatively, you can group the objects into an archive using whatever commands are provided by the operating system. In the GNAT environment, a library has these components: * Source files, * :file:`ALI` files (see :ref:`The_Ada_Library_Information_Files`), and * Object files, an archive, or a shared library. A GNAT library may expose all its source files, which is useful for documentation purposes. Alternatively, it may expose only the units needed by an external user to make use of the library, in other words, the specs reflecting the library services along with all the units needed to compile those specs, which can include generic bodies or any body implementing an inlined routine. In the case of *stand-alone libraries* those exposed units are called *interface units* (:ref:`Stand-alone_Ada_Libraries`). All compilation units comprising an application, including those in a library, need to be elaborated in an order partially defined by Ada's semantics. GNAT computes the elaboration order from the :file:`ALI` files and this is why they constitute a mandatory part of GNAT libraries. *Stand-alone libraries* are the exception to this rule because a specific library elaboration routine is produced independently of the application(s) using the library. .. _General_Ada_Libraries: General Ada Libraries --------------------- .. _Building_a_library: Building a library ^^^^^^^^^^^^^^^^^^ The easiest way to build a library is to use the Project Manager, which supports a special type of project called a *Library Project* (see the *Library Projects* section in the *GNAT Project Manager* chapter of the :title:`GPRbuild User's Guide`). A project is considered a library project when two project-level attributes are defined in it: ``Library_Name`` and ``Library_Dir``. In order to control different aspects of library configuration, you can specify additional optional project-level attributes: * ``Library_Kind`` This attribute controls whether the library is to be static or dynamic * ``Library_Version`` This attribute specifies the library version. Its value is used during dynamic linking of shared libraries to determine if the currently installed versions of the binaries are compatible. * ``Library_Options`` * ``Library_GCC`` These attributes specify additional low-level options to be used during library generation and the commands used to generate the library. The GNAT Project Manager takes complete care of the library maintenance task, including recompilation of the source files for which objects do not exist or are not up to date, assembly of the library archive, and installation of the library (i.e., copying associated source, object and :file:`ALI` files to the specified location). Here's a simple library project file: .. code-block:: gpr project My_Lib is for Source_Dirs use ("src1", "src2"); for Object_Dir use "obj"; for Library_Name use "mylib"; for Library_Dir use "lib"; for Library_Kind use "dynamic"; end My_lib; and the compilation command to build and install the library: .. code-block:: sh $ gnatmake -Pmy_lib It's complex to manually perform all the steps required to produce a library, so we recommend you use the GNAT Project Manager for this task. In case this is not desired, we discuss the necessary steps below. There are various possibilities for compiling the units that make up the library: for example with a :file:`Makefile` (:ref:`Using_the_GNU_make_Utility`) or with a conventional script. For simple libraries, you can also create a dummy main program that depends upon all the packages that comprise the interface of the library. You can then pass this dummy main program to ``gnatmake``, which will ensure all necessary objects are built. After the above has been accomplished, you should follow the standard procedure of the underlying operating system to produce the static or shared library. Here's an example of such a dummy program: .. code-block:: ada with My_Lib.Service1; with My_Lib.Service2; with My_Lib.Service3; procedure My_Lib_Dummy is begin null; end; Here are the generic commands that will build an archive or a shared library. .. code-block:: sh # compiling the library $ gnatmake -c my_lib_dummy.adb # we don't need the dummy object itself $ rm my_lib_dummy.o my_lib_dummy.ali # create an archive with the remaining objects $ ar rc libmy_lib.a *.o # some systems may require "ranlib" to be run as well # or create a shared library $ gcc -shared -o libmy_lib.so *.o # some systems may require the code to have been compiled with -fPIC # remove the object files that are now in the library $ rm *.o # Make the ALI files read-only so that gnatmake will not try to # regenerate the objects that are in the library $ chmod -w *.ali Please note that the library must have a name of the form :file:`lib{xxx}.a` or :file:`lib{xxx}.so` (or :file:`lib{xxx}.dll` on Windows) in order to be accessed by the :switch:`-l{xxx}` switch at link time. .. _Installing_a_library: Installing a library ^^^^^^^^^^^^^^^^^^^^ .. index:: ADA_PROJECT_PATH .. index:: GPR_PROJECT_PATH If you use project files, library installation is part of the library build process (see the *Installing a Library with Project Files* section of the *GNAT Project Manager* chapter of the :title:`GPRbuild User's Guide`). When you're not able to use project files for some reason, you can also install the library so that the sources needed to use the library are on the Ada source path and the ALI files & libraries be on the Ada Object path (see :ref:`Search_Paths_and_the_Run-Time_Library_RTL`), but we don't recommend doing this. Alternatively, the system administrator can place general-purpose libraries in the default compiler paths, by specifying the libraries' location in the configuration files :file:`ada_source_path` and :file:`ada_object_path`. These configuration files must be located in the GNAT installation tree at the same place as the ``gcc`` spec file. The location of the ``gcc`` spec file can be determined as follows: .. code-block:: sh $ gcc -v The configuration files mentioned above have a simple format: each line must contain one unique directory name. Those names are added to the corresponding path in their order of appearance in the file. The names can be either absolute or relative; in the latter case, they are relative to where theses files are located. The files :file:`ada_source_path` and :file:`ada_object_path` might not be present in a GNAT installation, in which case, GNAT looks for its run-time library in the directories :file:`adainclude` (for the sources) and :file:`adalib` (for the objects and :file:`ALI` files). When the files exist, the compiler does not look in :file:`adainclude` and :file:`adalib`, and thus the :file:`ada_source_path` file must contain the location for the GNAT run-time sources (which can simply be :file:`adainclude`). In the same way, the :file:`ada_object_path` file must contain the location for the GNAT run-time objects (which can simply be :file:`adalib`). You can also specify a new default path to the run-time library at compilation time with the :switch:`--RTS=rts-path` switch. You can thus choose the run-time library you want your program to be compiled with. This switch is recognized by ``gcc``, ``gnatmake``, ``gnatbind``, ``gnatls``, and all project aware tools. You can install a library before or after the standard GNAT library by selecting the ordering the lines in the configuration files. In general, a library must be installed before the GNAT library if it redefines any part of it. .. _Using_a_library: Using a library ^^^^^^^^^^^^^^^ Once again, the project facility greatly simplifies the use of libraries. In this context, using a library is just a matter of adding a |with| clause in your project. For example, to make use of the library ``My_Lib`` shown in examples in earlier sections, you can write: .. code-block:: gpr with "my_lib"; project My_Proj is ... end My_Proj; Even if you have a third-party, non-Ada library, you can still use GNAT's Project Manager facility to provide a wrapper for it. For example, the following project, when |withed| by your main project, will link with the third-party library :file:`liba.a`: .. code-block:: gpr project Liba is for Externally_Built use "true"; for Source_Files use (); for Library_Dir use "lib"; for Library_Name use "a"; for Library_Kind use "static"; end Liba; This is an alternative to the use of ``pragma Linker_Options``. It is especially interesting in the context of systems with several interdependent static libraries where finding a proper linker order is not easy and best be left to the tools having visibility over project dependence information. In order to use an Ada library manually, you need to make sure that this library is on both your source and object path (see :ref:`Search_Paths_and_the_Run-Time_Library_RTL` and :ref:`Search_Paths_for_gnatbind`). Furthermore, when the objects are grouped in an archive or a shared library, you need to specify the desired library at link time. For example, you can use the library :file:`mylib` installed in :file:`/dir/my_lib_src` and :file:`/dir/my_lib_obj` with the following commands: .. code-block:: sh $ gnatmake -aI/dir/my_lib_src -aO/dir/my_lib_obj my_appl \\ -largs -lmy_lib This can be expressed more simply: .. code-block:: sh $ gnatmake my_appl when the following conditions are met: * :file:`/dir/my_lib_src` has been added by the user to the environment variable :envvar:`ADA_INCLUDE_PATH`, or by the administrator to the file :file:`ada_source_path` * :file:`/dir/my_lib_obj` has been added by the user to the environment variable :envvar:`ADA_OBJECTS_PATH`, or by the administrator to the file :file:`ada_object_path` * a pragma ``Linker_Options`` has been added to one of the sources. For example: .. code-block:: ada pragma Linker_Options ("-lmy_lib"); Note that you may also load a library dynamically at run time given its filename, as illustrated in the GNAT :file:`plugins` example in the directory :file:`share/examples/gnat/plugins` within the GNAT install area. .. _Stand-alone_Ada_Libraries: Stand-alone Ada Libraries ------------------------- .. index:: ! Stand-alone libraries .. _Introduction_to_Stand-alone_Libraries: Introduction to Stand-alone Libraries ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ A Stand-alone Library (abbreviated 'SAL') is a library that contains the necessary code to elaborate the Ada units that are included in the library. In contrast with an ordinary library, which consists of all sources, objects and :file:`ALI` files of the library, a SAL may specify a restricted subset of compilation units to serve as a library interface. In this case, the fully self-sufficient set of files will normally consist of an objects archive, the sources of interface units' specs, and the :file:`ALI` files of interface units. If an interface spec contains a generic unit or an inlined subprogram, you must also provide the body's source; if the units that must be provided in the source form depend on other units, you must also provide the source and :file:`ALI` files of those units. The main purpose of a SAL is to minimize the recompilation overhead of client applications when a new version of the library is installed. Specifically, if the interface sources have not changed, client applications don't need to be recompiled. If, furthermore, a SAL is provided in the shared form and its version, controlled by ``Library_Version`` attribute, is not changed, the clients also do not need to be relinked. SALs also allow the library providers to minimize the amount of library source text exposed to the clients. Such 'information hiding' might be useful or necessary for various reasons. Stand-alone libraries are also well suited to be used in an executable whose main routine is not written in Ada. .. _Building_a_Stand-alone_Library: Building a Stand-alone Library ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ GNAT's Project facility provides a simple way of building and installing stand-alone libraries; see the *Stand-alone Library Projects* section in the *GNAT Project Manager* chapter of the :title:`GPRbuild User's Guide`. To be a Stand-alone Library Project, in addition to the two attributes that make a project a Library Project (``Library_Name`` and ``Library_Dir``; see the *Library Projects* section in the *GNAT Project Manager* chapter of the *GPRbuild User's Guide*), you must define the attribute ``Library_Interface``. For example: .. code-block:: gpr for Library_Dir use "lib_dir"; for Library_Name use "dummy"; for Library_Interface use ("int1", "int1.child"); Attribute ``Library_Interface`` has a non-empty string list value, each string in the list designating a unit contained in an immediate source of the project file. When a Stand-alone Library is built, the binder is first invoked to build a package whose name depends on the library name (:file:`b~dummy.ads/b` in the example above). This binder-generated package includes initialization and finalization procedures whose names depend on the library name (``dummyinit`` and ``dummyfinal`` in the example above). The object corresponding to this package is included in the library. You must ensure timely (e.g., prior to any use of interfaces in the SAL) calling of these procedures if a static SAL is built, or if a shared SAL is built with the project-level attribute ``Library_Auto_Init`` set to ``"false"``. For a Stand-Alone Library, only the :file:`ALI` files of the Interface Units (those that are listed in attribute ``Library_Interface``) are copied to the Library Directory. As a consequence, only the Interface Units may be imported from Ada units outside of the library. If other units are imported, the binding phase will fail. You can also build an encapsulated library where not only the code to elaborate and finalize the library is embedded but also ensure that the library is linked only against static libraries. That means that an encapsulated library only depends on system libraries: all other code, including the GNAT runtime, is embedded. To build an encapsulated library you must set attribute ``Library_Standalone`` to ``encapsulated``: .. code-block:: gpr for Library_Dir use "lib_dir"; for Library_Name use "dummy"; for Library_Kind use "dynamic"; for Library_Interface use ("int1", "int1.child"); for Library_Standalone use "encapsulated"; The default value for this attribute is ``standard`` in which case a stand-alone library is built. You may specify the attribute ``Library_Src_Dir`` for a Stand-Alone Library. ``Library_Src_Dir`` has a single string value. Its value must be the path (absolute or relative to the project directory) of an existing directory. This directory cannot be the object directory or one of the source directories, but it can be the same as the library directory. The sources of the Interface Units of the library that are needed by an Ada client of the library are copied to the designated directory, called the Interface Copy directory, when the library is built. These sources include the specs of the Interface Units, but they may also include bodies and subunits when pragmas ``Inline`` or ``Inline_Always`` are used or when there is a generic unit in the spec. Before the sources are copied to the Interface Copy directory, the building process makes an attempt to delete all files in the Interface Copy directory. Building stand-alone libraries by hand is somewhat tedious, but for those occasions when it is necessary here are the steps that you need to perform: * Compile all library sources. * Invoke the binder with the switch :switch:`-n` (No Ada main program), with all the :file:`ALI` files of the interfaces, and with the switch :switch:`-L` to give specific names to the ``init`` and ``final`` procedures. For example: .. code-block:: sh $ gnatbind -n int1.ali int2.ali -Lsal1 * Compile the binder generated file: .. code-block:: sh $ gcc -c b~int2.adb * Link the dynamic library with all the necessary object files, passing to the linker the names of the ``init`` (and possibly ``final``) procedures for automatic initialization (and finalization). You should place the built library in a different directory than the object files. * Copy the ``ALI`` files of the interface to the library directory, add in this copy an indication that it is an interface to a SAL (i.e., add a word ``SL`` on the line in the :file:`ALI` file that starts with letter 'P') and make the modified copy of the :file:`ALI` file read-only. Using SALs is not different from using other libraries (see :ref:`Using_a_library`). .. _Creating_a_Stand-alone_Library_to_be_used_in_a_non-Ada_context: Creating a Stand-alone Library to be used in a non-Ada context ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ It's easy for you to adapt the SAL build procedure discussed above for use of a SAL in a non-Ada context. The only extra step required is to ensure that library interface subprograms are compatible with the main program, by means of ``pragma Export`` or ``pragma Convention``. Here's an example of simple library interface for use with C main program: .. code-block:: ada package My_Package is procedure Do_Something; pragma Export (C, Do_Something, "do_something"); procedure Do_Something_Else; pragma Export (C, Do_Something_Else, "do_something_else"); end My_Package; On the C side, you must provide a 'foreign' view of the library interface; remember that it should contain elaboration routines in addition to interface subprograms. The example below shows the content of :file:`mylib_interface.h` (note that there is no rule for the naming of this file, any name can be used) .. code-block:: c /* the library elaboration procedure */ extern void mylibinit (void); /* the library finalization procedure */ extern void mylibfinal (void); /* the interface exported by the library */ extern void do_something (void); extern void do_something_else (void); Libraries built as explained above can be used from any program, provided the elaboration procedures (named ``mylibinit`` in the previous example) are called before any library services are used. Any number of libraries can be called from a single executable as long as the elaboration procedure of each library is called. Below is an example of a C program that uses the ``mylib`` library. .. code-block:: c #include "mylib_interface.h" int main (void) { /* First, elaborate the library before using it */ mylibinit (); /* Main program, using the library exported entities */ do_something (); do_something_else (); /* Library finalization at the end of the program */ mylibfinal (); return 0; } Note that invoking any library finalization procedure generated by ``gnatbind`` shuts down the Ada run-time environment. Consequently, the finalization of all Ada libraries must be performed at the end of the program. No call to these libraries or to the Ada run-time library should be made after the finalization phase. Information on limitations of binding Ada code in non-Ada contexts can be found under :ref:`Binding_with_Non-Ada_Main_Programs`. Note also that you must take special care with multi-tasking applications. In that case, the initialization and finalization routines are not protected against concurrent access. If you need such requirement, you must ensure it at the application level using a specific operating system services like a mutex or a critical-section. .. _Restrictions_in_Stand-alone_Libraries: Restrictions in Stand-alone Libraries ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ You should use the pragmas listed below with caution inside libraries, since they can create incompatibilities with other Ada libraries: * pragma ``Locking_Policy`` * pragma ``Partition_Elaboration_Policy`` * pragma ``Queuing_Policy`` * pragma ``Task_Dispatching_Policy`` * pragma ``Unreserve_All_Interrupts`` When using a library that contains such pragmas, the user of the library must ensure that all libraries use the same pragmas with the same values. Otherwise, ``Program_Error`` will be raised during the elaboration of the conflicting libraries. You should document the usage of these pragmas and its consequences for the user. Similarly, the traceback in the exception occurrence mechanism should be enabled or disabled in a consistent manner across all libraries. Otherwise, ``Program_Error`` will be raised during the elaboration of the conflicting libraries. If you use the ``Version`` or ``Body_Version`` attributes inside a library, you need to perform a ``gnatbind`` step that specifies all :file:`ALI` files in all libraries so that version identifiers can be properly computed. In practice these attributes are rarely used, so this is unlikely to be a consideration. .. _Rebuilding_the_GNAT_Run-Time_Library: Rebuilding the GNAT Run-Time Library ------------------------------------ .. index:: GNAT Run-Time Library, rebuilding .. index:: Building the GNAT Run-Time Library .. index:: Rebuilding the GNAT Run-Time Library .. index:: Run-Time Library, rebuilding You may need to recompile the GNAT library in various debugging or experimentation contexts. The GNAT distribution provides a project file called :file:`libada.gpr` to do that; it can be found in the directory containing the GNAT library. The location of this directory depends on the way the GNAT environment has been installed and can be determined by means of the command: .. code-block:: sh $ gnatls -v The last entry in the source search path usually contains the GNAT library (the :file:`adainclude` directory). This project file contains its own documentation and, in particular, the set of instructions needed to rebuild a new library and to use it. Note that rebuilding the GNAT Run-Time is only recommended for temporary experiments or debugging and is not supported for other purposes. .. index:: ! Conditional compilation .. _Conditional_Compilation: Conditional Compilation ======================= This section presents some guidelines for modeling conditional compilation in Ada and describes the gnatprep preprocessor utility. .. index:: ! Conditional compilation .. _Modeling_Conditional_Compilation_in_Ada: Modeling Conditional Compilation in Ada --------------------------------------- You may want to arrange for a single source program to serve multiple purposes, where it is compiled in different ways to achieve these different goals. Some examples of the need for this feature are * Adapting a program to a different hardware environment * Adapting a program to a different target architecture * Turning debugging features on and off * Arranging for a program to compile with different compilers In C, or C++, the typical approach is to use the preprocessor defined as part of the language. The Ada language does not contain such a feature. This is not an oversight, but rather a very deliberate design decision, based on the experience that overuse of the preprocessing features in C and C++ can result in programs that are extremely difficult to maintain. For example, if we have ten switches that can be on or off, this means that there are a thousand separate programs, any one of which might not even be syntactically correct, and, even if syntactically correct, might not work correctly. Testing all combinations can quickly become impossible. Nevertheless, the need to tailor programs certainly exists and in this section we will discuss how this can be achieved using Ada in general and GNAT in particular. .. _Use_of_Boolean_Constants: Use of Boolean Constants ^^^^^^^^^^^^^^^^^^^^^^^^ In the case where the difference is simply which code sequence is executed, the cleanest solution is to use Boolean constants to control which code is executed. .. code-block:: ada FP_Initialize_Required : constant Boolean := True; ... if FP_Initialize_Required then ... end if; Not only will the code inside the ``if`` statement not be executed if the constant Boolean is ``False``, but it will also be completely deleted from the program. However, the code is only deleted after the ``if`` statement block has been checked for syntactic and semantic correctness. (In contrast, with preprocessors the code is deleted before the compiler ever gets to see it, so it is not checked until the switch is turned on.) .. index:: Preprocessors (contrasted with conditional compilation) Typically the Boolean constants will be in a separate package, something like: .. code-block:: ada package Config is FP_Initialize_Required : constant Boolean := True; Reset_Available : constant Boolean := False; ... end Config; You would write the ``Config`` package multiple forms for various targets, with an appropriate script selecting the version of ``Config`` needed. Then, any other unit requiring conditional compilation can do a |with| of ``Config`` to make the constants visible. .. _Debugging_-_A_Special_Case: Debugging - A Special Case ^^^^^^^^^^^^^^^^^^^^^^^^^^ A common use of conditional code is to execute statements (for example dynamic checks, or output of intermediate results) under control of a debug switch, so that the debugging behavior can be turned on and off. You can do this by using a Boolean constant to control whether the debug code is active: .. code-block:: ada if Debugging then Put_Line ("got to the first stage!"); end if; or .. code-block:: ada if Debugging and then Temperature > 999.0 then raise Temperature_Crazy; end if; .. index:: pragma Assert Since this is a common case, GNAT provides special features to deal with this in a convenient manner. For the case of tests, Ada 2005 has added a pragma ``Assert`` that you can use for such tests. This pragma is modeled on the ``Assert`` pragma that has always been available in GNAT, so you can use this feature with GNAT even if you are not using Ada 2005 features. The use of pragma ``Assert`` is described in the :title:`GNAT_Reference_Manual`, but as an example, the last test could be written: .. code-block:: ada pragma Assert (Temperature <= 999.0, "Temperature Crazy"); or simply .. code-block:: ada pragma Assert (Temperature <= 999.0); In both cases, if assertions are active and the temperature is excessive, the exception ``Assert_Failure`` is raised with the exception message using the specified string in the first case or a string indicating the location of the pragma in the second case. .. index:: pragma Assertion_Policy You can turn assertions on and off by using the ``Assertion_Policy`` pragma. .. index:: -gnata switch This is an Ada 2005 pragma that is implemented in all modes by GNAT. Alternatively, you can use the :switch:`-gnata` switch to enable assertions from the command line, which also applies to all versions of Ada. .. index:: pragma Debug For the example above with the ``Put_Line``, the GNAT-specific pragma ``Debug`` can be used: .. code-block:: ada pragma Debug (Put_Line ("got to the first stage!")); If debug pragmas are enabled, the argument, which must be of the form of a procedure call, is executed (in this case, ``Put_Line`` is called). You can specify only one call, but you can of course include a special debugging procedure containing any code you like in the program and call it in a pragma ``Debug`` argument as needed. One advantage of pragma ``Debug`` over the ``if Debugging then`` construct is that pragma ``Debug`` can appear in declarative contexts, such as at the very beginning of a procedure, before local declarations have been elaborated. .. index:: pragma Debug_Policy You can enable debug pragmas using either the :switch:`-gnata` switch that also controls assertions, or with a separate Debug_Policy pragma. The latter pragma is new in the Ada 2005 versions of GNAT (but it can be used in Ada 95 and Ada 83 programs as well) and is analogous to pragma ``Assertion_Policy`` to control assertions. ``Assertion_Policy`` and ``Debug_Policy`` are configuration pragmas, and thus can appear in :file:`gnat.adc` if you are not using a project file or in the file designated to contain configuration pragmas in a project file. They then apply to all subsequent compilations. In practice the use of the :switch:`-gnata` switch is often the most convenient method of controlling the status of these pragmas. Note that a pragma is not a statement, so in contexts where a statement sequence is required, you can't just write a pragma on its own. You have to add a ``null`` statement. .. code-block:: ada if ... then ... -- some statements else pragma Assert (Num_Cases < 10); null; end if; .. _Conditionalizing_Declarations: Conditionalizing Declarations ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ In some cases it may be necessary to conditionalize declarations to meet different requirements. For example we might want a bit string whose length is set to meet some hardware message requirement. This may be possible using declare blocks controlled by conditional constants: .. code-block:: ada if Small_Machine then declare X : Bit_String (1 .. 10); begin ... end; else declare X : Large_Bit_String (1 .. 1000); begin ... end; end if; Note that in this approach, both declarations are analyzed by the compiler so this can only be used where both declarations are legal, even though one of them will not be used. Another approach is to define integer constants, e.g., ``Bits_Per_Word``, or Boolean constants, e.g., ``Little_Endian``, and then write declarations that are parameterized by these constants. For example .. code-block:: ada for Rec use Field1 at 0 range Boolean'Pos (Little_Endian) * 10 .. Bits_Per_Word; end record; If ``Bits_Per_Word`` is set to 32, this generates either .. code-block:: ada for Rec use Field1 at 0 range 0 .. 32; end record; for the big endian case, or .. code-block:: ada for Rec use record Field1 at 0 range 10 .. 32; end record; for the little endian case. Since a powerful subset of Ada expression notation is usable for creating static constants, clever use of this feature can often solve quite difficult problems in conditionalizing compilation (note incidentally that in Ada 95, the little endian constant was introduced as ``System.Default_Bit_Order``, so you don't need to define this one yourself). .. _Use_of_Alternative_Implementations: Use of Alternative Implementations ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ In some cases, none of the approaches described above are adequate. This can occur, for example, if the set of declarations required is radically different for two different configurations. In this situation, the official Ada way of dealing with conditionalizing such code is to write separate units for the different cases. As long as this doesn't result in excessive duplication of code, you can do this without creating maintenance problems. The approach is to share common code as far as possible and then isolate the code and declarations that are different. Subunits are often a convenient method for breaking out a piece of a unit that you need to be conditionalized, with separate files for different versions of the subunit for different targets, where the build script selects the right one to give to the compiler. .. index:: Subunits (and conditional compilation) As an example, consider a situation where a new feature in Ada 2005 allows something to be done in a really nice way. But your code must be able to compile with an Ada 95 compiler. Conceptually you want to say: .. code-block:: ada if Ada_2005 then ... neat Ada 2005 code else ... not quite as neat Ada 95 code end if; where ``Ada_2005`` is a Boolean constant. But this won't work when ``Ada_2005`` is set to ``False``, since the ``then`` clause will be illegal for an Ada 95 compiler. (Recall that although such unreachable code would eventually be deleted by the compiler, it still needs to be legal. If it uses features introduced in Ada 2005, it's still illegal in Ada 95.) So instead, we write .. code-block:: ada procedure Insert is separate; Then we have two files for the subunit ``Insert``, with the two sets of code. If the package containing this is called ``File_Queries``, then we might have two files * :file:`file_queries-insert-2005.adb` * :file:`file_queries-insert-95.adb` and the build script renames the appropriate file to :file:`file_queries-insert.adb` and then carries out the compilation. This can also be done with project files' naming schemes. For example: .. code-block:: gpr for body ("File_Queries.Insert") use "file_queries-insert-2005.ada"; Note also that with project files, you should use a different extension than :file:`ads` / :file:`adb` for alternative versions. Otherwise, a naming conflict may arise through another commonly used feature: declaring as part of the project a set of directories containing all the sources obeying the default naming scheme. The use of alternative units is certainly feasible in all situations, and for example the Ada part of the GNAT run-time is conditionalized based on the target architecture using this approach. As a specific example, consider the implementation of the AST feature in VMS. There is one spec: :file:`s-asthan.ads` which is the same for all architectures, and three bodies: * :file:`s-asthan.adb` used for all non-VMS operating systems * :file:`s-asthan-vms-alpha.adb` used for VMS on the Alpha * :file:`s-asthan-vms-ia64.adb` used for VMS on the ia64 The dummy version :file:`s-asthan.adb` simply raises exceptions noting that this operating system feature is not available and the two remaining versions interface with the corresponding versions of VMS to provide VMS-compatible AST handling. The GNAT build script knows the architecture and operating system, and automatically selects the right version, renaming it if necessary to :file:`s-asthan.adb` before the run-time build. Another style for arranging alternative implementations is through Ada's access-to-subprogram facility. In case some functionality is to be conditionally included, you can declare an access-to-procedure variable ``Ref`` that is initialized to designate a 'do nothing' procedure, and then invoke ``Ref.all`` when appropriate. Then, in, some library package, set ``Ref`` to ``Proc'Access`` for some procedure ``Proc`` that performs the relevant processing. The initialization only occurs if the library package is included in the program. The same idea can also be implemented using tagged types and dispatching calls. .. _Preprocessing: Preprocessing ^^^^^^^^^^^^^ .. index:: Preprocessing Although it is quite possible to conditionalize code without the use of C-style preprocessing, as described in the cases above, it is nevertheless convenient in some cases to use the C approach. Moreover, older Ada compilers have often provided some preprocessing capability, so legacy code may depend on this approach, even though it is not standard. To accommodate such use, GNAT provides a preprocessor (modeled to a large extent on the various preprocessors that have been used with legacy code on other compilers, to enable easier transition). .. index:: gnatprep You can use the preprocessor used in two different modes. You can use it separately from the compiler to generate a separate output source file, which you then feed to the compiler as a separate step. This is the ``gnatprep`` utility, whose use is fully described in :ref:`Preprocessing_with_gnatprep`. The preprocessing language allows such constructs as .. code-block:: c #if DEBUG or else (PRIORITY > 4) then sequence of declarations #else completely different sequence of declarations #end if; The values of the symbols ``DEBUG`` and ``PRIORITY`` can be defined either on the command line or in a separate file. The other way of running the preprocessor is even closer to the C style and often more convenient. In this approach, the preprocessing is integrated into the compilation process. You pass the compiler the preprocessor input, which includes ``#if`` lines etc, and the compiler carries out the preprocessing internally and compiles the resulting output. For more details on this approach, see :ref:`Integrated_Preprocessing`. .. _Preprocessing_with_gnatprep: Preprocessing with ``gnatprep`` ------------------------------- .. index:: ! gnatprep .. index:: Preprocessing (gnatprep) This section discusses how to you can use GNAT's ``gnatprep`` utility for simple preprocessing. Although designed for use with GNAT, ``gnatprep`` does not depend on any special GNAT features. For further discussion of conditional compilation in general, see :ref:`Conditional_Compilation`. .. _Preprocessing_Symbols: Preprocessing Symbols ^^^^^^^^^^^^^^^^^^^^^ Preprocessing symbols are defined in *definition files* and referenced in the sources to be preprocessed. A preprocessing symbol is an identifier, following normal Ada (case-insensitive) rules for its syntax, with the restriction that all characters need to be in the ASCII set (no accented letters). .. _Using_gnatprep: Using ``gnatprep`` ^^^^^^^^^^^^^^^^^^ To call ``gnatprep`` use: .. code-block:: sh $ gnatprep [ switches ] infile outfile [ deffile ] where * *switches* is an optional sequence of switches as described in the next section. * *infile* is the full name of the input file, which is an Ada source file containing preprocessor directives. * *outfile* is the full name of the output file, which is an Ada source in standard Ada form. When used with GNAT, this file name will normally have an ``ads`` or ``adb`` suffix. * ``deffile`` is the full name of a text file containing definitions of preprocessing symbols to be referenced by the preprocessor. You can omit this argument and instead use the :switch:`-D` switch. .. _Switches_for_gnatprep: Switches for ``gnatprep`` ^^^^^^^^^^^^^^^^^^^^^^^^^ .. index:: --version (gnatprep) :switch:`--version` Display copyright and version, then exit, disregarding all other options. .. index:: --help (gnatprep) :switch:`--help` If :switch:`--version` was not used, display usage and then exit, disregarding all other options. .. index:: -b (gnatprep) :switch:`-b` Causes both preprocessor lines and the lines deleted by preprocessing to be replaced by blank lines in the output source file, preserving line numbers in the output file. .. index:: -c (gnatprep) :switch:`-c` Causes both preprocessor lines and the lines deleted by preprocessing to be retained in the output source as comments marked with the special string ``"--! "``. This option also results in line numbers being preserved in the output file. .. index:: -C (gnatprep) :switch:`-C` Causes comments to be scanned. Normally comments are ignored by ``gnatprep``. If you specify this option, ``gnatprep`` scans comments and any $symbol substitutions performed as in program text. You will find this particularly useful when structured comments are used (e.g., for programs written in a pre-2014 version of the SPARK Ada subset). This switch is not available when doing integrated preprocessing (it would be useless in this context since comments are always ignored by the compiler). .. index:: -D (gnatprep) :switch:`-D{symbol}[={value}]` Defines a new preprocessing symbol with the specified value. If you don't specify a value, the symbol is defined to be ``True``. You can use this switch instead of providing a definition file. .. index:: -e (gnatprep) :switch:`-e` Causes both preprocessor lines and the lines deleted by preprocessing to be replaced by empty comment lines marked with ``"--!"`` (and no other text) in the output source file, preserving line numbers in the output file. This option can be useful as an alternative to :switch:`-b` and :switch:`-c` when compilation style switches like :switch:`-gnatyu` or :switch:`-gnatyM` are used (to avoid warnings about multiple blank lines or lines too long). .. index:: -r (gnatprep) :switch:`-r` Causes ``gnatprep`` to generate a ``Source_Reference`` pragma that references the original input file, so that error messages will use the file name of this original file. The use of this switch implies that preprocessor lines are not to be removed from the file, so the :switch:`-b` and :switch:`-c` are always enabled. If the file to be preprocessed contains multiple units, you must call ``gnatchop`` on the the output file from ``gnatprep``. If a ``Source_Reference`` pragma is present in the preprocessed file, it will be respected by ``gnatchop -r`` so that the final chopped files will correctly refer to the original input source file passed to ``gnatprep``. .. index:: -s (gnatprep) :switch:`-s` Causes a sorted list of symbol names and values to be listed on the standard output file. .. index:: -T (gnatprep) :switch:`-T` Use LF as line terminators when writing files. By default the line terminator of the host (LF under unix, CR/LF under Windows) is used. .. index:: -u (gnatprep) :switch:`-u` Causes undefined symbols to be treated as having the value ``False`` in the context of a preprocessor test. If you don't specify this switch, ``gnatprep`` will treat an undefined symbol in a ``#if`` or ``#elsif`` test as an error. .. index:: -v (gnatprep) :switch:`-v` Verbose mode: generates more output about what is done. Note: if you don't specify either :switch:`-b` or :switch:`-c`, then preprocessor lines and deleted lines are completely removed from the output, unless you specify :switch:`-r`, in which case ``gnatprep`` enables the :switch:`-b` switch. .. _Form_of_Definitions_File: Form of Definitions File ^^^^^^^^^^^^^^^^^^^^^^^^ The definitions file contains lines of the form:: symbol := value where ``symbol`` is a preprocessing symbol, and ``value`` is one of the following: * Empty, corresponding to a null substitution, * A string literal using normal Ada syntax, or * Any sequence of characters from the set {letters, digits, period, underline}. You may also place comment lines in the definitions file, starting with the usual ``--`` and comments may be added to the end of each definition line. .. _Form_of_Input_Text_for_gnatprep: Form of Input Text for ``gnatprep`` ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The input text contains preprocessor conditional inclusion lines as well as general symbol substitution sequences. Preprocessor conditional inclusion commands have the form: .. code-block:: c #if [then] lines #elsif [then] lines #elsif [then] lines ... #else lines #end if; In this example, is defined by the following grammar:: ::= ::= = "" ::= = ::= = ::= > ::= >= ::= < ::= <= ::= 'Defined ::= not ::= and ::= or ::= and then ::= or else ::= ( ) For the first test, ( ::= ), the symbol must have either the value true or false. The right-hand of the symbol definition must be one of the (case-insensitive) literals ``True`` or ``False``. If the value is true, the corresponding lines are included and if the value is false, they are excluded. When comparing a symbol to an integer, the integer is any non negative literal integer as defined in the Ada Reference Manual, such as 3, 16#FF# or 2#11#. The symbol value must also be a non negative integer. Integer values in the range 0 .. 2**31-1 are supported. The test ( ::= 'Defined) is true only if the symbol has been defined in the definition file or by a :switch:`-D` switch on the command line. Otherwise, the test is false. The equality tests are case insensitive, as are all the preprocessor lines. If the symbol referenced is not defined in the symbol definitions file, the result depends on whether or not you have specified the :switch:`-u` switch. If you have, the symbol is treated as if it had the value false and the test fails. If not, it's an error to reference an undefined symbol. It's also an error to reference a symbol that you have defined with a value other than ``True`` or ``False``. The use of the ``not`` operator inverts the sense of this logical test. You can't combine the ``not`` operator with the ``or`` or ``and`` operators without parentheses. For example, you can't write "if not X or Y then" allowed, but can write either "if (not X) or Y then" or "if not (X or Y) then". The ``then`` keyword is optional, as shown. You must place the ``#`` in the first non-blank character on a line, i.e., it must be preceded only by spaces or horizontal tabs, but otherwise the format is free form. You may place spaces or tabs between the ``#`` and the keyword. The keywords and the symbols are case insensitive, as in normal Ada code. You can write comments on a preprocessor line, but other than that, you can't place any other tokens on a preprocessor line. You can have any number of ``elsif`` clauses, including none at all. The ``else`` is optional, as in Ada. You obtain symbol substitution outside of preprocessor lines by using the sequence:: $symbol anywhere within a source line, except in a comment or within a string literal. The identifier following the ``$`` must match one of the symbols defined in the symbol definition file and the resulting output substitutes the value of the symbol in place of ``$symbol`` in the output file. Note that although you can't substitute strings within a string literal, you can have a symbol whose defined value is a string literal. So instead of setting XYZ to ``hello`` and writing: .. code-block:: ada Header : String := "$XYZ"; you should set XYZ to ``"hello"`` and write: .. code-block:: ada Header : String := $XYZ; and then the substitution will occur as desired. .. _Integrated_Preprocessing: Integrated Preprocessing ------------------------ As noted above, a file to be preprocessed consists of Ada source code in which preprocessing lines have been inserted. However, instead of using ``gnatprep`` to explicitly preprocess a file as a separate step before compilation, you can carry out the preprocessing implicitly as part of compilation. Such *integrated preprocessing*, which is the common style with C, is performed when you pass either or both of the following switches to the compiler: * :switch:`-gnatep`, which specifies the *preprocessor data file*. This file dictates how the source files will be preprocessed (e.g., which symbol definition files apply to which sources). * :switch:`-gnateD`, which defines values for preprocessing symbols. Integrated preprocessing applies only to Ada source files; it's not available for configuration pragma files. With integrated preprocessing, GNAT doesn't write the output from the preprocessor, by default, to any external file. Instead it's passed internally to the compiler. To preserve the result of preprocessing in a file, either run ``gnatprep`` in standalone mode or supply the :switch:`-gnateG` switch to the compiler. When using project files: * you should use the builder switch :switch:`-x` if any Ada source is compiled with :switch:`gnatep=` so that the compiler finds the *preprocessor data file*. * you should place the preprocessing data file and the symbol definition files in the source directories of the project. Note that the ``gnatmake`` switch :switch:`-m` will almost always trigger recompilation for sources that are preprocessed, because ``gnatmake`` cannot compute the checksum of the source after preprocessing. The actual preprocessing function is described in detail in :ref:`Preprocessing_with_gnatprep`. This section explains the switches that relate to integrated preprocessing. .. index:: -gnatep (gcc) :switch:`-gnatep={preprocessor_data_file}` This switch specifies the file name (without directory information) of the preprocessor data file. Either place this file in one of the source directories, or, when using project files, reference the project file's directory via the ``project_name'Project_Dir`` project attribute; e.g: .. code-block:: gpr project Prj is package Compiler is for Switches ("Ada") use ("-gnatep=" & Prj'Project_Dir & "prep.def"); end Compiler; end Prj; A preprocessor data file is a text file that contains *preprocessor control lines*. A preprocessor control line directs the preprocessing of either a particular source file, or, analogous to ``others`` in Ada, all sources not specified elsewhere in the preprocessor data file. A preprocessor control line can optionally identify a *definition file* that assigns values to preprocessor symbols, as well as a list of switches that relate to preprocessing. You can also include empty lines and comments (using Ada syntax), with no semantic effect. Here's an example of a preprocessor data file: .. code-block:: ada "toto.adb" "prep.def" -u -- Preprocess toto.adb, using definition file prep.def -- Undefined symbols are treated as False * -c -DVERSION=V101 -- Preprocess all other sources without using a definition file -- Suppressed lined are commented -- Symbol VERSION has the value V101 "tata.adb" "prep2.def" -s -- Preprocess tata.adb, using definition file prep2.def -- List all symbols with their values A preprocessor control line has the following syntax: :: ::= [ ] { } ::= | '*' ::= := := (See below for list) Thus, you start each preprocessor control line either a literal string or the character '*': * A literal string is the file name (without directory information) of the source file that will be input to the preprocessor. * The character '*' is a wild-card indicator; the additional parameters on the line indicate the preprocessing for all the sources that are not specified explicitly on other lines (the order of the lines is not significant). You cannot have two lines with the same file name or two lines starting with the '*' character. After the file name or '*', you can place an optional literal string to specify the name of the definition file to be used for preprocessing (:ref:`Form_of_Definitions_File`). The definition files are found by the compiler in one of the source directories. In some cases, when compiling a source in a directory other than the current directory, if the definition file is in the current directory, you may need to add the current directory as a source directory through the :switch:`-I` switch; otherwise the compiler would not find the definition file. Finally, switches similar to those of ``gnatprep`` may optionally appear: :switch:`-b` Causes both preprocessor lines and the lines deleted by preprocessing to be replaced by blank lines, preserving the line number. This switch is always implied; however, if specified after :switch:`-c` or :switch:`-e` it cancels the effect of those switches. :switch:`-c` Causes both preprocessor lines and the lines deleted by preprocessing to be retained as comments marked with the special string '`--!`'. :switch:`-D{symbol}={new_value}` Define or redefine ``symbol`` to have ``new_value`` as its value. You can write ``symbol`` as either an Ada identifier or any Ada reserved word aside from ``if``, ``else``, ``elsif``, ``end``, ``and``, ``or`` and ``then``. You can write ``new_value`` as a literal string, an Ada identifier or any Ada reserved word. A symbol declared with this switch replaces a symbol with the same name defined in a definition file. :switch:`-e` Causes both preprocessor lines and the lines deleted by preprocessing to be replaced by empty comment lines marked with '`--!`' (and no other text) in the output source file, :switch:`-s` Causes a sorted list of symbol names and values to be listed on the standard output file. :switch:`-u` Causes undefined symbols to be treated as having the value ``FALSE`` in the context of a preprocessor test. If you don't specify this switch, an undefined symbol in a ``#if`` or ``#elsif`` test is treated as an error. .. index:: -gnateD (gcc) :switch:`-gnateD{symbol}[={new_value}]` Define or redefine ``symbol`` to have ``new_value`` as its value. If you don't specify a value, the value of ``symbol`` is ``True``. You write ``symbol`` as an identifier, following normal Ada (case-insensitive) rules for its syntax, and ``new_value`` as either an arbitrary string between double quotes or any sequence (including an empty sequence) of characters from the set (letters, digits, period, underline). Ada reserved words may be used as symbols, with the exceptions of ``if``, ``else``, ``elsif``, ``end``, ``and``, ``or`` and ``then``. Examples: :: -gnateDToto=Tata -gnateDFoo -gnateDFoo=\"Foo-Bar\" A symbol declared with this switch on the command line replaces a symbol with the same name either in a definition file or specified with a switch :switch:`-D` in the preprocessor data file. This switch is similar to switch :switch:`-D` of ``gnatprep``. .. index:: -gnateG (gcc) :switch:`-gnateG[bce]` When integrated preprocessing is performed on source file :file:`filename.extension`, create or overwrite :file:`filename.extension.prep` to contain the result of the preprocessing. For example if the source file is :file:`foo.adb` then the output file is :file:`foo.adb.prep`. An optional character (b, c, or e) can be appended to indicate that filtered lines are to be replaced by blank lines, comments, or empty comments (see documentation above about :switch:`-b`, :switch:`-c`, and :switch:`-e`). If one of those switches is given in a preprocessor data file, then it will override any option included with :switch:`-gnateG`. .. _Mixed_Language_Programming: Mixed Language Programming ========================== .. index:: Mixed Language Programming This section describes how to develop a mixed-language program, with a focus on combining Ada with C or C++. .. _Interfacing_to_C: Interfacing to C ---------------- Interfacing Ada with a foreign language such as C involves using compiler directives to import and/or export entity definitions in each language -- using ``extern`` statements in C, for example, and the ``Import``, ``Export``, and ``Convention`` pragmas in Ada. A full treatment of these topics is provided in Appendix B, section 1 of the Ada Reference Manual. There are two ways to build a program using GNAT that contains some Ada sources and some foreign language sources, depending on whether or not the main subprogram is written in Ada. Here's an example with the main subprogram in Ada: .. code-block:: c /* file1.c */ #include void print_num (int num) { printf ("num is %d.\\n", num); return; } .. code-block:: c /* file2.c */ /* num_from_Ada is declared in my_main.adb */ extern int num_from_Ada; int get_num (void) { return num_from_Ada; } .. code-block:: ada -- my_main.adb procedure My_Main is -- Declare then export an Integer entity called num_from_Ada My_Num : Integer := 10; pragma Export (C, My_Num, "num_from_Ada"); -- Declare an Ada function spec for Get_Num, then use -- C function get_num for the implementation. function Get_Num return Integer; pragma Import (C, Get_Num, "get_num"); -- Declare an Ada procedure spec for Print_Num, then use -- C function print_num for the implementation. procedure Print_Num (Num : Integer); pragma Import (C, Print_Num, "print_num"); begin Print_Num (Get_Num); end My_Main; To build this example: * First compile the foreign language files to generate object files: .. code-block:: sh $ gcc -c file1.c $ gcc -c file2.c * Then compile the Ada units to produce a set of object files and ALI files: .. code-block:: sh $ gnatmake -c my_main.adb * Run the Ada binder on the Ada main program: .. code-block:: sh $ gnatbind my_main.ali * Link the Ada main program, the Ada objects, and the other language objects: .. code-block:: sh $ gnatlink my_main.ali file1.o file2.o You can merge the last three steps into a single command: .. code-block:: sh $ gnatmake my_main.adb -largs file1.o file2.o .. index:: Binder output file If the main program is in a language other than Ada, you may have more than one entry point into the Ada subsystem. You must use a special binder option to generate callable routines that initialize and finalize the Ada units (:ref:`Binding_with_Non-Ada_Main_Programs`). You must insert calls to the initialization and finalization routines in the main program or some other appropriate point in the code. You must place the call to initialize the Ada units so that it occurs before the first Ada subprogram is called and must place the call to finalize the Ada units so it occurs after the last Ada subprogram returns. The binder places the initialization and finalization subprograms into the :file:`b~xxx.adb` file, where they can be accessed by your C sources. To illustrate, we have the following example: .. code-block:: c /* main.c */ extern void adainit (void); extern void adafinal (void); extern int add (int, int); extern int sub (int, int); int main (int argc, char *argv[]) { int a = 21, b = 7; adainit(); /* Should print "21 + 7 = 28" */ printf ("%d + %d = %d\\n", a, b, add (a, b)); /* Should print "21 - 7 = 14" */ printf ("%d - %d = %d\\n", a, b, sub (a, b)); adafinal(); } .. code-block:: ada -- unit1.ads package Unit1 is function Add (A, B : Integer) return Integer; pragma Export (C, Add, "add"); end Unit1; .. code-block:: ada -- unit1.adb package body Unit1 is function Add (A, B : Integer) return Integer is begin return A + B; end Add; end Unit1; .. code-block:: ada -- unit2.ads package Unit2 is function Sub (A, B : Integer) return Integer; pragma Export (C, Sub, "sub"); end Unit2; .. code-block:: ada -- unit2.adb package body Unit2 is function Sub (A, B : Integer) return Integer is begin return A - B; end Sub; end Unit2; The build procedure for this application is similar to the last example's: * First, compile the foreign language files to generate object files: .. code-block:: sh $ gcc -c main.c * Next, compile the Ada units to produce a set of object files and ALI files: .. code-block:: sh $ gnatmake -c unit1.adb $ gnatmake -c unit2.adb * Run the Ada binder on every generated ALI file. Make sure to use the :switch:`-n` option to specify a foreign main program: .. code-block:: sh $ gnatbind -n unit1.ali unit2.ali * Link the Ada main program, the Ada objects and the foreign language objects. You need only list the last ALI file here: .. code-block:: sh $ gnatlink unit2.ali main.o -o exec_file This procedure yields a binary executable called :file:`exec_file`. Depending on the circumstances (for example when your non-Ada main object does not provide symbol ``main``), you may also need to instruct the GNAT linker not to include the standard startup objects by passing the :switch:`-nostartfiles` switch to ``gnatlink``. .. _Calling_Conventions: Calling Conventions ------------------- .. index:: Foreign Languages .. index:: Calling Conventions GNAT follows standard calling sequence conventions and will interface to any other language that also follows these conventions. The following Convention identifiers are recognized by GNAT: .. index:: Interfacing to Ada .. index:: Other Ada compilers .. index:: Convention Ada ``Ada`` This indicates that the standard Ada calling sequence is used and all Ada data items may be passed without any limitations in the case where GNAT is used to generate both the caller and callee. You can also mix GNAT generated code and code generated by another Ada compiler. In this case, you should restrict the data types to simple cases, including primitive types. Whether complex data types can be passed depends on the situation. It is probably safe to pass simple arrays, such as arrays of integers or floats. Records may or may not work, depending on whether both compilers lay them out identically. Complex structures involving variant records, access parameters, tasks, or protected types, are unlikely to be able to be passed. If output from two different compilers is mixed, you are responsible for dealing with elaboration issues. Probably the safest approach is to write the main program in the version of Ada other than GNAT, so it takes care of its own elaboration requirements, and call the GNAT-generated adainit procedure to ensure elaboration of the GNAT components. Consult the documentation of the other Ada compiler for further details on elaboration. .. index:: Interfacing to Assembly .. index:: Convention Assembler ``Assembler`` Specifies assembler as the convention. In practice this has the same effect as convention Ada (but is not equivalent in the sense of being considered the same convention). .. index:: Convention Asm .. index:: Asm ``Asm`` Equivalent to Assembler. .. index:: Interfacing to COBOL .. index:: Convention COBOL .. index:: COBOL ``COBOL`` Data is passed according to the conventions described in section B.4 of the Ada Reference Manual. .. index:: C .. index:: Interfacing to C .. index:: Convention C ``C`` Data is passed according to the conventions described in section B.3 of the Ada Reference Manual. A note on interfacing to a C 'varargs' function: .. index:: C varargs function .. index:: Interfacing to C varargs function .. index:: varargs function interfaces In C, ``varargs`` allows a function to take a variable number of arguments. There is no direct equivalent in this to Ada. One approach that you can use is to create a C wrapper for each different profile and then interface to this C wrapper. For example, to print an ``int`` value using ``printf``, create a C function ``printfi`` that takes two arguments, a pointer to a string and an int, and calls ``printf``. Then in the Ada program, use pragma ``Import`` to interface to ``printfi``. It may work on some platforms to directly interface to a ``varargs`` function by providing a specific Ada profile for a particular call. However, this does not work on all platforms since there is no guarantee that the calling sequence for a two-argument normal C function is the same as for calling a ``varargs`` C function with the same two arguments. .. index:: Convention Default .. index:: Default ``Default`` Equivalent to C. .. index:: Convention External .. index:: External ``External`` Equivalent to C. .. index:: C++ .. index:: Interfacing to C++ .. index:: Convention C++ ``C_Plus_Plus`` (or ``CPP``) This stands for C++. For most purposes, this is identical to C. See the separate description of the specialized GNAT pragmas relating to C++ interfacing for further details. .. index:: Fortran .. index:: Interfacing to Fortran .. index:: Convention Fortran ``Fortran`` Data is passed according to the conventions described in section B.5 of the Ada Reference Manual. ``Intrinsic`` This applies to an intrinsic operation, as defined in the Ada Reference Manual. If a pragma Import (Intrinsic) applies to a subprogram, it means the body of the subprogram is provided by the compiler itself, usually by means of an efficient code sequence, and that you don't supply an explicit body for it. In an application program, the pragma may be applied to the following sets of names: * ``Rotate_Left``, ``Rotate_Right``, ``Shift_Left``, ``Shift_Right``, ``Shift_Right_Arithmetic``. The corresponding subprogram declaration must have two formal parameters. The first must be a signed integer type or a modular type with a binary modulus and the second parameter must be of type Natural. The return type must be the same as the type of the first argument. The size of this type can only be 8, 16, 32, or 64. * Binary arithmetic operators: '+', '-', '*', '/'. The corresponding operator declaration must have parameters and result type that have the same root numeric type (for example, all three are long_float types). This simplifies the definition of operations that use type checking to perform dimensional checks: .. code-block:: ada type Distance is new Long_Float; type Time is new Long_Float; type Velocity is new Long_Float; function "/" (D : Distance; T : Time) return Velocity; pragma Import (Intrinsic, "/"); You often program this common idiom with a generic definition and an explicit body. The pragma makes it simpler to introduce such declarations. It incurs no overhead in compilation time or code size because it is implemented as a single machine instruction. * General subprogram entities. This is used to bind an Ada subprogram declaration to a compiler builtin by name with back ends where such interfaces are available. A typical example is the set of ``__builtin`` functions exposed by the ``gcc`` back end, as in the following example: .. code-block:: ada function builtin_sqrt (F : Float) return Float; pragma Import (Intrinsic, builtin_sqrt, "__builtin_sqrtf"); Most of the ``gcc`` builtins are accessible this way, and as for other import conventions (e.g. C), it is the user's responsibility to ensure that the Ada subprogram profile matches the underlying builtin expectations. .. index:: Stdcall .. index:: Convention Stdcall ``Stdcall`` This is relevant only to Windows implementations of GNAT and specifies that the ``Stdcall`` calling sequence is used, as defined by the NT API. To simplify building cross-platform bindings, this convention is handled as a ``C`` calling convention on non-Windows platforms. .. index:: DLL .. index:: Convention DLL ``DLL`` This is equivalent to ``Stdcall``. .. index:: Win32 .. index:: Convention Win32 ``Win32`` This is equivalent to ``Stdcall``. .. index:: Stubbed .. index:: Convention Stubbed ``Stubbed`` This is a special convention that indicates that the compiler should provide a stub body that raises ``Program_Error``. GNAT additionally provides a useful pragma ``Convention_Identifier`` that you can use to parameterize conventions and allow additional synonyms. For example, if you have legacy code in which the convention identifier Fortran77 was used for Fortran, you can use the configuration pragma: .. code-block:: ada pragma Convention_Identifier (Fortran77, Fortran); And from now on, you can use the identifier ``Fortran77`` as a convention identifier (for example in an ``Import`` pragma) with the same meaning as ``Fortran``. .. _Building_Mixed_Ada_and_C++_Programs: Building Mixed Ada and C++ Programs ----------------------------------- If you are inexperienced with mixed-language development, you may find that building an application containing both Ada and C++ code can be a challenge. This section gives a few hints that should make this task easier. .. _Interfacing_to_C++: Interfacing to C++ ^^^^^^^^^^^^^^^^^^ GNAT supports interfacing with the G++ compiler (or any C++ compiler generating code that is compatible with the G++ Application Binary Interface ---see http://itanium-cxx-abi.github.io/cxx-abi/abi.html). You can do interfacing at three levels: simple data, subprograms, and classes. In the first two cases, GNAT offers a specific ``Convention C_Plus_Plus`` (or ``CPP``) that behaves exactly like ``Convention C``. Usually, C++ mangles the names of subprograms. To generate proper mangled names automatically, see :ref:`Generating_Ada_Bindings_for_C_and_C++_headers`). You can also address this problem addressed manually in two ways: * by modifying the C++ code in order to force a C convention using the ``extern "C"`` syntax. * by figuring out the mangled name (using e.g. ``nm`` or by looking at the assembly code generated by the C++ compiler) and using it as the ``Link_Name`` argument of the pragma ``Import``. You can achieve interfacing at the class level by using the GNAT specific pragmas such as ``CPP_Constructor``. See the :title:`GNAT_Reference_Manual` for additional information. .. _Linking_a_Mixed_C++_and_Ada_Program: Linking a Mixed C++ & Ada Program ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Usually the linker, of the C++ development system must be used to link mixed applications because most C++ systems resolve elaboration issues (such as calling constructors on global class instances) transparently during the link phase. GNAT has been adapted to ease the use of a foreign linker for the last phase. We consider three cases: * Using GNAT and G++ (GNU C++ compiler) from the same GCC installation: You can call the C++ linker by using the C++ specific driver called ``g++``. If the C++ code uses inline functions that you plan to call from Ada, you need to compile your C++ code with the :switch:`-fkeep-inline-functions` so ``g++`` doesn't delete these functions. .. code-block:: sh $ g++ -c -fkeep-inline-functions file1.C $ g++ -c -fkeep-inline-functions file2.C $ gnatmake ada_unit -largs file1.o file2.o --LINK=g++ * Using GNAT and G++ from two different GCC installations: If both compilers are on the :envvar:`PATH`, you may use the previous method. However, environment variables such as :envvar:`C_INCLUDE_PATH`, :envvar:`GCC_EXEC_PREFIX`, :envvar:`BINUTILS_ROOT`, and :envvar:`GCC_ROOT` affect both compilers at the same time and may make one of the two compilers operate improperly if set during invocation of the wrong compiler. It is also very important that the linker uses the proper :file:`libgcc.a` ``gcc`` library -- that is, the one from the C++ compiler installation. You can replace the implicit link command as suggested in the ``gnatmake`` command from the former example with an explicit link command with the full-verbosity option in order to verify which library is used: .. code-block:: sh $ gnatbind ada_unit $ gnatlink -v -v ada_unit file1.o file2.o --LINK=c++ If there's a problem due to interfering environment variables, you can work around it by using an intermediate script. The following example shows the proper script to use when GNAT has not been installed at its default location and ``g++`` has been installed at its default location: .. code-block:: sh $ cat ./my_script #!/bin/sh unset BINUTILS_ROOT unset GCC_ROOT c++ $* $ gnatlink -v -v ada_unit file1.o file2.o --LINK=./my_script * Using a non-GNU C++ compiler: You can use the commands previously described used to insure that the C++ linker is used. Nonetheless, you need to add a few more parameters to the link command line, depending on the exception mechanism used. If you are using the ``setjmp`` / ``longjmp`` exception mechanism, you need only include the paths to the ``libgcc`` libraries: .. code-block:: sh $ cat ./my_script #!/bin/sh CC $* gcc -print-file-name=libgcc.a gcc -print-file-name=libgcc_eh.a $ gnatlink ada_unit file1.o file2.o --LINK=./my_script where CC is the name of the non-GNU C++ compiler. If you are using the "zero cost" exception mechanism and the platform supports automatic registration of exception tables (e.g., Solaris), you need to include paths to more objects: .. code-block:: sh $ cat ./my_script #!/bin/sh CC gcc -print-file-name=crtbegin.o $* \\ gcc -print-file-name=libgcc.a gcc -print-file-name=libgcc_eh.a \\ gcc -print-file-name=crtend.o $ gnatlink ada_unit file1.o file2.o --LINK=./my_script If you are using the "zero cost exception" mechanism is used and the platform doesn't support automatic registration of exception tables (e.g., HP-UX or AIX), the simple approach described above won't work and a you will need to preform a pre-linking phase using GNAT. Another alternative is to use the :command:`gprbuild` multi-language builder which has a large knowledge base and knows how to link Ada and C++ code together automatically in most cases. .. _A_Simple_Example: A Simple Example ^^^^^^^^^^^^^^^^ The following example, provided as part of the GNAT examples, shows how to achieve procedural interfacing between Ada and C++ in both directions. The C++ class ``A`` has two methods. The first method is exported to Ada by the means of an extern C wrapper function. The second method calls an Ada subprogram. On the Ada side, the C++ calls are modelled by a limited record with a layout comparable to the C++ class. The Ada subprogram, in turn, calls the C++ method. So, starting from the C++ main program, execution passes back and forth between the two languages. Here are the compilation commands: .. code-block:: sh $ gnatmake -c simple_cpp_interface $ g++ -c cpp_main.C $ g++ -c ex7.C $ gnatbind -n simple_cpp_interface $ gnatlink simple_cpp_interface -o cpp_main --LINK=g++ -lstdc++ ex7.o cpp_main.o Here are the corresponding sources: .. code-block:: cpp //cpp_main.C #include "ex7.h" extern "C" { void adainit (void); void adafinal (void); void method1 (A *t); } void method1 (A *t) { t->method1 (); } int main () { A obj; adainit (); obj.method2 (3030); adafinal (); } .. code-block:: cpp //ex7.h class Origin { public: int o_value; }; class A : public Origin { public: void method1 (void); void method2 (int v); A(); int a_value; }; .. code-block:: cpp //ex7.C #include "ex7.h" #include extern "C" { void ada_method2 (A *t, int v);} void A::method1 (void) { a_value = 2020; printf ("in A::method1, a_value = %d \\n",a_value); } void A::method2 (int v) { ada_method2 (this, v); printf ("in A::method2, a_value = %d \\n",a_value); } A::A(void) { a_value = 1010; printf ("in A::A, a_value = %d \\n",a_value); } .. code-block:: ada -- simple_cpp_interface.ads with System; package Simple_Cpp_Interface is type A is limited record Vptr : System.Address; O_Value : Integer; A_Value : Integer; end record; pragma Convention (C, A); procedure Method1 (This : in out A); pragma Import (C, Method1); procedure Ada_Method2 (This : in out A; V : Integer); pragma Export (C, Ada_Method2); end Simple_Cpp_Interface; .. code-block:: ada -- simple_cpp_interface.adb package body Simple_Cpp_Interface is procedure Ada_Method2 (This : in out A; V : Integer) is begin Method1 (This); This.A_Value := V; end Ada_Method2; end Simple_Cpp_Interface; .. _Interfacing_with_C++_constructors: Interfacing with C++ constructors ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ To interface with C++ constructors GNAT provides the ``pragma CPP_Constructor`` (see the :title:`GNAT_Reference_Manual` for additional information). In this section, we present some common uses of C++ constructors in mixed-languages programs in GNAT. Let us assume we need to interface with the following C++ class: .. code-block:: cpp class Root { public: int a_value; int b_value; virtual int Get_Value (); Root(); // Default constructor Root(int v); // 1st non-default constructor Root(int v, int w); // 2nd non-default constructor }; For this purpose, we can write the following package spec (further information on how to build this spec is available in :ref:`Interfacing_with_C++_at_the_Class_Level` and :ref:`Generating_Ada_Bindings_for_C_and_C++_headers`). .. code-block:: ada with Interfaces.C; use Interfaces.C; package Pkg_Root is type Root is tagged limited record A_Value : int; B_Value : int; end record; pragma Import (CPP, Root); function Get_Value (Obj : Root) return int; pragma Import (CPP, Get_Value); function Constructor return Root; pragma Cpp_Constructor (Constructor, "_ZN4RootC1Ev"); function Constructor (v : Integer) return Root; pragma Cpp_Constructor (Constructor, "_ZN4RootC1Ei"); function Constructor (v, w : Integer) return Root; pragma Cpp_Constructor (Constructor, "_ZN4RootC1Eii"); end Pkg_Root; On the Ada side, the constructor is represented by a function (whose name is arbitrary) that returns the classwide type corresponding to the imported C++ class. Although the constructor is described as a function, it's typically a procedure with an extra implicit argument (the object being initialized) at the implementation level. GNAT issues the appropriate call, whatever it is, to get the object properly initialized. Constructors can only appear in the following contexts: * On the right side of an initialization of an object of type. * On the right side of an initialization of a record component of type. * In an Ada 2005 limited aggregate. * In an Ada 2005 nested limited aggregate. * In an Ada 2005 limited aggregate that initializes an object built in place by an extended return statement. In a declaration of an object whose type is a class imported from C++, either the default C++ constructor is implicitly called by GNAT or you must explicitly call the required C++ constructor in the expression that initializes the object. For example: .. code-block:: ada Obj1 : Root; Obj2 : Root := Constructor; Obj3 : Root := Constructor (v => 10); Obj4 : Root := Constructor (30, 40); The first two declarations are equivalent: in both cases the default C++ constructor is invoked (in the former case the call to the constructor is implicit and in the latter case the call is explicit in the object declaration). ``Obj3`` is initialized by the C++ non-default constructor that takes an integer argument and ``Obj4`` is initialized by the non-default C++ constructor that takes two integers. Let's derive the imported C++ class in the Ada side. For example: .. code-block:: ada type DT is new Root with record C_Value : Natural := 2009; end record; In this case, you must initialize the components ``DT`` inherited from the C++ side by a C++ constructor and the additional Ada components of type DT are initialized by GNAT. The initialization of such an object is done either by default, or by means of a function returning an aggregate of type DT, or by means of an extension aggregate. .. code-block:: ada Obj5 : DT; Obj6 : DT := Function_Returning_DT (50); Obj7 : DT := (Constructor (30,40) with C_Value => 50); The declaration of ``Obj5`` invokes the default constructors: the C++ default constructor of the parent type takes care of the initialization of the components inherited from Root and GNAT takes care of the default initialization of the additional Ada components of type DT (that is, ``C_Value`` is initialized to value 2009). The order of invocation of the constructors is consistent with the order of elaboration required by Ada and C++. That is, the constructor of the parent type is always called before the constructor of the derived type. Let's now consider a record that has components whose type is imported from C++. For example: .. code-block:: ada type Rec1 is limited record Data1 : Root := Constructor (10); Value : Natural := 1000; end record; type Rec2 (D : Integer := 20) is limited record Rec : Rec1; Data2 : Root := Constructor (D, 30); end record; The initialization of an object of type ``Rec2`` calls the non-default C++ constructors specified for the imported components. For example: .. code-block:: ada Obj8 : Rec2 (40); Using Ada 2005, we can use limited aggregates to initialize an object invoking C++ constructors that differ from those specified in the type declarations. For example: .. code-block:: ada Obj9 : Rec2 := (Rec => (Data1 => Constructor (15, 16), others => <>), others => <>); The above declaration uses an Ada 2005 limited aggregate to initialize ``Obj9`` and the C++ constructor that has two integer arguments is invoked to initialize the ``Data1`` component instead of the constructor specified in the declaration of type ``Rec1``. In Ada 2005, the box in the aggregate indicates that unspecified components are initialized using the expression (if any) available in the component declaration. That is, in this case discriminant ``D`` is initialized to value ``20``, ``Value`` is initialized to value 1000, and the non-default C++ constructor that handles two integers takes care of initializing component ``Data2`` with values ``20,30``. In Ada 2005, we can use the extended return statement to build the Ada equivalent to C++ non-default constructors. For example: .. code-block:: ada function Constructor (V : Integer) return Rec2 is begin return Obj : Rec2 := (Rec => (Data1 => Constructor (V, 20), others => <>), others => <>) do -- Further actions required for construction of -- objects of type Rec2 ... end record; end Constructor; In this example, we use the extended return statement construct to build in place the returned object whose components are initialized by means of a limited aggregate. We could also place any further action associated with the constructor inside the construct. .. _Interfacing_with_C++_at_the_Class_Level: Interfacing with C++ at the Class Level ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ In this section, we demonstrate the GNAT features for interfacing with C++ by means of an example making use of Ada 2005 abstract interface types. This example consists of a classification of animals; classes have been used to model our main classification of animals and interfaces provide support for the management of secondary classifications. We first demonstrate a case in which the types and constructors are defined on the C++ side and imported from the Ada side and then the reverse case. The root of our derivation is the ``Animal`` class, with a single private attribute (the ``Age`` of the animal), a constructor, and two public primitives to set and get the value of this attribute. .. code-block:: cpp class Animal { public: virtual void Set_Age (int New_Age); virtual int Age (); Animal() {Age_Count = 0;}; private: int Age_Count; }; Abstract interface types are defined in C++ by means of classes with pure virtual functions and no data members. In our example we use two interfaces that provide support for the common management of ``Carnivore`` and ``Domestic`` animals: .. code-block:: cpp class Carnivore { public: virtual int Number_Of_Teeth () = 0; }; class Domestic { public: virtual void Set_Owner (char* Name) = 0; }; Using these declarations, we can now say that a ``Dog`` is an animal that is both Carnivore and Domestic, that is: .. code-block:: cpp class Dog : Animal, Carnivore, Domestic { public: virtual int Number_Of_Teeth (); virtual void Set_Owner (char* Name); Dog(); // Constructor private: int Tooth_Count; char *Owner; }; In the following examples we assume that the previous declarations are located in a file named :file:`animals.h`. The following package demonstrates how to import these C++ declarations from the Ada side: .. code-block:: ada with Interfaces.C.Strings; use Interfaces.C.Strings; package Animals is type Carnivore is limited interface; pragma Convention (C_Plus_Plus, Carnivore); function Number_Of_Teeth (X : Carnivore) return Natural is abstract; type Domestic is limited interface; pragma Convention (C_Plus_Plus, Domestic); procedure Set_Owner (X : in out Domestic; Name : Chars_Ptr) is abstract; type Animal is tagged limited record Age : Natural; end record; pragma Import (C_Plus_Plus, Animal); procedure Set_Age (X : in out Animal; Age : Integer); pragma Import (C_Plus_Plus, Set_Age); function Age (X : Animal) return Integer; pragma Import (C_Plus_Plus, Age); function New_Animal return Animal; pragma CPP_Constructor (New_Animal); pragma Import (CPP, New_Animal, "_ZN6AnimalC1Ev"); type Dog is new Animal and Carnivore and Domestic with record Tooth_Count : Natural; Owner : Chars_Ptr; end record; pragma Import (C_Plus_Plus, Dog); function Number_Of_Teeth (A : Dog) return Natural; pragma Import (C_Plus_Plus, Number_Of_Teeth); procedure Set_Owner (A : in out Dog; Name : Chars_Ptr); pragma Import (C_Plus_Plus, Set_Owner); function New_Dog return Dog; pragma CPP_Constructor (New_Dog); pragma Import (CPP, New_Dog, "_ZN3DogC2Ev"); end Animals; Thanks to the compatibility between GNAT run-time structures and the C++ ABI, interfacing with these C++ classes is easy. The only requirement is that you must declare all the primitives and components exactly in the same order in the two languages. Regarding the abstract interfaces, we must indicate to the GNAT compiler, by means of a ``pragma Convention (C_Plus_Plus)``, that the convention used to pass the arguments to the called primitives will be the same as for C++. For the imported classes, we use ``pragma Import`` with convention ``C_Plus_Plus`` to indicate they have been defined on the C++ side; this is required because the dispatch table associated with these tagged types will be built in the C++ side and therefore will not contain the predefined Ada primitives which Ada would otherwise expect. As the reader can see, there is no need to indicate the C++ mangled names associated with each subprogram because it is assumed that all the calls to these primitives will be dispatching calls. The only exception is the constructor, which we must register with the compiler by means of ``pragma CPP_Constructor`` and we need to provide its associated C++ mangled name because the Ada compiler generates direct calls to it. With the above packages, we can now declare objects of type ``Dog`` on the Ada side and dispatch calls to the corresponding subprograms on the C++ side. We can also extend the tagged type ``Dog`` with further fields and primitives and override some of its C++ primitives on the Ada side. For example, here we have a type derivation defined on the Ada side that inherits all the dispatching primitives of the ancestor from the C++ side. .. code-block:: ada with Animals; use Animals; package Vaccinated_Animals is type Vaccinated_Dog is new Dog with null record; function Vaccination_Expired (A : Vaccinated_Dog) return Boolean; end Vaccinated_Animals; It is important to note that, because of the ABI compatibility, we don't need to add any further information to indicate either the object layout or the dispatch table entry associated with each dispatching operation. Now let's define all the types and constructors on the Ada side and export them to C++, using the same hierarchy of our previous example: .. code-block:: ada with Interfaces.C.Strings; use Interfaces.C.Strings; package Animals is type Carnivore is limited interface; pragma Convention (C_Plus_Plus, Carnivore); function Number_Of_Teeth (X : Carnivore) return Natural is abstract; type Domestic is limited interface; pragma Convention (C_Plus_Plus, Domestic); procedure Set_Owner (X : in out Domestic; Name : Chars_Ptr) is abstract; type Animal is tagged record Age : Natural; end record; pragma Convention (C_Plus_Plus, Animal); procedure Set_Age (X : in out Animal; Age : Integer); pragma Export (C_Plus_Plus, Set_Age); function Age (X : Animal) return Integer; pragma Export (C_Plus_Plus, Age); function New_Animal return Animal'Class; pragma Export (C_Plus_Plus, New_Animal); type Dog is new Animal and Carnivore and Domestic with record Tooth_Count : Natural; Owner : String (1 .. 30); end record; pragma Convention (C_Plus_Plus, Dog); function Number_Of_Teeth (A : Dog) return Natural; pragma Export (C_Plus_Plus, Number_Of_Teeth); procedure Set_Owner (A : in out Dog; Name : Chars_Ptr); pragma Export (C_Plus_Plus, Set_Owner); function New_Dog return Dog'Class; pragma Export (C_Plus_Plus, New_Dog); end Animals; Compared with our previous example the only differences are the use of ``pragma Convention`` (instead of ``pragma Import``) and the use of ``pragma Export`` to indicate to the GNAT compiler that the primitives will be available to C++. Thanks to the ABI compatibility, on the C++ side there is nothing else to be done; as explained above, the only requirement is that all the primitives and components are declared in exactly the same order. For completeness, let us see a brief C++ main program that uses the declarations available in :file:`animals.h` (presented in our first example) to import and use the declarations from the Ada side, properly initializing and finalizing the Ada run-time system along the way: .. code-block:: cpp #include "animals.h" #include using namespace std; void Check_Carnivore (Carnivore *obj) {...} void Check_Domestic (Domestic *obj) {...} void Check_Animal (Animal *obj) {...} void Check_Dog (Dog *obj) {...} extern "C" { void adainit (void); void adafinal (void); Dog* new_dog (); } void test () { Dog *obj = new_dog(); // Ada constructor Check_Carnivore (obj); // Check secondary DT Check_Domestic (obj); // Check secondary DT Check_Animal (obj); // Check primary DT Check_Dog (obj); // Check primary DT } int main () { adainit (); test(); adafinal (); return 0; } .. _Partition_Wide_Settings: Partition-Wide Settings ----------------------- When building a mixed-language application, you must be aware that Ada enforces some partition-wide settings that may implicitly impact the behavior of the other languages. This is the case for certain signals that are reserved to the implementation to implement proper Ada semantics (such as the behavior of ``abort`` statements). It means that the Ada part of the application may override signal handlers that were previously installed by either the system or by other user code. If your application requires that either system or user signals be preserved, you need to instruct the Ada part not to install its own signal handler. You do this using ``pragma Interrupt_State`` that provides a general mechanism for overriding such uses of interrupts. Additionally, you can use pragma ``Interrupts_System_By_Default`` to default all interrupts to System. The set of interrupts for which the Ada run-time library sets a specific signal handler is the following: * Ada.Interrupts.Names.SIGSEGV * Ada.Interrupts.Names.SIGBUS * Ada.Interrupts.Names.SIGFPE * Ada.Interrupts.Names.SIGILL * Ada.Interrupts.Names.SIGABRT You can instruct the run-time library not to install its signal handler for a particular signal by using the configuration pragma ``Interrupt_State`` in the Ada code. For example: .. code-block:: ada pragma Interrupt_State (Ada.Interrupts.Names.SIGSEGV, System); pragma Interrupt_State (Ada.Interrupts.Names.SIGBUS, System); pragma Interrupt_State (Ada.Interrupts.Names.SIGFPE, System); pragma Interrupt_State (Ada.Interrupts.Names.SIGILL, System); pragma Interrupt_State (Ada.Interrupts.Names.SIGABRT, System); Obviously, if the Ada run-time system cannot set these handlers it comes with the drawback of not fully preserving Ada semantics. ``SIGSEGV``, ``SIGBUS``, ``SIGFPE`` and ``SIGILL`` are used to raise corresponding Ada exceptions in the application, while ``SIGABRT`` is used to asynchronously abort an action or a task. .. _Generating_Ada_Bindings_for_C_and_C++_headers: Generating Ada Bindings for C and C++ headers --------------------------------------------- .. index:: Binding generation (for C and C++ headers) .. index:: C headers (binding generation) .. index:: C++ headers (binding generation) GNAT includes a binding generator for C and C++ headers, which is intended to do 95% of the tedious work of generating Ada specs from C or C++ header files. This capability is not intended to generate 100% correct Ada specs and it will in some cases require you to make manual adjustments, although it can often be used out of the box in practice. Some of the known limitations include: * only very simple character constant macros are translated into Ada constants. Function macros (macros with arguments) are partially translated as comments, to be completed manually if needed. * some extensions (e.g. vector types) are not supported * pointers to pointers are mapped to ``System.Address`` * identifiers with names that are identical except for casing may generate compilation errors (e.g. ``shm_get`` vs ``SHM_GET``). The code is generated using Ada 2012 syntax, which makes it easier to interface with other languages. In most cases, you can still use the generated binding even if your code is compiled using earlier versions of Ada (e.g. ``-gnat95``). .. _Running_the_binding_generator: Running the Binding Generator ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The binding generator is part of the ``gcc`` compiler and you can invoke it via the :switch:`-fdump-ada-spec` switch, which generates Ada spec files for the header files specified on the command line and all header files needed by these files transitively. For example: .. code-block:: sh $ gcc -c -fdump-ada-spec -C /usr/include/time.h $ gcc -c *.ads generates, under GNU/Linux, the following files: :file:`time_h.ads`, :file:`bits_time_h.ads`, :file:`stddef_h.ads`, :file:`bits_types_h.ads` which correspond to the files :file:`/usr/include/time.h`, and :file:`/usr/include/bits/time.h` and then compile these Ada specs. The name of the Ada specs is consistent with the relative path under :file:`/usr/include/` of the header files. This behavior is specific to paths ending with :file:`/include/`; in all the other cases, the name of the Ada specs is derived from the simple name of the header files instead. The :switch:`-C` switch tells ``gcc`` to extract comments from headers, and attempt to generate corresponding Ada comments. If you want to generate a single Ada file and not the transitive closure, you can use instead the :switch:`-fdump-ada-spec-slim` switch. You can optionally specify a parent unit, of which all generated units will be children, using :switch:`-fada-spec-parent={unit}`. The simple ``gcc``-based command works only for C headers. For C++ headers you need to use either the ``g++`` command or the combination ``gcc -x c++``. In some cases, the generated bindings will be more complete or more meaningful when defining some macros, which you can do via the :switch:`-D` switch. This is for example the case with :file:`Xlib.h` under GNU/Linux: .. code-block:: sh $ gcc -c -fdump-ada-spec -DXLIB_ILLEGAL_ACCESS -C /usr/include/X11/Xlib.h The above generates more complete bindings than a call without the :switch:`-DXLIB_ILLEGAL_ACCESS` switch. In other cases, you can't parse a header file in a stand-alone manner because other include files need to be included first. In this case, the solution is to create a small header file including the needed ``#include`` and possible ``#define`` directives. For example, to generate Ada bindings for :file:`readline/readline.h`, you need to first include :file:`stdio.h`, so you can create a file with the following two lines in e.g. :file:`readline1.h`: .. code-block:: cpp #include #include and then generate Ada bindings from this file: .. code-block:: sh $ gcc -c -fdump-ada-spec readline1.h .. _Generating_bindings_for_C++_headers: Generating Bindings for C++ Headers ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Generating bindings for C++ headers is done using the same options, but with the ``g++`` compiler. Note that generating Ada spec from C++ headers is a much more complex job and support for C++ headers is much more limited that support for C headers. As a result, you will need to modify the resulting bindings by hand more extensively when using C++ headers. In this mode, C++ classes are mapped to Ada tagged types, constructors are mapped using the ``CPP_Constructor`` pragma, and when possible, multiple inheritance of abstract classes are mapped to Ada interfaces (see the *Interfacing to C++* section in the :title:`GNAT Reference Manual` for additional information on interfacing to C++). For example, given the following C++ header file: .. code-block:: cpp class Carnivore { public: virtual int Number_Of_Teeth () = 0; }; class Domestic { public: virtual void Set_Owner (char* Name) = 0; }; class Animal { public: int Age_Count; virtual void Set_Age (int New_Age); }; class Dog : Animal, Carnivore, Domestic { public: int Tooth_Count; char *Owner; virtual int Number_Of_Teeth (); virtual void Set_Owner (char* Name); Dog(); }; The corresponding Ada code is generated: .. code-block:: ada package Class_Carnivore is type Carnivore is limited interface; pragma Import (CPP, Carnivore); function Number_Of_Teeth (this : access Carnivore) return int is abstract; end; use Class_Carnivore; package Class_Domestic is type Domestic is limited interface; pragma Import (CPP, Domestic); procedure Set_Owner (this : access Domestic; Name : Interfaces.C.Strings.chars_ptr) is abstract; end; use Class_Domestic; package Class_Animal is type Animal is tagged limited record Age_Count : aliased int; end record; pragma Import (CPP, Animal); procedure Set_Age (this : access Animal; New_Age : int); pragma Import (CPP, Set_Age, "_ZN6Animal7Set_AgeEi"); end; use Class_Animal; package Class_Dog is type Dog is new Animal and Carnivore and Domestic with record Tooth_Count : aliased int; Owner : Interfaces.C.Strings.chars_ptr; end record; pragma Import (CPP, Dog); function Number_Of_Teeth (this : access Dog) return int; pragma Import (CPP, Number_Of_Teeth, "_ZN3Dog15Number_Of_TeethEv"); procedure Set_Owner (this : access Dog; Name : Interfaces.C.Strings.chars_ptr); pragma Import (CPP, Set_Owner, "_ZN3Dog9Set_OwnerEPc"); function New_Dog return Dog; pragma CPP_Constructor (New_Dog); pragma Import (CPP, New_Dog, "_ZN3DogC1Ev"); end; use Class_Dog; .. _Switches_for_Ada_Binding_Generation: Switches ^^^^^^^^ .. index:: -fdump-ada-spec (gcc) :switch:`-fdump-ada-spec` Generate Ada spec files for the given header files transitively (including all header files that these headers depend upon). .. index:: -fdump-ada-spec-slim (gcc) :switch:`-fdump-ada-spec-slim` Only generate Ada spec files for the header files specified on the command line. .. index:: -fada-spec-parent (gcc) :switch:`-fada-spec-parent={unit}` Specifies that all files generated by :switch:`-fdump-ada-spec` are to be child units of the specified parent unit. .. index:: -C (gcc) :switch:`-C` Extract comments from headers and generate Ada comments in the Ada spec files. .. _Generating_C_Headers_for_Ada_Specifications: Generating C Headers for Ada Specifications ------------------------------------------- .. index:: Binding generation (for Ada specs) .. index:: C headers (binding generation) GNAT includes a C header generator for Ada specifications that supports Ada types that have a direct mapping to C types. This specifically includes support for: * Scalar types * Constrained arrays * Records (untagged) * Composition of the above types * Constant declarations * Object declarations * Subprogram declarations Running the C Header Generator ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The C header generator is part of the GNAT compiler and can be invoked via the :switch:`-gnatceg` switch, which generates a :file:`.h` file corresponding to the given input file (Ada spec or body). Note that only spec files are processed, so giving a spec or a body file as input is equivalent. For example: .. code-block:: sh $ gcc -c -gnatceg pack1.ads generates a self-contained file called :file:`pack1.h` including common definitions from the Ada Standard package followed by the definitions included in :file:`pack1.ads` as well as all the other units withed by this file. For instance, given the following Ada files: .. code-block:: ada package Pack2 is type Int is range 1 .. 10; end Pack2; .. code-block:: ada with Pack2; package Pack1 is type Rec is record Field1, Field2 : Pack2.Int; end record; Global : Rec := (1, 2); procedure Proc1 (R : Rec); procedure Proc2 (R : in out Rec); end Pack1; The above ``gcc`` command generates the following :file:`pack1.h` file: .. code-block:: c /* Standard definitions skipped */ #ifndef PACK2_ADS #define PACK2_ADS typedef short_short_integer pack2__TintB; typedef pack2__TintB pack2__int; #endif /* PACK2_ADS */ #ifndef PACK1_ADS #define PACK1_ADS typedef struct _pack1__rec { pack2__int field1; pack2__int field2; } pack1__rec; extern pack1__rec pack1__global; extern void pack1__proc1(const pack1__rec r); extern void pack1__proc2(pack1__rec *r); #endif /* PACK1_ADS */ You can then ``include`` :file:`pack1.h` from a C source file and use the types, call subprograms, reference objects, and constants. .. _GNAT_and_Other_Compilation_Models: GNAT and Other Compilation Models ================================= This section compares the GNAT model with the approaches taken in other environments: first the C/C++ model and then the mechanism that has been used in other Ada systems, in particular those traditionally used for Ada 83. .. _Comparison_between_GNAT_and_C/C++_Compilation_Models: Comparison between GNAT and C/C++ Compilation Models ---------------------------------------------------- The GNAT compilation model is close to the C and C++ models. You can think of Ada specs as corresponding to header files in C. As in C, you don't need to compile specs; they are compiled when they are used. The Ada |with| is similar in effect to the ``#include`` of a C header. One notable difference is that, in Ada, you may compile specs separately to check them for semantic and syntactic accuracy. This is not always possible with C headers because they are fragments of programs that have less specific syntactic or semantic rules. The other major difference is the requirement for running the binder, which performs two important functions. First, it checks for consistency. In C or C++, the only defense against assembling inconsistent programs lies outside the compiler, in a :file:`makefile`, for example. The binder satisfies the Ada requirement that it be impossible to construct an inconsistent program when the compiler is used in normal mode. .. index:: Elaboration order control The other important function of the binder is to deal with elaboration issues. There are also elaboration issues in C++ that are handled automatically. This automatic handling has the advantage of being simpler to use, but the C++ programmer has no control over elaboration. Where ``gnatbind`` might complain there was no valid order of elaboration, a C++ compiler would simply construct a program that malfunctioned at run time. .. _Comparison_between_GNAT_and_Conventional_Ada_Library_Models: Comparison between GNAT and Conventional Ada Library Models ----------------------------------------------------------- This section is intended for Ada programmers who have used an Ada compiler implementing the traditional Ada library model, as described in the Ada Reference Manual. .. index:: GNAT library In GNAT, there is no 'library' in the normal sense. Instead, the set of source files themselves acts as the library. Compiling Ada programs does not generate any centralized information, but rather an object file and a :file:`.ali` file, which are of interest only to the binder and linker. In a traditional system, the compiler reads information not only from the source file being compiled but also from the centralized library. This means that the effect of a compilation depends on what has been previously compiled. In particular: * When a unit is |withed|, the unit seen by the compiler corresponds to the version of the unit most recently compiled into the library. * Inlining is effective only if the necessary body has already been compiled into the library. * Compiling a unit may obsolete other units in the library. In GNAT, compiling one unit never affects the compilation of any other units because the compiler reads only source files. Only changes to source files can affect the results of a compilation. In particular: * When a unit is |withed|, the unit seen by the compiler corresponds to the source version of the unit that is currently accessible to the compiler. .. index:: Inlining * Inlining requires the appropriate source files for the package or subprogram bodies to be available to the compiler. Inlining is always effective, independent of the order in which units are compiled. * Compiling a unit never affects any other compilations. The editing of sources may cause previous compilations to be out of date if they depended on the source file being modified. The most important result of these differences is that order of compilation is never significant in GNAT. There is no situation in which you are required to do one compilation before another. What shows up as order of compilation requirements in the traditional Ada library becomes, in GNAT, simple source dependencies; in other words, there is only a set of rules saying what source files must be present when a file is compiled. .. _Using_GNAT_Files_with_External_Tools: Using GNAT Files with External Tools ==================================== This section explains how files that are produced by GNAT may be used with tools designed for other languages. .. _Using_Other_Utility_Programs_with_GNAT: Using Other Utility Programs with GNAT -------------------------------------- The object files generated by GNAT are in standard system format and, in particular, the debugging information uses this format. This means programs generated by GNAT can be used with existing utilities that depend on these formats. In general, any utility program that works with C will also often work with Ada programs generated by GNAT. This includes software utilities such as ``gprof`` (a profiling program), ``gdb`` (the FSF debugger), and utilities such as ``Purify``. .. _The_External_Symbol_Naming_Scheme_of_GNAT: The External Symbol Naming Scheme of GNAT ----------------------------------------- To interpret the output from GNAT when using tools that are originally intended for use with other languages, you need to understand the conventions used to generate link names from the Ada entity names. All link names are in all lowercase. With the exception of library procedure names, the mechanism used is simply to use the full expanded Ada name with dots replaced by double underscores. For example, suppose we have the following package spec: .. code-block:: ada package QRS is MN : Integer; end QRS; .. index:: pragma Export The variable ``MN`` has a full expanded Ada name of ``QRS.MN``, so the corresponding link name is ``qrs__mn``. Of course if you use a pragma ``Export``, you maye override this: .. code-block:: ada package Exports is Var1 : Integer; pragma Export (Var1, C, External_Name => "var1_name"); Var2 : Integer; pragma Export (Var2, C, Link_Name => "var2_link_name"); end Exports; In this case, the link name for ``Var1`` is whatever link name the C compiler would assign for the C function ``var1_name``. This typically would be either ``var1_name`` or ``_var1_name``, depending on operating system conventions, but other possibilities exist. The link name for ``Var2`` is ``var2_link_name``, and this is not operating system dependent. One exception occurs for library level procedures. A potential ambiguity arises between the required name ``_main`` for the C main program, and the name we would otherwise assign to an Ada library level procedure called ``Main`` (which might well not be the main program). To avoid this ambiguity, GNAT adds the prefix ``_ada_`` to such names. So if we have a library level procedure such as: .. code-block:: ada procedure Hello (S : String); the external name of this procedure is ``_ada_hello``. --- ### Gcc Doc Libgdiagnostics Topics Compatibility (`gcc/doc/libgdiagnostics/topics/compatibility.rst`) .. Copyright (C) 2015-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see . .. default-domain:: c ABI and API compatibility ========================= The libgdiagnostics developers strive for ABI and API backward-compatibility: programs built against libgdiagnostics.so stand a good chance of running without recompilation against newer versions of libgdiagnostics.so, and ought to recompile without modification against newer versions of libgdiagnostics.h. .. note:: The libgdiagnostics++.h C++ API is more experimental, and less locked-down at this time. API compatibility is achieved by extending the API rather than changing it. For ABI compatibility, we avoid bumping the SONAME, and instead use symbol versioning to tag each symbol, so that a binary linked against libgdiagnostics.so is tagged according to the symbols that it uses. For example, :func:`diagnostic_logical_location_get_kind` was added in ``LIBGDIAGNOSTICS_ABI_1``. If a client program uses it, this can be detected from metadata by using ``objdump``: .. code-block:: bash $ objdump -p testsuite/libgdiagnostics/test-logical-location.c.exe | tail -n 7 Version References: required from libc.so.6: 0x09691a75 0x00 04 GLIBC_2.2.5 required from libgdiagnostics.so.0: 0x0ec567d1 0x00 03 LIBGDIAGNOSTICS_ABI_1 0x0ec567d0 0x00 02 LIBGDIAGNOSTICS_ABI_0 You can see the symbol tags provided by libgdiagnostics.so using ``objdump``: .. code-block:: bash $ objdump -p libgdiagnostics.so | less [...snip...] Version definitions: 1 0x01 0x099ea4b0 libgdiagnostics.so.0 2 0x00 0x0ec567d0 LIBGDIAGNOSTICS_ABI_0 3 0x00 0x0ec567d1 LIBGDIAGNOSTICS_ABI_1 LIBGDIAGNOSTICS_ABI_0 [...snip...] ABI symbol tags *************** .. _LIBGDIAGNOSTICS_ABI_0: ``LIBGDIAGNOSTICS_ABI_0`` ------------------------- All entrypoints in the initial release of libgdiagnostics (in GCC 15) are tagged with ``LIBGDIAGNOSTICS_ABI_0``; these entrypoints are: * :func:`diagnostic_manager_new` * :func:`diagnostic_manager_release` * :func:`diagnostic_manager_set_tool_name` * :func:`diagnostic_manager_set_full_name` * :func:`diagnostic_manager_set_version_string` * :func:`diagnostic_manager_set_version_url` * :func:`diagnostic_manager_add_text_sink` * :func:`diagnostic_text_sink_set_source_printing_enabled` * :func:`diagnostic_text_sink_set_colorize` * :func:`diagnostic_text_sink_set_labelled_source_colorization_enabled` * :func:`diagnostic_manager_add_sarif_sink` * :func:`diagnostic_manager_write_patch` * :func:`diagnostic_manager_new_file` * :func:`diagnostic_file_set_buffered_content` * :func:`diagnostic_manager_debug_dump_file` * :func:`diagnostic_manager_new_location_from_file_and_line` * :func:`diagnostic_manager_new_location_from_file_line_column` * :func:`diagnostic_manager_new_location_from_range` * :func:`diagnostic_manager_debug_dump_location` * :func:`diagnostic_manager_new_logical_location` * :func:`diagnostic_manager_debug_dump_logical_location` * :func:`diagnostic_manager_begin_group` * :func:`diagnostic_manager_end_group` * :func:`diagnostic_begin` * :func:`diagnostic_set_cwe` * :func:`diagnostic_add_rule` * :func:`diagnostic_set_location` * :func:`diagnostic_set_location_with_label` * :func:`diagnostic_add_location` * :func:`diagnostic_add_location_with_label` * :func:`diagnostic_set_logical_location` * :func:`diagnostic_add_fix_it_hint_insert_before` * :func:`diagnostic_add_fix_it_hint_insert_after` * :func:`diagnostic_add_fix_it_hint_replace` * :func:`diagnostic_add_fix_it_hint_delete` * :func:`diagnostic_add_execution_path` * :func:`diagnostic_manager_new_execution_path` * :func:`diagnostic_take_execution_path` * :func:`diagnostic_execution_path_release` * :func:`diagnostic_execution_path_add_event` * :func:`diagnostic_execution_path_add_event_va` * :func:`diagnostic_finish` * :func:`diagnostic_finish_va` * :func:`diagnostic_physical_location_get_file` .. _LIBGDIAGNOSTICS_ABI_1: ``LIBGDIAGNOSTICS_ABI_1`` ------------------------- ``LIBGDIAGNOSTICS_ABI_1`` covers the addition of these functions for accessing values within a :type:`diagnostic_logical_location`: * :func:`diagnostic_logical_location_get_kind` * :func:`diagnostic_logical_location_get_parent` * :func:`diagnostic_logical_location_get_short_name` * :func:`diagnostic_logical_location_get_fully_qualified_name` * :func:`diagnostic_logical_location_get_decorated_name` .. _LIBGDIAGNOSTICS_ABI_2: ``LIBGDIAGNOSTICS_ABI_2`` ------------------------- ``LIBGDIAGNOSTICS_ABI_2`` covers the addition of these functions for supporting command-line options and SARIF playback: * :func:`diagnostic_manager_add_sink_from_spec` * :func:`diagnostic_manager_set_analysis_target` .. _LIBGDIAGNOSTICS_ABI_3: ``LIBGDIAGNOSTICS_ABI_3`` ------------------------- ``LIBGDIAGNOSTICS_ABI_3`` covers the addition of these functions for working with directed graphs: * :func:`diagnostic_manager_new_graph` * :func:`diagnostic_manager_take_global_graph` * :func:`diagnostic_take_graph` * :func:`diagnostic_graph_release` * :func:`diagnostic_graph_set_description` * :func:`diagnostic_graph_add_node` * :func:`diagnostic_graph_add_edge` * :func:`diagnostic_graph_get_node_by_id` * :func:`diagnostic_graph_get_edge_by_id` * :func:`diagnostic_node_set_label` * :func:`diagnostic_node_set_location` * :func:`diagnostic_node_set_logical_location` .. _LIBGDIAGNOSTICS_ABI_4: ``LIBGDIAGNOSTICS_ABI_4`` ------------------------- ``LIBGDIAGNOSTICS_ABI_4`` covers the addition of these functions for working with :type:`diagnostic_message_buffer`. * :func:`diagnostic_message_buffer_new` * :func:`diagnostic_message_buffer_release` * :func:`diagnostic_message_buffer_append_str` * :func:`diagnostic_message_buffer_append_text` * :func:`diagnostic_message_buffer_append_byte` * :func:`diagnostic_message_buffer_append_printf` * :func:`diagnostic_message_buffer_append_event_id` * :func:`diagnostic_message_buffer_begin_url` * :func:`diagnostic_message_buffer_end_url` * :func:`diagnostic_message_buffer_begin_quote` * :func:`diagnostic_message_buffer_end_quote` * :func:`diagnostic_message_buffer_begin_color` * :func:`diagnostic_message_buffer_end_color` * :func:`diagnostic_message_buffer_dump` * :func:`diagnostic_finish_via_msg_buf` * :func:`diagnostic_add_location_with_label_via_msg_buf` * :func:`diagnostic_execution_path_add_event_via_msg_buf` .. _LIBGDIAGNOSTICS_ABI_5: ``LIBGDIAGNOSTICS_ABI_5`` ------------------------- ``LIBGDIAGNOSTICS_ABI_5`` covers the addition of this function: * :func:`diagnostic_manager_set_debug_physical_locations` --- ### Gcc Doc Libgdiagnostics Topics Diagnostic Manager (`gcc/doc/libgdiagnostics/topics/diagnostic-manager.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see . .. default-domain:: c Diagnostic Managers =================== .. type:: diagnostic_manager; A :type:`diagnostic_manager` is an opaque bundle of state for a client of libgdiagnostics. It has zero of more "output sinks" to which diagnostics are emitted. Responsibilities include: * location-management * caching of source file content * patch generation .. function:: diagnostic_manager *diagnostic_manager_new (void) Create a new diagnostic_manager. The caller will need to call :func:`diagnostic_release_manager` on it at some point. .. note:: No output sinks are created by default; so you will want to create one with something like: .. code-block:: diagnostic_manager_add_text_sink (diag_mgr, stderr, DIAGNOSTIC_COLORIZE_IF_TTY); .. function:: void diagnostic_manager_release (diagnostic_manager *diag_mgr) Release a diagnostic_manager. This will flush output to all of the output sinks, and clean up. The parameter must be non-NULL. .. function:: int diagnostic_manager_add_sink_from_spec (diagnostic_manager *affected_mgr, \ const char *option_name, \ const char *spec, \ diagnostic_manager *control_mgr) This function can be used to support option processing similar to GCC's `-fdiagnostics-add-output= `_. This allows command-line tools to support the same domain-specific language for specifying output sinks as GCC does. The function will attempt to parse ``spec`` as if it were an argument to GCC's `-fdiagnostics-add-output= `_. If successful, it will add an output sink to ``affected_mgr`` and return zero. Otherwise, it will emit an error diagnostic to ``control_mgr`` and return non-zero. ``affected_mgr`` and ``control_mgr`` can be the same manager, or be different managers. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_2`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_manager_add_sink_from_spec .. function:: void diagnostic_manager_set_analysis_target (diagnostic_manager *mgr, \ const diagnostic_file *file) This function sets the "main input file" of ``mgr`` to be ``file``. This affects the :code:`` of generated HTML and the :code:`role` of the :code:`artifact` in SARIF output (`SARIF v2.1.0 section 3.24.6 <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790867>`_). This function was added in :ref:`LIBGDIAGNOSTICS_ABI_2`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_manager_set_analysis_target --- ### Gcc Doc Libgdiagnostics Topics Diagnostics (`gcc/doc/libgdiagnostics/topics/diagnostics.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Diagnostics =========== .. type:: diagnostic A :type:`diagnostic` is an opaque bundle of state for a particular diagnostic that is being constructed in memory. Lifecycle of a diagnostic ************************* Diagnostics are * *created* from a :type:`diagnostic_manager` by using :func:`diagnostic_begin`, then * *populated* with data, such as physical locations, logical locations, metadata, execution paths, or fix-it hints, then * *finished*, in which a formatting string and arguments are given, via a call to :func:`diagnostic_finish` or :func:`diagnostic_finish_va`. The :type:`diagnostic_manager` will emit the diagnostic to all of the manager's output sinks (either immediately, or at some later time, depending on the sink). Once a :type:`diagnostic` has had one of these "finish" functions called on it, it is freed, and is no longer valid for use. The formatting strings use their own syntax; see :doc:`message-formatting`. .. function:: diagnostic *diagnostic_begin (diagnostic_manager *diag_mgr, \ enum diagnostic_level level) Create a new :type:`diagnostic` associated with the given :type:`diagnostic_manager`. The parameter ``diag_mgr`` must be non-NULL. The parameter ``level`` describes the severity of the diagnostic. .. enum:: diagnostic_level This enum describes the severity of a particular diagnostic. .. macro:: DIAGNOSTIC_LEVEL_ERROR A problem sufficiently severe that the program cannot successfully complete, or where the input being analyzed is definitely wrong (e.g. malformed). .. macro:: DIAGNOSTIC_LEVEL_WARNING A problem where the input is technically correct, but is likely not what the user intended, such as common mistakes, or other unusual conditions that *may* indicate trouble, such as use of obsolete features. .. macro:: DIAGNOSTIC_LEVEL_NOTE A supplementary message added to another :type:`diagnostic`, giving extra information that may help the user understand it. .. macro:: DIAGNOSTIC_LEVEL_SORRY A problem where the input is valid, but the tool isn't able to handle it. .. function:: void diagnostic_finish (diagnostic *diag, const char *fmt, ...) Emit ``diag`` to all sinks of its manager, and release ``diag``. It is not valid to use ``diag`` after this call. Use parameter ``fmt`` for the message. Note that this uses gcc's pretty-print format, which is *not* printf. See :doc:`message-formatting`. Both ``diag`` and ``fmt`` must be non-NULL. TODO: who is responsible for putting FMT through gettext? .. function:: void diagnostic_finish_va (diagnostic *diag, const char *fmt, va_list *args) This is equivalent to :func:`diagnostic_finish`, but using a :type:`va_list` rather than directly taking variadic arguments. All three parameters must be non-NULL. .. function:: void diagnostic_finish_via_msg_buf (diagnostic *diag, \ diagnostic_message_buffer *msg_buf) This is equivalent to :func:`diagnostic_finish`, but using a message buffer rather than a format string and variadic arguments. ``diag`` and ``msg_buf`` must both be non-NULL. Calling this function transfers ownership of ``msg_buf`` to the diagnostic - do not call :func:`diagnostic_message_buffer_release` on it. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer Diagnostic groups ***************** See :doc:`the "adding notes" section of the tutorial <../tutorial/04-notes>` for an example of a diagnostic group. .. function:: void diagnostic_manager_begin_group (diagnostic_manager *diag_mgr) Begin a diagnostic group. All diagnostics emitted within ``diag_mgr`` after the first one will be treated as additional information relating to the initial diagnostic. The parameter ``diag_mgr`` must be non-NULL. .. function:: void diagnostic_manager_end_group (diagnostic_manager *diag_mgr) Finish a diagnostic group. The parameter ``diag_mgr`` must be non-NULL. --- ### Gcc Doc Libgdiagnostics Topics Execution Paths (`gcc/doc/libgdiagnostics/topics/execution-paths.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Execution paths =============== .. type:: diagnostic_execution_path A :type:`diagnostic` can optionally contain a :type:`diagnostic_execution_path` describing a path of execution through code. .. function:: diagnostic_execution_path * diagnostic_add_execution_path (diagnostic *diag) Create and borrow a pointer to an execution path for ``diag``. The path is automatically cleaned up when ``diag`` is finished. ``diag`` must be non-NULL. .. function:: diagnostic_execution_path * diagnostic_manager_new_execution_path (diagnostic_manager *diag_mgr) Create a new execution path. This is owned by the caller and must have either :func:`diagnostic_take_execution_path` or :func:`diagnostic_execution_path_release` called on it. ``diag_mgr`` must be non-NULL. .. function:: void diagnostic_take_execution_path (diagnostic *diag, diagnostic_execution_path *path) Set ``diag`` to use ``path`` as its execution path, taking ownership of ``path``. Both parameters must be non-NULL. .. function:: void diagnostic_execution_path_release (diagnostic_execution_path *path) Release ownership of ``path``, which must not have been taken by a diagnostic. .. type:: diagnostic_event_id A :type:`diagnostic_event_id` identifies a particular event within a :type:`diagnostic_execution_path` and can be used for expressing cross-references between events. In particular FIXME .. function:: diagnostic_event_id diagnostic_execution_path_add_event (diagnostic_execution_path *path, \ const diagnostic_physical_location *physical_loc, \ const diagnostic_logical_location *logical_loc, \ unsigned stack_depth, \ const char *fmt, ...) Append an event to the end of ``path``, which must be non-NULL. ``physical_loc`` can be NULL, or non-NULL to associate the event with a :type:`diagnostic_physical_location`. ``logical_loc`` can be NULL, or non-NULL to associate the event with a :type:`diagnostic_logical_location`. ``stack_depth`` is for use in interprocedural paths and identifies the depth of the stack at the event. Purely intraprocedural paths should use a stack depth of 1 for their events ``fmt`` must be non-NULL. See :doc:`message-formatting` for details of how to use it. .. function:: diagnostic_event_id diagnostic_execution_path_add_event_va (diagnostic_execution_path *path, \ const diagnostic_physical_location *physical_loc, \ const diagnostic_logical_location *logical_loc, \ unsigned stack_depth, \ const char *fmt, \ va_list *args) Equivalent to :func:`diagnostic_execution_path_add_event`, but using a :type:`va_list` rather than directly taking variadic arguments. .. function:: diagnostic_event_id diagnostic_execution_path_add_event_via_msg_buf (diagnostic_execution_path *path, \ const diagnostic_physical_location *physical_loc, \ const diagnostic_logical_location *logical_loc, \ unsigned stack_depth, \ diagnostic_message_buffer *msg_buf) This is equivalent to :func:`diagnostic_execution_path_add_event` but using a message buffer rather than a format string and variadic arguments. ``path`` and ``msg_buf`` must both be non-NULL. Calling this function transfers ownership of ``msg_buf`` to the path - do not call :func:`diagnostic_message_buffer_release` on it. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer Paths are printed to text sinks, and for SARIF sinks each path is added as a ``codeFlow`` object (see SARIF 2.1.0 `§3.36 codeFlow object <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790990>`_). --- ### Gcc Doc Libgdiagnostics Topics Fix It Hints (`gcc/doc/libgdiagnostics/topics/fix-it-hints.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Fix-it hints ============ Adding fix-it hints to a diagnostic *********************************** A :type:`diagnostic` can contain "fix-it hints", giving suggestions for the user on how to edit their code to fix a problem. These can be expressed as insertions, replacements, and removals of text. There is only limited support for newline characters in fix-it hints: only hints with newlines which insert an entire new line are permitted, inserting at the start of a line, and finishing with a newline (with no interior newline characters). Other attempts to add fix-it hints containing newline characters will fail. Similarly, attempts to delete or replace a range *affecting* multiple lines will fail. The API handles these failures gracefully, so that diagnostics can attempt to add fix-it hints without each needing extensive checking. Fix-it hints are printed to text sinks, and are emitted by SARIF sinks as ``fix`` objects (see SARIF 2.1.0 `§3.55 fix object <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141791131>`_). Fix-it hints within a :type:`diagnostic` are "atomic": if any hints can't be applied, none of them will be, and no fix-its hints will be displayed for that diagnostic. This implies that diagnostic messages need to be worded in such a way that they make sense whether or not the fix-it hints are displayed. All fix-it hints within one :type:`diagnostic` must affect the same :type:`diagnostic_file`. .. function:: void diagnostic_add_fix_it_hint_insert_before (diagnostic *diag, \ const diagnostic_physical_location *loc, \ const char *addition) Add a fix-it hint to ``diag`` suggesting the insertion of the string ``addition`` before ``LOC``. For example:: ptr = arr[0]; ^~~~~~ & This :type:`diagnostic` has a single location covering ``arr[0]``, with the caret at the start. It has a single insertion fix-it hint, inserting ``&`` before the start of ``loc``. .. function:: void diagnostic_add_fix_it_hint_insert_after (diagnostic *diag, \ const diagnostic_physical_location *loc, \ const char *addition) Add a fix-it hint to ``diag`` suggesting the insertion of the string ``addition`` after the end of ``LOC``. For example, in:: #define FN(ARG0, ARG1, ARG2) fn(ARG0, ARG1, ARG2) ^~~~ ^~~~ ^~~~ ( ) ( ) ( ) the :type:`diagnostic` has three physical locations, covering ``ARG0``, ``ARG1``, and ``ARG2``, and 6 insertion fix-it hints: each arg has a pair of insertion fix-it hints, suggesting wrapping them with parentheses: one a '(' inserted before, the other a ')' inserted after. .. function:: void diagnostic_add_fix_it_hint_replace (diagnostic *diag, \ const diagnostic_physical_location *loc, \ const char *replacement) Add a fix-it hint to ``diag`` suggesting the replacement of the text at ``LOC`` with the string ``replacement``. For example, in:: c = s.colour; ^~~~~~ color This :type:`diagnostic` has a single physical location covering ``colour``, and a single "replace" fix-it hint, covering the same range, suggesting replacing it with ``color``. .. function:: void diagnostic_add_fix_it_hint_delete (diagnostic *diag, \ const diagnostic_physical_location *loc) Add a fix-it hint to ``diag`` suggesting the deletion of the text at ``LOC``. For example, in:: struct s {int i};; ^ - This :type:`diagnostic` has a single physical location at the stray trailing semicolon, along with a single removal fix-it hint, covering the same location. Generating patches ****************** .. function:: void diagnostic_manager_write_patch (diagnostic_manager *diag_mgr, \ FILE *dst_stream) Write a patch to ``dst_stream`` consisting of the effect of all fix-it hints on all diagnostics that have been finished on ``diag_mgr``. Both parameters must be non-NULL. --- ### Gcc Doc Libgdiagnostics Topics Graphs (`gcc/doc/libgdiagnostics/topics/graphs.rst`) .. Copyright (C) 2025-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Graphs ====== .. type:: diagnostic_graph SARIF has support for capturing directed graphs (such as callgraphs and control flow graphs), both at the level of the run as a whole, and at the level of individual results. libgdiagnostics supports this with the following entrypoints, allowing directed graphs to be * created (with :func:`diagnostic_manager_new_graph`) * reported "globally" (with :func:`diagnostic_manager_take_global_graph`) * reported as part of a :type:`diagnostic` (with :func:`diagnostic_take_graph`), or * discarded (with :func:`diagnostic_graph_release`). .. function:: diagnostic_graph * diagnostic_manager_new_graph (diagnostic_manager *manager) Create a new directed graph. The resulting graph is owned by the caller and must have one of :func:`diagnostic_manager_take_global_graph`, :func:`diagnostic_take_graph`, or :func:`diagnostic_graph_release` called on it to avoid leaks. The parameter ``manager`` must be non-null. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. .. function:: void diagnostic_manager_take_global_graph (diagnostic_manager *manager, \ diagnostic_graph *graph) Report this graph "globally", taking ownership of it. This won't appear in text sinks, but in SARIF sinks the graph will be added to theRun.graphs (SARIF v2.1.0 3.14.20). Parameters ``manager`` and ``graph`` must both be non-null. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. .. function:: void diagnostic_take_graph (diagnostic *diag, \ diagnostic_graph *graph) Add this graph to ``diag``, transferring ownership of it to ``diag``. This won't appear in text sinks, but in SARIF sinks the graph will be added to theResult.graphs (SARIF v2.1.0 3.27.19). Parameters ``diag`` and ``graph`` must both be non-null. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. .. function:: void diagnostic_graph_release (diagnostic_graph *graph) Release ``graph`` which must still be owned by the caller i.e. it must *not* have had :func:`diagnostic_manager_take_global_graph` or :func:`diagnostic_take_graph` called on it. Parameters ``graph`` can be null. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. .. function:: void diagnostic_graph_set_description (diagnostic_graph *graph, \ const char *description) Set the description of ``graph`` for use in the value of the SARIF ``description`` property (SARIF v2.1.0 section 3.39.2). The parameter ``graph`` must be non-null. The parameter ``description`` can be null, for clearing any existing description. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. .. function:: diagnostic_node * diagnostic_graph_add_node (diagnostic_graph *graph, \ const char *node_id, \ diagnostic_node *parent_node) Create and add a new node within ``graph`` with the given `id``. The id must be unique within nodes in ``graph``. The parameters ``graph`` and ``id`` must be non-null. ``parent_node`` can be NULL (for a top-level node in the graph), or non-null for a child node, allowing for arbitrary nesting of nodes. The new node is owned by ``graph``. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. .. function:: diagnostic_edge * diagnostic_graph_add_edge (diagnostic_graph *graph, \ const char *edge_id, \ diagnostic_node *src_node, \ diagnostic_node *dst_node, \ const char *label) Create and add a new edge within ``graph``. The parameters ``graph``, ``src_node`` and ``dest_node`` must be non-null. If non-null, then ``edge_id`` must be unique within ``graph``; if ``edge_id`` is null then a unique id of the form "edge0", "edge1", etc will be used automatically. If non-null, then ``label`` will be used for the value of the SARIF ``label`` property (SARIF v2.1.0 section 3.41.3). The new edge is owned by ``graph``. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. .. function:: diagnostic_node *diagnostic_graph_get_node_by_id (diagnostic_graph *graph, \ const char *node_id) Get the node in ``graph`` with the given id, or null. The parameters ``graph`` and ``node_id`` must be non-null. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. .. function:: diagnostic_edge *diagnostic_graph_get_edge_by_id (diagnostic_graph *graph, \ const char *edge_id) Get the edge in ``graph`` with the given id, or null. The parameters ``graph`` and ``edge_id`` must be non-null. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. .. type:: diagnostic_node .. function:: void diagnostic_node_set_label (diagnostic_node *node, \ const char *label) Set the label of ``node`` for use in the value of the SARIF ``label`` property (SARIF v2.1.0 section 3.40.3). The parameter ``node`` must be non-null. The parameter ``label`` can be null, for clearing any existing label. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. .. function:: void diagnostic_node_set_location (diagnostic_node *node, \ const diagnostic_physical_location *loc) Set the physical location of ``node``, if any. The parameter ``node`` must be non-null. The parameter ``loc`` can be null, for clearing any existing location. If set, the value will be used by SARIF sinks within the ``location`` property (SARIF v2.1.0 section 3.40.4). This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. .. function:: void diagnostic_node_set_logical_location (diagnostic_node *node, \ const diagnostic_logical_location *logical_loc) Set the logical location of ``node``, if any. The parameter ``node`` must be non-null. The parameter ``logical_loc`` _can be null, for clearing any existing location. If set, the value will be used by SARIF sinks within the ``location`` property (SARIF v2.1.0 section 3.40.4). This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`. --- ### Gcc Doc Libgdiagnostics Topics Index (`gcc/doc/libgdiagnostics/topics/index.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Topic reference =============== .. toctree:: :maxdepth: 2 retrofitting.rst diagnostic-manager.rst diagnostics.rst message-formatting.rst message-buffers.rst physical-locations.rst logical-locations.rst metadata.rst fix-it-hints.rst execution-paths.rst text-output.rst sarif.rst graphs.rst ux.rst compatibility.rst --- ### Gcc Doc Libgdiagnostics Topics Logical Locations (`gcc/doc/libgdiagnostics/topics/logical-locations.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Logical locations ================= A "logical" location is a location expressed in terms of construct in a programming language, such as ``within function 'foo'`` (as opposed to a :doc:`"physical" location <physical-locations>`, which refers to a specific file, and line(s) and/or column(s)) Creating location information ***************************** .. type:: diagnostic_logical_location A :type:`diagnostic_logical_location` is an opaque type describing a "logical" source location .. function:: const diagnostic_logical_location * diagnostic_manager_new_logical_location (diagnostic_manager *diag_mgr, \ enum diagnostic_logical_location_kind_t kind, \ const diagnostic_logical_location *parent, \ const char *short_name, \ const char *fully_qualified_name, \ const char *decorated_name) Create a :type:`diagnostic_logical_location`. ``diag_mgr`` must be non-NULL. ``kind`` describes the kind of logical location: .. enum:: diagnostic_logical_location_kind_t This roughly corresponds to the ``kind`` property in SARIF v2.1.0 (`§3.33.7 <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790976>`_). Kinds within executable code: .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_FUNCTION .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_MEMBER .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_MODULE .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_NAMESPACE .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_TYPE .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_RETURN_TYPE .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_PARAMETER .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_VARIABLE Kinds within XML or HTML documents: .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_ELEMENT .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_ATTRIBUTE .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_TEXT .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_COMMENT .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_PROCESSING_INSTRUCTION .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_DTD .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_DECLARATION Kinds within JSON documents: .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_OBJECT .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_ARRAY .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_PROPERTY .. macro:: DIAGNOSTIC_LOGICAL_LOCATION_KIND_VALUE ``parent`` can be NULL; if non-NULL it can be used to express tree-like nesting of logical locations, such as in:: namespace foo { namespace bar { class baz { baz (); }; } } where a diagnostic within ``baz``'s constructor could be reported as being within ``foo::bar::baz::baz`` where the logical locations are two namespaces, a type, and a member, respectively. ``short_name`` can be NULL, or else a string suitable for use by the SARIF logicalLocation ``name`` property (SARIF v2.1.0 `§3.33.4 <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790973>`_). ``fully_qualified_name`` can be NULL or else a string suitable for use by the SARIF logicalLocation ``fullyQualifiedName`` property (SARIF v2.1.0 `§3.33.5 <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790974>`_). ``decorated_name`` can be NULL or else a string suitable for use by the SARIF logicalLocation ``decoratedName`` property (SARIF v2.1.0 `§3.33.6 <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790975>`_). Repeated calls to :func:`diagnostic_manager_new_logical_location` with "equal" input values on the same :type:`diagnostic_manager` will return the same instance of :type:`diagnostic_logical_location`. "Equal" here includes different string buffers that compare as equal with :func:`strcmp`. .. function:: void diagnostic_manager_debug_dump_logical_location (const diagnostic_manager *diag_mgr, \ const diagnostic_logical_location *loc, \ FILE *out) Write a representation of ``file`` to ``out``, for debugging. Both ``diag_mgr`` and ``out`` must be non-NULL. ``file`` may be NULL. TODO: example of output Associating diagnostics with locations ************************************** .. function:: void diagnostic_set_logical_location (diagnostic *diag, \ const diagnostic_logical_location *logical_loc) Set the logical location of ``diag``. ``diag`` must be non-NULL; ``logical_loc`` can be NULL. Accessors ********* The following functions can be used to access the data that was passed to a :type:`diagnostic_logical_location` when it was created. In each case, the ``loc`` parameter must be non-NULL. The return values will point at *copies* of the original buffer owned by the :type:`diagnostic_logical_location`, or be null. .. function:: enum diagnostic_logical_location_kind_t diagnostic_logical_location_get_kind (const diagnostic_logical_location *loc) .. function:: const diagnostic_logical_location *diagnostic_logical_location_get_parent (const diagnostic_logical_location *loc) .. function:: const char *diagnostic_logical_location_get_short_name (const diagnostic_logical_location *loc) .. function:: const char *diagnostic_logical_location_get_fully_qualified_name (const diagnostic_logical_location *loc) .. function:: const char *diagnostic_logical_location_get_decorated_name (const diagnostic_logical_location *loc) The above accessors were added in :ref:`LIBGDIAGNOSTICS_ABI_1`; you can test for their presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_LOGICAL_LOCATION_ACCESSORS --- ### Gcc Doc Libgdiagnostics Topics Message Buffers (`gcc/doc/libgdiagnostics/topics/message-buffers.rst`) .. Copyright (C) 2025-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Message buffers =============== .. type:: diagnostic_message_buffer A :type:`diagnostic_message_buffer` is a buffer into which text can be accumulated, before being used: * as the message of a diagnostic, using :func:`diagnostic_finish_via_msg_buf` * as the text of a label for a :type:`diagnostic_physical_location` using :func:`diagnostic_add_location_with_label_via_msg_buf` * as the text of an event within a :type:`diagnostic_execution_path` using :func:`diagnostic_execution_path_add_event_via_msg_buf` This is to allow more flexible creation of messages than a "format string plus variadic arguments" API. .. function:: diagnostic_message_buffer * diagnostic_message_buffer_new (void) This function creates a new :type:`diagnostic_message_buffer`. The caller is responsible for cleaning it up, either by handing it off to one of the API entrypoints that takes ownership of it (such as :func:`diagnostic_finish_via_msg_buf`), or by calling :func:`diagnostic_message_buffer_release` on it. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer .. function:: void diagnostic_message_buffer_release (diagnostic_message_buffer *msg_buf) This function releases ``msg_buf``. Typically you don't need to call this, but instead will pass the buffer to one of the API entrypoints that takes over ownership of it (such as :func:`diagnostic_finish_via_msg_buf`); calling it after this would lead to a double-free bug, as you no longer "own" the buffer. ``msg_buf`` must be non-NULL. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer .. function:: void diagnostic_message_buffer_append_str (diagnostic_message_buffer *msg_buf, \ const char *p) This function appends the null-terminated string ``p`` to the buffer. The string is assumed to be UTF-8 encoded. ``msg_buf`` and ``p`` must both be non-NULL. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer .. function:: void diagnostic_message_buffer_append_text (diagnostic_message_buffer *msg_buf, \ const char *p, \ size_t len) This function appends ``len`` bytes from ``p`` to the buffer. The bytes are assumed to be UTF-8 encoded. ``msg_buf`` and ``p`` must both be non-NULL. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer .. function:: void diagnostic_message_buffer_append_byte (diagnostic_message_buffer *msg_buf,\ char ch) This function appends ``ch`` to the buffer. This should be either ASCII, or part of UTF-8 encoded text. ``msg_buf`` must be non-NULL. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer .. function:: void diagnostic_message_buffer_append_printf (diagnostic_message_buffer *msg_buf, \ const char *fmt, ...) This function appends a formatted string to the buffer, using the formatting rules for ``printf``. The string is assumed to be UTF-8 encoded. ``msg_buf`` and ``fmt`` must both be non-NULL. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer .. function:: void diagnostic_message_buffer_append_event_id (diagnostic_message_buffer *msg_buf, \ diagnostic_event_id event_id) This function appends a :type:`diagnostic_event_id` to the buffer. ``msg_buf`` must be non-NULL. For text output, the event will be printed in the form ``(1)``. This is analogous to the :doc:`%@ message formatting code <message-formatting>`. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer Hyperlink support ***************** .. function:: void diagnostic_message_buffer_begin_url (diagnostic_message_buffer *msg_buf, \ const char *url) This function indicates the beginning of a run of text that should be associated with the given URL. The run of text should be closed with a matching call to :func:`diagnostic_message_buffer_end_url`. ``msg_buf`` and ``url`` must both be non-NULL. For text output in a suitably modern terminal, the run of text will be emitted as a clickable hyperlink to the URL. For SARIF sinks, the run of text will be emitted using SARIF's embedded link syntax. This is analogous to the :doc:`%{ message formatting code <message-formatting>`. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer .. function:: void diagnostic_message_buffer_end_url (diagnostic_message_buffer *msg_buf) This function ends a run of text within the buffer started with :func:`diagnostic_message_buffer_begin_url`. ``msg_buf`` must be non-NULL. This is analogous to the :doc:`%} message formatting code <message-formatting>`. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer Quoted text *********** .. function:: void diagnostic_message_buffer_begin_quote (diagnostic_message_buffer *msg_buf) This function indicates the beginning of a run of text that should be printed in quotes. The run of text should be closed with a matching call to :func:`diagnostic_message_buffer_end_quote`. ``msg_buf`` must be non-NULL. For text output in a suitably modern terminal, the run of text will appear in bold. be emitted as a clickable hyperlink to the URL. For SARIF sinks, the run of text will be emitted using SARIF's embedded link syntax. This is analogous to the ``%<``:doc:`message formatting code <message-formatting>`. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer .. function:: void diagnostic_message_buffer_end_quote (diagnostic_message_buffer *msg_buf) This function ends a run of text started with :func:`diagnostic_message_buffer_begin_quote`. ``msg_buf`` must be non-NULL. This is analogous to the :doc:`%> message formatting code <message-formatting>`. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer Color ***** .. function:: void diagnostic_message_buffer_begin_color (diagnostic_message_buffer *msg_buf, \ const char *color) This function indicates the beginning of a run of text that should be colorized as the given color. The run of text should be closed with a matching call to :func:`diagnostic_message_buffer_end_color`. The precise set of available color names is currently undocumented. ``msg_buf`` and ``color`` must both be non-NULL. For text output in a suitable terminal, the run of text will be colorized. For SARIF sinks, the run of text will be emitted using SARIF's embedded link syntax. This is analogous to the :doc:`%r message formatting code <message-formatting>`. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer .. function:: void diagnostic_message_buffer_end_color (diagnostic_message_buffer *msg_buf) This function ends a run of text within the buffer started with :func:`diagnostic_message_buffer_begin_color`. ``msg_buf`` must be non-NULL. This is analogous to the :doc:`%R message formatting code <message-formatting>`. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer Debugging a message buffer ************************** .. function:: void diagnostic_message_buffer_dump (const diagnostic_message_buffer *msg_buf, \ FILE *outf) This function writes a representation of the contents of ``msg_buf`` to ``outf``, for debugging. ``msg_buf`` can be NULL or non-NULL. ``outf`` must be non-NULL. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer --- ### Gcc Doc Libgdiagnostics Topics Message Formatting (`gcc/doc/libgdiagnostics/topics/message-formatting.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Message formatting ================== Various libgdiagnostics entrypoints take a format string and variadic arguments. .. note:: See also :type:`diagnostic_message_buffer`, which offers an alternative way to build up messages. The format strings take codes prefixed by ``%``, or ``%q`` to put the result in quotes. For example:: "hello %s", "world" would print:: hello world whereas:: "hello %qs", "world" would print:: hello `world' where ```world'`` would be displayed in bold if colorization were enabled in the terminal. The following format specifiers are accepted: Numbers ******* ``d`` and ``i`` (``signed int``), ``u`` (``unsigned int``) ``%d``, ``%i``, and ``%u`` print integers in base ten. For example:: "the answer is %i", 42 would print:: the answer is 42 ``o`` (``unsigned int``) Print the integer in base eight ``x`` (``unsigned int``) Print the integer in base sixteen The above can be prefixed with ``l`` and ``ll`` prefixes to take ``long`` and ``long long`` values of the appropriate signedness. For example:: "address: %lx", (unsigned long)0x108b516 would print:: address: 108b516 Similarly, the prefix ``z`` can be used for ``size_t``:: "size: %zd", sizeof(struct foo) size: 32 and ``t`` for ptrdiff_t. ``f`` (``double``) ``%f`` prints a floating-point value. For example:: "value: %f", 1.0 might print:: value: 1.000000 Strings ******* ``c`` (``char``) ``%c`` prints a single character. ``s`` (``const char *``) ``%s`` prints a string. Note that if the string refers to something that might appear in the input file (such as the name of a function), it's better to quote the value; for example:: "unrecognized identifier: %qs", "foo" might print:: unrecognized identifier: `foo' ``m`` (no argument) Prints ``strerror(errno)``, for example:: "can't open %qs: %m" might print:: can't open `foo.txt': No such file or directory ``%`` (no argument) ``%%`` prints a `%` character, for example:: "8%% of 75 is 75%% of 8, and is thus 6" prints:: 8% of 75 is 75% of 8, and is thus 6 ``'`` (no argument) ``%'`` prints an apostrophe. This should only be used in untranslated messages; translations should use appropriate punctuation directly. Other format specifiers *********************** ``p`` (pointer) ``%p`` prints a pointer, although the precise format is implementation-defined. ``r`` (``const char *``) ``%r`` starts colorization on suitable text sinks, where the argument specifies the name of the kind of entity to be colored, such as ``error``. ``R`` (no argument) ``%R`` stops colorization ``<`` and ``>`` (no arguments) ``%<`` adds an opening quote and ``%>`` a closing quote, such as:: "missing element %<%s:%s%>", ns, name which might be printed as:: missing element `xhtml:head' If the thing to be quoted can be handled with another format specifier, then it's simpler to use ``q`` with it. For example, it's much simpler to print a ``const char *`` in quotes via:: "%qs", str rather than the error-prone:: "%<%s%>", str ``{`` (``const char *``) ``%{`` starts a link; the argument is the URL. This will be displayed in a suitably-capable terminal if a text sink is directly connected to a tty, and will be captured in SARIF output. ``}`` (no argument) ``%}`` stops a link started with ``%{``. For example:: "for more information see %{the documentation%}", "https://example.com" would be printed as:: for more information see the documentation with the URL emitted in suitable output sinks. ``@`` (``diagnostic_event_id *``) ``%@`` prints a reference to an event in a :type:`diagnostic_execution_path`, where the :type:`diagnostic_event_id` is passed by pointer. For example, if ``event_id`` refers to the first event in a path, then:: "double-%qs of %qs; first %qs was at %@", function, ptr, function, &event_id might print:: double-`free' of `p'; first `free` was at (1) .. : TODO: %.*s: a substring the length of which is specified by an argument integer. %Ns: likewise, but length specified as constant in the format string. %Z: Requires two arguments - array of int, and len. Prints elements of the array. %e: Consumes a pp_element * argument. Arguments can be used sequentially, or through %N$ resp. *N$ notation Nth argument after the format string. If %N$ / *N$ notation is used, it must be used for all arguments, except %m, %%, %<, %>, %} and %', which may not have a number, as they do not consume an argument. When %M$.*N$s is used, M must be N + 1. (This may also be written %M$.*s, provided N is not otherwise used.) The format string must have conversion specifiers with argument numbers 1 up to highest argument; each argument may only be used once. A format string can have at most 30 arguments. */ --- ### Gcc Doc Libgdiagnostics Topics Metadata (`gcc/doc/libgdiagnostics/topics/metadata.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Adding metadata =============== Tool metadata ************* It's possible to set up various metadata on the :type:`diagnostic_manager` as a whole, describing the program creating the diagnostics. .. note:: It's not required to set up any of this up on a :type:`diagnostic_manager`. However, if you are doing :doc:`SARIF output <sarif>`, then you need to at least call :func:`diagnostic_manager_set_tool_name` or the generated ``.sarif`` file will not validate against the schema. .. function:: void diagnostic_manager_set_tool_name (diagnostic_manager *diag_mgr, \ const char *value) Set a string for the name of the tool emitting the diagnostics. Both parameters must be non-NULL. If set, this string will be used * by :doc:`text output sinks <text-output>` as a prefix for output when no physical location is available, replacing ``progname`` in the following: .. code-block:: console $ ./tut01-hello-world progname: error: I'm sorry Dave, I'm afraid I can't do that * by :doc:`SARIF output sinks <sarif>` as the value for the ``name`` property of the ``driver`` (`SARIF v2.1.0 §3.19.8 <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790791>`_). .. function:: void diagnostic_manager_set_full_name (diagnostic_manager *diag_mgr, \ const char *value) Set a string giving the name of the tool along with the its version and other useful information:: diagnostic_manager_set_full_name (diag_mgr, "FooChecker 0.1 (en_US)"); If set, this string will be used by :doc:`SARIF output sinks <sarif>` as the value for the ``fullName`` property of the ``driver`` (`SARIF v2.1.0 §3.19.9 <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790792>`_). Both parameters must be non-NULL. .. function:: void diagnostic_manager_set_version_string (diagnostic_manager *diag_mgr, \ const char *value) Set a string suitable for use as the value of the SARIF ``version`` property of the ``driver``. (`SARIF v2.1.0 §3.19.13 <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790796>`_):: diagnostic_manager_set_version_string (diag_mgr, "0.1"); Both parameters must be non-NULL. .. function:: void diagnostic_manager_set_version_url (diagnostic_manager *diag_mgr, \ const char *value) Set a string suitable for use as the value of the SARIF ``informationUri`` property of the ``driver``. (`SARIF v2.1.0 §3.19.17 <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790800>`_):: diagnostic_manager_set_version_url (diag_mgr, "https://www.example.com/foo-checker/releases/0.1/"); Both parameters must be non-NULL. Adding metadata to a diagnostic ******************************* .. function:: void diagnostic_set_cwe (diagnostic *diag, \ unsigned cwe_id) Associate ``diag`` with the given ID within the `Common Weakness Enumeration <https://cwe.mitre.org/>`_:: /* CWE-242: Use of Inherently Dangerous Function. */ diagnostic_set_cwe (d, 242); ``diag`` must be non-NULL. The CWE value will be printed by text sinks after the message:: test-metadata.c:21:3: warning: never use 'gets' [CWE-242] and in a sufficiently-capable terminal will be a link to documentation about the CWE. .. function:: void diagnostic_add_rule (diagnostic *diag, \ const char *title, \ const char *url) Associate this :type:`diagnostic` with a particular rule that has been violated (such as in a coding standard, or within a specification). A diagnostic can be associated with zero or more rules. ``diag`` must be non-NULL. The rule must have at least one of a title and a URL, but these can be NULL. For example, given:: diagnostic_add_rule (d, "MSC24-C", "https://wiki.sei.cmu.edu/confluence/display/c/MSC24-C.+Do+not+use+deprecated+or+obsolescent+functions"); the rule name will be printed by text sinks after the message:: test-metadata.c:21:3: warning: never use 'gets' [MSC24-C] 21 | gets (buf); | ^~~~~~~~~~ and if so, the URL will be available in a sufficiently capable terminal. This can be used in conjunction with :func:`diagnostic_set_cwe`, giving output like this:: test-metadata.c:21:3: warning: never use 'gets' [CWE-242] [MSC24-C] 21 | gets (buf); | ^~~~~~~~~~ --- ### Gcc Doc Libgdiagnostics Topics Physical Locations (`gcc/doc/libgdiagnostics/topics/physical-locations.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Physical locations ================== A "physical" source location is a location expressed in terms of a specific file, and line(s) and column(s) (as opposed to a :doc:`"logical" location <logical-locations>`, which refers to semantic constructs in a programming language). Creating location information ***************************** The :type:`diagnostic_manager` manages various objects relating to locations. .. type:: diagnostic_file A :type:`diagnostic_file` is an opaque type describing a particular input file. .. function:: diagnostic_file * diagnostic_manager_new_file (diagnostic_manager *diag_mgr, \ const char *name, \ const char *sarif_source_language) Create a new :type:`diagnostic_file` for file ``name``. Repeated calls with strings that match ``name`` will return the same object. Both ``diag_mgr`` and ``name`` must be non-NULL. If ``sarif_source_language`` is non-NULL, it specifies a ``sourceLanguage`` value for the file for use when writing :doc:`SARIF <sarif>` (`SARIF v2.1.0 §3.24.10 <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790871>`_). See `SARIF v2.1.0 Appendix J <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141791197>`_ for suggested values for various programming languages. For example, this creates a :type:`diagnostic_file` for ``foo.c`` and identifies it as C source code:: foo_c = diagnostic_manager_new_file (diag_mgr, "foo.c", "c" /* source_language */); .. function:: void diagnostic_manager_debug_dump_file (diagnostic_manager *diag_mgr, \ const diagnostic_file *file, \ FILE *out) Write a representation of ``file`` to ``out``, for debugging. Both ``diag_mgr`` and ``out`` must be non-NULL. `file`` may be NULL. For example:: diagnostic_manager_debug_dump_file (diag_mgr, foo_c, stderr); might lead to this output:: file(name="foo.c", sarif_source_language="c") .. type:: diagnostic_line_num_t A :type:`diagnostic_line_num_t` is used for representing line numbers within text files. libgdiagnostics treats the first line of a text file as line 1. .. type:: diagnostic_column_num_t A :type:`diagnostic_column_num_t` is used for representing column numbers within text files. libgdiagnostics treats the first column of a text line as column 1, **not** column 0. .. note:: Both libgdiagnostics and Emacs number source *lines* starting at 1, but they have differing conventions for *columns*. libgdiagnostics uses a 1-based convention for source columns, whereas Emacs's ``M-x column-number-mode`` uses a 0-based convention. For example, an error in the initial, left-hand column of source line 3 is reported by libgdiagnostics as:: some-file.c:3:1: error: ...etc... On navigating to the location of that error in Emacs (e.g. via ``next-error``), the locus is reported in the Mode Line (assuming ``M-x column-number-mode``) as:: some-file.c 10% (3, 0) i.e. ``3:1:`` in libgdiagnostics corresponds to ``(3, 0)`` in Emacs. .. type:: diagnostic_physical_location A :type:`diagnostic_physical_location` is an opaque type representing a key into a database of source locations within a :type:`diagnostic_manager`. :type:`diagnostic_physical_location` instances are created by various API calls into the :type:`diagnostic_manager` expressing source code points and ranges. They persist until the :type:`diagnostic_manager` is released, which cleans them up. A ``NULL`` value means "unknown", and can be returned by the :type:`diagnostic_manager` as a fallback when a problem occurs (e.g. too many locations). A :type:`diagnostic_physical_location` can be a single point within the source code, such as here (at the the '"' at the start of the string literal):: int i = "foo"; ^ or be a range with a start and finish, and a "caret" location:: a = (foo && bar) ~~~~~^~~~~~~ where the caret here is at the first "&", and the start and finish are at the parentheses. .. function:: const diagnostic_physical_location *diagnostic_manager_new_location_from_file_and_line (diagnostic_manager *diag_mgr, \ const diagnostic_file *file, \ diagnostic_line_num_t line_num) Attempt to create a :type:`diagnostic_physical_location` representing ``FILENAME:LINE_NUM``, with no column information (thus representing the whole of the given line. Both ``diag_mgr`` and ``file`` must be non-NULL. .. function:: const diagnostic_physical_location * diagnostic_manager_new_location_from_file_line_column (diagnostic_manager *diag_mgr, \ const diagnostic_file *file, \ diagnostic_line_num_t line_num, \ diagnostic_column_num_t column_num) Attempt to create a :type:`diagnostic_physical_location` for ``FILENAME:LINE_NUM:COLUMN_NUM`` representing a particular point in the source file. Both ``diag_mgr`` and ``file`` must be non-NULL. .. function:: const diagnostic_physical_location *diagnostic_manager_new_location_from_range (diagnostic_manager *diag_mgr,\ const diagnostic_physical_location *loc_caret,\ const diagnostic_physical_location *loc_start,\ const diagnostic_physical_location *loc_end) Attempt to create a diagnostic_physical_location representing a range within a source file, with a highlighted "caret" location. All must be within the same file, but they can be on different lines. For example, consider the location of the binary expression below:: ...|__________1111111112222222 ...|12345678901234567890123456 ...| 521|int sum (int foo, int bar) 522|{ 523| return foo + bar; ...| ~~~~^~~~~ 524|} The location's caret is at the "+", line 523 column 15, but starts earlier, at the "f" of "foo" at column 11. The finish is at the "r" of "bar" at column 19. ``diag_mgr`` must be non-NULL. .. function:: void diagnostic_manager_debug_dump_location (const diagnostic_manager *diag_mgr,\ const diagnostic_physical_location *loc, \ FILE *out) Write a representation of ``loc`` to ``out``, for debugging. Both ``diag_mgr`` and ``out`` must be non-NULL. `loc`` may be NULL. TODO: example of output .. function:: diagnostic_file *diagnostic_physical_location_get_file (const diagnostic_physical_location *physical_loc) Get the :type:`diagnostic_file` associated with a given physical location. Associating diagnostics with locations ************************************** A :type:`diagnostic` has an optional primary physical location and zero or more secondary physical locations. For example:: a = (foo && bar) ~~~~~^~~~~~~ This diagnostic has a single :type:`diagnostic_physical_location`, with the caret at the first "&", and the start/finish at the parentheses. Contrast with:: a = (foo && bar) ~~~ ^~ ~~~ This diagnostic has three locations * The primary location (at "&&") has its caret and start location at the first "&" and end at the second "&. * The secondary location for "foo" has its start and finish at the "f" and "o" of "foo"; the caret is not displayed, but is perhaps at the "f" of "foo". * Similarly, the other secondary location (for "bar") has its start and finish at the "b" and "r" of "bar"; the caret is not displayed, but is perhaps at the"b" of "bar". .. function:: void diagnostic_set_location (diagnostic *diag, \ const diagnostic_physical_location * loc) Set the primary location of ``diag``. ``diag`` must be non-NULL; ``loc`` can be NULL. .. function:: void diagnostic_set_location_with_label (diagnostic *diag, \ const diagnostic_physical_location *loc, \ const char *fmt, ...) Set the primary location of ``diag``, with a label. The label is formatted as per the rules FIXME ``diag`` and ``fmt`` must be non-NULL; ``loc`` can be NULL. See :doc:`message-formatting` for details of how to use ``fmt``. TODO: example of use .. function:: void diagnostic_add_location (diagnostic *diag, \ const diagnostic_physical_location * loc) Add a secondary location to ``diag``. ``diag`` must be non-NULL; ``loc`` can be NULL. .. function:: void diagnostic_add_location_with_label (diagnostic *diag, \ const diagnostic_physical_location *loc, \ const char *text) Add a secondary location to ``diag``, with a label. The label is formatted as per the rules FIXME ``diag`` and ``fmt`` must be non-NULL; ``loc`` can be NULL. For example, .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-labelled-ranges.c :language: c :start-after: /* begin quoted source */ :end-before: /* end quoted source */ might give this text output:: test-labelled-ranges.c:9:6: error: mismatching types: 'int' and 'const char *' 19 | 42 + "foo" | ~~ ^ ~~~~~ | | | | int const char * .. function:: void diagnostic_add_location_with_label_via_msg_buf (diagnostic *diag, \ const diagnostic_physical_location *loc, \ diagnostic_message_buffer *msg_buf) This is equivalent to :func:`diagnostic_add_location_with_label` but using a message buffer rather than a text string. ``diag`` and ``msg_buf`` must both be non-NULL. Calling this function transfers ownership of ``msg_buf`` to the diagnostic - do not call :func:`diagnostic_message_buffer_release` on it. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_3`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_message_buffer .. function:: void diagnostic_manager_set_debug_physical_locations (diagnostic_manager *mgr, \ int value) Calling ``diagnostic_manager_set_debug_physical_locations (mgr, 1);`` will lead to debugging information being printed to ``stderr`` when creating :type:`diagnostic_physical_location` instances. The precise format of these messages is subject to change. This function was added in :ref:`LIBGDIAGNOSTICS_ABI_5`; you can test for its presence using .. code-block:: c #ifdef LIBDIAGNOSTICS_HAVE_diagnostic_manager_set_debug_physical_locations --- ### Gcc Doc Libgdiagnostics Topics Retrofitting (`gcc/doc/libgdiagnostics/topics/retrofitting.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Adding libgdiagnostics to an existing project ============================================= TODO --- ### Gcc Doc Libgdiagnostics Topics Sarif (`gcc/doc/libgdiagnostics/topics/sarif.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c SARIF support ============= `SARIF <https://sarif.info>`_ is a machine-readable format, originally designed for the output of static analysis tools, but which can be used for diagnostics in general. .. function:: void diagnostic_manager_add_sarif_sink (diagnostic_manager *diag_mgr, \ FILE *dst_stream, \ const diagnostic_file *main_input_file, \ enum diagnostic_sarif_version version) Add a new output sink to ``diag_mgr``, which writes SARIF of the given version to ``dst_stream``. The output is not written until ``diag_mgr`` is released. ``dst_stream`` is borrowed, and must outlive ``diag_mgr``. For the result to be a valid SARIF file according to the schema, ``diag_mgr`` must have had :func:`diagnostic_manager_set_tool_name` called on it. ``diag_mgr``, ``dst_stream``, and ``main_input_file`` must all be non-NULL. .. enum:: diagnostic_sarif_version An enum for choosing the SARIF version for a SARIF output sink. .. macro:: DIAGNOSTIC_SARIF_VERSION_2_1_0 .. macro:: DIAGNOSTIC_SARIF_VERSION_2_2_PRERELEASE --- ### Gcc Doc Libgdiagnostics Topics Text Output (`gcc/doc/libgdiagnostics/topics/text-output.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Text output =========== .. type:: diagnostic_text_sink .. function:: diagnostic_text_sink * diagnostic_manager_add_text_sink (diagnostic_manager *diag_mgr,\ FILE *dst_stream, \ enum diagnostic_colorize colorize) Add a new output sink to ``diag_mgr``, which writes GCC-style diagnostics to ``dst_stream``. Return a borrowed pointer to the sink, which is cleaned up when ``diag_mgr`` is released. ``diag_mgr`` must be non-NULL. ``dst_stream`` must be non-NULL. It is borrowed and must outlive ``DIAG_MGR``. The output for each diagnostic is written and flushed as each :type:`diagnostic` is finished. .. enum:: diagnostic_colorize An enum for determining if we should colorize a text output sink. .. macro:: DIAGNOSTIC_COLORIZE_IF_TTY Diagnostics should be colorized if the destination stream is directly connected to a tty. .. macro:: DIAGNOSTIC_COLORIZE_NO Diagnostics should not be colorized. .. macro:: DIAGNOSTIC_COLORIZE_YES Diagnostics should be colorized. .. function:: void diagnostic_text_sink_set_source_printing_enabled (diagnostic_text_sink *text_sink, \ int value) Enable or disable printing of source text in the text sink. ``text_sink`` must be non-NULL. Default: enabled. .. function:: void diagnostic_text_sink_set_colorize (diagnostic_text_sink *text_sink, \ enum diagnostic_colorize colorize) Update colorization of text sink. ``text_sink`` must be non-NULL. .. function:: void diagnostic_text_sink_set_labelled_source_colorization_enabled (diagnostic_text_sink *text_sink, \ int value) ``text_sink`` must be non-NULL. Enable or disable colorization of the characters of source text that are underlined. This should be true for clients that generate range information (so that the ranges of code are colorized), and false for clients that merely specify points within the source code (to avoid e.g. colorizing just the first character in a token, which would look strange). Default: enabled. --- ### Gcc Doc Libgdiagnostics Topics Ux (`gcc/doc/libgdiagnostics/topics/ux.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c User Experience =============== Refer to `GCC's user experience guidelines <https://gcc.gnu.org/onlinedocs/gccint/User-Experience-Guidelines.html>`_ for notes on `what makes a good diagnostic <https://gcc.gnu.org/onlinedocs/gccint/Guidelines-for-Diagnostics.html>`_. --- ### Gcc Doc Libgdiagnostics Tutorial 01 Hello World (`gcc/doc/libgdiagnostics/tutorial/01-hello-world.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Tutorial part 1: "Hello world" ============================== Before we look at the details of the API, let's look at building and running programs that use the library. Here's a toy program that uses libgdiagnostics to emit an error message to stderr. .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-example-1.c :language: c :start-after: /* begin quoted source */ :end-before: /* end quoted source */ Copy the above to `tut01-hello-world.c`. Assuming you have libgdiagnostics installed, build the test program using: .. code-block:: console $ gcc \ tut01-hello-world.c \ -o tut01-hello-world \ -lgdiagnostics You should then be able to run the built program: .. code-block:: console $ ./tut01-hello-world progname: error: I'm sorry Dave, I'm afraid I can't do that If stderr is connected to a terminal, you should get colorized output (using `SGR control codes <https://en.wikipedia.org/wiki/ANSI_escape_code>`_). .. image:: example-1.png Otherwise, the output will be plain text. Obviously a trivial example like the above could be done using ``fprintf`` on stderr, and it's fairly easy to colorize text at the terminal. In :doc:`the next part of the tutorial <02-physical-locations>` we'll add file/location information to our error messages, and libgdiagnostics will quote the pertinent parts of the file, underlining them, which is less trivial to reimplement. libgdiagnostics gives us many other such abilities, such as fix-it hints and execution paths, which we'll cover in the following tutorials. Also, once a program's diagnostics are using libgdiagnostics, it is trivial to add support for outputting them in machine-readable form as :doc:`SARIF <../topics/sarif>`. Structure ********* The above example shows the typical structure of a program using libgdiagnostics: * **initialization**: create a :type:`diagnostic_manager` instance, and create an output sink for it, and other one-time initialization * **emission**: create various :type:`diagnostic` instances, populating them with data, and calling "finish" once they're ready to be emitted. :doc:`Text sinks <../topics/text-output>` emit their diagnostics as soon as "finish" is called on them. * **cleanup**: call :func:`diagnostic_manager_release` on the :type:`diagnostic_manager` to finish and free up resources. :doc:`SARIF sinks <../topics/sarif>` write their output when :func:`diagnostic_manager_release` is called on the manager. For non-trivial examples we'll also want to create location information, which could happen during initialization, or during a parsing phase of the program using libgdiagnostics. See :doc:`02-physical-locations` for more information. Formatted messages ****************** The above example uses :func:`diagnostic_finish`, which takes a format string and arguments. libgdiagnostics has its own style of format string arguments used for :func:`diagnostic_finish` and some other entrypoints. .. note:: The format syntax is *not* the same as ``printf``; see :doc:`supported formatting options <../topics/message-formatting>`. You can use the ``q`` modifier on arguments to quote them, so, for example ``%qs`` is a quoted string, consuming a ``const char *`` argument:: diagnostic_finish (d, "can't find %qs", "foo"); This gives output like this: .. code-block:: console progname: error: can't find ‘foo’ where the quoted string will appear in bold in a suitably-capable terminal, and the quotes will be internationalized, so that e.g. with ``LANG=fr_FR.UTF8`` we might get: .. code-block:: console progname: erreur: can't find « foo » Note that: * the string ``error`` has been localized by libgdiagnostics to ``erreur``, * locale-specific quoting has been used (``«`` and ``»`` rather than ``‘`` and ``’``), * ``foo`` hasn't been localized - you would typically use quoted strings for referring to identifiers in the input language (such as function names in code, property names in JSON, etc), * the message itself hasn't been localized: you are responsible for passing a translated format string to :func:`diagnostic_finish` if you want to internationalize the output. There are many :doc:`supported formatting options <../topics/message-formatting>`. Naming the program ****************** In the above output the message was preceded with ``progname``. This appears for diagnostics that don't have any location information associated with them. We'll look at setting up location information in the :doc:`next tutorial <02-physical-locations>`, but we can override this default name via :func:`diagnostic_manager_set_tool_name`:: diagnostic_manager_set_tool_name (diag_mgr, "my-awesome-checker"); leading to output like this:: my-awesome-checker: error: can't find ‘foo’ There are various other functions for :doc:`supplying metadata to libgdiagnostics <../../topics/metadata>`. Moving beyond trivial examples ****************************** Obviously it's not very useful if we can't refer to specific files and specific locations in those files in our diagnostics, so read :doc:`part 2 of the tutorial <02-physical-locations>` for information on how to do this. --- ### Gcc Doc Libgdiagnostics Tutorial 02 Physical Locations (`gcc/doc/libgdiagnostics/tutorial/02-physical-locations.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Tutorial part 2: physical locations =================================== libgdiagnostics has two kinds of location: * *physical locations* expressed in terms of a specific file, and line(s) and perhaps column(s), such as ``some-file.c:3:1``, or a range of columns, such as in:: test-typo.c:19:13: error: unknown field 'colour' 19 | return p->colour; | ^~~~~~ or even a range spanning multiple lines of a file. All of these are instances of :type:`diagnostic_physical_location`. * *logical locations* which refers to semantic constructs in the input, such as ``within function 'foo'``, or within namespace ``foo``'s class ``bar``'s member function ``get_color``. These are instances of :type:`diagnostic_logical_location`, A :type:`diagnostic` can have zero or more physical locations, and optionally have a logical location. Let's extend the previous example to add a physical location to the :type:`diagnostic`; we'll cover logical locations in the :doc:`next section <03-logical-locations>`. Source files ************ Given these declarations:: static diagnostic_manager *diag_mgr; static diagnostic_file *main_file; we can create a :type:`diagnostic_file` describing an input file ``foo.c`` via :func:`diagnostic_manager_new_file`:: foo_c = diagnostic_manager_new_file (diag_mgr, "foo.c", "c" /* source_language */); You can use :func:`diagnostic_manager_debug_dump_file` to print a representation of a :type:`diagnostic_file` for debugging. For example:: diagnostic_manager_debug_dump_file (diag_mgr, foo_c, stderr); might lead to this output on ``stderr``:: file(name="foo.c", sarif_source_language="c") Once we have a :type:`diagnostic_file` we can use it to create instances of :type:`diagnostic_physical_location` within the :type:`diagnostic_manager`. These are owned by the :type:`diagnostic_manager` and cleaned up automatically when :func:`diagnostic_manager_release` is called. Instances of :type:`diagnostic_physical_location` can refer to * a source line as a whole, created via :func:`diagnostic_manager_new_location_from_file_and_line`. * a particular point within a source file (line/column), created via :func:`diagnostic_manager_new_location_from_file_line_column`. * a range of text within of source file, created via :func:`diagnostic_manager_new_location_from_range`. Diagnostics affecting a whole source line ***************************************** If we want a diagnostic to refer to an entire source line, we can use :func:`diagnostic_manager_new_location_from_file_and_line`. For example, given this example input where the tool can't find the header:: #include <foo.h> we could complain about it via libgdiagnostics via: .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-no-column.c :language: c :start-after: /* begin quoted source */ :end-before: /* end quoted source */ leading to output like this:: foo.c:17: error: can't find 'foo.h'" 17 | #include <foo.h> where libgdiagnostics will attempt to load ``foo.c`` and quote the pertinent line. If libgdiagnostics cannot open ``foo.c``, it will merely print:: foo.c:17: error: can't find 'foo.h' You can use :func:`diagnostic_manager_debug_dump_location` to dump a :type:`diagnostic_physical_location`. For the above example:: diagnostic_manager_debug_dump_location (diag_mgr, loc, stderr); might print:: foo.c:17 to stderr. Columns and ranges ****************** If we want to generate output like this:: foo.c:17:11: error: can't find 'foo'" 17 | #include <foo.h> | ^~~~~ where the diagnostic is marked as relating to the above range of characters in line 17, we need to express the range of characters within the line of interest. We can do this by creating a :type:`diagnostic_physical_location` for the start of the range, another one for the end of the range, and then using these two to create a :type:`diagnostic_physical_location` for the range as a whole: .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-error.c :language: c :start-after: /* begin quoted source */ :end-before: /* end quoted source */ On compiling and running the program, we should get this output:: foo.c:17:11: error: can't find 'foo.h' 17 | #include <foo.h> | ^~~~~ where libgdiagnostics will attempt to load ``foo.c`` and underling the pertinent part of the given line. If libgdiagnostics cannot open ``foo.c``, it will merely print:: foo.c:17:8: error: can't find 'foo.h' A range can span multiple lines within the same file, but cannot span multiple files. As before, you can use :func:`diagnostic_manager_debug_dump_location` to dump the locations. For the above example:: diagnostic_manager_debug_dump_location (diag_mgr, loc_start, stderr); and:: diagnostic_manager_debug_dump_location (diag_mgr, loc_range, stderr); might print:: foo.c:17:11 to stderr, whereas:: diagnostic_manager_debug_dump_location (diag_mgr, loc_end, stderr); might print:: foo.c:17:15 Multiple locations ****************** As well as the primary physical location seen above, a :type:`diagnostic` can have additional physical locations. You can add these secondary locations via :func:`diagnostic_add_location`. For example, for this valid but suspicious-looking C code:: const char *strs[3] = {"foo", "bar" "baz"}; the following :type:`diagnostic` has its primary location where the missing comma should be, and secondary locations for each of the string literals ``"foo"``, ``"bar"``, and ``"baz"``, added via :func:`diagnostic_add_location`: .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-multiple-lines.c :language: c :start-after: /* begin quoted source */ :end-before: /* end quoted source */ where the text output might be:: test-multiple-lines.c:23:29: warning: missing comma 22 | const char *strs[3] = {"foo", | ~~~~~ 23 | "bar" | ~~~~~^ 24 | "baz"}; | ~~~~~ Labelling locations ******************* You can give the locations labels using :func:`diagnostic_set_location_with_label` and :func:`diagnostic_add_location_with_label`. Consider emitting a "type mismatch" diagnostic for:: 42 + "foo" where the primary location is on the ``+``, with secondary locations on the``42`` and the ``"foo"``: .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-labelled-ranges.c :language: c :start-after: /* begin quoted source */ :end-before: /* end quoted source */ giving this text output:: test-labelled-ranges.c:9:6: error: mismatching types: 'int' and 'const char *' 19 | 42 + "foo" | ~~ ^ ~~~~~ | | | | int const char * More on locations ***************** For more details on the above, see :doc:`../topics/physical-locations`. Otherwise the :doc:`next part of the tutorial <03-logical-locations>` covers logical locations. --- ### Gcc Doc Libgdiagnostics Tutorial 03 Logical Locations (`gcc/doc/libgdiagnostics/tutorial/03-logical-locations.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Tutorial part 3: logical locations ================================== Let's extend the previous example to add a :doc:`logical location <../topics/logical-locations>` to the :type:`diagnostic`. First we create a :type:`diagnostic_logical_location` representing a particular function:: const diagnostic_logical_location *logical_loc = diagnostic_manager_new_logical_location (diag_mgr, DIAGNOSTIC_LOGICAL_LOCATION_KIND_FUNCTION, NULL, /* parent */ "foo", NULL, NULL); In this simple example we specify that it is a function, and just give it a name (``foo``). For more complicated cases we can set up tree-like hierarchies of logical locations, set qualified names, "mangled" names, and so on; see :func:`diagnostic_manager_new_logical_location` for details. Once we have :type:`diagnostic_logical_location` we can associate it with a :type:`diagnostic` with :func:`diagnostic_set_logical_location`:: diagnostic_set_logical_location (d, logical_loc); The logical location will be printed by text output sinks like this:: In function 'foo': and will be captured in :doc:`SARIF <../topics/sarif>` output. Find out more ************* For more details on the above, see :doc:`../topics/logical-locations`. Otherwise the :doc:`next part of the tutorial <04-notes>` covers adding supplementary "notes" to a :type:`diagnostic`. --- ### Gcc Doc Libgdiagnostics Tutorial 04 Notes (`gcc/doc/libgdiagnostics/tutorial/04-notes.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Tutorial part 4: adding notes ============================= Let's further extend the previous example to add a "note" to it. We want to generate output like this:: test-with-note.c:17:11: error: can't find 'foo.h' 17 | #include <foo.h> | ^~~~~ test-with-note.c:17:11: note: have you looked behind the couch? The "error" and "note" are both instances of :type:`diagnostic`. We want to let libgdiagnostics know that they are grouped together. The way to do this is to use :func:`diagnostic_manager_begin_group` and :func:`diagnostic_manager_end_group` around the "finish" calls to the diagnostics. .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-error-with-note.c :language: c :start-after: /* begin quoted source */ :end-before: /* end quoted source */ On compiling and running the program, we should get the desired output:: test-with-note.c:17:11: error: can't find 'foo.h' 17 | #include <foo.h> | ^~~~~ test-with-note.c:17:11: note: have you looked behind the couch? The grouping doesn't affect text output sinks, but a :doc:`SARIF sink <../topics/sarif>` will group the note within the error (via the ``relatedLocations`` property of ``result`` objects; see SARIF v2.1.0 `§3.27.22 <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790910>`_). In the above, the note had the same physical location as the error (``loc_range``). This can be useful for splitting up a message into two parts to make localization easier, but they could have different locations, such as in:: test.xml:10:2: error: 'foo' is not valid here test.xml:5:1: note: within element 'bar' where each :type:`diagnostic` had its own :type:`diagnostic_physical_location`. In :doc:`the next tutorial <05-warnings>` we'll look at issuing warnings, rather than errors. --- ### Gcc Doc Libgdiagnostics Tutorial 05 Warnings (`gcc/doc/libgdiagnostics/tutorial/05-warnings.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Tutorial part 5: warnings ========================= So far we've only emitted errors, but other kinds of diagnostic are possible, such as warnings. We can select different kinds of diagnostic via :enum:`diagnostic_level` when calling :func:`diagnostic_begin`: .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-warning.c :language: c :start-after: /* begin quoted source */ :end-before: /* end quoted source */ On compiling and running the program, we should get output similar to:: test-warning.c:17:11: warning: this is a warning 17 | #include <foo.h> | ^~~~~ Various severities are possible, see :enum:`diagnostic_level` for more information. In :doc:`the next section of the tutorial <06-fix-it-hints>` we'll look at adding fix-it hints to diagnostics. --- ### Gcc Doc Libgdiagnostics Tutorial 06 Fix It Hints (`gcc/doc/libgdiagnostics/tutorial/06-fix-it-hints.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Tutorial part 6: fix-it hints ============================= libgdiagnostics supports adding "fix-it hints" to a :type:`diagnostic`: suggestions for the user on how to edit their code to fix a problem. These can be expressed as insertions, replacements, and removals of text. For example, here we use :func:`diagnostic_add_fix_it_hint_replace` to add a replacement fix-it hint to a diagnostic: .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-fix-it-hint.c :language: c :start-after: /* begin quoted source */ :end-before: /* end quoted source */ On compiling and running the program, we should get output similar to:: test-fix-it-hint.c:19:13: error: unknown field 'colour'; did you mean 'color' 19 | return p->colour; | ^~~~~~ | color We can also add a call to :func:`diagnostic_manager_write_patch` to the program cleanup code:: diagnostic_manager_write_patch (diag_mgr, stderr); This will write a patch to the stream (here ``stderr``) giving the effect of all fix-it hints on all diagnostics emitted by the :type:`diagnostic_manager`, giving something like:: @@ -16,7 +16,7 @@ struct rgb get_color (struct object *p) { - return p->colour; + return p->color; } See the :doc:`guide to fix-it hints <../topics/fix-it-hints>` for more information, or go on to :doc:`the next section of the tutorial <07-execution-paths>`. --- ### Gcc Doc Libgdiagnostics Tutorial 07 Execution Paths (`gcc/doc/libgdiagnostics/tutorial/07-execution-paths.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Tutorial part 7: execution paths ================================ A :type:`diagnostic` can optionally have a :type:`diagnostic_execution_path` describing a path of execution through code. For example, let's pretend we're writing a static analysis tool for finding bugs in `CPython extension code <https://docs.python.org/3/c-api/index.html>`_. Let's say we're analyzing this code: .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-warning-with-path.c :language: c :start-after: begin fake source :end-before: end fake source This code attempts to take an Python integer parameter and then build a list of that length, containing random integers. However, there are **numerous bugs** in this code: a type mismatch, mistakes in reference-counting, and an almost total lack of error-handling. For example, ``PyList_Append`` requires a non-NULL first parameter (``list``), but ``PyList_New`` can fail, returning NULL, and this isn't checked for, which would lead to a segfault if ``PyList_New`` fails. We can add a :type:`diagnostic_execution_path` to the :type:`diagnostic` via :func:`diagnostic_add_execution_path`, and then add events to it using :func:`diagnostic_execution_path_add_event`. For example, with:: diagnostic_event_id alloc_event_id = diagnostic_execution_path_add_event (path, loc_call_to_PyList_New, logical_loc, 0, "when %qs fails, returning NULL", "PyList_New"); we create an event that will be worded as:: (1) when `PyList_New' fails, returning NULL Note that :func:`diagnostic_execution_path_add_event` returns a :type:`diagnostic_event_id`. We can use this to refer to this event in another event using the ``%@`` format code in its message, which takes the address of a :type:`diagnostic_event_id`:: diagnostic_execution_path_add_event (path, loc_call_to_PyList_Append, logical_loc, 0, "when calling %qs, passing NULL from %@ as argument %i", "PyList_Append", &alloc_event_id, 1); where the latter event will be worded as:: (2) when calling `PyList_Append', passing NULL from (1) as argument 1 where the ``%@`` reference to the other event has been printed as ``(1)``. In SARIF output the text "(1)" will have a embedded link referring within the sarif log to the ``threadFlowLocation`` object for the other event, via JSON pointer (see `§3.10.3 "URIs that use the sarif scheme" <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790707>`_). Let's add an event between these describing control flow, creating three events in all: .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-warning-with-path.c :language: c :start-after: begin path creation :end-before: end path creation Assuming we also gave it :type:`diagnostic_logical_location` with: .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-warning-with-path.c :language: c :start-after: begin create logical locs :end-before: end create logical locs and finish the :type:`diagnostic` with :func:`diagnostic_finish` like this:: diagnostic_finish (d, "passing NULL as argument %i to %qs" " which requires a non-NULL parameter", 1, "PyList_Append"); then we should get output to text sinks similar to the following:: In function 'make_a_list_of_random_ints_badly': test-warning-with-path.c:30:5: warning: passing NULL as argument 1 to 'PyList_Append' which requires a non-NULL parameter" 30 | PyList_Append(list, item); | ^~~~~~~~~~~~~~~~~~~~~~~~~ make_a_list_of_random_ints_badly': events 1-3 26 | list = PyList_New(0); | ^~~~~~~~~~~~~ | | | (1) when 'PyList_New' fails, returning NULL 27 | 28 | for (i = 0; i < count; i++) { | ~~~~~~~~~ | | | (2) when 'i < count' 29 | item = PyLong_FromLong(random()); 30 | PyList_Append(list, item); | ~~~~~~~~~~~~~~~~~~~~~~~~~ | | | (3) when calling 'PyList_Append', passing NULL from (1) as argument 1 and for SARIF sinks the path will be added as a ``codeFlow`` object (see SARIF 2.1.0 `§3.36 codeFlow object <https://docs.oasis-open.org/sarif/sarif/v2.1.0/errata01/os/sarif-v2.1.0-errata01-os-complete.html#_Toc141790990>`_). Here's the above example in full: .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-warning-with-path.c :language: c :start-after: begin full example :end-before: end full example See the :doc:`guide to execution paths <../topics/execution-paths>` for more information, or go on to :doc:`the next section of the tutorial <08-message-buffers>`. --- ### Gcc Doc Libgdiagnostics Tutorial 08 Message Buffers (`gcc/doc/libgdiagnostics/tutorial/08-message-buffers.rst`) .. Copyright (C) 2025-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Tutorial part 8: message buffers ================================ In previous examples, we finished a diagnostic with a call to :func:`diagnostic_finish`, which takes a format string and arguments to determine the text message of the diagnostic. Sometimes this approach is inconvenient, such as where you might want to build up a message programmatically from a series of components. Additionally, you might have existing code that uses ``fprintf``, whereas :func:`diagnostic_finish` has its :doc:`own formatting conventions <../topics/message-formatting>` which are :strong:`not` the same as printf. For this reason libgdiagnostics (from ``LIBGDIAGNOSTICS_ABI_3`` onwards) supports :type:`diagnostic_message_buffer`, which can be used to accumulate a message before using it. You create a :type:`diagnostic_message_buffer` using :func:`diagnostic_message_buffer_new`. There are various API entrypoints for accumulating text into the buffer. For example: .. literalinclude:: ../../../testsuite/libgdiagnostics.dg/test-message-buffer.c :language: c :start-after: /* begin quoted source */ :end-before: /* end quoted source */ Running this will produce this text output:: .. code-block:: console $ ./test-message-buffer.c.exe ./test-message-buffer.c.exe: error: this is a string; foo; int: 42 str: mostly harmless; this is a link 'this is quoted' highlight A highlight B (1). where in a suitably-capable terminal if a text sink is directly connected to a tty: * the ``this is a link`` will be a clickable hyperlink (and the URL will be captured in SARIF output). * the quoted text will be in bold * the ``highlight A`` and ``highlight B`` text will be colorized * the event ID will be colorized (and will be a URL in SARIF output if used within a :type:`diagnostic_execution_path`). Moving on ********* That's the end of the tutorial. For more information on libgdiagnostics, see the :doc:`topic guide <../topics/index>`. --- ### Gcc Doc Libgdiagnostics Tutorial Index (`gcc/doc/libgdiagnostics/tutorial/index.rst`) .. Copyright (C) 2024-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. Tutorial ======== The following tutorial gives an overview of how to use libgdiagnostics. .. toctree:: :maxdepth: 2 01-hello-world.rst 02-physical-locations.rst 03-logical-locations.rst 04-notes.rst 05-warnings.rst 06-fix-it-hints.rst 07-execution-paths.rst 08-message-buffers.rst --- ### Gcc Jit Docs Cp Index (`gcc/jit/docs/cp/index.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. C++ bindings for libgccjit ========================== This document describes the C++ bindings to `libgccjit <https://gcc.gnu.org/wiki/JIT>`_, an API for embedding GCC inside programs and libraries. The C++ bindings consist of a single header file ``libgccjit++.h``. This is a collection of "thin" wrapper classes around the C API. Everything is an inline function, implemented in terms of the C API, so there is nothing extra to link against. Contents: .. toctree:: :maxdepth: 2 intro/index.rst topics/index.rst --- ### Gcc Jit Docs Internals Index (`gcc/jit/docs/internals/index.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. Internals ========= Working on the JIT library -------------------------- Having checked out the source code (to "src"), you can configure and build the JIT library like this: .. code-block:: bash mkdir build mkdir install PREFIX=$(pwd)/install cd build ../src/configure \ --enable-host-shared \ --enable-languages=jit,c++ \ --disable-bootstrap \ --enable-checking=release \ --prefix=$PREFIX nice make -j4 # altering the "4" to however many cores you have This should build a libgccjit.so within jit/build/gcc: .. code-block:: console [build] $ file gcc/libgccjit.so* gcc/libgccjit.so: symbolic link to `libgccjit.so.0' gcc/libgccjit.so.0: symbolic link to `libgccjit.so.0.0.1' gcc/libgccjit.so.0.0.1: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, not stripped Here's what those configuration options mean: .. option:: --enable-host-shared Configuring with this option means that the compiler is built as position-independent code, which incurs a slight performance hit, but it necessary for a shared library. .. option:: --enable-languages=jit,c++ This specifies which frontends to build. The JIT library looks like a frontend to the rest of the code. The C++ portion of the JIT test suite requires the C++ frontend to be enabled at configure-time, or you may see errors like this when running the test suite: .. code-block:: console xgcc: error: /home/david/jit/src/gcc/testsuite/jit.dg/test-quadratic.cc: C++ compiler not installed on this system c++: error trying to exec 'cc1plus': execvp: No such file or directory .. option:: --disable-bootstrap For hacking on the "jit" subdirectory, performing a full bootstrap can be overkill, since it's unused by a bootstrap. However, when submitting patches, you should remove this option, to ensure that the compiler can still bootstrap itself. .. option:: --enable-checking=release The compile can perform extensive self-checking as it runs, useful when debugging, but slowing things down. For maximum speed, configure with ``--enable-checking=release`` to disable this self-checking. Running the test suite ---------------------- .. code-block:: console [build] $ cd gcc [gcc] $ make check-jit RUNTESTFLAGS="-v -v -v" A summary of the tests can then be seen in: .. code-block:: console jit/build/gcc/testsuite/jit/jit.sum and detailed logs in: .. code-block:: console jit/build/gcc/testsuite/jit/jit.log The test executables are normally deleted after each test is run. For debugging, they can be preserved by setting :envvar:`PRESERVE_EXECUTABLES` in the environment. If so, they can then be seen as: .. code-block:: console jit/build/gcc/testsuite/jit/*.exe which can be run independently. You can compile and run individual tests by passing "jit.exp=TESTNAME" to RUNTESTFLAGS e.g.: .. code-block:: console [gcc] $ PRESERVE_EXECUTABLES= \ make check-jit \ RUNTESTFLAGS="-v -v -v jit.exp=test-factorial.c" and once a test has been compiled, you can debug it directly: .. code-block:: console [gcc] $ PATH=.:$PATH \ LD_LIBRARY_PATH=. \ LIBRARY_PATH=. \ gdb --args \ testsuite/jit/test-factorial.c.exe Running under valgrind ********************** The jit testsuite detects if :envvar:`RUN_UNDER_VALGRIND` is present in the environment (with any value). If it is present, it runs the test client code under `valgrind <https://valgrind.org>`_, specifically, the default `memcheck <https://valgrind.org/docs/manual/mc-manual.html>`_ tool with `--leak-check=full <https://valgrind.org/docs/manual/mc-manual.html#opt.leak-check>`_. It automatically parses the output from valgrind, injecting XFAIL results if any issues are found, or PASS results if the output is clean. The output is saved to ``TESTNAME.exe.valgrind.txt``. For example, the following invocation verbosely runs the testcase ``test-sum-of-squares.c`` under valgrind, showing an issue: .. code-block:: console $ RUN_UNDER_VALGRIND= \ make check-jit \ RUNTESTFLAGS="-v -v -v jit.exp=test-sum-of-squares.c" (...verbose log contains detailed valgrind errors, if any...) === jit Summary === # of expected passes 28 # of expected failures 2 $ less testsuite/jit/jit.sum (...other results...) XFAIL: jit.dg/test-sum-of-squares.c: test-sum-of-squares.c.exe.valgrind.txt: definitely lost: 8 bytes in 1 blocks XFAIL: jit.dg/test-sum-of-squares.c: test-sum-of-squares.c.exe.valgrind.txt: unsuppressed errors: 1 (...other results...) $ less testsuite/jit/test-sum-of-squares.c.exe.valgrind.txt (...shows full valgrind report for this test case...) When running under valgrind, it's best to have configured gcc with :option:`--enable-valgrind-annotations`, which automatically suppresses various known false positives. Environment variables --------------------- When running client code against a locally-built libgccjit, three environment variables need to be set up: .. envvar:: LD_LIBRARY_PATH `libgccjit.so` is dynamically linked into client code, so if running against a locally-built library, ``LD_LIBRARY_PATH`` needs to be set up appropriately. The library can be found within the "gcc" subdirectory of the build tree: .. code-block:: console $ file libgccjit.so* libgccjit.so: symbolic link to `libgccjit.so.0' libgccjit.so.0: symbolic link to `libgccjit.so.0.0.1' libgccjit.so.0.0.1: ELF 64-bit LSB shared object, x86-64, version 1 (GNU/Linux), dynamically linked, not stripped .. envvar:: PATH The library uses a driver executable for converting from .s assembler files to .so shared libraries. Specifically, it looks for a name expanded from ``${target_noncanonical}-gcc-${gcc_BASEVER}${exeext}`` such as ``x86_64-unknown-linux-gnu-gcc-5.0.0``. Hence ``PATH`` needs to include a directory where the library can locate this executable. The executable is normally installed to the installation bindir (e.g. /usr/bin), but a copy is also created within the "gcc" subdirectory of the build tree for running the testsuite, and for ease of development. .. envvar:: LIBRARY_PATH The driver executable invokes the linker, and the latter needs to locate support libraries needed by the generated code, or you will see errors like: .. code-block:: console ld: cannot find crtbeginS.o: No such file or directory ld: cannot find -lgcc ld: cannot find -lgcc_s Hence if running directly from a locally-built copy (without installing), ``LIBRARY_PATH`` needs to contain the "gcc" subdirectory of the build tree. For example, to run a binary that uses the library against a non-installed build of the library in LIBGCCJIT_BUILD_DIR you need an invocation of the client code like this, to prepend the dir to each of the environment variables: .. code-block:: console $ LD_LIBRARY_PATH=$(LIBGCCJIT_BUILD_DIR):$(LD_LIBRARY_PATH) \ PATH=$(LIBGCCJIT_BUILD_DIR):$(PATH) \ LIBRARY_PATH=$(LIBGCCJIT_BUILD_DIR):$(LIBRARY_PATH) \ ./jit-hello-world hello world Packaging notes --------------- The configure-time option :option:`--enable-host-shared` is needed when building the jit in order to get position-independent code. This will slow down the regular compiler by a few percent. Hence when packaging gcc with libgccjit, please configure and build twice: * once without :option:`--enable-host-shared` for most languages, and * once with :option:`--enable-host-shared` for the jit For example: .. code-block:: bash # Configure and build with --enable-host-shared # for the jit: mkdir configuration-for-jit pushd configuration-for-jit $(SRCDIR)/configure \ --enable-host-shared \ --enable-languages=jit \ --prefix=$(DESTDIR) make popd # Configure and build *without* --enable-host-shared # for maximum speed: mkdir standard-configuration pushd standard-configuration $(SRCDIR)/configure \ --enable-languages=all \ --prefix=$(DESTDIR) make popd # Both of the above are configured to install to $(DESTDIR) # Install the configuration with --enable-host-shared first # *then* the one without, so that the faster build # of "cc1" et al overwrites the slower build. pushd configuration-for-jit make install popd pushd standard-configuration make install popd Overview of code structure -------------------------- The library is implemented in C++. * ``libgccjit.cc`` implements the API entrypoints. It performs error checking, then calls into classes of the gcc::jit::recording namespace within ``jit-recording.cc`` and ``jit-recording.h``. * The gcc::jit::recording classes (within ``jit-recording.cc`` and ``jit-recording.h``) record the API calls that are made: .. literalinclude:: ../../jit-common.h :start-after: /* Recording types. */ :end-before: /* End of recording types. */ :language: c++ * When the context is compiled, the gcc::jit::playback classes (within ``jit-playback.cc`` and ``jit-playback.h``) replay the API calls within langhook:parse_file: .. literalinclude:: ../../jit-common.h :start-after: /* Playback types. */ :end-before: /* End of playback types. */ :language: c++ .. literalinclude:: ../../notes.txt :lines: 1- Here is a high-level summary from ``jit-common.h``: .. include:: ../../jit-common.h :start-after: This comment is included by the docs. :end-before: End of comment for inclusion in the docs. */ .. _example-of-log-file: Another way to understand the structure of the code is to enable logging, via :c:func:`gcc_jit_context_set_logfile`. Here is an example of a log generated via this call: .. literalinclude:: test-hello-world.exe.log.txt :lines: 1- Design notes ------------ It should not be possible for client code to cause an internal compiler error. If this *does* happen, the root cause should be isolated (perhaps using :c:func:`gcc_jit_context_dump_reproducer_to_file`) and the cause should be rejected via additional checking. The checking ideally should be within the libgccjit API entrypoints in libgccjit.cc, since this is as close as possible to the error; failing that, a good place is within ``recording::context::validate ()`` in jit-recording.cc. Submitting patches ------------------ Please read the contribution guidelines for gcc at https://gcc.gnu.org/contribute.html. Patches for the jit should be sent to both the gcc-patches@gcc.gnu.org and jit@gcc.gnu.org mailing lists, with "jit" and "PATCH" in the Subject line. You don't need to do a full bootstrap for code that just touches the ``jit`` and ``testsuite/jit.dg`` subdirectories. However, please run ``make check-jit`` before submitting the patch, and mention the results in your email (along with the host triple that the tests were run on). A good patch should contain the information listed in the gcc contribution guide linked to above; for a ``jit`` patch, the patch should contain: * the code itself (for example, a new API entrypoint will typically touch ``libgccjit.h`` and ``.c``, along with support code in ``jit-recording.[ch]`` and ``jit-playback.[ch]`` as appropriate) * test coverage * documentation for the C API * documentation for the C++ API A patch that adds new API entrypoints should also contain: * a feature macro in ``libgccjit.h`` so that client code that doesn't use a "configure" mechanism can still easily detect the presence of the entrypoint. See e.g. ``LIBGCCJIT_HAVE_SWITCH_STATEMENTS`` (for a category of entrypoints) and ``LIBGCCJIT_HAVE_gcc_jit_context_set_bool_allow_unreachable_blocks`` (for an individual entrypoint). * a new ABI tag containing the new symbols (in ``libgccjit.map``), so that we can detect client code that uses them * Support for :c:func:`gcc_jit_context_dump_reproducer_to_file`. Most jit testcases attempt to dump their contexts to a .c file; ``jit.exp`` then sanity-checks the generated c by compiling them (though not running them). A new API entrypoint needs to "know" how to write itself back out to C (by implementing ``gcc::jit::recording::memento::write_reproducer`` for the appropriate ``memento`` subclass). * C++ bindings for the new entrypoints (see ``libgccjit++.h``); ideally with test coverage, though the C++ API test coverage is admittedly spotty at the moment * documentation for the new C entrypoints * documentation for the new C++ entrypoints * documentation for the new ABI tag (see ``topics/compatibility.rst``). Depending on the patch you can either extend an existing test case, or add a new test case. If you add an entirely new testcase: ``jit.exp`` expects jit testcases to begin with ``test-``, or ``test-error-`` (for a testcase that generates an error on a :c:type:`gcc_jit_context`). Every new testcase that doesn't generate errors should also touch ``gcc/testsuite/jit.dg/all-non-failing-tests.h``: * Testcases that don't generate errors should ideally be added to the ``testcases`` array in that file; this means that, in addition to being run standalone, they also get run within ``test-combination.c`` (which runs all successful tests inside one big :c:type:`gcc_jit_context`), and ``test-threads.c`` (which runs all successful tests in one process, each one running in a different thread on a different :c:type:`gcc_jit_context`). .. note:: Given that exported functions within a :c:type:`gcc_jit_context` must have unique names, and most testcases are run within ``test-combination.c``, this means that every jit-compiled test function typically needs a name that's unique across the entire test suite. * Testcases that aren't to be added to the ``testcases`` array should instead add a comment to the file clarifying why they're not in that array. See the file for examples. Typically a patch that touches the .rst documentation will also need the texinfo to be regenerated. You can do this with `Sphinx 1.0 <https://sphinx-doc.org/>`_ or later by running ``make texinfo`` within ``SRCDIR/gcc/jit/docs``. Don't do this within the patch sent to the mailing list; it can often be relatively large and inconsequential (e.g. anchor renumbering), rather like generated "configure" changes from configure.ac. You can regenerate it when committing to svn. --- ### Gcc Jit Docs Internals Test Hello World.Exe.Log (`gcc/jit/docs/internals/test-hello-world.exe.log.txt`) JIT: libgccjit (GCC) version 6.0.0 20150803 (experimental) (x86_64-pc-linux-gnu) JIT: compiled by GNU C version 4.8.3 20140911 (Red Hat 4.8.3-7), GMP version 5.1.2, MPFR version 3.1.2, MPC version 1.0.1 JIT: entering: gcc_jit_context_set_str_option JIT: GCC_JIT_STR_OPTION_PROGNAME: "./test-hello-world.c.exe" JIT: exiting: gcc_jit_context_set_str_option JIT: entering: gcc_jit_context_set_int_option JIT: GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL: 3 JIT: exiting: gcc_jit_context_set_int_option JIT: entering: gcc_jit_context_set_bool_option JIT: GCC_JIT_BOOL_OPTION_DEBUGINFO: true JIT: exiting: gcc_jit_context_set_bool_option JIT: entering: gcc_jit_context_set_bool_option JIT: GCC_JIT_BOOL_OPTION_DUMP_INITIAL_TREE: false JIT: exiting: gcc_jit_context_set_bool_option JIT: entering: gcc_jit_context_set_bool_option JIT: GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE: false JIT: exiting: gcc_jit_context_set_bool_option JIT: entering: gcc_jit_context_set_bool_option JIT: GCC_JIT_BOOL_OPTION_SELFCHECK_GC: true JIT: exiting: gcc_jit_context_set_bool_option JIT: entering: gcc_jit_context_set_bool_option JIT: GCC_JIT_BOOL_OPTION_DUMP_SUMMARY: false JIT: exiting: gcc_jit_context_set_bool_option JIT: entering: gcc_jit_context_get_type JIT: exiting: gcc_jit_context_get_type JIT: entering: gcc_jit_context_get_type JIT: exiting: gcc_jit_context_get_type JIT: entering: gcc_jit_context_new_param JIT: exiting: gcc_jit_context_new_param JIT: entering: gcc_jit_context_new_function JIT: exiting: gcc_jit_context_new_function JIT: entering: gcc_jit_context_new_param JIT: exiting: gcc_jit_context_new_param JIT: entering: gcc_jit_context_get_type JIT: exiting: gcc_jit_context_get_type JIT: entering: gcc_jit_context_new_function JIT: exiting: gcc_jit_context_new_function JIT: entering: gcc_jit_context_new_string_literal JIT: exiting: gcc_jit_context_new_string_literal JIT: entering: gcc_jit_function_new_block JIT: exiting: gcc_jit_function_new_block JIT: entering: gcc_jit_block_add_comment JIT: exiting: gcc_jit_block_add_comment JIT: entering: gcc_jit_context_new_call JIT: exiting: gcc_jit_context_new_call JIT: entering: gcc_jit_block_add_eval JIT: exiting: gcc_jit_block_add_eval JIT: entering: gcc_jit_block_end_with_void_return JIT: exiting: gcc_jit_block_end_with_void_return JIT: entering: gcc_jit_context_dump_reproducer_to_file JIT: entering: void gcc::jit::recording::context::dump_reproducer_to_file(const char*) JIT: exiting: void gcc::jit::recording::context::dump_reproducer_to_file(const char*) JIT: exiting: gcc_jit_context_dump_reproducer_to_file JIT: entering: gcc_jit_context_compile JIT: in-memory compile of ctxt: 0x1283e20 JIT: entering: gcc::jit::result* gcc::jit::recording::context::compile() JIT: GCC_JIT_STR_OPTION_PROGNAME: "./test-hello-world.c.exe" JIT: GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL: 3 JIT: GCC_JIT_BOOL_OPTION_DEBUGINFO: true JIT: GCC_JIT_BOOL_OPTION_DUMP_INITIAL_TREE: false JIT: GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE: false JIT: GCC_JIT_BOOL_OPTION_DUMP_GENERATED_CODE: false JIT: GCC_JIT_BOOL_OPTION_DUMP_SUMMARY: false JIT: GCC_JIT_BOOL_OPTION_DUMP_EVERYTHING: false JIT: GCC_JIT_BOOL_OPTION_SELFCHECK_GC: true JIT: GCC_JIT_BOOL_OPTION_KEEP_INTERMEDIATES: false JIT: gcc_jit_context_set_bool_allow_unreachable_blocks: false JIT: gcc_jit_context_set_bool_use_external_driver: false JIT: entering: void gcc::jit::recording::context::validate() JIT: exiting: void gcc::jit::recording::context::validate() JIT: entering: gcc::jit::playback::context::context(gcc::jit::recording::context*) JIT: exiting: gcc::jit::playback::context::context(gcc::jit::recording::context*) JIT: entering: gcc::jit::playback::compile_to_memory::compile_to_memory(gcc::jit::recording::context*) JIT: exiting: gcc::jit::playback::compile_to_memory::compile_to_memory(gcc::jit::recording::context*) JIT: entering: void gcc::jit::playback::context::compile() JIT: entering: gcc::jit::tempdir::tempdir(gcc::jit::logger*, int) JIT: exiting: gcc::jit::tempdir::tempdir(gcc::jit::logger*, int) JIT: entering: bool gcc::jit::tempdir::create() JIT: m_path_template: /tmp/libgccjit-XXXXXX JIT: m_path_tempdir: /tmp/libgccjit-CKq1M9 JIT: exiting: bool gcc::jit::tempdir::create() JIT: entering: void gcc::jit::playback::context::acquire_mutex() JIT: exiting: void gcc::jit::playback::context::acquire_mutex() JIT: entering: void gcc::jit::playback::context::make_fake_args(vec<char*>*, const char*, vec<gcc::jit::recording::requested_dump>*) JIT: reusing cached configure-time options JIT: configure_time_options[0]: -mtune=generic JIT: configure_time_options[1]: -march=x86-64 JIT: exiting: void gcc::jit::playback::context::make_fake_args(vec<char*>*, const char*, vec<gcc::jit::recording::requested_dump>*) JIT: entering: toplev::main JIT: argv[0]: ./test-hello-world.c.exe JIT: argv[1]: /tmp/libgccjit-CKq1M9/fake.c JIT: argv[2]: -fPIC JIT: argv[3]: -O3 JIT: argv[4]: -g JIT: argv[5]: -quiet JIT: argv[6]: --param JIT: argv[7]: ggc-min-expand=0 JIT: argv[8]: --param JIT: argv[9]: ggc-min-heapsize=0 JIT: argv[10]: -mtune=generic JIT: argv[11]: -march=x86-64 JIT: entering: bool jit_langhook_init() JIT: exiting: bool jit_langhook_init() JIT: entering: void gcc::jit::playback::context::replay() JIT: entering: void gcc::jit::recording::context::replay_into(gcc::jit::replayer*) JIT: exiting: void gcc::jit::recording::context::replay_into(gcc::jit::replayer*) JIT: entering: void gcc::jit::recording::context::disassociate_from_playback() JIT: exiting: void gcc::jit::recording::context::disassociate_from_playback() JIT: entering: void gcc::jit::playback::context::handle_locations() JIT: exiting: void gcc::jit::playback::context::handle_locations() JIT: entering: void gcc::jit::playback::function::build_stmt_list() JIT: exiting: void gcc::jit::playback::function::build_stmt_list() JIT: entering: void gcc::jit::playback::function::build_stmt_list() JIT: exiting: void gcc::jit::playback::function::build_stmt_list() JIT: entering: void gcc::jit::playback::function::postprocess() JIT: exiting: void gcc::jit::playback::function::postprocess() JIT: entering: void gcc::jit::playback::function::postprocess() JIT: exiting: void gcc::jit::playback::function::postprocess() JIT: exiting: void gcc::jit::playback::context::replay() JIT: exiting: toplev::main JIT: entering: void gcc::jit::playback::context::extract_any_requested_dumps(vec<gcc::jit::recording::requested_dump>*) JIT: exiting: void gcc::jit::playback::context::extract_any_requested_dumps(vec<gcc::jit::recording::requested_dump>*) JIT: entering: toplev::finalize JIT: exiting: toplev::finalize JIT: entering: virtual void gcc::jit::playback::compile_to_memory::postprocess(const char*) JIT: entering: void gcc::jit::playback::context::convert_to_dso(const char*) JIT: entering: void gcc::jit::playback::context::invoke_driver(const char*, const char*, const char*, timevar_id_t, bool, bool) JIT: entering: void gcc::jit::playback::context::add_multilib_driver_arguments(vec<char*>*) JIT: exiting: void gcc::jit::playback::context::add_multilib_driver_arguments(vec<char*>*) JIT: argv[0]: x86_64-unknown-linux-gnu-gcc-6.0.0 JIT: argv[1]: -m64 JIT: argv[2]: -shared JIT: argv[3]: /tmp/libgccjit-CKq1M9/fake.s JIT: argv[4]: -o JIT: argv[5]: /tmp/libgccjit-CKq1M9/fake.so JIT: argv[6]: -fno-use-linker-plugin JIT: entering: void gcc::jit::playback::context::invoke_embedded_driver(const vec<char*>*) JIT: exiting: void gcc::jit::playback::context::invoke_embedded_driver(const vec<char*>*) JIT: exiting: void gcc::jit::playback::context::invoke_driver(const char*, const char*, const char*, timevar_id_t, bool, bool) JIT: exiting: void gcc::jit::playback::context::convert_to_dso(const char*) JIT: entering: gcc::jit::result* gcc::jit::playback::context::dlopen_built_dso() JIT: GCC_JIT_BOOL_OPTION_DEBUGINFO was set: handing over tempdir to jit::result JIT: entering: gcc::jit::result::result(gcc::jit::logger*, void*, gcc::jit::tempdir*) JIT: exiting: gcc::jit::result::result(gcc::jit::logger*, void*, gcc::jit::tempdir*) JIT: exiting: gcc::jit::result* gcc::jit::playback::context::dlopen_built_dso() JIT: exiting: virtual void gcc::jit::playback::compile_to_memory::postprocess(const char*) JIT: entering: void gcc::jit::playback::context::release_mutex() JIT: exiting: void gcc::jit::playback::context::release_mutex() JIT: exiting: void gcc::jit::playback::context::compile() JIT: entering: gcc::jit::playback::context::~context() JIT: exiting: gcc::jit::playback::context::~context() JIT: exiting: gcc::jit::result* gcc::jit::recording::context::compile() JIT: gcc_jit_context_compile: returning (gcc_jit_result *)0x12f75d0 JIT: exiting: gcc_jit_context_compile JIT: entering: gcc_jit_result_get_code JIT: locating fnname: hello_world JIT: entering: void* gcc::jit::result::get_code(const char*) JIT: exiting: void* gcc::jit::result::get_code(const char*) JIT: gcc_jit_result_get_code: returning (void *)0x7ff6b8cd87f0 JIT: exiting: gcc_jit_result_get_code JIT: entering: gcc_jit_context_release JIT: deleting ctxt: 0x1283e20 JIT: entering: gcc::jit::recording::context::~context() JIT: exiting: gcc::jit::recording::context::~context() JIT: exiting: gcc_jit_context_release JIT: entering: gcc_jit_result_release JIT: deleting result: 0x12f75d0 JIT: entering: virtual gcc::jit::result::~result() JIT: entering: gcc::jit::tempdir::~tempdir() JIT: unlinking .s file: /tmp/libgccjit-CKq1M9/fake.s JIT: unlinking .so file: /tmp/libgccjit-CKq1M9/fake.so JIT: removing tempdir: /tmp/libgccjit-CKq1M9 JIT: exiting: gcc::jit::tempdir::~tempdir() JIT: exiting: virtual gcc::jit::result::~result() JIT: exiting: gcc_jit_result_release JIT: gcc::jit::logger::~logger() --- ### Gcc Jit Docs Intro Index (`gcc/jit/docs/intro/index.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. Tutorial ======== .. toctree:: :maxdepth: 2 tutorial01.rst tutorial02.rst tutorial03.rst tutorial04.rst tutorial05.rst --- ### Gcc Jit Docs Intro Tutorial01 (`gcc/jit/docs/intro/tutorial01.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Tutorial part 1: "Hello world" ============================== Before we look at the details of the API, let's look at building and running programs that use the library. Here's a toy "hello world" program that uses the library to synthesize a call to `printf` and uses it to write a message to stdout. Don't worry about the content of the program for now; we'll cover the details in later parts of this tutorial. .. literalinclude:: ../examples/tut01-hello-world.c :language: c Copy the above to `tut01-hello-world.c`. Assuming you have the jit library installed, build the test program using: .. code-block:: console $ gcc \ tut01-hello-world.c \ -o tut01-hello-world \ -lgccjit You should then be able to run the built program: .. code-block:: console $ ./tut01-hello-world hello world --- ### Gcc Jit Docs Intro Tutorial02 (`gcc/jit/docs/intro/tutorial02.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Tutorial part 2: Creating a trivial machine code function --------------------------------------------------------- Consider this C function: .. code-block:: c int square (int i) { return i * i; } How can we construct this at run-time using libgccjit? First we need to include the relevant header: .. code-block:: c #include <libgccjit.h> All state associated with compilation is associated with a :c:expr:`gcc_jit_context *`. Create one using :c:func:`gcc_jit_context_acquire`: .. code-block:: c gcc_jit_context *ctxt; ctxt = gcc_jit_context_acquire (); The JIT library has a system of types. It is statically-typed: every expression is of a specific type, fixed at compile-time. In our example, all of the expressions are of the C `int` type, so let's obtain this from the context, as a :c:expr:`gcc_jit_type *`, using :c:func:`gcc_jit_context_get_type`: .. code-block:: c gcc_jit_type *int_type = gcc_jit_context_get_type (ctxt, GCC_JIT_TYPE_INT); :c:expr:`gcc_jit_type *` is an example of a "contextual" object: every entity in the API is associated with a :c:expr:`gcc_jit_context *`. Memory management is easy: all such "contextual" objects are automatically cleaned up for you when the context is released, using :c:func:`gcc_jit_context_release`: .. code-block:: c gcc_jit_context_release (ctxt); so you don't need to manually track and cleanup all objects, just the contexts. Although the API is C-based, there is a form of class hierarchy, which looks like this:: +- gcc_jit_object +- gcc_jit_location +- gcc_jit_type +- gcc_jit_struct +- gcc_jit_field +- gcc_jit_function +- gcc_jit_block +- gcc_jit_rvalue +- gcc_jit_lvalue +- gcc_jit_param There are casting methods for upcasting from subclasses to parent classes. For example, :c:func:`gcc_jit_type_as_object`: .. code-block:: c gcc_jit_object *obj = gcc_jit_type_as_object (int_type); One thing you can do with a :c:expr:`gcc_jit_object *` is to ask it for a human-readable description, using :c:func:`gcc_jit_object_get_debug_string`: .. code-block:: c printf ("obj: %s\n", gcc_jit_object_get_debug_string (obj)); giving this text on stdout: .. code-block:: bash obj: int This is invaluable when debugging. Let's create the function. To do so, we first need to construct its single parameter, specifying its type and giving it a name, using :c:func:`gcc_jit_context_new_param`: .. code-block:: c gcc_jit_param *param_i = gcc_jit_context_new_param (ctxt, NULL, int_type, "i"); Now we can create the function, using :c:func:`gcc_jit_context_new_function`: .. code-block:: c gcc_jit_function *func = gcc_jit_context_new_function (ctxt, NULL, GCC_JIT_FUNCTION_EXPORTED, int_type, "square", 1, ¶m_i, 0); To define the code within the function, we must create basic blocks containing statements. Every basic block contains a list of statements, eventually terminated by a statement that either returns, or jumps to another basic block. Our function has no control-flow, so we just need one basic block: .. code-block:: c gcc_jit_block *block = gcc_jit_function_new_block (func, NULL); Our basic block is relatively simple: it immediately terminates by returning the value of an expression. We can build the expression using :c:func:`gcc_jit_context_new_binary_op`: .. code-block:: c gcc_jit_rvalue *expr = gcc_jit_context_new_binary_op ( ctxt, NULL, GCC_JIT_BINARY_OP_MULT, int_type, gcc_jit_param_as_rvalue (param_i), gcc_jit_param_as_rvalue (param_i)); A :c:expr:`gcc_jit_rvalue *` is another example of a :c:expr:`gcc_jit_object *` subclass. We can upcast it using :c:func:`gcc_jit_rvalue_as_object` and as before print it with :c:func:`gcc_jit_object_get_debug_string`. .. code-block:: c printf ("expr: %s\n", gcc_jit_object_get_debug_string ( gcc_jit_rvalue_as_object (expr))); giving this output: .. code-block:: bash expr: i * i Creating the expression in itself doesn't do anything; we have to add this expression to a statement within the block. In this case, we use it to build a return statement, which terminates the basic block: .. code-block:: c gcc_jit_block_end_with_return (block, NULL, expr); OK, we've populated the context. We can now compile it using :c:func:`gcc_jit_context_compile`: .. code-block:: c gcc_jit_result *result; result = gcc_jit_context_compile (ctxt); and get a :c:expr:`gcc_jit_result *`. At this point we're done with the context; we can release it: .. code-block:: c gcc_jit_context_release (ctxt); We can now use :c:func:`gcc_jit_result_get_code` to look up a specific machine code routine within the result, in this case, the function we created above. .. code-block:: c void *fn_ptr = gcc_jit_result_get_code (result, "square"); if (!fn_ptr) { fprintf (stderr, "NULL fn_ptr"); goto error; } We can now cast the pointer to an appropriate function pointer type, and then call it: .. code-block:: c typedef int (*fn_type) (int); fn_type square = (fn_type)fn_ptr; printf ("result: %d", square (5)); .. code-block:: bash result: 25 Once we're done with the code, we can release the result: .. code-block:: c gcc_jit_result_release (result); We can't call ``square`` anymore once we've released ``result``. Error-handling ************** Various kinds of errors are possible when using the API, such as mismatched types in an assignment. You can only compile and get code from a context if no errors occur. Errors are printed on stderr; they typically contain the name of the API entrypoint where the error occurred, and pertinent information on the problem: .. code-block:: console ./buggy-program: error: gcc_jit_block_add_assignment: mismatching types: assignment to i (type: int) from "hello world" (type: const char *) The API is designed to cope with errors without crashing, so you can get away with having a single error-handling check in your code: .. code-block:: c void *fn_ptr = gcc_jit_result_get_code (result, "square"); if (!fn_ptr) { fprintf (stderr, "NULL fn_ptr"); goto error; } For more information, see the :ref:`error-handling guide <error-handling>` within the Topic reference. Options ******* To get more information on what's going on, you can set debugging flags on the context using :c:func:`gcc_jit_context_set_bool_option`. .. (I'm deliberately not mentioning :c:macro:`GCC_JIT_BOOL_OPTION_DUMP_INITIAL_TREE` here since I think it's probably more of use to implementors than to users) Setting :c:macro:`GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE` will dump a C-like representation to stderr when you compile (GCC's "GIMPLE" representation): .. code-block:: c gcc_jit_context_set_bool_option ( ctxt, GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE, 1); result = gcc_jit_context_compile (ctxt); .. code-block:: c square (signed int i) { signed int D.260; entry: D.260 = i * i; return D.260; } We can see the generated machine code in assembler form (on stderr) by setting :c:macro:`GCC_JIT_BOOL_OPTION_DUMP_GENERATED_CODE` on the context before compiling: .. code-block:: c gcc_jit_context_set_bool_option ( ctxt, GCC_JIT_BOOL_OPTION_DUMP_GENERATED_CODE, 1); result = gcc_jit_context_compile (ctxt); .. code-block:: gas .file "fake.c" .text .globl square .type square, @function square: .LFB6: .cfi_startproc pushq %rbp .cfi_def_cfa_offset 16 .cfi_offset 6, -16 movq %rsp, %rbp .cfi_def_cfa_register 6 movl %edi, -4(%rbp) .L14: movl -4(%rbp), %eax imull -4(%rbp), %eax popq %rbp .cfi_def_cfa 7, 8 ret .cfi_endproc .LFE6: .size square, .-square .ident "GCC: (GNU) 4.9.0 20131023 (Red Hat 0.2)" .section .note.GNU-stack,"",@progbits By default, no optimizations are performed, the equivalent of GCC's `-O0` option. We can turn things up to e.g. `-O3` by calling :c:func:`gcc_jit_context_set_int_option` with :c:macro:`GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL`: .. code-block:: c gcc_jit_context_set_int_option ( ctxt, GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL, 3); .. code-block:: gas .file "fake.c" .text .p2align 4,,15 .globl square .type square, @function square: .LFB7: .cfi_startproc .L16: movl %edi, %eax imull %edi, %eax ret .cfi_endproc .LFE7: .size square, .-square .ident "GCC: (GNU) 4.9.0 20131023 (Red Hat 0.2)" .section .note.GNU-stack,"",@progbits Naturally this has only a small effect on such a trivial function. Full example ************ Here's what the above looks like as a complete program: .. literalinclude:: ../examples/tut02-square.c :lines: 1- :language: c Building and running it: .. code-block:: console $ gcc \ tut02-square.c \ -o tut02-square \ -lgccjit # Run the built program: $ ./tut02-square result: 25 --- ### Gcc Jit Docs Intro Tutorial03 (`gcc/jit/docs/intro/tutorial03.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. Tutorial part 3: Loops and variables ------------------------------------ Consider this C function: .. code-block:: c int loop_test (int n) { int sum = 0; for (int i = 0; i < n; i++) sum += i * i; return sum; } This example demonstrates some more features of libgccjit, with local variables and a loop. To break this down into libgccjit terms, it's usually easier to reword the `for` loop as a `while` loop, giving: .. code-block:: c int loop_test (int n) { int sum = 0; int i = 0; while (i < n) { sum += i * i; i++; } return sum; } Here's what the final control flow graph will look like: .. figure:: sum-of-squares.png :alt: image of a control flow graph As before, we include the libgccjit header and make a :c:expr:`gcc_jit_context *`. .. code-block:: c #include <libgccjit.h> void test (void) { gcc_jit_context *ctxt; ctxt = gcc_jit_context_acquire (); The function works with the C `int` type: .. code-block:: c gcc_jit_type *the_type = gcc_jit_context_get_type (ctxt, GCC_JIT_TYPE_INT); gcc_jit_type *return_type = the_type; though we could equally well make it work on, say, `double`: .. code-block:: c gcc_jit_type *the_type = gcc_jit_context_get_type (ctxt, GCC_JIT_TYPE_DOUBLE); Let's build the function: .. code-block:: c gcc_jit_param *n = gcc_jit_context_new_param (ctxt, NULL, the_type, "n"); gcc_jit_param *params[1] = {n}; gcc_jit_function *func = gcc_jit_context_new_function (ctxt, NULL, GCC_JIT_FUNCTION_EXPORTED, return_type, "loop_test", 1, params, 0); Expressions: lvalues and rvalues ******************************** The base class of expression is the :c:expr:`gcc_jit_rvalue *`, representing an expression that can be on the *right*-hand side of an assignment: a value that can be computed somehow, and assigned *to* a storage area (such as a variable). It has a specific :c:expr:`gcc_jit_type *`. Another important class is :c:expr:`gcc_jit_lvalue *`. A :c:expr:`gcc_jit_lvalue *`. is something that can of the *left*-hand side of an assignment: a storage area (such as a variable). In other words, every assignment can be thought of as: .. code-block:: c LVALUE = RVALUE; Note that :c:expr:`gcc_jit_lvalue *` is a subclass of :c:expr:`gcc_jit_rvalue *`, where in an assignment of the form: .. code-block:: c LVALUE_A = LVALUE_B; the `LVALUE_B` implies reading the current value of that storage area, assigning it into the `LVALUE_A`. So far the only expressions we've seen are `i * i`: .. code-block:: c gcc_jit_rvalue *expr = gcc_jit_context_new_binary_op ( ctxt, NULL, GCC_JIT_BINARY_OP_MULT, int_type, gcc_jit_param_as_rvalue (param_i), gcc_jit_param_as_rvalue (param_i)); which is a :c:expr:`gcc_jit_rvalue *`, and the various function parameters: `param_i` and `param_n`, instances of :c:expr:`gcc_jit_param *`, which is a subclass of :c:expr:`gcc_jit_lvalue *` (and, in turn, of :c:expr:`gcc_jit_rvalue *`): we can both read from and write to function parameters within the body of a function. Our new example has a couple of local variables. We create them by calling :c:func:`gcc_jit_function_new_local`, supplying a type and a name: .. code-block:: c /* Build locals: */ gcc_jit_lvalue *i = gcc_jit_function_new_local (func, NULL, the_type, "i"); gcc_jit_lvalue *sum = gcc_jit_function_new_local (func, NULL, the_type, "sum"); These are instances of :c:expr:`gcc_jit_lvalue *` - they can be read from and written to. Note that there is no precanned way to create *and* initialize a variable like in C: .. code-block:: c int i = 0; Instead, having added the local to the function, we have to separately add an assignment of `0` to `local_i` at the beginning of the function. Control flow ************ This function has a loop, so we need to build some basic blocks to handle the control flow. In this case, we need 4 blocks: 1. before the loop (initializing the locals) 2. the conditional at the top of the loop (comparing `i < n`) 3. the body of the loop 4. after the loop terminates (`return sum`) so we create these as :c:expr:`gcc_jit_block *` instances within the :c:expr:`gcc_jit_function *`: .. code-block:: c gcc_jit_block *b_initial = gcc_jit_function_new_block (func, "initial"); gcc_jit_block *b_loop_cond = gcc_jit_function_new_block (func, "loop_cond"); gcc_jit_block *b_loop_body = gcc_jit_function_new_block (func, "loop_body"); gcc_jit_block *b_after_loop = gcc_jit_function_new_block (func, "after_loop"); We now populate each block with statements. The entry block `b_initial` consists of initializations followed by a jump to the conditional. We assign `0` to `i` and to `sum`, using :c:func:`gcc_jit_block_add_assignment` to add an assignment statement, and using :c:func:`gcc_jit_context_zero` to get the constant value `0` for the relevant type for the right-hand side of the assignment: .. code-block:: c /* sum = 0; */ gcc_jit_block_add_assignment ( b_initial, NULL, sum, gcc_jit_context_zero (ctxt, the_type)); /* i = 0; */ gcc_jit_block_add_assignment ( b_initial, NULL, i, gcc_jit_context_zero (ctxt, the_type)); We can then terminate the entry block by jumping to the conditional: .. code-block:: c gcc_jit_block_end_with_jump (b_initial, NULL, b_loop_cond); The conditional block is equivalent to the line `while (i < n)` from our C example. It contains a single statement: a conditional, which jumps to one of two destination blocks depending on a boolean :c:expr:`gcc_jit_rvalue *`, in this case the comparison of `i` and `n`. We build the comparison using :c:func:`gcc_jit_context_new_comparison`: .. code-block:: c /* (i >= n) */ gcc_jit_rvalue *guard = gcc_jit_context_new_comparison ( ctxt, NULL, GCC_JIT_COMPARISON_GE, gcc_jit_lvalue_as_rvalue (i), gcc_jit_param_as_rvalue (n)); and can then use this to add `b_loop_cond`'s sole statement, via :c:func:`gcc_jit_block_end_with_conditional`: .. code-block:: c /* Equivalent to: if (guard) goto after_loop; else goto loop_body; */ gcc_jit_block_end_with_conditional ( b_loop_cond, NULL, guard, b_after_loop, /* on_true */ b_loop_body); /* on_false */ Next, we populate the body of the loop. The C statement `sum += i * i;` is an assignment operation, where an lvalue is modified "in-place". We use :c:func:`gcc_jit_block_add_assignment_op` to handle these operations: .. code-block:: c /* sum += i * i */ gcc_jit_block_add_assignment_op ( b_loop_body, NULL, sum, GCC_JIT_BINARY_OP_PLUS, gcc_jit_context_new_binary_op ( ctxt, NULL, GCC_JIT_BINARY_OP_MULT, the_type, gcc_jit_lvalue_as_rvalue (i), gcc_jit_lvalue_as_rvalue (i))); The `i++` can be thought of as `i += 1`, and can thus be handled in a similar way. We use :c:func:`gcc_jit_context_one` to get the constant value `1` (for the relevant type) for the right-hand side of the assignment. .. code-block:: c /* i++ */ gcc_jit_block_add_assignment_op ( b_loop_body, NULL, i, GCC_JIT_BINARY_OP_PLUS, gcc_jit_context_one (ctxt, the_type)); .. note:: For numeric constants other than 0 or 1, we could use :c:func:`gcc_jit_context_new_rvalue_from_int` and :c:func:`gcc_jit_context_new_rvalue_from_double`. The loop body completes by jumping back to the conditional: .. code-block:: c gcc_jit_block_end_with_jump (b_loop_body, NULL, b_loop_cond); Finally, we populate the `b_after_loop` block, reached when the loop conditional is false. We want to generate the equivalent of: .. code-block:: c return sum; so the block is just one statement: .. code-block:: c /* return sum */ gcc_jit_block_end_with_return ( b_after_loop, NULL, gcc_jit_lvalue_as_rvalue (sum)); .. note:: You can intermingle block creation with statement creation, but given that the terminator statements generally include references to other blocks, I find it's clearer to create all the blocks, *then* all the statements. We've finished populating the function. As before, we can now compile it to machine code: .. code-block:: c gcc_jit_result *result; result = gcc_jit_context_compile (ctxt); typedef int (*loop_test_fn_type) (int); loop_test_fn_type loop_test = (loop_test_fn_type)gcc_jit_result_get_code (result, "loop_test"); if (!loop_test) goto error; printf ("result: %d", loop_test (10)); .. code-block:: bash result: 285 Visualizing the control flow graph ********************************** You can see the control flow graph of a function using :c:func:`gcc_jit_function_dump_to_dot`: .. code-block:: c gcc_jit_function_dump_to_dot (func, "/tmp/sum-of-squares.dot"); giving a .dot file in GraphViz format. You can convert this to an image using `dot`: .. code-block:: bash $ dot -Tpng /tmp/sum-of-squares.dot -o /tmp/sum-of-squares.png or use a viewer (my preferred one is xdot.py; see https://github.com/jrfonseca/xdot.py; on Fedora you can install it with `yum install python-xdot`): .. figure:: sum-of-squares.png :alt: image of a control flow graph Full example ************ .. literalinclude:: ../examples/tut03-sum-of-squares.c :lines: 1- :language: c Building and running it: .. code-block:: console $ gcc \ tut03-sum-of-squares.c \ -o tut03-sum-of-squares \ -lgccjit # Run the built program: $ ./tut03-sum-of-squares loop_test returned: 285 --- ### Gcc Jit Docs Intro Tutorial04 (`gcc/jit/docs/intro/tutorial04.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. Tutorial part 4: Adding JIT-compilation to a toy interpreter ------------------------------------------------------------ In this example we construct a "toy" interpreter, and add JIT-compilation to it. Our toy interpreter ******************* It's a stack-based interpreter, and is intended as a (very simple) example of the kind of bytecode interpreter seen in dynamic languages such as Python, Ruby etc. For the sake of simplicity, our toy virtual machine is very limited: * The only data type is `int` * It can only work on one function at a time (so that the only function call that can be made is to recurse). * Functions can only take one parameter. * Functions have a stack of `int` values. * We'll implement function call within the interpreter by calling a function in our implementation, rather than implementing our own frame stack. * The parser is only good enough to get the examples to work. Naturally, a real interpreter would be much more complicated that this. The following operations are supported: .. list-table:: :header-rows: 1 :widths: 35 50 25 25 * - Operation - Meaning - Old Stack - New Stack * - DUP - Duplicate top of stack. - ``[..., x]`` - ``[..., x, x]`` * - ROT - Swap top two elements of stack. - ``[..., x, y]`` - ``[..., y, x]`` * - BINARY_ADD - Add the top two elements on the stack. - ``[..., x, y]`` - ``[..., (x+y)]`` * - BINARY_SUBTRACT - Likewise, but subtract. - ``[..., x, y]`` - ``[..., (x-y)]`` * - BINARY_MULT - Likewise, but multiply. - ``[..., x, y]`` - ``[..., (x*y)]`` * - BINARY_COMPARE_LT - Compare the top two elements on the stack and push a nonzero/zero if (x<y). - ``[..., x, y]`` - ``[..., (x<y)]`` * - RECURSE - Recurse, passing the top of the stack, and popping the result. - ``[..., x]`` - ``[..., fn(x)]`` * - RETURN - Return the top of the stack. - ``[x]`` - ``[]`` * - PUSH_CONST `arg` - Push an int const. - ``[...]`` - ``[..., arg]`` * - JUMP_ABS_IF_TRUE `arg` - Pop; if top of stack was nonzero, jump to ``arg``. - ``[..., x]`` - ``[...]`` Programs can be interpreted, disassembled, and compiled to machine code. The interpreter reads ``.toy`` scripts. Here's what a simple recursive factorial program looks like, the script ``factorial.toy``. The parser ignores lines beginning with a `#`. .. literalinclude:: ../examples/tut04-toyvm/factorial.toy :lines: 1- :language: c The interpreter is a simple infinite loop with a big ``switch`` statement based on what the next opcode is: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Execute the given function. */ :end-before: /* JIT compilation. */ :language: c Compiling to machine code ************************* We want to generate machine code that can be cast to this type and then directly executed in-process: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Functions are compiled to this function ptr type. */ :end-before: enum opcode :language: c The lifetime of the code is tied to that of a :c:expr:`gcc_jit_result *`. We'll handle this by bundling them up in a structure, so that we can clean them up together by calling :c:func:`gcc_jit_result_release`: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* A struct to hold the compilation results. */ :end-before: /* The main compilation hook. */ :language: c Our compiler isn't very sophisticated; it takes the implementation of each opcode above, and maps it directly to the operations supported by the libgccjit API. How should we handle the stack? In theory we could calculate what the stack depth will be at each opcode, and optimize away the stack manipulation "by hand". We'll see below that libgccjit is able to do this for us, so we'll implement stack manipulation in a direct way, by creating a ``stack`` array and ``stack_depth`` variables, local within the generated function, equivalent to this C code: .. code-block:: c int stack_depth; int stack[MAX_STACK_DEPTH]; We'll also have local variables ``x`` and ``y`` for use when implementing the opcodes, equivalent to this: .. code-block:: c int x; int y; This means our compiler has the following state: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* JIT compilation. */ :end-before: /* Stack manipulation. */ :language: c Setting things up ***************** First we create our types: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Create types. */ :end-before: /* The constant value 1. */ :language: c along with extracting a useful `int` constant: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* The constant value 1. */ :end-before: /* Create locations. */ :language: c We'll implement push and pop in terms of the ``stack`` array and ``stack_depth``. Here are helper functions for adding statements to a block, implementing pushing and popping values: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Stack manipulation. */ :end-before: /* A struct to hold the compilation results. */ :language: c We will support single-stepping through the generated code in the debugger, so we need to create :c:type:`gcc_jit_location` instances, one per operation in the source code. These will reference the lines of e.g. ``factorial.toy``. .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Create locations. */ :end-before: /* Creating the function. */ :language: c Let's create the function itself. As usual, we create its parameter first, then use the parameter to create the function: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Creating the function. */ :end-before: /* Create stack lvalues. */ :language: c We create the locals within the function. .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Create stack lvalues. */ :end-before: /* 1st pass: create blocks, one per opcode. :language: c Populating the function *********************** There's some one-time initialization, and the API treats the first block you create as the entrypoint of the function, so we need to create that block first: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: first. */ :end-before: /* Create a block per operation. */ :language: c We can now create blocks for each of the operations. Most of these will be consolidated into larger blocks when the optimizer runs. .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Create a block per operation. */ :end-before: /* Populate the initial block. */ :language: c Now that we have a block it can jump to when it's done, we can populate the initial block: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Populate the initial block. */ :end-before: /* 2nd pass: fill in instructions. */ :language: c We can now populate the blocks for the individual operations. We loop through them, adding instructions to their blocks: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* 2nd pass: fill in instructions. */ :end-before: /* Helper macros. */ :language: c We're going to have another big ``switch`` statement for implementing the opcodes, this time for compiling them, rather than interpreting them. It's helpful to have macros for implementing push and pop, so that we can make the ``switch`` statement that's coming up look as much as possible like the one above within the interpreter: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Helper macros. */ :end-before: gcc_jit_block_add_comment :language: c .. note:: A particularly clever implementation would have an *identical* ``switch`` statement shared by the interpreter and the compiler, with some preprocessor "magic". We're not doing that here, for the sake of simplicity. When I first implemented this compiler, I accidentally missed an edit when copying and pasting the ``Y_EQUALS_POP`` macro, so that popping the stack into ``y`` instead erroneously assigned it to ``x``, leaving ``y`` uninitialized. To track this kind of thing down, we can use :c:func:`gcc_jit_block_add_comment` to add descriptive comments to the internal representation. This is invaluable when looking through the generated IR for, say ``factorial``: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: PUSH_RVALUE (gcc_jit_lvalue_as_rvalue (state.y)) :end-before: /* Handle the individual opcodes. */ :language: c We can now write the big ``switch`` statement that implements the individual opcodes, populating the relevant block with statements: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Handle the individual opcodes. */ :end-before: /* Go to the next block. */ :language: c Every block must be terminated, via a call to one of the ``gcc_jit_block_end_with_`` entrypoints. This has been done for two of the opcodes, but we need to do it for the other ones, by jumping to the next block. .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* Go to the next block. */ :end-before: /* end of loop on PC locations. */ :language: c This is analogous to simply incrementing the program counter. Verifying the control flow graph ******************************** Having finished looping over the blocks, the context is complete. As before, we can verify that the control flow and statements are sane by using :c:func:`gcc_jit_function_dump_to_dot`: .. code-block:: c gcc_jit_function_dump_to_dot (state.fn, "/tmp/factorial.dot"); and viewing the result. Note how the label names, comments, and variable names show up in the dump, to make it easier to spot errors in our compiler. .. figure:: factorial.png :alt: image of a control flow graph Compiling the context ********************* Having finished looping over the blocks and populating them with statements, the context is complete. We can now compile it, and extract machine code from the result: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* We've now finished populating the context. Compile it. */ :end-before: /* (this leaks "jit_result" and "funcname") */ :language: c We can now run the result: .. literalinclude:: ../examples/tut04-toyvm/toyvm.c :start-after: /* JIT-compilation. */ :end-before: return 0; :language: c Single-stepping through the generated code ****************************************** It's possible to debug the generated code. To do this we need to both: * Set up source code locations for our statements, so that we can meaningfully step through the code. We did this above by calling :c:func:`gcc_jit_context_new_location` and using the results. * Enable the generation of debugging information, by setting :c:macro:`GCC_JIT_BOOL_OPTION_DEBUGINFO` on the :c:type:`gcc_jit_context` via :c:func:`gcc_jit_context_set_bool_option`: .. code-block:: c gcc_jit_context_set_bool_option ( ctxt, GCC_JIT_BOOL_OPTION_DEBUGINFO, 1); Having done this, we can put a breakpoint on the generated function: .. code-block:: console $ gdb --args ./toyvm factorial.toy 10 (gdb) break factorial Function "factorial" not defined. Make breakpoint pending on future shared library load? (y or [n]) y Breakpoint 1 (factorial) pending. (gdb) run Breakpoint 1, factorial (arg=10) at factorial.toy:14 14 DUP We've set up location information, which references ``factorial.toy``. This allows us to use e.g. ``list`` to see where we are in the script: .. code-block:: console (gdb) list 9 10 # Initial state: 11 # stack: [arg] 12 13 # 0: 14 DUP 15 # stack: [arg, arg] 16 17 # 1: 18 PUSH_CONST 2 and to step through the function, examining the data: .. code-block:: console (gdb) n 18 PUSH_CONST 2 (gdb) n 22 BINARY_COMPARE_LT (gdb) print stack $5 = {10, 10, 2, 0, -7152, 32767, 0, 0} (gdb) print stack_depth $6 = 3 You'll see that the parts of the ``stack`` array that haven't been touched yet are uninitialized. .. note:: Turning on optimizations may lead to unpredictable results when stepping through the generated code: the execution may appear to "jump around" the source code. This is analogous to turning up the optimization level in a regular compiler. Examining the generated code **************************** How good is the optimized code? We can turn up optimizations, by calling :c:func:`gcc_jit_context_set_int_option` with :c:macro:`GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL`: .. code-block:: c gcc_jit_context_set_int_option ( ctxt, GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL, 3); One of GCC's internal representations is called "gimple". A dump of the initial gimple representation of the code can be seen by setting: .. code-block:: c gcc_jit_context_set_bool_option (ctxt, GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE, 1); With optimization on and source locations displayed, this gives: .. We'll use "c" for gimple dumps .. code-block:: c factorial (signed int arg) { <unnamed type> D.80; signed int D.81; signed int D.82; signed int D.83; signed int D.84; signed int D.85; signed int y; signed int x; signed int stack_depth; signed int stack[8]; try { initial: stack_depth = 0; stack[stack_depth] = arg; stack_depth = stack_depth + 1; goto instr0; instr0: /* DUP */: stack_depth = stack_depth + -1; x = stack[stack_depth]; stack[stack_depth] = x; stack_depth = stack_depth + 1; stack[stack_depth] = x; stack_depth = stack_depth + 1; goto instr1; instr1: /* PUSH_CONST */: stack[stack_depth] = 2; stack_depth = stack_depth + 1; goto instr2; /* etc */ You can see the generated machine code in assembly form via: .. code-block:: c gcc_jit_context_set_bool_option ( ctxt, GCC_JIT_BOOL_OPTION_DUMP_GENERATED_CODE, 1); result = gcc_jit_context_compile (ctxt); which shows that (on this x86_64 box) the compiler has unrolled the loop and is using MMX instructions to perform several multiplications simultaneously: .. code-block:: gas .file "fake.c" .text .Ltext0: .p2align 4,,15 .globl factorial .type factorial, @function factorial: .LFB0: .file 1 "factorial.toy" .loc 1 14 0 .cfi_startproc .LVL0: .L2: .loc 1 26 0 cmpl $1, %edi jle .L13 leal -1(%rdi), %edx movl %edx, %ecx shrl $2, %ecx leal 0(,%rcx,4), %esi testl %esi, %esi je .L14 cmpl $9, %edx jbe .L14 leal -2(%rdi), %eax movl %eax, -16(%rsp) leal -3(%rdi), %eax movd -16(%rsp), %xmm0 movl %edi, -16(%rsp) movl %eax, -12(%rsp) movd -16(%rsp), %xmm1 xorl %eax, %eax movl %edx, -16(%rsp) movd -12(%rsp), %xmm4 movd -16(%rsp), %xmm6 punpckldq %xmm4, %xmm0 movdqa .LC1(%rip), %xmm4 punpckldq %xmm6, %xmm1 punpcklqdq %xmm0, %xmm1 movdqa .LC0(%rip), %xmm0 jmp .L5 # etc - edited for brevity This is clearly overkill for a function that will likely overflow the ``int`` type before the vectorization is worthwhile - but then again, this is a toy example. Turning down the optimization level to 2: .. code-block:: c gcc_jit_context_set_int_option ( ctxt, GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL, 3); yields this code, which is simple enough to quote in its entirety: .. code-block:: gas .file "fake.c" .text .p2align 4,,15 .globl factorial .type factorial, @function factorial: .LFB0: .cfi_startproc .L2: cmpl $1, %edi jle .L8 movl $1, %edx jmp .L4 .p2align 4,,10 .p2align 3 .L6: movl %eax, %edi .L4: .L5: leal -1(%rdi), %eax imull %edi, %edx cmpl $1, %eax jne .L6 .L3: .L7: imull %edx, %eax ret .L8: movl %edi, %eax movl $1, %edx jmp .L7 .cfi_endproc .LFE0: .size factorial, .-factorial .ident "GCC: (GNU) 4.9.0 20131023 (Red Hat 0.2)" .section .note.GNU-stack,"",@progbits Note that the stack pushing and popping have been eliminated, as has the recursive call (in favor of an iteration). Putting it all together *********************** The complete example can be seen in the source tree at ``gcc/jit/docs/examples/tut04-toyvm/toyvm.c`` along with a Makefile and a couple of sample .toy scripts: .. code-block:: console $ ls -al drwxrwxr-x. 2 david david 4096 Sep 19 17:46 . drwxrwxr-x. 3 david david 4096 Sep 19 15:26 .. -rw-rw-r--. 1 david david 615 Sep 19 12:43 factorial.toy -rw-rw-r--. 1 david david 834 Sep 19 13:08 fibonacci.toy -rw-rw-r--. 1 david david 238 Sep 19 14:22 Makefile -rw-rw-r--. 1 david david 16457 Sep 19 17:07 toyvm.c $ make toyvm g++ -Wall -g -o toyvm toyvm.c -lgccjit $ ./toyvm factorial.toy 10 interpreter result: 3628800 compiler result: 3628800 $ ./toyvm fibonacci.toy 10 interpreter result: 55 compiler result: 55 Behind the curtain: How does our code get optimized? **************************************************** Our example is done, but you may be wondering about exactly how the compiler turned what we gave it into the machine code seen above. We can examine what the compiler is doing in detail by setting: .. code-block:: c gcc_jit_context_set_bool_option (state.ctxt, GCC_JIT_BOOL_OPTION_DUMP_EVERYTHING, 1); gcc_jit_context_set_bool_option (state.ctxt, GCC_JIT_BOOL_OPTION_KEEP_INTERMEDIATES, 1); This will dump detailed information about the compiler's state to a directory under ``/tmp``, and keep it from being cleaned up. The precise names and their formats of these files is subject to change. Higher optimization levels lead to more files. Here's what I saw (edited for brevity; there were almost 200 files): .. code-block:: console intermediate files written to /tmp/libgccjit-KPQbGw $ ls /tmp/libgccjit-KPQbGw/ fake.c.000i.cgraph fake.c.000i.type-inheritance fake.c.004t.gimple fake.c.007t.omplower fake.c.008t.lower fake.c.011t.eh fake.c.012t.cfg fake.c.014i.visibility fake.c.015i.early_local_cleanups fake.c.016t.ssa # etc The gimple code is converted into Static Single Assignment form, with annotations for use when generating the debuginfo: .. code-block:: console $ less /tmp/libgccjit-KPQbGw/fake.c.016t.ssa .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) factorial (signed int arg) { signed int stack[8]; signed int stack_depth; signed int x; signed int y; <unnamed type> _20; signed int _21; signed int _38; signed int _44; signed int _51; signed int _56; initial: stack_depth_3 = 0; # DEBUG stack_depth => stack_depth_3 stack[stack_depth_3] = arg_5(D); stack_depth_7 = stack_depth_3 + 1; # DEBUG stack_depth => stack_depth_7 # DEBUG instr0 => NULL # DEBUG /* DUP */ => NULL stack_depth_8 = stack_depth_7 + -1; # DEBUG stack_depth => stack_depth_8 x_9 = stack[stack_depth_8]; # DEBUG x => x_9 stack[stack_depth_8] = x_9; stack_depth_11 = stack_depth_8 + 1; # DEBUG stack_depth => stack_depth_11 stack[stack_depth_11] = x_9; stack_depth_13 = stack_depth_11 + 1; # DEBUG stack_depth => stack_depth_13 # DEBUG instr1 => NULL # DEBUG /* PUSH_CONST */ => NULL stack[stack_depth_13] = 2; /* etc; edited for brevity */ We can perhaps better see the code by turning off :c:macro:`GCC_JIT_BOOL_OPTION_DEBUGINFO` to suppress all those ``DEBUG`` statements, giving: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.016t.ssa .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) factorial (signed int arg) { signed int stack[8]; signed int stack_depth; signed int x; signed int y; <unnamed type> _20; signed int _21; signed int _38; signed int _44; signed int _51; signed int _56; initial: stack_depth_3 = 0; stack[stack_depth_3] = arg_5(D); stack_depth_7 = stack_depth_3 + 1; stack_depth_8 = stack_depth_7 + -1; x_9 = stack[stack_depth_8]; stack[stack_depth_8] = x_9; stack_depth_11 = stack_depth_8 + 1; stack[stack_depth_11] = x_9; stack_depth_13 = stack_depth_11 + 1; stack[stack_depth_13] = 2; stack_depth_15 = stack_depth_13 + 1; stack_depth_16 = stack_depth_15 + -1; y_17 = stack[stack_depth_16]; stack_depth_18 = stack_depth_16 + -1; x_19 = stack[stack_depth_18]; _20 = x_19 < y_17; _21 = (signed int) _20; stack[stack_depth_18] = _21; stack_depth_23 = stack_depth_18 + 1; stack_depth_24 = stack_depth_23 + -1; x_25 = stack[stack_depth_24]; if (x_25 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: stack_depth_26 = stack_depth_24 + -1; x_27 = stack[stack_depth_26]; stack[stack_depth_26] = x_27; stack_depth_29 = stack_depth_26 + 1; stack[stack_depth_29] = x_27; stack_depth_31 = stack_depth_29 + 1; stack[stack_depth_31] = 1; stack_depth_33 = stack_depth_31 + 1; stack_depth_34 = stack_depth_33 + -1; y_35 = stack[stack_depth_34]; stack_depth_36 = stack_depth_34 + -1; x_37 = stack[stack_depth_36]; _38 = x_37 - y_35; stack[stack_depth_36] = _38; stack_depth_40 = stack_depth_36 + 1; stack_depth_41 = stack_depth_40 + -1; x_42 = stack[stack_depth_41]; _44 = factorial (x_42); stack[stack_depth_41] = _44; stack_depth_46 = stack_depth_41 + 1; stack_depth_47 = stack_depth_46 + -1; y_48 = stack[stack_depth_47]; stack_depth_49 = stack_depth_47 + -1; x_50 = stack[stack_depth_49]; _51 = x_50 * y_48; stack[stack_depth_49] = _51; stack_depth_53 = stack_depth_49 + 1; # stack_depth_1 = PHI <stack_depth_24(2), stack_depth_53(3)> instr9: /* RETURN */: stack_depth_54 = stack_depth_1 + -1; x_55 = stack[stack_depth_54]; _56 = x_55; stack ={v} {CLOBBER}; return _56; } Note in the above how all the :c:type:`gcc_jit_block` instances we created have been consolidated into just 3 blocks in GCC's internal representation: ``initial``, ``instr4`` and ``instr9``. Optimizing away stack manipulation ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Recall our simple implementation of stack operations. Let's examine how the stack operations are optimized away. After a pass of constant-propagation, the depth of the stack at each opcode can be determined at compile-time: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.021t.ccp1 .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) factorial (signed int arg) { signed int stack[8]; signed int stack_depth; signed int x; signed int y; <unnamed type> _20; signed int _21; signed int _38; signed int _44; signed int _51; initial: stack[0] = arg_5(D); x_9 = stack[0]; stack[0] = x_9; stack[1] = x_9; stack[2] = 2; y_17 = stack[2]; x_19 = stack[1]; _20 = x_19 < y_17; _21 = (signed int) _20; stack[1] = _21; x_25 = stack[1]; if (x_25 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: x_27 = stack[0]; stack[0] = x_27; stack[1] = x_27; stack[2] = 1; y_35 = stack[2]; x_37 = stack[1]; _38 = x_37 - y_35; stack[1] = _38; x_42 = stack[1]; _44 = factorial (x_42); stack[1] = _44; y_48 = stack[1]; x_50 = stack[0]; _51 = x_50 * y_48; stack[0] = _51; instr9: /* RETURN */: x_55 = stack[0]; x_56 = x_55; stack ={v} {CLOBBER}; return x_56; } Note how, in the above, all those ``stack_depth`` values are now just constants: we're accessing specific stack locations at each opcode. The "esra" pass ("Early Scalar Replacement of Aggregates") breaks out our "stack" array into individual elements: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.024t.esra .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) Created a replacement for stack offset: 0, size: 32: stack$0 Created a replacement for stack offset: 32, size: 32: stack$1 Created a replacement for stack offset: 64, size: 32: stack$2 Symbols to be put in SSA form { D.89 D.90 D.91 } Incremental SSA update started at block: 0 Number of blocks in CFG: 5 Number of blocks to update: 4 ( 80%) factorial (signed int arg) { signed int stack$2; signed int stack$1; signed int stack$0; signed int stack[8]; signed int stack_depth; signed int x; signed int y; <unnamed type> _20; signed int _21; signed int _38; signed int _44; signed int _51; initial: stack$0_45 = arg_5(D); x_9 = stack$0_45; stack$0_39 = x_9; stack$1_32 = x_9; stack$2_30 = 2; y_17 = stack$2_30; x_19 = stack$1_32; _20 = x_19 < y_17; _21 = (signed int) _20; stack$1_28 = _21; x_25 = stack$1_28; if (x_25 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: x_27 = stack$0_39; stack$0_22 = x_27; stack$1_14 = x_27; stack$2_12 = 1; y_35 = stack$2_12; x_37 = stack$1_14; _38 = x_37 - y_35; stack$1_10 = _38; x_42 = stack$1_10; _44 = factorial (x_42); stack$1_6 = _44; y_48 = stack$1_6; x_50 = stack$0_22; _51 = x_50 * y_48; stack$0_1 = _51; # stack$0_52 = PHI <stack$0_39(2), stack$0_1(3)> instr9: /* RETURN */: x_55 = stack$0_52; x_56 = x_55; stack ={v} {CLOBBER}; return x_56; } Hence at this point, all those pushes and pops of the stack are now simply assignments to specific temporary variables. After some copy propagation, the stack manipulation has been completely optimized away: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.026t.copyprop1 .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) factorial (signed int arg) { signed int stack$2; signed int stack$1; signed int stack$0; signed int stack[8]; signed int stack_depth; signed int x; signed int y; <unnamed type> _20; signed int _21; signed int _38; signed int _44; signed int _51; initial: stack$0_39 = arg_5(D); _20 = arg_5(D) <= 1; _21 = (signed int) _20; if (_21 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: _38 = arg_5(D) + -1; _44 = factorial (_38); _51 = arg_5(D) * _44; stack$0_1 = _51; # stack$0_52 = PHI <arg_5(D)(2), _51(3)> instr9: /* RETURN */: stack ={v} {CLOBBER}; return stack$0_52; } Later on, another pass finally eliminated ``stack_depth`` local and the unused parts of the `stack`` array altogether: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.036t.release_ssa .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) Released 44 names, 314.29%, removed 44 holes factorial (signed int arg) { signed int stack$0; signed int mult_acc_1; <unnamed type> _5; signed int _6; signed int _7; signed int mul_tmp_10; signed int mult_acc_11; signed int mult_acc_13; # arg_9 = PHI <arg_8(D)(0)> # mult_acc_13 = PHI <1(0)> initial: <bb 5>: # arg_4 = PHI <arg_9(2), _7(3)> # mult_acc_1 = PHI <mult_acc_13(2), mult_acc_11(3)> _5 = arg_4 <= 1; _6 = (signed int) _5; if (_6 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: _7 = arg_4 + -1; mult_acc_11 = mult_acc_1 * arg_4; goto <bb 5>; # stack$0_12 = PHI <arg_4(5)> instr9: /* RETURN */: mul_tmp_10 = mult_acc_1 * stack$0_12; return mul_tmp_10; } Elimination of tail recursion ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Another significant optimization is the detection that the call to ``factorial`` is tail recursion, which can be eliminated in favor of an iteration: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.030t.tailr1 .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) Symbols to be put in SSA form { D.88 } Incremental SSA update started at block: 0 Number of blocks in CFG: 5 Number of blocks to update: 4 ( 80%) factorial (signed int arg) { signed int stack$2; signed int stack$1; signed int stack$0; signed int stack[8]; signed int stack_depth; signed int x; signed int y; signed int mult_acc_1; <unnamed type> _20; signed int _21; signed int _38; signed int mul_tmp_44; signed int mult_acc_51; # arg_5 = PHI <arg_39(D)(0), _38(3)> # mult_acc_1 = PHI <1(0), mult_acc_51(3)> initial: _20 = arg_5 <= 1; _21 = (signed int) _20; if (_21 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: _38 = arg_5 + -1; mult_acc_51 = mult_acc_1 * arg_5; goto <bb 2> (initial); # stack$0_52 = PHI <arg_5(2)> instr9: /* RETURN */: stack ={v} {CLOBBER}; mul_tmp_44 = mult_acc_1 * stack$0_52; return mul_tmp_44; } --- ### Gcc Jit Docs Intro Tutorial05 (`gcc/jit/docs/intro/tutorial05.rst`) .. Copyright (C) 2015-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. Tutorial part 5: Implementing an Ahead-of-Time compiler ------------------------------------------------------- If you have a pre-existing language frontend that's compatible with libgccjit's license, it's possible to hook it up to libgccjit as a backend. In the previous example we showed how to do that for in-memory JIT-compilation, but libgccjit can also compile code directly to a file, allowing you to implement a more traditional ahead-of-time compiler ("JIT" is something of a misnomer for this use-case). The essential difference is to compile the context using :c:func:`gcc_jit_context_compile_to_file` rather than :c:func:`gcc_jit_context_compile`. The "brainf" language ********************* In this example we use libgccjit to construct an ahead-of-time compiler for an esoteric programming language that we shall refer to as "brainf". brainf scripts operate on an array of bytes, with a notional data pointer within the array. brainf is hard for humans to read, but it's trivial to write a parser for it, as there is no lexing; just a stream of bytes. The operations are: .. list-table:: :header-rows: 1 * - Character - Meaning * - ``>`` - ``idx += 1`` * - ``<`` - ``idx -= 1`` * - ``+`` - ``data[idx] += 1`` * - ``-`` - ``data[idx] -= 1`` * - ``.`` - ``output (data[idx])`` * - ``,`` - ``data[idx] = input ()`` * - ``[`` - loop until ``data[idx] == 0`` * - ``]`` - end of loop * - Anything else - ignored Unlike the previous example, we'll implement an ahead-of-time compiler, which reads ``.bf`` scripts and outputs executables (though it would be trivial to have it run them JIT-compiled in-process). Here's what a simple ``.bf`` script looks like: .. literalinclude:: ../examples/emit-alphabet.bf :lines: 1- .. note:: This example makes use of whitespace and comments for legibility, but could have been written as:: ++++++++++++++++++++++++++ >+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++< [>.+<-] It's not a particularly useful language, except for providing compiler-writers with a test case that's easy to parse. The point is that you can use :c:func:`gcc_jit_context_compile_to_file` to use libgccjit as a backend for a pre-existing language frontend (provided that the pre-existing frontend is compatible with libgccjit's license). Converting a brainf script to libgccjit IR ****************************************** As before we write simple code to populate a :c:expr:`gcc_jit_context *`. .. literalinclude:: ../examples/tut05-bf.c :start-after: #define MAX_OPEN_PARENS 16 :end-before: /* Entrypoint to the compiler. */ :language: c Compiling a context to a file ***************************** Unlike the previous tutorial, this time we'll compile the context directly to an executable, using :c:func:`gcc_jit_context_compile_to_file`: .. code-block:: c gcc_jit_context_compile_to_file (ctxt, GCC_JIT_OUTPUT_KIND_EXECUTABLE, output_file); Here's the top-level of the compiler, which is what actually calls into :c:func:`gcc_jit_context_compile_to_file`: .. literalinclude:: ../examples/tut05-bf.c :start-after: /* Entrypoint to the compiler. */ :end-before: /* Use the built compiler to compile the example to an executable: :language: c Note how once the context is populated you could trivially instead compile it to memory using :c:func:`gcc_jit_context_compile` and run it in-process as in the previous tutorial. To create an executable, we need to export a ``main`` function. Here's how to create one from the JIT API: .. literalinclude:: ../examples/tut05-bf.c :start-after: #include "libgccjit.h" :end-before: #define MAX_OPEN_PARENS 16 :language: c .. note:: The above implementation ignores ``argc`` and ``argv``, but you could make use of them by exposing ``param_argc`` and ``param_argv`` to the caller. Upon compiling this C code, we obtain a bf-to-machine-code compiler; let's call it ``bfc``: .. code-block:: console $ gcc \ tut05-bf.c \ -o bfc \ -lgccjit We can now use ``bfc`` to compile .bf files into machine code executables: .. code-block:: console $ ./bfc \ emit-alphabet.bf \ a.out which we can run directly: .. code-block:: console $ ./a.out ABCDEFGHIJKLMNOPQRSTUVWXYZ Success! We can also inspect the generated executable using standard tools: .. code-block:: console $ objdump -d a.out |less which shows that libgccjit has managed to optimize the function somewhat (for example, the runs of 26 and 65 increment operations have become integer constants 0x1a and 0x41): .. code-block:: console 0000000000400620 <main>: 400620: 80 3d 39 0a 20 00 00 cmpb $0x0,0x200a39(%rip) # 601060 <data 400627: 74 07 je 400630 <main 400629: eb fe jmp 400629 <main+0x9> 40062b: 0f 1f 44 00 00 nopl 0x0(%rax,%rax,1) 400630: 48 83 ec 08 sub $0x8,%rsp 400634: 0f b6 05 26 0a 20 00 movzbl 0x200a26(%rip),%eax # 601061 <data_cells+0x1> 40063b: c6 05 1e 0a 20 00 1a movb $0x1a,0x200a1e(%rip) # 601060 <data_cells> 400642: 8d 78 41 lea 0x41(%rax),%edi 400645: 40 88 3d 15 0a 20 00 mov %dil,0x200a15(%rip) # 601061 <data_cells+0x1> 40064c: 0f 1f 40 00 nopl 0x0(%rax) 400650: 40 0f b6 ff movzbl %dil,%edi 400654: e8 87 fe ff ff callq 4004e0 <putchar@plt> 400659: 0f b6 05 01 0a 20 00 movzbl 0x200a01(%rip),%eax # 601061 <data_cells+0x1> 400660: 80 2d f9 09 20 00 01 subb $0x1,0x2009f9(%rip) # 601060 <data_cells> 400667: 8d 78 01 lea 0x1(%rax),%edi 40066a: 40 88 3d f0 09 20 00 mov %dil,0x2009f0(%rip) # 601061 <data_cells+0x1> 400671: 75 dd jne 400650 <main+0x30> 400673: 31 c0 xor %eax,%eax 400675: 48 83 c4 08 add $0x8,%rsp 400679: c3 retq 40067a: 66 0f 1f 44 00 00 nopw 0x0(%rax,%rax,1) We also set up debugging information (via :c:func:`gcc_jit_context_new_location` and :c:macro:`GCC_JIT_BOOL_OPTION_DEBUGINFO`), so it's possible to use ``gdb`` to singlestep through the generated binary and inspect the internal state ``idx`` and ``data_cells``: .. code-block:: console (gdb) break main Breakpoint 1 at 0x400790 (gdb) run Starting program: a.out Breakpoint 1, 0x0000000000400790 in main (argc=1, argv=0x7fffffffe448) (gdb) stepi 0x0000000000400797 in main (argc=1, argv=0x7fffffffe448) (gdb) stepi 0x00000000004007a0 in main (argc=1, argv=0x7fffffffe448) (gdb) stepi 9 >+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++< (gdb) list 4 5 cell 0 = 26 6 ++++++++++++++++++++++++++ 7 8 cell 1 = 65 9 >+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++< 10 11 while cell#0 != 0 12 [ 13 > (gdb) n 6 ++++++++++++++++++++++++++ (gdb) n 9 >+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++< (gdb) p idx $1 = 1 (gdb) p data_cells $2 = "\032", '\000' <repeats 29998 times> (gdb) p data_cells[0] $3 = 26 '\032' (gdb) p data_cells[1] $4 = 0 '\000' (gdb) list 4 5 cell 0 = 26 6 ++++++++++++++++++++++++++ 7 8 cell 1 = 65 9 >+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++< 10 11 while cell#0 != 0 12 [ 13 > Other forms of ahead-of-time-compilation **************************************** The above demonstrates compiling a :c:expr:`gcc_jit_context *` directly to an executable. It's also possible to compile it to an object file, and to a dynamic library. See the documentation of :c:func:`gcc_jit_context_compile_to_file` for more information. --- ### Gcc Jit Docs Topics Asm (`gcc/jit/docs/topics/asm.rst`) .. Copyright (C) 2020-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Using Assembly Language with libgccjit ====================================== libgccjit has some support for directly embedding assembler instructions. This is based on GCC's support for inline ``asm`` in C code, and the following assumes a familiarity with that functionality. See `How to Use Inline Assembly Language in C Code <https://gcc.gnu.org/onlinedocs/gcc/Using-Assembly-Language-with-C.html>`_ in GCC's documentation, the "Extended Asm" section in particular. These entrypoints were added in :ref:`LIBGCCJIT_ABI_15`; you can test for their presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_ASM_STATEMENTS Adding assembler instructions within a function *********************************************** .. type:: gcc_jit_extended_asm A `gcc_jit_extended_asm` represents an extended ``asm`` statement: a series of low-level instructions inside a function that convert inputs to outputs. To avoid having an API entrypoint with a very large number of parameters, an extended ``asm`` statement is made in stages: an initial call to create the :type:`gcc_jit_extended_asm`, followed by calls to add operands and set other properties of the statement. There are two API entrypoints for creating a :type:`gcc_jit_extended_asm`: * :func:`gcc_jit_block_add_extended_asm` for an ``asm`` statement with no control flow, and * :func:`gcc_jit_block_end_with_extended_asm_goto` for an ``asm goto``. For example, to create the equivalent of: .. literalinclude:: ../../../testsuite/jit.dg/test-asm.c :start-after: // Quote from here in docs/topics/asm.rst: example 1: C :end-before: // Quote up to here in docs/topics/asm.rst: example 1: C :language: c the following API calls could be used: .. literalinclude:: ../../../testsuite/jit.dg/test-asm.c :start-after: /* Quote from here in docs/topics/asm.rst: example 1: jit. */ :end-before: /* Quote up to here in docs/topics/asm.rst: example 1: jit. */ :language: c .. warning:: When considering the numbering of operands within an extended ``asm`` statement (e.g. the ``%0`` and ``%1`` above), the equivalent to the C syntax is followed i.e. all output operands, then all input operands, regardless of what order the calls to :func:`gcc_jit_extended_asm_add_output_operand` and :func:`gcc_jit_extended_asm_add_input_operand` were made in. As in the C syntax, operands can be given symbolic names to avoid having to number them. For example, to create the equivalent of: .. literalinclude:: ../../../testsuite/jit.dg/test-asm.c :start-after: // Quote from here in docs/topics/asm.rst: example 2: C :end-before: // Quote up to here in docs/topics/asm.rst: example 2: C :language: c the following API calls could be used: .. literalinclude:: ../../../testsuite/jit.dg/test-asm.c :start-after: /* Quote from here in docs/topics/asm.rst: example 2: jit. */ :end-before: /* Quote up to here in docs/topics/asm.rst: example 2: jit. */ :language: c .. function:: gcc_jit_extended_asm *\ gcc_jit_block_add_extended_asm (gcc_jit_block *block,\ gcc_jit_location *loc,\ const char *asm_template) Create a :type:`gcc_jit_extended_asm` for an extended ``asm`` statement with no control flow (i.e. without the ``goto`` qualifier). The parameter ``asm_template`` corresponds to the `AssemblerTemplate` within C's extended ``asm`` syntax. It must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. .. function:: gcc_jit_extended_asm *\ gcc_jit_block_end_with_extended_asm_goto (gcc_jit_block *block,\ gcc_jit_location *loc,\ const char *asm_template,\ int num_goto_blocks,\ gcc_jit_block **goto_blocks,\ gcc_jit_block *fallthrough_block) Create a :type:`gcc_jit_extended_asm` for an extended ``asm`` statement that may perform jumps, and use it to terminate the given block. This is equivalent to the ``goto`` qualifier in C's extended ``asm`` syntax. For example, to create the equivalent of: .. literalinclude:: ../../../testsuite/jit.dg/test-asm.c :start-after: // Quote from here in docs/topics/asm.rst: example 3b: C :end-before: // Quote up to here in docs/topics/asm.rst: example 3b: C :language: c the following API calls could be used: .. literalinclude:: ../../../testsuite/jit.dg/test-asm.c :start-after: /* Quote from here in docs/topics/asm.rst: example 3: jit. */ :end-before: /* Quote up to here in docs/topics/asm.rst: example 3: jit. */ :language: c here referencing a :type:`gcc_jit_block` named "carry". ``num_goto_blocks`` must be >= 0. ``goto_blocks`` must be non-NULL. This corresponds to the ``GotoLabels`` parameter within C's extended ``asm`` syntax. The block names can be referenced within the assembler template. ``fallthrough_block`` can be NULL. If non-NULL, it specifies the block to fall through to after the statement. .. note:: This is needed since each :type:`gcc_jit_block` must have a single exit point, as a basic block: you can't jump from the middle of a block. A "goto" is implicitly added after the asm to handle the fallthrough case, which is equivalent to what would have happened in the C case. .. function:: void\ gcc_jit_extended_asm_set_volatile_flag (gcc_jit_extended_asm *ext_asm,\ int flag) Set whether the :type:`gcc_jit_extended_asm` has side-effects, equivalent to the `volatile <https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html#Volatile>`_ qualifier in C's extended asm syntax. For example, to create the equivalent of: .. code-block:: c asm volatile ("rdtsc\n\t" // Returns the time in EDX:EAX. "shl $32, %%rdx\n\t" // Shift the upper bits left. "or %%rdx, %0" // 'Or' in the lower bits. : "=a" (msr) : : "rdx"); the following API calls could be used: .. literalinclude:: ../../../testsuite/jit.dg/test-asm.c :start-after: /* Quote from here in docs/topics/asm.rst: example 4: jit. */ :end-before: /* Quote up to here in docs/topics/asm.rst: example 4: jit. */ :language: c where the :type:`gcc_jit_extended_asm` is flagged as volatile. .. function:: void\ gcc_jit_extended_asm_set_inline_flag (gcc_jit_extended_asm *ext_asm,\ int flag) Set the equivalent of the `inline <https://gcc.gnu.org/onlinedocs/gcc/Size-of-an-asm.html#Size-of-an-asm>`_ qualifier in C's extended ``asm`` syntax. .. function:: void\ gcc_jit_extended_asm_add_output_operand (gcc_jit_extended_asm *ext_asm,\ const char *asm_symbolic_name,\ const char *constraint,\ gcc_jit_lvalue *dest) Add an output operand to the extended ``asm`` statement. See the `Output Operands <https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html#OutputOperands>`_ section of the documentation of the C syntax. ``asm_symbolic_name`` corresponds to the ``asmSymbolicName`` component of C's extended ``asm`` syntax. It can be NULL. If non-NULL it specifies the symbolic name for the operand. ``constraint`` corresponds to the ``constraint`` component of C's extended ``asm`` syntax. It must be non-NULL. ``dest`` corresponds to the ``cvariablename`` component of C's extended ``asm`` syntax. It must be non-NULL. .. code-block:: c // Example with a NULL symbolic name, the equivalent of: // : "=r" (dst) gcc_jit_extended_asm_add_output_operand (ext_asm, NULL, "=r", dst); // Example with a symbolic name ("aIndex"), the equivalent of: // : [aIndex] "=r" (index) gcc_jit_extended_asm_add_output_operand (ext_asm, "aIndex", "=r", index); This function can't be called on an ``asm goto`` as such instructions can't have outputs; see the `Goto Labels <https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html#GotoLabels>`_ section of GCC's "Extended Asm" documentation. .. function:: void\ gcc_jit_extended_asm_add_input_operand (gcc_jit_extended_asm *ext_asm,\ const char *asm_symbolic_name,\ const char *constraint,\ gcc_jit_rvalue *src) Add an input operand to the extended ``asm`` statement. See the `Input Operands <https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html#InputOperands>`_ section of the documentation of the C syntax. ``asm_symbolic_name`` corresponds to the ``asmSymbolicName`` component of C's extended ``asm`` syntax. It can be NULL. If non-NULL it specifies the symbolic name for the operand. ``constraint`` corresponds to the ``constraint`` component of C's extended ``asm`` syntax. It must be non-NULL. ``src`` corresponds to the ``cexpression`` component of C's extended ``asm`` syntax. It must be non-NULL. .. code-block:: c // Example with a NULL symbolic name, the equivalent of: // : "r" (src) gcc_jit_extended_asm_add_input_operand (ext_asm, NULL, "r", gcc_jit_lvalue_as_rvalue (src)); // Example with a symbolic name ("aMask"), the equivalent of: // : [aMask] "r" (Mask) gcc_jit_extended_asm_add_input_operand (ext_asm, "aMask", "r", gcc_jit_lvalue_as_rvalue (mask)); .. function:: void\ gcc_jit_extended_asm_add_clobber (gcc_jit_extended_asm *ext_asm,\ const char *victim) Add `victim` to the list of registers clobbered by the extended ``asm`` statement. It must be non-NULL. See the `Clobbers and Scratch Registers <https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html#Clobbers-and-Scratch-Registers#>`_ section of the documentation of the C syntax. Statements with multiple clobbers will require multiple calls, one per clobber. For example: .. code-block:: c gcc_jit_extended_asm_add_clobber (ext_asm, "r0"); gcc_jit_extended_asm_add_clobber (ext_asm, "cc"); gcc_jit_extended_asm_add_clobber (ext_asm, "memory"); A :type:`gcc_jit_extended_asm` is a :type:`gcc_jit_object` "owned" by the block's context. The following upcast is available: .. function:: gcc_jit_object *\ gcc_jit_extended_asm_as_object (gcc_jit_extended_asm *ext_asm) Upcast from extended ``asm`` to object. Adding top-level assembler statements ************************************* In addition to creating extended ``asm`` instructions within a function, there is support for creating "top-level" assembler statements, outside of any function. .. function:: void \ gcc_jit_context_add_top_level_asm (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ const char *asm_stmts) Create a set of top-level asm statements, analogous to those created by GCC's "basic" ``asm`` syntax in C at file scope. For example, to create the equivalent of: .. literalinclude:: ../../../testsuite/jit.dg/test-asm.c :start-after: // Quote from here in docs/topics/asm.rst: example 5: C :end-before: // Quote up to here in docs/topics/asm.rst: example 5: C :language: c the following API calls could be used: .. literalinclude:: ../../../testsuite/jit.dg/test-asm.c :start-after: /* Quote from here in docs/topics/asm.rst: example 5: jit. */ :end-before: /* Quote up to here in docs/topics/asm.rst: example 5: jit. */ :language: c --- ### Gcc Jit Docs Topics Compatibility (`gcc/jit/docs/topics/compatibility.rst`) .. Copyright (C) 2015-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c ABI and API compatibility ========================= The libgccjit developers strive for ABI and API backward-compatibility: programs built against libgccjit.so stand a good chance of running without recompilation against newer versions of libgccjit.so, and ought to recompile without modification against newer versions of libgccjit.h. .. note:: The libgccjit++.h C++ API is more experimental, and less locked-down at this time. API compatibility is achieved by extending the API rather than changing it. For ABI compatibility, we avoid bumping the SONAME, and instead use symbol versioning to tag each symbol, so that a binary linked against libgccjit.so is tagged according to the symbols that it uses. For example, :func:`gcc_jit_context_add_command_line_option` was added in ``LIBGCCJIT_ABI_1``. If a client program uses it, this can be detected from metadata by using ``objdump``: .. code-block:: bash $ objdump -p testsuite/jit/test-extra-options.c.exe | tail -n 8 Version References: required from libgccjit.so.0: 0x00824161 0x00 04 LIBGCCJIT_ABI_1 0x00824160 0x00 03 LIBGCCJIT_ABI_0 required from libc.so.6: You can see the symbol tags provided by libgccjit.so using ``objdump``: .. code-block:: bash $ objdump -p libgccjit.so | less [...snip...] Version definitions: 1 0x01 0x0ff81f20 libgccjit.so.0 2 0x00 0x00824160 LIBGCCJIT_ABI_0 3 0x00 0x00824161 LIBGCCJIT_ABI_1 LIBGCCJIT_ABI_0 [...snip...] Programmatically checking version ********************************* Client code can programmatically check libgccjit version using: .. function:: int gcc_jit_version_major (void) Return libgccjit major version. This is analogous to __GNUC__ in C code. .. function:: int gcc_jit_version_minor (void) Return libgccjit minor version. This is analogous to __GNUC_MINOR__ in C code. .. function:: int gcc_jit_version_patchlevel (void) Return libgccjit patchlevel version. This is analogous to __GNUC_PATCHLEVEL__ in C code. .. note:: These entry points has been added with ``LIBGCCJIT_ABI_13`` (see below). ABI symbol tags *************** The initial release of libgccjit (in gcc 5.1) did not use symbol versioning. Newer releases use the following tags. .. _LIBGCCJIT_ABI_0: ``LIBGCCJIT_ABI_0`` ------------------- All entrypoints in the initial release of libgccjit are tagged with ``LIBGCCJIT_ABI_0``, to signify the transition to symbol versioning. Binaries built against older copies of ``libgccjit.so`` should continue to work, with this being handled transparently by the linker (see `this post <https://gcc.gnu.org/ml/gcc-patches/2015-06/msg02126.html>`_) .. _LIBGCCJIT_ABI_1: ``LIBGCCJIT_ABI_1`` ------------------- ``LIBGCCJIT_ABI_1`` covers the addition of :func:`gcc_jit_context_add_command_line_option` .. _LIBGCCJIT_ABI_2: ``LIBGCCJIT_ABI_2`` ------------------- ``LIBGCCJIT_ABI_2`` covers the addition of :func:`gcc_jit_context_set_bool_allow_unreachable_blocks` .. _LIBGCCJIT_ABI_3: ``LIBGCCJIT_ABI_3`` ------------------- ``LIBGCCJIT_ABI_3`` covers the addition of switch statements via API entrypoints: * :func:`gcc_jit_block_end_with_switch` * :func:`gcc_jit_case_as_object` * :func:`gcc_jit_context_new_case` .. _LIBGCCJIT_ABI_4: ``LIBGCCJIT_ABI_4`` ------------------- ``LIBGCCJIT_ABI_4`` covers the addition of timers via API entrypoints: * :func:`gcc_jit_context_get_timer` * :func:`gcc_jit_context_set_timer` * :func:`gcc_jit_timer_new` * :func:`gcc_jit_timer_release` * :func:`gcc_jit_timer_push` * :func:`gcc_jit_timer_pop` * :func:`gcc_jit_timer_print` .. _LIBGCCJIT_ABI_5: ``LIBGCCJIT_ABI_5`` ------------------- ``LIBGCCJIT_ABI_5`` covers the addition of :func:`gcc_jit_context_set_bool_use_external_driver` .. _LIBGCCJIT_ABI_6: ``LIBGCCJIT_ABI_6`` ------------------- ``LIBGCCJIT_ABI_6`` covers the addition of :func:`gcc_jit_rvalue_set_bool_require_tail_call` .. _LIBGCCJIT_ABI_7: ``LIBGCCJIT_ABI_7`` ------------------- ``LIBGCCJIT_ABI_7`` covers the addition of :func:`gcc_jit_type_get_aligned` .. _LIBGCCJIT_ABI_8: ``LIBGCCJIT_ABI_8`` ------------------- ``LIBGCCJIT_ABI_8`` covers the addition of :func:`gcc_jit_type_get_vector` .. _LIBGCCJIT_ABI_9: ``LIBGCCJIT_ABI_9`` ------------------- ``LIBGCCJIT_ABI_9`` covers the addition of :func:`gcc_jit_function_get_address` .. _LIBGCCJIT_ABI_10: ``LIBGCCJIT_ABI_10`` -------------------- ``LIBGCCJIT_ABI_10`` covers the addition of :func:`gcc_jit_context_new_rvalue_from_vector` .. _LIBGCCJIT_ABI_11: ``LIBGCCJIT_ABI_11`` -------------------- ``LIBGCCJIT_ABI_11`` covers the addition of :func:`gcc_jit_context_add_driver_option` .. _LIBGCCJIT_ABI_12: ``LIBGCCJIT_ABI_12`` -------------------- ``LIBGCCJIT_ABI_12`` covers the addition of :func:`gcc_jit_context_new_bitfield` .. _LIBGCCJIT_ABI_13: ``LIBGCCJIT_ABI_13`` -------------------- ``LIBGCCJIT_ABI_13`` covers the addition of version functions via API entrypoints: * :func:`gcc_jit_version_major` * :func:`gcc_jit_version_minor` * :func:`gcc_jit_version_patchlevel` .. _LIBGCCJIT_ABI_14: ``LIBGCCJIT_ABI_14`` -------------------- ``LIBGCCJIT_ABI_14`` covers the addition of :func:`gcc_jit_global_set_initializer` .. _LIBGCCJIT_ABI_15: ``LIBGCCJIT_ABI_15`` -------------------- ``LIBGCCJIT_ABI_15`` covers the addition of API entrypoints for directly embedding assembler instructions: * :func:`gcc_jit_block_add_extended_asm` * :func:`gcc_jit_block_end_with_extended_asm_goto` * :func:`gcc_jit_extended_asm_as_object` * :func:`gcc_jit_extended_asm_set_volatile_flag` * :func:`gcc_jit_extended_asm_set_inline_flag` * :func:`gcc_jit_extended_asm_add_output_operand` * :func:`gcc_jit_extended_asm_add_input_operand` * :func:`gcc_jit_extended_asm_add_clobber` * :func:`gcc_jit_context_add_top_level_asm` .. _LIBGCCJIT_ABI_16: ``LIBGCCJIT_ABI_16`` -------------------- ``LIBGCCJIT_ABI_16`` covers the addition of reflection functions via API entrypoints: * :func:`gcc_jit_function_get_return_type` * :func:`gcc_jit_function_get_param_count` * :func:`gcc_jit_type_dyncast_array` * :func:`gcc_jit_type_is_bool` * :func:`gcc_jit_type_is_integral` * :func:`gcc_jit_type_is_pointer` * :func:`gcc_jit_type_is_struct` * :func:`gcc_jit_type_dyncast_vector` * :func:`gcc_jit_type_unqualified` * :func:`gcc_jit_type_dyncast_function_ptr_type` * :func:`gcc_jit_function_type_get_return_type` * :func:`gcc_jit_function_type_get_param_count` * :func:`gcc_jit_function_type_get_param_type` * :func:`gcc_jit_vector_type_get_num_units` * :func:`gcc_jit_vector_type_get_element_type` * :func:`gcc_jit_struct_get_field` * :func:`gcc_jit_struct_get_field_count` .. _LIBGCCJIT_ABI_17: ``LIBGCCJIT_ABI_17`` ----------------------- ``LIBGCCJIT_ABI_17`` covers the addition of an API entrypoint to set the thread-local storage model of a variable: * :func:`gcc_jit_lvalue_set_tls_model` .. _LIBGCCJIT_ABI_18: ``LIBGCCJIT_ABI_18`` ----------------------- ``LIBGCCJIT_ABI_18`` covers the addition of an API entrypoint to set the link section of a variable: * :func:`gcc_jit_lvalue_set_link_section` .. _LIBGCCJIT_ABI_19: ``LIBGCCJIT_ABI_19`` ----------------------- ``LIBGCCJIT_ABI_19`` covers the addition of API entrypoints to set the initial value of a global with an rvalue and to use constructors: * :func:`gcc_jit_context_new_array_constructor` * :func:`gcc_jit_context_new_struct_constructor` * :func:`gcc_jit_context_new_union_constructor` * :func:`gcc_jit_global_set_initializer_rvalue` .. _LIBGCCJIT_ABI_20: ``LIBGCCJIT_ABI_20`` ----------------------- ``LIBGCCJIT_ABI_20`` covers the addition of sized integer types, including 128-bit integers and helper functions for types: * :func:`gcc_jit_compatible_types` * :func:`gcc_jit_type_get_size` * :c:macro:`GCC_JIT_TYPE_UINT8_T` * :c:macro:`GCC_JIT_TYPE_UINT16_T` * :c:macro:`GCC_JIT_TYPE_UINT32_T` * :c:macro:`GCC_JIT_TYPE_UINT64_T` * :c:macro:`GCC_JIT_TYPE_UINT128_T` * :c:macro:`GCC_JIT_TYPE_INT8_T` * :c:macro:`GCC_JIT_TYPE_INT16_T` * :c:macro:`GCC_JIT_TYPE_INT32_T` * :c:macro:`GCC_JIT_TYPE_INT64_T` * :c:macro:`GCC_JIT_TYPE_INT128_T` .. _LIBGCCJIT_ABI_21: ``LIBGCCJIT_ABI_21`` -------------------- ``LIBGCCJIT_ABI_21`` covers the addition of an API entrypoint to bitcast a value from one type to another: * :func:`gcc_jit_context_new_bitcast` .. _LIBGCCJIT_ABI_22: ``LIBGCCJIT_ABI_22`` -------------------- ``LIBGCCJIT_ABI_22`` covers the addition of an API entrypoint to set the register name of a variable: * :func:`gcc_jit_lvalue_set_register_name` .. _LIBGCCJIT_ABI_23: ``LIBGCCJIT_ABI_23`` -------------------- ``LIBGCCJIT_ABI_23`` covers the addition of an API entrypoint to hide stderr logs: * :func:`gcc_jit_context_set_bool_print_errors_to_stderr` .. _LIBGCCJIT_ABI_24: ``LIBGCCJIT_ABI_24`` -------------------- ``LIBGCCJIT_ABI_24`` covers the addition of functions to get and set the alignment of a variable: * :func:`gcc_jit_lvalue_set_alignment` * :func:`gcc_jit_lvalue_get_alignment` .. _LIBGCCJIT_ABI_25: ``LIBGCCJIT_ABI_25`` -------------------- ``LIBGCCJIT_ABI_25`` covers the addition of :func:`gcc_jit_type_get_restrict` .. _LIBGCCJIT_ABI_26: ``LIBGCCJIT_ABI_26`` -------------------- ``LIBGCCJIT_ABI_26`` covers the addition of functions to set attributes on functions and variables: * :func:`gcc_jit_function_add_attribute` * :func:`gcc_jit_function_add_string_attribute` * :func:`gcc_jit_function_add_integer_array_attribute` * :func:`gcc_jit_lvalue_add_string_attribute` .. _LIBGCCJIT_ABI_27: ``LIBGCCJIT_ABI_27`` -------------------- ``LIBGCCJIT_ABI_27`` covers the addition of :func:`gcc_jit_context_new_sizeof` .. _LIBGCCJIT_ABI_28: ``LIBGCCJIT_ABI_28`` -------------------- ``LIBGCCJIT_ABI_28`` covers the addition of :func:`gcc_jit_context_new_alignof` .. _LIBGCCJIT_ABI_29: ``LIBGCCJIT_ABI_29`` -------------------- ``LIBGCCJIT_ABI_29`` covers the addition of :func:`gcc_jit_global_set_readonly` .. _LIBGCCJIT_ABI_30: ``LIBGCCJIT_ABI_30`` -------------------- ``LIBGCCJIT_ABI_30`` covers the addition of :func:`gcc_jit_context_convert_vector` .. _LIBGCCJIT_ABI_31: ``LIBGCCJIT_ABI_31`` -------------------- ``LIBGCCJIT_ABI_31`` covers the addition of functions to manipulate vectors: * :func:`gcc_jit_context_new_rvalue_vector_perm` * :func:`gcc_jit_context_new_vector_access` .. _LIBGCCJIT_ABI_32: ``LIBGCCJIT_ABI_32`` -------------------- ``LIBGCCJIT_ABI_32`` covers the addition of a function to get target builtins: * :func:`gcc_jit_context_get_target_builtin_function` .. _LIBGCCJIT_ABI_33: ``LIBGCCJIT_ABI_33`` -------------------- ``LIBGCCJIT_ABI_33`` covers the addition of a function to create a new temporary variable: * :func:`gcc_jit_function_new_temp` .. _LIBGCCJIT_ABI_34: ``LIBGCCJIT_ABI_34`` -------------------- ``LIBGCCJIT_ABI_34`` covers the addition of * :func:`gcc_jit_context_set_output_ident` .. _LIBGCCJIT_ABI_35: ``LIBGCCJIT_ABI_35`` -------------------- ``LIBGCCJIT_ABI_35`` covers the addition of functions to query the target information: * :func:`gcc_jit_context_get_target_info` * :func:`gcc_jit_target_info_release` * :func:`gcc_jit_target_info_cpu_supports` * :func:`gcc_jit_target_info_arch` * :func:`gcc_jit_target_info_supports_target_dependent_type` .. _LIBGCCJIT_ABI_36: ``LIBGCCJIT_ABI_36`` -------------------- ``LIBGCCJIT_ABI_36`` covers the addition of * :func:`gcc_jit_context_set_abort_on_unsupported_target_builtin` .. _LIBGCCJIT_ABI_37: ``LIBGCCJIT_ABI_37`` -------------------- ``LIBGCCJIT_ABI_37`` covers the addition of :func:`gcc_jit_context_new_array_type_u64` .. _LIBGCCJIT_ABI_38: ``LIBGCCJIT_ABI_38`` -------------------- ``LIBGCCJIT_ABI_38`` covers the addition of * :func:`gcc_jit_type_is_floating_point` --- ### Gcc Jit Docs Topics Compilation (`gcc/jit/docs/topics/compilation.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Compiling a context =================== Once populated, a :c:expr:`gcc_jit_context *` can be compiled to machine code, either in-memory via :c:func:`gcc_jit_context_compile` or to disk via :c:func:`gcc_jit_context_compile_to_file`. You can compile a context multiple times (using either form of compilation), although any errors that occur on the context will prevent any future compilation of that context. In-memory compilation ********************* .. function:: gcc_jit_result *\ gcc_jit_context_compile (gcc_jit_context *ctxt) This calls into GCC and builds the code, returning a `gcc_jit_result *`. If the result is non-NULL, the caller becomes responsible for calling :func:`gcc_jit_result_release` on it once they're done with it. .. type:: gcc_jit_result A `gcc_jit_result` encapsulates the result of compiling a context in-memory, and the lifetimes of any machine code functions or globals that are within the result. .. function:: void *\ gcc_jit_result_get_code (gcc_jit_result *result,\ const char *funcname) Locate a given function within the built machine code. Functions are looked up by name. For this to succeed, a function with a name matching `funcname` must have been created on `result`'s context (or a parent context) via a call to :func:`gcc_jit_context_new_function` with `kind` :macro:`GCC_JIT_FUNCTION_EXPORTED`: .. code-block:: c gcc_jit_context_new_function (ctxt, any_location, /* or NULL */ /* Required for func to be visible to gcc_jit_result_get_code: */ GCC_JIT_FUNCTION_EXPORTED, any_return_type, /* Must string-compare equal: */ funcname, /* etc */); If such a function is not found (or `result` or `funcname` are ``NULL``), an error message will be emitted on stderr and ``NULL`` will be returned. If the function is found, the result will need to be cast to a function pointer of the correct type before it can be called. Note that the resulting machine code becomes invalid after :func:`gcc_jit_result_release` is called on the :expr:`gcc_jit_result *`; attempting to call it after that may lead to a segmentation fault. .. function:: void *\ gcc_jit_result_get_global (gcc_jit_result *result,\ const char *name) Locate a given global within the built machine code. Globals are looked up by name. For this to succeed, a global with a name matching `name` must have been created on `result`'s context (or a parent context) via a call to :func:`gcc_jit_context_new_global` with `kind` :macro:`GCC_JIT_GLOBAL_EXPORTED`. If the global is found, the result will need to be cast to a pointer of the correct type before it can be called. This is a *pointer* to the global, so e.g. for an :expr:`int` this is an :expr:`int *`. For example, given an ``int foo;`` created this way: .. code-block:: c gcc_jit_lvalue *exported_global = gcc_jit_context_new_global (ctxt, any_location, /* or NULL */ GCC_JIT_GLOBAL_EXPORTED, int_type, "foo"); we can access it like this: .. code-block:: c int *ptr_to_foo = (int *)gcc_jit_result_get_global (result, "foo"); If such a global is not found (or `result` or `name` are ``NULL``), an error message will be emitted on stderr and ``NULL`` will be returned. Note that the resulting address becomes invalid after :func:`gcc_jit_result_release` is called on the :expr:`gcc_jit_result *`; attempting to use it after that may lead to a segmentation fault. .. function:: void\ gcc_jit_result_release (gcc_jit_result *result) Once we're done with the code, this unloads the built .so file. This cleans up the result; after calling this, it's no longer valid to use the result, or any code or globals that were obtained by calling :func:`gcc_jit_result_get_code` or :func:`gcc_jit_result_get_global` on it. Ahead-of-time compilation ************************* Although libgccjit is primarily aimed at just-in-time compilation, it can also be used for implementing more traditional ahead-of-time compilers, via the :c:func:`gcc_jit_context_compile_to_file` API entrypoint. For linking in object files, use :c:func:`gcc_jit_context_add_driver_option`. .. function:: void \ gcc_jit_context_compile_to_file (gcc_jit_context *ctxt, \ enum gcc_jit_output_kind output_kind,\ const char *output_path) Compile the :c:expr:`gcc_jit_context *` to a file of the given kind. :c:func:`gcc_jit_context_compile_to_file` ignores the suffix of ``output_path``, and insteads uses the given :c:enum:`gcc_jit_output_kind` to decide what to do. .. note:: This is different from the ``gcc`` program, which does make use of the suffix of the output file when determining what to do. .. enum:: gcc_jit_output_kind The available kinds of output are: .. list-table:: :header-rows: 1 * - Output kind - Typical suffix * - :c:macro:`GCC_JIT_OUTPUT_KIND_ASSEMBLER` - .s * - :c:macro:`GCC_JIT_OUTPUT_KIND_OBJECT_FILE` - .o * - :c:macro:`GCC_JIT_OUTPUT_KIND_DYNAMIC_LIBRARY` - .so or .dll * - :c:macro:`GCC_JIT_OUTPUT_KIND_EXECUTABLE` - None, or .exe .. c:macro:: GCC_JIT_OUTPUT_KIND_ASSEMBLER Compile the context to an assembler file. .. c:macro:: GCC_JIT_OUTPUT_KIND_OBJECT_FILE Compile the context to an object file. .. c:macro:: GCC_JIT_OUTPUT_KIND_DYNAMIC_LIBRARY Compile the context to a dynamic library. .. c:macro:: GCC_JIT_OUTPUT_KIND_EXECUTABLE Compile the context to an executable. Getting information about a target ********************************** You can query the target information by using the following API: .. function:: gcc_jit_target_info * \ gcc_jit_context_get_target_info (gcc_jit_context *ctxt) Compute the information about a target. If the result is non-NULL, the caller becomes responsible for calling :func:`gcc_jit_target_info_release` on it once they're done with it. .. function:: void \ gcc_jit_target_info_release (gcc_jit_target_info *info) This function releases all resources associated with the given target info. .. function:: int \ gcc_jit_target_info_cpu_supports (gcc_jit_target_info *info, const char *feature) Check if the specified target INFO supports the cpu FEATURE. .. function:: const char * \ gcc_jit_target_info_arch (gcc_jit_target_info *info) Get the architecture of the currently running CPU, e.g. the value of -march equivalent to -march=native. Ex.: znver3 The underlying buffer is only valid until the gcc_jit_target_info is released. .. function:: int \ gcc_jit_target_info_supports_target_dependent_type (gcc_jit_target_info *info, enum gcc_jit_types type) Check if the specified target INFO supports target-dependent types like 128-bit integers. The API entrypoints relating to the target info: * :c:func:`gcc_jit_context_get_target_info` * :c:func:`gcc_jit_target_info_release` * :c:func:`gcc_jit_target_info_cpu_supports` * :c:func:`gcc_jit_target_info_arch` * :c:func:`gcc_jit_target_info_supports_target_dependent_type` were added in :ref:`LIBGCCJIT_ABI_35`; you can test for their presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_TARGET_INFO_API --- ### Gcc Jit Docs Topics Contexts (`gcc/jit/docs/topics/contexts.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Compilation contexts ==================== .. type:: gcc_jit_context The top-level of the API is the :c:type:`gcc_jit_context` type. A :c:type:`gcc_jit_context` instance encapsulates the state of a compilation. You can set up options on it, and add types, functions and code. Invoking :c:func:`gcc_jit_context_compile` on it gives you a :c:type:`gcc_jit_result`. Lifetime-management ------------------- Contexts are the unit of lifetime-management within the API: objects have their lifetime bounded by the context they are created within, and cleanup of such objects is done for you when the context is released. .. function:: gcc_jit_context *gcc_jit_context_acquire (void) This function acquires a new :c:expr:`gcc_jit_context *` instance, which is independent of any others that may be present within this process. .. function:: void gcc_jit_context_release (gcc_jit_context *ctxt) This function releases all resources associated with the given context. Both the context itself and all of its :c:expr:`gcc_jit_object *` instances are cleaned up. It should be called exactly once on a given context. It is invalid to use the context or any of its "contextual" objects after calling this. .. code-block:: c gcc_jit_context_release (ctxt); .. function:: gcc_jit_context * gcc_jit_context_new_child_context (gcc_jit_context *parent_ctxt) Given an existing JIT context, create a child context. The child inherits a copy of all option-settings from the parent. The child can reference objects created within the parent, but not vice-versa. The lifetime of the child context must be bounded by that of the parent: you should release a child context before releasing the parent context. If you use a function from a parent context within a child context, you have to compile the parent context before you can compile the child context, and the gcc_jit_result of the parent context must outlive the gcc_jit_result of the child context. This allows caching of shared initializations. For example, you could create types and declarations of global functions in a parent context once within a process, and then create child contexts whenever a function or loop becomes hot. Each such child context can be used for JIT-compiling just one function or loop, but can reference types and helper functions created within the parent context. Contexts can be arbitrarily nested, provided the above rules are followed, but it's probably not worth going above 2 or 3 levels, and there will likely be a performance hit for such nesting. Thread-safety ------------- Instances of :c:expr:`gcc_jit_context *` created via :c:func:`gcc_jit_context_acquire` are independent from each other: only one thread may use a given context at once, but multiple threads could each have their own contexts without needing locks. Contexts created via :c:func:`gcc_jit_context_new_child_context` are related to their parent context. They can be partitioned by their ultimate ancestor into independent "family trees". Only one thread within a process may use a given "family tree" of such contexts at once, and if you're using multiple threads you should provide your own locking around entire such context partitions. .. _error-handling: Error-handling -------------- Various kinds of errors are possible when using the API, such as mismatched types in an assignment. You can only compile and get code from a context if no errors occur. Errors are printed on stderr and can be queried using :c:func:`gcc_jit_context_get_first_error`. They typically contain the name of the API entrypoint where the error occurred, and pertinent information on the problem: .. code-block:: console ./buggy-program: error: gcc_jit_block_add_assignment: mismatching types: assignment to i (type: int) from "hello world" (type: const char *) In general, if an error occurs when using an API entrypoint, the entrypoint returns NULL. You don't have to check everywhere for NULL results, since the API handles a NULL being passed in for any argument by issuing another error. This typically leads to a cascade of followup error messages, but is safe (albeit verbose). The first error message is usually the one to pay attention to, since it is likely to be responsible for all of the rest: .. function:: const char *\ gcc_jit_context_get_first_error (gcc_jit_context *ctxt) Returns the first error message that occurred on the context. The returned string is valid for the rest of the lifetime of the context. If no errors occurred, this will be NULL. If you are wrapping the C API for a higher-level language that supports exception-handling, you may instead be interested in the last error that occurred on the context, so that you can embed this in an exception: .. function:: const char *\ gcc_jit_context_get_last_error (gcc_jit_context *ctxt) Returns the last error message that occurred on the context. If no errors occurred, this will be NULL. If non-NULL, the returned string is only guaranteed to be valid until the next call to libgccjit relating to this context. Debugging --------- .. function:: void\ gcc_jit_context_dump_to_file (gcc_jit_context *ctxt,\ const char *path,\ int update_locations) To help with debugging: dump a C-like representation to the given path, describing what's been set up on the context. If "update_locations" is true, then also set up :type:`gcc_jit_location` information throughout the context, pointing at the dump file as if it were a source file. This may be of use in conjunction with :macro:`GCC_JIT_BOOL_OPTION_DEBUGINFO` to allow stepping through the code in a debugger. .. function:: void\ gcc_jit_context_set_logfile (gcc_jit_context *ctxt,\ FILE *logfile,\ int flags,\ int verbosity) To help with debugging; enable ongoing logging of the context's activity to the given file. For example, the following will enable logging to stderr. .. code-block:: c gcc_jit_context_set_logfile (ctxt, stderr, 0, 0); Examples of information logged include: * API calls * the various steps involved within compilation * activity on any :c:type:`gcc_jit_result` instances created by the context * activity within any child contexts An example of a log can be seen :ref:`here <example-of-log-file>`, though the precise format and kinds of information logged is subject to change. The caller remains responsible for closing `logfile`, and it must not be closed until all users are released. In particular, note that child contexts and :c:type:`gcc_jit_result` instances created by the context will use the logfile. There may a performance cost for logging. You can turn off logging on `ctxt` by passing `NULL` for `logfile`. Doing so only affects the context; it does not affect child contexts or :c:type:`gcc_jit_result` instances already created by the context. The parameters "flags" and "verbosity" are reserved for future expansion, and must be zero for now. To contrast the above: :c:func:`gcc_jit_context_dump_to_file` dumps the current state of a context to the given path, whereas :c:func:`gcc_jit_context_set_logfile` enables on-going logging of future activities on a context to the given `FILE *`. .. function:: void\ gcc_jit_context_dump_reproducer_to_file (gcc_jit_context *ctxt,\ const char *path) Write C source code into `path` that can be compiled into a self-contained executable (i.e. with libgccjit as the only dependency). The generated code will attempt to replay the API calls that have been made into the given context. This may be useful when debugging the library or client code, for reducing a complicated recipe for reproducing a bug into a simpler form. For example, consider client code that parses some source file into some internal representation, and then walks this IR, calling into libgccjit. If this encounters a bug, a call to `gcc_jit_context_dump_reproducer_to_file` will write out C code for a much simpler executable that performs the equivalent calls into libgccjit, without needing the client code and its data. Typically you need to supply :option:`-Wno-unused-variable` when compiling the generated file (since the result of each API call is assigned to a unique variable within the generated C source, and not all are necessarily then used). .. function:: void\ gcc_jit_context_enable_dump (gcc_jit_context *ctxt,\ const char *dumpname, \ char **out_ptr) Enable the dumping of a specific set of internal state from the compilation, capturing the result in-memory as a buffer. Parameter "dumpname" corresponds to the equivalent gcc command-line option, without the "-fdump-" prefix. For example, to get the equivalent of :option:`-fdump-tree-vrp1`, supply ``"tree-vrp1"``: .. code-block:: c static char *dump_vrp1; void create_code (gcc_jit_context *ctxt) { gcc_jit_context_enable_dump (ctxt, "tree-vrp1", &dump_vrp1); /* (other API calls omitted for brevity) */ } The context directly stores the dumpname as a ``(const char *)``, so the passed string must outlive the context. :func:`gcc_jit_context_compile` will capture the dump as a dynamically-allocated buffer, writing it to ``*out_ptr``. The caller becomes responsible for calling: .. code-block:: c free (*out_ptr) each time that :func:`gcc_jit_context_compile` is called. ``*out_ptr`` will be written to, either with the address of a buffer, or with ``NULL`` if an error occurred. .. warning:: This API entrypoint is likely to be less stable than the others. In particular, both the precise dumpnames, and the format and content of the dumps are subject to change. It exists primarily for writing the library's own test suite. Options ------- Options present in the initial release of libgccjit were handled using enums, whereas those added subsequently have their own per-option API entrypoints. Adding entrypoints for each new option means that client code that use the new options can be identified directly from binary metadata, which would not be possible if we instead extended the various ``enum gcc_jit_*_option``. String Options ************** .. function:: void gcc_jit_context_set_str_option(gcc_jit_context *ctxt, \ enum gcc_jit_str_option opt, \ const char *value) Set a string option of the context. .. enum:: gcc_jit_str_option The parameter ``value`` can be NULL. If non-NULL, the call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. .. macro:: GCC_JIT_STR_OPTION_PROGNAME The name of the program, for use as a prefix when printing error messages to stderr. If `NULL`, or default, "libgccjit.so" is used. .. macro:: GCC_JIT_STR_OPTION_SPECIAL_CHARS_IN_FUNC_NAMES Alphabetic ASCII characters and underscores are always valid. Numeric ASCII characters are always valid after the initial character of the string. Use this entrypoint to specify additional ASCII characters that are valid throughout function names (ex.: ".$"). Boolean options *************** .. function:: void gcc_jit_context_set_bool_option(gcc_jit_context *ctxt, \ enum gcc_jit_bool_option opt, \ int value) Set a boolean option of the context. Zero is "false" (the default), non-zero is "true". .. enum:: gcc_jit_bool_option .. macro:: GCC_JIT_BOOL_OPTION_DEBUGINFO If true, :func:`gcc_jit_context_compile` will attempt to do the right thing so that if you attach a debugger to the process, it will be able to inspect variables and step through your code. Note that you can't step through code unless you set up source location information for the code (by creating and passing in :type:`gcc_jit_location` instances). .. macro:: GCC_JIT_BOOL_OPTION_DUMP_INITIAL_TREE If true, :func:`gcc_jit_context_compile` will dump its initial "tree" representation of your code to stderr (before any optimizations). Here's some sample output (from the `square` example):: <statement_list 0x7f4875a62cc0 type <void_type 0x7f4875a64bd0 VOID align 8 symtab 0 alias set -1 canonical type 0x7f4875a64bd0 pointer_to_this <pointer_type 0x7f4875a64c78>> side-effects head 0x7f4875a761e0 tail 0x7f4875a761f8 stmts 0x7f4875a62d20 0x7f4875a62d00 stmt <label_expr 0x7f4875a62d20 type <void_type 0x7f4875a64bd0> side-effects arg 0 <label_decl 0x7f4875a79080 entry type <void_type 0x7f4875a64bd0> VOID file (null) line 0 col 0 align 1 context <function_decl 0x7f4875a77500 square>>> stmt <return_expr 0x7f4875a62d00 type <integer_type 0x7f4875a645e8 public SI size <integer_cst 0x7f4875a623a0 constant 32> unit size <integer_cst 0x7f4875a623c0 constant 4> align 32 symtab 0 alias set -1 canonical type 0x7f4875a645e8 precision 32 min <integer_cst 0x7f4875a62340 -2147483648> max <integer_cst 0x7f4875a62360 2147483647> pointer_to_this <pointer_type 0x7f4875a6b348>> side-effects arg 0 <modify_expr 0x7f4875a72a78 type <integer_type 0x7f4875a645e8> side-effects arg 0 <result_decl 0x7f4875a7a000 D.54> arg 1 <mult_expr 0x7f4875a72a50 type <integer_type 0x7f4875a645e8> arg 0 <parm_decl 0x7f4875a79000 i> arg 1 <parm_decl 0x7f4875a79000 i>>>>> .. macro:: GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE If true, :func:`gcc_jit_context_compile` will dump the "gimple" representation of your code to stderr, before any optimizations are performed. The dump resembles C code: .. code-block:: c square (signed int i) { signed int D.56; entry: D.56 = i * i; return D.56; } .. macro:: GCC_JIT_BOOL_OPTION_DUMP_GENERATED_CODE If true, :func:`gcc_jit_context_compile` will dump the final generated code to stderr, in the form of assembly language: .. code-block:: gas .file "fake.c" .text .globl square .type square, @function square: .LFB0: .cfi_startproc pushq %rbp .cfi_def_cfa_offset 16 .cfi_offset 6, -16 movq %rsp, %rbp .cfi_def_cfa_register 6 movl %edi, -4(%rbp) .L2: movl -4(%rbp), %eax imull -4(%rbp), %eax popq %rbp .cfi_def_cfa 7, 8 ret .cfi_endproc .LFE0: .size square, .-square .ident "GCC: (GNU) 4.9.0 20131023 (Red Hat 0.2)" .section .note.GNU-stack,"",@progbits .. macro:: GCC_JIT_BOOL_OPTION_DUMP_SUMMARY If true, :func:`gcc_jit_context_compile` will print information to stderr on the actions it is performing. .. macro:: GCC_JIT_BOOL_OPTION_DUMP_EVERYTHING If true, :func:`gcc_jit_context_compile` will dump copious amount of information on what it's doing to various files within a temporary directory. Use :macro:`GCC_JIT_BOOL_OPTION_KEEP_INTERMEDIATES` (see below) to see the results. The files are intended to be human-readable, but the exact files and their formats are subject to change. .. macro:: GCC_JIT_BOOL_OPTION_SELFCHECK_GC If true, libgccjit will aggressively run its garbage collector, to shake out bugs (greatly slowing down the compile). This is likely to only be of interest to developers *of* the library. It is used when running the selftest suite. .. macro:: GCC_JIT_BOOL_OPTION_KEEP_INTERMEDIATES If true, the :type:`gcc_jit_context` will not clean up intermediate files written to the filesystem, and will display their location on stderr. .. function:: void \ gcc_jit_context_set_bool_allow_unreachable_blocks (gcc_jit_context *ctxt, \ int bool_value) By default, libgccjit will issue an error about unreachable blocks within a function. This entrypoint can be used to disable that error. This entrypoint was added in :ref:`LIBGCCJIT_ABI_2`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_set_bool_allow_unreachable_blocks .. function:: void \ gcc_jit_context_set_bool_use_external_driver (gcc_jit_context *ctxt, \ int bool_value) libgccjit internally generates assembler, and uses "driver" code for converting it to other formats (e.g. shared libraries). By default, libgccjit will use an embedded copy of the driver code. This option can be used to instead invoke an external driver executable as a subprocess. This entrypoint was added in :ref:`LIBGCCJIT_ABI_5`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_set_bool_use_external_driver .. function:: void \ gcc_jit_context_set_bool_print_errors_to_stderr (gcc_jit_context *ctxt, \ int enabled) By default, libgccjit will print errors to stderr. This entrypoint can be used to disable the printing. This entrypoint was added in :ref:`LIBGCCJIT_ABI_23`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_set_bool_print_errors_to_stderr .. function:: void \ gcc_jit_context_set_abort_on_unsupported_target_builtin (gcc_jit_context *ctxt) By default, libgccjit will silently ignore when a target builtin has an unsupported type. This entrypoint can be used to make it abort when the specified context encounters such a target builtin. This entrypoint was added in :ref:`LIBGCCJIT_ABI_36`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_set_abort_on_unsupported_target_builtin Integer options *************** .. function:: void gcc_jit_context_set_int_option (gcc_jit_context *ctxt, \ enum gcc_jit_int_option opt, \ int value) Set an integer option of the context. .. enum:: gcc_jit_int_option There is just one integer option specified this way: .. macro:: GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL How much to optimize the code. Valid values are 0-3, corresponding to GCC's command-line options -O0 through -O3. The default value is 0 (unoptimized). Additional command-line options ******************************* .. function:: void gcc_jit_context_add_command_line_option (gcc_jit_context *ctxt,\ const char *optname) Add an arbitrary gcc command-line option to the context, for use by :func:`gcc_jit_context_compile` and :func:`gcc_jit_context_compile_to_file`. If you need to add more than one such option, each should be done via its own call to `gcc_jit_context_add_command_line_option`. The parameter ``optname`` must be non-NULL. The underlying buffer is copied, so that it does not need to outlive the call. Extra options added by calls to `gcc_jit_context_add_command_line_option` are applied *after* the regular options above, potentially overriding them. Options from parent contexts are inherited by child contexts; options from the parent are applied *before* those from the child. For example: .. code-block:: c gcc_jit_context_add_command_line_option (ctxt, "-ffast-math"); gcc_jit_context_add_command_line_option (ctxt, "-fverbose-asm"); Note that only some options are likely to be meaningful; there is no "frontend" within libgccjit, so typically only those affecting optimization and code-generation are likely to be useful. This entrypoint was added in :ref:`LIBGCCJIT_ABI_1`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_add_command_line_option .. function:: void gcc_jit_context_add_driver_option (gcc_jit_context *ctxt,\ const char *optname) Add an arbitrary gcc driver option to the context, for use by :func:`gcc_jit_context_compile` and :func:`gcc_jit_context_compile_to_file`. The parameter ``optname`` must be non-NULL. The underlying buffer is copied, so that it does not need to outlive the call. If you need to add more than one such option, each should be done via its own call to `gcc_jit_context_add_driver_option`. Extra options added by calls to `gcc_jit_context_add_driver_option` are applied *after* all other options potentially overriding them. Options from parent contexts are inherited by child contexts; options from the parent are applied *before* those from the child. For example: .. code-block:: c gcc_jit_context_add_driver_option (ctxt, "-lm"); gcc_jit_context_add_driver_option (ctxt, "-fuse-linker-plugin"); gcc_jit_context_add_driver_option (ctxt, "obj.o"); gcc_jit_context_add_driver_option (ctxt, "-L."); gcc_jit_context_add_driver_option (ctxt, "-lwhatever"); Note that only some options are likely to be meaningful; there is no "frontend" within libgccjit, so typically only those affecting assembler and linker are likely to be useful. This entrypoint was added in :ref:`LIBGCCJIT_ABI_11`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_add_driver_option Output options ************** .. function:: void gcc_jit_context_set_output_ident (gcc_jit_context *ctxt,\ const char* output_ident) Set the identifier to write in the .comment section of the output file to ``output_ident``. The parameter ``output_ident`` must be non-NULL. This only works on some target, as you can see here: https://gcc.gnu.org/onlinedocs/cpp/Other-Directives.html Analogous to: .. code-block:: c #ident "My comment" in C. This entrypoint was added in :ref:`LIBGCCJIT_ABI_34`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_set_output_ident --- ### Gcc Jit Docs Topics Expressions (`gcc/jit/docs/topics/expressions.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Expressions =========== Rvalues ------- .. type:: gcc_jit_rvalue A :c:type:`gcc_jit_rvalue` is an expression that can be computed. It can be simple, e.g.: * an integer value e.g. `0` or `42` * a string literal e.g. `"Hello world"` * a variable e.g. `i`. These are also lvalues (see below). or compound e.g.: * a unary expression e.g. `!cond` * a binary expression e.g. `(a + b)` * a function call e.g. `get_distance (&player_ship, &target)` * etc. Every rvalue has an associated type, and the API will check to ensure that types match up correctly (otherwise the context will emit an error). .. function:: gcc_jit_type *gcc_jit_rvalue_get_type (gcc_jit_rvalue *rvalue) Get the type of this rvalue. .. function:: gcc_jit_object *gcc_jit_rvalue_as_object (gcc_jit_rvalue *rvalue) Upcast the given rvalue to be an object. Simple expressions ****************** .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_rvalue_from_int (gcc_jit_context *ctxt, \ gcc_jit_type *numeric_type, \ int value) Given a numeric type (integer or floating point), build an rvalue for the given constant :expr:`int` value. .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_rvalue_from_long (gcc_jit_context *ctxt, \ gcc_jit_type *numeric_type, \ long value) Given a numeric type (integer or floating point), build an rvalue for the given constant :expr:`long` value. .. function:: gcc_jit_rvalue *gcc_jit_context_zero (gcc_jit_context *ctxt, \ gcc_jit_type *numeric_type) Given a numeric type (integer or floating point), get the rvalue for zero. Essentially this is just a shortcut for: .. code-block:: c gcc_jit_context_new_rvalue_from_int (ctxt, numeric_type, 0) .. function:: gcc_jit_rvalue *gcc_jit_context_one (gcc_jit_context *ctxt, \ gcc_jit_type *numeric_type) Given a numeric type (integer or floating point), get the rvalue for one. Essentially this is just a shortcut for: .. code-block:: c gcc_jit_context_new_rvalue_from_int (ctxt, numeric_type, 1) .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_rvalue_from_double (gcc_jit_context *ctxt, \ gcc_jit_type *numeric_type, \ double value) Given a numeric type (integer or floating point), build an rvalue for the given constant :expr:`double` value. .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_rvalue_from_ptr (gcc_jit_context *ctxt, \ gcc_jit_type *pointer_type, \ void *value) Given a pointer type, build an rvalue for the given address. .. function:: gcc_jit_rvalue *gcc_jit_context_null (gcc_jit_context *ctxt, \ gcc_jit_type *pointer_type) Given a pointer type, build an rvalue for ``NULL``. Essentially this is just a shortcut for: .. code-block:: c gcc_jit_context_new_rvalue_from_ptr (ctxt, pointer_type, NULL) .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_string_literal (gcc_jit_context *ctxt, \ const char *value) Generate an rvalue for the given NIL-terminated string, of type :c:data:`GCC_JIT_TYPE_CONST_CHAR_PTR`. The parameter ``value`` must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_sizeof (gcc_jit_context *ctxt, \ gcc_jit_type *type) Generate an rvalue that is equal to the size of ``type``. The parameter ``type`` must be non-NULL. This is equivalent to this C code: .. code-block:: c sizeof (type) .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_alignof (gcc_jit_context *ctxt, \ gcc_jit_type *type) Generate an rvalue that is equal to the alignment of ``type``. The parameter ``type`` must be non-NULL. This is equivalent to this C code: .. code-block:: c _Alignof (type) Constructor expressions *********************** The following functions make constructors for array, struct and union types. The constructor rvalue can be used for assignment to locals. It can be used to initialize global variables with :func:`gcc_jit_global_set_initializer_rvalue`. It can also be used as a temporary value for function calls and return values, but its address can't be taken. Note that arrays in libgccjit do not collapse to pointers like in C. I.e. if an array constructor is used as e.g. a return value, the whole array would be returned by value - array constructors can be assigned to array variables. The constructor can contain nested constructors. Note that a string literal rvalue can't be used to construct a char array; the latter needs one rvalue for each char. These entrypoints were added in :ref:`LIBGCCJIT_ABI_19`; you can test for their presence using: .. code-block:: c #ifdef LIBGCCJIT_HAVE_CTORS .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_array_constructor (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_type *type,\ size_t num_values,\ gcc_jit_rvalue **values) Create a constructor for an array as an rvalue. Returns NULL on error. ``values`` are copied and do not have to outlive the context. ``type`` specifies what the constructor will build and has to be an array. ``num_values`` specifies the number of elements in ``values`` and it can't have more elements than the array type. Each value in ``values`` sets the corresponding value in the array. If the array type itself has more elements than ``values``, the left-over elements will be zeroed. Each value in ``values`` need to be the same unqualified type as the array type's element type. If ``num_values`` is 0, the ``values`` parameter will be ignored and zero initialization will be used. This entrypoint was added in :ref:`LIBGCCJIT_ABI_19`; you can test for its presence using: .. code-block:: c #ifdef LIBGCCJIT_HAVE_CTORS .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_struct_constructor (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_type *type,\ size_t num_values,\ gcc_jit_field **fields,\ gcc_jit_rvalue **values) Create a constructor for a struct as an rvalue. Returns NULL on error. The two parameter arrays are copied and do not have to outlive the context. ``type`` specifies what the constructor will build and has to be a struct. ``num_values`` specifies the number of elements in ``values``. ``fields`` need to have the same length as ``values``, or be NULL. If ``fields`` is null, the values are applied in definition order. Otherwise, each field in ``fields`` specifies which field in the struct to set to the corresponding value in ``values``. ``fields`` and ``values`` are paired by index. The fields in ``fields`` have to be in definition order, but there can be gaps. Any field in the struct that is not specified in ``fields`` will be zeroed. The fields in ``fields`` need to be the same objects that were used to create the struct. Each value has to have the same unqualified type as the field it is applied to. A NULL value element in ``values`` is a shorthand for zero initialization of the corresponding field. If ``num_values`` is 0, the array parameters will be ignored and zero initialization will be used. This entrypoint was added in :ref:`LIBGCCJIT_ABI_19`; you can test for its presence using: .. code-block:: c #ifdef LIBGCCJIT_HAVE_CTORS .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_union_constructor (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_type *type,\ gcc_jit_field *field,\ gcc_jit_rvalue *value) Create a constructor for a union as an rvalue. Returns NULL on error. ``type`` specifies what the constructor will build and has to be an union. ``field`` specifies which field to set. If it is NULL, the first field in the union will be set.``field`` need to be the same object that were used to create the union. ``value`` specifies what value to set the corresponding field to. If ``value`` is NULL, zero initialization will be used. Each value has to have have the same unqualified type as the field it is applied to. This entrypoint was added in :ref:`LIBGCCJIT_ABI_19`; you can test for its presence using: .. code-block:: c #ifdef LIBGCCJIT_HAVE_CTORS Vector expressions ****************** .. function:: gcc_jit_rvalue * \ gcc_jit_context_new_rvalue_from_vector (gcc_jit_context *ctxt, \ gcc_jit_location *loc, \ gcc_jit_type *vec_type, \ size_t num_elements, \ gcc_jit_rvalue **elements) Build a vector rvalue from an array of elements. "vec_type" should be a vector type, created using :func:`gcc_jit_type_get_vector`. "num_elements" should match that of the vector type. This entrypoint was added in :ref:`LIBGCCJIT_ABI_10`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_new_rvalue_from_vector .. function:: gcc_jit_rvalue * \ gcc_jit_context_new_rvalue_vector_perm (gcc_jit_context *ctxt, \ gcc_jit_location *loc, \ gcc_jit_rvalue *elements1, \ gcc_jit_rvalue *elements2, \ gcc_jit_rvalue *mask); Build a permutation of two vectors. "elements1" and "elements2" should have the same type. The length of "mask" and "elements1" should be the same. The element type of "mask" should be integral. The size of the element type of "mask" and "elements1" should be the same. This entrypoint was added in :ref:`LIBGCCJIT_ABI_31`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_VECTOR_OPERATIONS Analogous to: .. code-block:: c __builtin_shuffle (elements1, elements2, mask) in C. Unary Operations **************** .. function:: gcc_jit_rvalue * \ gcc_jit_context_new_unary_op (gcc_jit_context *ctxt, \ gcc_jit_location *loc, \ enum gcc_jit_unary_op op, \ gcc_jit_type *result_type, \ gcc_jit_rvalue *rvalue) Build a unary operation out of an input rvalue. The parameter ``result_type`` must be a numeric type. .. enum:: gcc_jit_unary_op The available unary operations are: .. list-table:: :header-rows: 1 * - Unary Operation - C equivalent * - :c:macro:`GCC_JIT_UNARY_OP_MINUS` - `-(EXPR)` * - :c:macro:`GCC_JIT_UNARY_OP_BITWISE_NEGATE` - `~(EXPR)` * - :c:macro:`GCC_JIT_UNARY_OP_LOGICAL_NEGATE` - `!(EXPR)` * - :c:macro:`GCC_JIT_UNARY_OP_ABS` - `abs (EXPR)` .. c:macro:: GCC_JIT_UNARY_OP_MINUS Negate an arithmetic value; analogous to: .. code-block:: c -(EXPR) in C. .. c:macro:: GCC_JIT_UNARY_OP_BITWISE_NEGATE Bitwise negation of an integer value (one's complement); analogous to: .. code-block:: c ~(EXPR) in C. .. c:macro:: GCC_JIT_UNARY_OP_LOGICAL_NEGATE Logical negation of an arithmetic or pointer value; analogous to: .. code-block:: c !(EXPR) in C. .. c:macro:: GCC_JIT_UNARY_OP_ABS Absolute value of an arithmetic expression; analogous to: .. code-block:: c abs (EXPR) in C. Binary Operations ***************** .. function:: gcc_jit_rvalue *gcc_jit_context_new_binary_op (gcc_jit_context *ctxt, \ gcc_jit_location *loc, \ enum gcc_jit_binary_op op, \ gcc_jit_type *result_type, \ gcc_jit_rvalue *a, gcc_jit_rvalue *b) Build a binary operation out of two constituent rvalues. The parameter ``result_type`` must be a numeric type. .. enum:: gcc_jit_binary_op The available binary operations are: .. list-table:: :header-rows: 1 * - Binary Operation - C equivalent * - :c:macro:`GCC_JIT_BINARY_OP_PLUS` - `x + y` * - :c:macro:`GCC_JIT_BINARY_OP_MINUS` - `x - y` * - :c:macro:`GCC_JIT_BINARY_OP_MULT` - `x * y` * - :c:macro:`GCC_JIT_BINARY_OP_DIVIDE` - `x / y` * - :c:macro:`GCC_JIT_BINARY_OP_MODULO` - `x % y` * - :c:macro:`GCC_JIT_BINARY_OP_BITWISE_AND` - `x & y` * - :c:macro:`GCC_JIT_BINARY_OP_BITWISE_XOR` - `x ^ y` * - :c:macro:`GCC_JIT_BINARY_OP_BITWISE_OR` - `x | y` * - :c:macro:`GCC_JIT_BINARY_OP_LOGICAL_AND` - `x && y` * - :c:macro:`GCC_JIT_BINARY_OP_LOGICAL_OR` - `x || y` * - :c:macro:`GCC_JIT_BINARY_OP_LSHIFT` - `x << y` * - :c:macro:`GCC_JIT_BINARY_OP_RSHIFT` - `x >> y` .. c:macro:: GCC_JIT_BINARY_OP_PLUS Addition of arithmetic values; analogous to: .. code-block:: c (EXPR_A) + (EXPR_B) in C. For pointer addition, use :c:func:`gcc_jit_context_new_array_access`. .. c:macro:: GCC_JIT_BINARY_OP_MINUS Subtraction of arithmetic values; analogous to: .. code-block:: c (EXPR_A) - (EXPR_B) in C. .. c:macro:: GCC_JIT_BINARY_OP_MULT Multiplication of a pair of arithmetic values; analogous to: .. code-block:: c (EXPR_A) * (EXPR_B) in C. .. c:macro:: GCC_JIT_BINARY_OP_DIVIDE Quotient of division of arithmetic values; analogous to: .. code-block:: c (EXPR_A) / (EXPR_B) in C. The result type affects the kind of division: if the result type is integer-based, then the result is truncated towards zero, whereas a floating-point result type indicates floating-point division. .. c:macro:: GCC_JIT_BINARY_OP_MODULO Remainder of division of arithmetic values; analogous to: .. code-block:: c (EXPR_A) % (EXPR_B) in C. .. c:macro:: GCC_JIT_BINARY_OP_BITWISE_AND Bitwise AND; analogous to: .. code-block:: c (EXPR_A) & (EXPR_B) in C. .. c:macro:: GCC_JIT_BINARY_OP_BITWISE_XOR Bitwise exclusive OR; analogous to: .. code-block:: c (EXPR_A) ^ (EXPR_B) in C. .. c:macro:: GCC_JIT_BINARY_OP_BITWISE_OR Bitwise inclusive OR; analogous to: .. code-block:: c (EXPR_A) | (EXPR_B) in C. .. c:macro:: GCC_JIT_BINARY_OP_LOGICAL_AND Logical AND; analogous to: .. code-block:: c (EXPR_A) && (EXPR_B) in C. .. c:macro:: GCC_JIT_BINARY_OP_LOGICAL_OR Logical OR; analogous to: .. code-block:: c (EXPR_A) || (EXPR_B) in C. .. c:macro:: GCC_JIT_BINARY_OP_LSHIFT Left shift; analogous to: .. code-block:: c (EXPR_A) << (EXPR_B) in C. .. c:macro:: GCC_JIT_BINARY_OP_RSHIFT Right shift; analogous to: .. code-block:: c (EXPR_A) >> (EXPR_B) in C. Comparisons *********** .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_comparison (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ enum gcc_jit_comparison op,\ gcc_jit_rvalue *a, gcc_jit_rvalue *b) Build a boolean rvalue out of the comparison of two other rvalues. .. enum:: gcc_jit_comparison .. list-table:: :header-rows: 1 * - Comparison - C equivalent * - :c:macro:`GCC_JIT_COMPARISON_EQ` - `x == y` * - :c:macro:`GCC_JIT_COMPARISON_NE` - `x != y` * - :c:macro:`GCC_JIT_COMPARISON_LT` - `x < y` * - :c:macro:`GCC_JIT_COMPARISON_LE` - `x <= y` * - :c:macro:`GCC_JIT_COMPARISON_GT` - `x > y` * - :c:macro:`GCC_JIT_COMPARISON_GE` - `x >= y` Function calls ************** .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_call (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_function *func,\ int numargs , gcc_jit_rvalue **args) Given a function and the given table of argument rvalues, construct a call to the function, with the result as an rvalue. .. note:: :c:func:`gcc_jit_context_new_call` merely builds a :c:type:`gcc_jit_rvalue` i.e. an expression that can be evaluated, perhaps as part of a more complicated expression. The call *won't* happen unless you add a statement to a function that evaluates the expression. For example, if you want to call a function and discard the result (or to call a function with ``void`` return type), use :c:func:`gcc_jit_block_add_eval`: .. code-block:: c /* Add "(void)printf (arg0, arg1);". */ gcc_jit_block_add_eval ( block, NULL, gcc_jit_context_new_call ( ctxt, NULL, printf_func, 2, args)); .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_call_through_ptr (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_rvalue *fn_ptr,\ int numargs, \ gcc_jit_rvalue **args) Given an rvalue of function pointer type (e.g. from :c:func:`gcc_jit_context_new_function_ptr_type`), and the given table of argument rvalues, construct a call to the function pointer, with the result as an rvalue. .. note:: The same caveat as for :c:func:`gcc_jit_context_new_call` applies. .. function:: void\ gcc_jit_rvalue_set_bool_require_tail_call (gcc_jit_rvalue *call,\ int require_tail_call) Given an :c:type:`gcc_jit_rvalue` for a call created through :c:func:`gcc_jit_context_new_call` or :c:func:`gcc_jit_context_new_call_through_ptr`, mark/clear the call as needing tail-call optimization. The optimizer will attempt to optimize the call into a jump instruction; if it is unable to do do, an error will be emitted. This may be useful when implementing functions that use the continuation-passing style (e.g. for functional programming languages), in which every function "returns" by calling a "continuation" function pointer. This call must be guaranteed to be implemented as a jump, otherwise the program could consume an arbitrary amount of stack space as it executed. This entrypoint was added in :ref:`LIBGCCJIT_ABI_6`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_rvalue_set_bool_require_tail_call Function pointers ***************** Function pointers can be obtained: * from a :c:type:`gcc_jit_function` using :c:func:`gcc_jit_function_get_address`, or * from an existing function using :c:func:`gcc_jit_context_new_rvalue_from_ptr`, using a function pointer type obtained using :c:func:`gcc_jit_context_new_function_ptr_type`. Type-coercion ************* .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_cast (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_rvalue *rvalue,\ gcc_jit_type *type) Given an rvalue of T, construct another rvalue of another type. Currently only a limited set of conversions are possible: * int <-> float * int <-> bool * P* <-> Q*, for pointer types P and Q .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_bitcast (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_rvalue *rvalue,\ gcc_jit_type *type) Given an rvalue of T, bitcast it to another type, meaning that this will generate a new rvalue by interpreting the bits of ``rvalue`` to the layout of ``type``. The type of rvalue must be the same size as the size of ``type``. This entrypoint was added in :ref:`LIBGCCJIT_ABI_21`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_new_bitcast .. function:: gcc_jit_rvalue * gcc_jit_context_convert_vector (gcc_jit_context *ctxt, \ gcc_jit_location *loc, \ gcc_jit_rvalue *vector, \ gcc_jit_type *type) Given a vector rvalue, cast it to the type ``type``, doing an element-wise conversion. The number of elements in ``vector`` and ``type`` must match. The ``type`` must be a vector type. This entrypoint was added in :ref:`LIBGCCJIT_ABI_30`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_convert_vector Lvalues ------- .. type:: gcc_jit_lvalue An lvalue is something that can of the *left*-hand side of an assignment: a storage area (such as a variable). It is also usable as an rvalue, where the rvalue is computed by reading from the storage area. .. function:: gcc_jit_object *\ gcc_jit_lvalue_as_object (gcc_jit_lvalue *lvalue) Upcast an lvalue to be an object. .. function:: gcc_jit_rvalue *\ gcc_jit_lvalue_as_rvalue (gcc_jit_lvalue *lvalue) Upcast an lvalue to be an rvalue. .. function:: gcc_jit_rvalue *\ gcc_jit_lvalue_get_address (gcc_jit_lvalue *lvalue,\ gcc_jit_location *loc) Take the address of an lvalue; analogous to: .. code-block:: c &(EXPR) in C. .. function:: void\ gcc_jit_lvalue_set_tls_model (gcc_jit_lvalue *lvalue,\ enum gcc_jit_tls_model model) Make a variable a thread-local variable. The "model" parameter determines the thread-local storage model of the "lvalue": .. enum:: gcc_jit_tls_model .. c:macro:: GCC_JIT_TLS_MODEL_NONE Don't set the TLS model. .. c:macro:: GCC_JIT_TLS_MODEL_GLOBAL_DYNAMIC .. c:macro:: GCC_JIT_TLS_MODEL_LOCAL_DYNAMIC .. c:macro:: GCC_JIT_TLS_MODEL_INITIAL_EXEC .. c:macro:: GCC_JIT_TLS_MODEL_LOCAL_EXEC This is analogous to: .. code-block:: c _Thread_local int foo __attribute__ ((tls_model("MODEL"))); in C. This entrypoint was added in :ref:`LIBGCCJIT_ABI_17`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_lvalue_set_tls_model .. function:: void\ gcc_jit_lvalue_set_link_section (gcc_jit_lvalue *lvalue,\ const char *section_name) Set the link section of a variable. The parameter ``section_name`` must be non-NULL and must contain the leading dot. Analogous to: .. code-block:: c int variable __attribute__((section(".section"))); in C. This entrypoint was added in :ref:`LIBGCCJIT_ABI_18`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_lvalue_set_link_section .. function:: void\ gcc_jit_lvalue_set_register_name (gcc_jit_lvalue *lvalue,\ const char *reg_name); Set the register name of a variable. The parameter ``reg_name`` must be non-NULL. Analogous to: .. code-block:: c register int variable asm ("r12"); in C. This entrypoint was added in :ref:`LIBGCCJIT_ABI_22`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_lvalue_set_register_name .. function:: void\ gcc_jit_lvalue_set_alignment (gcc_jit_lvalue *lvalue,\ unsigned bytes) Set the alignment of a variable, in bytes. Analogous to: .. code-block:: c int variable __attribute__((aligned (16))); in C. This entrypoint was added in :ref:`LIBGCCJIT_ABI_24`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_ALIGNMENT .. function:: unsigned\ gcc_jit_lvalue_get_alignment (gcc_jit_lvalue *lvalue) Return the alignment of a variable set by ``gcc_jit_lvalue_set_alignment``. Return 0 if the alignment was not set. Analogous to: .. code-block:: c _Alignof (variable) in C. This entrypoint was added in :ref:`LIBGCCJIT_ABI_24`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_ALIGNMENT Global variables **************** .. function:: gcc_jit_lvalue *\ gcc_jit_context_new_global (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ enum gcc_jit_global_kind kind,\ gcc_jit_type *type,\ const char *name) Add a new global variable of the given type and name to the context. The parameter ``type`` must be non-`void`. The parameter ``name`` must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. The "kind" parameter determines the visibility of the "global" outside of the :c:type:`gcc_jit_result`: .. enum:: gcc_jit_global_kind .. c:macro:: GCC_JIT_GLOBAL_EXPORTED Global is defined by the client code and is visible by name outside of this JIT context via :c:func:`gcc_jit_result_get_global` (and this value is required for the global to be accessible via that entrypoint). .. c:macro:: GCC_JIT_GLOBAL_INTERNAL Global is defined by the client code, but is invisible outside of it. Analogous to a "static" global within a .c file. Specifically, the variable will only be visible within this context and within child contexts. .. c:macro:: GCC_JIT_GLOBAL_IMPORTED Global is not defined by the client code; we're merely referring to it. Analogous to using an "extern" global from a header file. .. function:: gcc_jit_lvalue *\ gcc_jit_global_set_initializer (gcc_jit_lvalue *global,\ const void *blob,\ size_t num_bytes) Set an initializer for ``global`` using the memory content pointed by ``blob`` for ``num_bytes``. ``global`` must be an array of an integral type. Return the global itself. The parameter ``blob`` must be non-NULL. The call copies the memory pointed by ``blob`` for ``num_bytes`` bytes, so it is valid to pass in a pointer to an on-stack buffer. The content will be stored in the compilation unit and used as initialization value of the array. This entrypoint was added in :ref:`LIBGCCJIT_ABI_14`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_global_set_initializer .. function:: gcc_jit_lvalue *\ gcc_jit_global_set_initializer_rvalue (gcc_jit_lvalue *global,\ gcc_jit_rvalue *init_value) Set the initial value of a global with an rvalue. The rvalue needs to be a constant expression, e.g. no function calls. The global can't have the ``kind`` :c:macro:`GCC_JIT_GLOBAL_IMPORTED`. As a non-comprehensive example it is OK to do the equivalent of: .. code-block:: c int foo = 3 * 2; /* rvalue from gcc_jit_context_new_binary_op. */ int arr[] = {1,2,3,4}; /* rvalue from gcc_jit_context_new_constructor. */ int *bar = &arr[2] + 1; /* rvalue from nested "get address" of "array access". */ const int baz = 3; /* rvalue from gcc_jit_context_rvalue_from_int. */ int boz = baz; /* rvalue from gcc_jit_lvalue_as_rvalue. */ Use together with :c:func:`gcc_jit_context_new_struct_constructor`, :c:func:`gcc_jit_context_new_union_constructor`, :c:func:`gcc_jit_context_new_array_constructor` to initialize structs, unions and arrays. On success, returns the ``global`` parameter unchanged. Otherwise, ``NULL``. This entrypoint was added in :ref:`LIBGCCJIT_ABI_19`; you can test for its presence using: .. code-block:: c #ifdef LIBGCCJIT_HAVE_CTORS Variables ********* .. function:: void\ gcc_jit_lvalue_add_string_attribute (gcc_jit_lvalue *variable, enum gcc_jit_variable_attribute attribute, const char *value) Add an attribute ``attribute`` with value ``value`` to a variable ``variable``. This entrypoint was added in :ref:`LIBGCCJIT_ABI_26`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_ATTRIBUTES .. function:: void\ gcc_jit_global_set_readonly (gcc_jit_lvalue *global) Set the global variable as read-only, meaning you cannot assign to this variable. This entrypoint was added in :ref:`LIBGCCJIT_ABI_29`; you can test for its presence using: .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_global_set_readonly Working with pointers, structs and unions ----------------------------------------- .. function:: gcc_jit_lvalue *\ gcc_jit_rvalue_dereference (gcc_jit_rvalue *rvalue,\ gcc_jit_location *loc) Given an rvalue of pointer type ``T *``, dereferencing the pointer, getting an lvalue of type ``T``. Analogous to: .. code-block:: c *(EXPR) in C. Field access is provided separately for both lvalues and rvalues. .. function:: gcc_jit_lvalue *\ gcc_jit_lvalue_access_field (gcc_jit_lvalue *struct_,\ gcc_jit_location *loc,\ gcc_jit_field *field) Given an lvalue of struct or union type, access the given field, getting an lvalue of the field's type. Analogous to: .. code-block:: c (EXPR).field = ...; in C. .. function:: gcc_jit_rvalue *\ gcc_jit_rvalue_access_field (gcc_jit_rvalue *struct_,\ gcc_jit_location *loc,\ gcc_jit_field *field) Given an rvalue of struct or union type, access the given field as an rvalue. Analogous to: .. code-block:: c (EXPR).field in C. .. function:: gcc_jit_lvalue *\ gcc_jit_rvalue_dereference_field (gcc_jit_rvalue *ptr,\ gcc_jit_location *loc,\ gcc_jit_field *field) Given an rvalue of pointer type ``T *`` where T is of struct or union type, access the given field as an lvalue. Analogous to: .. code-block:: c (EXPR)->field in C, itself equivalent to ``(*EXPR).FIELD``. .. function:: gcc_jit_lvalue *\ gcc_jit_context_new_array_access (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_rvalue *ptr,\ gcc_jit_rvalue *index) Given an rvalue of pointer type ``T *``, get at the element `T` at the given index, using standard C array indexing rules i.e. each increment of ``index`` corresponds to ``sizeof(T)`` bytes. Analogous to: .. code-block:: c PTR[INDEX] in C (or, indeed, to ``PTR + INDEX``). .. function:: gcc_jit_lvalue *\ gcc_jit_context_new_vector_access (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_rvalue *vector,\ gcc_jit_rvalue *index) Given an rvalue of vector type ``T __attribute__ ((__vector_size__ (SIZE)))``, get the element `T` at the given index. This entrypoint was added in :ref:`LIBGCCJIT_ABI_31`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_VECTOR_OPERATIONS Analogous to: .. code-block:: c VECTOR[INDEX] in C. --- ### Gcc Jit Docs Topics Function Pointers (`gcc/jit/docs/topics/function-pointers.rst`) .. Copyright (C) 2017-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Function pointers ================= You can generate calls that use a function pointer via :c:func:`gcc_jit_context_new_call_through_ptr`. To do requires a :c:type:`gcc_jit_rvalue` of the correct function pointer type. Function pointers for a :c:type:`gcc_jit_function` can be obtained via :c:func:`gcc_jit_function_get_address`. .. function:: gcc_jit_rvalue *\ gcc_jit_function_get_address (gcc_jit_function *fn,\ gcc_jit_location *loc) Get the address of a function as an rvalue, of function pointer type. This entrypoint was added in :ref:`LIBGCCJIT_ABI_9`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_function_get_address Alternatively, given an existing function, you can obtain a pointer to it in :c:type:`gcc_jit_rvalue` form using :c:func:`gcc_jit_context_new_rvalue_from_ptr`, using a function pointer type obtained using :c:func:`gcc_jit_context_new_function_ptr_type`. Here's an example of creating a function pointer type corresponding to C's :expr:`void (*) (int, int, int)`: .. code-block:: c gcc_jit_type *void_type = gcc_jit_context_get_type (ctxt, GCC_JIT_TYPE_VOID); gcc_jit_type *int_type = gcc_jit_context_get_type (ctxt, GCC_JIT_TYPE_INT); /* Build the function ptr type. */ gcc_jit_type *param_types[3]; param_types[0] = int_type; param_types[1] = int_type; param_types[2] = int_type; gcc_jit_type *fn_ptr_type = gcc_jit_context_new_function_ptr_type (ctxt, NULL, void_type, 3, param_types, 0); .. function:: gcc_jit_type *\ gcc_jit_context_new_function_ptr_type (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_type *return_type,\ int num_params,\ gcc_jit_type **param_types,\ int is_variadic) Generate a :c:type:`gcc_jit_type` for a function pointer with the given return type and parameters. Each of `param_types` must be non-`void`; `return_type` may be `void`. --- ### Gcc Jit Docs Topics Functions (`gcc/jit/docs/topics/functions.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Creating and using functions ============================ Params ------ .. type:: gcc_jit_param A `gcc_jit_param` represents a parameter to a function. .. function:: gcc_jit_param *\ gcc_jit_context_new_param (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_type *type,\ const char *name) In preparation for creating a function, create a new parameter of the given type and name. The parameter ``type`` must be non-`void`. The parameter ``name`` must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. Parameters are lvalues, and thus are also rvalues (and objects), so the following upcasts are available: .. function:: gcc_jit_lvalue *\ gcc_jit_param_as_lvalue (gcc_jit_param *param) Upcasting from param to lvalue. .. function:: gcc_jit_rvalue *\ gcc_jit_param_as_rvalue (gcc_jit_param *param) Upcasting from param to rvalue. .. function:: gcc_jit_object *\ gcc_jit_param_as_object (gcc_jit_param *param) Upcasting from param to object. Functions --------- .. type:: gcc_jit_function A `gcc_jit_function` represents a function - either one that we're creating ourselves, or one that we're referencing. .. function:: gcc_jit_function *\ gcc_jit_context_new_function (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ enum gcc_jit_function_kind kind,\ gcc_jit_type *return_type,\ const char *name,\ int num_params,\ gcc_jit_param **params,\ int is_variadic) Create a gcc_jit_function with the given name and parameters. .. enum:: gcc_jit_function_kind This enum controls the kind of function created, and has the following values: .. macro:: GCC_JIT_FUNCTION_EXPORTED Function is defined by the client code and visible by name outside of the JIT. This value is required if you want to extract machine code for this function from a :type:`gcc_jit_result` via :func:`gcc_jit_result_get_code`. .. macro:: GCC_JIT_FUNCTION_INTERNAL Function is defined by the client code, but is invisible outside of the JIT. Analogous to a "static" function. .. macro:: GCC_JIT_FUNCTION_IMPORTED Function is not defined by the client code; we're merely referring to it. Analogous to using an "extern" function from a header file. .. macro:: GCC_JIT_FUNCTION_ALWAYS_INLINE Function is only ever inlined into other functions, and is invisible outside of the JIT. Analogous to prefixing with ``inline`` and adding ``__attribute__((always_inline))`` Inlining will only occur when the optimization level is above 0; when optimization is off, this is essentially the same as GCC_JIT_FUNCTION_INTERNAL. The parameter ``name`` must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. .. function:: gcc_jit_function *\ gcc_jit_context_get_builtin_function (gcc_jit_context *ctxt,\ const char *name) Get the :type:`gcc_jit_function` for the built-in function with the given name. For example: .. code-block:: c gcc_jit_function *fn = gcc_jit_context_get_builtin_function (ctxt, "__builtin_memcpy"); .. note:: Due to technical limitations with how libgccjit interacts with the insides of GCC, not all built-in functions are supported. More precisely, not all types are supported for parameters of built-in functions from libgccjit. Attempts to get a built-in function that uses such a parameter will lead to an error being emitted within the context. .. function:: gcc_jit_function *\ gcc_jit_context_get_target_builtin_function (gcc_jit_context *ctxt,\ const char *name) Get the :type:`gcc_jit_function` for the built-in function (sometimes called intrinsic functions) with the given name. For example: .. code-block:: c gcc_jit_function *fn = gcc_jit_context_get_target_builtin_function (ctxt, "__builtin_ia32_pmuldq512_mask"); .. note:: Due to technical limitations with how libgccjit interacts with the insides of GCC, not all built-in functions are supported. More precisely, not all types are supported for parameters of built-in functions from libgccjit. Attempts to get a built-in function that uses such a parameter will lead to an error being emitted within the context. .. function:: gcc_jit_object *\ gcc_jit_function_as_object (gcc_jit_function *func) Upcasting from function to object. .. function:: gcc_jit_param *\ gcc_jit_function_get_param (gcc_jit_function *func, int index) Get the param of the given index (0-based). .. function:: void \ gcc_jit_function_dump_to_dot (gcc_jit_function *func,\ const char *path) Emit the function in graphviz format to the given path. .. function:: gcc_jit_lvalue *\ gcc_jit_function_new_local (gcc_jit_function *func,\ gcc_jit_location *loc,\ gcc_jit_type *type,\ const char *name) Create a new local variable within the function, of the given type and name. The parameter ``type`` must be non-`void`. The parameter ``name`` must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. .. function:: gcc_jit_lvalue *\ gcc_jit_function_new_temp (gcc_jit_function *func,\ gcc_jit_location *loc,\ gcc_jit_type *type) Create a new local variable within the function, of the given type. This function is similar to :func:`gcc_jit_function_new_local`, but it is to be used for compiler-generated variables (as opposed to user-defined variables in the language to be compiled) and these variables won't show up in the debug info. The parameter ``type`` must be non-`void`. This entrypoint was added in :ref:`LIBGCCJIT_ABI_33`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_function_new_temp .. function:: size_t \ gcc_jit_function_get_param_count (gcc_jit_function *func) Get the number of parameters of the function. .. function:: gcc_jit_type *\ gcc_jit_function_get_return_type (gcc_jit_function *func) Get the return type of the function. The API entrypoints relating to getting info about parameters and return types: * :c:func:`gcc_jit_function_get_return_type` * :c:func:`gcc_jit_function_get_param_count` were added in :ref:`LIBGCCJIT_ABI_16`; you can test for their presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_REFLECTION .. type:: gcc_jit_case .. function:: void\ gcc_jit_function_add_attribute (gcc_jit_function *func, enum gcc_jit_fn_attribute attribute) Add an attribute ``attribute`` to a function ``func``. This is equivalent to the following code: .. code-block:: c __attribute__((always_inline)) This entrypoint was added in :ref:`LIBGCCJIT_ABI_26`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_ATTRIBUTES .. function:: void\ gcc_jit_function_add_string_attribute (gcc_jit_function *func, enum gcc_jit_fn_attribute attribute, const char *value) Add a string attribute ``attribute`` with value ``value`` to a function ``func``. This is equivalent to the following code: .. code-block:: c __attribute__ ((alias ("xxx"))) This entrypoint was added in :ref:`LIBGCCJIT_ABI_26`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_ATTRIBUTES .. function:: void\ gcc_jit_function_add_integer_array_attribute (gcc_jit_function *func, enum gcc_jit_fn_attribute attribute, const int *value, size_t length) Add an attribute ``attribute`` with ``length`` integer values ``value`` to a function ``func``. The integer values must be the same as you would write them in a C code. This is equivalent to the following code: .. code-block:: c __attribute__ ((nonnull (1, 2))) This entrypoint was added in :ref:`LIBGCCJIT_ABI_26`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_ATTRIBUTES Blocks ------ .. type:: gcc_jit_block A `gcc_jit_block` represents a basic block within a function i.e. a sequence of statements with a single entry point and a single exit point. The first basic block that you create within a function will be the entrypoint. Each basic block that you create within a function must be terminated, either with a conditional, a jump, a return, or a switch. It's legal to have multiple basic blocks that return within one function. .. function:: gcc_jit_block *\ gcc_jit_function_new_block (gcc_jit_function *func,\ const char *name) Create a basic block of the given name. The name may be NULL, but providing meaningful names is often helpful when debugging: it may show up in dumps of the internal representation, and in error messages. It is copied, so the input buffer does not need to outlive the call; you can pass in a pointer to an on-stack buffer, e.g.: .. code-block:: c for (pc = 0; pc < fn->fn_num_ops; pc++) { char buf[16]; sprintf (buf, "instr%i", pc); state.op_blocks[pc] = gcc_jit_function_new_block (state.fn, buf); } .. function:: gcc_jit_object *\ gcc_jit_block_as_object (gcc_jit_block *block) Upcast from block to object. .. function:: gcc_jit_function *\ gcc_jit_block_get_function (gcc_jit_block *block) Which function is this block within? Statements ---------- .. function:: void\ gcc_jit_block_add_eval (gcc_jit_block *block,\ gcc_jit_location *loc,\ gcc_jit_rvalue *rvalue) Add evaluation of an rvalue, discarding the result (e.g. a function call that "returns" void). This is equivalent to this C code: .. code-block:: c (void)expression; .. function:: void\ gcc_jit_block_add_assignment (gcc_jit_block *block,\ gcc_jit_location *loc,\ gcc_jit_lvalue *lvalue,\ gcc_jit_rvalue *rvalue) Add evaluation of an rvalue, assigning the result to the given lvalue. This is roughly equivalent to this C code: .. code-block:: c lvalue = rvalue; .. function:: void\ gcc_jit_block_add_assignment_op (gcc_jit_block *block,\ gcc_jit_location *loc,\ gcc_jit_lvalue *lvalue,\ enum gcc_jit_binary_op op,\ gcc_jit_rvalue *rvalue) Add evaluation of an rvalue, using the result to modify an lvalue. This is analogous to "+=" and friends: .. code-block:: c lvalue += rvalue; lvalue *= rvalue; lvalue /= rvalue; etc. For example: .. code-block:: c /* "i++" */ gcc_jit_block_add_assignment_op ( loop_body, NULL, i, GCC_JIT_BINARY_OP_PLUS, gcc_jit_context_one (ctxt, int_type)); .. function:: void\ gcc_jit_block_add_comment (gcc_jit_block *block,\ gcc_jit_location *loc,\ const char *text) Add a no-op textual comment to the internal representation of the code. It will be optimized away, but will be visible in the dumps seen via :macro:`GCC_JIT_BOOL_OPTION_DUMP_INITIAL_TREE` and :macro:`GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE`, and thus may be of use when debugging how your project's internal representation gets converted to the libgccjit IR. The parameter ``text`` must be non-NULL. It is copied, so the input buffer does not need to outlive the call. For example: .. code-block:: c char buf[100]; snprintf (buf, sizeof (buf), "op%i: %s", pc, opcode_names[op->op_opcode]); gcc_jit_block_add_comment (block, loc, buf); .. function:: void\ gcc_jit_block_end_with_conditional (gcc_jit_block *block,\ gcc_jit_location *loc,\ gcc_jit_rvalue *boolval,\ gcc_jit_block *on_true,\ gcc_jit_block *on_false) Terminate a block by adding evaluation of an rvalue, branching on the result to the appropriate successor block. This is roughly equivalent to this C code: .. code-block:: c if (boolval) goto on_true; else goto on_false; block, boolval, on_true, and on_false must be non-NULL. .. function:: void\ gcc_jit_block_end_with_jump (gcc_jit_block *block,\ gcc_jit_location *loc,\ gcc_jit_block *target) Terminate a block by adding a jump to the given target block. This is roughly equivalent to this C code: .. code-block:: c goto target; .. function:: void\ gcc_jit_block_end_with_return (gcc_jit_block *block,\ gcc_jit_location *loc,\ gcc_jit_rvalue *rvalue) Terminate a block by adding evaluation of an rvalue, returning the value. This is roughly equivalent to this C code: .. code-block:: c return expression; .. function:: void\ gcc_jit_block_end_with_void_return (gcc_jit_block *block,\ gcc_jit_location *loc) Terminate a block by adding a valueless return, for use within a function with "void" return type. This is equivalent to this C code: .. code-block:: c return; .. function:: void\ gcc_jit_block_end_with_switch (gcc_jit_block *block,\ gcc_jit_location *loc,\ gcc_jit_rvalue *expr,\ gcc_jit_block *default_block,\ int num_cases,\ gcc_jit_case **cases) Terminate a block by adding evaluation of an rvalue, then performing a multiway branch. This is roughly equivalent to this C code: .. code-block:: c switch (expr) { default: goto default_block; case C0.min_value ... C0.max_value: goto C0.dest_block; case C1.min_value ... C1.max_value: goto C1.dest_block; ...etc... case C[N - 1].min_value ... C[N - 1].max_value: goto C[N - 1].dest_block; } ``block``, ``expr``, ``default_block`` and ``cases`` must all be non-NULL. ``expr`` must be of the same integer type as all of the ``min_value`` and ``max_value`` within the cases. ``num_cases`` must be >= 0. The ranges of the cases must not overlap (or have duplicate values). The API entrypoints relating to switch statements and cases: * :c:func:`gcc_jit_block_end_with_switch` * :c:func:`gcc_jit_case_as_object` * :c:func:`gcc_jit_context_new_case` were added in :ref:`LIBGCCJIT_ABI_3`; you can test for their presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_SWITCH_STATEMENTS .. type:: gcc_jit_case A `gcc_jit_case` represents a case within a switch statement, and is created within a particular :c:type:`gcc_jit_context` using :c:func:`gcc_jit_context_new_case`. Each case expresses a multivalued range of integer values. You can express single-valued cases by passing in the same value for both `min_value` and `max_value`. .. function:: gcc_jit_case *\ gcc_jit_context_new_case (gcc_jit_context *ctxt,\ gcc_jit_rvalue *min_value,\ gcc_jit_rvalue *max_value,\ gcc_jit_block *dest_block) Create a new gcc_jit_case instance for use in a switch statement. `min_value` and `max_value` must be constants of an integer type, which must match that of the expression of the switch statement. `dest_block` must be within the same function as the switch statement. .. function:: gcc_jit_object *\ gcc_jit_case_as_object (gcc_jit_case *case_) Upcast from a case to an object. Here's an example of creating a switch statement: .. literalinclude:: ../../../testsuite/jit.dg/test-switch.c :start-after: /* Quote from here in docs/topics/functions.rst. */ :end-before: /* Quote up to here in docs/topics/functions.rst. */ :language: c See also :type:`gcc_jit_extended_asm` for entrypoints for adding inline assembler statements to a function. --- ### Gcc Jit Docs Topics Index (`gcc/jit/docs/topics/index.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. Topic Reference =============== .. toctree:: :maxdepth: 2 contexts.rst objects.rst types.rst expressions.rst functions.rst function-pointers.rst locations.rst compilation.rst compatibility.rst performance.rst asm.rst --- ### Gcc Jit Docs Topics Locations (`gcc/jit/docs/topics/locations.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Source Locations ================ .. type:: gcc_jit_location A `gcc_jit_location` encapsulates a source code location, so that you can (optionally) associate locations in your language with statements in the JIT-compiled code, allowing the debugger to single-step through your language. `gcc_jit_location` instances are optional: you can always pass NULL to any API entrypoint accepting one. You can construct them using :c:func:`gcc_jit_context_new_location`. You need to enable :c:macro:`GCC_JIT_BOOL_OPTION_DEBUGINFO` on the :c:type:`gcc_jit_context` for these locations to actually be usable by the debugger: .. code-block:: c gcc_jit_context_set_bool_option ( ctxt, GCC_JIT_BOOL_OPTION_DEBUGINFO, 1); .. function:: gcc_jit_location *\ gcc_jit_context_new_location (gcc_jit_context *ctxt,\ const char *filename,\ int line,\ int column) Create a `gcc_jit_location` instance representing the given source location. The parameter ``filename`` must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. Faking it --------- If you don't have source code for your internal representation, but need to debug, you can generate a C-like representation of the functions in your context using :c:func:`gcc_jit_context_dump_to_file()`: .. code-block:: c gcc_jit_context_dump_to_file (ctxt, "/tmp/something.c", 1 /* update_locations */); This will dump C-like code to the given path. If the `update_locations` argument is true, this will also set up `gcc_jit_location` information throughout the context, pointing at the dump file as if it were a source file, giving you *something* you can step through in the debugger. --- ### Gcc Jit Docs Topics Objects (`gcc/jit/docs/topics/objects.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Objects ======= .. type:: gcc_jit_object Almost every entity in the API (with the exception of :c:expr:`gcc_jit_context *` and :c:expr:`gcc_jit_result *`) is a "contextual" object, a :c:expr:`gcc_jit_object *` A JIT object: * is associated with a :c:expr:`gcc_jit_context *`. * is automatically cleaned up for you when its context is released so you don't need to manually track and cleanup all objects, just the contexts. Although the API is C-based, there is a form of class hierarchy, which looks like this:: +- gcc_jit_object +- gcc_jit_location +- gcc_jit_type +- gcc_jit_struct +- gcc_jit_field +- gcc_jit_function +- gcc_jit_block +- gcc_jit_rvalue +- gcc_jit_lvalue +- gcc_jit_param +- gcc_jit_case +- gcc_jit_extended_asm There are casting methods for upcasting from subclasses to parent classes. For example, :c:func:`gcc_jit_type_as_object`: .. code-block:: c gcc_jit_object *obj = gcc_jit_type_as_object (int_type); The object "base class" has the following operations: .. function:: gcc_jit_context *gcc_jit_object_get_context (gcc_jit_object *obj) Which context is "obj" within? .. function:: const char *gcc_jit_object_get_debug_string (gcc_jit_object *obj) Generate a human-readable description for the given object. For example, .. code-block:: c printf ("obj: %s\n", gcc_jit_object_get_debug_string (obj)); might give this text on stdout: .. code-block:: bash obj: 4.0 * (float)i .. note:: If you call this on an object, the `const char *` buffer is allocated and generated on the first call for that object, and the buffer will have the same lifetime as the object i.e. it will exist until the object's context is released. --- ### Gcc Jit Docs Topics Performance (`gcc/jit/docs/topics/performance.rst`) .. Copyright (C) 2015-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Performance =========== The timing API -------------- As of GCC 6, libgccjit exposes a timing API, for printing reports on how long was spent in different parts of code. You can create a :c:type:`gcc_jit_timer` instance, which will measure time spent since its creation. The timer maintains a stack of "timer items": as control flow moves through your code, you can push and pop named items relating to your code onto the stack, and the timer will account the time spent accordingly. You can also associate a timer with a :c:type:`gcc_jit_context`, in which case the time spent inside compilation will be subdivided. For example, the following code uses a timer, recording client items "create_code", "compile", and "running code": .. code-block:: c /* Create a timer. */ gcc_jit_timer *timer = gcc_jit_timer_new (); if (!timer) { error ("gcc_jit_timer_new failed"); return -1; } /* Let's repeatedly compile and run some code, accumulating it all into the timer. */ for (int i = 0; i < num_iterations; i++) { /* Create a context and associate it with the timer. */ gcc_jit_context *ctxt = gcc_jit_context_acquire (); if (!ctxt) { error ("gcc_jit_context_acquire failed"); return -1; } gcc_jit_context_set_timer (ctxt, timer); /* Populate the context, timing it as client item "create_code". */ gcc_jit_timer_push (timer, "create_code"); create_code (ctxt); gcc_jit_timer_pop (timer, "create_code"); /* Compile the context, timing it as client item "compile". */ gcc_jit_timer_push (timer, "compile"); result = gcc_jit_context_compile (ctxt); gcc_jit_timer_pop (timer, "compile"); /* Run the generated code, timing it as client item "running code". */ gcc_jit_timer_push (timer, "running code"); run_the_code (ctxt, result); gcc_jit_timer_pop (timer, "running code"); /* Clean up. */ gcc_jit_context_release (ctxt); gcc_jit_result_release (result); } /* Print the accumulated timings. */ gcc_jit_timer_print (timer, stderr); gcc_jit_timer_release (timer); giving output like this, showing the internal GCC items at the top, then client items, then the total:: Execution times (seconds) GCC items: phase setup : 0.29 (14%) usr 0.00 ( 0%) sys 0.32 ( 5%) wall 10661 kB (50%) ggc phase parsing : 0.02 ( 1%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 653 kB ( 3%) ggc phase finalize : 0.01 ( 1%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 0 kB ( 0%) ggc dump files : 0.02 ( 1%) usr 0.00 ( 0%) sys 0.01 ( 0%) wall 0 kB ( 0%) ggc callgraph construction : 0.02 ( 1%) usr 0.01 ( 6%) sys 0.01 ( 0%) wall 242 kB ( 1%) ggc callgraph optimization : 0.03 ( 2%) usr 0.00 ( 0%) sys 0.02 ( 0%) wall 142 kB ( 1%) ggc trivially dead code : 0.01 ( 1%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 0 kB ( 0%) ggc df scan insns : 0.01 ( 1%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 9 kB ( 0%) ggc df live regs : 0.01 ( 1%) usr 0.00 ( 0%) sys 0.01 ( 0%) wall 0 kB ( 0%) ggc inline parameters : 0.02 ( 1%) usr 0.00 ( 0%) sys 0.01 ( 0%) wall 82 kB ( 0%) ggc tree CFG cleanup : 0.01 ( 1%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 0 kB ( 0%) ggc tree PHI insertion : 0.01 ( 1%) usr 0.00 ( 0%) sys 0.02 ( 0%) wall 64 kB ( 0%) ggc tree SSA other : 0.01 ( 1%) usr 0.00 ( 0%) sys 0.01 ( 0%) wall 18 kB ( 0%) ggc expand : 0.01 ( 1%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 398 kB ( 2%) ggc jump : 0.01 ( 1%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 0 kB ( 0%) ggc loop init : 0.01 ( 0%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 67 kB ( 0%) ggc integrated RA : 0.02 ( 1%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 2468 kB (12%) ggc thread pro- & epilogue : 0.01 ( 1%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 162 kB ( 1%) ggc final : 0.01 ( 1%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 216 kB ( 1%) ggc rest of compilation : 1.37 (69%) usr 0.00 ( 0%) sys 1.13 (18%) wall 1391 kB ( 6%) ggc assemble JIT code : 0.01 ( 1%) usr 0.00 ( 0%) sys 4.04 (66%) wall 0 kB ( 0%) ggc load JIT result : 0.02 ( 1%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 0 kB ( 0%) ggc JIT client code : 0.00 ( 0%) usr 0.01 ( 6%) sys 0.00 ( 0%) wall 0 kB ( 0%) ggc Client items: create_code : 0.00 ( 0%) usr 0.01 ( 6%) sys 0.00 ( 0%) wall 0 kB ( 0%) ggc compile : 0.36 (18%) usr 0.15 (83%) sys 0.86 (14%) wall 14939 kB (70%) ggc running code : 0.00 ( 0%) usr 0.00 ( 0%) sys 0.00 ( 0%) wall 0 kB ( 0%) ggc TOTAL : 2.00 0.18 6.12 21444 kB The exact format is intended to be human-readable, and is subject to change. .. macro:: LIBGCCJIT_HAVE_TIMING_API The timer API was added to libgccjit in GCC 6. This macro is only defined in versions of libgccjit.h which have the timer API, and so can be used to guard code that may need to compile against earlier releases:: #ifdef LIBGCCJIT_HAVE_TIMING_API gcc_jit_timer *t = gcc_jit_timer_new (); gcc_jit_context_set_timer (ctxt, t); #endif .. type:: gcc_jit_timer .. function:: gcc_jit_timer * gcc_jit_timer_new(void) Create a :c:type:`gcc_jit_timer` instance, and start timing:: gcc_jit_timer *t = gcc_jit_timer_new (); This API entrypoint was added in :ref:`LIBGCCJIT_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_TIMING_API .. function:: void gcc_jit_timer_release(gcc_jit_timer *timer) Release a :c:type:`gcc_jit_timer` instance:: gcc_jit_timer_release (t); This should be called exactly once on a timer. This API entrypoint was added in :ref:`LIBGCCJIT_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_TIMING_API .. function:: void gcc_jit_context_set_timer(gcc_jit_context *ctxt, \ gcc_jit_timer *timer) Associate a :c:type:`gcc_jit_timer` instance with a context:: gcc_jit_context_set_timer (ctxt, t); A timer instance can be shared between multiple :c:type:`gcc_jit_context` instances. Timers have no locking, so if you have a multithreaded program, you must provide your own locks if more than one thread could be working with the same timer via timer-associated contexts. This API entrypoint was added in :ref:`LIBGCCJIT_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_TIMING_API .. function:: gcc_jit_timer *gcc_jit_context_get_timer(gcc_jit_context *ctxt) Get the timer associated with a context (if any). This API entrypoint was added in :ref:`LIBGCCJIT_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_TIMING_API .. function:: void gcc_jit_timer_push(gcc_jit_timer *timer, \ const char *item_name) Push the given item onto the timer's stack:: gcc_jit_timer_push (t, "running code"); run_the_code (ctxt, result); gcc_jit_timer_pop (t, "running code"); This API entrypoint was added in :ref:`LIBGCCJIT_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_TIMING_API .. function:: void gcc_jit_timer_pop(gcc_jit_timer *timer, \ const char *item_name) Pop the top item from the timer's stack. If "item_name" is provided, it must match that of the top item. Alternatively, ``NULL`` can be passed in, to suppress checking. This API entrypoint was added in :ref:`LIBGCCJIT_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_TIMING_API .. function:: void gcc_jit_timer_print(gcc_jit_timer *timer, \ FILE *f_out) Print timing information to the given stream about activity since the timer was started. This API entrypoint was added in :ref:`LIBGCCJIT_ABI_4`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_TIMING_API --- ### Gcc Jit Docs Topics Types (`gcc/jit/docs/topics/types.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: c Types ===== .. c:type:: gcc_jit_type gcc_jit_type represents a type within the library. .. function:: gcc_jit_object *gcc_jit_type_as_object (gcc_jit_type *type) Upcast a type to an object. Types can be created in several ways: * fundamental types can be accessed using :func:`gcc_jit_context_get_type`: .. code-block:: c gcc_jit_type *int_type = gcc_jit_context_get_type (ctxt, GCC_JIT_TYPE_INT); See :func:`gcc_jit_context_get_type` for the available types. * derived types can be accessed by using functions such as :func:`gcc_jit_type_get_pointer` and :func:`gcc_jit_type_get_const`: .. code-block:: c gcc_jit_type *const_int_star = gcc_jit_type_get_pointer (gcc_jit_type_get_const (int_type)); gcc_jit_type *int_const_star = gcc_jit_type_get_const (gcc_jit_type_get_pointer (int_type)); * by creating structures (see below). Standard types -------------- .. function:: gcc_jit_type *gcc_jit_context_get_type (gcc_jit_context *ctxt, \ enum gcc_jit_types type_) Access a specific type. The available types are: .. list-table:: :header-rows: 1 * - `enum gcc_jit_types` value - Meaning * - :c:data:`GCC_JIT_TYPE_VOID` - C's ``void`` type. * - :c:data:`GCC_JIT_TYPE_VOID_PTR` - C's ``void *``. * - :c:data:`GCC_JIT_TYPE_BOOL` - C++'s ``bool`` type; also C99's ``_Bool`` type, aka ``bool`` if using stdbool.h. * - :c:data:`GCC_JIT_TYPE_CHAR` - C's ``char`` (of some signedness) * - :c:data:`GCC_JIT_TYPE_SIGNED_CHAR` - C's ``signed char`` * - :c:data:`GCC_JIT_TYPE_UNSIGNED_CHAR` - C's ``unsigned char`` * - :c:data:`GCC_JIT_TYPE_SHORT` - C's ``short`` (signed) * - :c:data:`GCC_JIT_TYPE_UNSIGNED_SHORT` - C's ``unsigned short`` * - :c:data:`GCC_JIT_TYPE_INT` - C's ``int`` (signed) * - :c:data:`GCC_JIT_TYPE_UNSIGNED_INT` - C's ``unsigned int`` * - :c:data:`GCC_JIT_TYPE_LONG` - C's ``long`` (signed) * - :c:data:`GCC_JIT_TYPE_UNSIGNED_LONG` - C's ``unsigned long`` * - :c:data:`GCC_JIT_TYPE_LONG_LONG` - C99's ``long long`` (signed) * - :c:data:`GCC_JIT_TYPE_UNSIGNED_LONG_LONG` - C99's ``unsigned long long`` * - :c:data:`GCC_JIT_TYPE_UINT8_T` - C99's ``uint8_t`` * - :c:data:`GCC_JIT_TYPE_UINT16_T` - C99's ``uint16_t`` * - :c:data:`GCC_JIT_TYPE_UINT32_T` - C99's ``uint32_t`` * - :c:data:`GCC_JIT_TYPE_UINT64_T` - C99's ``uint64_t`` * - :c:data:`GCC_JIT_TYPE_UINT128_T` - C99's ``__uint128_t`` * - :c:data:`GCC_JIT_TYPE_INT8_T` - C99's ``int8_t`` * - :c:data:`GCC_JIT_TYPE_INT16_T` - C99's ``int16_t`` * - :c:data:`GCC_JIT_TYPE_INT32_T` - C99's ``int32_t`` * - :c:data:`GCC_JIT_TYPE_INT64_T` - C99's ``int64_t`` * - :c:data:`GCC_JIT_TYPE_INT128_T` - C99's ``__int128_t`` * - :c:data:`GCC_JIT_TYPE_FLOAT` - * - :c:data:`GCC_JIT_TYPE_BFLOAT16` - C's ``__bfloat16`` * - :c:data:`GCC_JIT_TYPE_DOUBLE` - * - :c:data:`GCC_JIT_TYPE_LONG_DOUBLE` - * - :c:data:`GCC_JIT_TYPE_FLOAT16` - * - :c:data:`GCC_JIT_TYPE_FLOAT32` - * - :c:data:`GCC_JIT_TYPE_FLOAT64` - * - :c:data:`GCC_JIT_TYPE_FLOAT128` - * - :c:data:`GCC_JIT_TYPE_CONST_CHAR_PTR` - C type: ``(const char *)`` * - :c:data:`GCC_JIT_TYPE_SIZE_T` - C's ``size_t`` type * - :c:data:`GCC_JIT_TYPE_FILE_PTR` - C type: ``(FILE *)`` * - :c:data:`GCC_JIT_TYPE_COMPLEX_FLOAT` - C99's ``_Complex float`` * - :c:data:`GCC_JIT_TYPE_COMPLEX_DOUBLE` - C99's ``_Complex double`` * - :c:data:`GCC_JIT_TYPE_COMPLEX_LONG_DOUBLE` - C99's ``_Complex long double`` .. function:: gcc_jit_type *\ gcc_jit_context_get_int_type (gcc_jit_context *ctxt, \ int num_bytes, int is_signed) Access the integer type of the given size. Pointers, `const`, and `volatile` --------------------------------- .. function:: gcc_jit_type *gcc_jit_type_get_pointer (gcc_jit_type *type) Given type "T", get type "T*". .. function:: gcc_jit_type *gcc_jit_type_get_const (gcc_jit_type *type) Given type "T", get type "const T". .. function:: gcc_jit_type *gcc_jit_type_get_volatile (gcc_jit_type *type) Given type "T", get type "volatile T". .. function:: gcc_jit_type *\ gcc_jit_context_new_array_type (gcc_jit_context *ctxt, \ gcc_jit_location *loc, \ gcc_jit_type *element_type, \ int num_elements) Given non-`void` type "T", get type "T[N]" (for a constant N). .. function:: gcc_jit_type *\ gcc_jit_context_new_array_type_u64 (gcc_jit_context *ctxt, \ gcc_jit_location *loc, \ gcc_jit_type *element_type, \ uint64_t num_elements) Given non-`void` type "T", get type "T[N]" (for a constant N). This is the same as gcc_jit_context_new_array_type, but the type of ``num_elements` different and thus allows creating bigger array types. This API entrypoint was added in :ref:`LIBGCCJIT_ABI_37`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_new_array_type_u64 .. function:: gcc_jit_type *\ gcc_jit_type_get_aligned (gcc_jit_type *type, \ size_t alignment_in_bytes) Given non-`void` type "T", get type: .. code-block:: c T __attribute__ ((aligned (ALIGNMENT_IN_BYTES))) The alignment must be a power of two. This entrypoint was added in :ref:`LIBGCCJIT_ABI_7`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_type_get_aligned Vector types ------------ .. function:: gcc_jit_type *\ gcc_jit_type_get_vector (gcc_jit_type *type, \ size_t num_units) Given type "T", get type: .. code-block:: c T __attribute__ ((vector_size (sizeof(T) * num_units)) T must be integral or floating point; num_units must be a power of two. This can be used to construct a vector type in which operations are applied element-wise. The compiler will automatically use SIMD instructions where possible. See: https://gcc.gnu.org/onlinedocs/gcc/Vector-Extensions.html For example, assuming 4-byte ``ints``, then: .. code-block:: c typedef int v4si __attribute__ ((vector_size (16))); can be obtained using: .. code-block:: c gcc_jit_type *int_type = gcc_jit_context_get_type (ctxt, GCC_JIT_TYPE_INT); gcc_jit_type *v4si_type = gcc_jit_type_get_vector (int_type, 4); This API entrypoint was added in :ref:`LIBGCCJIT_ABI_8`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_type_get_vector Vector rvalues can be generated using :func:`gcc_jit_context_new_rvalue_from_vector`. Structures and unions --------------------- .. c:type:: gcc_jit_struct A compound type analogous to a C `struct`. .. c:type:: gcc_jit_field A field within a :c:type:`gcc_jit_struct`. You can model C `struct` types by creating :c:type:`gcc_jit_struct` and :c:type:`gcc_jit_field` instances, in either order: * by creating the fields, then the structure. For example, to model: .. code-block:: c struct coord {double x; double y; }; you could call: .. code-block:: c gcc_jit_field *field_x = gcc_jit_context_new_field (ctxt, NULL, double_type, "x"); gcc_jit_field *field_y = gcc_jit_context_new_field (ctxt, NULL, double_type, "y"); gcc_jit_field *fields[2] = {field_x, field_y}; gcc_jit_struct *coord = gcc_jit_context_new_struct_type (ctxt, NULL, "coord", 2, fields); * by creating the structure, then populating it with fields, typically to allow modelling self-referential structs such as: .. code-block:: c struct node { int m_hash; struct node *m_next; }; like this: .. code-block:: c gcc_jit_type *node = gcc_jit_context_new_opaque_struct (ctxt, NULL, "node"); gcc_jit_type *node_ptr = gcc_jit_type_get_pointer (node); gcc_jit_field *field_hash = gcc_jit_context_new_field (ctxt, NULL, int_type, "m_hash"); gcc_jit_field *field_next = gcc_jit_context_new_field (ctxt, NULL, node_ptr, "m_next"); gcc_jit_field *fields[2] = {field_hash, field_next}; gcc_jit_struct_set_fields (node, NULL, 2, fields); .. function:: gcc_jit_field *\ gcc_jit_context_new_field (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_type *type,\ const char *name) Construct a new field, with the given type and name. The parameter ``type`` must be non-`void`. The parameter ``name`` must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. .. function:: gcc_jit_field *\ gcc_jit_context_new_bitfield (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_type *type,\ int width,\ const char *name) Construct a new bit field, with the given type width and name. The parameter ``name`` must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. The parameter ``type`` must be an integer type. The parameter ``width`` must be a positive integer that does not exceed the size of ``type``. This API entrypoint was added in :ref:`LIBGCCJIT_ABI_12`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_new_bitfield .. function:: gcc_jit_object *\ gcc_jit_field_as_object (gcc_jit_field *field) Upcast from field to object. .. function:: gcc_jit_struct *\ gcc_jit_context_new_struct_type (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ const char *name,\ int num_fields,\ gcc_jit_field **fields) Construct a new struct type, with the given name and fields. The parameter ``name`` must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. .. function:: gcc_jit_struct *\ gcc_jit_context_new_opaque_struct (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ const char *name) Construct a new struct type, with the given name, but without specifying the fields. The fields can be omitted (in which case the size of the struct is not known), or later specified using :c:func:`gcc_jit_struct_set_fields`. The parameter ``name`` must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. .. function:: gcc_jit_type *\ gcc_jit_struct_as_type (gcc_jit_struct *struct_type) Upcast from struct to type. .. function:: void\ gcc_jit_struct_set_fields (gcc_jit_struct *struct_type,\ gcc_jit_location *loc,\ int num_fields,\ gcc_jit_field **fields) Populate the fields of a formerly-opaque struct type. This can only be called once on a given struct type. .. function:: gcc_jit_type *\ gcc_jit_context_new_union_type (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ const char *name,\ int num_fields,\ gcc_jit_field **fields) Construct a new union type, with the given name and fields. The parameter ``name`` must be non-NULL. It is copied, so the input buffer does not need to outlive the call. Example of use: .. literalinclude:: ../../../testsuite/jit.dg/test-accessing-union.c :start-after: /* Quote from here in docs/topics/types.rst. */ :end-before: /* Quote up to here in docs/topics/types.rst. */ :language: c Function pointer types ---------------------- Function pointer types can be created using :c:func:`gcc_jit_context_new_function_ptr_type`. Reflection API -------------- .. function:: gcc_jit_type *\ gcc_jit_type_dyncast_array (gcc_jit_type *type) Get the element type of an array type or NULL if it's not an array. .. function:: int\ gcc_jit_type_is_bool (gcc_jit_type *type) Return non-zero if the type is a bool. .. function:: gcc_jit_function_type *\ gcc_jit_type_dyncast_function_ptr_type (gcc_jit_type *type) Return the function type if it is one or NULL. .. function:: gcc_jit_type *\ gcc_jit_function_type_get_return_type (gcc_jit_function_type *function_type) Given a function type, return its return type. .. function:: size_t\ gcc_jit_function_type_get_param_count (gcc_jit_function_type *function_type) Given a function type, return its number of parameters. .. function:: gcc_jit_type *\ gcc_jit_function_type_get_param_type (gcc_jit_function_type *function_type,\ size_t index) Given a function type, return the type of the specified parameter. .. function:: int\ gcc_jit_type_is_integral (gcc_jit_type *type) Return non-zero if the type is an integral. .. function:: int\ gcc_jit_type_is_floating_point (gcc_jit_type *type) Return non-zero if the type is floating point. This entrypoint was added in :ref:`LIBGCCJIT_ABI_38`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_type_is_floating_point .. function:: gcc_jit_type *\ gcc_jit_type_is_pointer (gcc_jit_type *type) Return the type pointed by the pointer type or NULL if it's not a pointer. .. function:: gcc_jit_vector_type *\ gcc_jit_type_dyncast_vector (gcc_jit_type *type) Given a type, return a dynamic cast to a vector type or NULL. .. function:: gcc_jit_struct *\ gcc_jit_type_is_struct (gcc_jit_type *type) Given a type, return a dynamic cast to a struct type or NULL. .. function:: size_t\ gcc_jit_vector_type_get_num_units (gcc_jit_vector_type *vector_type) Given a vector type, return the number of units it contains. .. function:: gcc_jit_type *\ gcc_jit_vector_type_get_element_type (gcc_jit_vector_type * vector_type) Given a vector type, return the type of its elements. .. function:: gcc_jit_type *\ gcc_jit_type_unqualified (gcc_jit_type *type) Given a type, return the unqualified type, removing "const", "volatile" and alignment qualifiers. .. function:: gcc_jit_field *\ gcc_jit_struct_get_field (gcc_jit_struct *struct_type,\ size_t index) Get a struct field by index. .. function:: size_t\ gcc_jit_struct_get_field_count (gcc_jit_struct *struct_type) Get the number of fields in the struct. The API entrypoints related to the reflection API: * :c:func:`gcc_jit_function_type_get_return_type` * :c:func:`gcc_jit_function_type_get_param_count` * :c:func:`gcc_jit_function_type_get_param_type` * :c:func:`gcc_jit_type_unqualified` * :c:func:`gcc_jit_type_dyncast_array` * :c:func:`gcc_jit_type_is_bool` * :c:func:`gcc_jit_type_dyncast_function_ptr_type` * :c:func:`gcc_jit_type_is_integral` * :c:func:`gcc_jit_type_is_pointer` * :c:func:`gcc_jit_type_dyncast_vector` * :c:func:`gcc_jit_vector_type_get_element_type` * :c:func:`gcc_jit_vector_type_get_num_units` * :c:func:`gcc_jit_struct_get_field` * :c:func:`gcc_jit_type_is_struct` * :c:func:`gcc_jit_struct_get_field_count` were added in :ref:`LIBGCCJIT_ABI_16`; you can test for their presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_REFLECTION .. type:: gcc_jit_case .. function:: int\ gcc_jit_compatible_types (gcc_jit_type *ltype,\ gcc_jit_type *rtype) Return non-zero if the two types are compatible. For instance, if :c:data:`GCC_JIT_TYPE_UINT64_T` and :c:data:`GCC_JIT_TYPE_UNSIGNED_LONG` are the same size on the target, this will return non-zero. The parameters ``ltype`` and ``rtype`` must be non-NULL. Return 0 on errors. This entrypoint was added in :ref:`LIBGCCJIT_ABI_20`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_SIZED_INTEGERS .. function:: ssize_t\ gcc_jit_type_get_size (gcc_jit_type *type) Return the size of a type, in bytes. It only works on integer types for now. The parameter ``type`` must be non-NULL. Return -1 on errors. This entrypoint was added in :ref:`LIBGCCJIT_ABI_20`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_SIZED_INTEGERS .. function:: gcc_jit_type *\ gcc_jit_type_get_restrict (gcc_jit_type *type) Given type "T", get type "restrict T". This entrypoint was added in :ref:`LIBGCCJIT_ABI_25`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_type_get_restrict --- ### Gcc Jit Docs Cp Intro Index (`gcc/jit/docs/cp/intro/index.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. Tutorial ======== .. toctree:: :maxdepth: 2 tutorial01.rst tutorial02.rst tutorial03.rst tutorial04.rst --- ### Gcc Jit Docs Cp Intro Tutorial01 (`gcc/jit/docs/cp/intro/tutorial01.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Tutorial part 1: "Hello world" ============================== Before we look at the details of the API, let's look at building and running programs that use the library. Here's a toy "hello world" program that uses the library's C++ API to synthesize a call to `printf` and uses it to write a message to stdout. Don't worry about the content of the program for now; we'll cover the details in later parts of this tutorial. .. literalinclude:: ../../examples/tut01-hello-world.cc :language: c++ Copy the above to `tut01-hello-world.cc`. Assuming you have the jit library installed, build the test program using: .. code-block:: console $ gcc \ tut01-hello-world.cc \ -o tut01-hello-world \ -lgccjit You should then be able to run the built program: .. code-block:: console $ ./tut01-hello-world hello world --- ### Gcc Jit Docs Cp Intro Tutorial02 (`gcc/jit/docs/cp/intro/tutorial02.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Tutorial part 2: Creating a trivial machine code function --------------------------------------------------------- Consider this C function: .. code-block:: c int square (int i) { return i * i; } How can we construct this at run-time using libgccjit's C++ API? First we need to include the relevant header: .. code-block:: c++ #include <libgccjit++.h> All state associated with compilation is associated with a :type:`gccjit::context`, which is a thin C++ wrapper around the C API's :c:expr:`gcc_jit_context *`. Create one using :func:`gccjit::context::acquire`: .. code-block:: c++ gccjit::context ctxt; ctxt = gccjit::context::acquire (); The JIT library has a system of types. It is statically-typed: every expression is of a specific type, fixed at compile-time. In our example, all of the expressions are of the C `int` type, so let's obtain this from the context, as a :type:`gccjit::type`, using :func:`gccjit::context::get_type`: .. code-block:: c++ gccjit::type int_type = ctxt.get_type (GCC_JIT_TYPE_INT); :type:`gccjit::type` is an example of a "contextual" object: every entity in the API is associated with a :type:`gccjit::context`. Memory management is easy: all such "contextual" objects are automatically cleaned up for you when the context is released, using :func:`gccjit::context::release`: .. code-block:: c++ ctxt.release (); so you don't need to manually track and cleanup all objects, just the contexts. All of the C++ classes in the API are thin wrappers around pointers to types in the C API. The C++ class hierarchy within the ``gccjit`` namespace looks like this:: +- object +- location +- type +- struct +- field +- function +- block +- rvalue +- lvalue +- param One thing you can do with a :type:`gccjit::object` is to ask it for a human-readable description as a :type:`std::string`, using :func:`gccjit::object::get_debug_string`: .. code-block:: c++ printf ("obj: %s\n", obj.get_debug_string ().c_str ()); giving this text on stdout: .. code-block:: bash obj: int This is invaluable when debugging. Let's create the function. To do so, we first need to construct its single parameter, specifying its type and giving it a name, using :func:`gccjit::context::new_param`: .. code-block:: c++ gccjit::param param_i = ctxt.new_param (int_type, "i"); and we can then make a vector of all of the params of the function, in this case just one: .. code-block:: c++ std::vector<gccjit::param> params; params.push_back (param_i); Now we can create the function, using :cpp:func:`gccjit::context::new_function`: .. code-block:: c++ gccjit::function func = ctxt.new_function (GCC_JIT_FUNCTION_EXPORTED, int_type, "square", params, 0); To define the code within the function, we must create basic blocks containing statements. Every basic block contains a list of statements, eventually terminated by a statement that either returns, or jumps to another basic block. Our function has no control-flow, so we just need one basic block: .. code-block:: c++ gccjit::block block = func.new_block (); Our basic block is relatively simple: it immediately terminates by returning the value of an expression. We can build the expression using :func:`gccjit::context::new_binary_op`: .. code-block:: c++ gccjit::rvalue expr = ctxt.new_binary_op ( GCC_JIT_BINARY_OP_MULT, int_type, param_i, param_i); A :type:`gccjit::rvalue` is another example of a :type:`gccjit::object` subclass. As before, we can print it with :func:`gccjit::object::get_debug_string`. .. code-block:: c++ printf ("expr: %s\n", expr.get_debug_string ().c_str ()); giving this output: .. code-block:: bash expr: i * i Note that :type:`gccjit::rvalue` provides numerous overloaded operators which can be used to dramatically reduce the amount of typing needed. We can build the above binary operation more directly with this one-liner: .. code-block:: c++ gccjit::rvalue expr = param_i * param_i; Creating the expression in itself doesn't do anything; we have to add this expression to a statement within the block. In this case, we use it to build a return statement, which terminates the basic block: .. code-block:: c++ block.end_with_return (expr); OK, we've populated the context. We can now compile it using :func:`gccjit::context::compile`: .. code-block:: c++ gcc_jit_result *result; result = ctxt.compile (); and get a :c:expr:`gcc_jit_result *`. We can now use :c:func:`gcc_jit_result_get_code` to look up a specific machine code routine within the result, in this case, the function we created above. .. code-block:: c++ void *fn_ptr = gcc_jit_result_get_code (result, "square"); if (!fn_ptr) { fprintf (stderr, "NULL fn_ptr"); goto error; } We can now cast the pointer to an appropriate function pointer type, and then call it: .. code-block:: c++ typedef int (*fn_type) (int); fn_type square = (fn_type)fn_ptr; printf ("result: %d", square (5)); .. code-block:: bash result: 25 Options ******* To get more information on what's going on, you can set debugging flags on the context using :func:`gccjit::context::set_bool_option`. .. (I'm deliberately not mentioning :c:macro:`GCC_JIT_BOOL_OPTION_DUMP_INITIAL_TREE` here since I think it's probably more of use to implementors than to users) Setting :c:macro:`GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE` will dump a C-like representation to stderr when you compile (GCC's "GIMPLE" representation): .. code-block:: c++ ctxt.set_bool_option (GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE, 1); result = ctxt.compile (); .. code-block:: c square (signed int i) { signed int D.260; entry: D.260 = i * i; return D.260; } We can see the generated machine code in assembler form (on stderr) by setting :c:macro:`GCC_JIT_BOOL_OPTION_DUMP_GENERATED_CODE` on the context before compiling: .. code-block:: c++ ctxt.set_bool_option (GCC_JIT_BOOL_OPTION_DUMP_GENERATED_CODE, 1); result = ctxt.compile (); .. code-block:: gas .file "fake.c" .text .globl square .type square, @function square: .LFB6: .cfi_startproc pushq %rbp .cfi_def_cfa_offset 16 .cfi_offset 6, -16 movq %rsp, %rbp .cfi_def_cfa_register 6 movl %edi, -4(%rbp) .L14: movl -4(%rbp), %eax imull -4(%rbp), %eax popq %rbp .cfi_def_cfa 7, 8 ret .cfi_endproc .LFE6: .size square, .-square .ident "GCC: (GNU) 4.9.0 20131023 (Red Hat 0.2)" .section .note.GNU-stack,"",@progbits By default, no optimizations are performed, the equivalent of GCC's `-O0` option. We can turn things up to e.g. `-O3` by calling :func:`gccjit::context::set_int_option` with :c:macro:`GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL`: .. code-block:: c++ ctxt.set_int_option (GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL, 3); .. code-block:: gas .file "fake.c" .text .p2align 4,,15 .globl square .type square, @function square: .LFB7: .cfi_startproc .L16: movl %edi, %eax imull %edi, %eax ret .cfi_endproc .LFE7: .size square, .-square .ident "GCC: (GNU) 4.9.0 20131023 (Red Hat 0.2)" .section .note.GNU-stack,"",@progbits Naturally this has only a small effect on such a trivial function. Full example ************ Here's what the above looks like as a complete program: .. literalinclude:: ../../examples/tut02-square.cc :lines: 1- :language: c++ Building and running it: .. code-block:: console $ gcc \ tut02-square.cc \ -o tut02-square \ -lgccjit # Run the built program: $ ./tut02-square result: 25 --- ### Gcc Jit Docs Cp Intro Tutorial03 (`gcc/jit/docs/cp/intro/tutorial03.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Tutorial part 3: Loops and variables ------------------------------------ Consider this C function: .. code-block:: c int loop_test (int n) { int sum = 0; for (int i = 0; i < n; i++) sum += i * i; return sum; } This example demonstrates some more features of libgccjit, with local variables and a loop. To break this down into libgccjit terms, it's usually easier to reword the `for` loop as a `while` loop, giving: .. code-block:: c int loop_test (int n) { int sum = 0; int i = 0; while (i < n) { sum += i * i; i++; } return sum; } Here's what the final control flow graph will look like: .. figure:: ../../intro/sum-of-squares.png :alt: image of a control flow graph As before, we include the libgccjit++ header and make a :type:`gccjit::context`. .. code-block:: c++ #include <libgccjit++.h> void test (void) { gccjit::context ctxt; ctxt = gccjit::context::acquire (); The function works with the C `int` type. In the previous tutorial we acquired this via .. code-block:: c++ gccjit::type the_type = ctxt.get_type (ctxt, GCC_JIT_TYPE_INT); though we could equally well make it work on, say, `double`: .. code-block:: c++ gccjit::type the_type = ctxt.get_type (ctxt, GCC_JIT_TYPE_DOUBLE); For integer types we can use :func:`gccjit::context::get_int_type<T>` to directly bind a specific type: .. code-block:: c++ gccjit::type the_type = ctxt.get_int_type <int> (); Let's build the function: .. code-block:: c++ gcc_jit_param n = ctxt.new_param (the_type, "n"); std::vector<gccjit::param> params; params.push_back (n); gccjit::function func = ctxt.new_function (GCC_JIT_FUNCTION_EXPORTED, return_type, "loop_test", params, 0); Expressions: lvalues and rvalues ******************************** The base class of expression is the :type:`gccjit::rvalue`, representing an expression that can be on the *right*-hand side of an assignment: a value that can be computed somehow, and assigned *to* a storage area (such as a variable). It has a specific :type:`gccjit::type`. Another important class is :type:`gccjit::lvalue`. A :type:`gccjit::lvalue`. is something that can of the *left*-hand side of an assignment: a storage area (such as a variable). In other words, every assignment can be thought of as: .. code-block:: c LVALUE = RVALUE; Note that :type:`gccjit::lvalue` is a subclass of :type:`gccjit::rvalue`, where in an assignment of the form: .. code-block:: c LVALUE_A = LVALUE_B; the `LVALUE_B` implies reading the current value of that storage area, assigning it into the `LVALUE_A`. So far the only expressions we've seen are from the previous tutorial: 1. the multiplication `i * i`: .. code-block:: c++ gccjit::rvalue expr = ctxt.new_binary_op ( GCC_JIT_BINARY_OP_MULT, int_type, param_i, param_i); /* Alternatively, using operator-overloading: */ gccjit::rvalue expr = param_i * param_i; which is a :type:`gccjit::rvalue`, and 2. the various function parameters: `param_i` and `param_n`, instances of :type:`gccjit::param`, which is a subclass of :type:`gccjit::lvalue` (and, in turn, of :type:`gccjit::rvalue`): we can both read from and write to function parameters within the body of a function. Our new example has a new kind of expression: we have two local variables. We create them by calling :func:`gccjit::function::new_local`, supplying a type and a name: .. code-block:: c++ /* Build locals: */ gccjit::lvalue i = func.new_local (the_type, "i"); gccjit::lvalue sum = func.new_local (the_type, "sum"); These are instances of :type:`gccjit::lvalue` - they can be read from and written to. Note that there is no precanned way to create *and* initialize a variable like in C: .. code-block:: c int i = 0; Instead, having added the local to the function, we have to separately add an assignment of `0` to `local_i` at the beginning of the function. Control flow ************ This function has a loop, so we need to build some basic blocks to handle the control flow. In this case, we need 4 blocks: 1. before the loop (initializing the locals) 2. the conditional at the top of the loop (comparing `i < n`) 3. the body of the loop 4. after the loop terminates (`return sum`) so we create these as :type:`gccjit::block` instances within the :type:`gccjit::function`: .. code-block:: c++ gccjit::block b_initial = func.new_block ("initial"); gccjit::block b_loop_cond = func.new_block ("loop_cond"); gccjit::block b_loop_body = func.new_block ("loop_body"); gccjit::block b_after_loop = func.new_block ("after_loop"); We now populate each block with statements. The entry block `b_initial` consists of initializations followed by a jump to the conditional. We assign `0` to `i` and to `sum`, using :func:`gccjit::block::add_assignment` to add an assignment statement, and using :func:`gccjit::context::zero` to get the constant value `0` for the relevant type for the right-hand side of the assignment: .. code-block:: c++ /* sum = 0; */ b_initial.add_assignment (sum, ctxt.zero (the_type)); /* i = 0; */ b_initial.add_assignment (i, ctxt.zero (the_type)); We can then terminate the entry block by jumping to the conditional: .. code-block:: c++ b_initial.end_with_jump (b_loop_cond); The conditional block is equivalent to the line `while (i < n)` from our C example. It contains a single statement: a conditional, which jumps to one of two destination blocks depending on a boolean :type:`gccjit::rvalue`, in this case the comparison of `i` and `n`. We could build the comparison using :func:`gccjit::context::new_comparison`: .. code-block:: c++ gccjit::rvalue guard = ctxt.new_comparison (GCC_JIT_COMPARISON_GE, i, n); and can then use this to add `b_loop_cond`'s sole statement, via :func:`gccjit::block::end_with_conditional`: .. code-block:: c++ b_loop_cond.end_with_conditional (guard, b_after_loop, // on_true b_loop_body); // on_false However :type:`gccjit::rvalue` has overloaded operators for this, so we express the conditional as .. code-block:: c++ gccjit::rvalue guard = (i >= n); and hence we can write the block more concisely as: .. code-block:: c++ b_loop_cond.end_with_conditional ( i >= n, b_after_loop, // on_true b_loop_body); // on_false Next, we populate the body of the loop. The C statement `sum += i * i;` is an assignment operation, where an lvalue is modified "in-place". We use :func:`gccjit::block::add_assignment_op` to handle these operations: .. code-block:: c++ /* sum += i * i */ b_loop_body.add_assignment_op (sum, GCC_JIT_BINARY_OP_PLUS, i * i); The `i++` can be thought of as `i += 1`, and can thus be handled in a similar way. We use :c:func:`gcc_jit_context_one` to get the constant value `1` (for the relevant type) for the right-hand side of the assignment. .. code-block:: c++ /* i++ */ b_loop_body.add_assignment_op (i, GCC_JIT_BINARY_OP_PLUS, ctxt.one (the_type)); .. note:: For numeric constants other than 0 or 1, we could use :func:`gccjit::context::new_rvalue`, which has overloads for both ``int`` and ``double``. The loop body completes by jumping back to the conditional: .. code-block:: c++ b_loop_body.end_with_jump (b_loop_cond); Finally, we populate the `b_after_loop` block, reached when the loop conditional is false. We want to generate the equivalent of: .. code-block:: c++ return sum; so the block is just one statement: .. code-block:: c++ /* return sum */ b_after_loop.end_with_return (sum); .. note:: You can intermingle block creation with statement creation, but given that the terminator statements generally include references to other blocks, I find it's clearer to create all the blocks, *then* all the statements. We've finished populating the function. As before, we can now compile it to machine code: .. code-block:: c++ gcc_jit_result *result; result = ctxt.compile (); ctxt.release (); if (!result) { fprintf (stderr, "NULL result"); return 1; } typedef int (*loop_test_fn_type) (int); loop_test_fn_type loop_test = (loop_test_fn_type)gcc_jit_result_get_code (result, "loop_test"); if (!loop_test) { fprintf (stderr, "NULL loop_test"); gcc_jit_result_release (result); return 1; } printf ("result: %d", loop_test (10)); .. code-block:: bash result: 285 Visualizing the control flow graph ********************************** You can see the control flow graph of a function using :func:`gccjit::function::dump_to_dot`: .. code-block:: c++ func.dump_to_dot ("/tmp/sum-of-squares.dot"); giving a .dot file in GraphViz format. You can convert this to an image using `dot`: .. code-block:: bash $ dot -Tpng /tmp/sum-of-squares.dot -o /tmp/sum-of-squares.png or use a viewer (my preferred one is xdot.py; see https://github.com/jrfonseca/xdot.py; on Fedora you can install it with `yum install python-xdot`): .. figure:: ../../intro/sum-of-squares.png :alt: image of a control flow graph Full example ************ .. literalinclude:: ../../examples/tut03-sum-of-squares.cc :lines: 1- :language: c++ Building and running it: .. code-block:: console $ gcc \ tut03-sum-of-squares.cc \ -o tut03-sum-of-squares \ -lgccjit # Run the built program: $ ./tut03-sum-of-squares loop_test returned: 285 --- ### Gcc Jit Docs Cp Intro Tutorial04 (`gcc/jit/docs/cp/intro/tutorial04.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Tutorial part 4: Adding JIT-compilation to a toy interpreter ------------------------------------------------------------ In this example we construct a "toy" interpreter, and add JIT-compilation to it. Our toy interpreter ******************* It's a stack-based interpreter, and is intended as a (very simple) example of the kind of bytecode interpreter seen in dynamic languages such as Python, Ruby etc. For the sake of simplicity, our toy virtual machine is very limited: * The only data type is `int` * It can only work on one function at a time (so that the only function call that can be made is to recurse). * Functions can only take one parameter. * Functions have a stack of `int` values. * We'll implement function call within the interpreter by calling a function in our implementation, rather than implementing our own frame stack. * The parser is only good enough to get the examples to work. Naturally, a real interpreter would be much more complicated that this. The following operations are supported: .. list-table:: :header-rows: 1 :widths: 35 50 25 25 * - Operation - Meaning - Old Stack - New Stack * - DUP - Duplicate top of stack. - ``[..., x]`` - ``[..., x, x]`` * - ROT - Swap top two elements of stack. - ``[..., x, y]`` - ``[..., y, x]`` * - BINARY_ADD - Add the top two elements on the stack. - ``[..., x, y]`` - ``[..., (x+y)]`` * - BINARY_SUBTRACT - Likewise, but subtract. - ``[..., x, y]`` - ``[..., (x-y)]`` * - BINARY_MULT - Likewise, but multiply. - ``[..., x, y]`` - ``[..., (x*y)]`` * - BINARY_COMPARE_LT - Compare the top two elements on the stack and push a nonzero/zero if (x<y). - ``[..., x, y]`` - ``[..., (x<y)]`` * - RECURSE - Recurse, passing the top of the stack, and popping the result. - ``[..., x]`` - ``[..., fn(x)]`` * - RETURN - Return the top of the stack. - ``[x]`` - ``[]`` * - PUSH_CONST `arg` - Push an int const. - ``[...]`` - ``[..., arg]`` * - JUMP_ABS_IF_TRUE `arg` - Pop; if top of stack was nonzero, jump to ``arg``. - ``[..., x]`` - ``[...]`` Programs can be interpreted, disassembled, and compiled to machine code. The interpreter reads ``.toy`` scripts. Here's what a simple recursive factorial program looks like, the script ``factorial.toy``. The parser ignores lines beginning with a `#`. .. literalinclude:: ../../examples/tut04-toyvm/factorial.toy :lines: 1- :language: c The interpreter is a simple infinite loop with a big ``switch`` statement based on what the next opcode is: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Execute the given function. */ :end-before: /* JIT compilation. */ :language: c++ Compiling to machine code ************************* We want to generate machine code that can be cast to this type and then directly executed in-process: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Functions are compiled to this function ptr type. */ :end-before: enum opcode :language: c++ Our compiler isn't very sophisticated; it takes the implementation of each opcode above, and maps it directly to the operations supported by the libgccjit API. How should we handle the stack? In theory we could calculate what the stack depth will be at each opcode, and optimize away the stack manipulation "by hand". We'll see below that libgccjit is able to do this for us, so we'll implement stack manipulation in a direct way, by creating a ``stack`` array and ``stack_depth`` variables, local within the generated function, equivalent to this C code: .. code-block:: c int stack_depth; int stack[MAX_STACK_DEPTH]; We'll also have local variables ``x`` and ``y`` for use when implementing the opcodes, equivalent to this: .. code-block:: c int x; int y; This means our compiler has the following state: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* State. */ :end-before: }; :language: c++ Setting things up ***************** First we create our types: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Create types. */ :end-before: /* The constant value 1. */ :language: c++ along with extracting a useful `int` constant: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* The constant value 1. */ :end-before: /* Create locations. */ :language: c++ We'll implement push and pop in terms of the ``stack`` array and ``stack_depth``. Here are helper functions for adding statements to a block, implementing pushing and popping values: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Stack manipulation. */ :end-before: /* Create the context. */ :language: c++ We will support single-stepping through the generated code in the debugger, so we need to create :type:`gccjit::location` instances, one per operation in the source code. These will reference the lines of e.g. ``factorial.toy``. .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Create locations. */ :end-before: /* Creating the function. */ :language: c++ Let's create the function itself. As usual, we create its parameter first, then use the parameter to create the function: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Creating the function. */ :end-before: /* Create stack lvalues. */ :language: c++ We create the locals within the function. .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Create stack lvalues. */ :end-before: /* 1st pass: create blocks, one per opcode. :language: c++ Populating the function *********************** There's some one-time initialization, and the API treats the first block you create as the entrypoint of the function, so we need to create that block first: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: first. */ :end-before: /* Create a block per operation. */ :language: c++ We can now create blocks for each of the operations. Most of these will be consolidated into larger blocks when the optimizer runs. .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Create a block per operation. */ :end-before: /* Populate the initial block. */ :language: c++ Now that we have a block it can jump to when it's done, we can populate the initial block: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Populate the initial block. */ :end-before: /* 2nd pass: fill in instructions. */ :language: c++ We can now populate the blocks for the individual operations. We loop through them, adding instructions to their blocks: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* 2nd pass: fill in instructions. */ :end-before: /* Helper macros. */ :language: c++ We're going to have another big ``switch`` statement for implementing the opcodes, this time for compiling them, rather than interpreting them. It's helpful to have macros for implementing push and pop, so that we can make the ``switch`` statement that's coming up look as much as possible like the one above within the interpreter: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Helper macros. */ :end-before: block.add_comment :language: c++ .. note:: A particularly clever implementation would have an *identical* ``switch`` statement shared by the interpreter and the compiler, with some preprocessor "magic". We're not doing that here, for the sake of simplicity. When I first implemented this compiler, I accidentally missed an edit when copying and pasting the ``Y_EQUALS_POP`` macro, so that popping the stack into ``y`` instead erroneously assigned it to ``x``, leaving ``y`` uninitialized. To track this kind of thing down, we can use :func:`gccjit::block::add_comment` to add descriptive comments to the internal representation. This is invaluable when looking through the generated IR for, say ``factorial``: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: PUSH_RVALUE (y) :end-before: /* Handle the individual opcodes. */ :language: c++ We can now write the big ``switch`` statement that implements the individual opcodes, populating the relevant block with statements: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Handle the individual opcodes. */ :end-before: /* Go to the next block. */ :language: c++ Every block must be terminated, via a call to one of the ``gccjit::block::end_with_`` entrypoints. This has been done for two of the opcodes, but we need to do it for the other ones, by jumping to the next block. .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Go to the next block. */ :end-before: /* end of loop on PC locations. */ :language: c++ This is analogous to simply incrementing the program counter. Verifying the control flow graph ******************************** Having finished looping over the blocks, the context is complete. As before, we can verify that the control flow and statements are sane by using :func:`gccjit::function::dump_to_dot`: .. code-block:: c++ fn.dump_to_dot ("/tmp/factorial.dot"); and viewing the result. Note how the label names, comments, and variable names show up in the dump, to make it easier to spot errors in our compiler. .. figure:: ../../intro/factorial.png :alt: image of a control flow graph Compiling the context ********************* Having finished looping over the blocks and populating them with statements, the context is complete. We can now compile it, extract machine code from the result, and run it: .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* Wrapper around a gcc_jit_result *. */ :end-before: /* Functions are compiled to this function ptr type. */ :language: c++ .. literalinclude:: ../../examples/tut04-toyvm/toyvm.cc :start-after: /* JIT-compilation. */ :end-before: return 0; :language: c++ Single-stepping through the generated code ****************************************** It's possible to debug the generated code. To do this we need to both: * Set up source code locations for our statements, so that we can meaningfully step through the code. We did this above by calling :func:`gccjit::context::new_location` and using the results. * Enable the generation of debugging information, by setting :c:macro:`GCC_JIT_BOOL_OPTION_DEBUGINFO` on the :type:`gccjit::context` via :func:`gccjit::context::set_bool_option`: .. code-block:: c++ ctxt.set_bool_option (GCC_JIT_BOOL_OPTION_DEBUGINFO, 1); Having done this, we can put a breakpoint on the generated function: .. code-block:: console $ gdb --args ./toyvm factorial.toy 10 (gdb) break factorial Function "factorial" not defined. Make breakpoint pending on future shared library load? (y or [n]) y Breakpoint 1 (factorial) pending. (gdb) run Breakpoint 1, factorial (arg=10) at factorial.toy:14 14 DUP We've set up location information, which references ``factorial.toy``. This allows us to use e.g. ``list`` to see where we are in the script: .. code-block:: console (gdb) list 9 10 # Initial state: 11 # stack: [arg] 12 13 # 0: 14 DUP 15 # stack: [arg, arg] 16 17 # 1: 18 PUSH_CONST 2 and to step through the function, examining the data: .. code-block:: console (gdb) n 18 PUSH_CONST 2 (gdb) n 22 BINARY_COMPARE_LT (gdb) print stack $5 = {10, 10, 2, 0, -7152, 32767, 0, 0} (gdb) print stack_depth $6 = 3 You'll see that the parts of the ``stack`` array that haven't been touched yet are uninitialized. .. note:: Turning on optimizations may lead to unpredictable results when stepping through the generated code: the execution may appear to "jump around" the source code. This is analogous to turning up the optimization level in a regular compiler. Examining the generated code **************************** How good is the optimized code? We can turn up optimizations, by calling :cpp:func:`gccjit::context::set_int_option` with :c:macro:`GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL`: .. code-block:: c++ ctxt.set_int_option (GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL, 3); One of GCC's internal representations is called "gimple". A dump of the initial gimple representation of the code can be seen by setting: .. code-block:: c++ ctxt.set_bool_option (GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE, 1); With optimization on and source locations displayed, this gives: .. We'll use "c" for gimple dumps .. code-block:: c factorial (signed int arg) { <unnamed type> D.80; signed int D.81; signed int D.82; signed int D.83; signed int D.84; signed int D.85; signed int y; signed int x; signed int stack_depth; signed int stack[8]; try { initial: stack_depth = 0; stack[stack_depth] = arg; stack_depth = stack_depth + 1; goto instr0; instr0: /* DUP */: stack_depth = stack_depth + -1; x = stack[stack_depth]; stack[stack_depth] = x; stack_depth = stack_depth + 1; stack[stack_depth] = x; stack_depth = stack_depth + 1; goto instr1; instr1: /* PUSH_CONST */: stack[stack_depth] = 2; stack_depth = stack_depth + 1; goto instr2; /* etc */ You can see the generated machine code in assembly form via: .. code-block:: c++ ctxt.set_bool_option (GCC_JIT_BOOL_OPTION_DUMP_GENERATED_CODE, 1); result = ctxt.compile (); which shows that (on this x86_64 box) the compiler has unrolled the loop and is using MMX instructions to perform several multiplications simultaneously: .. code-block:: gas .file "fake.c" .text .Ltext0: .p2align 4,,15 .globl factorial .type factorial, @function factorial: .LFB0: .file 1 "factorial.toy" .loc 1 14 0 .cfi_startproc .LVL0: .L2: .loc 1 26 0 cmpl $1, %edi jle .L13 leal -1(%rdi), %edx movl %edx, %ecx shrl $2, %ecx leal 0(,%rcx,4), %esi testl %esi, %esi je .L14 cmpl $9, %edx jbe .L14 leal -2(%rdi), %eax movl %eax, -16(%rsp) leal -3(%rdi), %eax movd -16(%rsp), %xmm0 movl %edi, -16(%rsp) movl %eax, -12(%rsp) movd -16(%rsp), %xmm1 xorl %eax, %eax movl %edx, -16(%rsp) movd -12(%rsp), %xmm4 movd -16(%rsp), %xmm6 punpckldq %xmm4, %xmm0 movdqa .LC1(%rip), %xmm4 punpckldq %xmm6, %xmm1 punpcklqdq %xmm0, %xmm1 movdqa .LC0(%rip), %xmm0 jmp .L5 # etc - edited for brevity This is clearly overkill for a function that will likely overflow the ``int`` type before the vectorization is worthwhile - but then again, this is a toy example. Turning down the optimization level to 2: .. code-block:: c++ ctxt.set_int_option (GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL, 2); yields this code, which is simple enough to quote in its entirety: .. code-block:: gas .file "fake.c" .text .p2align 4,,15 .globl factorial .type factorial, @function factorial: .LFB0: .cfi_startproc .L2: cmpl $1, %edi jle .L8 movl $1, %edx jmp .L4 .p2align 4,,10 .p2align 3 .L6: movl %eax, %edi .L4: .L5: leal -1(%rdi), %eax imull %edi, %edx cmpl $1, %eax jne .L6 .L3: .L7: imull %edx, %eax ret .L8: movl %edi, %eax movl $1, %edx jmp .L7 .cfi_endproc .LFE0: .size factorial, .-factorial .ident "GCC: (GNU) 4.9.0 20131023 (Red Hat 0.2)" .section .note.GNU-stack,"",@progbits Note that the stack pushing and popping have been eliminated, as has the recursive call (in favor of an iteration). Putting it all together *********************** The complete example can be seen in the source tree at ``gcc/jit/docs/examples/tut04-toyvm/toyvm.cc`` along with a Makefile and a couple of sample .toy scripts: .. code-block:: console $ ls -al drwxrwxr-x. 2 david david 4096 Sep 19 17:46 . drwxrwxr-x. 3 david david 4096 Sep 19 15:26 .. -rw-rw-r--. 1 david david 615 Sep 19 12:43 factorial.toy -rw-rw-r--. 1 david david 834 Sep 19 13:08 fibonacci.toy -rw-rw-r--. 1 david david 238 Sep 19 14:22 Makefile -rw-rw-r--. 1 david david 16457 Sep 19 17:07 toyvm.cc $ make toyvm g++ -Wall -g -o toyvm toyvm.cc -lgccjit $ ./toyvm factorial.toy 10 interpreter result: 3628800 compiler result: 3628800 $ ./toyvm fibonacci.toy 10 interpreter result: 55 compiler result: 55 Behind the curtain: How does our code get optimized? **************************************************** Our example is done, but you may be wondering about exactly how the compiler turned what we gave it into the machine code seen above. We can examine what the compiler is doing in detail by setting: .. code-block:: c++ state.ctxt.set_bool_option (GCC_JIT_BOOL_OPTION_DUMP_EVERYTHING, 1); state.ctxt.set_bool_option (GCC_JIT_BOOL_OPTION_KEEP_INTERMEDIATES, 1); This will dump detailed information about the compiler's state to a directory under ``/tmp``, and keep it from being cleaned up. The precise names and their formats of these files is subject to change. Higher optimization levels lead to more files. Here's what I saw (edited for brevity; there were almost 200 files): .. code-block:: console intermediate files written to /tmp/libgccjit-KPQbGw $ ls /tmp/libgccjit-KPQbGw/ fake.c.000i.cgraph fake.c.000i.type-inheritance fake.c.004t.gimple fake.c.007t.omplower fake.c.008t.lower fake.c.011t.eh fake.c.012t.cfg fake.c.014i.visibility fake.c.015i.early_local_cleanups fake.c.016t.ssa # etc The gimple code is converted into Static Single Assignment form, with annotations for use when generating the debuginfo: .. code-block:: console $ less /tmp/libgccjit-KPQbGw/fake.c.016t.ssa .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) factorial (signed int arg) { signed int stack[8]; signed int stack_depth; signed int x; signed int y; <unnamed type> _20; signed int _21; signed int _38; signed int _44; signed int _51; signed int _56; initial: stack_depth_3 = 0; # DEBUG stack_depth => stack_depth_3 stack[stack_depth_3] = arg_5(D); stack_depth_7 = stack_depth_3 + 1; # DEBUG stack_depth => stack_depth_7 # DEBUG instr0 => NULL # DEBUG /* DUP */ => NULL stack_depth_8 = stack_depth_7 + -1; # DEBUG stack_depth => stack_depth_8 x_9 = stack[stack_depth_8]; # DEBUG x => x_9 stack[stack_depth_8] = x_9; stack_depth_11 = stack_depth_8 + 1; # DEBUG stack_depth => stack_depth_11 stack[stack_depth_11] = x_9; stack_depth_13 = stack_depth_11 + 1; # DEBUG stack_depth => stack_depth_13 # DEBUG instr1 => NULL # DEBUG /* PUSH_CONST */ => NULL stack[stack_depth_13] = 2; /* etc; edited for brevity */ We can perhaps better see the code by turning off :c:macro:`GCC_JIT_BOOL_OPTION_DEBUGINFO` to suppress all those ``DEBUG`` statements, giving: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.016t.ssa .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) factorial (signed int arg) { signed int stack[8]; signed int stack_depth; signed int x; signed int y; <unnamed type> _20; signed int _21; signed int _38; signed int _44; signed int _51; signed int _56; initial: stack_depth_3 = 0; stack[stack_depth_3] = arg_5(D); stack_depth_7 = stack_depth_3 + 1; stack_depth_8 = stack_depth_7 + -1; x_9 = stack[stack_depth_8]; stack[stack_depth_8] = x_9; stack_depth_11 = stack_depth_8 + 1; stack[stack_depth_11] = x_9; stack_depth_13 = stack_depth_11 + 1; stack[stack_depth_13] = 2; stack_depth_15 = stack_depth_13 + 1; stack_depth_16 = stack_depth_15 + -1; y_17 = stack[stack_depth_16]; stack_depth_18 = stack_depth_16 + -1; x_19 = stack[stack_depth_18]; _20 = x_19 < y_17; _21 = (signed int) _20; stack[stack_depth_18] = _21; stack_depth_23 = stack_depth_18 + 1; stack_depth_24 = stack_depth_23 + -1; x_25 = stack[stack_depth_24]; if (x_25 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: stack_depth_26 = stack_depth_24 + -1; x_27 = stack[stack_depth_26]; stack[stack_depth_26] = x_27; stack_depth_29 = stack_depth_26 + 1; stack[stack_depth_29] = x_27; stack_depth_31 = stack_depth_29 + 1; stack[stack_depth_31] = 1; stack_depth_33 = stack_depth_31 + 1; stack_depth_34 = stack_depth_33 + -1; y_35 = stack[stack_depth_34]; stack_depth_36 = stack_depth_34 + -1; x_37 = stack[stack_depth_36]; _38 = x_37 - y_35; stack[stack_depth_36] = _38; stack_depth_40 = stack_depth_36 + 1; stack_depth_41 = stack_depth_40 + -1; x_42 = stack[stack_depth_41]; _44 = factorial (x_42); stack[stack_depth_41] = _44; stack_depth_46 = stack_depth_41 + 1; stack_depth_47 = stack_depth_46 + -1; y_48 = stack[stack_depth_47]; stack_depth_49 = stack_depth_47 + -1; x_50 = stack[stack_depth_49]; _51 = x_50 * y_48; stack[stack_depth_49] = _51; stack_depth_53 = stack_depth_49 + 1; # stack_depth_1 = PHI <stack_depth_24(2), stack_depth_53(3)> instr9: /* RETURN */: stack_depth_54 = stack_depth_1 + -1; x_55 = stack[stack_depth_54]; _56 = x_55; stack ={v} {CLOBBER}; return _56; } Note in the above how all the :type:`gccjit::block` instances we created have been consolidated into just 3 blocks in GCC's internal representation: ``initial``, ``instr4`` and ``instr9``. Optimizing away stack manipulation ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Recall our simple implementation of stack operations. Let's examine how the stack operations are optimized away. After a pass of constant-propagation, the depth of the stack at each opcode can be determined at compile-time: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.021t.ccp1 .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) factorial (signed int arg) { signed int stack[8]; signed int stack_depth; signed int x; signed int y; <unnamed type> _20; signed int _21; signed int _38; signed int _44; signed int _51; initial: stack[0] = arg_5(D); x_9 = stack[0]; stack[0] = x_9; stack[1] = x_9; stack[2] = 2; y_17 = stack[2]; x_19 = stack[1]; _20 = x_19 < y_17; _21 = (signed int) _20; stack[1] = _21; x_25 = stack[1]; if (x_25 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: x_27 = stack[0]; stack[0] = x_27; stack[1] = x_27; stack[2] = 1; y_35 = stack[2]; x_37 = stack[1]; _38 = x_37 - y_35; stack[1] = _38; x_42 = stack[1]; _44 = factorial (x_42); stack[1] = _44; y_48 = stack[1]; x_50 = stack[0]; _51 = x_50 * y_48; stack[0] = _51; instr9: /* RETURN */: x_55 = stack[0]; x_56 = x_55; stack ={v} {CLOBBER}; return x_56; } Note how, in the above, all those ``stack_depth`` values are now just constants: we're accessing specific stack locations at each opcode. The "esra" pass ("Early Scalar Replacement of Aggregates") breaks out our "stack" array into individual elements: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.024t.esra .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) Created a replacement for stack offset: 0, size: 32: stack$0 Created a replacement for stack offset: 32, size: 32: stack$1 Created a replacement for stack offset: 64, size: 32: stack$2 Symbols to be put in SSA form { D.89 D.90 D.91 } Incremental SSA update started at block: 0 Number of blocks in CFG: 5 Number of blocks to update: 4 ( 80%) factorial (signed int arg) { signed int stack$2; signed int stack$1; signed int stack$0; signed int stack[8]; signed int stack_depth; signed int x; signed int y; <unnamed type> _20; signed int _21; signed int _38; signed int _44; signed int _51; initial: stack$0_45 = arg_5(D); x_9 = stack$0_45; stack$0_39 = x_9; stack$1_32 = x_9; stack$2_30 = 2; y_17 = stack$2_30; x_19 = stack$1_32; _20 = x_19 < y_17; _21 = (signed int) _20; stack$1_28 = _21; x_25 = stack$1_28; if (x_25 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: x_27 = stack$0_39; stack$0_22 = x_27; stack$1_14 = x_27; stack$2_12 = 1; y_35 = stack$2_12; x_37 = stack$1_14; _38 = x_37 - y_35; stack$1_10 = _38; x_42 = stack$1_10; _44 = factorial (x_42); stack$1_6 = _44; y_48 = stack$1_6; x_50 = stack$0_22; _51 = x_50 * y_48; stack$0_1 = _51; # stack$0_52 = PHI <stack$0_39(2), stack$0_1(3)> instr9: /* RETURN */: x_55 = stack$0_52; x_56 = x_55; stack ={v} {CLOBBER}; return x_56; } Hence at this point, all those pushes and pops of the stack are now simply assignments to specific temporary variables. After some copy propagation, the stack manipulation has been completely optimized away: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.026t.copyprop1 .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) factorial (signed int arg) { signed int stack$2; signed int stack$1; signed int stack$0; signed int stack[8]; signed int stack_depth; signed int x; signed int y; <unnamed type> _20; signed int _21; signed int _38; signed int _44; signed int _51; initial: stack$0_39 = arg_5(D); _20 = arg_5(D) <= 1; _21 = (signed int) _20; if (_21 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: _38 = arg_5(D) + -1; _44 = factorial (_38); _51 = arg_5(D) * _44; stack$0_1 = _51; # stack$0_52 = PHI <arg_5(D)(2), _51(3)> instr9: /* RETURN */: stack ={v} {CLOBBER}; return stack$0_52; } Later on, another pass finally eliminated ``stack_depth`` local and the unused parts of the `stack`` array altogether: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.036t.release_ssa .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) Released 44 names, 314.29%, removed 44 holes factorial (signed int arg) { signed int stack$0; signed int mult_acc_1; <unnamed type> _5; signed int _6; signed int _7; signed int mul_tmp_10; signed int mult_acc_11; signed int mult_acc_13; # arg_9 = PHI <arg_8(D)(0)> # mult_acc_13 = PHI <1(0)> initial: <bb 5>: # arg_4 = PHI <arg_9(2), _7(3)> # mult_acc_1 = PHI <mult_acc_13(2), mult_acc_11(3)> _5 = arg_4 <= 1; _6 = (signed int) _5; if (_6 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: _7 = arg_4 + -1; mult_acc_11 = mult_acc_1 * arg_4; goto <bb 5>; # stack$0_12 = PHI <arg_4(5)> instr9: /* RETURN */: mul_tmp_10 = mult_acc_1 * stack$0_12; return mul_tmp_10; } Elimination of tail recursion ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Another significant optimization is the detection that the call to ``factorial`` is tail recursion, which can be eliminated in favor of an iteration: .. code-block:: console $ less /tmp/libgccjit-1Hywc0/fake.c.030t.tailr1 .. code-block:: c ;; Function factorial (factorial, funcdef_no=0, decl_uid=53, symbol_order=0) Symbols to be put in SSA form { D.88 } Incremental SSA update started at block: 0 Number of blocks in CFG: 5 Number of blocks to update: 4 ( 80%) factorial (signed int arg) { signed int stack$2; signed int stack$1; signed int stack$0; signed int stack[8]; signed int stack_depth; signed int x; signed int y; signed int mult_acc_1; <unnamed type> _20; signed int _21; signed int _38; signed int mul_tmp_44; signed int mult_acc_51; # arg_5 = PHI <arg_39(D)(0), _38(3)> # mult_acc_1 = PHI <1(0), mult_acc_51(3)> initial: _20 = arg_5 <= 1; _21 = (signed int) _20; if (_21 != 0) goto <bb 4> (instr9); else goto <bb 3> (instr4); instr4: /* DUP */: _38 = arg_5 + -1; mult_acc_51 = mult_acc_1 * arg_5; goto <bb 2> (initial); # stack$0_52 = PHI <arg_5(2)> instr9: /* RETURN */: stack ={v} {CLOBBER}; mul_tmp_44 = mult_acc_1 * stack$0_52; return mul_tmp_44; } --- ### Gcc Jit Docs Cp Topics Asm (`gcc/jit/docs/cp/topics/asm.rst`) .. Copyright (C) 2020-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Using Assembly Language with libgccjit++ ======================================== libgccjit has some support for directly embedding assembler instructions. This is based on GCC's support for inline ``asm`` in C code, and the following assumes a familiarity with that functionality. See `How to Use Inline Assembly Language in C Code <https://gcc.gnu.org/onlinedocs/gcc/Using-Assembly-Language-with-C.html>`_ in GCC's documentation, the "Extended Asm" section in particular. These entrypoints were added in :ref:`LIBGCCJIT_ABI_15`; you can test for their presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_ASM_STATEMENTS Adding assembler instructions within a function *********************************************** .. class:: gccjit::extended_asm A `gccjit::extended_asm` represents an extended ``asm`` statement: a series of low-level instructions inside a function that convert inputs to outputs. :class:`gccjit::extended_asm` is a subclass of :class:`gccjit::object`. It is a thin wrapper around the C API's :c:expr:`gcc_jit_extended_asm *`. To avoid having an API entrypoint with a very large number of parameters, an extended ``asm`` statement is made in stages: an initial call to create the :type:`gccjit::extended_asm`, followed by calls to add operands and set other properties of the statement. There are two API entrypoints for creating a :type:`gccjit::extended_asm`: * :func:`gccjit::block::add_extended_asm` for an ``asm`` statement with no control flow, and * :func:`gccjit::block::end_with_extended_asm_goto` for an ``asm goto``. For example, to create the equivalent of: .. literalinclude:: ../../../../testsuite/jit.dg/test-asm.cc :start-after: // Quote from here in docs/cp/topics/asm.rst: example 1: C :end-before: // Quote up to here in docs/cp/topics/asm.rst: example 1: C :language: c the following API calls could be used: .. literalinclude:: ../../../../testsuite/jit.dg/test-asm.cc :start-after: /* Quote from here in docs/cp/topics/asm.rst: example 1: jit. */ :end-before: /* Quote up to here in docs/cp/topics/asm.rst: example 1: jit. */ :language: c .. warning:: When considering the numbering of operands within an extended ``asm`` statement (e.g. the ``%0`` and ``%1`` above), the equivalent to the C syntax is followed i.e. all output operands, then all input operands, regardless of what order the calls to :func:`gccjit::extended_asm::add_output_operand` and :func:`gccjit::extended_asm::add_input_operand` were made in. As in the C syntax, operands can be given symbolic names to avoid having to number them. For example, to create the equivalent of: .. literalinclude:: ../../../../testsuite/jit.dg/test-asm.cc :start-after: // Quote from here in docs/cp/topics/asm.rst: example 2: C :end-before: // Quote up to here in docs/cp/topics/asm.rst: example 2: C :language: c the following API calls could be used: .. literalinclude:: ../../../../testsuite/jit.dg/test-asm.cc :start-after: /* Quote from here in docs/cp/topics/asm.rst: example 2: jit. */ :end-before: /* Quote up to here in docs/cp/topics/asm.rst: example 2: jit. */ :language: c .. function:: extended_asm \ gccjit::block::add_extended_asm (const std::string &asm_template,\ gccjit::location loc = location ()) Create a :type:`gccjit::extended_asm` for an extended ``asm`` statement with no control flow (i.e. without the ``goto`` qualifier). The parameter ``asm_template`` corresponds to the `AssemblerTemplate` within C's extended ``asm`` syntax. It must be non-NULL. The call takes a copy of the underlying string, so it is valid to pass in a pointer to an on-stack buffer. .. function:: extended_asm\ gccjit::block::end_with_extended_asm_goto (const std::string &asm_template,\ std::vector<block> goto_blocks,\ block *fallthrough_block,\ location loc = location ()) Create a :type:`gccjit::extended_asm` for an extended ``asm`` statement that may perform jumps, and use it to terminate the given block. This is equivalent to the ``goto`` qualifier in C's extended ``asm`` syntax. For example, to create the equivalent of: .. literalinclude:: ../../../../testsuite/jit.dg/test-asm.cc :start-after: // Quote from here in docs/cp/topics/asm.rst: example 3b: C :end-before: // Quote up to here in docs/cp/topics/asm.rst: example 3b: C :language: c the following API calls could be used: .. literalinclude:: ../../../../testsuite/jit.dg/test-asm.cc :start-after: /* Quote from here in docs/cp/topics/asm.rst: example 3: jit. */ :end-before: /* Quote up to here in docs/cp/topics/asm.rst: example 3: jit. */ :language: c here referencing a :type:`gcc_jit_block` named "carry". ``num_goto_blocks`` corresponds to the ``GotoLabels`` parameter within C's extended ``asm`` syntax. The block names can be referenced within the assembler template. ``fallthrough_block`` can be NULL. If non-NULL, it specifies the block to fall through to after the statement. .. note:: This is needed since each :type:`gccjit::block` must have a single exit point, as a basic block: you can't jump from the middle of a block. A "goto" is implicitly added after the asm to handle the fallthrough case, which is equivalent to what would have happened in the C case. .. function:: gccjit::extended_asm &\ gccjit::extended_asm::set_volatile_flag (bool flag) Set whether the :type:`gccjit::extended_asm` has side-effects, equivalent to the `volatile <https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html#Volatile>`_ qualifier in C's extended asm syntax. For example, to create the equivalent of: .. code-block:: c asm volatile ("rdtsc\n\t" // Returns the time in EDX:EAX. "shl $32, %%rdx\n\t" // Shift the upper bits left. "or %%rdx, %0" // 'Or' in the lower bits. : "=a" (msr) : : "rdx"); the following API calls could be used: .. literalinclude:: ../../../../testsuite/jit.dg/test-asm.cc :start-after: /* Quote from here in docs/cp/topics/asm.rst: example 4: jit. */ :end-before: /* Quote up to here in docs/cp/topics/asm.rst: example 4: jit. */ :language: c where the :type:`gccjit::extended_asm` is flagged as volatile. .. function:: gccjit::extended_asm &\ gccjit::extended_asm::set_inline_flag (bool flag) Set the equivalent of the `inline <https://gcc.gnu.org/onlinedocs/gcc/Size-of-an-asm.html#Size-of-an-asm>`_ qualifier in C's extended ``asm`` syntax. .. function:: gccjit::extended_asm&\ gccjit::extended_asm::add_output_operand (const std::string &asm_symbolic_name,\ const std::string &constraint,\ gccjit::lvalue dest) Add an output operand to the extended ``asm`` statement. See the `Output Operands <https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html#OutputOperands>`_ section of the documentation of the C syntax. ``asm_symbolic_name`` corresponds to the ``asmSymbolicName`` component of C's extended ``asm`` syntax, and specifies the symbolic name for the operand. See the overload below for an alternative that does not supply a symbolic name. ``constraint`` corresponds to the ``constraint`` component of C's extended ``asm`` syntax. ``dest`` corresponds to the ``cvariablename`` component of C's extended ``asm`` syntax. .. code-block:: c++ // Example with a symbolic name ("aIndex"), the equivalent of: // : [aIndex] "=r" (index) ext_asm.add_output_operand ("aIndex", "=r", index); This function can't be called on an ``asm goto`` as such instructions can't have outputs; see the `Goto Labels <https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html#GotoLabels>`_ section of GCC's "Extended Asm" documentation. .. function:: gccjit::extended_asm&\ gccjit::extended_asm::add_output_operand (const std::string &constraint,\ gccjit::lvalue dest) As above, but don't supply a symbolic name for the operand. .. code-block:: c++ // Example without a symbolic name, the equivalent of: // : "=r" (dst) ext_asm.add_output_operand ("=r", dst); .. function:: gccjit::extended_asm&\ gccjit::extended_asm::add_input_operand (const std::string &asm_symbolic_name, \ const std::string &constraint, \ gccjit::rvalue src) Add an input operand to the extended ``asm`` statement. See the `Input Operands <https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html#InputOperands>`_ section of the documentation of the C syntax. ``asm_symbolic_name`` corresponds to the ``asmSymbolicName`` component of C's extended ``asm`` syntax. See the overload below for an alternative that does not supply a symbolic name. ``constraint`` corresponds to the ``constraint`` component of C's extended ``asm`` syntax. ``src`` corresponds to the ``cexpression`` component of C's extended ``asm`` syntax. .. code-block:: c++ // Example with a symbolic name ("aMask"), the equivalent of: // : [aMask] "r" (Mask) ext_asm.add_input_operand ("aMask", "r", mask); .. function:: gccjit::extended_asm&\ gccjit::extended_asm::add_input_operand (const std::string &constraint,\ gccjit::rvalue src) As above, but don't supply a symbolic name for the operand. .. code-block:: c++ // Example without a symbolic name, the equivalent of: // : "r" (src) ext_asm.add_input_operand ("r", src); .. function:: gccjit::extended_asm&\ gccjit::extended_asm::add_clobber (const std::string &victim) Add `victim` to the list of registers clobbered by the extended ``asm`` statement. See the `Clobbers and Scratch Registers <https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html#Clobbers-and-Scratch-Registers#>`_ section of the documentation of the C syntax. Statements with multiple clobbers will require multiple calls, one per clobber. For example: .. code-block:: c++ ext_asm.add_clobber ("r0").add_clobber ("cc").add_clobber ("memory"); Adding top-level assembler statements ************************************* In addition to creating extended ``asm`` instructions within a function, there is support for creating "top-level" assembler statements, outside of any function. .. function:: void\ gccjit::context::add_top_level_asm (const char *asm_stmts,\ gccjit::location loc = location ()) Create a set of top-level asm statements, analogous to those created by GCC's "basic" ``asm`` syntax in C at file scope. For example, to create the equivalent of: .. literalinclude:: ../../../../testsuite/jit.dg/test-asm.cc :start-after: // Quote from here in docs/cp/topics/asm.rst: example 5: C :end-before: // Quote up to here in docs/cp/topics/asm.rst: example 5: C :language: c the following API calls could be used: .. literalinclude:: ../../../../testsuite/jit.dg/test-asm.cc :start-after: /* Quote from here in docs/cp/topics/asm.rst: example 5: jit. */ :end-before: /* Quote up to here in docs/cp/topics/asm.rst: example 5: jit. */ :language: c --- ### Gcc Jit Docs Cp Topics Compilation (`gcc/jit/docs/cp/topics/compilation.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Compiling a context =================== Once populated, a :class:`gccjit::context` can be compiled to machine code, either in-memory via :func:`gccjit::context::compile` or to disk via :func:`gccjit::context::compile_to_file`. You can compile a context multiple times (using either form of compilation), although any errors that occur on the context will prevent any future compilation of that context. In-memory compilation ********************* .. function:: gcc_jit_result *\ gccjit::context::compile () This calls into GCC and builds the code, returning a `gcc_jit_result *`. This is a thin wrapper around the :c:func:`gcc_jit_context_compile` API entrypoint. Ahead-of-time compilation ************************* Although libgccjit is primarily aimed at just-in-time compilation, it can also be used for implementing more traditional ahead-of-time compilers, via the :func:`gccjit::context::compile_to_file` method. .. function:: void \ gccjit::context::compile_to_file (enum gcc_jit_output_kind,\ const char *output_path) Compile the :class:`gccjit::context` to a file of the given kind. This is a thin wrapper around the :c:func:`gcc_jit_context_compile_to_file` API entrypoint. --- ### Gcc Jit Docs Cp Topics Contexts (`gcc/jit/docs/cp/topics/contexts.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Compilation contexts ==================== .. class:: gccjit::context The top-level of the C++ API is the :class:`gccjit::context` type. A :class:`gccjit::context` instance encapsulates the state of a compilation. You can set up options on it, and add types, functions and code. Invoking :func:`gccjit::context::compile` on it gives you a :c:expr:`gcc_jit_result *`. It is a thin wrapper around the C API's :c:expr:`gcc_jit_context *`. Lifetime-management ------------------- Contexts are the unit of lifetime-management within the API: objects have their lifetime bounded by the context they are created within, and cleanup of such objects is done for you when the context is released. .. function:: gccjit::context gccjit::context::acquire () This function acquires a new :class:`gccjit::context` instance, which is independent of any others that may be present within this process. .. function:: void gccjit::context::release () This function releases all resources associated with the given context. Both the context itself and all of its :expr:`gccjit::object *` instances are cleaned up. It should be called exactly once on a given context. It is invalid to use the context or any of its "contextual" objects after calling this. .. code-block:: c++ ctxt.release (); .. function:: gccjit::context \ gccjit::context::new_child_context () Given an existing JIT context, create a child context. The child inherits a copy of all option-settings from the parent. The child can reference objects created within the parent, but not vice-versa. The lifetime of the child context must be bounded by that of the parent: you should release a child context before releasing the parent context. If you use a function from a parent context within a child context, you have to compile the parent context before you can compile the child context, and the gccjit::result of the parent context must outlive the gccjit::result of the child context. This allows caching of shared initializations. For example, you could create types and declarations of global functions in a parent context once within a process, and then create child contexts whenever a function or loop becomes hot. Each such child context can be used for JIT-compiling just one function or loop, but can reference types and helper functions created within the parent context. Contexts can be arbitrarily nested, provided the above rules are followed, but it's probably not worth going above 2 or 3 levels, and there will likely be a performance hit for such nesting. Thread-safety ------------- Instances of :class:`gccjit::context` created via :func:`gccjit::context::acquire` are independent from each other: only one thread may use a given context at once, but multiple threads could each have their own contexts without needing locks. Contexts created via :func:`gccjit::context::new_child_context` are related to their parent context. They can be partitioned by their ultimate ancestor into independent "family trees". Only one thread within a process may use a given "family tree" of such contexts at once, and if you're using multiple threads you should provide your own locking around entire such context partitions. Error-handling -------------- .. FIXME: How does error-handling work for C++ API? You can only compile and get code from a context if no errors occur. In general, if an error occurs when using an API entrypoint, it returns NULL. You don't have to check everywhere for NULL results, since the API gracefully handles a NULL being passed in for any argument. Errors are printed on stderr and can be queried using :func:`gccjit::context::get_first_error`. .. function:: const char *\ gccjit::context::get_first_error (gccjit::context *ctxt) Returns the first error message that occurred on the context. The returned string is valid for the rest of the lifetime of the context. If no errors occurred, this will be NULL. Debugging --------- .. function:: void\ gccjit::context::dump_to_file (const std::string &path, \ int update_locations) To help with debugging: dump a C-like representation to the given path, describing what's been set up on the context. If "update_locations" is true, then also set up :class:`gccjit::location` information throughout the context, pointing at the dump file as if it were a source file. This may be of use in conjunction with :c:macro:`GCC_JIT_BOOL_OPTION_DEBUGINFO` to allow stepping through the code in a debugger. .. function:: void\ gccjit::context::dump_reproducer_to_file (gcc_jit_context *ctxt,\ const char *path) This is a thin wrapper around the C API :c:func:`gcc_jit_context_dump_reproducer_to_file`, and hence works the same way. Note that the generated source is C code, not C++; this might be of use for seeing what the C++ bindings are doing at the C level. Options ------- String Options ************** .. function:: void \ gccjit::context::set_str_option (enum gcc_jit_str_option, \ const char *value) Set a string option of the context. This is a thin wrapper around the C API :c:func:`gcc_jit_context_set_str_option`; the options have the same meaning. Boolean options *************** .. function:: void \ gccjit::context::set_bool_option(enum gcc_jit_bool_option, \ int value) Set a boolean option of the context. This is a thin wrapper around the C API :c:func:`gcc_jit_context_set_bool_option`; the options have the same meaning. .. function:: void \ gccjit::context::set_bool_allow_unreachable_blocks (int bool_value) By default, libgccjit will issue an error about unreachable blocks within a function. This entrypoint can be used to disable that error; it is a thin wrapper around the C API :c:func:`gcc_jit_context_set_bool_allow_unreachable_blocks`. This entrypoint was added in :ref:`LIBGCCJIT_ABI_2`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_set_bool_allow_unreachable_blocks .. function:: void \ gccjit::context::set_bool_use_external_driver (int bool_value) libgccjit internally generates assembler, and uses "driver" code for converting it to other formats (e.g. shared libraries). By default, libgccjit will use an embedded copy of the driver code. This option can be used to instead invoke an external driver executable as a subprocess; it is a thin wrapper around the C API :c:func:`gcc_jit_context_set_bool_use_external_driver`. This entrypoint was added in :ref:`LIBGCCJIT_ABI_5`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_set_bool_use_external_driver Integer options *************** .. function:: void \ gccjit::context::set_int_option (enum gcc_jit_int_option, \ int value) Set an integer option of the context. This is a thin wrapper around the C API :c:func:`gcc_jit_context_set_int_option`; the options have the same meaning. Additional command-line options ******************************* .. function:: void \ gccjit::context::add_command_line_option (const char *optname) Add an arbitrary gcc command-line option to the context for use when compiling. This is a thin wrapper around the C API :c:func:`gcc_jit_context_add_command_line_option`. This entrypoint was added in :ref:`LIBGCCJIT_ABI_1`; you can test for its presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_gcc_jit_context_add_command_line_option --- ### Gcc Jit Docs Cp Topics Expressions (`gcc/jit/docs/cp/topics/expressions.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Expressions =========== Rvalues ------- .. class:: gccjit::rvalue A :class:`gccjit::rvalue` is an expression that can be computed. It is a subclass of :class:`gccjit::object`, and is a thin wrapper around :c:expr:`gcc_jit_rvalue *` from the C API. It can be simple, e.g.: * an integer value e.g. `0` or `42` * a string literal e.g. `"Hello world"` * a variable e.g. `i`. These are also lvalues (see below). or compound e.g.: * a unary expression e.g. `!cond` * a binary expression e.g. `(a + b)` * a function call e.g. `get_distance (&player_ship, &target)` * etc. Every rvalue has an associated type, and the API will check to ensure that types match up correctly (otherwise the context will emit an error). .. function:: gccjit::type gccjit::rvalue::get_type () Get the type of this rvalue. Simple expressions ****************** .. function:: gccjit::rvalue \ gccjit::context::new_rvalue (gccjit::type numeric_type, \ int value) const Given a numeric type (integer or floating point), build an rvalue for the given constant :expr:`int` value. .. function:: gccjit::rvalue \ gccjit::context::new_rvalue (gccjit::type numeric_type, \ long value) const Given a numeric type (integer or floating point), build an rvalue for the given constant :expr:`long` value. .. function:: gccjit::rvalue \ gccjit::context::zero (gccjit::type numeric_type) const Given a numeric type (integer or floating point), get the rvalue for zero. Essentially this is just a shortcut for: .. code-block:: c++ ctxt.new_rvalue (numeric_type, 0) .. function:: gccjit::rvalue \ gccjit::context::one (gccjit::type numeric_type) const Given a numeric type (integer or floating point), get the rvalue for one. Essentially this is just a shortcut for: .. code-block:: c++ ctxt.new_rvalue (numeric_type, 1) .. function:: gccjit::rvalue \ gccjit::context::new_rvalue (gccjit::type numeric_type, \ double value) const Given a numeric type (integer or floating point), build an rvalue for the given constant :expr:`double` value. .. function:: gccjit::rvalue \ gccjit::context::new_rvalue (gccjit::type pointer_type, \ void *value) const Given a pointer type, build an rvalue for the given address. .. function:: gccjit::rvalue \ gccjit::context::new_rvalue (const std::string &value) const Generate an rvalue of type :c:data:`GCC_JIT_TYPE_CONST_CHAR_PTR` for the given string. This is akin to a string literal. Vector expressions ****************** .. function:: gccjit::rvalue \ gccjit::context::new_rvalue (gccjit::type vector_type, \ std::vector<gccjit::rvalue> elements) const Given a vector type, and a vector of scalar rvalue elements, generate a vector rvalue. The number of elements needs to match that of the vector type. Unary Operations **************** .. function:: gccjit::rvalue \ gccjit::context::new_unary_op (enum gcc_jit_unary_op, \ gccjit::type result_type, \ gccjit::rvalue rvalue, \ gccjit::location loc) Build a unary operation out of an input rvalue. Parameter ``loc`` is optional. This is a thin wrapper around the C API's :c:func:`gcc_jit_context_new_unary_op` and the available unary operations are documented there. There are shorter ways to spell the various specific kinds of unary operation: .. function:: gccjit::rvalue \ gccjit::context::new_minus (gccjit::type result_type, \ gccjit::rvalue a, \ gccjit::location loc) Negate an arithmetic value; for example: .. code-block:: c++ gccjit::rvalue negpi = ctxt.new_minus (t_double, pi); builds the equivalent of this C expression: .. code-block:: c -pi .. function:: gccjit::rvalue \ new_bitwise_negate (gccjit::type result_type, \ gccjit::rvalue a, \ gccjit::location loc) Bitwise negation of an integer value (one's complement); for example: .. code-block:: c++ gccjit::rvalue mask = ctxt.new_bitwise_negate (t_int, a); builds the equivalent of this C expression: .. code-block:: c ~a .. function:: gccjit::rvalue \ new_logical_negate (gccjit::type result_type, \ gccjit::rvalue a, \ gccjit::location loc) Logical negation of an arithmetic or pointer value; for example: .. code-block:: c++ gccjit::rvalue guard = ctxt.new_logical_negate (t_bool, cond); builds the equivalent of this C expression: .. code-block:: c !cond The most concise way to spell them is with overloaded operators: .. function:: gccjit::rvalue operator- (gccjit::rvalue a) .. code-block:: c++ gccjit::rvalue negpi = -pi; .. function:: gccjit::rvalue operator~ (gccjit::rvalue a) .. code-block:: c++ gccjit::rvalue mask = ~a; .. function:: gccjit::rvalue operator! (gccjit::rvalue a) .. code-block:: c++ gccjit::rvalue guard = !cond; Binary Operations ***************** .. function:: gccjit::rvalue\ gccjit::context::new_binary_op (enum gcc_jit_binary_op, \ gccjit::type result_type, \ gccjit::rvalue a, \ gccjit::rvalue b, \ gccjit::location loc) Build a binary operation out of two constituent rvalues. Parameter ``loc`` is optional. This is a thin wrapper around the C API's :c:func:`gcc_jit_context_new_binary_op` and the available binary operations are documented there. There are shorter ways to spell the various specific kinds of binary operation: .. function:: gccjit::rvalue \ gccjit::context::new_plus (gccjit::type result_type, \ gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_minus (gccjit::type result_type, \ gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_mult (gccjit::type result_type, \ gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_divide (gccjit::type result_type, \ gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_modulo (gccjit::type result_type, \ gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_bitwise_and (gccjit::type result_type, \ gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_bitwise_xor (gccjit::type result_type, \ gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_bitwise_or (gccjit::type result_type, \ gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_logical_and (gccjit::type result_type, \ gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_logical_or (gccjit::type result_type, \ gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) The most concise way to spell them is with overloaded operators: .. function:: gccjit::rvalue operator+ (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue sum = a + b; .. function:: gccjit::rvalue operator- (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue diff = a - b; .. function:: gccjit::rvalue operator* (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue prod = a * b; .. function:: gccjit::rvalue operator/ (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue result = a / b; .. function:: gccjit::rvalue operator% (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue mod = a % b; .. function:: gccjit::rvalue operator& (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue x = a & b; .. function:: gccjit::rvalue operator^ (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue x = a ^ b; .. function:: gccjit::rvalue operator| (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue x = a | b; .. function:: gccjit::rvalue operator&& (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue cond = a && b; .. function:: gccjit::rvalue operator|| (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue cond = a || b; These can of course be combined, giving a terse way to build compound expressions: .. code-block:: c++ gccjit::rvalue discriminant = (b * b) - (four * a * c); Comparisons *********** .. function:: gccjit::rvalue \ gccjit::context::new_comparison (enum gcc_jit_comparison,\ gccjit::rvalue a, \ gccjit::rvalue b, \ gccjit::location loc) Build a boolean rvalue out of the comparison of two other rvalues. Parameter ``loc`` is optional. This is a thin wrapper around the C API's :c:func:`gcc_jit_context_new_comparison` and the available kinds of comparison are documented there. There are shorter ways to spell the various specific kinds of binary operation: .. function:: gccjit::rvalue \ gccjit::context::new_eq (gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_ne (gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_lt (gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_le (gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_gt (gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) gccjit::rvalue \ gccjit::context::new_ge (gccjit::rvalue a, gccjit::rvalue b, \ gccjit::location loc) The most concise way to spell them is with overloaded operators: .. function:: gccjit::rvalue \ operator== (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue cond = (a == ctxt.zero (t_int)); .. function:: gccjit::rvalue \ operator!= (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue cond = (i != j); .. function:: gccjit::rvalue \ operator< (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue cond = i < n; .. function:: gccjit::rvalue \ operator<= (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue cond = i <= n; .. function:: gccjit::rvalue \ operator> (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue cond = (ch > limit); .. function:: gccjit::rvalue \ operator>= (gccjit::rvalue a, gccjit::rvalue b) .. code-block:: c++ gccjit::rvalue cond = (score >= ctxt.new_rvalue (t_int, 100)); .. TODO: beyond this point Function calls ************** .. function:: gcc_jit_rvalue *\ gcc_jit_context_new_call (gcc_jit_context *ctxt,\ gcc_jit_location *loc,\ gcc_jit_function *func,\ int numargs , gcc_jit_rvalue **args) Given a function and the given table of argument rvalues, construct a call to the function, with the result as an rvalue. .. note:: :func:`gccjit::context::new_call` merely builds a :class:`gccjit::rvalue` i.e. an expression that can be evaluated, perhaps as part of a more complicated expression. The call *won't* happen unless you add a statement to a function that evaluates the expression. For example, if you want to call a function and discard the result (or to call a function with ``void`` return type), use :func:`gccjit::block::add_eval`: .. code-block:: c++ /* Add "(void)printf (arg0, arg1);". */ block.add_eval (ctxt.new_call (printf_func, arg0, arg1)); Function pointers ***************** .. function:: gccjit::rvalue \ gccjit::function::get_address (gccjit::location loc) Get the address of a function as an rvalue, of function pointer type. Type-coercion ************* .. function:: gccjit::rvalue \ gccjit::context::new_cast (gccjit::rvalue rvalue,\ gccjit::type type, \ gccjit::location loc) Given an rvalue of T, construct another rvalue of another type. Currently only a limited set of conversions are possible: * int <-> float * int <-> bool * P* <-> Q*, for pointer types P and Q Lvalues ------- .. class:: gccjit::lvalue An lvalue is something that can of the *left*-hand side of an assignment: a storage area (such as a variable). It is a subclass of :class:`gccjit::rvalue`, where the rvalue is computed by reading from the storage area. It iss a thin wrapper around :c:expr:`gcc_jit_lvalue *` from the C API. .. function:: gccjit::rvalue \ gccjit::lvalue::get_address (gccjit::location loc) Take the address of an lvalue; analogous to: .. code-block:: c &(EXPR) in C. Parameter "loc" is optional. Global variables **************** .. function:: gccjit::lvalue \ gccjit::context::new_global (enum gcc_jit_global_kind,\ gccjit::type type, \ const char *name, \ gccjit::location loc) Add a new global variable of the given type and name to the context. This is a thin wrapper around :c:func:`gcc_jit_context_new_global` from the C API; the "kind" parameter has the same meaning as there. Working with pointers, structs and unions ----------------------------------------- .. function:: gccjit::lvalue \ gccjit::rvalue::dereference (gccjit::location loc) Given an rvalue of pointer type ``T *``, dereferencing the pointer, getting an lvalue of type ``T``. Analogous to: .. code-block:: c++ *(EXPR) in C. Parameter "loc" is optional. If you don't need to specify the location, this can also be expressed using an overloaded operator: .. function:: gccjit::lvalue \ gccjit::rvalue::operator* () .. code-block:: c++ gccjit::lvalue content = *ptr; Field access is provided separately for both lvalues and rvalues: .. function:: gccjit::lvalue \ gccjit::lvalue::access_field (gccjit::field field, \ gccjit::location loc) Given an lvalue of struct or union type, access the given field, getting an lvalue of the field's type. Analogous to: .. code-block:: c++ (EXPR).field = ...; in C. .. function:: gccjit::rvalue \ gccjit::rvalue::access_field (gccjit::field field, \ gccjit::location loc) Given an rvalue of struct or union type, access the given field as an rvalue. Analogous to: .. code-block:: c++ (EXPR).field in C. .. function:: gccjit::lvalue \ gccjit::rvalue::dereference_field (gccjit::field field, \ gccjit::location loc) Given an rvalue of pointer type ``T *`` where T is of struct or union type, access the given field as an lvalue. Analogous to: .. code-block:: c++ (EXPR)->field in C, itself equivalent to ``(*EXPR).FIELD``. .. function:: gccjit::lvalue \ gccjit::context::new_array_access (gccjit::rvalue ptr, \ gccjit::rvalue index, \ gccjit::location loc) Given an rvalue of pointer type ``T *``, get at the element `T` at the given index, using standard C array indexing rules i.e. each increment of ``index`` corresponds to ``sizeof(T)`` bytes. Analogous to: .. code-block:: c++ PTR[INDEX] in C (or, indeed, to ``PTR + INDEX``). Parameter "loc" is optional. For array accesses where you don't need to specify a :class:`gccjit::location`, two overloaded operators are available: gccjit::lvalue gccjit::rvalue::operator[] (gccjit::rvalue index) .. code-block:: c++ gccjit::lvalue element = array[idx]; gccjit::lvalue gccjit::rvalue::operator[] (int index) .. code-block:: c++ gccjit::lvalue element = array[0]; --- ### Gcc Jit Docs Cp Topics Functions (`gcc/jit/docs/cp/topics/functions.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Creating and using functions ============================ Params ------ .. class:: gccjit::param A `gccjit::param` represents a parameter to a function. .. function:: gccjit::param \ gccjit::context::new_param (gccjit::type type,\ const char *name, \ gccjit::location loc) In preparation for creating a function, create a new parameter of the given type and name. :class:`gccjit::param` is a subclass of :class:`gccjit::lvalue` (and thus of :class:`gccjit::rvalue` and :class:`gccjit::object`). It is a thin wrapper around the C API's :c:expr:`gcc_jit_param *`. Functions --------- .. class:: gccjit::function A `gccjit::function` represents a function - either one that we're creating ourselves, or one that we're referencing. .. function:: gccjit::function \ gccjit::context::new_function (enum gcc_jit_function_kind,\ gccjit::type return_type, \ const char *name, \ std::vector<param> ¶ms, \ int is_variadic, \ gccjit::location loc) Create a gcc_jit_function with the given name and parameters. Parameters "is_variadic" and "loc" are optional. This is a wrapper around the C API's :c:func:`gcc_jit_context_new_function`. .. function:: gccjit::function \ gccjit::context::get_builtin_function (const char *name) This is a wrapper around the C API's :c:func:`gcc_jit_context_get_builtin_function`. .. function:: gccjit::param \ gccjit::function::get_param (int index) const Get the param of the given index (0-based). .. function:: void \ gccjit::function::dump_to_dot (const char *path) Emit the function in graphviz format to the given path. .. function:: gccjit::lvalue \ gccjit::function::new_local (gccjit::type type,\ const char *name, \ gccjit::location loc) Create a new local variable within the function, of the given type and name. Blocks ------ .. class:: gccjit::block A `gccjit::block` represents a basic block within a function i.e. a sequence of statements with a single entry point and a single exit point. :class:`gccjit::block` is a subclass of :class:`gccjit::object`. The first basic block that you create within a function will be the entrypoint. Each basic block that you create within a function must be terminated, either with a conditional, a jump, a return, or a switch. It's legal to have multiple basic blocks that return within one function. .. function:: gccjit::block \ gccjit::function::new_block (const char *name) Create a basic block of the given name. The name may be NULL, but providing meaningful names is often helpful when debugging: it may show up in dumps of the internal representation, and in error messages. Statements ---------- .. function:: void\ gccjit::block::add_eval (gccjit::rvalue rvalue, \ gccjit::location loc) Add evaluation of an rvalue, discarding the result (e.g. a function call that "returns" void). This is equivalent to this C code: .. code-block:: c (void)expression; .. function:: void\ gccjit::block::add_assignment (gccjit::lvalue lvalue, \ gccjit::rvalue rvalue, \ gccjit::location loc) Add evaluation of an rvalue, assigning the result to the given lvalue. This is roughly equivalent to this C code: .. code-block:: c lvalue = rvalue; .. function:: void\ gccjit::block::add_assignment_op (gccjit::lvalue lvalue, \ enum gcc_jit_binary_op, \ gccjit::rvalue rvalue, \ gccjit::location loc) Add evaluation of an rvalue, using the result to modify an lvalue. This is analogous to "+=" and friends: .. code-block:: c lvalue += rvalue; lvalue *= rvalue; lvalue /= rvalue; etc. For example: .. code-block:: c /* "i++" */ loop_body.add_assignment_op ( i, GCC_JIT_BINARY_OP_PLUS, ctxt.one (int_type)); .. function:: void\ gccjit::block::add_comment (const char *text, \ gccjit::location loc) Add a no-op textual comment to the internal representation of the code. It will be optimized away, but will be visible in the dumps seen via :c:macro:`GCC_JIT_BOOL_OPTION_DUMP_INITIAL_TREE` and :c:macro:`GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE`, and thus may be of use when debugging how your project's internal representation gets converted to the libgccjit IR. Parameter "loc" is optional. .. function:: void\ gccjit::block::end_with_conditional (gccjit::rvalue boolval,\ gccjit::block on_true,\ gccjit::block on_false, \ gccjit::location loc) Terminate a block by adding evaluation of an rvalue, branching on the result to the appropriate successor block. This is roughly equivalent to this C code: .. code-block:: c if (boolval) goto on_true; else goto on_false; block, boolval, on_true, and on_false must be non-NULL. .. function:: void\ gccjit::block::end_with_jump (gccjit::block target, \ gccjit::location loc) Terminate a block by adding a jump to the given target block. This is roughly equivalent to this C code: .. code-block:: c goto target; .. function:: void\ gccjit::block::end_with_return (gccjit::rvalue rvalue, \ gccjit::location loc) Terminate a block. Both params are optional. An rvalue must be provided for a function returning non-void, and must not be provided by a function "returning" `void`. If an rvalue is provided, the block is terminated by evaluating the rvalue and returning the value. This is roughly equivalent to this C code: .. code-block:: c return expression; If an rvalue is not provided, the block is terminated by adding a valueless return, for use within a function with "void" return type. This is equivalent to this C code: .. code-block:: c return; .. class:: gccjit::case_ A `gccjit::case_` represents a case within a switch statement, and is created within a particular :class:`gccjit::context` using :func:`gccjit::context::new_case`. It is a subclass of :class:`gccjit::object`. Each case expresses a multivalued range of integer values. You can express single-valued cases by passing in the same value for both `min_value` and `max_value`. .. function:: gccjit::case_ *\ gccjit::context::new_case (gccjit::rvalue min_value,\ gccjit::rvalue max_value,\ gccjit::block dest_block) Create a new gccjit::case for use in a switch statement. `min_value` and `max_value` must be constants of an integer type, which must match that of the expression of the switch statement. `dest_block` must be within the same function as the switch statement. .. function:: void\ gccjit::block::end_with_switch (gccjit::rvalue expr,\ gccjit::block default_block,\ std::vector <gccjit::case_> cases,\ gccjit::location loc) Terminate a block by adding evaluation of an rvalue, then performing a multiway branch. This is roughly equivalent to this C code: .. code-block:: c switch (expr) { default: goto default_block; case C0.min_value ... C0.max_value: goto C0.dest_block; case C1.min_value ... C1.max_value: goto C1.dest_block; ...etc... case C[N - 1].min_value ... C[N - 1].max_value: goto C[N - 1].dest_block; } ``expr`` must be of the same integer type as all of the ``min_value`` and ``max_value`` within the cases. The ranges of the cases must not overlap (or have duplicate values). The API entrypoints relating to switch statements and cases: * :func:`gccjit::block::end_with_switch` * :func:`gccjit::context::new_case` were added in :ref:`LIBGCCJIT_ABI_3`; you can test for their presence using .. code-block:: c #ifdef LIBGCCJIT_HAVE_SWITCH_STATEMENTS Here's an example of creating a switch statement: .. literalinclude:: ../../../../testsuite/jit.dg/test-switch.cc :start-after: /* Quote from here in docs/cp/topics/functions.rst. */ :end-before: /* Quote up to here in docs/cp/topics/functions.rst. */ :language: c++ --- ### Gcc Jit Docs Cp Topics Index (`gcc/jit/docs/cp/topics/index.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. Topic Reference =============== .. toctree:: :maxdepth: 2 contexts.rst objects.rst types.rst expressions.rst functions.rst locations.rst compilation.rst asm.rst --- ### Gcc Jit Docs Cp Topics Locations (`gcc/jit/docs/cp/topics/locations.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Source Locations ================ .. class:: gccjit::location A `gccjit::location` encapsulates a source code location, so that you can (optionally) associate locations in your language with statements in the JIT-compiled code, allowing the debugger to single-step through your language. `gccjit::location` instances are optional: you can always omit them from any C++ API entrypoint accepting one. You can construct them using :func:`gccjit::context::new_location`. You need to enable :c:macro:`GCC_JIT_BOOL_OPTION_DEBUGINFO` on the :class:`gccjit::context` for these locations to actually be usable by the debugger: .. code-block:: cpp ctxt.set_bool_option (GCC_JIT_BOOL_OPTION_DEBUGINFO, 1); .. function:: gccjit::location \ gccjit::context::new_location (const char *filename, \ int line, \ int column) Create a `gccjit::location` instance representing the given source location. Faking it --------- If you don't have source code for your internal representation, but need to debug, you can generate a C-like representation of the functions in your context using :func:`gccjit::context::dump_to_file()`: .. code-block:: cpp ctxt.dump_to_file ("/tmp/something.c", 1 /* update_locations */); This will dump C-like code to the given path. If the `update_locations` argument is true, this will also set up `gccjit::location` information throughout the context, pointing at the dump file as if it were a source file, giving you *something* you can step through in the debugger. --- ### Gcc Jit Docs Cp Topics Objects (`gcc/jit/docs/cp/topics/objects.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Objects ======= .. class:: gccjit::object Almost every entity in the API (with the exception of :class:`gccjit::context` and :c:expr:`gcc_jit_result *`) is a "contextual" object, a :class:`gccjit::object`. A JIT object: * is associated with a :class:`gccjit::context`. * is automatically cleaned up for you when its context is released so you don't need to manually track and cleanup all objects, just the contexts. The C++ class hierarchy within the ``gccjit`` namespace looks like this:: +- object +- location +- type +- struct +- field +- function +- block +- rvalue +- lvalue +- param +- case_ The :class:`gccjit::object` base class has the following operations: .. function:: gccjit::context gccjit::object::get_context () const Which context is the obj within? .. function:: std::string gccjit::object::get_debug_string () const Generate a human-readable description for the given object. For example, .. code-block:: c++ printf ("obj: %s\n", obj.get_debug_string ().c_str ()); might give this text on stdout: .. code-block:: bash obj: 4.0 * (float)i --- ### Gcc Jit Docs Cp Topics Types (`gcc/jit/docs/cp/topics/types.rst`) .. Copyright (C) 2014-2026 Free Software Foundation, Inc. Originally contributed by David Malcolm <dmalcolm@redhat.com> This is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>. .. default-domain:: cpp Types ===== .. class:: gccjit::type gccjit::type represents a type within the library. It is a subclass of :class:`gccjit::object`. Types can be created in several ways: * fundamental types can be accessed using :func:`gccjit::context::get_type`: .. code-block:: c++ gccjit::type int_type = ctxt.get_type (GCC_JIT_TYPE_INT); or using the :func:`gccjit::context::get_int_type<T>` template: .. code-block:: c++ gccjit::type t = ctxt.get_int_type <unsigned short> (); See :c:func:`gcc_jit_context_get_type` for the available types. * derived types can be accessed by using functions such as :func:`gccjit::type::get_pointer` and :func:`gccjit::type::get_const`: .. code-block:: c++ gccjit::type const_int_star = int_type.get_const ().get_pointer (); gccjit::type int_const_star = int_type.get_pointer ().get_const (); * by creating structures (see below). Standard types -------------- .. function:: gccjit::type gccjit::context::get_type (enum gcc_jit_types) Access a specific type. This is a thin wrapper around :c:func:`gcc_jit_context_get_type`; the parameter has the same meaning. .. function:: gccjit::type \ gccjit::context::get_int_type (size_t num_bytes, int is_signed) Access the integer type of the given size. .. function:: template<typename T> gccjit::type \ gccjit::context::get_int_type <T> () Access the given integer type. For example, you could map the ``unsigned short`` type into a gccjit::type via: .. code-block:: c++ gccjit::type t = ctxt.get_int_type <unsigned short> (); Pointers, `const`, and `volatile` --------------------------------- .. function:: gccjit::type gccjit::type::get_pointer () Given type "T", get type "T*". .. function:: gccjit::type gccjit::type::get_const () Given type "T", get type "const T". .. function:: gccjit::type gccjit::type::get_volatile () Given type "T", get type "volatile T". .. function:: gccjit::type gccjit::type::get_aligned (size_t alignment_in_bytes) Given type "T", get type: .. code-block:: c T __attribute__ ((aligned (ALIGNMENT_IN_BYTES))) The alignment must be a power of two. .. function:: gccjit::type \ gccjit::context::new_array_type (gccjit::type element_type, \ int num_elements, \ gccjit::location loc) Given type "T", get type "T[N]" (for a constant N). Param "loc" is optional. Vector types ------------ .. function:: gccjit::type gccjit::type::get_vector (size_t num_units) Given type "T", get type: .. code-block:: c T __attribute__ ((vector_size (sizeof(T) * num_units)) T must be integral or floating point; num_units must be a power of two. Structures and unions --------------------- .. class:: gccjit::struct_ A compound type analogous to a C `struct`. :class:`gccjit::struct_` is a subclass of :class:`gccjit::type` (and thus of :class:`gccjit::object` in turn). .. class:: gccjit::field A field within a :class:`gccjit::struct_`. :class:`gccjit::field` is a subclass of :class:`gccjit::object`. You can model C `struct` types by creating :class:`gccjit::struct_` and :class:`gccjit::field` instances, in either order: * by creating the fields, then the structure. For example, to model: .. code-block:: c struct coord {double x; double y; }; you could call: .. code-block:: c++ gccjit::field field_x = ctxt.new_field (double_type, "x"); gccjit::field field_y = ctxt.new_field (double_type, "y"); std::vector fields; fields.push_back (field_x); fields.push_back (field_y); gccjit::struct_ coord = ctxt.new_struct_type ("coord", fields); * by creating the structure, then populating it with fields, typically to allow modelling self-referential structs such as: .. code-block:: c struct node { int m_hash; struct node *m_next; }; like this: .. code-block:: c++ gccjit::struct_ node = ctxt.new_opaque_struct_type ("node"); gccjit::type node_ptr = node.get_pointer (); gccjit::field field_hash = ctxt.new_field (int_type, "m_hash"); gccjit::field field_next = ctxt.new_field (node_ptr, "m_next"); std::vector fields; fields.push_back (field_hash); fields.push_back (field_next); node.set_fields (fields); .. FIXME: the above API doesn't seem to exist yet .. function:: gccjit::field \ gccjit::context::new_field (gccjit::type type,\ const char *name, \ gccjit::location loc) Construct a new field, with the given type and name. .. function:: gccjit::struct_ \ gccjit::context::new_struct_type (const std::string &name,\ std::vector<field> &fields,\ gccjit::location loc) Construct a new struct type, with the given name and fields. .. function:: gccjit::struct_ \ gccjit::context::new_opaque_struct (const std::string &name, \ gccjit::location loc) Construct a new struct type, with the given name, but without specifying the fields. The fields can be omitted (in which case the size of the struct is not known), or later specified using :c:func:`gcc_jit_struct_set_fields`. --- --- METRICS --- - Files Extracted: 96 - Estimated Token Budget: ~596112 tokens - Canonical Reference: https://codewiki.google/github.com/gcc-mirror/gcc