HACKING - master - niceSOFT/gettext

HACKING
558 строк21 KB
Bruno Haible
HACKING: Explain how to use the undefined-behaviour sanitizer.
05 май 2026, 18:15
05 май 2026, 18:15d760a40
All you need to know when hacking (modifying) GNU gettext or when building
it off the git repository.


Requirements
============

You will need reasonably recent versions of the build tools:

  * A C compiler; on Cygwin or mingw also a C++ compiler. Such as GNU GCC.
    + Homepage:
      https://gcc.gnu.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: gcc, gcc-doc,
      - On Red Hat distributions: gcc.
      - Other: https://repology.org/project/gcc/versions

  * GNU automake 1.13 or newer
    + Homepage:
      https://www.gnu.org/software/automake/
    + Pre-built package name:
      - On Debian and Debian-based systems: automake,
      - On Red Hat distributions: automake.
      - Other: https://repology.org/project/automake/versions

  * GNU autoconf
    + Homepage:
      https://www.gnu.org/software/autoconf/
    + Pre-built package name:
      - On Debian and Debian-based systems: autoconf, autoconf-doc,
      - On Red Hat distributions: autoconf.
      - Other: https://repology.org/project/autoconf/versions

  * GNU m4
    + Homepage:
      https://www.gnu.org/software/m4/
    + Pre-built package name:
      - On Debian and Debian-based systems: m4, m4-doc,
      - On Red Hat distributions: m4.
      - Other: https://repology.org/project/m4/versions

  * GNU bison 3.0 or newer
    + Homepage:
      https://www.gnu.org/software/bison/
    + Pre-built package name:
      - On Debian and Debian-based systems: bison, bison-doc,
      - On Red Hat distributions: bison.
      - Other: https://repology.org/project/bison/versions

  * GNU gperf
    + Homepage:
      https://www.gnu.org/software/gperf/
    + Pre-built package name:
      - On Debian and Debian-based systems: gperf,
      - On Red Hat distributions: gperf.
      - Other: https://repology.org/project/gperf/versions

  * GNU groff 1.17 or newer
    + Homepage:
      https://www.gnu.org/software/groff/
    + Pre-built package name:
      - On Debian and Debian-based systems: groff,
      - On Red Hat distributions: groff.
      - Other: https://repology.org/project/groff/versions

  * GNU texinfo
    + Homepage:
      https://www.gnu.org/software/texinfo/
    + Pre-built package name:
      - On Debian and Debian-based systems: texinfo,
      - On Red Hat distributions: texinfo.
      - Other: https://repology.org/project/texinfo/versions

  * GNU emacs
    + Homepage:
      https://www.gnu.org/software/emacs/
    + Pre-built package name:
      - On Debian and Debian-based systems: emacs,
      - On Red Hat distributions: emacs.
      - Other: https://repology.org/project/emacs/versions

  * Perl
    + Homepage:
      https://www.perl.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: perl, perl-doc,
      - On Red Hat distributions: perl.
      - Other: https://repology.org/project/perl/versions

  * The Free Pascal compiler ppc386 or ppcx64
    + Homepage: https://www.freepascal.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: fp-compiler, fp-docs,
      - On Red Hat distributions: fpc (EPEL).
      - Other: https://repology.org/project/fpc/versions
  * The Free Pascal Free Component Library
    + Homepage: https://www.freepascal.org/fcl/fcl.var
    + Pre-built package name:
      - On Debian and Debian-based systems: fp-units-fcl,
      - On Red Hat distributions: --.
      - Other: https://repology.org/project/fpc/versions

  * Either an internet connection or a recent copy of GNU gnulib.
    + Homepage:
      https://www.gnu.org/software/gnulib/

  * TeX (for making the doc in .dvi, .ps or .pdf format)
    + Homepage:
      https://tug.org/texlive/
    + Pre-built package name:
      - On Debian and Debian-based systems: texlive-base, texlive-latex-base,
      - On Red Hat distributions: texlive-base.
      - Other: https://repology.org/project/texlive/versions

  * GNU tar (for creating distributable tarballs)
    + Homepage:
      https://www.gnu.org/software/tar/
    + Pre-built package name:
      - On Debian and Debian-based systems: tar, tar-doc,
      - On Red Hat distributions: tar.
      - Other: https://repology.org/project/tar/versions

  * GNU Wget
    + Homepage:
      https://www.gnu.org/software/wget/
    + Pre-built package name:
      - On Debian and Debian-based systems: wget,
      - On Red Hat distributions: wget.
      - Other: https://repology.org/project/wget/versions

And, of course, the packages listed in the DEPENDENCIES file.


Additional requirements for maintainers
---------------------------------------

If you are a maintainer, you will want to minimize the number of tests that
are skipped. To this effect, you need to install also:

  * The xmllint program
    + Homepage: https://gitlab.gnome.org/GNOME/libxml2/-/wikis/home
    + Pre-built package name:
      - On Debian and Debian-based systems: libxml2-utils,
      - On Red Hat distributions: libxml2.
      - Other: https://repology.org/project/libxml2/versions
  * The Xerces2 Java XML parser (binary distribution named 'xml-schema-1.1')
    + Homepage: https://xerces.apache.org/xerces2-j/
    + Pre-built package name: --
    + See gettext-tools/build-aux/xml-validate-11.in for more details.

  * A C++ compiler
    + Homepage: https://gcc.gnu.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: g++,
      - On Red Hat distributions: gcc-c++.
      - Other: https://repology.org/project/gcc/versions
  * An Objective C compiler
    + Homepage: https://gcc.gnu.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: gobjc,
      - On Red Hat distributions: gcc-objc.
      - Other: https://repology.org/project/gcc/versions
  * The Boost C++ library
    + Homepage: https://www.boost.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: libboost1.83-dev,
      - On Red Hat distributions: boost.
      - Other: https://repology.org/project/boost/versions

  * Python
    + Homepage: https://www.python.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: python3, python3-doc,
      - On Red Hat distributions: python3.
      - Other: https://repology.org/project/python/versions

  * A Java compiler
    + Homepage: http://openjdk.java.net/
    + Pre-built package name:
      - On Debian and Debian-based systems:
        openjdk-11-jdk or openjdk-8-jdk, default-jdk,
      - On Red Hat distributions:
        java-11-openjdk-devel or java-1.8.0-openjdk-devel.
      - Other: https://repology.org/project/openjdk/versions

  * A C# compiler
    Either Mono or dotnet.
    o Mono
      + Wikipedia: https://en.wikipedia.org/wiki/Mono_(software)
      + Homepage: https://www.mono-project.com/
      + Pre-built package name:
        - On Debian and Debian-based systems: mono-mcs,
        - On Red Hat distributions: mono-devel (EPEL).
        - Other: https://repology.org/project/mono/versions
    o dotnet
      + Wikipedia: https://en.wikipedia.org/wiki/.NET
      + Homepage: https://dotnet.microsoft.com/
      + Pre-built package name:
        - On Debian and Debian-based systems: dotnet7 or dotnet8,
        - On Red Hat distributions: dotnet7.0 or dotnet8.0.
        - Other: https://repology.org/project/dotnet/versions

  * The GNOME/Mozilla JavaScript interpreter
    + Homepage: https://wiki.gnome.org/Projects/Gjs
    + Pre-built package name:
      - On Debian and Debian-based systems: gjs, gir1.2-gtk-3.0,
      - On Red Hat distributions: gjs.
      - Other: https://repology.org/project/gjs/versions

  * GNU guile
    + Homepage: https://www.gnu.org/software/guile/
    + Pre-built package name:
      - On Debian and Debian-based systems: guile-2.2, guile-2.2-doc,
        or:                                 guile-3.0, guile-3.0-doc,
      - On Red Hat distributions: guile.
      - Other: https://repology.org/project/guile/versions

  * GNU clisp
    + Homepage: http://clisp.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: clisp, clisp-doc,
      - On Red Hat distributions: --.
      - Other: https://repology.org/project/clisp/versions

  * librep
    + Homepage: http://librep.sourceforge.net/
    + Pre-built package name:
      - On Debian and Debian-based systems: rep, rep-doc,
      - On Red Hat distributions: --.
      - Other: https://repology.org/project/librep/versions

  * Rust
    + Homepage: https://www.rust-lang.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: rustc, cargo,
      - On Red Hat distributions: rust, rust-cargo.
      - Other: https://repology.org/project/rust/versions
               https://repology.org/project/cargo/versions

  * Go
    Either the golang Go implementation or the GCC Go implementation.
    o golang Go implementation
      + Recommended where available.
      + Homepage: https://go.dev/
      + Documentation: https://go.dev/doc/
      + Pre-built package name:
        - On Debian and Debian-based systems: golang-go,
        - On Red Hat distributions: golang.
        - Other: https://repology.org/project/go/versions
    o GCC Go implementation
      + Recommended on those platforms where the golang Go implementation
        is not available. This includes in particular:
        - Linux/alpha, Linux/hppa, Linux/m68k, Linux/sparc, Linux/powerpc (32-bit)
        - GNU/Hurd
        - Solaris/sparc
        - Haiku
      + Homepage: https://gcc.gnu.org/
      + Documentation:
        - https://go.dev/doc/install/gccgo
        - https://gcc.gnu.org/onlinedocs/gccgo/index.html
      + Pre-built package name:
        - On Debian and Debian-based systems: gccgo,
        - On Red Hat distributions: gcc-go.
        - Other: https://repology.org/project/gcc/versions
                 https://repology.org/project/gccgo/versions

  * Ruby
    + Homepage: https://www.ruby-lang.org/en/
    + Pre-built package name:
      - On Debian and Debian-based systems: ruby,
      - On Red Hat distributions: ruby.
      - Other: https://repology.org/project/ruby/versions
  * The ruby-gettext package
    + Homepage: https://ruby-gettext.github.io/
    + Pre-built package name:
      - On Debian and Debian-based systems: ruby-gettext,
      - On Red Hat distributions: ruby-gettext.
      - Other: https://repology.org/project/ruby:gettext/versions

  * GNU awk
    + Homepage: https://www.gnu.org/software/gawk/
    + Pre-built package name:
      - On Debian and Debian-based systems: gawk, gawk-doc,
      - On Red Hat distributions: gawk.
      - Other: https://repology.org/project/gawk/versions

  * lua
    + Homepage: https://www.lua.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: lua5.4 or lua5.3 or lua5.2 or lua5.1,
      - On Red Hat distributions: lua.
      - Other: https://repology.org/project/lua/versions
  * The lua-gettext package
    + Homepage: https://gitlab.com/sukhichev/lua-gettext/blob/master/README.us.md
    + Debian and Ubuntu packages:
      Download (following the hyperlinks in the homepage), then install through
      $ sudo dpkg -i lua-gettext_0.0+nmu1_amd64.deb
    + Pre-built package name:
      - On Debian and Debian-based systems: --,
      - On Red Hat distributions: --.
      - Other: https://repology.org/project/lua:gettext/versions

  * OCaml
    + Homepage: https://ocaml.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: ocaml,
      - On Red Hat distributions: ocaml.
      - Other: https://repology.org/project/ocaml/versions
  * The OCaml package manager
    + Homepage: https://opam.ocaml.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: opam,
      - On Red Hat distributions: --.
      - Other: https://repology.org/project/opam/versions

  * GNU smalltalk
    + Homepage: http://smalltalk.gnu.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: gnu-smalltalk, gnu-smalltalk-doc
        (only up to Debian 11),
      - On Red Hat distributions: gnu-smalltalk (EPEL).
      - Other: https://repology.org/project/smalltalk/versions

  * The GNOME Vala compiler
    + Homepage: https://wiki.gnome.org/Projects/Vala
    + Pre-built package name:
      - On Debian and Debian-based systems: valac,
      - On Red Hat distributions: vala.
      - Other: https://repology.org/project/vala/versions

  * Tcl
    + Homepage: https://www.tcl.tk/
    + Pre-built package name:
      - On Debian and Debian-based systems: tcl, tcl-doc,
      - On Red Hat distributions: tcl.
      - Other: https://repology.org/project/tcl/versions

  * Perl
    + Homepage: https://www.perl.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: perl, perl-doc,
      - On Red Hat distributions: perl.
      - Other: https://repology.org/project/perl/versions
  * The Perl packages Locale::Messages and libintl-perl
    + Homepage: http://search.cpan.org/~guido/libintl-perl/
    + Pre-built package name:
      - On Debian and Debian-based systems: libintl-perl,
      - On Red Hat distributions: perl-libintl.
      - Other: https://repology.org/project/perl:libintl-perl/versions

  * php
    + Homepage: http://www.php.net/
    + Pre-built package name:
      - On Debian and Debian-based systems: php,
      - On Red Hat distributions: php.
      - Other: https://repology.org/project/php/versions

If you want to exercise the sample programs in gettext-tools/examples/, you
also need to install:

  * The Gtk3 GUI toolkit library
    + Homepage: https://www.gtk.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: libgtk-3-dev,
      - On Red Hat distributions: gtk3-devel.
      - Other: https://repology.org/project/gtk/versions
  * Its C++ bindings
    + Homepage: https://gtkmm.gnome.org/
      and:      https://www.gtk.org/docs/language-bindings/cpp
    + Pre-built package name:
      - On Debian and Debian-based systems: libgtkmm-3.0-dev,
      - On Red Hat distributions: gtkmm3.0-devel.
      - Other: https://repology.org/project/gtkmm/versions

  * The Qt GUI toolkit library
    + Homepage: https://www.qt.io/
    + Pre-built package name:
      - On Debian and Debian-based systems: libqt5gui5, qt5-doc-html,
        or:                                 libqt4-dev,
      - On Red Hat distributions: --.
      - Other: https://repology.org/project/qt/versions

  * The wxWidgets GUI toolkit library
    + Homepage: https://www.wxwidgets.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: libwxgtk3.0-gtk3-dev, wx3.0-doc,
        or:                                 libwxgtk3.0-dev,
        or:                                 libwxgtk3.2-dev,
      - On Red Hat distributions: --.
      - Other: https://repology.org/project/wxwidgets/versions

  * The GNUstep toolkit and development environment
    + Homepage: http://www.gnustep.org/
    + Pre-built package name:
      - On Debian and Debian-based systems: gnustep-devel,
      - On Red Hat distributions: gnustep-base (EPEL).
      - Other: https://repology.org/project/gnustep-base/versions

  * The C# System.Windows.Forms implementation
    + Homepage: https://www.mono-project.com/
    + Pre-built package name:
      - On Debian and Debian-based systems: libmono-system-windows-forms4.0-cil,
      - On Red Hat distributions: mono-winforms (EPEL).
      - Other: https://pkgs.org/search/?q=mono-winforms

If you are going to make releases, you will also need, for creating a small
autopoint archive file:

  * XZ Utils
    + Homepage:
      https://tukaani.org/xz/
    + Pre-built package name:
      - On Debian and Debian-based systems: xz-utils,
      - On Red Hat distributions: xz.
      - Other: https://repology.org/project/xz/versions

And you need to install a couple of locales, used by the tests.
By default, only those enabled in /etc/locale.gen are preinstalled.

$ sudo localedef -i ar_SA -f ISO-8859-6 ar_SA.ISO-8859-6
$ sudo localedef -i de_DE -f ISO-8859-1 de_DE.ISO-8859-1
$ sudo localedef -i de_DE -f UTF-8 de_DE.UTF-8
$ sudo localedef -i es_ES -f UTF-8 es_ES.UTF-8
$ sudo localedef -i fa_IR -f UTF-8 fa_IR
$ sudo localedef -i fr_FR -f ISO-8859-1 fr_FR.ISO-8859-1
$ sudo localedef -i fr_FR -f UTF-8 fr_FR.UTF-8
$ sudo localedef -i ja_JP -f EUC-JP ja_JP.EUC-JP
$ sudo localedef -i tr_TR -f UTF-8 tr_TR.UTF-8
$ sudo localedef -i zh_CN -f GB18030 zh_CN.GB18030
$ sudo localedef -i zh_HK -f BIG5-HKSCS zh_HK.BIG5-HKSCS


Building off the Git repository
===============================

Access to the Git repository is described at
https://savannah.gnu.org/git/?group=gettext .

After fetching the sources from the Git repository, peek at the comments in
autogen.sh, then run
  ./autopull.sh
  ./autogen.sh
Then you can proceed with "./configure" as usual.

Each time you want to update the source, do not only "git pull".  Instead do
  git pull && ./autopull.sh
  ./autogen.sh


Continuous integration
======================

The package is built automatically, at regular intervals.  You find the latest
build results here:
  https://gitlab.com/gnu-gettext/ci-distcheck/pipelines
  https://gitlab.com/gnu-gettext/ci-distcheck/-/jobs?scope=finished


Coding Style
============

For C source code:

* Follow the portability guidelines
  https://www.gnu.org/software/gnulib/manual/html_node/Portability-guidelines.html

* Use mixed declarations and statements, as appropriate. (This is supported
  since C99.)

* Note: You can't put a label before a declaration, such as in
    retry:
      int foo = ...;
  This is supported only in C23 or newer. Instead, add an empty statement:
    retry: ;
      int foo = ...;
  Similarly, you can't put a declaration immediately after a case label, such
  as in
    case 1:
      int foo = ...;
      ...
      break;
  Again, this is supported only in C23 or newer. Instead, use a sub-block:
    case 1:
      {
        int foo = ...;
        ...
      }
      break;

* In function definitions and blocks, the pre-C99 style "declare all variables
  upfront" is discouraged.
  Instead, the following style elements are encouraged (for local non-static
  variables):
    - Move the declaration of a variable to its initialization.
    - Reduce the scope of variables:
      . If a variable is only used in a sub-block, declare it in that sub-block.
      . If a variable is only used in several sub-blocks and has a separate
        initialization in each of the sub-blocks, declare it in each of these
        sub-blocks. It's actually several independent variables.
      . If a variable is only used in a small part of a function definition or
        block (say, 1/3 of it or less), you may surround that part (including
        the variable's initialization) with braces, so that it becomes clear
        that the variable is not used in the rest of the function / block.
    - Usually, try to declare and initialize variables right when they are
      needed for the first time, not several computation steps before.
  The rationale is:
    - It follows the general principle "Put related things close together."
    - A function with 3 blocks that each has 4 variables is easier to
      understand than a function that has 12 variables, and where the
      reader has to find out which variable is used where.
    - Variables may represent state, and it is the state and the control flow
      that contribute to the complexity of a function.
  Only static variables are better kept declared upfront in a function or block.


Customizing the Undefined Behaviour Sanitizer (UBSAN)
=====================================================

Using gcc with UBSAN is as usual:

  CC="gcc -fsanitize=undefined,signed-integer-overflow,shift,integer-divide-by-zero -fno-sanitize-recover=undefined"
  CXX="g++ -fsanitize=undefined,signed-integer-overflow,shift,integer-divide-by-zero -fno-sanitize-recover=undefined"
  CFLAGS="-O1 -fno-omit-frame-pointer -ggdb"
  CXXFLAGS="-O1 -fno-omit-frame-pointer -ggdb"
  export CC CXX CFLAGS CXXFLAGS

Using clang with UBSAN needs a customization, to avoid errors such as
  "runtime error: call to function get_control_data through pointer to incorrect function type"
coming from libtextstyle, namely

  CC="clang -fsanitize=undefined,signed-integer-overflow,shift,integer-divide-by-zero -fno-sanitize=function -fno-sanitize-recover=undefined"
  CXX="clang++ -fsanitize=undefined,signed-integer-overflow,shift,integer-divide-by-zero -fno-sanitize=function -fno-sanitize-recover=undefined"
  CFLAGS="-O1 -fno-omit-frame-pointer -ggdb"
  CXXFLAGS="-O1 -fno-omit-frame-pointer -ggdb"
  export CC CXX CFLAGS CXXFLAGS


Submitting patches
==================

Patches should be sent to bug-gettext@gnu.org, the bug/feature mailing
list.  You can subscribe to the mailing list, or see the list
archives, by following links from
https://savannah.gnu.org/mail/?group=gettext .

To email a patch you can use a shell command like 'git format-patch -1'
to create a file, and then attach the file to your email.

GNU gettext development no longer stores descriptions of new changes
in ChangeLog files.  Instead, a single ChangeLog file is generated
from the commit messages when a release is prepared.  So changes you
commit should not touch any of the ChangeLog files in the repository,
but instead should contain the log entries in the commit message.

For the style of a ChangeLog entry, see the "Change Logs" section of
the GNU coding standards:

  https://www.gnu.org/prep/standards/html_node/Change-Logs.html