path: root/libstdc++-v3/doc/xml/manual/build_hacking.xml

<section xmlns="http://docbook.org/ns/docbook" version="5.0"
	 xml:id="appendix.porting.build_hacking" xreflabel="Build Hacking">
<?dbhtml filename="build_hacking.html"?>

<info><title>Configure and Build Hacking</title>
  <keywordset>
    <keyword>C++</keyword>
    <keyword>build</keyword>
    <keyword>configure</keyword>
    <keyword>hacking</keyword>
    <keyword>version</keyword>
    <keyword>dynamic</keyword>
    <keyword>shared</keyword>
  </keywordset>
</info>

<section xml:id="build_hacking.prereq"><info><title>Prerequisites</title></info>

  <para>
    As noted <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://gcc.gnu.org/install/prerequisites.html">previously</link>,
    certain other tools are necessary for hacking on files that
    control configure (<code>configure.ac</code>,
    <code>acinclude.m4</code>) and make
    (<code>Makefile.am</code>). These additional tools
    (<code>automake</code>, and <code>autoconf</code>) are further
    described in detail in their respective manuals. All the libraries
    in GCC try to stay in sync with each other in terms of versions of
    the auto-tools used, so please try to play nicely with the
    neighbors.
  </para>
</section>

<section xml:id="build_hacking.overview">
<info><title>Overview</title></info>

<section xml:id="build_hacking.overview.basic">
<info><title>General Process</title></info>

<para>
  The configure process begins the act of building libstdc++, and is
  started via:
</para>

<screen>
<computeroutput>
configure
</computeroutput>
</screen>

<para>
The <filename>configure</filename> file is a script generated (via
<command>autoconf</command>) from the file
<filename>configure.ac</filename>.
</para>


<para>
  After the configure process is complete, 
</para>

<screen>
<computeroutput>
make all
</computeroutput>
</screen>

<para>
in the build directory starts the build process. The <literal>all</literal> target comes from the <filename>Makefile</filename> file, which is  generated via <command>configure</command> from the <filename>Makefile.in</filename> file, which is in turn generated (via
<command>automake</command>) from the file
<filename>Makefile.am</filename>.
</para>

</section>


<section xml:id="build_hacking.overview.map"><info><title>What Comes from Where</title></info>


  <figure xml:id="fig.build_hacking.deps">
    <title>Configure and Build File Dependencies</title>
  <mediaobject>
    <imageobject>
      <imagedata align="center" format="PDF" scale="75" fileref="../images/confdeps.pdf"/>
    </imageobject>
    <imageobject>
      <imagedata align="center" format="PNG" scale="100" fileref="../images/confdeps.png"/>
    </imageobject>
    <textobject>
      <phrase>Dependency Graph for Configure and Build Files</phrase>
    </textobject>
  </mediaobject>
  </figure>

  <para>
    Regenerate all generated files by using the command 
    <command>autoreconf</command> at the top level of the libstdc++ source
    directory.
  </para>
</section>

</section> <!-- overview -->


<section xml:id="build_hacking.configure">
<info><title>Configure</title></info>

<section xml:id="build_hacking.configure.scripts"><info><title>Storing Information in non-AC files (like configure.host)</title></info>


  <para>
    Until that glorious day when we can use <literal>AC_TRY_LINK</literal>
    with a cross-compiler, we have to hardcode the results of what the tests
    would have shown if they could be run.  So we have an inflexible
    mess like <filename>crossconfig.m4</filename>.
  </para>

  <para>
    Wouldn't it be nice if we could store that information in files
    like configure.host, which can be modified without needing to
    regenerate anything, and can even be tweaked without really
    knowing how the configury all works?  Perhaps break the pieces of
    <filename>crossconfig.m4</filename> out and place them in their appropriate
    <filename class="directory">config/{cpu,os}</filename> directory.
  </para>

  <para>
    Alas, writing macros like
    "<code>AC_DEFINE(HAVE_A_NICE_DAY)</code>" can only be done inside
    files which are passed through autoconf.  Files which are pure
    shell script can be source'd at configure time.  Files which
    contain autoconf macros must be processed with autoconf.  We could
    still try breaking the pieces out into "config/*/cross.m4" bits,
    for instance, but then we would need arguments to aclocal/autoconf
    to properly find them all when generating configure.  I would
    discourage that.
</para>
</section>

<section xml:id="build_hacking.configure.conventions"><info><title>Coding and Commenting Conventions</title></info>


  <para>
    Most comments should use {octothorpes, shibboleths, hash marks,
    pound signs, whatever} rather than "<literal>dnl</literal>".
    Nearly all comments in <filename>configure.ac</filename> should.
    Comments inside macros written in ancillary
    <filename class="extension">.m4</filename> files should.
    About the only comments which should <emphasis>not</emphasis>
    use <literal>#</literal>, but use <literal>dnl</literal> instead,
    are comments <emphasis>outside</emphasis> our own macros in the ancillary
    files.  The difference is that <literal>#</literal> comments show up in
    <filename>configure</filename> (which is most helpful for debugging),
    while <literal>dnl</literal>'d lines just vanish.  Since the macros
    in ancillary files generate code which appears in odd places,
    their "outside" comments tend to not be useful while reading
    <filename>configure</filename>.
  </para>

  <para>
    Do not use any <code>$target*</code> variables, such as
    <varname>$target_alias</varname>.  The single exception is in
    <filename>configure.ac</filename>, for automake+dejagnu's sake.
  </para>
</section>

<section xml:id="build_hacking.configure.acinclude"><info><title>The acinclude.m4 layout</title></info>

  <para>
    The nice thing about
    <filename>acinclude.m4</filename>/<filename>aclocal.m4</filename>
    is that macros aren't
    actually performed/called/expanded/whatever here, just loaded.  So
    we can arrange the contents however we like.  As of this writing,
    <filename>acinclude.m4</filename> is arranged as follows:
  </para>
  <programlisting>
    GLIBCXX_CHECK_HOST
    GLIBCXX_TOPREL_CONFIGURE
    GLIBCXX_CONFIGURE
  </programlisting>
  <para>
    All the major variable "discovery" is done here.
    <varname>CXX</varname>, multilibs,
    etc.
  </para>
  <programlisting>
    fragments included from elsewhere
  </programlisting>
  <para>
    Right now, "fragments" == "the math/linkage bits".
  </para>
<programlisting>
    GLIBCXX_CHECK_COMPILER_FEATURES
    GLIBCXX_CHECK_LINKER_FEATURES
    GLIBCXX_CHECK_WCHAR_T_SUPPORT
</programlisting>
<para>
  Next come extra compiler/linker feature tests.  Wide character
  support was placed here because I couldn't think of another place
  for it.  It will probably get broken apart like the math tests,
  because we're still disabling wchars on systems which could actually
  support them.
</para>
<programlisting>
    GLIBCXX_CHECK_SETRLIMIT_ancilliary
    GLIBCXX_CHECK_SETRLIMIT
    GLIBCXX_CHECK_S_ISREG_OR_S_IFREG
    GLIBCXX_CHECK_POLL
    GLIBCXX_CHECK_WRITEV

    GLIBCXX_CONFIGURE_TESTSUITE
</programlisting>
<para>
  Feature tests which only get used in one place.  Here, things used
  only in the testsuite, plus a couple bits used in the guts of I/O.
</para>
<programlisting>
    GLIBCXX_EXPORT_INCLUDES
    GLIBCXX_EXPORT_FLAGS
    GLIBCXX_EXPORT_INSTALL_INFO
</programlisting>
<para>
  Installation variables, multilibs, working with the rest of the
  compiler.  Many of the critical variables used in the makefiles are
  set here.
</para>
<programlisting>
    GLIBGCC_ENABLE
    GLIBCXX_ENABLE_C99
    GLIBCXX_ENABLE_CHEADERS
    GLIBCXX_ENABLE_CLOCALE
    GLIBCXX_ENABLE_CONCEPT_CHECKS
    GLIBCXX_ENABLE_CSTDIO
    GLIBCXX_ENABLE_CXX_FLAGS
    GLIBCXX_ENABLE_C_MBCHAR
    GLIBCXX_ENABLE_DEBUG
    GLIBCXX_ENABLE_DEBUG_FLAGS
    GLIBCXX_ENABLE_LONG_LONG
    GLIBCXX_ENABLE_PCH
    GLIBCXX_ENABLE_SYMVERS
    GLIBCXX_ENABLE_THREADS
</programlisting>
<para>
  All the features which can be controlled with enable/disable
  configure options.  Note how they're alphabetized now?  Keep them
  like that.  :-)
</para>
<programlisting>
    AC_LC_MESSAGES
    libtool bits
</programlisting>
<para>
  Things which we don't seem to use directly, but just has to be
  present otherwise stuff magically goes wonky.
</para>

</section>

<section xml:id="build_hacking.configure.enable"><info><title><constant>GLIBCXX_ENABLE</constant>, the <literal>--enable</literal> maker</title></info>


  <para>
    All the <literal>GLIBCXX_ENABLE_FOO</literal> macros use a common
    helper, <literal>GLIBCXX_ENABLE</literal>.  (You don't have to use
    it, but it's easy.)  The helper does two things for us:
  </para>

<orderedlist>
 <listitem>
   <para>
     Builds the call to the <literal>AC_ARG_ENABLE</literal> macro, with
     <option>--help</option> text
     properly quoted and aligned.  (Death to changequote!)
   </para>
 </listitem>
 <listitem>
   <para>
     Checks the result against a list of allowed possibilities, and
     signals a fatal error if there's no match.  This means that the
     rest of the <literal>GLIBCXX_ENABLE_FOO</literal> macro doesn't need to test for
     strange arguments, nor do we need to protect against
     empty/whitespace strings with the <code>"x$foo" = "xbar"</code>
     idiom.
   </para>
 </listitem>
</orderedlist>

<para>Doing these things correctly takes some extra autoconf/autom4te code,
   which made our macros nearly illegible.  So all the ugliness is factored
   out into this one helper macro.
</para>

<para>Many of the macros take an argument, passed from when they are expanded
   in configure.ac.  The argument controls the default value of the
   enable/disable switch.  Previously, the arguments themselves had defaults.
   Now they don't, because that's extra complexity with zero gain for us.
</para>

<para>There are three "overloaded signatures".  When reading the descriptions
   below, keep in mind that the brackets are autoconf's quotation characters,
   and that they will be stripped.  Examples of just about everything occur
   in acinclude.m4, if you want to look.
</para>

<programlisting>
    GLIBCXX_ENABLE (FEATURE, DEFAULT, HELP-ARG, HELP-STRING)
    GLIBCXX_ENABLE (FEATURE, DEFAULT, HELP-ARG, HELP-STRING, permit a|b|c)
    GLIBCXX_ENABLE (FEATURE, DEFAULT, HELP-ARG, HELP-STRING, SHELL-CODE-HANDLER)
</programlisting>

<itemizedlist>
 <listitem>
   <para>
     <literal>FEATURE</literal> is the string that follows
     <option>--enable</option>.  The results of the
     test (such as it is) will be in the variable
     <varname>$enable_FEATURE</varname>,
     where <literal>FEATURE</literal> has been squashed.  Example:
     <code>[extra-foo]</code>, controlled by the
     <option>--enable-extra-foo</option>
     option and stored in <varname>$enable_extra_foo</varname>.
   </para>
 </listitem>
 <listitem>
   <para>
     <literal>DEFAULT</literal> is the value to store in
     <varname>$enable_FEATURE</varname> if the user does
     not pass <option>--enable</option>/<option>--disable</option>.
     It should be one of the permitted values passed later.
     Examples: <code>[yes]</code>, or <code>[bar]</code>, or
     <code>[$1]</code> (which passes the argument given to the
     <literal>GLIBCXX_ENABLE_FOO</literal> macro as the default).
   </para>
   <para>
     For cases where we need to probe for particular models of things,
     it is useful to have an undocumented "auto" value here (see
     <literal>GLIBCXX_ENABLE_CLOCALE</literal> for an example).
   </para>
 </listitem>
 <listitem>
   <para>
     <literal>HELP-ARG</literal> is any text to append to the option string
     itself in the <option>--help</option> output.  Examples:
     <code>[]</code> (i.e., an empty string, which appends nothing),
     <code>[=BAR]</code>, which produces <code>--enable-extra-foo=BAR</code>,
     and <code>[@&lt;:@=BAR@:&gt;@]</code>, which produces
     <code>--enable-extra-foo[=BAR]</code>.  See the difference?  See
     what it implies to the user?
   </para>
   <para>
     If you're wondering what that line noise in the last example was,
     that's how you embed autoconf special characters in output text.
     They're called <link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="http://www.gnu.org/software/autoconf/manual/autoconf.html#Quadrigraphs"><emphasis>quadrigraphs</emphasis></link>
     and you should use them whenever necessary.
 </para>
 </listitem>
 <listitem>
   <para><literal>HELP-STRING</literal> is what you think it is.  Do not include the
   "default" text like we used to do; it will be done for you by
   <literal>GLIBCXX_ENABLE</literal>.  By convention, these are not full English
   sentences.  Example: <literal>[turn on extra foo]</literal>
   </para>
 </listitem>
</itemizedlist>

<para>
  With no other arguments, only the standard autoconf patterns are
  allowed: "<option>--{enable,disable}-foo[={yes,no}]</option>" The
  <varname>$enable_FEATURE</varname> variable is guaranteed to equal
  either "<literal>yes</literal>" or "<literal>no</literal>"
  after the macro.  If the user tries to pass something else, an
  explanatory error message will be given, and configure will halt.
</para>

<para>
  The second signature takes a fifth argument, "<code>[permit
  a | b | c | ...]</code>"
  This allows <emphasis>a</emphasis> or <emphasis>b</emphasis> or
  ... after the equals sign in the option, and
  <varname>$enable_FEATURE</varname> is
  guaranteed to equal one of them after the macro.  Note that if you
  want to allow plain <option>--enable</option>/<option>--disable</option>
  with no "<literal>=whatever</literal>", you must
  include "<literal>yes</literal>" and "<literal>no</literal>" in the
  list of permitted values.  Also note that whatever you passed as
  <literal>DEFAULT</literal> must be in the list.  If the
  user tries to pass something not on the list, a semi-explanatory
  error message will be given, and configure will halt.  Example:
  <code>[permit generic|gnu|ieee_1003.1-2001|yes|no|auto]</code>
</para>

<para>
  The third signature takes a fifth argument.  It is arbitrary shell
  code to execute if the user actually passes the enable/disable
  option.  (If the user does not, the default is used.  Duh.)  No
  argument checking at all is done in this signature.  See
  <literal>GLIBCXX_ENABLE_CXX_FLAGS</literal> for an example of handling,
  and an error message.
</para>

</section>

<section xml:id="build_hacking.configure.version"><info><title>Shared Library Versioning</title></info>

<para>
The <filename class="library">libstdc++.so</filename> shared library must
be carefully managed to maintain binary compatible with older versions
of the library. This ensures a new version of the library is still usable by
programs that were linked against an older version.
</para>

<para>
Dependent on the target supporting it, the library uses <link
xmlns:xlink="http://www.w3.org/1999/xlink"
xlink:href="https://www.akkadia.org/drepper/symbol-versioning">ELF
symbol versioning</link> for all exported symbols. The symbol versions
are defined by a <link xmlns:xlink="http://www.w3.org/1999/xlink"
xlink:href="https://sourceware.org/binutils/docs/ld/VERSION.html">linker
script</link> that assigns a version to every symbol.
The set of symbols in each version is fixed when a GCC
release is made, and must not change after that.
</para>

<para> When new symbols are added to the library they must be added
to a new symbol version, which must be created the first time new symbols
are added after a release. Adding a new symbol version involves the
following steps:
</para>

<itemizedlist>
<listitem><para>
Edit <filename>acinclude.m4</filename> to update the "revision" value of
<varname>libtool_VERSION</varname>, e.g. from <literal>6:22:0</literal>
to <literal>6:23:0</literal>, which will cause the shared library to be
built as <filename class="library">libstdc++.so.6.0.23</filename>.
</para>
</listitem>
<listitem><para>
Regenerate the <filename>configure</filename> script by running the
<command>autoreconf</command> tool from the correct version of the Autoconf
package (as dictated by the <link xmlns:xlink="http://www.w3.org/1999/xlink"
xlink:href="https://gcc.gnu.org/install/prerequisites.html">GCC
prerequisites</link>).
</para>
</listitem>
<listitem><para>
Edit the file <filename>config/abi/pre/gnu.ver</filename> to
add a new version node after the last new node. The node name should be
<literal>GLIBCXX_3.4.X</literal> where <literal>X</literal> is the new
revision set in <filename>acinclude.m4</filename>, and the node should
depend on the previous version e.g.
<programlisting>
    GLIBCXX_3.4.23 {

    } GLIBCXX_3.4.22;
</programlisting>
For symbols in the ABI runtime, libsupc++, the symbol version naming uses
<literal>CXXABI_1.3.Y</literal> where <literal>Y</literal> increases
monotonically with each new version. Again, the new node must depend on the
previous version node e.g.
<programlisting>
    CXXABI_1.3.11 {

    } CXXABI_1.3.10;
</programlisting>
</para>
</listitem>
<listitem><para>
In order for the <link linkend="test.run.variations">check-abi</link> test
target to pass the testsuite must be updated to know about the new symbol
version(s). Edit the file <filename>testsuite/util/testsuite_abi.cc</filename>
file to add the new versions to the <varname>known_versions</varname> list,
and update the checks for the latest versions that set the
<varname>latestp</varname> variable).
</para>
</listitem>
<listitem><para>
Add the library (<filename class="library">libstdc++.so.6.0.X</filename>)
and symbols versions
(<literal>GLIBCXX_3.4.X</literal> and <literal>CXXABI_1.3.Y</literal>)
to the <link linkend="abi.versioning.history">History</link> section in
<filename>doc/xml/manual/abi.xml</filename> at the relevant places.
</para>
</listitem>
</itemizedlist>

<para>
Once the new symbol version has been added you can add the names of your new
symbols in the new version node:
<programlisting>
    GLIBCXX_3.4.23 {

      # basic_string&lt;C, T, A&gt;::_Alloc_hider::_Alloc_hider(C*, A&amp;&amp;)
      _ZNSt7__cxx1112basic_stringI[cw]St11char_traitsI[cw]ESaI[cw]EE12_Alloc_hiderC[12]EP[cw]OS3_;

    } GLIBCXX_3.4.22;
</programlisting>
You can either use mangled names, or demangled names inside an
<literal>extern "C++"</literal> block. You might find that the new symbol
matches an existing pattern in an old symbol version (causing the
<literal>check-abi</literal> test target to fail). If that happens then the
existing pattern must be adjusted to be more specific so that it doesn't
match the new symbol.
</para>

<para>
For an example of these steps, including adjusting old patterns to be less
greedy, see <link xmlns:xlink="http://www.w3.org/1999/xlink"
xlink:href="https://gcc.gnu.org/ml/gcc-patches/2016-07/msg01926.html">https://gcc.gnu.org/ml/gcc-patches/2016-07/msg01926.html</link>
and the attached patch.
</para>

<para>
If it wasn't done for the last release, you might also need to regenerate
the <filename>baseline_symbols.txt</filename> file that defines the set
of expected symbols for old symbol versions. A new baseline file can be
generated by running <userinput>make new-abi-baseline</userinput> in the
<filename class="directory"><replaceable>libbuilddir</replaceable>/testsuite</filename>
directory. Be sure to generate the baseline from a clean build using
unmodified sources, or you will incorporate your local changes into the
baseline file.
</para>

</section>
</section> <!-- configure -->

<section xml:id="build_hacking.make"><info><title>Make</title></info>

  <para>
    The build process has to make all of object files needed for
    static or shared libraries, but first it has to generate some
    include files. The general order is as follows:
  </para>

<orderedlist>
 <listitem>
   <para>
     make include files, make pre-compiled headers
   </para>
 </listitem>
 <listitem>
   <para>
     make libsupc++
   </para>
   <para>
     Generates a libtool convenience library,
     <filename>libsupc++convenience</filename> with language-support
     routines. Also generates a freestanding static library,
     <filename>libsupc++.a</filename>.
   </para>
 </listitem>
 <listitem>
   <para>
     make src
   </para>
   <para>
     Generates two convenience libraries, one for C++98 and one for
     C++11, various compatibility files for shared and static
     libraries, and then collects all the generated bits and creates
     the final libstdc++ libraries.
  </para>
<orderedlist>
 <listitem>
   <para>
     make src/c++98
   </para>
   <para>
     Generates a libtool convenience library,
     <filename>libc++98convenience</filename> with language-support
     routines. Uses the <option>-std=gnu++98</option> dialect.
   </para>
 </listitem>
 <listitem>
   <para>
     make src/c++11
   </para>
   <para>
     Generates a libtool convenience library,
     <filename>libc++11convenience</filename> with language-support
     routines. Uses the <option>-std=gnu++11</option> dialect.
   </para>
 </listitem>
 <listitem>
   <para>
     make src
   </para>
   <para>
     Generates needed compatibility objects for shared and static
     libraries. Shared-only code is seggregated at compile-time via
     the macro <literal>_GLIBCXX_SHARED</literal>.
   </para>

   <para>
     Then, collects all the generated convenience libraries, adds in
     any required compatibility objects, and creates the final shared
     and static libraries: <filename>libstdc++.so</filename> and
     <filename>libstdc++.a</filename>.
   </para>

 </listitem>
</orderedlist>
 </listitem>
</orderedlist>

</section> <!-- make -->

</section>