source: chapter05/gcc-pass2.xml@ 297d9f78

<?xml version="1.0" encoding="ISO-8859-1"?>
<!DOCTYPE sect1 PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN"
  "http://www.oasis-open.org/docbook/xml/4.5/docbookx.dtd" [
  <!ENTITY % general-entities SYSTEM "../general.ent">
  %general-entities;
]>

<sect1 id="ch-tools-gcc-pass2" role="wrap">
  <?dbhtml filename="gcc-pass2.html"?>

  <sect1info condition="script">
    <productname>gcc-pass2</productname>
    <productnumber>&gcc-version;</productnumber>
    <address>&gcc-url;</address>
  </sect1info>

  <title>GCC-&gcc-version; - Pass 2</title>

  <indexterm zone="ch-tools-gcc-pass2">
    <primary sortas="a-GCC">GCC</primary>
    <secondary>tools, pass 2</secondary>
  </indexterm>

  <sect2 role="package">
    <title/>

    <xi:include xmlns:xi="http://www.w3.org/2001/XInclude"
    href="../chapter06/gcc.xml"
    xpointer="xpointer(/sect1/sect2[1]/para[1])"/>

    <segmentedlist>
      <segtitle>&buildtime;</segtitle>
      <segtitle>&diskspace;</segtitle>

      <seglistitem>
        <seg>&gcc-ch5p2-sbu;</seg>
        <seg>&gcc-ch5p2-du;</seg>
      </seglistitem>
    </segmentedlist>

  </sect2>

  <sect2 role="installation">
    <title>Installation of GCC</title>

    <para>Our first build of GCC has installed a couple of internal system
    headers.  Normally one of them, <filename>limits.h</filename>, will in turn
    include the corresponding system <filename>limits.h</filename> header, in
    this case, <filename>/tools/include/limits.h</filename>. However, at the
    time of the first build of gcc <filename>/tools/include/limits.h</filename>
    did not exist, so the internal header that GCC installed is a partial,
    self-contained file and does not include the extended features of the
    system header. This was adequate for building the temporary libc, but this
    build of GCC now requires the full internal header.  Create a full version
    of the internal header using a command that is identical to what the GCC
    build system does in normal circumstances:</para>

<screen><userinput remap="pre">cat gcc/limitx.h gcc/glimits.h gcc/limity.h &gt; \
  `dirname $($LFS_TGT-gcc -print-libgcc-file-name)`/include-fixed/limits.h</userinput></screen>
<!--
    <para>For x86 machines, the limited number of registers is a bottleneck
    for the system.  Free one up by not using a frame pointer that is not
    needed:</para>

<screen><userinput remap="pre">case `uname -m` in
  i?86) sed -i 's/^T_CFLAGS =$/&amp; -fomit-frame-pointer/' gcc/Makefile.in ;;
esac</userinput></screen>
-->
    <para>Once again, change the location of GCC's default dynamic linker to
    use the one installed in <filename
    class="directory">/tools</filename>.</para>

<screen><userinput remap="pre">for file in gcc/config/{linux,i386/linux{,64}}.h
do
  cp -uv $file{,.orig}
  sed -e 's@/lib\(64\)\?\(32\)\?\(x32\)\?/ld@/tools&amp;@g' \
      -e 's@/usr@/tools@g' $file.orig &gt; $file
  echo '
#undef STANDARD_STARTFILE_PREFIX_1
#undef STANDARD_STARTFILE_PREFIX_2
#define STANDARD_STARTFILE_PREFIX_1 "/tools/lib/"
#define STANDARD_STARTFILE_PREFIX_2 ""' &gt;&gt; $file
  touch $file.orig
done</userinput></screen>

    <para arch="default">If building on x86_64, change the default directory
    name for 64-bit libraries to <quote>lib</quote>:</para>

<screen arch="default"><userinput remap="pre">case $(uname -m) in
  x86_64)
    sed -e '/m64=/s/lib64/lib/' \
        -i.orig gcc/config/i386/t-linux64
  ;;
esac</userinput></screen>

    <para arch="ml_32,ml_x32,ml_all">Change the default directory name for 64-bit
    libraries to <quote>lib</quote>:</para>

<screen arch="ml_32,ml_x32,ml_all"><userinput remap="pre">sed -e '/m64=/s/lib64/lib/'  \
    -e '/m32=/s/m32=.*/m32=..\/lib32$(call if_multiarch,:i386-linux-gnu)/' \
    -i.orig gcc/config/i386/t-linux64</userinput></screen>

    <para arch="default">As in the first build of GCC it requires the GMP,
    MPFR, and MPC packages. Unpack the tarballs and move them into the
    required directory names:</para>

    <para arch="ml_32,ml_x32,ml_all">As in the first build of GCC it requires the GMP,
    ISL, MPFR, and MPC packages. Unpack the tarballs and move them into the
    required directory names:</para>

<screen arch="default"><userinput remap="pre">tar -xf ../mpfr-&mpfr-version;.tar.xz
mv -v mpfr-&mpfr-version; mpfr
tar -xf ../gmp-&gmp-version;.tar.xz
mv -v gmp-&gmp-version; gmp
tar -xf ../mpc-&mpc-version;.tar.gz
mv -v mpc-&mpc-version; mpc</userinput></screen>
<screen arch="ml_32,ml_x32,ml_all"><userinput remap="pre">tar -xf ../mpfr-&mpfr-version;.tar.xz
mv -v mpfr-&mpfr-version; mpfr
tar -xf ../gmp-&gmp-version;.tar.xz
mv -v gmp-&gmp-version; gmp
tar -xf ../mpc-&mpc-version;.tar.gz
mv -v mpc-&mpc-version; mpc
tar -xf ../isl-&isl-version;.tar.xz
mv -v isl-&isl-version; isl</userinput></screen>

    <!-- Following patch might be obsolete with gcc >= 8.2.1 --> 
    <para arch="ml_32,ml_x32,ml_all">Fix an issue with isl-&isl-version;:</para>

<screen arch="ml_32,ml_x32,ml_all"><userinput remap="pre">sed -e "/#include &lt;isl\/schedule_node.h&gt;/ a#include &lt;isl/id.h&gt;\n#include &lt;isl/space.h&gt;" \
    -i gcc/graphite.h</userinput></screen>

    <para>Create a separate build directory again:</para>

<screen><userinput remap="pre">mkdir -v build
cd       build</userinput></screen>

    <para>Before starting to build GCC, remember to unset any environment
    variables that override the default optimization flags.</para>

    <para>Now prepare GCC for compilation:</para>

<screen arch="default"><userinput remap="configure">CC=$LFS_TGT-gcc                                    \
CXX=$LFS_TGT-g++                                   \
AR=$LFS_TGT-ar                                     \
RANLIB=$LFS_TGT-ranlib                             \
../configure                                       \
    --prefix=/tools                                \
    --with-local-prefix=/tools                     \
    --with-native-system-header-dir=/tools/include \
    --enable-languages=c,c++                       \
    --disable-libstdcxx-pch                        \
    --disable-multilib                             \
    --disable-bootstrap                            \
    --disable-libgomp</userinput></screen>

<screen arch="ml_32,ml_x32,ml_all"><userinput remap="configure">mlist="m64"</userinput>
<userinput arch="ml_32,ml_all" remap="configure">mlist="$mlist,m32"</userinput>
<userinput arch="ml_x32,ml_all" remap="configure">mlist="$mlist,mx32"</userinput>
<userinput remap="configure">CC=$LFS_TGT-gcc                                    \
CXX=$LFS_TGT-g++                                   \
AR=$LFS_TGT-ar                                     \
RANLIB=$LFS_TGT-ranlib                             \
../configure                                       \
    --prefix=/tools                                \
    --with-local-prefix=/tools                     \
    --with-native-system-header-dir=/tools/include \
    --enable-languages=c,c++                       \
    --disable-libstdcxx-pch                        \
    --enable-multilib                              \
    --with-multilib-list=$mlist                    \
    --with-system-zlib                             \
    --disable-bootstrap                            \
    --disable-libgomp</userinput></screen>

    <variablelist>
      <title>The meaning of the new configure options:</title>

      <varlistentry>
        <term><parameter>--enable-languages=c,c++</parameter></term>
        <listitem>
          <para>This option ensures that both the C and C++ compilers are
          built.</para>
        </listitem>
      </varlistentry>

      <varlistentry>
        <term><parameter>--disable-libstdcxx-pch</parameter></term>
        <listitem>
          <para>Do not build the pre-compiled header (PCH) for
          <filename class="libraryfile">libstdc++</filename>. It takes up a
          lot of space, and we have no use for it.</para>
        </listitem>
      </varlistentry>

      <varlistentry>
        <term><parameter>--disable-bootstrap</parameter></term>
        <listitem>
          <para>For native builds of GCC, the default is to do a "bootstrap"
          build. This does not just compile GCC, but compiles it several times.
          It uses the programs compiled in a first round to compile itself a
          second time, and then again a third time.  The second and third
          iterations are compared to make sure it can reproduce itself
          flawlessly. This also implies that it was compiled correctly.
          However, the LFS build method should provide a solid compiler
          without the need to bootstrap each time.</para>
        </listitem>
      </varlistentry>

    </variablelist>

    <para>Compile the package:</para>

<screen><userinput remap="make">make</userinput></screen>

    <para>Install the package:</para>

<screen><userinput remap="install">make install</userinput></screen>

    <para>As a finishing touch, create a symlink. Many programs and scripts
    run <command>cc</command> instead of <command>gcc</command>, which is
    used to keep programs generic and therefore usable on all kinds of UNIX
    systems where the GNU C compiler is not always installed. Running
    <command>cc</command> leaves the system administrator free to decide
    which C compiler to install:</para>

<screen><userinput remap="install">ln -sv gcc /tools/bin/cc</userinput></screen>

  <caution>
    <para>At this point, it is imperative to stop and ensure that the basic
    functions (compiling and linking) of the new toolchain are working as
    expected. To perform a sanity check, run the following commands:</para>

<screen><userinput>echo 'int main(){}' &gt; dummy.c
cc dummy.c
readelf -l a.out | grep ': /tools'</userinput></screen>

    <para>If everything is working correctly, there should be no errors,
    and the output of the last command will be of the form:</para>

<screen><computeroutput>[Requesting program interpreter: /tools/lib64/ld-linux-x86-64.so.2]</computeroutput></screen>


    <para arch="ml_32,ml_all">Repeat the test for 32-bit:</para>

<screen arch="ml_32,ml_all"><userinput>cc -m32 dummy.c
readelf -l a.out | grep ': /tools'</userinput></screen>

    <para arch="ml_32,ml_all">In this case, the output shoud be:</para>

<screen arch="ml_32,ml_all"><computeroutput>[Requesting program interpreter: /tools/lib/ld-linux.so.2]</computeroutput></screen>

    <para arch="ml_x32,ml_all">Repeat the test for x32-bit:</para>

<screen arch="ml_x32,ml_all"><userinput>cc -mx32 dummy.c
readelf -l a.out | grep ': /tools'</userinput></screen>

    <para arch="ml_x32,ml_all">In this case, the output shoud be:</para>

<screen arch="ml_x32,ml_all"><computeroutput>[Requesting program interpreter: /tools/libx32/ld-linux-x32.so.2]</computeroutput></screen>

    <para arch="default">Note that the dynamic linker will be
    /tools/lib/ld-linux.so.2 for 32-bit machines.</para>

    <para>If the output is not shown as above or there was no output at all,
    then something is wrong. Investigate and retrace the steps to find out
    where the problem is and correct it. This issue must be resolved before
    continuing on. First, perform the sanity check again, using
    <command>gcc</command> instead of <command>cc</command>. If this works,
    then the <filename class="symlink">/tools/bin/cc</filename> symlink is
    missing. Install the symlink as per above.
    Next, ensure that the <envar>PATH</envar> is correct. This
    can be checked by running <command>echo $PATH</command> and verifying that
    <filename class="directory">/tools/bin</filename> is at the head of the
    list. If the <envar>PATH</envar> is wrong it could mean that you are not
    logged in as user <systemitem class="username">lfs</systemitem> or that
    something went wrong back in <xref linkend="ch-tools-settingenviron"
    role="."/></para>

    <para>Once all is well, clean up the test files:</para>

<screen><userinput>rm -v dummy.c a.out</userinput></screen>

  </caution>

  </sect2>

  <sect2 role="content">
    <title/>

    <para>Details on this package are located in
    <xref linkend="contents-gcc" role="."/></para>

  </sect2>

</sect1>