commit:     105fbc76f5dc6eba9ca69e8c6365a961e6a7a9ee
Author:     Alfred Wingate <parona <AT> protonmail <DOT> com>
AuthorDate: Fri Jan 16 21:57:17 2026 +0000
Commit:     Sam James <sam <AT> gentoo <DOT> org>
CommitDate: Sat Jan 17 14:01:52 2026 +0000
URL:        https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=105fbc76

sys-libs/libstdc++-v3: workaround gcc-15 issues

Bug: https://bugs.gentoo.org/944234
Bug: https://bugs.gentoo.org/940229
Signed-off-by: Alfred Wingate <parona <AT> protonmail.com>
Part-of: https://github.com/gentoo/gentoo/pull/45408
Closes: https://github.com/gentoo/gentoo/pull/45408
Signed-off-by: Sam James <sam <AT> gentoo.org>

 sys-libs/libstdc++-v3/libstdc++-v3-3.3.6-r5.ebuild | 206 +++++++++++++++++++++
 1 file changed, 206 insertions(+)

diff --git a/sys-libs/libstdc++-v3/libstdc++-v3-3.3.6-r5.ebuild 
b/sys-libs/libstdc++-v3/libstdc++-v3-3.3.6-r5.ebuild
new file mode 100644
index 000000000000..ff65a54f9620
--- /dev/null
+++ b/sys-libs/libstdc++-v3/libstdc++-v3-3.3.6-r5.ebuild
@@ -0,0 +1,206 @@
+# Copyright 1999-2026 Gentoo Authors
+# Distributed under the terms of the GNU General Public License v2
+
+EAPI=8
+
+inherit flag-o-matic libtool multilib toolchain-funcs
+
+PATCH_VER="4"
+
+DESCRIPTION="Compatibility package for binaries linked against a pre gcc 3.4 
libstdc++"
+HOMEPAGE="https://gcc.gnu.org/libstdc++/";
+SRC_URI="
+       https://gcc.gnu.org/pub/gcc/releases/gcc-${PV}/gcc-${PV}.tar.bz2
+       
https://dev.gentoo.org/~sam/distfiles/gcc-${PV}-patches-${PATCH_VER}.tar.bz2
+"
+
+LICENSE="GPL-2 LGPL-2.1"
+SLOT="5"
+KEYWORDS="~amd64 ~x86"
+IUSE="multilib nls"
+
+RDEPEND="virtual/zlib:="
+DEPEND="
+       ${RDEPEND}
+       app-alternatives/yacc"
+
+transform_known_flags() {
+       declare setting
+
+       # and on x86, we just need to filter the 3.4 specific amd64 -marchs
+       replace-cpu-flags k8 athlon64 opteron x86-64
+
+       # gcc 3.3 doesn't support -march=pentium-m
+       replace-cpu-flags pentium-m pentium3m pentium3
+
+       #GCC 3.3 does not understand G3, G4, G5 on ppc
+       replace-cpu-flags G3 750
+       replace-cpu-flags G4 7400
+       replace-cpu-flags G5 7400
+
+       filter-flags -fdiagnostics-show-option
+}
+
+is_arch_allowed() {
+       i386_processor_table="i386 i486 i586 pentium pentium-mmx winchip-c6 \
+               winchip2 c3 i686 pentiumpro pentium2 pentium3 pentium4 prescott 
\
+               nocona k6 k6-2 k6-3 athlon athlon-tbird x86-64 athlon-4 
athlon-xp \
+               athlon-mp"
+
+       for proc in ${i386_processor_table} ; do
+               [[ "${proc}" == "${1}" ]] && return 0
+       done
+
+       mips_processor_table="mips1 mips2 mips3 mips4 mips32 mips64 r3000 r2000 
\
+               r3900 r6000 r4000 vr4100 vr4111 vr4120 vr4300 r4400 r4600 orion 
\
+               r4650 r8000 vr5000 vr5400 vr5500 4kc 4kp 5kc 20kc sr71000 sb1"
+
+       for proc in ${mips_processor_table} ; do
+               [[ "${proc}" == "${1}" ]] && return 0
+       done
+
+       rs6000_processor_table="common power power2 power3 power4 powerpc \
+               powerpc64 rios rios1 rsc rsc1 rios2 rs64a 401 403 405 505 601 
602 \
+               603 603e ec603e 604 604e 620 630 740 750 7400 7450 8540 801 821 
823 \
+               860"
+
+       for proc in ${rs6000_processor_table} ; do
+               [[ "${proc}" == "${1}" ]] && return 0
+       done
+
+       return 1
+}
+
+do_filter_flags() {
+       declare setting newflags
+
+       # In general gcc does not like optimization, and add -O1 where
+       # it is safe.  This is especially true for gcc 3.3 + 3.4
+       # Compiler crash with -O2, bug #940229
+       if is-flagq -O?; then
+               newflags+=" -O1"
+       fi
+
+       # gcc 3.3 doesn't support -mtune on numerous archs, so xgcc will fail
+       setting="`get-flag mtune`"
+       [[ ! -z "${setting}" ]] && filter-flags -mtune="${setting}"
+
+       # in gcc 3.3 there is a bug on ppc64 where if -mcpu is used
+       # the compiler incorrectly assumes the code you are about to build
+       # is 32 bit
+       use ppc64 && setting="`get-flag mcpu`"
+       [[ ! -z "${setting}" ]] && filter-flags -mcpu="${setting}"
+
+       # only allow the flags that we -know- are supported
+       transform_known_flags
+       setting="`get-flag march`"
+       if [[ ! -z "${setting}" ]] ; then
+               is_arch_allowed "${setting}" && newflags+=" -march=${setting}"
+       fi
+       setting="`get-flag mcpu`"
+       if [[ ! -z "${setting}" ]] ; then
+               is_arch_allowed "${setting}" && newflags+=" -mcpu=${setting}"
+       fi
+
+       # There is lots of brittle old code that violates the aliasing rules. 
GCC
+       # 3.3 supports disabling this optimization.
+       newflags+=" -fno-strict-aliasing"
+
+       # Force older c standard due to incompatibilities, bug #944234
+       newflags+=" -std=gnu99"
+
+       # xgcc wont understand gcc 3.4 flags... in fact it won't understand most
+       # things or have most patches, regardless of what the real GCC 
understands.
+       # A random collection of bugs:
+       # #269433 #290202 #442784 #610064 #879775 #919184 #832016
+       #
+       # There's some extensive discussion at bug #923112, ultimately the only
+       # practical approach is to simply reject *all* flags unless we 
handpicked
+       # them to allow them. Check in "${S}"/gcc/doc/gcc.1 before proceeding.
+       export CFLAGS="${newflags}"
+       export CXXFLAGS="${newflags}"
+       unset LDFLAGS
+}
+
+S=${WORKDIR}/gcc-${PV}
+
+src_prepare() {
+       eapply "${WORKDIR}"/patch/*.patch
+
+       default
+
+       elibtoolize --portage --shallow
+       ./contrib/gcc_update --touch
+
+       if use multilib && [[ ${SYMLINK_LIB} == "yes" ]] ; then
+               # ugh, this shit has to match the way we've hacked gcc else
+               # the build falls apart #259215
+               sed -i \
+                       -e 's:\(MULTILIB_OSDIRNAMES = \).*:\1../lib64 
../lib32:' \
+                       "${S}"/gcc/config/i386/t-linux64 \
+                       || die "sed failed!"
+       fi
+
+       tc-export AR CC RANLIB NM
+
+       # newer versions of GCC add default werrors that we need to disable for
+       # this very old and brittle code. But adding it to CFLAGS doesn't work,
+       # since GCC creates xgcc and uses that to compile libstdc++, and the
+       # ancient xgcc doesn't understand the flags we need.
+       mkdir "${T}/conservative-compiler" || die
+       export PATH="${$}/conservative-compiler:${PATH}"
+
+       local realcc=$(type -P "${CC}") || die
+       export CC="${T}/conservative-compiler/${CC##*/}"
+       cat > "${CC}" <<- __EOF__ || die
+               #!/bin/sh
+               "${realcc}" -Wno-error=implicit-function-declaration 
-Wno-error=implicit-int -Wno-error=int-conversion 
-Wno-error=incompatible-pointer-types "\$@"
+       __EOF__
+
+       chmod +x "${CC}" || die
+}
+
+src_configure() {
+       mkdir -p "${WORKDIR}"/build
+       cd "${WORKDIR}"/build
+       do_filter_flags
+       ECONF_SOURCE=${S} \
+       econf \
+               --enable-shared \
+               --with-system-zlib \
+               --enable-languages=c++ \
+               --enable-threads=posix \
+               --enable-long-long \
+               --disable-checking \
+               --enable-cstdio=stdio \
+               --enable-__cxa_atexit \
+               $(use_enable multilib) \
+               $(use_enable nls) \
+               $(use_with !nls included-gettext)
+
+       touch "${S}"/gcc/c-gperf.h
+}
+
+src_compile() {
+       emake \
+               -C "${WORKDIR}"/build all-target-libstdc++-v3 \
+               AR="$(tc-getAR)" \
+               NM="$(tc-getNM)"
+}
+
+src_install() {
+       emake -j1 \
+               -C "${WORKDIR}"/build \
+               AR="$(tc-getAR)" \
+               NM="$(tc-getNM)" \
+               DESTDIR="${D}" \
+               install-target-libstdc++-v3
+
+       # scrub everything but the library we care about
+       pushd "${D}" >/dev/null
+       mv usr/lib* . || die
+       rm -rf usr
+       rm -f lib*/*.{a,la,so} || die
+       dodir /usr
+       mv lib* usr/ || die
+}

Reply via email to