[Bug ld/31824] New: gcc 14 fails during link step on Windows when bootstrap-lto is specified
https://sourceware.org/bugzilla/show_bug.cgi?id=31824
Bug ID: 31824
Summary: gcc 14 fails during link step on Windows when
bootstrap-lto is specified
Product: binutils
Version: unspecified
Status: UNCONFIRMED
Severity: normal
Priority: P2
Component: ld
Assignee: unassigned at sourceware dot org
Reporter: tanksherman27 at gmail dot com
Target Milestone: ---
gcc 14 seems to be unable to link itself when the bootstrap config that is
active is bootstrap-lto, to turn on Link Time Optimization for the gcc being
compiled. The following is the configure command:
# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105507#c3
# At least with mingw32 + dwarf-2 exceptions there can only be one libgcc in
# process, or exceptions will no longer work. Since some of the gcc deps are
# linked dynamically like gmp/zlib/zstd and those pull in libgcc we can't
# allow libgcc to be linked statically. The default is "-static-libstdc++
# -static-libgcc" for both, so we drop "-static-libgcc" here:
# GCC 14 Update: Since we dropped 32bit Ada with GCC 14.1 and GCC doesn't use
# exceptions elsewhere this could in theory be removed, in case it makes
problems.
_extra_config+=(
'--with-boot-ldflags="-static-libstdc++"'
'--with-stage1-ldflags="-static-libstdc++"'
)
# In addition adaint.c does `#include ` which pulls in msxml.h,
hacky hack:
CPPFLAGS+=" -DCOM_NO_WINDOWS_H"
../${_sourcedir}/configure \
--prefix=${MINGW_PREFIX} \
--with-local-prefix=${MINGW_PREFIX}/local \
--with-build-config='bootstrap-lto' \
--build=${MINGW_CHOST} \
--host=${MINGW_CHOST} \
--target=${MINGW_CHOST} \
--with-native-system-header-dir=${MINGW_PREFIX}/include \
--libexecdir=${MINGW_PREFIX}/lib \
--enable-bootstrap \
--enable-checking=release \
--with-arch=${_arch} \
--with-tune=generic \
--enable-languages=${_languages} \
--enable-shared \
--enable-static \
--enable-libatomic \
--enable-threads=${_threads} \
--enable-graphite \
--enable-fully-dynamic-string \
--enable-libstdcxx-threads \
--enable-libstdcxx-filesystem-ts \
--enable-libstdcxx-time \
--disable-libstdcxx-pch \
--enable-lto \
--enable-libgomp \
--disable-libssp \
--disable-multilib \
--disable-rpath \
--disable-win32-registry \
--disable-nls \
--disable-werror \
--disable-symvers \
--with-libiconv \
--with-system-zlib \
--with-{gmp,mpfr,mpc,isl}=${MINGW_PREFIX} \
--with-pkgversion="Rev${pkgrel}, Built by MSYS2 project" \
--with-bugurl="https://github.com/msys2/MINGW-packages/issues"; \
--with-gnu-as \
--with-gnu-ld \
"${_extra_config[@]}"
...and the invocation of make:
# https://gcc.gnu.org/onlinedocs/gccint/Makefile.html
make -O STAGE1_CFLAGS="-O2" \
--jobs profiledbootstrap
This fails with the following error:
C:\msys64\ucrt64\x86_64-w64-mingw32\bin\ld.exe: error: export ordinal too
large: 91917
C:\msys64\ucrt64\x86_64-w64-mingw32\bin\ld.exe: error: export ordinal too
large: 91820
collect2.exe: error: ld returned 1 exit status
make[3]: *** [../../gcc-14.1.0/gcc/lto/Make-lang.in:102: lto-dump.exe] Error 1
make[3]: *** Waiting for unfinished jobs
collect2.exe: error: ld returned 1 exit status
make[3]: *** [../../gcc-14.1.0/gcc/lto/Make-lang.in:96: lto1.exe] Error 1
C:\msys64\ucrt64\x86_64-w64-mingw32\bin\ld.exe: error: export ordinal too
large: 94928
collect2.exe: error: ld returned 1 exit status
make[3]: *** [../../gcc-14.1.0/gcc/c/Make-lang.in:87: cc1.exe] Error 1
C:\msys64\ucrt64\x86_64-w64-mingw32\bin\ld.exe: error: export ordinal too
large: 97453
collect2.exe: error: ld returned 1 exit status
make[3]: *** [../../gcc-14.1.0/gcc/cp/Make-lang.in:145: cc1plus.exe] Error 1
rm gcc.pod
make[2]: *** [Makefile:5228: all-stageprofile-gcc] Error 2
make[1]: *** [Makefile:23910: stageprofile-bubble] Error 2
make: *** [Makefile:24164: profiledbootstrap] Error 2
==> ERROR: A failure occurred in build().
Aborting...
--
You are receiving this mail because:
You are on the CC list for the bug.
PE section extension
Hello, I'd like to extend the ld generated PE section with a custom section called .store in analogy to .idata and I would love a little guidance where to make this change. The new field must be generated after the idata has been generated. In addition I need to know where the dllname name, foo.dll is generated as I need to change this as well to reserve like 300 more chars for later changes of said dllname. I found pe-dll.c but so far didn't find where the dllname is translated into the .idata section of the PE header. Regards, Joachim -- Sent from my Android device with K-9 Mail. Please excuse my brevity.
[Bug binutils/17466] objcopy corrupts Windows executables when adding a section
https://sourceware.org/bugzilla/show_bug.cgi?id=17466
Joachim Schiele changed:
What|Removed |Added
CC||js at lastlog dot de
--- Comment #2 from Joachim Schiele ---
Maybe someone else reads your ticket therefore this is what I did instead:
https://stackoverflow.com/questions/78555828/adding-custom-section-to-pe-with-gcc-mingw
Basically I had the same issue like you but found a workaround:
$ echo 'int main() { puts ("Hello world"); }' | gcc -x c - -c -o main.o
$ echo "this is my special data" >mydata
$ objcopy --add-section .mydata=mydata main.o
$ gcc main.o -o main
$ objdump -h main.exe
The difference is one adds the new section to the .o file instead the exe.
--
You are receiving this mail because:
You are on the CC list for the bug.
[Bug ld/31828] New: Lack of support for HEASLR (high-entropy-va) for PE targets
https://sourceware.org/bugzilla/show_bug.cgi?id=31828 Bug ID: 31828 Summary: Lack of support for HEASLR (high-entropy-va) for PE targets Product: binutils Version: 2.39 Status: UNCONFIRMED Severity: normal Priority: P2 Component: ld Assignee: unassigned at sourceware dot org Reporter: ralf.habacker at freenet dot de Target Milestone: --- Created attachment 15547 --> https://sourceware.org/bugzilla/attachment.cgi?id=15547&action=edit Patch adding high-entropy-va support For bug 19011, patches were added to add hardening options to ld, but these were not fully included in the git repo. The attached patch adds the missing support for HEASLR (high-entropy-va). -- You are receiving this mail because: You are on the CC list for the bug.
[Bug gold/31830] New: error: y.o: multiple definition of 'foo'
https://sourceware.org/bugzilla/show_bug.cgi?id=31830
Bug ID: 31830
Summary: error: y.o: multiple definition of 'foo'
Product: binutils
Version: 2.43 (HEAD)
Status: NEW
Severity: normal
Priority: P2
Component: gold
Assignee: ccoutant at gmail dot com
Reporter: hjl.tools at gmail dot com
CC: ian at airs dot com
Target Milestone: ---
[hjl@gnu-tgl-3 gold-1]$ cat x.c
extern __inline __attribute__((__gnu_inline__)) void foo(void) {}
__attribute__((__symver__("foo@GLIBC_2.2.5")))
int __collector_foo_2_2(void) {}
[hjl@gnu-tgl-3 gold-1]$ cat y.c
void foo(void) {}
[hjl@gnu-tgl-3 gold-1]$ cat x.t
GLIBC_2.2.5 {
global:
foo;
};
[hjl@gnu-tgl-3 gold-1]$ make
gcc -B./ -O2 -fPIC -c -o x.o x.c
gcc -B./ -O2 -fPIC -c -o y.o y.c
./ld --version-script=x.t -shared -o x.so x.o y.o
./ld: error: y.o: multiple definition of 'foo'
./ld: x.o: previous definition here
make: *** [Makefile:13: x.so] Error 1
[hjl@gnu-tgl-3 gold-1]$
--
You are receiving this mail because:
You are on the CC list for the bug.
[Bug binutils/31800] src-release.sh recursively changes permissions of everything in to 0777
https://sourceware.org/bugzilla/show_bug.cgi?id=31800 --- Comment #5 from Rostislav Krasny --- Created attachment 15549 --> https://sourceware.org/bugzilla/attachment.cgi?id=15549&action=edit formatted patch with the fix -- You are receiving this mail because: You are on the CC list for the bug.
[Bug binutils/31800] src-release.sh recursively changes permissions of everything in to 0777
https://sourceware.org/bugzilla/show_bug.cgi?id=31800 --- Comment #6 from Rostislav Krasny --- The 'git reset --hard' works not stable. Sometimes it changes permissions according to umask or to the git configuration, sometimes it doesn't. So I made another fix. I've attached a formatted patch with my fix. Please review and merge it. I tested it on my Fedora 40. -- You are receiving this mail because: You are on the CC list for the bug.
[Bug ld/30281] error: multiple definition of `pwrite@GLIBC_2.2' with LTO and symver attribute
https://sourceware.org/bugzilla/show_bug.cgi?id=30281 --- Comment #18 from H.J. Lu --- (In reply to Andrey Sotnikov from comment #17) > (In reply to H.J. Lu from comment #16) > > You were using gold, not ld. Please pass -fuse-ld=bfd to GCC. > > I am not sure I understand. I used ld. I verified with strace that there > were no other processes spawned. It just seems that ld was gold. This is a different bug. Please try https://sourceware.org/pipermail/binutils/2024-June/134459.html -- You are receiving this mail because: You are on the CC list for the bug.
[Bug ld/30281] error: multiple definition of `pwrite@GLIBC_2.2' with LTO and symver attribute
https://sourceware.org/bugzilla/show_bug.cgi?id=30281 H.J. Lu changed: What|Removed |Added See Also||https://sourceware.org/bugz ||illa/show_bug.cgi?id=31830 -- You are receiving this mail because: You are on the CC list for the bug.
[Bug gold/31830] error: y.o: multiple definition of 'foo'
https://sourceware.org/bugzilla/show_bug.cgi?id=31830 H.J. Lu changed: What|Removed |Added See Also||https://sourceware.org/bugz ||illa/show_bug.cgi?id=30281 -- You are receiving this mail because: You are on the CC list for the bug.
