mold
mold copied to clipboard
rui314
Is s390x supported?
I noticed one can't build mold on s390x due to:
[ 33s] make[3]: Entering directory '/home/abuild/rpmbuild/BUILD/mold-1.0.999+git.20220109.b18d4881/out/mimalloc'
[ 33s] [100%] Built target mimalloc-static
[ 33s] make[3]: Leaving directory '/home/abuild/rpmbuild/BUILD/mold-1.0.999+git.20220109.b18d4881/out/mimalloc'
[ 33s] make[2]: Entering directory '/home/abuild/rpmbuild/BUILD/mold-1.0.999+git.20220109.b18d4881/out/mimalloc'
[ 33s] /usr/bin/cmake -E cmake_progress_start /home/abuild/rpmbuild/BUILD/mold-1.0.999+git.20220109.b18d4881/out/mimalloc/CMakeFiles 0
[ 33s] make[2]: Leaving directory '/home/abuild/rpmbuild/BUILD/mold-1.0.999+git.20220109.b18d4881/out/mimalloc'
[ 33s] g++ -DMOLD_VERSION=\"1.0.1\" -DLIBDIR="\"/usr/lib64\"" -Ithird-party/mimalloc/include -O2 -g -m64 -fmessage-length=0 -D_FORTIFY_SOURCE=2 -fstack-protector -funwind-tables -fasynchronous-unwind-tables -Wno-sign-compare -pthread -fPIE -fno-unwind-tables -fno-asynchronous-unwind-tables -std=c++20 -fno-exceptions -c -o out/perf.o perf.cc
[ 36s] g++ -DMOLD_VERSION=\"1.0.1\" -DLIBDIR="\"/usr/lib64\"" -Ithird-party/mimalloc/include -O2 -g -m64 -fmessage-length=0 -D_FORTIFY_SOURCE=2 -fstack-protector -funwind-tables -fasynchronous-unwind-tables -Wno-sign-compare -pthread -fPIE -fno-unwind-tables -fno-asynchronous-unwind-tables -std=c++20 -fno-exceptions -c -o out/elf/arch-i386.o elf/arch-i386.cc
[ 37s] g++ -DMOLD_VERSION=\"1.0.1\" -DLIBDIR="\"/usr/lib64\"" -Ithird-party/mimalloc/include -O2 -g -m64 -fmessage-length=0 -D_FORTIFY_SOURCE=2 -fstack-protector -funwind-tables -fasynchronous-unwind-tables -Wno-sign-compare -pthread -fPIE -fno-unwind-tables -fno-asynchronous-unwind-tables -std=c++20 -fno-exceptions -c -o out/elf/arch-arm64.o elf/arch-arm64.cc
[ 37s] g++ -DMOLD_VERSION=\"1.0.1\" -DLIBDIR="\"/usr/lib64\"" -Ithird-party/mimalloc/include -O2 -g -m64 -fmessage-length=0 -D_FORTIFY_SOURCE=2 -fstack-protector -funwind-tables -fasynchronous-unwind-tables -Wno-sign-compare -pthread -fPIE -fno-unwind-tables -fno-asynchronous-unwind-tables -std=c++20 -fno-exceptions -c -o out/elf/arch-x86-64.o elf/arch-x86-64.cc
[ 38s] g++ -DMOLD_VERSION=\"1.0.1\" -DLIBDIR="\"/usr/lib64\"" -Ithird-party/mimalloc/include -O2 -g -m64 -fmessage-length=0 -D_FORTIFY_SOURCE=2 -fstack-protector -funwind-tables -fasynchronous-unwind-tables -Wno-sign-compare -pthread -fPIE -fno-unwind-tables -fno-asynchronous-unwind-tables -std=c++20 -fno-exceptions -c -o out/elf/gc-sections.o elf/gc-sections.cc
[ 38s] elf/gc-sections.cc:30:19: error: 'tbb::feeder' has not been declared
[ 38s] 30 | tbb::feeder<InputSection<E> *> &feeder, i64 depth) {
[ 38s] | ^~~
[ 38s] elf/gc-sections.cc:30:30: error: expected ',' or '...' before '<' token
[ 38s] 30 | tbb::feeder<InputSection<E> *> &feeder, i64 depth) {
[ 38s] | ^
[ 38s] elf/gc-sections.cc: In function 'void mold::elf::visit(mold::elf::Context<E>&, mold::elf::InputSection<E>*, int)':
[ 38s] elf/gc-sections.cc:47:11: error: 'feeder' was not declared in this scope
[ 38s] 47 | feeder.add(sym->input_section);
[ 38s] | ^~~~~~
[ 38s] elf/gc-sections.cc:64:9: error: 'depth' was not declared in this scope
[ 38s] 64 | if (depth < 3)
[ 38s] | ^~~~~
[ 38s] elf/gc-sections.cc:65:37: error: 'feeder' was not declared in this scope
[ 38s] 65 | visit(ctx, sym.input_section, feeder, depth + 1);
[ 38s] | ^~~~~~
[ 38s] elf/gc-sections.cc:67:7: error: 'feeder' was not declared in this scope
[ 38s] 67 | feeder.add(sym.input_section);
[ 38s] | ^~~~~~
[ 38s] elf/gc-sections.cc: In function 'void mold::elf::mark(mold::elf::Context<E>&, tbb::concurrent_vector<mold::elf::InputSection<E>*>&)':
[ 38s] elf/gc-sections.cc:145:37: error: 'tbb::feeder' has not been declared
[ 38s] 145 | tbb::feeder<InputSection<E> *> &feeder) {
[ 38s] | ^~~
[ 38s] elf/gc-sections.cc:145:48: error: expected ',' or '...' before '<' token
[ 38s] 145 | tbb::feeder<InputSection<E> *> &feeder) {
[ 38s] | ^
[ 38s] elf/gc-sections.cc: In lambda function:
[ 38s] elf/gc-sections.cc:146:22: error: 'feeder' was not declared in this scope
[ 38s] 146 | visit(ctx, isec, feeder, 0);
[ 38s] | ^~~~~~
[ 38s] In file included from elf/gc-sections.cc:8:
[ 38s] /usr/include/tbb/parallel_for_each.h: In instantiation of 'void tbb::internal::parallel_for_each_body_for<Function, Iterator>::operator()(tbb::blocked_range<Index>) const [with Function = mold::elf::mark<mold::elf::ARM64>(mold::elf::Context<mold::elf::ARM64>&, tbb::concurrent_vector<mold::elf::InputSection<mold::elf::ARM64>*, tbb::cache_aligned_allocator<mold::elf::InputSection<mold::elf::ARM64>*> >&)::<lambda(mold::elf::InputSection<mold::elf::ARM64>*, int)>; Iterator = tbb::internal::vector_iterator<tbb::concurrent_vector<mold::elf::InputSection<mold::elf::ARM64>*, tbb::cache_aligned_allocator<mold::elf::InputSection<mold::elf::ARM64>*> >, mold::elf::InputSection<mold::elf::ARM64>*>]':
Currently I have only x86-64 and ARM64 machines, so I'd say s390 isn't on our supported host list. If you can fix this compilation error, I'm happy to apply a patch though.
I ran into this same issue. By the way I'm packaging mold for Fedora at https://copr.fedorainfracloud.org/coprs/ktdreyer/mold/ , are you a Fedora packager who can review this if I submit it to the distro?
I'm not a Fedora packager nor a Fedora user, and I know nothing about its packaging system. So I don't think I'm a right person to review your change.
@ktdreyer, I'm in the process of packaging mold for Fedora. The current progress is being tracked here: https://bugzilla.redhat.com/show_bug.cgi?id=2036468
As for s390x support, I'm not sure it makes a lot of sense to package mold for a platform for which it cannot produce native binaries. That should probably be debated over at the Red Hat Bugzilla rather than here.
@rui314 Can you please describe if you plan to support this target? Note it's a target most commercial distros (RHEL, SLES) do support.
I'm trying to get a support to support s390 and POWER (especially POWER10) from IBM. I'll update this bug once the situation changes.
Great. Do you speak about getting access to a machine you can use for testing?
I just started working on it yesterday. Until then I knew nothing about s390x and zArchitecture, but somehow I managed to self-host mold on a real s390x machine today! So I think this item is done.
Nice work! I've just read the comment at elf/arch-s390.cc and yes, we as SUSE have quite some customers who run safety-critical systems on the platform. Mostly being used by banks and similar types of clients.
Yup. I hope to make support contracts with such users to make this project viable.