Commit graph

792 commits

Author SHA1 Message Date
Nikita Popov
fb09e28335 Move llvm-config handling 2024-09-03 08:15:00 +02:00
Nikita Popov
b99883cb6f Fix binary symlinks 2024-09-03 08:14:54 +02:00
Nikita Popov
e481d54052 Fix name of macros file 2024-09-03 08:14:48 +02:00
Nikita Popov
1fc607d3b7 Fix inverted resource dir condition 2024-09-03 08:14:43 +02:00
Nikita Popov
a9bfa39083 Avoid duplicate llvm-config files 2024-09-03 08:14:34 +02:00
Nikita Popov
f18108da40 Fix manpages 2024-09-03 08:14:29 +02:00
Nikita Popov
8fe6e008ba Fix prep 2024-09-03 08:14:23 +02:00
Nikita Popov
1434c4a0c2 Fix some paths 2024-09-03 08:14:12 +02:00
Nikita Popov
1ad15a1548 Don't try to create versioned symlinks multiple times 2024-09-03 08:14:05 +02:00
Nikita Popov
593bda8820 Fix build_libddir 2024-09-03 08:13:55 +02:00
Nikita Popov
8e0f66c2d2 Misc fixes and cleanup 2024-09-03 08:13:44 +02:00
Nikita Popov
318d6afcaa Add libomp compat package support
This is not part of current rawhide sources, but based on libomp17
instead.
2024-09-03 08:13:36 +02:00
Nikita Popov
5868bd75d6 Port compiler-rt changes
From: f0ee285ff1
2024-09-03 08:13:29 +02:00
Nikita Popov
52ddcd59f5 Add versioned macros
From: 1382930328
2024-09-03 08:13:23 +02:00
Nikita Popov
ba868932ac Port clang changes
From: bec3936967
2024-09-03 08:13:15 +02:00
Nikita Popov
c3ed99b303 Port llvm changes
From: d5ddbea65f
2024-09-03 08:13:08 +02:00
Nikita Popov
360dd3bc68 Remove _smp_mflags override on ppc64le
Using a fixed -j argument is pretty dubious and may have the
opposite of the desired effect. E.g. on a recent copr build I
see ppc64le having 5 cpus, so this actually increases the level
of parallelism (and oversubscribes...) rather than decreasing.

I don't think we have evidence that this particular workaround is
still needed, so drop it.
2024-09-03 08:13:03 +02:00
Nikita Popov
1e0387cf85 Disable python-lit and lldb for compat builds 2024-09-03 08:12:54 +02:00
Timm Bäder
b6b17b013e Update to 19.1.0-rc3 2024-09-03 08:00:21 +02:00
Timm Bäder
ac897c6595 Revert "llvm-omp-device-info -> llvm-offload-device-info"
This reverts commit 62d7c09c1c.
2024-09-02 15:45:30 +02:00
Timm Bäder
8d1fbf35a9 Revert "Add llvm-ctxprof-util to file list"
This reverts commit 26a4e937ba.
2024-09-02 15:45:30 +02:00
Nikita Popov
00f2957d29 Add llvm-ctxprof-util to file list
Added in https://github.com/llvm/llvm-project/pull/100379.
2024-09-02 15:45:30 +02:00
Nikita Popov
e7a1f94f29 Don't explicitly set LIBOMP_HAVE_VERSION_SCRIPT_FLAG
From: ee22a23839
2024-09-02 15:45:30 +02:00
Nikita Popov
ced5c791da Remove FEDORA_LLVM_LIB_SUFFIX handling
From: 6f21bc0e93
2024-09-02 15:45:30 +02:00
Nikita Popov
a576e25924 Also fix lld-devel(major) provide
This is incorrect in rawhide compat rpms, fix it for the future.
2024-09-02 15:45:30 +02:00
Nikita Popov
d77a012eb7 Fix clang(major) provides
From: 46db6738a3
From: ffe2656114
2024-09-02 15:45:30 +02:00
Nikita Popov
88414b3581 Ship hwasan_symbolize on riscv64
From: 19722bfcff
2024-09-02 15:45:30 +02:00
Nikita Popov
760435b3cf Add compiler-rt(major) provide
From: 7d3d8b8866
2024-09-02 15:45:30 +02:00
Nikita Popov
6ce190142d llvm-omp-device-info -> llvm-offload-device-info
Was renamed in https://github.com/llvm/llvm-project/pull/100309.
2024-09-02 15:45:30 +02:00
Nikita Popov
8879ae32d2 Add clang-nvlink-wrapper to file list
Introduced in https://github.com/llvm/llvm-project/pull/96561.

I've added the file to clang-tools-extra, as we also put many
similar binaries like clang-linker-wrapper in there.
2024-09-02 15:45:30 +02:00
Nikita Popov
b2926ef5b4 Move patches and sources into main spec file
This way all the dependencies are listed in the main spec file.
Given how small these are, I don't think there's benefit to
having them split out.
2024-09-02 15:45:30 +02:00
Nikita Popov
2e22c397b6 Add python3-lit to big-merge 2024-09-02 15:45:30 +02:00
Nikita Popov
fd8973dfc0 Update Python 3.6 patch
Port the changes from https://github.com/llvm/llvm-project/pull/89490
to Python 3.6.
2024-09-02 15:45:30 +02:00
Timm Bäder
f6e24c047d Require llvm-libs from clang-libs 2024-09-02 15:45:30 +02:00
Nikita Popov
16b62060ba Restore reduced debuginfo on i686
At least on f40 this resulted in a segfault during gdb-add-index.
2024-09-02 15:45:30 +02:00
Nikita Popov
542735e3c4 Decrease debuginfo verbosity on RHEL 8
To see whether this fixes the OOMs. This also drops the change on
i686 for other OSs, to see whether that works or not.

On s390x this was already not effective due to a mismatch with
the condition in the cmake invocation.
2024-09-02 15:45:30 +02:00
Nikita Popov
1612a9fa8f Explicitly list lld libraries
Otherwise we end up shipping liblldb.so and liblldbIntelFeatures.so
as part of lld-libs/lld-devel.
2024-09-02 15:45:30 +02:00
Nikita Popov
712adc2eb0 Patch back in Python 3.6 support on RHEL 8 2024-09-02 15:45:30 +02:00
Nikita Popov
7d226f4859 Fix py_byte_compile invocation 2024-09-02 15:45:30 +02:00
Nikita Popov
690b69906a Set LLVM_RAM_PER_COMPILE_JOB on RHEL 8
RHEL 8 uses gcc instead of clang as the host compiler, which
uses more memory.
2024-09-02 15:45:30 +02:00
Nikita Popov
61f67fb135 Use Python3_EXECUTABLE instead of PYTHON_EXECUTABLE
Also drop the unnecessary python version variables.

If there are multiple python versions installed, the wrong one may be
picked. PYTHON_EXECUTABLE tries to avoid that problem, but the correct
way to spell this nowadays is Python3_EXECUTABLE.
2024-09-02 15:45:30 +02:00
Nikita Popov
724af11e09 Undefine __cmake_in_source_build on RHEL 8
So we get the same behavior as in newer versions.
2024-09-02 15:45:30 +02:00
Nikita Popov
961276e5f3 Disable marshalparser on RHEL 8 2024-09-02 15:45:30 +02:00
Nikita Popov
4cca104ede Run check-clang-tools
These were lost during the big-merge migration, because they
have their own check target.

I locally confirmed that the XFAIL for i686 is still needed.
2024-09-02 15:45:30 +02:00
Nikita Popov
f631698c03 Filter out omp_collapse tests on ppc64le, s390x and i686
These tests also fail on i686: https://download.copr.fedorainfracloud.org/results/@fedora-llvm-team/llvm-snapshots-big-merge-20240717/fedora-39-i386/07740274-llvm/builder-live.log.gz

And yesterday they failed on ppc64le: https://download.copr.fedorainfracloud.org/results/@fedora-llvm-team/llvm-snapshots-big-merge-20240717/fedora-39-ppc64le/07740275-llvm/builder-live.log.gz

While the day before they succeeded on ppc64le.

As such, ignore them instead of marking them as xfail.
2024-09-02 15:45:30 +02:00
Konrad Kleine
6ec634c521 Remove conditions that tried to support LLVM 18 in big-merge 2024-09-02 15:45:30 +02:00
Konrad Kleine
b6c196c8a3 Remove not needed condition in version.spec.inc 2024-09-02 15:45:30 +02:00
Nikita Popov
3aac563374 Don't use wildcard for lld binaries
Otherwise we'll also ship lldb. Instead explicitly list lld and
lld-link.
2024-09-02 15:45:30 +02:00
Nikita Popov
f553e4d7d0 Don't xfail omp_collapse tests on ppc64le
The tests pass: https://download.copr.fedorainfracloud.org/results/@fedora-llvm-team/llvm-snapshots-big-merge-20240716/fedora-rawhide-ppc64le/07737001-llvm/builder-live.log.gz

The tests do fail on s390x. Note that the thread_state_{1,2}.c tests
are already part of the ppc64le exclusive xfail list above, so
we can make this one specific to s390x.
2024-09-02 15:45:30 +02:00
Nikita Popov
9552a8f6c7 Actually disable lldb tests
Make sure to use %% for macros inside comments.
2024-09-02 15:45:30 +02:00