Is this still true with the new runtime?

In D146471#4208808, @dvyukov wrote:

Remove weak, they are from headers now

Should 'weak' be in interface headers?
All user definitions of the callbacks must be non-weak to override our weak definitions, and the interface header is for users.

Remove weak, they are from headers now

what do you think? (as a follow-up)

In D145799#4187988, @vitalybuka wrote:

This one is flaky on power pc
https://lab.llvm.org/buildbot/#/builders/19
https://lab.llvm.org/buildbot/#/builders/18

The code looks sane to me. But I never looked at this part of the code before.
Wonder if this also fixes debug info?

kubamracek added a reviewer: dvyukov.

Nice!

TSan pass has these checks, probably can just copy-paste these:

We also discussed that msan/asan shadow accesses need to be ignored.
MSan accesses can lead to false positives, ASan just to unnecessary checking.
It can also make sense to exclude accesses to global const objects.

Nice!

also don't run on OpenBSD

run metadata tests only if COMPILER_RT_BUILD_SANITIZERS=ON

In D106948#4062454, @protze.joachim wrote:

I'm posting here, because this change broke LLVM tests on a machine only supporting SSE3.

The feature check for COMPILER_RT_HAS_MSSE4_2_FLAG does not really check, whether the current architecture supports SSE 4.2, but rather whether the compiler can target such architecture. Would it be reasonable to change the feature check to something like the following (assuming LLVM_TARGET_ARCH=host)?

clang -march=native -dM -E - < /dev/null | grep __SSE4_2__

otherwise, LLVM_TARGET_ARCH could be used for -march. This would avoid to generate illegal instructions for the target architecture.

Rubber stamp if mapping tests pass.

PTAL

addressed the comments

llvm::isAllocaPromotable

I don't have cmake readability, but looks reasonable.
Better to merge after the fix for metadata tests.

FTR 2 breakage fixes merged:
https://github.com/llvm/llvm-project/commit/2a05bd212e3e8aaed818ee23464f4d1fe0b0596d
https://github.com/llvm/llvm-project/commit/08742e72a34e835e6fc3c696eabe6045c78d6289

This is already fixed: https://github.com/llvm/llvm-project/commit/2a05bd212e3e8aaed818ee23464f4d1fe0b0596d

In D136078#3970792, @thakis wrote:

This seems to break tests: http://45.33.8.238/linux/93224/step_12.txt

Can you take a look?

Fix for:

In D136078#3970792, @thakis wrote:

This seems to break tests: http://45.33.8.238/linux/93224/step_12.txt

Can you take a look?

PTAL

Moved tests to compiler-rt and rebased to HEAD.

Re this failure:

@melver Re this failure:

restrict tests to linux

moved *.cpp tests to clang/test/
updated tests that hardcode pipeline passes
fixed metadata update

moved *.cpp tests to clang/test/
updated tests that hardcode pipeline passes
fixed metadata update

In D136078#3959643, @dvyukov wrote:

FTR ./llvm/test/CodeGen/AMDGPU/llc-pipeline.ll also failed as it hardcodes all passes:
https://lab.llvm.org/buildbot/#/builders/123/builds/14397

FTR ./llvm/test/CodeGen/AMDGPU/llc-pipeline.ll also failed as it hardcodes all passes:
https://lab.llvm.org/buildbot/#/builders/123/builds/14397

Also is this return value correct?

fixed return value

Kazu, thanks for taking care of the revert.

fixed the debug build