HomePhabricator

[TSan] Attempt to fix linker error for Linux on AArch64

Description

[TSan] Attempt to fix linker error for Linux on AArch64

Introduced in 5be69ebe121d5b6ea284c3dc6d1cd5098c94f353.

Details

Committed
ylnJun 28 2019, 5:40 PM
Parents
rL364699: AMDGPU/GlobalISel: RegBankSelect for atomic.inc/atomic.dec
Branches
Unknown
Tags
Unknown

Event Timeline

We are still seeing liner errors in the AArch64 build bots (http://lab.llvm.org:8011/builders/clang-cmake-aarch64-full/builds/7487/steps/ninja%20check%202/logs/stdio), are you looking into this? If there's not a quick, obvious fix, it would be better to revert the original commit to get the bots back to green.

yln added a comment.Jul 8 2019, 11:48 AM

We are still seeing liner errors in the AArch64 build bots (http://lab.llvm.org:8011/builders/clang-cmake-aarch64-full/builds/7487/steps/ninja%20check%202/logs/stdio), are you looking into this? If there's not a quick, obvious fix, it would be better to revert the original commit to get the bots back to green.

Thanks for pointing this out to me; and apologies for the churn. I pushed a revert/attempted fix.

I want to fix this instead of a blanket revert, because it allows me to ensure that the code that is going to replace the assembly produces the same result on architectures I cannot test directly.
https://reviews.llvm.org/D64092

InitializeGuardPtr();
InitializeLongjmpXorKey();
CHECK_EQ(longjmp_xor_key, _tsan_pointer_chk_guard);
yln added a comment.Jul 9 2019, 9:48 AM

Yes, it is. I am going to land the linked revision now, which will hopefully fix it.