[PATCH] D52927: [CodeGen] Fix for PR39094.

[Hsiangkai Wang via Phabricator via llvm-commits]

HsiangKai added a comment.

I am sorry that I misunderstood the meaning of the status. I thought that if the status is ‘ACCEPTED’, I could upstream the commit.

I have verified that the commit could fix PR39094. It also passed LLVM and Clang regression tests. I assumed the patch will not broken LLVM. So, I upstreamed the patch.

According to the test case provided in PR39094, dbg.label is put as the first instruction in some bundles. So, it triggered the assertion failed when users use getBundleStart() to get SlotIndex for the queried MI. I am not familiar about instruction bundling. I do not know under what conditions instructions will be bundled together. So, I could not answer your questions clearly.

I understood that it is better to have a test case for the patch. I will send another patch to provide the test case if I could find one afterward. Thanks for your kindly reminding.


Repository:
  rL LLVM

https://reviews.llvm.org/D52927