This is an archive of the discontinued LLVM Phabricator instance.

[InstCombine][NFC] Rename check run label
ClosedPublic

Authored by piotr on Jul 10 2023, 2:38 AM.

Details

Summary

The existing check run label "DEFAULT" may clash with
the label from switch "DEFAULT", so renaming it for clarity.

Diff Detail

Event Timeline

piotr created this revision.Jul 10 2023, 2:38 AM
Herald added a project: Restricted Project. · View Herald TranscriptJul 10 2023, 2:38 AM
piotr requested review of this revision.Jul 10 2023, 2:38 AM
Herald added a project: Restricted Project. · View Herald TranscriptJul 10 2023, 2:38 AM
piotr added a comment.Jul 10 2023, 2:40 AM

Note: I am aware the "DEFAULT" prefix is currently unused, but was not sure if the intention was for it to be used in the near future - so I am leaving it there for now.

piotr added a reviewer: nikic.Jul 10 2023, 2:40 AM
nikic added a comment.Jul 10 2023, 2:50 AM

The existing check run label "DEFAULT" may clash with
the label from switch "DEFAULT", so renaming it for clarity.

Not sure I understand the clash. The check-prefixes and the variable names shouldn't clash, right?

piotr added inline comments.Jul 10 2023, 3:07 AM
llvm/test/Transforms/InstCombine/unreachable-code.ll
146

Part of this line contains a substring that could potentially be matched with the check run line "DEFAULT:".

nikic accepted this revision.Jul 11 2023, 7:55 AM

LGTM -- The described file check behavior seems very weird to me, but changing the prefix doesn't hurt either.

llvm/test/Transforms/InstCombine/unreachable-code.ll
146

Is that intentional FileCheck behavior rather than a bug? It seems pretty odd to me that a check prefix would get matched inside a placeholder declaration of a different check prefix.

This revision is now accepted and ready to land.Jul 11 2023, 7:55 AM
This revision was automatically updated to reflect the committed changes.
piotr added inline comments.Jul 12 2023, 1:38 AM
llvm/test/Transforms/InstCombine/unreachable-code.ll
146

Not sure - I was surprised to see this failing too. Maybe --match-full-lines would've avoided the failure, although I didn't check.