This is an archive of the discontinued LLVM Phabricator instance.

[libc++] Add XFAIL annotations for older Clangs
ClosedPublic

Authored by ldionne on Nov 3 2021, 8:30 AM.

Details

Summary

Those were missed by https://llvm.org/D111477, which broke the CI.

Diff Detail

Event Timeline

ldionne requested review of this revision.Nov 3 2021, 8:30 AM
ldionne created this revision.
Herald added a project: Restricted Project. · View Herald TranscriptNov 3 2021, 8:30 AM
Herald added a reviewer: Restricted Project. · View Herald Transcript

Since we install the nightly ToT Clang (which is clang-14) on our CI machines, it turns out that this still fails for clang-14. It will fail until the Docker images pick up a ToT Clang that contains D111477, but we don't really have control over when that happens.

That's not a great story for rolling out changes to Clang and libc++ in lockstep, we need to figure out how to do that better. Ideally we'd build Clang ToT in each job before running the tests, but that is going to be prohibitively expensive.

For now, I'm going to mark the test as UNSUPPORTED on clang-14 until the builders pick up something recent enough. Not great but it should get the CI green again.

ldionne updated this revision to Diff 384470.Nov 3 2021, 8:49 AM

Try to address failures on Clang "ToT"

Does the matching here support regex? We could make the tests format agnostic instead? (hmm, maybe the internal stuff in lldb isn't format agnostic, though... )

Does the matching here support regex? We could make the tests format agnostic instead? (hmm, maybe the internal stuff in lldb isn't format agnostic, though... )

Good suggestion.

ldionne updated this revision to Diff 384490.Nov 3 2021, 9:30 AM

Use regexes instead.

dblaikie accepted this revision.Nov 3 2021, 9:38 AM

Awesome! :) Thanks for the fix!

This revision was not accepted when it landed; it landed in state Needs Review.Nov 3 2021, 10:02 AM
This revision was landed with ongoing or failed builds.
This revision was automatically updated to reflect the committed changes.

"This revision was not accepted when it landed; it landed in state Needs Review."

I guess since libc++'s Phab is configured to require review from a libc++ maintainer (Since I did approve it, but that didn't take it out of the "Needs Review" state, I guess?) ? Could that be fixed in some way, since it doesn't seem to match the reality of what's expected here?