This is an archive of the discontinued LLVM Phabricator instance.

Mark trivial_abi unique_ptr tests as unsupported in clang-11.
AbandonedPublic

Authored by zoecarver on Nov 16 2020, 8:51 PM.

Details

Reviewers
ldionne
Group Reviewers
Restricted Project
Summary

It would appear that these tests fail for clang-11. See the CI and Compiler Explorer. This was exposed while testing D66262.

Diff Detail

Event Timeline

zoecarver created this revision.Nov 16 2020, 8:51 PM
Herald added a project: Restricted Project. · View Herald TranscriptNov 16 2020, 8:51 PM
Herald added a reviewer: Restricted Project. · View Herald Transcript
zoecarver requested review of this revision.Nov 16 2020, 8:51 PM
Mordante added inline comments.
libcxx/test/libcxx/memory/trivial_abi/unique_ptr_arg.pass.cpp
15

Can you update this comment in all modified tests?

ldionne requested changes to this revision.Nov 17 2020, 1:08 PM

It seems to me like the most likely issue is just that D66262 breaks these tests -- is that possible?

This revision now requires changes to proceed.Nov 17 2020, 1:08 PM

It seems to me like the most likely issue is just that D66262 breaks these tests -- is that possible?

Because our CI is using Clang 11.0.1, and it hasn't been failing because of this so far AFAICT.

It seems to me like the most likely issue is just that D66262 breaks these tests -- is that possible?

That's what I originally thought (because that's the only patch causing CI failures). But check out the Godbolt link in the description, it also fails there which makes me think maybe it's unrelated.

Is it possible that the difference lies between 11.0.0 (your godbolt link) and 11.0.1 (the CI)?

Is it possible that the difference lies between 11.0.0 (your godbolt link) and 11.0.1 (the CI)?

Godbolt also fails on trunk, it would be odd (but not impossible) for it to be fixed only for 11.0.1 and then broken again for trunk. Also, that patch that's failing is a super small change, it seems very unlikely that it is responsible for the failures.

Is it possible that the difference lies between 11.0.0 (your godbolt link) and 11.0.1 (the CI)?

Godbolt also fails on trunk, it would be odd (but not impossible) for it to be fixed only for 11.0.1 and then broken again for trunk. Also, that patch that's failing is a super small change, it seems very unlikely that it is responsible for the failures.

It doesn't fail on trunk when you actually use libc++! https://godbolt.org/z/af66T1

Clang on Godbolt uses libstdc++ by default (probably because it runs on some Unix).

Is it possible that the difference lies between 11.0.0 (your godbolt link) and 11.0.1 (the CI)?

Godbolt also fails on trunk, it would be odd (but not impossible) for it to be fixed only for 11.0.1 and then broken again for trunk. Also, that patch that's failing is a super small change, it seems very unlikely that it is responsible for the failures.

It doesn't fail on trunk when you actually use libc++! https://godbolt.org/z/af66T1

Clang on Godbolt uses libstdc++ by default (probably because it runs on some Unix).

Ah, you're right! I always forget. So it might be resolved in 11.0.1. I'll pull down the Docker image and do a full investigation :)

zoecarver abandoned this revision.Nov 23 2020, 12:19 PM