Skip to content

Conversation

@alexrp
Copy link
Member

@alexrp alexrp commented Sep 30, 2025

No description provided.

@alexrp alexrp force-pushed the libc-test-ci branch 2 times, most recently from 9d0a38c to a6e8578 Compare September 30, 2025 19:59
@alexrp
Copy link
Member Author

alexrp commented Sep 30, 2025

Looks like we might have to just disable malloc-oom. It does kind of make sense that a test like that has the potential to break QEMU. Thoughts @rpkak?

alexrp added 3 commits October 1, 2025 03:25
This causes QEMU to OOM in CI.
1.4G was the highest value I observed for any test case on x86_64-linux. This
change should prevent OOM conditions in CI.
@rpkak
Copy link
Contributor

rpkak commented Oct 1, 2025

The error of this ci failure is the same error which I get locally when running the tests which I disabled with the comment QEMU OOM. When I worked on #25270, I already wondered, why don't get this error when running malloc-oom.c, but get this error when running any of the other test cases using t_memfill.

@alexrp
Copy link
Member Author

alexrp commented Oct 1, 2025

It looks like this adds about 40 minutes to x86_64-linux-release. I'll let @andrewrk decide if he thinks that's acceptable. FWIW, even with this PR, x86_64-linux-release remains faster than x86_64-linux-debug-llvm by a fairly large margin.

@alexrp alexrp requested a review from andrewrk October 1, 2025 11:15
@alexrp alexrp merged commit 4edebf4 into ziglang:master Oct 14, 2025
9 checks passed
@alexrp alexrp deleted the libc-test-ci branch October 14, 2025 19:30
@alexrp
Copy link
Member Author

alexrp commented Oct 15, 2025

Reverting fcfdf99 for now as it's causing timeouts all over the place. I don't think there's anything actually wrong with it; it just appears to have added more load to the x86_64-linux machines than I had expected. I'll reevaluate adding it back after I do some other work to reduce our CI run times.

BratishkaErik added a commit to BratishkaErik/gentoo that referenced this pull request Oct 23, 2025
Upstream changed behavior so that the "max_rss" field
in one line of their build.zig is now required:
ziglang/zig#25402

Drop the command removing it, as it is no longer needed.
"max_rss" errors were downgraded to warnings in:
ziglang/zig#23525

Upstream suggestion:
ziglang/zig#25659

Closes: https://bugs.gentoo.org/964953
Signed-off-by: Eric Joldasov <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants