Ensure we take compiler-provided declaration of posix_memalign

[asl]

Some systems (e.g. those with GNU libc) declare posix_memalign with an exception specifier. Compilers (clang including) might have special handling of this to allow posix_memalign redeclaration with / without exception specifier. As we override posix_memalign for GC purposes we really need to ensure the proper include order to workaround this weirdness.

[fruffy]

Thanks, we can merge this instead of #4595. They are doing similar things.

[asl]

@fruffy Yes, I am checking different approaches and how the problem does manifest.

[fruffy]

Conditionally approving assuming COMPILE_WITH_CLANG: ON is removed.

[asl]

@fruffy How about adding builder that would build p4c using clang on Ubuntu?

[vlstill]

@fruffy How about adding builder that would build p4c using clang on Ubuntu?

Considering the discussion I had with @fruffy regarding the static builds, this should be probably part of nightly builds not PR suite (the risk of breakage would be relatively small I guess, especially breakage just under clang and not on macOS). I think having a clang build in CI would be useful.

[fruffy]

We do already have a Clang nightly build, it is the sanitizer one. It just does not use unity builds.

https://github.com/p4lang/p4c/blob/main/.github/workflows/ci-ubuntu-20-sanitizer-nightly.yml#L21

[asl]

We do already have a Clang nightly build, it is the sanitizer one. It just does not use unity builds.

Ah, great. How does it do notifications?

[fruffy]

We do already have a Clang nightly build, it is the sanitizer one. It just does not use unity builds.

Ah, great. How does it do notifications?

Notifications are kind of strange here: https://docs.github.com/en/actions/monitoring-and-troubleshooting-workflows/notifications-for-workflow-runs

The last user that modified the workflow (me) gets notified of failures.

[asl]

@fruffy Ok, I think this is the least bad approach, still not perfect, but we'd have to live with this as-is. Merging when ready.

[asl]

@fruffy Looks like ptf-ebpf fails due to some network issue?

[fruffy]

@fruffy Looks like ptf-ebpf fails due to some network issue?

Looks like a spurious failure, I restarted the test. It's optional so it should not block merging anyway.

[asl]

Looks like a spurious failure, I restarted the test. It's optional so it should not block merging anyway.

I already restarted it 3 or 4 times...

[fruffy]

Looks like a spurious failure, I restarted the test. It's optional so it should not block merging anyway.

I already restarted it 3 or 4 times...

Ah just saw. Hmm, there is nothing we can really do about it for now. I would ignore it. If the issue persists we can start to look into it. This particular test is not super important.