fix builds.sr.ht failing

[jakkdl]

it was failing on every PR, and been doing so for a while. Configuration failure or necessary to replace it?

EDIT: temporarily paused until someone gains motivation to think about this!

[A5rocks]

It's easy enough to disable but I'm not sure if we want to try doing any deeper analysis first.

[A5rocks]

I'm going to leave this open but disable the integration for now.

[jakkdl]

What was the purpose of the builds.sr.ht job? Does it do anything special not covered by other tests?

[A5rocks]

Fedora and freebsd testing (and maybe another platform), both of which GitHub actions can do but only by literally just using qemu :(

[jakkdl]

Ah. Seems unlikely that fedora would behave much differently than ubuntu, but losing out on freebsd isn't great.

[njsmith]

Originally fedora was added because ubuntu's kernel wasn't new enough to test pidfd, but that might not be an issue anymore.

On Wed, Sep 13, 2023 at 5:04 AM John Litborn @.***> wrote:

Ah. Seems unlikely that fedora would behave much differently than ubuntu, but losing out on freebsd isn't great.

— Reply to this email directly, view it on GitHub https://github.com/python-trio/trio/issues/2792#issuecomment-1717497194, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAEU42FAOVMGI6YCQJ6SY7LX2GONHANCNFSM6AAAAAA4LQLEHM . You are receiving this because you are subscribed to this thread.Message ID: @.***>

-- Nathaniel J. Smith -- https://vorpus.org http://vorpus.org

[A5rocks]

For what it's worth, I've when thinking about builds.sr.ht replacements and seen how e.g. snmalloc uses qemu to run freebsd stuff in github actions: https://github.com/microsoft/snmalloc/blob/6b8f3338c71fb06a3609ef935a8579b7bda7e07e/.github/workflows/main.yml#L117

[jakkdl]

Originally fedora was added because ubuntu's kernel wasn't new enough to test pidfd, but that might not be an issue anymore.

If you click through from a test run you get to https://github.com/actions/runner-images/blob/ubuntu22/20230911.1/images/linux/Ubuntu2204-Readme.md which says

Kernel Version: 6.2.0-1011-azure

and https://docs.python.org/3/library/os.html#os.pidfd_open says availability linux >= 5.3, so it shouldn't be a problem. We still need the funky workaround code in _subprocess until we drop 3.8 though, as pidfd_open is only available on >=3.9. I quickly wrote up #2799 to document this better in the code.

[jakkdl]

For what it's work, I've when thinking about builds.sr.ht replacements seen how e.g. snmalloc uses qemu to run freebsd stuff in github actions: https://github.com/microsoft/snmalloc/blob/6b8f3338c71fb06a3609ef935a8579b7bda7e07e/.github/workflows/main.yml#L117

that doesn't look too nasty, worth giving a try.