[pre-commit.ci] pre-commit autoupdate

[pre-commit-ci]

updates:

• github.com/astral-sh/ruff-pre-commit: v0.8.2 → v0.8.3

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.62%. Comparing base (c4c8ce4) to head (1bd004a).
Report is 3 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3155   +/-   ##
=======================================
  Coverage   99.62%   99.62%           
=======================================
  Files         122      122           
  Lines       18405    18407    +2     
  Branches     1226     1226           
=======================================
+ Hits        18336    18338    +2     
  Misses         47       47           
  Partials       22       22           

Files with missing lines	Coverage Δ
src/trio/_tests/test_highlevel_socket.py	100.00% <100.00%> (ø)
src/trio/_tests/test_socket.py	99.84% <100.00%> (+<0.01%)	⬆️

[A5rocks]

I wonder why 3.13 macOS is all of a sudden failing consistently...

[A5rocks]

OK so presumably something here is causing it: https://github.com/actions/runner-images/releases/tag/macos-14-arm64%2F20241211.616. I assume it's a matter of CPython 3.13.0 -> 3.13.1, but it's weird for that to break anything and specifically on macOS.

---

I checked and yes, that does seem to be the change: https://github.com/A5rocks/trio/actions/runs/12366588124/job/34513521105. I wonder what happened -- I assume they must have made error checking somewhere more strict. I'll check the changelog I suppose.

[webknjaz]

@A5rocks I suggest pinning it, temporarily, and debugging in a separate “unpinning” PR.

[jakkdl]

OK so presumably something here is causing it: https://github.com/actions/runner-images/releases/tag/macos-14-arm64%2F20241211.616. I assume it's a matter of CPython 3.13.0 -> 3.13.1, but it's weird for that to break anything and specifically on macOS.

I checked and yes, that does seem to be the change: https://github.com/A5rocks/trio/actions/runs/12366588124/job/34513521105. I wonder what happened -- I assume they must have made error checking somewhere more strict. I'll check the changelog I suppose.

It's not terribly surprising, nor alarming, that this code broke

trio/src/trio/_tests/test_socket.py

Lines 467 to 473 in c4c8ce4

	def setsockopt_tests(sock: SocketType | SocketStream) -> None:
	"""Extract these out, to be reused for SocketStream also."""
	# specifying optlen. Not supported on pypy, and I couldn't find
	# valid calls on darwin or win32.
	if hasattr(tsocket, "SO_BINDTODEVICE"):
	sock.setsockopt(tsocket.SOL_SOCKET, tsocket.SO_BINDTODEVICE, None, 0)

Author of the comments is me in cf1f3c7

but looking deeper SO_BINDTODEVICE should not be set on mac, afaict it's a linux-only option, so it seems it's now set, but not supported. cpython doesn't set the constant themselves as it's a linux kernel header constant that gets passed through, so I'm not even sure it's cpython/3.13.1 that is to blame....

[A5rocks]

so I'm not even sure it's cpython/3.13.1 that is to blame....

Setting the version back to 3.13.0 makes CI pass on same GHA runner images. Maybe it's something about the build environment but still weird.

[jakkdl]

asdfffff. codecov/merge queue is being very irritating. I have no clue why codecov says the diff hit is <100%

[A5rocks]

Let's disable codecov in merge queue then!

[webknjaz]

@A5rocks how?

@jakkdl I think it's because of the default branch, probably.

[A5rocks]

@A5rocks how?

I thought what you thought up on Gitter sounds fine, ie just skipping the codecov step based on the GitHub event in question.