Closed A5rocks closed 2 weeks ago
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 99.63%. Comparing base (
3350c11
) to head (f62e81e
). Report is 5 commits behind head on master.
Seems like with these changes CancelScope.__exit__
doesn't need to inline the KI protection decorator anymore, so I pushed a commit on that.
I think this is ready for review/merge. While this applied as a patch doesn't seem to work, I can't seem to reproduce that and I'm sure it's just some other PR that changed things just enough for this subtle locals stuff to work (which can be located probably just using a git bisect
?).
I think you can get 3.13 to pass in CI if you explicitly install cffi==1.17.0rc1
as a dependency: https://github.com/python-cffi/cffi/releases/tag/v1.17.0rc1
Since it's a release candidate pip doesn't automatically target it.
I wrote a PR to get trio-websocket to support trio>0.25 https://github.com/python-trio/trio-websocket/pull/188, so once merged that may fully resolve the issues encountered by the distro maintainers.
Alright, this is now running the whole test suite (rather than the part that doesn't require the test requirements).
Fixes https://github.com/python-trio/trio/issues/3004, also refs https://github.com/python-trio/trio/issues/2885
@njsmith could you check my naive
locals()
->sys._getframe().f_locals
replacement? Unfortunately 3.13 makeslocals()
mutations not mutate actual local variables, as part of PEP 667. I'm especially not sure about the replacement in_run.py
wherelocals()
used to clear something (not sure what).