make KeyboardInterrupt cancel pytest

[arvidn]

No description provided.

[altendky]

It looks like these may all be available in the public API instead of _pytest. pytest-dev/pytest#7469

Perhaps not applicable after checking above, but I expect isort might prefer these to be all together? If not we may want to adjust a setting for that.

[arvidn]

in the version of pytest I have they are not exposed under pytest. That was my first attempt

[arvidn]

I didn't pay attention whether pre-commit ran isort for me or not. But it should have, right?

[arvidn]

I find I frequently have to explicitly run pre-commit run --all to have checks be run..

[altendky]

Python 3.9.5 (default, Jun  3 2021, 15:18:23) 
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pytest
>>> pytest.__version__
'7.0.1'
>>> from pytest import CallInfo, CollectReport, TestReport
>>>

If you have latest pytest and it isn't working for you I would be curious to know. We could dig into that on keybase.

Suggested change

	from _pytest.runner import CallInfo
	from _pytest.runner import CollectReport
	from _pytest.runner import TestReport
	from pytest import CallInfo, CollectReport, TestReport

---

On the pre-commit, it's default is to only run on 'modified' files. I'm not sure the definition of that. I personally just manually run it with --all.

[altendky]

https://docs.pytest.org/en/6.2.x/reference.html#pytest.hookspec.pytest_exception_interact

It sounds like this would terminate tests early on an uncaught exception in a test? This is really required as opposed to just the sys.exit() removal above?

[arvidn]

the sys.exit() code that I removed did not contribute meaningfully to anything. That exception was also caught by pytest, just like KeyboardInterrupt is. This hook is called later, for tests (or collections) that resulted in an exception. pytest.exit() throws a special exception that pytest recognizes means it should shut down.

Are you suggesting I make this conditional on KeyboardInterrupt specifically?

[arvidn]

actually, I think this is only called for "interactive" exceptions. This is what the docs say.

Called when an exception was raised which can potentially be interactively handled.

[arvidn]

fixed

[altendky]

I don't recall pytest itself having an issue with terminating on ctrl+c. Rather, I had just assumed that the issue was that we were consuming the KeyboardInterrupt ourselves. We had at least one place that we just claimed a signal handler without any cooperative considerations when starting a service. That was changed recently in #10163 with a Windows regression addressed in #10263. Maybe there's something with pytest-asyncio related to this so I'll look there to get a bit more confidence I understand what is going on.

I saw that phrase you quoted about interactively handleable errors. I have no idea what it means. I'll ask around and see what I can figure out.

[github-actions]

This PR has been flagged as stale due to no activity for over 60 days. It will not be automatically closed, but it has been given a stale-pr label and should be manually reviewed by the relevant parties.

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.