Open toomanybrians opened 1 year ago
Do we know when this started happening? Have these always failed?
This only happens with the updated MPF 0.57 which runs on Python 3.10+. That required making some minor changes all throughout MPF for asyncio. I also got rid of the custom YAML processing since the way that worked changed in more recent version of rumael.yaml which were not compatible with the latest versions of python.
So I think the OPP test was always fine, but now with these changes something randomly breaks. Like I said, I just disabled the test since it's not a showstopper, and the tests still run fine on the stable MPF 0.56.x. This is more if just an FYI that something in those tests in funky with the latest versions of python.
We're getting random failures of the tests running on Windows, across all versions of Python (3.8-3.11). It's totally random, but consistent. Different windows tests fail, and if you rerun tests, ones that failed might pass and vice versa. I created the
investigate-github-tests
branch to investigate:https://github.com/missionpinball/mpf/tree/investigate-github-tests
Here are the workflow actions that have run on that branch. You can download the logs and see the details. (I enabled verbose logging for the tests, and also disabled the non-Windows tests.) https://github.com/missionpinball/mpf/actions?query=branch%3Ainvestigate-github-tests
I've narrowed it down to the
test_OPP.TestOPPStm32.testOpp()
test. (This test has been disabled in the dev branch, and the latest on that branch as of this writing.)Here's a log of a test that fails. (Actually, it just hangs, not a failure. The keyboard interrupt is what comes in when the job is canceled, whether canceled via the UI, or canceled automatically after 6 hours by GitHub.) The logs for all tests that fail are identical here, same lines, same hang:
I assume this is something with the serial port stopping hanging up? Dunno if it's an asyncio thing or what?
I'm going to assign this to @cobra18t since you maintain the OPP platform. Feel free to keep using this
investigate-github-tests
branch to figure this out. I don't think this is a showstopper, so take your time. I've disabled this test in dev so it doesn't hold us up.