search

Smithay / calloop

A callback-based Event Loop
MIT License
178 stars 36 forks source link

Investigate non-deterministic timer test failure on macos #84

Closed elinorbgr closed 2 years ago

elinorbgr commented 2 years ago

The current timer implementation on macos randomly fails the CI in a non-deterministic way. It'd be nice if someone with access to a macos computer could investigate it.

i509VCB commented 2 years ago

Interestingly this does not happen on an m1 mac mini, not sure why the CI environment keeps having these issues.

detly commented 2 years ago

Is it the threaded implementation for MacOS?

elinorbgr commented 2 years ago

Currently yes

detly commented 2 years ago

Interestingly this does not happen on an m1 mac mini, not sure why the CI environment keeps having these issues.

It's probably due to speed or number of cores, I had a similar dog of a time trying to figure out the story behind #79 until I deliberately slowed it down by running under Tarpaulin.

detly commented 2 years ago

Do you need to re-enable CI on OS X now that #89 is merged?

elinorbgr commented 2 years ago

I don't think is was ever disabled...?

detly commented 2 years ago

I thought it was removed as part of 2c40aec276ef76d96cfa9af4032d1d962d78b254. There's only a macos-latest entry for the coverage job. Same if I look under the actions.

elinorbgr commented 2 years ago

Ah yeah, I reorganized the CI when changing the coverage system. You think we need more CI runs for macos?

detly commented 2 years ago

Not if we have one, I just couldn't find it in the GH actions.

elinorbgr commented 2 years ago

Not sure I understand your point, coverage jobs run the test suite and fail if the test suite fails, so I suppose all is good...?

detly commented 2 years ago

Yes, exactly, it's all fine.