yakut
yakut copied to clipboard
e2e tests are flaky
End-to-end tests are time-sensitive and so they expect a certain degree of responsiveness from the OS and the Python runtime. The CI environment is not always responsive enough and so tests tend to fail randomly. Obviously, each separate test may seem robust enough, but we have 102 of them run separately per Python version, so at the moment CI jobs are rarely green ever.
We could go the hard way and improve the tests, which I don't want to do unless I must, or we could set up a self-hosted system that will hopefully perform better than the cloud runner.
This is probably linked to: https://github.com/OpenCyphal/pycyphal/issues/286