I would need to go back over some (ideally merged) pull requests to check, but my strong impression compared to when we ran TravisCI/CircleCI/AppVeyor to now with GitHub Actions/CircleCI/AppVeyor, the unit test coverage reported via CodeCov is much more random.
So, it that really the case?
If so, how can we fix it? One hypothesis is that CodeCov is reporting before all the CI tests have finished? e.g. If it only waits for at least one GitHub Actions report, rather than all the GitHub Actions reports?
Read more here: Source link