Fix CI

[mikemhenry]

Working on getting CI to pass, looks like it is do to an import for openforcefield failing, which may be due to this PR

openforcefield/openff-toolkit#803

[jchodera]

@mikemhenry : We should upgrade the CI to GitHub Actions after updating to the new MolSSI Cookiecutter.

I'd reach out to @jaimergp to have him give you pointers on this process the first time if you haven't done either a Cookiecutter upgrade or a GitHub Actions update.

Let's also start tagging issues with their assignees, labels, and eventually milestones/projects. Perhaps you can come up with a labeling scheme that would make some sense?

Let me know if you need any repo or org permissions!

[jaimergp]

This repo was already migrated to GH Actions (see here). There might be some PRs opened that predate that contribution, so they'd need a rebase in that case. The openforcefield migration is surely a cause of failures. Check this for instructions: openforcefield/openff-toolkit#819

[mikemhenry]

Something I didn't quite understand (but I have figured it now when switching over to pytest) is that on our CI we skip tets marked with @skipIf(running_on_github_actions) and @attr('advanced') so while things are passing, we have tests that fail but were not caught by CI. I found these when running the tests locally since I was not on GitHub actions and I just used pytest -v to run the tests. I believe that many of the tests that fail locally likely just need to be updated and have fallen behind as the API has changed.

This also explains why CI takes 20 minutes but when running all tests it took 3 hours!

[jaimergp]

I think those were too resource intensive to be run on Travis (where the limit was 1h), and I just converted those skipIf directly without reconsidering a "promotion".

Or maybe some might need a GPU too. 🤷