Fix some usage of deprecated imp module

[asottile]

Summary of changes

Partial resolution of #479 -- the other usage is a little more tricky and will require some more clever fixes

Pull Request Checklist

• Changes have tests
• News fragment added in changelog.d. See documentation for details

[asottile]

not sure why travis didn't report status -- it is complete and successful 🤷‍♂️

[pganssle]

@asottile I don't have any objections to this, but when running the test suite against master with DeprecationWarning set to raise errors on master, I'm not getting any failures, but coverage.py is saying all these lines are covered. Any idea what's going on there?

[asottile]

@asottile I don't have any objections to this, but when running the test suite against master with DeprecationWarning set to raise errors on master, I'm not getting any failures, but coverage.py is saying all these lines are covered. Any idea what's going on there?

depending on the version of pytest you're using (I've only fixed pytest in 5.x) imp will get imported as a side-effect of pytest's initialization (and so imports in user-space code won't trigger that warning) -- this is probably what you're seeing if I had to guess?

[pganssle]

Hm.. That seems like a huge problem with pytest, that's super annoying.

[asottile]

hehe yeah that's why I fixed it 🙃

[asottile]

@pganssle @jaraco any updates on this? I think this is ready for merge

[pganssle]

Sorry to let this languish for so long. I am made mildly uncomfortable by the fact that there's no regression test in this (and thus no automated verification that this fixes the problem), but given the issue with pytest and the fact that we can't really turn on warnings-as-errors yet (see #1823), I think we should just go ahead and merge this.