Empty test directories cause tests to fail

[Conduitry]

A mild annoyance, but one that has come up several times for me. If a directory that's supposed to contain a test is empty (which can happen when the test was removed remotely, and then upon pull, git decides to not remove the empty directory), we still try to run that test, but fail when its main.html/whatever cannot be opened.

Not sure what would be optimal here. Ignoring empty directories? Maybe even removing them when running tests?

[Rich-Harris]

yeah, it is a nuisance. Also not sure what the right move is. Feels like we could easily run into false positives/negatives if we deleted or ignored them, and we'd have to define 'empty directory' differently for each type of test since they're actually not empty — they just only contain gitignored files.

[Conduitry]

Oh right, I wasn't thinking about the gitignored files. Yeah, can't really blame git for this one. It's just an unfortunate confluence

[Conduitry]

I think I'm just going to close this. Rather than trying to get fancy in the test code, this can be fixed by a simple git clean -dfx test before running tests.