Drop Tox?

[benkiel]

Currently causing an issue with setuptools_scm, not used for our automated tests, maybe we just drop? @roberto-arista what do you think?

[benkiel]

See #648 (comment) and #884 (comment)

[benkiel]

Also pinging @alerque for thoughts as he is running into this with Arch

[justvanrossum]

+1 from me. GH Actions matrix + pytest is nice and light. I tend to not test multiple Python versions locally until the CI uncovers an issue.

[benkiel]

Yeah, I pretty much agree with that. Tox had it's place, but it's not worth it

[roberto-arista]

+1

[benkiel]

Ok, the 'joy' here I'm realizing is going to be with uploading the coverage data, tox kinda abstracted that away. I have a WIP PR (#908) if anyone has ideas/experience there

[alerque]

Yes drop Tox. Or if you keep Tox, keep it for developers only. It's a local testing tool that does not (never has, still doesn't) play nice with any kind of end to end testing — which is what makes sense to do if tests are included in release packaging. When packaging for Arch Linux we try to verify the test suite to make sure that the packaged build plays nice with the system dependencies. Tox has never allowed that because it only plays in it's own sandbox. It's fine to use if you really want to for upstream development (although I think there is better tooling out there these days) but keep it in its own lane. It shouldn't dictate how the published assets have to look.