Comment as on IRC:

This should probably also add a --check run of black, and flake8, to make test.

Potentially we could then drop black and flake8 as separate GitHub actions because the main testing would run them, but then we'd lose inline notes from TrueBrain's action.

I added them to "make test".
But now "make test" depends on "flake8" and "black" being installed.

I don't think that's an issue really; black at least you pretty much need to make any significant changes without upsetting the commit-checker.

Perhaps the people building distro packages will mind?

In principle we could have a test script that checks whether they're installed and warns but succeeds otherwise, but that seems TMWFTLB.

I guess this should add black and flake8 to "Optional Dependencies" of README.md.

Not sure I like this being rolled into the make test rule. A separate "check" rule, perhaps?

Not sure I like this being rolled into the make test rule. A separate "check" rule, perhaps?

As much as a regression shows a runtime error, flake shows possible runtime errors. One can debate if black is really a test, but flake for sure is. And with having flake under make test, you might as well add black too, as .. why not! :D

Conclusion: noone likes pyproject.toml