MNT: Do not completely take over test collection ignore

[pllim]

Main motivation is to fix compatibility with pytest 7.4.0.dev though theoretically this plugin should not completely take over the ignore part of test collection? See pytest-dev/pytest#11082 (comment) for more details. Affected issue:

• TST: extern/configobj no longer excluded by pytest-dev astropy/astropy#14930

I am adding a test first to make sure it fails in pytest-dev before trying out the patch. I am also using pytester instead of testdir, see #200 .

[pllim]

Indeed the test added here threw error with pytest-dev without any patch:

E     File ".../test_norecursedirs0/bad_dir/test_foobar.py", line 2
E       def f():
E   IndentationError: unexpected indent

With the patch, error is gone.

[bsipocz]

extreme nitpick, otherwise looks good.

[bsipocz]

Spell it out pls :)

[pllim]

OK, done! 😸

[bsipocz]

Unrelated, but I suppose we should maybe test all OSes with pytest-dev? The full CI matrix is done in 4minutes, so it wouldn't add much additional burden.

[bsipocz]

Thx 😉 !

[pllim]

test all OSes with pytest-dev?

But what value does that add? We don't do that with other packages right now, and hopefully pytest itself would test across different platforms. Does not hurt but I also don't see a real need, so up to you.

[bsipocz]

I would just like to have a more comprehensive test suite (I do recall where we indeed do an equal matrix on the 3 OSes), at worst, we waste ~5min of CI time once a week. We do walk through all the old versions already, but not that pedantic with newer versions, e.g. all are missing between 7.0 and dev.

[pllim]

Actually, maybe that is a very good idea especially since doctest deals with paths a lot and Windows is always so special. But perhaps cron instead of per commit? I opened #202 about this. Thanks!

[bsipocz]

ATM I prefer the per-commit option (and run all from cron), at least until the notifications are fixed for cron jobs.

[The-Compiler]

FWIW, I'd use an explicit return None here - from PEP8:

Be consistent in return statements. Either all return statements in a function should return an expression, or none of them should. If any return statement returns an expression, any return statements where no value is returned should explicitly state this as return None, and an explicit return statement should be present at the end of the function (if reachable)

[pllim]

Thanks, @The-Compiler ! I have done that in 7b605ef directly in main since we're about to move the repo and I don't want extra overhead for a trivial change.