MNT: move pytest.ini configs to .toml

[story645]

PR summary

Fixes the pytest asserts not showing using @QuLogic 's suggestion of turning on the import-mode=importlib and moves the contents of pytest.ini into pytest.toml to consolidate since we're not configuring a lot.

PR checklist

• "closes #0000" is in the body of the PR description to link the related issue
• new and changed code is tested
• Plotting related features are demonstrated in an example
• New Features and API Changes are noted with a directive and release note
• Documentation complies with general and docstring guidelines

[rcomer]

I can reproduce the "partially initialized" errors locally against main with

python -mpytest lib/matplotlib/tests/test_basic.py --import-mode importlib

I note that CI uses python -mpytest

matplotlib/.github/workflows/tests.yml

Line 311 in 112a225

python -mpytest -raR -n auto \

Pytest docs say this about invoking that way.

I also note that the following gives me a bunch of failures that I can't begin to debug

pytest lib/matplotlib/tests/test_backends_interactive.py --import-mode importlib

[story645]

Oh, do we not actually use our .toml file on CI? ☹️

[ksunden]

I can confirm locally that pytest ... works but python -m pytest ... fails with this branch checked out.

setting the pythonpath argument in config has no effect. (Didn't really expect it to, but hoped a little bit.)

However running as python -I -m pytest ... seems to elide that problem (-I means "run in isolated mode", ignoring both PYTHON* env vars and not adding the current directory to sys.path), thus it also escapes the "I have PYTHONPATH set" problem, which I also encountered locally (though I will make it so I can just run pytest ... because that is how I prefer it... but that much can be on me)

[ksunden]

note that in azure we set PYTHONFAULTHANDLER=1, and -I will ignore this, so we would also want to add -X to enable it via flag instead of envvar. (if we want to keep it)

[QuLogic]

Might setting -P / PYTHONSAFEPATH work here? python/cpython#31542 I don't have an environment with the new version setup to test right now.

[ksunden]

-P helps with the "I have an env var set" part of the problem, but not the "current directory is added to the path" part, which is what is actually causing the main part of the problem when running with python -m

[QuLogic]

But the latter is the point of -P...

[ksunden]

Ah, sorry, I had it confused with -E, so I didn't override that part and it still failed... but yeah, -I is essentially just a combination of -E and -P (well, I guess also -s, which disables user site-packages which could actually be problematic in some installs...)...

-E helps with "I have PYTHONPATH set and it fails" -P helps with "Current directory is causing it to fail"

[ksunden]

Somehow I get the same error missing either of those two (with PYTHONPATH set)

[timhoffm]

This is the default nowadays and can thus be omitted:

https://docs.pytest.org/en/7.1.x/reference/reference.html#confval-junit_family

[story645]

so is python_files =test_*py so I'm also gonna omit that
https://docs.pytest.org/en/7.1.x/reference/reference.html#confval-python_files

[timhoffm]

Not exactly.

By default, files matching test_*.py and *_test.py will be considered test modules.

We don't have any *_test.py files under lib (and only one in the whole project: galleries/examples/units/evans_test.py). So, effectively, you'll detect the same files when leaving it out.

[QuLogic]

I didn't go back in the history, but I'm pretty sure we added python_files first to avoid running evans_test.py, but later we added testpaths = lib, which excluded it in a different manner.

[timhoffm]

I'm +0.05 on leaving python_files =test_*py to be very explicit what we consider tests. (But OTOH, I likely would not have intoduced it if it was not there).

[QuLogic]

To be clear, I don't think it needs to stay either.

[story645]

my bias is to not have two configurations that we think are doing the same thing cause I think that'll be harder to debug when everything breaks

[QuLogic]

I think at least the beginning of this comment is important to preserve.

[story645]

Moved to .toml