Open
Description
We have a recurring problem where tests are skipped without us realizing. See the discussion on #26852 and elsewhere.
Can we think of a way to make this more robust? Perhaps by adding an additional command line option to our pytest calls
pytest pandas --no-skip-xclip
Then we would update our calls to pytest.importorskip
to use a pandas' wrapper that checks the config options. By default, it'll still skip, but if the --no-skip-{dep}
flag is passed we would fail.
This would complicate the actual pytest
call in the CI scripts, though it's already quite complicated.