✔️ Linting Passed

All linting checks passed. Your pull request is in excellent shape! ☀️

_{Generated for commit: 32ae3d0. Link to the linter CI: here}

Argh, reviewed twice I see now - I thought this was new. Scikit-learn devs: I suggest to merge these cleanup PRs from @mtsokol rapidly; it helps avoid failures when we merge the corresponding cleanups in numpy's main branch.

I think this PR can be closed since #27190 got merged.

The only thing that could still be useful are the changes to sklearn/utils/tests/test_validation.py. What do you think @lesteve ?

Sorry I did not see this PR indeed before I opened #27190, which started as a single fix and ended up as "let's try to fix as many numpy dev and scipy dev issues while I am at it".

I think this can be closed. AFAICT the changes in sklearn/utils/tests/test_validation.py are reordering, for example:

• ("uint16", "ushort", np.unsignedinteger) instead of ("ushort", "uint16", np.unsignedinteger). @mtsokol was there a good reason to change the ordering? I would expect the test to be symmetric but maybe I am missing something ...
• ("longdouble", "longdouble", np.floating) instead of ("longfloat", "longdouble", np.floating). For this one I maintained a different dtype IIRC since I think this is the test aims to test different dtypes.

I am going to close this one, but feel free to reopen if I missed something!

* `("uint16", "ushort", np.unsignedinteger)` instead of `("ushort", "uint16", np.unsignedinteger)`. @mtsokol was there a good reason to change the ordering?

@lesteve It's just a detail, to match other entires (first canonical name str, then sized alias str). I think it can be closed.