MAINT: Update dependency specifiers

[mne-bot]

Created by spec_zero GitHub action.

• Min versions asserted (should fail): b300f40 / https://github.com/mne-tools/mne-python/actions/runs/21381846599/job/61550187452?pr=13611#step:12:49
• deps/scripts fixed so assertions pass
• fix generation of tools/environment_old.yml Make a note that tools/environment_old.yml might need updating, since it'll probably be reworked in use a lockfile for "old" CI job #13490 anyway
• Adjustments made to shims in mne/fixes.py in this or another PR. git grep TODO VERSION is a good starting point for finding potential updates (esp. sklearn 1.4+)
• Create a doc/changes/dev/dependency.rst as part of action
• CIs green

[larsoner]

Good bot 🚀

But really, nice job @tsbinns ! 👏

[larsoner]

Okay failure in old is legit, need to figure out how to debug it!

[larsoner]

Okay looking here:

  Collecting numpy>=1.21.0 (from pyvista==0.47.dev0)
    Downloading numpy-2.2.6-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (62 kB)

it looks like we have two things to fix (at least)

1. We need to assert something somewhere that the min versions of at least a couple things (NumPy, SciPy) stay at their intended min versions. This could be a new tools/github_actions_ensure_old.py or so that could run after the tools/github_actions_install.sh step. It could read the YAML, import some libraries, and assert the expected version matches the installed mod.__version__.
2. We need to actually fix test_extra and/or the github_actions_install.sh step/script to make sure our old versions don't get uninstalled. In particular I don't think test_extra should install a dev version of pyvista... I think that was probably a workaround we needed at some point that we can now shouldn't need.

I'll try to push some commits for this...

[larsoner]

Oh... we have no code for automatically updating environment_old.yml! I'll update it manually for now but we should fix this...

[tsbinns]

Oh... we have no code for automatically updating environment_old.yml! I'll update it manually for now but we should fix this...

This would be addressed in #13490 when it lands, but is still a WIP

[larsoner]

In the meantime this should be good for review @drammock , I'll fix CIs shortly but the main ideas are in here!

[larsoner]

@tsbinns the old code both updated inplace and set vars, which is unnecessary (and led to some confusion where I assumed things were not updated inplace because the vars were set). I've changed it just to operate inplace where possible.

[larsoner]

@tsbinns actually I think this would be a good one for you to review + merge if you're happy!

[tsbinns]

@larsoner Sure, I'll have a look!