Mne-python: BUG: Big Sur compat

[larsoner]

More of an FYI than anything else -- PyQt5 appears not to work on Big Sur. I've tried 5.12 and 5.15.1, and both basically hang as soon as you make a mayavi, PyVista, or matplotlib Qt5Agg window.

So those thinking about upgrading, don't do so if you want to be able to use Python. For matplotlib at least we can use the MacOSX backend in the meantime, but 3D plotting is now (hopefully temporarily) totally broken for people who have upgraded. :(

Or maybe I'm the only one with this issue. If someone else upgrades, feel free to try it out and let me know!

[larsoner]

Closing because there isn't really anything for us to do. But we can keep any discussion about this here as things progress

[cbrnr]

Thanks for the heads up. I think this time I will upgrade as late as possible (this is the first time I haven't upgraded on day 1, and just like the new CPUs the macOS update doesn't really hype me).

[christianbrodbeck]

os.environ['QT_MAC_WANTS_LAYER'] = '1' (stackoverflow) at least gets the GUIs to open, will do some more testing.

[cbrnr]

I've done the upgrade and of course things don't work. Workaround for now: manually download the .whl files, change 10_9 to 11_0 and install with pip.
https://github.com/numpy/numpy/issues/17784
https://github.com/scipy/scipy/issues/13102
matplotlib
Pillow
PyQt5
PyQt5-sip
It seems like most packages on PyPI are incompatible with Big Sur, just because they have the wrong version number. Renaming works. Reopening because I'm expecting others will also run into these issues.
Edit: This is a pip issue (https://github.com/pypa/packaging/pull/319), hopefully everything will work with the next release (no idea when this is going to happen though).

[hoechenberger]

I have been running Big Sur for over a week or so and no problems besides the Qt thing. I'm using conda-forge.

[cbrnr]

Yes, it's a pip issue. I know we recommend/support conda, but pip is still the default package manager and currently that doesn't work.

[hoechenberger]

Yep. Didn't mean to suggest otherwise, just wanted to share that it works well for me with conda-forge :)

[cbrnr]

The latest releases of pip and Homebrew fixed all issue, so everything is working on macOS/Homebrew.