CI: Test circle

[larsoner]

CicleCI is failing on main, let's see what the actual error is.

[larsoner]

Failure replicated!

@GuillaumeFavelier can you replicate? I suspect this might be due to #10430

[GuillaumeFavelier]

can you replicate?

Yes I can replicate on this PR. I'm on it 👍

generating gallery for auto_tutorials/inverse... [ 27%] 30_mne_dspm_loreta.py                                                                                                                
Traceback (most recent call last):
  File "/home/guillaume/source/anaconda3/envs/mne-py38/lib/python3.8/site-packages/sphinx/events.py", line 101, in emit
    results.append(listener.handler(self.app, *args))
  File "/home/guillaume/source/anaconda3/envs/mne-py38/lib/python3.8/site-packages/sphinx_gallery/gen_gallery.py", line 514, in generate_gallery_rst
    ) = generate_dir_rst(
  File "/home/guillaume/source/anaconda3/envs/mne-py38/lib/python3.8/site-packages/sphinx_gallery/gen_rst.py", line 402, in generate_dir_rst
    intro, title, cost = generate_file_rst(
  File "/home/guillaume/source/anaconda3/envs/mne-py38/lib/python3.8/site-packages/sphinx_gallery/gen_rst.py", line 1069, in generate_file_rst
    clean_modules(gallery_conf, fname, 'after')
  File "/home/guillaume/source/anaconda3/envs/mne-py38/lib/python3.8/site-packages/sphinx_gallery/scrapers.py", line 596, in clean_modules
    reset_module(gallery_conf, fname, when=when)
  File "/home/guillaume/source/mne-python/doc/conf.py", line 340, in __call__
    _assert_no_instances(Brain, when)  # calls gc.collect()
  File "/home/guillaume/source/mne-python/mne/utils/misc.py", line 378, in _assert_no_instances
    assert n == 0, f'\n{n} {cls.__name__} @ {when}:\n' + '\n'.join(ref)
AssertionError: 
1 Brain @ mne/conf.py:Resetter.__call__:after:30_mne_dspm_loreta.py:
Brain._cleaned = None
method: <bound method Brain._play of <mne.viz._brain._brain.Brain object at 0x7ff68f8975e0>>
method: <bound method Brain.restore_user_scaling of <mne.viz._brain._brain.Brain object at 0x7ff68f8975e0>>

[larsoner]

One possibility is that the Brain is never properly/completely __init__ed, because I had to add the getattr for _cleaned, but this is set in __init__ I think.

[larsoner]

(and then the example fails, and during cleanup it does the GC check, which fails because cleanup was never attempted)

[GuillaumeFavelier]

This happens because we use partial in mne/viz/backends/_qt.py. I chose partial to patch the NotImplementedError in Brain.__hash__() (for PySide2), I'll have to find something else.

[larsoner]

I chose partial to patch the NotImplementedError in Brain.hash(), I'll have to find something else.

It would probably work to just have Brain.__hash__ return id(self). As long as we really do clean everything up so Qt doesn't try to call the method again, it hopefully wouldn't lead to segfaults :)

[GuillaumeFavelier]

It would probably work to just have Brain.__hash__ return id(self)

If you do not mind, I can try in this PR.

it hopefully wouldn't lead to segfaults

If it is the case, it is a sign that cleaning is not thorough anyway

[larsoner]

... actually, even better would be to just set self._hash = time.time_ns() during __init__.py, and have __hash__ return this value. There is no way in practice that we will instantiate two brain objects at the same time, right? :)

But really I don't think it's going to matter much, we just have to find a way to placate pyside2. Do you know offhand, does it hash the self of any bound method for some reason? I searched a tiny bit and didn't find anything.

If you do not mind, I can try in this PR.

Yes, please test locally then push here, and I'll merge if CIs are green.

[larsoner]

... please push with [circle front] or touch an example that uses Brain, though!

[larsoner]

Thanks for the quick fix @GuillaumeFavelier !