FIX: fix topo plot when large channel distance gives rise to invalid values for arcsin

[mmagnuski]

Fixes #9526 and #9511. I'll add the example posted by @agramfort as a test later.

[rob-luke]

Could this possibly fix/close #8291 too?

[mmagnuski]

Unfortunatelly #8291 is much more complex, this PR is a simple bug-fix. For #8291 we would first need some logic for detecting sensor groups.

[rob-luke]

Thanks for the clarification @mmagnuski, I was just hoping for a magic fix 😄

[agramfort]

@mmagnuski can you just a what's new entry? thx for the quick reaction to fix this one !

[mmagnuski]

No problem! what's new entry added, waiting for the CIs.

[mmagnuski]

One CI task (Azure Main Ultraslow) timed out, rest is green.

[larsoner]

FYI this is no longer necessary as we have a magic autouse mpl fixture that calls plt.close

mne-python/mne/conftest.py

Lines 131 to 137 in 2e200aa

	@pytest.fixture(autouse=True)
	def close_all():
	"""Close all matplotlib plots, regardless of test status."""
	# This adds < 1 µS in local testing, and we have ~2500 tests, so ~2 ms max
	import matplotlib.pyplot as plt
	yield
	plt.close('all')

[mmagnuski]

ok, good, I just followed what other tests there used (and actually thought about a decorator like that), I can change this in the next PR.

[larsoner]

I am slowly eliminating the ones at the end of functions as I see them. (The ones in the middle of tests are still useful to prevent too many MPL windows from popping up, eating memory, and possibly emitting warnings about 20+ windows.)

[larsoner]

Thanks @mmagnuski !

[mmagnuski]

Thanks @larsoner @agramfort! 🚀