MAINT: Ensure we dont download without permission

[larsoner]

Lots of failures locally with this.

Okay for me to put download=False in a bunch of data_path() calls and add requires_testing_data calls where necessary?

• Show failures: https://github.com/mne-tools/mne-bids/actions/runs/3152298983/jobs/5127340379
• Fix failures
• Revert change to use larsoner branch in favor of MNE-Python main once MAINT: Ensure no datasets are downloaded in tests mne-python#11213 is merged

[hoechenberger]

Good enough for me!

[larsoner]

A few things I found and fixed along the way:

1. Don't create data = np.empty(...) for RawArray in tests. The array can contain junk and lead to casting errors (e.g., to float32 when writing).
2. Don't use bare np.random.random(...) for the same reason. It can contain numbers that fit in float64 but not float32 leading to overflow errors when casting.
3. Added a bad_frame_xfail that we can remove once BUG: Add estimated fiducials when missing / assumed head coords mne-python#11212 is merged and any necesary changes are made here, in the meantime it allows tests here to be green
4. Added to CIs a run that executes tests without the testing data. This should be quick (locally takes 6 sec) and pass, and ensure that requires_testing_data decorators are used properly.

Pushing a commit now that hopefully will pass!

[larsoner]

... I also added dataset caching, and removed one use of sample data where we could just as easily use testing.

[codecov]

Codecov Report

Merging #1073 (888dfa7) into main (7be49cf) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##             main    #1073   +/-   ##
=======================================
  Coverage   94.87%   94.87%           
=======================================
  Files          24       24           
  Lines        3845     3845           
=======================================
  Hits         3648     3648           
  Misses        197      197           

Impacted Files	Coverage Δ
mne_bids/write.py	96.79% <100.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[agramfort]

shall we keep this?

[larsoner]

Yes it's just more specific/explicit and a tiny bit faster

[sappelhoff]

but we should probably remove the trailing -b main, right? that's just a single flag specified twice?

[larsoner]

All green, ready for review/merge from my end

[agramfort]

thanks a lot @larsoner 🙏

[sappelhoff]

Thanks!