test_egfxset.py is failing due to librosa API change

[dagett]

The test tests/datasets/test_egfxset.py is failing in checks for #576 #577. The test fails where the latest version of librosa 0.10.0 is used - py37 and py38 but not for py36 which is not supported by librosa so uses the old release, version 0.9.2.

It seems the call inside mirdata/annotations.py to librosa.hz_to_note() in latest librosa returns a ndarray instead as for 0.9.2, a list. The test expects a list so fails its assert in tests/test_utils.py and run_track_tests().

There might be many ways to fix this, a short-term one is to lock librosa to version 0.9.2, that should should allow test_egfxset.py to get the expected datatype for all Python versions, and pass those tests. Another is to remove support for Python 3.6 (already EOL) and update the test.

[genisplaja]

Hey @dagett! Thank you very much for bringing this up last month. That helped a lot! We will be removing 3.6 support and moving to 3.9 (and maybe 3.10 soon) to address this problem. I am closing this issue, thanks for your contribution!