- Adds tests for eot_track vs next_track as we were only testing the one code
path.
- Makes test_random_until_end_of_playlist_and_play_from_start actually test
random and not repeat.
- Adds test_play_track_then_enable_random which is the underlying bug covered
by the regression test for issue #17.
- Replace setting the random seed with mocking out shuffle with reverse. This
makes tests easier to reason about while still being able to test correct
behaviour.
- Increases number of tracks to four to avoid test passing due to symetry in
plain and reversed lists.
- Made test_eot_track_with_random_after_append_playlist actually test eot and
not next.
- Found a test failure in handling of test_random_until_end_of_playlist, fix in
next commit.
- Checks that adding a non local:track:... uri to a playlist and reading it
back works.
- Also fixes mistake in API usage where we assumed lookup still raised
LookupError.
Adds a fallback to `Track(uri=uri` when track lookup fails for playlists. This
means we can at least load metadata less tracks giving users functioning
playlists, instead of only supporting `local:track:...` style URIs.
Issue is not fixed, but this is sufficient to reduce priority until we get to
the larger planed refactor for this and other core API issues.
- New API wraps gst.pbutils.Discoverer for mopidy use.
- Custom scanner code removed.
- Callback based async API switched with simpler synchronous API
- Scanner is feed one and one URI and returns the result or raises an error.
- Adds new ScannerError exception for error handling in scanner
- Limits local/scan_timeout to 1000-3600000 i.e. 1s to 1h
- Files with zero duration are now excluded.
- Mtime handling moved out of scanner so we can use it for streams etc.
The test was modified to not fail after a refactoring, making it not test what
it was intended to test at all. This reverts the changes and updates the test
in another way, keeping the original intention.
