Uses a custom assertion to have more detailed output (which should help
with getting an idea what the difference between expected and actual
lyrics is), and use `subTest` to clearly associate test failure to a
backend.
In addition, this strip parenthesis from the lyrics' words. That helps
with the currently failing Tekstowo test, but doesn't entirely fix it.
Requires Python 3.4 for subTest.
Rectify a couple of things in that PR, pointed out here:
https://github.com/beetbox/beets/pull/4226#issuecomment-1011499620
- Undo the `pretend` sensitivity in the import path, because it's not
clear how this setting could ever be true.
- Preserve the log message in debug mode, even when quiet.
Test more combinations of tags that might initially be present and
expected tags. The R128 codepath and the case of having the wrong type
of tags wasn't really tested before.
Also, add a convenience function `store()` that dispatches two the
either of the two methods. This will be useful later, when rewriting the
parallel code (but doesn't simplify the code now).
Renames *GainHandler -> *Task and instead of having a singleton
instance, creates a *Task object for each album/item to process. The
advantage is that now, related data can be bundled in the instance,
instead of passing multiple arguments around.
The plugin has loads of indirection and nested functions which make it
really hard to reason about. The larger picture here is that I'd like to make
the code more manageable before reworking the parallelism issues.
In particular, instead of manually implementing an interface using a
function that returns a tuple of function pointers, this commit creates proper
classes. Again, no functionality is changed, this only moves code
around.
Another incorrect py2 -> py3 translation. Since python 3 attached the
traceback to the exception, this should preserve the traceback without
needing to resort to sys.exc_info
This is an incorrect translation of a python 2 reraise to python 3.
With python 3, however, we can just rely on exception chaining to get
the traceback, so get rid of the complicated re-raising entirely, with
the additional benefit that the exception from the tear-down is also
shown.
A recent change in Sphinx introduced a new warning about missed extlink
opportunities:
https://github.com/sphinx-doc/sphinx/issues/10112
These warnings are causing spurious CI failures. Some of these
suggestions are good but many of them are not, and there is not
currently a way to disable the warning (globally or locally). So the
only workable solution currently seems to be to pin an old version of
Sphinx in CI for now. Hopefully there will be an option to disable this
in 4.4.1, at which point we can unpin.
