This avoids naming conflicts in the source directory. In particular, when
encoding MP3 -> MP3, the previous scheme would overwrite the original file
(and hang ffmpeg waiting for input). This should also work in
situations where the source directory is read-only.
I introduced a regression a few commits ago when I started using
lib.destination with the basedir keyword argument as opposed to doing
os.path.join manually.
The major functional change here is how files move around when in keep_new
mode. Now, files are first moved to the destination directory and then
copied/transcoded back into the library.
This avoids problems where naming conflicts could occur when transcoding from
MP3 to MP3 (and thus not changing the filename).
Instead of flac and lame the convert plugin now uses ffmpeg. This adds
support for more input formats and simplifies the code. ffmpeg also uses
the lame encoder internally and has equivalents of all the -V<num>
presets which should be sufficient.
Paging @yagebu: I think the old version of the code would embed album art into
the wrong file. Please correct me (and accept my apologies) if I'm wrong
though.
A user reported a problem with one of the logging statements where .format()
tried to convert a Unicode string to bytes because the log message was '', not
u''. As a rule, we should ensure that all logging statements use Unicode
literals.
Eliminate the __subclasses__ trick for finding all plugins. Now we explicitly
look in each plugin module for a plugin class. This allows us to import plugin
modules with unintentionally loading them. This lets us reuse the image
embedding machinery without copypasta.
