Use GLib's G_GNUC_UNUSED instead of gcc.h's mpd_unused.

The old code casted it to a 32 bit integer, which cut off bits.
AVFormatContext.duration is a int64_t, so use this type.

Use NULL instead.  Found by sparse.

Refuse to play audio formats which are not supported by MPD.

The ffmpeg library provides some of the song metadata in the
AVFormatContext struct.  Pass it from there to MPD.

With whence==AVSEEK_SIZE, the seek function should return the file
size, not the current offset.  Check the return value of
input_stream_seek().

The stream_decode() and file_decode() methods returned a boolean,
indicating whether they were able to decode the song.  This is
redundant, since we already know that: if decoder_initialized() has
been called (and dc.state==DECODE), the plugin succeeded.  Change both
methods to return void.

Instead of having a seprate try_decode() method, let the
stream_decode() and file_decode() methods decide whether they are able
to decode the song.

Instead of checking the stream_types bit set, we can simply check
whether the methods stream_decode() and file_decode() are implemented.

The function decoder_read() already cares about the decoder command,
and loops until data is available.  Reduced mpd_ffmpeg_read() to no
more than the decoder_read() call.

That variable is never used except during initialization in
ffmpeg_helper().

Renamed variables.

ffmpeg_tag() deleted the tag when ffmpeg_helper() returned success.
The return value was interpreted incorrectly, it should return the tag
on success.

Rewinding the stream here is not useful, but may consume valuable
resources (and time).

decoder_read() checks the decoder command.  Without this patch, the
ffmpeg plugin could become unresponsive.

Ogg and ffmpeg detection was disabled when the stream was not
seekable, because the detection was too expensive.  Since the curl
input stream can now rewind the stream cheaply, we can re-enable
detection on streams.

Don't pass the "seekable" flag with every decoder_data() invocation.
Since that flag won't change within the file, it is enough to pass it
to decoder_initialized() once per file.

The strings were constant, but the pointers weren't.  C syntax is
somewhat tricky..

All decoder_plugin structs are initialized at compile time, and must
never change.

decoder_data() uses wait times to let the input stream continue its
transfer.

decoder_data() always returns the current command.  If we use this, we
can save a lot of decoder_get_command() calls.

Converted the runtime check to an assertion.

Remove one indent level.

Move code from ffmpeg_decode_internal() to make it smaller and more
readable.

The decoder API provides the function decoder_seek_error() to report
seek errors.  Use this function instead of logging the error.

The function mpdurl_read() is too complicated, and uses the wrong data
types.

tag objects must be freed with tag_free() to ensure that all resources
are freed.

The function ffmpeg_tag() already has the variable base.tag, which can
be used for this.

ffmpeg_try_decode() did not interpret ffmpeg_helper()'s return value
properly; migrate everything to bool to make it consistent.

We don't need those anymore, they just fill the log.

ffmpeg_tag() did not initialize base.decoder, which made valgrind
unhappy, and can lead to a egmentation fault.

Don't return 0/-1 on success/error, but true/false.  Instead of int,
use bool for storing flags.

A decoder_flush() invocation was missing in the FLAC plugin, resulting
in casual assertion failures due to a wrong assumption about the last
chunk's audio format.  It's much easier to remove that decoder_flush()
function and make the decoder thread call ob_flush().