Somehow we lost 600e1322 after renaming some variables in ed6f6046.

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

[mk: by definition, tag_new() cannot fail - removed check]

Define the special value "-1" as "unknown size".  Previously, there
was no indicator for streams with unknown size, which might confuse
some decoders.

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

neaacdec.h declares all arguments as "unsigned long", but internally
expects uint32_t pointers.  This triggers gcc warnings on 64 bit
architectures.  To avoid that, make configure.ac detect whether we're
using Debian's corrected headers or the original libfaad headers.  In
any case, pass a pointer to an uint32_t, conditionally casted to
"unsigned long*".

In 432da18e a dynamic buffer was replaced by a static one but some
frees were accidently left there which caused some segfaults.

The wavpack open function gives us an option called OPEN_STREAMING. This
provides more robust and error tolerant playback, but it automatically
disables seeking. (More exactly the wavpack lib will not return the
length information.) So, if the stream is already not seekable we can
use this option safely.

Wavpack plugin doesn't stop decoding if a block couldn't be fully
decoded, rather it tries to go on.

mp3 seeking was broken, because the command==SEEK check was never
reached.  Swap the command check order (==SEEK before !=NONE) to fix
that.

Unfortunately, ov_fopen() is not supported by libvorbis versions older
than 1.2.

When the buffer was full, but everything was already consumed,
fillAacBuffer() would not attempt to flush and refill it.

Allocate the input buffer within the AacBuffer struct.

The aac plugin does not support seeking.  Reject SEEK requests by
calling decoder_seek_error().  Quit the plugin's main loop only when
STOP is received.

Removed a superfluous decoder_get_command() call.

The audio_format variable is only used and initialized for
decoder_initialized().  Move it into that block to save some bytes on
the stack.

aac_stream_decode() was basically copy+pasted from aac_decode().
Since stream_decode() can also decode files, eliminate aac_decode().

Check whether enough data has been read yet.

If the stream is not seekable, don't try to decode all frames to find
out the total song time.

Use input_stream_eof() instead.

According to the documentation, mpc_decoder_decode() returns an
mpc_uint32_t.  Since the special return value (mpc_uint32_t)-1
translates to a very large long integer, this may cause segmentation
faults if not interpreted properly.

Don't hard-code the factor "2".

Removed 3 superfluous decoder_get_command() invocations.

mpc_to_mpd_buffer() converts a whole chunk at a time.  This eliminates
3 local variables in mpc_decode().

Don't split the buffer conversion loop.  When libmpcdec returns a
chunk, convert and send the whole chunk at a time.  This moves several
checks out of the loop, and greatly improves performance.

Don't use deprecated MPD libraries.

The buffer is always casted to int32_t* anyway; declare it as int32_t
array, and remove the cast.

Since each "eof=true" is followed by "break", the variable is
superfluous.

Due to an unnoticed merge error, there was a superfluous "}".  Remove
it.

Move the reponsibility for freeing the file handle to libvorbis.

Eliminate two decoder_get_command() invocations from the main loop.

Don't let the buffer grow until it is full, flush it whenever there is
data available.

Parse ID3 tags, even when they are in the middle of the stream.  Very
few streams provide embedded ID3 tags.  Most of them send only
Shoutcast "icy" tags, which limits the practical usefulness of this
patch.

When a command is received, decode_next_frame_header() and
decodeNextFrame() return DECODE_BREAK.  This is already checked by
both callers, which means that we can eliminate lots of
decoder_get_command() checks.