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().

There are no callers which pass audio_format==NULL, and it shouldn't
be allowed.

Request the next song from the playlist (by clearing pc.next_song)
only if the player command is empty.  If it is not, the player may be
clearing the song that has already been queued, leading to an
assertion failure.

Call ob_clear() in decoder_command_finished() instead of implementing
that call in every decoder plugin.

Remember the seek_where argument and call decoder_command_finished()
immediately.  This way, the player thread can continue working, and we
can receive more commands.

This also fixes several issues which resulted in broken frames,
leading to erroneos "elapsed" values: frames weren't parsed properly,
since the code was checking for command!=NONE.

Previously, the function would only return when a STOP was issued.  It
makes more sense to consider all possible commands.

Break the large function mp3_read() into smaller pieces.

Break the large function mp3_read() into smaller pieces.

Break the large function mp3_read() into smaller pieces.

Break the large function mp3_read() into smaller pieces.

Break the large function mp3_read() into smaller pieces.

dc_seek() won't send a SEEK command to the decoder thread unless the
stream is seekable.  No need to do another check; convert that to an
assertion.

need_chunks() returns a decoder_command enum.  Store its return value
as this type.

The function mp3_decode_first_frame() is too large.  Move some code to
separate smaller functions.

http://xkcd.com/292/

This removes the need for util.h.

Use the C99 bool data type for boolean values.

Renamed all functions and variables.  Also removed the mp3DecodeData
typedef.

Yet another superfluous buffering layer.  input_file was using FILE*,
but we're better off with unbuffered I/O using open(), read(), ...

size_t and long aren't 64 bit safe (i.e. files larger than 2 GB on a
32 bit OS).  Use off_t instead, which is a 64 bit integer if compiled
with large file support.

Remove duplicated code from MPD.

When the decoder failed to start, the function do_play() returned,
still having pc.command==PLAY.  This is because pc.command was reset
only when the decoder started up successfully.  Add another
player_command_finished() call in the error handler.

Don't attempt to open a HTTP URL as a local file, and don't send a
local path to libcurl.

Due to a missing "test", the "xyes" token was interpreted as a
command.

Don't compile the sources of disabled output plugins at all.

On some systems, string.h declares basename().  This emits a shadow
warning.  Change the variable name.

If the server sends the headers "icy-name", "ice-name",
"x-audiocast-name", set the stream title.

Replaced the local variable "colon" (which had only temporary meaning)
with the variable "value".  It is a pointer to the first byte of the
header value.

g_free() allows passing the NULL pointer.

For boolean values and success flags, use bool instead of integer (1/0
for true/false, 0/-1 for success/failure).

close() shouldn't fail with read-only streams.

Instead of managing a set of method pointers in each input_stream
struct, move these into the new input_plugin struct.  Each
input_stream has only a pointer to the plugin struct.  Pointers to all
implementations are kept in the array "input_plugins".

Renamed all functions and variables.

The global constructor is empty, and can be removed.

The methods are only used in inputStream_fileOpen(), and should not be
exported.

Everybody should use struct input_stream.

Renamed inputStream.c and inputStream_file.c.

MPD's HTTP client code has always been broken, no matter how effort
was put into fixing it.  Replace it with libcurl, which is known to be
quite stable.  This adds a fat library dependency, but only for people
who need streaming.