PA_SAMPLE_S16NE is the only sample format which is suported by both
MPD and pulseaudio.  Unfortunately, pulse does not accept 24 bit
samples.

Instead of bailing out with an error message, we should tell the MPD
core to convert all samples to 16 bit for pulse.

Eric is too busy with other projects and will remain inactive
indefinitely.

This bug caused the audio output devices to stay open, although MPD
wasn't playing: quitDecode() resetted player_control.command, assuming
that the command was STOP.  This way, player_task() didn't see the
CLOSE_AUDIO command, and the device was kept open.

Don't clear player_control.command in quitDecode().

These are results from failed merges which I didn't notice.

When the audio source provides 24 bit samples, don't bother to convert
(lossily) them to 16 bit before jack's floating point conversion - go
directly from 24 bit to float.

Move sample format dependent code to a separate function.

Renamed all variables and functions.  Add the prefix "mpd_jack_" to
function names.

We must never pass partial frames.  Added assertions to debug this.

Merge the variables "avail_data" and "avail_frames" into "available".
Both variables are never used at the same time.

The JACK documentation postulates that the process() callback must not
block, therefore locking is forbidden.  Anyway, the old code was racy.

Remove all locks, and don't wait for more data to become available -
just send to the port what is already in the buffer.

Don't wait until there is room for the full data chunk passed to
jack_playAudio().  Try to incrementally send as much as possible into
the ring buffer.

Don't hard-code a frame size of "4" (16 bit stereo), calculate the
sample size from sizeof(*buffer), and create the constant
"frame_size".

Indent with tabs.

libsamplerate 0.1.2 didn't have the 32 bit <-> float conversion
routines.  Emulate them in case they aren't supported.

Another partial frame fix: the silence buffer was 1020 bytes, which
had room for 127.5 24 bit stereo frames.  Don't send the partial last
frame in this case.

24 bit output is as important as 16 bit output.  Provide a
pcm_convert() implementation which can convert to 24 bit with as
little quality loss as possible.

The old pcm_convert_size() ignored most of the destination format,
e.g. it did not check its sample size, and assumed it is 16 bit.
Simplify and universalize it by using audio_format_frame_size().

pcm_convert() converted only to 16 bit.  To be able to support other
sample sizes, move that 16 bit specific code to a separate function.

The 24 bit implementation is mostly copy'n'paste of the 16 bit
version, except that the data type is int32_t instead of int16_t.

Separate code from pcm_utils.c to keep it small and simple.

Similar to pcm_resample_16(), implement pcm_resample_24().  The 24 bit
implementation is very similar, but it uses src_int_to_float_array()
instead of src_short_to_float_array() before sending data to
libsamplerate.

A future patch will implement a 24 bit resampler.  To unify code, move
code which can be shared to a separate function.

Copy from source to destination buffer directly, don't use the
temporary variables "lsample" and "rsample".

Added assertions which ensure that there are no partial samples in the
source buffer.

Use sizeof(sample) instead of hard-coding "2".  Although we're in 16
bit right now, this will make code sharing easier when we support
other sample sizes.

Separate the resampling code from the rest of pcm_utils.c.  Create two
sub-libraries: pcm_resample_libsamplerate.c and
pcm_resample_fallback.c.

Due to a logic error, no value was valid for the boolean value
parser.  Replace "||" with "&&".

libmad produces samples of more than 24 bit.  Rounding that down to 16
bits using dithering makes those people lose quality who have a 24 bit
capable sound device.  Send 24 bit PCM data, and let the receiver
decide whether to apply 16 bit dithering.

We are going to convert the code to 24 bit; don't hard-code a sample
size of 2 bytes.

Copied and adapted code from the mp3 decoder plugin.  This library now
replaces the old and low-quality function pcm_convert_24_to_16().

I added 24 bit support a while ago, but it wasn't possible to force 24
bit output.  Add 24 and 8 bit to the list of allowed sample sizes.
Although 8 bit audio isn't as widely used as 24 bit, there is no
reason to exclude it.

Splitting a frame between two buffer chunks causes distortion in the
output.  MPD used to assume that the chunk size 1020 would never cause
splitted frames, but that isn't the case for 24 bit stereo (127.5
frames), and even less for files with even more channels.

Don't attempt to map paths which are already absolute with
map_song_fs(): check with song_in_database() instead of
song_is_file().

The "rm" command did not send notifications to idle clients.  Add it.

The function deletePlaylist() shouldn't be in playlist.c.

Don't use our deprecated linked list library, use GLib's GPtrArray
instead.

The "long" result of strtol() was implicitly casted down to a 32 bit
integer.  Add some range checking instead.

Many command arguments must not be negative; add a separate
parser/checker function for that.  For the same reason, add
check_bool().  This eliminates two strange special cases handlers from
check_int().

Pass index arguments as unsigned integers.  They must not be negative,
and even if some caller accidently passes -1, it won't pass the bound
checks (since it's now 2**32-1).

There are some integers which have a "magic" -1 value which means
"undefined" or "nothing".  All others can be converted to unsigned,
since they must not contain a negative number.