Silently choosing default is misleading and can cause hours of investigation.
It's better to fail immediately telling user what is wrong with config.

If output plugin fails to init it will try to call ao_base_finish() immediately,
which segfaults because replay gain filters are not initialized yet and contain
garbage values.

It appears the musepack section has not really been reviewed in some
time, many parts unfunctional, others, just unnecessary.

Conflicts:
	Makefile.am
	NEWS
	configure.ac
	src/encoder/flac_encoder.c
	src/log.c
	src/pcm_buffer.c

Initialize the audio_format before calling avcodec_open(), because
avcodec_open() will fill bogus values.

Yet another common support case.

This is a common support case, and hopefully, the new error message
will allow the user to understand the error without requiring support.

Use stat() instead of g_file_test() to detect other types of errors,
such as "permission denied".

Convert to padded 24 bit samples, instead of falling back to 16 bit.

This fixes a buffer corruption bug; pcm_buffer is not designed to be a
persistent buffers, and will discard anything between two consecutive
calls.

This fixes a buffer corruption bug; pcm_buffer is not designed to be a
persistent buffers, and will discard anything between two consecutive
calls.

This fixes a buffer corruption bug; pcm_buffer is not designed to be a
persistent buffers, and will discard anything between two consecutive
calls.

For growing FIFO buffers.

This method is too complex for inlining.

"stat" is a macro on mingw32, which is a pretty stupid thing, and this
commit works around this build failure.

Fixes warning on mingw32.

Ensure that the io.thread variable is set before entering the event
loop.

Fixes assertion failure.

Commit 3a3158 introduced a small bug where the pulseaudio mixer would not link
with the main binary in some situations.

WinAPI explicitly declares filesystem encoding.
It can be determined by GetACP().
Use that instead of Glib routine that always "detects" UTF-8 on Win32,
which is incorrect for MPD case.

Eliminate the conversion to integer samples, the MPD core can do this
now.

For MPD's new floating point support: when a decoded wavpack file
needs to be resampled, don't convert float to int and back to float.

For future internal use.

Support for conversion from float to 16, 24 and 32 bit integer
samples.

Let the caller deal with a failure.