Keep the device open as long as the output is enabled, but initialize
it only when playback starts.

When we don't have enough data, generate some silence, hoping the
input buffer will fill soon.  Reducing the render buffer size is not
legal.

Blocking inside the render callback is forbidden, and this sleep call
didn't make any sense.

This plugin's use of the "Timer" library was wrong; it added the same
amount of virtual data in every iteration in _play(), but did not
actually play something.  This created an artificial, but useless,
delay.  This patch implements the method _cancel(), and implements
hard-coded sleep values.  This is only slightly better, but does not
attempt to look sane.

The implementation of cancel() did not work well: you cannot use
alSourceUnqueueBuffers() to unqueue queued buffers, and our function
openal_unqueue_buffers() left the OpenAL library in a rather undefined
state; nothing was supposed to be queued, but the "filled" variable
was not reset.

The expression "!format" does not make sense, and cannot occur.

Unnecessary overhead.

The OpenAL specification says that AL_FORMAT_MONO8 and
AL_FORMAT_STEREO8 expect unsigned 8 bit samples, but MPD uses unsigned
samples.

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

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

.. instead of failing playback completely.

Allow gcc to warn when a new format isn't supported.

The Naim Uniti does not appear to support icecast-style streaming of FLAC
music but does support the codec from a DLNA server. This change looks for
"transferMode.dlna.org: Streaming" in the HTTP request header and responds
with something the Uniti (and hopefully other DLNA clients) accepts.

The only difference in the DLNA streaming mode is the reponse header and
that icecast metadata is disabled. If a client request indicates both modes
are supported, the DLNA mode is preferred (as the Uniti says it supports
both but then rejects a FLAC ICY stream).

Note: This change may be specific to Naim equipment (the only device it was
tested on). E.g. the hardcoding of Content-Length which works but is not a
logically correct value. The change should be backwards-compatible, so
only those clients requesting a DLNA stream will see any difference.

On Win32, the third setsockopt parameter has type (char *) while on POSIX
systems it is (void *). However, given that it is a no-op cast to go from a
char pointer to a void pointer, we can cast to a char pointer (with a
possible const modifier) on all platforms and satisfy the compiler.
Signed-off-by: Dan McGee <dan@archlinux.org>

Wrap close(), use closesocket() on WIN32/WinSock.

The server_socket library (used by the httpd output plugin) didn't
check for WIN32, that's fixed now.

Pass audio_output objects around instead of void pointers.  This will
give some more control to the plugin, and prepares for non-blocking
audio outputs.

Allow enabling the plugin explicitly without running Solaris, to test
the build.

Fixes gcc warnings.

Wrap the locking code.

GLib has a built-in OOM handler.

Consistent naming.

Move the "extern" declarations from output_list.c, for more type
safety.

Provide _lock() and _unlock() to wrap all accesses from the mixer
plugin.

Use these instead of exposing the internal roar_t struct.