auhal: do not destroy "audio-device" when AudioHardwarePropertyListener failed, but try to use it anyway.

It allows multiple Safari webplugins to play audio at the sametime, just as 2 vlc instances would do.

It is the main cause for memory leaks of VLC browser plugin for MacOS X. Every time the browser plugin is reloaded, vlc core reloads the plugins which causes all global defined objects to initialize. Since there are objects in libmacosx_plugin.dylib that are supposed to live as long as the application these are never freed. Testing is done with --enable-debug.

This patch adds --rtsp-wmserver which is needed to support pause mode. WMServer does not report any of the known RTSP OPTIONS when asking for it. Instead it returns a list of non-standard options.

Conflicts:

	modules/demux/live555.cpp

(cherry picked from commit be262960c031fbab9e2600969b08e516f20b6679)

Conflicts:

	projects/mozilla/vlcshell.cpp

(cherry-picked from commit aed48363)

Conflict:
	projects/mozilla/control/npolibvlc.cpp

(cherry picked from commit 63ae92da)

However, it only works with the avcodec module. --codec avcodec if you want it.

only case without initialization is for a void variable in which case the val
variable isn't used)

I don't immediately see a solution for.

The .dll is still a bit big, because of static linking to libiconv, I think

For some reasons, this doesn't crash on Linux...
(cherry picked from commit 25d37a4930519c5bcabbe2c0fae8c86e38d9ea54)
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

As a side effect, fix a potential memleak if the variable disappeares in
var_SetChecked.

Destroying directly saves one Lookup.

Also remove an uneeded var_Destroy.

float, ...), so remove all the flags before the check.

created two times with the same type but different flags.

The system can use the hint to prefetch the data into memory, if
possible. Hopefully, this will reduce choppiness, as well as power
consumption.

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Fixes #2983 for good ths time hopefully...

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>