It closes #4585.

It close #4828 and #3685.

It will allow to share them with the AVI demuxer.

this breaks consitency with the other OS X intf code, but is far more readable

This would allow creating audio output objects in different context.

This should fix #6875

Anyway V4L1 is dead, so there is no ambiguity.

Cameras such as D-Link's DCS-932L don't provide the first video
boundary.

This fixes the freetype ATS lookup for some Asian fonts such as Hiragino Micho ProN with multiple weights.

freetype: in case the provided font name is neither found by name nor by family, let ATS check whether it is a correct PostScript name

This fixes issues with some Asian fonts as pointed by Naohiro Koriyama

freetype: provide a proper fix to prevent ATS from crashing when searching for the font family instead of the fontname

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

Authors are:
 - Jean-Paul Saman
 - Hugo Beauzée-Luyssen
 - Luc Saillard
 - Chris Smowton
 - Rafaël Carré
 - Rémi Duraffort
 - Rui Zhang
 - Eric Petit
 - Ilkka Ollakka
 - Jean-Baptiste Kempf
 - Jeroen Ost

X-Patched-By: Sergey Bolshakov <sbolshakov@altlinux.org>

This reverts commit 4f71d0bb.

Fix include/vlc_atomic.h:124: undefined reference to `___sync_bool_compare_and_swap_4' etc
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

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

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

It seems the volume is expected to be negative. At least, WinCE MSDN
and MingW and the Internets say so. Windows MSDN says the opposite...

This is as already done a long time ago for PulseAudio. VLC might
deliver audio a long time ahead, but there is no warranty thereof.
So a small latency must be assumed for the worst case.

libmad Decoder is necessary for that.