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.

The transform video_filter mixed up horizontal and vertical flip. They
were swapped. This patch corrects that.

Some AdaptationSet tags may have no attributes at all. Those should be
found in child nodes (most likely the <Representation> tags)