It fixes a segfault when the directx/direct3d device is lost at the wrong time
or cannot be restored soon enough (close #3647).

There is a little difference in Direct3DLockSurface as it now also
sets the height field, but it should have no functionnal changes.

IDirectDrawSurface2_UpdateOverlay may fails, in which case vlc should simply
retry later.

This reverts commit 7c93516a.
It is useless as 6adc3961 added it already.

Do not try to delete a callback and/or a mutex that do not exist.
The goto error paths still look a bit suspicious, someone should
double check them.

This doesn't make it usable though, but noone seems to care anyway...
WARNING: this is a HUGE hack. It will kill yourself and your kittens

This was fixed after 1.1.0-RC.

fullscreen is now managed (activation and deactivation) in skins2
 exactly as it is done in qt4.
    - activation when a VOUT_WINDOW_SETFULLSCREEN is received
    - deactivation when (and only when) a vout_window release is received.

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

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

WId is almost always an integer (except for Symbian)
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

If 0.8.8 is detected and --enable-mod is passed, it will error,
if --disable-mod is passed, nothing will happen
else it will autodetect. This is mostly like all other PKG_ENABLE_MODULES_VLC
detections

This was annoying because if you had a modplug.pc and didn't --enable-mod, it was auto-detected and then failed.

Should fix #3767

Close #3887

plughw provides transparent userspace format conversion if needed.
That is much safer. In fact, I doubt VLC would support all raw hardware
drivers - some have rather weird format requirements.