libvlc_video_take_snapshot API change (it would be nice to grep the
VLC source, or at least the bindings directory, for the symbol when
doing such a change).

MacOSX/Framework/VLCMediaPlayer.m: Missing a comment before the detach to match the attach of libvlc_MediaInstancePositionChanged.

Start splitting the ALSA and OSS code. The --v4l2-adev device will be tried first as an ALSA device, and then as OSS upon failure. If no audio device is specified, "default" is used for ALSA, "/dev/dsp" for OSS. You can disable/activate usage of ALSA and OSS using the --v4l2-audio-method option (see vlc -p v4l2 --advanced --help-verbose).
Next step will be to split the module in different source files, the current code is quite a mess.
Also fix compilation of the module if HAVE_ALSA isn't defined.

video_output/opengllayer.m: Don't use AGP texturing until we get a proper frames lifo stack. Set to the layer synchronous, and call setNeedsDisplay when needed.

video_output/opengllayer.m: Rename the layer from VLCVideoLayer to VLCVoutLayer, to avoid name collision with the framework.

MacOSX/Framework: Properly convert libvlc Exception to NSException. Rename quit_on_exception to catch_exception.

Limit the mouse-x and mouse-y beetween 0 and p_vout->fmt_in.i_visible_(width\|height) to avoid unsigned overflow. Feel free to revert if this commit is stupid.

Revert [19071] because it is broken. It doesn't allow channel or samplerate changes for fourcc mp4a even when explicitly requested by the user. Specify channels=2 and samplerate=NNNN for fourcc mp4a if you want AAC SBR.

Use id->p_decoder->fmt_out as input for id->p_encoder->fmt_in. Removes double initialisation of id->p_encoder->fmt_in (es_format_Init()).