GitLab

In many cases we don't know the value so it's better to use UNDEF.

A boolean also prevents from overwriting a value only when it's undefined.

The es_out display now separates the color space and the color range. One may
be known without the other.

Replaces the timestamp FIFO.

update the ffmpeg hash to use the latest API changes of dav1d

Samples here: http://download.opencontent.netflix.com/?prefix=AV1/Chimera/AVIF/

There is no point waiting when resetting pictures. The picture buffers
can outlive the video format being reset. We still to wait for buffers
at termination because they depend on the event queue.

Since there is something to free.

Since there is something to free.

Since there is something to free.

Since there is something to free.

Due to private members, manually allocating a picture has not been
possible for a while. So we can assume that all picture_t are allocated
with either picture_NewFromResource() or picture_NewFromFormat().
And we can then just factor the final free() call.

Fixes playback of files created with lavf 58.12. They contain no keyframe,
even though all frames are keyframes for MJPEG.

Only some common values exist in VLC so we map what we can.

And debug the value read.

And debug the value read from the file.

Signed-off-by: Steve Lhomme <robux4@ycbcr.xyz>

This poorly named flag was meant to disable double buffering. I doubt
it ever worked. But if it did, that was before the video output rework
nine years ago.

Providing a pool of one picture would make the core copy manually
before the prepare callback. There are no differences to copying
manually in the display callback, except for exposing the picture too
early in the frame buffer. Just merge the two modes.

Do not vlc_close(0) or whatever sys->fd was.
I want to keep my standard input, thank you very much.