[FFmpeg-devel] [PATCH 8/9] ffplay: move video filtering graph in the VideoContext

Stefano Sabatini stefasab at gmail.com
Mon Jun 25 21:03:22 CEST 2012

On date Saturday 2012-06-23 12:47:46 +0200, Marton Balint encoded:
> On Fri, 22 Jun 2012, Stefano Sabatini wrote:
> >Make room for some factorization, improve consistency.
> I'm actually not a big fan of moving stuff into VideoState. Graph is
> a video thread local variable. Also allocating and freeing it in the
> same function is probably better than hiding the allocation inside
> configure video filters. If we use the videoState variables
> everywhere it will make to code less reusable.
> So I'd rather not change this only to make it consistent with audio.
> Maybe making audio consistent with this should be the goal. Or is
> there a place where this version has some real advantage?

The video graph context is only accessed inside the video_thread(),
where it is also destroyed. On the other hand the audio context can be
handled inside a single function (there is nothing like video_thread()
for audio), so we need to reference it in stream_component_close(), so
we need to access it from the global VideoState (that should be rather
named PlayerContext), and it looked weird to have an agraph but not a
graph (for a simple question of simmetry).

But I'm fine with the latter if you prefer that.
FFmpeg = Foolish Fabulous Mastodontic Puritan Enigmatic God

More information about the ffmpeg-devel mailing list