[FFmpeg-devel] [PATCH 06/17] lavfi: add ff_inlink_process_timeline().

[Nicolas George]

Le decadi 10 nivôse, an CCXXV, Clement Boesch a écrit :
> You want a watermark to be effective only in a timerange, the second input
> (the watermark) would get ignored outside that range, and the first one
> pass through. This would be filter specific (FLAG_SUPPORT_TIMELINE_INTERNAL).
> 
> That's how overlay and dualinput actually works with a given outlink,
> doesn't it?

You are right, I missed that when replying to Michael. It works because
the inputs are clearly defined as primary and secondary. I believe I did
not break it with the current series; if there is a FATE test then I am
sure I did not.

Still, the current code is fragile: since the enable expression is
evaluated for both inputs at the time the frames arrive and are queued
in the dualinput/framesync context, it may cause slightly incorrect
results: the timestamp for enable/disable would be taken from the
incoming frame that allows to flush the current frame.

Still, I think fixing that is orthogonal to the current patch series.

Regards,

-- 
  Nicolas George
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20161230/393530f3/attachment.sig>