[vlc-devel] [PATCH] [RFC] apply the same flush on the sout than on aout/vout

[Steve Lhomme]

On Tue, Nov 10, 2015 at 5:05 PM, Rémi Denis-Courmont <remi at remlab.net> wrote:
> Le 2015-11-10 18:47, Steve Lhomme a écrit :
>>
>> When seeking during a sout session, we send discontinuities to the
>> receiver(s),
>> resulting in apps possibly stopping because of incoherent data.
>
>
> At this point, I'd rather add a flush callback to stream output plugins,
> than continue (ab)using block flags.

I agree, that would be cleaner. I'll see if it's not too big of a change.

>> There remains the question of pre-roll when seeking. Should we send it to
>> the
>> receiver(s) or not ?
>
>
> I may be wrong, but I would tihnk that the decoder underneath transcode
> needs pre-roll blocks (just like the normal decoder). So they have to be
> sent.

Right now, not sending them to the Chromecast results in distorted
display until the next keyframe. This is better than aborting playback
for unknown errors (before the flush). The Chromecast seem to be smart
about the seeking though, it waits until the actual timecode specified
in the seek command (always in the future) to display the frames. So
there might be a chance it works when sending the preroll.

Right now I have a frame from before the seek that is left in there
after the flush. Once I fix this I will know for sure. And since this
was working by chance before (and not always), I suggest we take an
approach that works well with the Chromecast.

> --
> Rémi Denis-Courmont
> http://www.remlab.net/
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel