[FFmpeg-devel] [PATCH] avcodec: only warn about hwaccel with frame threads

[wm4]

On Sat, 6 Feb 2016 18:46:59 +0100
Andreas Cadhalpun <andreas.cadhalpun at googlemail.com> wrote:

> On 06.02.2016 18:25, wm4 wrote:
> > On Sat, 6 Feb 2016 18:15:37 +0100
> > Andreas Cadhalpun <andreas.cadhalpun at googlemail.com> wrote:
> >   
> >> On 06.02.2016 18:02, wm4 wrote:  
> >>> On Sat, 6 Feb 2016 17:48:17 +0100
> >>> Andreas Cadhalpun <andreas.cadhalpun at googlemail.com> wrote:  
> >>>> This has nothing to do with my convenience. It is about fixing an
> >>>> important regression caused by a commit you reviewed.    
> >>>
> >>> It's not a regression in FFmpeg or its API.    
> >>
> >> That's splitting hairs.
> >> VLC's hwaccels worked before that commit and don't work after it,
> >> which is a clear regression.  
> > 
> > A regression in VLC. Make them fix it.  
> 
> You're not constructive.

Even if that is true, neither are you. You want to revert a change that
was often discussed and found to be the best way to go about the issue,
just because it's the way of least resistance for you.

While we normally would very much try not to break API users, including
VLC, this case has a bit of a history. Don't forget the behavior of the
VLC dev, who was never constructive about this either. On the Libav
side he AFAIK kept claiming it was not his problem. When the change was
made on the FFmpeg side, he just changed the VLC configure script to
reject FFmpeg, instead of bringing up this issue anywhere.

> I doubt you'd like being told that when a
> change in ffmpeg broke mpv in a similar way and couldn't easily be
> fixed there.

I doubt I could get a debian maintainer to waste the time of FFmpeg
developers just because I don't want to adapt in a specific case.

> >> All of that has been said before.  
> > 
> > Indeed.  
> 
> Thus nothing changed and I still intend to apply these patches soon.

No, you won't.