[FFmpeg-user] ffmpeg3.4 regression

[Piotr Oniszczuk]

> 
> VDPAU is fully supported via the hwaccel API and is usable in the ffmpeg program via the -hwaccel option.  The (deprecated for years) standalone VDPAU decoder which you are trying to use here was removed from master with the major version bump leading to ffmpeg 4.  It is plausible that it was broken before that point (because it never really fit as a decoder and I doubt it ever got much testing once the hwaccel was introduced), but you will likely have a hard time getting anyone to care about such a long-deprecated feature which doesn’t exist any more on master.
> 

Mark,

Probably we have a bit misunderstanding here…

My issue with h_t_t_p://warped.inet2.org/sample3.mkv is fact mythtv playback issue causes by ffmpeg3.4.1 regression (as recently mythtv upgrade build-in ffmpeg from 3.2.1 to 3.4.1)
IIRC mythtv uses hwaccel API for HW decoded vdpau playback. 

In discussion here I’m using ffplay to decouple ffmpeg from mythtv and demonstrate that root cause of failed playback is 3.4.1 regression - not something else (i.e. mythtv).

If You have better idea to address this regression - I’m open to any suggestion.

Currently we have proof that most recent ffmpeg release (3.4.2) has regression (as 3.2.1 plays perfectly).

Is this enough to look on problem?