[FFmpeg-devel] h264 PPS processing oddity

[Andreas Rheinhardt]

Alex Agranovsky:
> Hello,
> 
> 
> I've been chasing the problem with VideoToolbox failures to decode on
> iOS and Apple Silicone described in https://trac.ffmpeg.org/ticket/9713
> 
> As I've mentioned in the ticket, the problem seems to boil down to the
> last byte of PPS "disappearing" by the time we initialize VideoToolbox
> decoder. For some reason, Apple doesn't seem to mind on Intel, but fails
> on iOS/M1.
> 
> At this point I find myself in the area where I'm not 100% familiar with
> the code -- can't be sure whether the last byte of PPS isn't extracted
> and stored by design, or is it a bug.
> 
> Any chance I could get some assistance looking at this? I'm BCC'ing a
> couple of people who recently contributed to the relevant modules; hope
> it's not too much of an imposition.
> 
> Thanks in advance!
> 
> - Alex
> 

This patchset:
http://ffmpeg.org/pipermail/ffmpeg-devel/2022-May/296244.html (in
particular 6/6) look like they are intended to fix the issue you reported.

- Andreas