On Thu, Oct 30, 2014 at 4:45 PM, John Stebbins <[email protected]> wrote:
> On 10/30/2014 08:20 AM, Hendrik Leppkes wrote: > > On Thu, Oct 30, 2014 at 4:05 PM, John Stebbins <[email protected]> > > wrote: > > > >> SPS_EXT can come before the PPS and results in incomplete extradata. > >> > > If you want this to be really complete, may also want to add SEI (6) and > > subset SPS (15), I had some sort of crazy file where this was required to > > get the SPS/PPS into extradata, but I lost it. > > > > > > Does it make sense to drop all these NAL checks and just add another flag > has_pps that gets set when PPS is seen? Or is > this already known to cause problems? It would be a performance hit if > there is no PPS in the packet, but split isn't > called often. > > The whole idea is to stop working when it finds something that absolutely means there isn't going to be any metadata anymore, like, a slice. So the check does serve some purpose to avoid parsing through the entire frame, just needs to be a bit more relaxed and allow a few more metadata NALs. Additionally to stoping parsing, it also ensures a slice isn't copied into extradata in some crazy bitstream condition. - Hendrik _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
