> On Jun 17, 2016, at 10:36 AM, Clément Bœsch <u...@pkh.me> wrote: > > On Fri, Jun 17, 2016 at 09:43:09AM -0400, Richard Kern wrote: >> Since ca2f19b9cc (h264: switch to h2645_parse for NAL parsing) the hwaccel >> wouldn’t init (when the first h.264 frame is decoded). It was the same on OS >> X and iOS - creating the decoder would return a “decoder malfunction” error. >> Apple’s suggested way to create the decoder config for h.264 decoders is >> CMVideoFormatDescriptionCreateFromH264ParameterSets(), and switching to this >> function solved the issue on OS X and iOS. A user on IRC confirmed it worked >> for them too. >> > > So avc_info constructed from extradata in decoder_config_create() wasn't > enough? (Any idea why?)
I don’t know why. Unfortunately there isn’t much documentation, and the error message is vague. > > "the hwaccel wouldn't init", why? Is it because the SPS/PPS NALs were > somehow split out of the packets since ca2f19b9cc and thus VT never got > them, or I misunderstand? The same SPS and PPS from the H264Context are being used to create the avcC atom as are passed to CMVideoFormatDescriptionCreateFromH264ParameterSets()… I’ll need to do some digging in the commit that broke it this weekend. > > -- > Clément B. _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel