On Mon, Jun 13, 2016 at 12:23:07PM +0200, Hendrik Leppkes wrote: > On Mon, Jun 13, 2016 at 11:51 AM, Matthieu Bouron > <matthieu.bou...@gmail.com> wrote: > > On Fri, Jun 10, 2016 at 03:08:48PM +0200, Matthieu Bouron wrote: > >> From: Matthieu Bouron <matthieu.bou...@stupeflix.com> > >> > >> Fixes playback of HLS streams on MediaTek devices which requires PPS/SPS > >> to be set in their respective csd-{0,1} buffers. > >> --- > >> > >> Hello, > >> > >> The attached patch fixes playback of HLS streams on MediaTek devices which > >> requires PPS/SPS to be set in their respetive csd-{0,1} buffers (instead of > >> having sps+pps in the csd-0 which works on other devices). > >> > >> I'm not sure if I can use the ff_h264_decode_extradata this way (or at > >> least > >> initialize the H264Context with zeroes minus the avctx field). > > > > Rebased patch (after the h264 ps merged) attached. > > > > I still have the same question, is my use of > > H264Context + ff_h264_decode_extradata correct ? > > > > Using H264 decoder internals seems to be a rather unfortunate > solution, as its prone to breakage, often subtle, as the h264 decoder > gets changed and not all inter-module dependencies are known. > So if possible at all, not using something that uses H264Context for > example would be nice. > > For the record, ff_h264_decode_extradata is scheduled for refactoring > to make it independent of H264Context so it can be more easily shared > with the h264 decoder and the h264 parser. > Once that is done, it may give you a cleaner interface to use it from > mediacodec as well.
Ok. I can wait for the refactor to be merged but the MediaCodec decoder will remain broken on those devices. I'm not too happy about that if a release is to be made in those following days. Do we have an ETA ? I'm also not too happy to write the same parsing code as I did before for the AVCC format to split/extract the PPS/SPS. Or ..., I can push this code (if its use is valid) and update it when the merge lands (I'm helping Clément with the merges, so I will take care about this part). Matthieu [...] _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel