On Thu Aug 31 18:29:16 2023 +0000, Matteo Bruni wrote:
> That's 100% it :rolling_eyes:
> But yeah, that's what I mean when I say that d3d QXL test results ought
> to be ignored.
That sounds like a bug in QXL. Is there a QEmu bug report for this issue?
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/3694#note_45290
Hi Ziqing, I've been doing some research, and I believe patch 2/6 (committed as 320383c594c4) is incorrect. GStreamer's "codec-data" field is an AVCDecoderConfigurationRecord, as defined by ISO/IEC 14496-15 (which you can see at [1]). MF_MT_MPEG_SEQUENCE_HEADER, on the other hand, seems to contain the raw SPS/PPS NALUs, start codes and all.
Since Microsoft uses Annex B everywhere, it shouldn't need the codec-data anyway, so I'm curious now why this was added—was there a need for it, or did you just add it on the assumption that the two attributes mapped to each other?
[1] https://doc-kurento.readthedocs.io/en/latest/knowledge/h264.html#gstreamer-…
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/3636#note_45288