On 4/20/2021 7:10 AM, Nicolas George wrote:
Brad Hards (12021-04-20):
The existing function allows access the first instance of a given
type. Mostly that is OK, but some types can occur multiple times
(e.g. libx264 can write version info, VANC and UMID related data as
user data unregistered SEI.
This adds API to access additional instances of a given SEI type.
---
libavutil/frame.c | 19 +++++++++++++++++++
libavutil/frame.h | 20 ++++++++++++++++++++
2 files changed, 39 insertions(+)
IIRC, there was a discussion a long time ago that concluded that
multiple instances of the same side data should not happen.
Somebody remembers it?
I recall long ago i tried to make av_frame_new_side_data() replace the
existing element if it found one, effectively limiting side data to one
per type and copying the behavior of packet side data, but it was
rejected as it was argued more than one per type had some use. And
seeing that now decoders like h264 may attach more than one element of
AV_FRAME_DATA_SEI_UNREGISTERED type as contained inside the bitstream
AUs, it may be a good idea for it to be supported officially with our
accessors (Right now you need to manually traverse the side data array
to find all instances of a given type).
Regards,
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
