There still are some issues in the C side of this filter:
You're allocating a lot of stuff on init (integral_img, ...) but never free it. So the filter leaks those.
You define a list of supported formats at the top, listing practically all formats that can be in a CUDA frame as of right now. But then later infilter_frame you have a switch-case that returns AVERROR_BUG if it's anything else than NV12. So, either add support for all the other formats or don't claim support for more than NV12.
I'm also not sure if not taking an internal reference to the hw_frames_ctx is valid.
It might, but I'm not sure about the lifetime on the one on the output. Typically, filters also hold an internal reference and unref it on uninit. Generally:There's a bunch of trailing whitespaces everywhere, please tell your editor to not do that.
Also, when bumping minor version, the patch version drops back to 100.
smime.p7s
Description: S/MIME Cryptographic Signature
_______________________________________________ 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".
