Quoting James Almer (2020-06-01 22:55:12) > Also, i've noticed the way the buffer pools in some of the hwcontext > implementations are used is very wtf, being both extremely fragile/hacky > and depending on undocumented behavior. I didn't test, but i suspect > that configuring with --enable-memory-poisoning will break a few of > them, like d3d11, as they depend on the buffer returned by the pool to > be in the exact same state as it was last time it was returned to the > pool, which is not how the pools are meant to work. But i also don't > think this patch here would help much in that regard.
As I see it, that problem is rooted exactly in using AVBuffer as something it's not supposed to be. But changing that at this point would be very invasive. -- Anton Khirnov _______________________________________________ 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".
