28 Nov
2025
28 Nov
'25
6:35 p.m.
But shouldn't it already be at a keyframe due to being in thinned mode before?
Not necessarily from the demuxer point of view? I think we read samples from any stream then dispatch the sample to the right stream. If we are disabling thinning mode, we perhaps just read a keyframe for some stream but the next sample read might be for a different stream and might not be a keyframe itself, and might references samples before it that we didn't send to the decoder. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/8505#note_124034