On Tue Jul 8 20:42:25 2025 +0000, Elizabeth Figura wrote:
I don't think we should be adding this extra code to use probes in winegstreamer; rather we should be outputting compressed samples from the media source (5988), and then we can just discard any non-I-frames at the mfplat level.
what exactly does "at the mfplat level" mean here? the media source itself is responsible for dropping delta frames when thinned.
I agree that this feature would be much easier to implement with a separation between a demuxer source and a decoder MFT (like the winedmo stack does), instead of hacking it into the pipeline of a source that emits uncompressed samples.
Still, this is the best implementation I could come up with for thinning itself currently, I don't see an issue with doing this for now and then removing it when we have a better approach (iirc winedmo is supposed to eventually replace wg_parser anyway?). or perhaps it was a mistake to try to upstream this now - the winedmo/mfsrcsnk parts are needed for a proton game, so I suppose just putting those bits on proton for now wouldn't really affect anything that happens with winegstreamer later.