May 27, 2026
5:29 p.m.
Sorry about the delay, again >_> That explanation makes a lot of sense, thanks. The one thing I notice that worries me is that we're setting the base pts as the minimum, but we don't actually necessarily wait for a buffer from every stream (basically, we only actually wait for caps and tags, and rely on a buffer to tell us we won't get any more), so we could end up returning a sample and then base_pts gets updated later. I think the right solution here is to go ahead and make sure every stream has a buffer before we consider initialization done. For regression testing purposes I would highly recommend putting this in its own commit. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10864#note_141447