Some tests reuse the same MP4 stream and are currently passing only because we have the generic byte stream handler fallback. They pass on Windows too, so I'm assuming the stream is seeked before looking for the hints. We could also very well remove the position restore because the media sources are rewinding the stream later too, but I've kept it like it was.
--
v2: mfplat: Seek byte stream to the start for URL hint detection.
mfplat/tests: Test source resolver bytestream interactions.
https://gitlab.winehq.org/wine/wine/-/merge_requests/6373
--
v3: win32u: Use the current display mode as monitor rect.
win32u: Keep track of the display source physical display mode.
win32u: Add virtual modes when drivers report a single display mode.
win32u: Remove unnecessary GetCurrentDisplaySettings call.
win32u: Cache display source current display settings.
win32u: Remove now unnecessary GetDisplayDepth driver entry.
win32u: Keep the source depth separately from the current mode.
win32u: Always write the source current mode to the registry.
win32u: Make sure to load drivers when updating the display cache.
https://gitlab.winehq.org/wine/wine/-/merge_requests/6573