This website gittea.dev/yeyinlaoying/moonlight-qt/commits/commit/83ec929758c940fd3453a86a902aa410070b8323/app/streaming/session.cpp is currently offline. Cloudflare\'s Always Online™ shows a snapshot of this web page from the Internet Archive\'s Wayback Machine. To check for the live version, click Refresh.
This exposes all sorts of issues with full-screen transitions on Windows and macOS:
- Loss of proper window decorations when transitioning from full-screen to windowed on Windows
- Improper sizing when transitioning from true full-screen to windowed on macOS
This reverts commit 38396b2691.
Previously we opted to just lie to SDL and tell it that the current
GL context is actually GLES 3.0 when it wasn't. This meant that
we avoided our window being recreated for GLES usage on Wayland
and KMSDRM, but that meant our 10-bit color change didn't apply
either. I suspect this hackery is what led EGLRenderer to get
stuck in a state where SDL_CreateRenderer() always failed.
Now SDL will recreate our window once to configure a compatible GL
API for our renderer. If that fails, we'll move on to a different
renderer (which may recreate the window again).
Skipping size changed events can cause EGLRenderer to get stuck in
a state where it can't reset itself properly. This seems to happen
consistently when starting a maximized stream then transitioning
to full-screen.
This reverts commit e1b3f90f8b.
This allows us to keep asynchronous decoders like MMAL and V4L2M2M fed
while we're waiting for output frames. Behavior for synchronous decoders
should be identical.
Continuing to feed new data while waiting for output frames is crucial for
acceptable performance on 1080p video on the Raspberry Pi using V4L2M2M,
since it allows the decode and copy operations to be pipelined.
- HEVC forced with no host or client hw support
- H.264 forced with no client hw support
- Client has hw HEVC only and host only supports H.264
- Client has no hw decode at all
Cameron Gutman