Improve Windows raw input thread handling by whrvt · Pull Request #15239 · libsdl-org/SDL

whrvt · 2026-03-20T13:41:11Z

Description

This is a followup to commits 43e90c7 and b9c227e.

The thread is still created once and only shut down on WIN_VideoQuit,
but the thread will unregister the devices when they're disabled to avoid
wasting time processing raw input that won't be appreciated by anyone.

This also allows RIDEV_NOLEGACY to be set on the devices, otherwise they
swallow up the WM_MOUSE* events that would be going to the main thread
when relative mode is disabled. That's out of scope for this change, though.

Also, fixed up some existing issues in the first 2 commits.

whrvt · 2026-03-20T13:57:07Z

Just a small update to preemptively guard against an unlikely error scenario.

src/video/windows/SDL_windowsrawinput.c

whrvt · 2026-03-20T15:44:08Z

After testing this in my game, I realized that RegisterRawInputDevices itself is actually the culprit for the long delays, not the thread startup/shutdown itself. I'll update this with a slightly modified strategy that should work around that, though.

whrvt · 2026-03-20T16:06:46Z

Updated to avoid declaring variables between goto/label.

src/video/windows/SDL_windowsrawinput.c

slouken · 2026-03-20T20:53:18Z

src/video/windows/SDL_windowsrawinput.c

    return SDL_Unsupported();
 }

+bool WIN_SetRawKeyboardFlag_Inputsink(SDL_VideoDevice *_this, bool enabled)


Did you mean to include this here?

I sent it as a separate commit because I was already touching code near that and it's a minor change. Do you want me to drop it?
I did just notice that I messed up the splitting for the second commit, though, I'll fix that.

No, that's fine, I didn't realize it was just missing.

CreateEvent/CreateThread return NULL on failure.

The thread is still created once and only shut down on WIN_VideoQuit, but the thread will unregister the devices when they're disabled to avoid wasting time processing raw input that won't be appreciated by anyone. This also allows RIDEV_NOLEGACY to be set on the devices, otherwise they swallow up the WM_MOUSE* events that would be going to the main thread when relative mode is disabled. That's out of scope for this change, though.

slouken · 2026-03-23T00:57:09Z

Merged, thanks!

windowsrawinput: Add missing WIN_SetRawKeyboardFlag_Inputsink stub.

efff575

whrvt force-pushed the win-rawinput-register branch from 256413a to fc983e8 Compare March 20, 2026 13:55