[Windows] Report raw absolute mouse input as pen events when SDL_HINT_PEN_MOUSE_EVENTS is disabled

[Susko3]

Description

Raw absolute mouse input is currently understood as being tablet (pen) input and is converted to relative mouse input events. Raw mouse input is only reported when relative mouse mode is enabled. With this change, and when SDL_HINT_PEN_MOUSE_EVENTS is disabled, this input will be reported as pen motion events.

Some caveats:

• this leaks 1 byte of memory every time relative mode is (re-)enabled and a pen comes in proximity
  • allocating a handle is required by the pen subsystem for some reason
  • fixing this in a nice way would require exposing SDL_Pen * from SDL_pen.c
  • we could also store pointer to this memory in SDL_VideoData but I find it silly, and the resulting code is hard to read
• the fake pen will leave proximity only when relative mode is disabled, i.e. it'll be stuck in proximity
  • unsure if detecting proximity is even possible from raw mouse input
• the reported pen position is not bounded and can be outside the window bounds

Testing

I can confirm this fixes the original issue. Tested on Windows 11 24H2 with OpenTabletDriver and a Wacom CTL-6100WL.

Existing Issue(s)

• Closes Windows tablets don't report pen events when relative mouse mode is enabled #12324

[Joehuu]

Functionally-wise, LGTM.

On osu!-side though, as said on Discord, cursor sensitivity doesn't work for tablets anymore (compared to SDL2). Can be done osu!-side.

[slouken]

Can you split these into separate PRs? I want to merge the second commit, but the first one needs more review from @icculus

[Susko3]

The second commit only touches code added in the first, it's not possible to split it out.

[slouken]

Okay, never mind, we'll wait on @icculus then.

[icculus]

I'm fine with this, but there's a conflicting line...

        (SDL_TouchDevicesAvailable() && (rawmouse->ulExtraInformation & 0x80) == 0x80)) {

...@Susko3, can you make sure that won't mess up your patch (like, rawmouse->ulExtraInformation in these cases isn't going to cause this to throw away these events in this scenario)?

If that's going to be okay, I'll resolve the conflict and merge.

[Susko3]

I'm fine with this, but there's a conflicting line...

        (SDL_TouchDevicesAvailable() && (rawmouse->ulExtraInformation & 0x80) == 0x80)) {

...@Susko3, can you make sure that won't mess up your patch (like, rawmouse->ulExtraInformation in these cases isn't going to cause this to throw away these events in this scenario)?

If that's going to be okay, I'll resolve the conflict and merge.

This should be fine, as the ulExtraInformation will fail:

the eighth bit, masked by 0x80, is used to differentiate touch input from pen input (0 = pen, 1 = touch).

https://learn.microsoft.com/en-us/windows/win32/tablet/system-events-and-mouse-messages#distinguishing-pen-input-from-mouse-and-touch

[icculus]

Okay, we're rebased, the memory leak is resolved...if the builders turn green, we merge!

[slouken]

Did you check this with RDP? I thought mouse motion was sent as raw absolute in that case.

[icculus]

I thought mouse motion was sent as raw absolute in that case.

I had no idea...I'll try that out real fast.

[icculus]

(Also my rebase broke this, stand by on that, too.)

[Susko3]

Remote desktop should be unaffected, as I didn't touch the remote_desktop == true path.

[slouken]

Remote desktop should be unaffected, as I didn't touch the remote_desktop == true path.

Ah, good point.

[icculus]

Okay, so...merge?