More touchpad palm rejection options

[non-bin]

Good evening all! I am a little disappointed with the state of trackpad palm rejection in Linux, so I was hoping we could be the change!

At the moment, HL only has a toggle for input:touchpad:disable_while_typing with no adjustment for the delay.

I would like to suggest a few things in order of doability:

1. enable_firmware_palm_rejection - Listen to ABS_MT_TOOL_TYPE messages from the touchpad if they are supported
2. disable_while_typing_delay - Delay in ms to keep the touchpad disabled after a key press
3. disable_clicks_while_typing or allow_moving_while_typing - So you don't click and move the cursor, but don't have to wait for the delay before starting to move
4. palm_min_width - To only acknowledge touches small enough to reasonably be a finger (like synclient PalmMinWidth=X, explained here

More on ABS_MT_TOOL_TYPE

Running sudo libinput record -o toupad.log you can see what messages are sent by the touchpad, and in my case at least (Framework 16 with NixOS) it reports in the top info section that it supports the method, then each time the tool changes it reports it (the first touch is assumed to be a finger unless specified, key is 0=finger, 1=pen, 2=palm)

toupad.log

This event is described in the multi-touch-protocol documentation here, internal function descibed here

[non-bin]

After some digging, it seems we would need to add this functionality to Aquamarine as Hyprland doesn't interface directly with libinput. I'll put together some pseudocode of what would likely need to be changed, but I would love some feedback on which way this should be done (I'm leaning towards option 1, then option 0 once that works):

Option 1: Pass Palm Info Up to Hyprland

Aquamarine attaches a "palm" flag (e.g. isPalm or palm) to each event (move, click, etc). HL receives the flag and decides what to do

Pros

• Flexible
• You can tweak filtering logic in the HL without recompiling aquamarine.
• HL could selectively ignore palm status for some gestures, or even use it as a separate gesture

Cons

• More complex event structs and APIs.
• Requires all compositors to implement palm filtering logic. (I don't know if any others do)

Option 2 Aquamarine Does the Filtering

Aquamarine internally filters out palm events and never passes them up as pointer/touch events, unless explicitly told not to (via a config or runtime API from HL).

Pros

• Simplicity for HL
• DRY principle: If multiple compositors use aquamarine, they all get palm filtering for free.
• Event structs and API stay simpler.

Cons

• Less flexibility: The compositor cannot override or experiment with palm handling.

Option 0 Both

By default, pass palm info up, and optionally add a config to aquamarine to filter at the backend for users who want it.

[non-bin]

Never mind, libinput already filters out palm moves by default, if I try to move the mouse with my palm, nothing happens.

I can click with my palm though which is my issue, but libinput doesn't seem to pass on tool info on taps.

I think the allow_moving_while_typing option is best, I'll look into that now

[m-bartlett]

Something like allow_moving_while_typing and disable_while_typing_delay would be huge quality quality of life improvements, I'm constantly accidentally clicking while typing.

[non-bin]

This might be better as an option in libinput so I'll suggest it there too

I've never used c++, only c so sorry if this is wildly wrong but might this work?

diff --git a/src/managers/input/InputManager.cpp b/src/managers/input/InputManager.cpp
index ea16484..9629958 100644
--- a/src/managers/input/InputManager.cpp
+++ b/src/managers/input/InputManager.cpp
@@ -614,10 +614,15 @@ void CInputManager::onMouseButton(IPointer::SButtonEvent e) {
     EMIT_HOOK_EVENT_CANCELLABLE("mouseButton", e);
 
     if (e.mouse)
         recheckMouseWarpOnMouseInput();
 
+    // If the user is typing, ignore the click
+    if (e.device && e.device->m_isTouchpad && Time::millis(Time::steadyNow()) - g_pKeybindManager->m_timeLastMs <= 200) {
+        return
+    }
+
     m_lastCursorMovement.reset();
 
     if (e.state == WL_POINTER_BUTTON_STATE_PRESSED) {
         m_currentlyHeldButtons.push_back(e.button);
     } else {

[vaxerski]

isn't dwt already a thing in the config?

[non-bin]

it ignores all movement, not just taps. If palm rejection works then movment isn't a problem, but it doesn't reject palm taps.

[Christopher2K]

I need this so bad