Multi-keyboard Modifier Use Support

[641i130]

Hyprland Version

v0.33.1

Bug or Regression?

Bug

Description

Using modifiers on one keyboard and pressing keys on a different keyboard does not result in expected behavior.

i.e
SHIFT on keyboard A -> 8 on keyboard B
results in 8
when it should result in *

This behavior is not in sway or i3 for reference. I have a weird keyboard setup for accessibility reasons, this shouldn't be an issue.

How to reproduce

Plug in a different keyboard, and try to use the modifiers on one.

Crash reports, logs, images, videos

None.

[vaxerski]

and I assume it still happens in 0.48?

[rsammelson]

I believe I am having an issue with the same root cause, which is that the numlock on my numpad (separate from my keyboard) is reset whenever I press a key on the keyboard. The light on the numpad does not change, but keys start sending as KP_Left, KP_Right, etc instead of KP_1, KP_2, etc. In order to send numbers again I have to disable and then re-enable numlock on the numpad. This is on Hyprland 0.44.1.

My unsubstantiated thought is that it might have something to do with a keymap event being sent whenever I switch which keyboard I am using.

[rabiddoughnuts]

I have that same numlock issue, from what I can tell it is that when you type on the main keyboard, the apps for some reason associate the numlock state with that keyboard, and not the split numpad, like whats on the framework, so when you go back to the numpad it doesnt work

[khanhn201]

Same problem here, weirdly only with one of the recent updates.
Edit: Only happen on Wayland apps. XWayland works fine.

[khanhn201]

It was due to some logic around this line.

Hyprland/src/managers/input/InputManager.cpp

Line 1492 in 967c3c7

if (USEIME || !HASIME) {

It based on whether you are using IME or not, which also explains why in my case only Wayland apps is affected (my IME works fine on both Wayland and Xwayland, so the if check is just not passing for Wayland for some reasons).
Anyway, you can replace the if statement with true for a quick and dirty fix and rebuild it locally. For my use case, this works fine.
I don't understand the IME use case in the code enough to know what is the proper fix. It might be helpful to have this behavior configurable in the config file.
Edit: FYI I'm using fcitx5

[xzn]

A possible culprit may be this line:

Hyprland/src/protocols/InputMethodV2.cpp

Line 53 in b5a2ef7

sendMods(keyboard->m_modifiersState.depressed, keyboard->m_modifiersState.latched, keyboard->m_modifiersState.locked, keyboard->m_modifiersState.group);

It set the modifier state for ime from the current keyboard without the merged state.

I'll look at what's the best? way to go about this.

[xzn]

I tried connecting two separate physical keyboards and could not reproduce the issue.. I wonder what's the difference ?

[khanhn201]

Maybe the fcitx5 setup? It is also only in Wayland apps, not in XWayland. Also I'm on 0.52.1

env = QT_IM_MODULE,fcitx
env = XMODIFIERS,@im=fcitx

[khanhn201]

if (USEIME || !HASIME)
is false for Wayland app, but true for XWayland. Maybe IME detection not working properly?

[khanhn201]

Nvm the problem is not in main (b5a2ef7). Some commits between v0.52.1 and main must have fixed it. Thanks for looking into it.