[Linux client + MacOS server] Sticky modifier keys

[kiwiz]

Client (IO source): KDE/Wayland, Ubuntu, commit: 3922b45
Server: MacOS, commit: 640fa99

When performing key combinations (ex: meta+tab), the modifier key appears to get stuck down. Hitting the key again resets the state. FWIW, MacOS -> MacOS works perfectly fine.

[kiwiz]

Mac -> Mac

+ [2026-01-12T23:35:37Z WARN  input_emulation::macos] Event::Keyboard Keyboard(Key { time: 0, key: 125, state: 1 })
+ [2026-01-12T23:35:37Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(0x0) } 125
+ [2026-01-12T23:35:37Z WARN  input_emulation::macos] key event: 55 1
- [2026-01-12T23:35:37Z WARN  input_emulation::macos] modifiers updated: XMods(Mod4Mask)
...

+ [2026-01-12T23:35:39Z WARN  input_emulation::macos] Event::Keyboard Keyboard(Key { time: 0, key: 15, state: 1 })
+ [2026-01-12T23:35:39Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(Mod4Mask) } 15
+ [2026-01-12T23:35:39Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(Mod4Mask) } 55
+ [2026-01-12T23:35:39Z WARN  input_emulation::macos] key event: 55 0
+ [2026-01-12T23:35:39Z WARN  input_emulation::macos] key event: 48 1

+ [2026-01-12T23:35:39Z WARN  input_emulation::macos] Event::Keyboard Keyboard(Key { time: 0, key: 15, state: 0 })
+ [2026-01-12T23:35:39Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(Mod4Mask) } 15
+ [2026-01-12T23:35:39Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(Mod4Mask) } 48
+ [2026-01-12T23:35:39Z WARN  input_emulation::macos] key event: 48 0

+ [2026-01-12T23:35:39Z WARN  input_emulation::macos] Event::Keyboard Keyboard(Key { time: 0, key: 125, state: 0 })
+ [2026-01-12T23:35:39Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(Mod4Mask) } 125
- [2026-01-12T23:35:39Z WARN  input_emulation::macos] modifiers updated: XMods(0x0)

Linux -> Mac

+ [2026-01-12T23:39:41Z WARN  input_emulation::macos] Event::Keyboard Keyboard(Key { time: 1186694665, key: 125, state: 1 })
+ [2026-01-12T23:39:41Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(0x0) } 125
+ [2026-01-12T23:39:41Z WARN  input_emulation::macos] key event: 55 1
...

+ [2026-01-12T23:39:43Z WARN  input_emulation::macos] Event::Keyboard Keyboard(Key { time: 1186696549, key: 15, state: 1 })
+ [2026-01-12T23:39:43Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(Mod4Mask) } 15
+ [2026-01-12T23:39:43Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(Mod4Mask) } 55
+ [2026-01-12T23:39:43Z WARN  input_emulation::macos] key event: 55 0
+ [2026-01-12T23:39:43Z WARN  input_emulation::macos] key event: 48 1

+ [2026-01-12T23:39:44Z WARN  input_emulation::macos] Event::Keyboard Keyboard(Key { time: 1186697887, key: 15, state: 0 })
+ [2026-01-12T23:39:44Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(Mod4Mask) } 15
+ [2026-01-12T23:39:44Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(Mod4Mask) } 48
+ [2026-01-12T23:39:44Z WARN  input_emulation::macos] key event: 48 0

+ [2026-01-12T23:39:45Z WARN  input_emulation::macos] Event::Keyboard Keyboard(Key { time: 1186698053, key: 125, state: 0 })
+ [2026-01-12T23:39:45Z WARN  input_emulation::macos] update_modifiers Cell { value: XMods(Mod4Mask) } 125

It looks like modifier_event isn't being called in certain cases.

[feschber]

which input-capture backend is used on linux?

[kiwiz]

Wayland (KDE)

[kiwiz]

Ah sorry, it's libei

[kenichi]

Thanks for a great tool! I've been having this exact problem, except Gnome/Wayland, still libei. So far, this branch is working for me:

https://github.com/kenichi/lan-mouse/tree/linux_to_macos_fix

can test with cargo install --git https://github.com/kenichi/lan-mouse.git --branch linux_to_macos_fix

[feschber]

@kenichi are you open to making a PR?

[kenichi]

@feschber yes, no problem. though I'm not sure about affects to other OS combos 🤷🏽‍♂️

[feschber]

@feschber yes, no problem. though I'm not sure about affects to other OS combos 🤷🏽‍♂️

We might get duplicated modifier events but I don't think that should cause any problems.
Eventually I think its best to handle modifier events entirely client side but as a temporary fix this is certainly okay!