Keypresses are repeated as if key is stuck, ignores modifier

[Julian Andres Klode]

I switch into Mattermost with Windows+3 key combination, this causes an endless stream of 3 to be written into the mattermost field until I press another key (similarly for any other number key, I did not try non-numbers)

If we look at the keyboard input device with https://github.com/DIGImend/evdev-dump we can see that the key is not stuck, all the input events are correct:

/dev/input/event4 1725622057.447632 EV_MSC MSC_SCAN 0x000000DB
/dev/input/event4 1725622057.447632 EV_KEY KEY_LEFTMETA 0x00000002
^ press windows key
/dev/input/event4 1725622057.447632 EV_SYN SYN_REPORT 0x00000000
/dev/input/event4 1725622057.477343 EV_MSC MSC_SCAN 0x00000004
/dev/input/event4 1725622057.477343 EV_KEY KEY_3 0x00000001
^ press key 3
/dev/input/event4 1725622057.477343 EV_SYN SYN_REPORT 0x00000000
/dev/input/event4 1725622057.627896 EV_MSC MSC_SCAN 0x00000004
/dev/input/event4 1725622057.627896 EV_KEY KEY_3 0x00000000
^ release key 3
/dev/input/event4 1725622057.627896 EV_SYN SYN_REPORT 0x00000000
/dev/input/event4 1725622057.997495 EV_MSC MSC_SCAN 0x000000DB
/dev/input/event4 1725622057.997495 EV_KEY KEY_LEFTMETA 0x00000000
^ release windows key (now it starts endlessly dumping 3 into mattermost)

/dev/input/event4 1725622057.997495 EV_SYN SYN_REPORT 0x00000000
/dev/input/event4 1725622059.869769 EV_MSC MSC_SCAN 0x000000DB
/dev/input/event4 1725622059.869769 EV_KEY KEY_LEFTMETA 0x00000001
^ press windows key again
/dev/input/event4 1725622059.869769 EV_SYN SYN_REPORT 0x00000000
/dev/input/event4 1725622059.919605 EV_MSC MSC_SCAN 0x00000009
/dev/input/event4 1725622059.919605 EV_KEY KEY_8 0x00000001
^ press 8 to switch to the terminal window

This only affects Xwayland, native Wayland clients behave correctly.

ProblemType: Bug
DistroRelease: Ubuntu 24.10
Package: xwayland 2:24.1.2-1
ProcVersionSignature: Ubuntu 6.8.0-31.31-generic 6.8.1
Uname: Linux 6.8.0-31-generic x86_64
NonfreeKernelModules: zfs
ApportVersion: 2.30.0-0ubuntu1
Architecture: amd64
CasperMD5CheckResult: pass
CurrentDesktop: GNOME
Date: Fri Sep 6 13:26:18 2024
InstallationDate: Installed on 2022-11-26 (650 days ago)
InstallationMedia: Ubuntu 23.04 "Lunar Lobster" - Alpha amd64 (20221126)
SourcePackage: xwayland
UpgradeStatus: No upgrade log present (probably fresh install)

[Julian Andres Klode (juliank)]

Oddly I saw this in Wayland clients now too in plucky

[Julian Andres Klode (juliank)]

I'll try to see if I can run libinput debug-events next time it happens

[Julian Andres Klode (juliank)]

I am getting this multiple times a day with the new keyboard.

I pressed backspace for a bit today

Event: time 1732724530.116877, -------------- SYN_REPORT ------------
Event: time 1732724530.150868, type 1 (EV_KEY), code 14 (KEY_BACKSPACE), value 2
Event: time 1732724530.150868, -------------- SYN_REPORT ------------
Event: time 1732724530.163546, type 4 (EV_MSC), code 4 (MSC_SCAN), value 7002a
Event: time 1732724530.163546, type 1 (EV_KEY), code 14 (KEY_BACKSPACE), value 0

After this I pressed d a couple times.

It took multiple seconds after i stopped for the terminal window to suddenly register all the `d` inputs, it's as is wayland's input processing thread for keyboard is blocked from running. Mouse works.

[Julian Andres Klode (juliank)]

Reassigning to mutter, this is a bug in the key repeat implementation there, I just had

KEY_X(45) pressed
KEY_X(45) released

in libinput debug-events, but then the key kept repeating.

I think I need to set up mutter tracing.

[Julian Andres Klode (juliank)]

This sounds related:

https://gitlab.gnome.org/GNOME/mutter/-/issues/203

Basically since each keypress gets delivered to ibus it could be the ibus that's breaking things and causing its own repeat.