keyd: Keyd's virtual pointer emits wrong button codes

This is probably not a keyd issue, but nevertheless I’m recording it here: After having updated my laptop (running Archlinux), the buttons emitted by keyd’s virtual pointer are suddenly 18, 17 and 16 (leftmouse, middlemouse, rightmouse) instead of 1, 2 and 3. I had to fix this via xinput set-button-map. Is there any explanation for this strange behaviour?

About this issue

  • Original URL
  • State: closed
  • Created 4 years ago
  • Comments: 15 (9 by maintainers)

Commits related to this issue

Most upvoted comments

couldn’t find any documentation which corroborates this

====== < SARCASM > ====== Documentation is deprecated in systemd. Documentation is readable by humans, we only do binary, therefore can’t read nor write it. By not adhering to documentation, we are able to boot more quickly. ~~ Poettering ====== < /SARCASM > ======

I believe BTN_0 is just a generic button event that can be used by any mouse driver

Mice and similar devices use the buttons starting with BTN_LEFT and up. BTN_0 is really a “generic numbered button” and only tablet pads use those (statistically anyway) - you shouldn’t see real-world mice with those buttons. Mind you, there’s this systemd PR which may also fix this issue here I just haven’t had the time to look at that one yet.

@herrsimon Thanks.

I did a bit of digging and was able to narrow it down to this commit in systemd:

https://github.com/systemd/systemd/commit/0855ce6772#diff-e3c18b2fc4f50acbb9f2d66a1daf229fde99a5c8785f17aa6f69e6a9c41ae2c7R261

It seems devices which are capable of emitting BTN_0 (and wheel events) are now treated as ‘tablet pads’. This seems odd to me, since I believe BTN_0 is just a generic button event that can be used by any mouse driver, but I couldn’t find any documentation which corroborates this (perhaps @whot can shed some light).

In any case, the events are currently unused in keyd, so I have removed them in 0fb86d3a9, which should solve the problem.

@rvaiya

If so, my suspicion is that some mouse detection mapping/code has changed in X/libinput

Yes, that’s what I suspect too. Something like “if your pointer device has 17 buttons, it means it’s some new iThingy and has to behave differently”.

libinput debug-events, press keyboard “shortcut” for middle-mouse-click:

— original keyd (with both loops)

-event13  TABLET_PAD_BUTTON       +39.030s	12 pressed (mode 0)
 event13  TABLET_PAD_BUTTON       +39.157s	12 released (mode 0)

— modified keyd (second loop commented out)

 event13  POINTER_BUTTON          +0.000s	BTN_MIDDLE (274) pressed, seat count: 1
 event13  POINTER_BUTTON          +0.128s	BTN_MIDDLE (274) released, seat count: 0

This fixes the issue for me:

diff --git a/src/vkbd/uinput.c b/src/vkbd/uinput.c
index b98e585..d1a9fca 100644
--- a/src/vkbd/uinput.c
+++ b/src/vkbd/uinput.c
@@ -103,7 +103,6 @@ static int create_virtual_keyboard(const char *name)
 
 static int create_virtual_pointer(const char *name)
 {
-       uint16_t code;
        struct uinput_user_dev udev = {0};
 
        int fd = open("/dev/uinput", O_WRONLY | O_NONBLOCK | O_CLOEXEC);
@@ -126,11 +125,11 @@ static int create_virtual_pointer(const char *name)
        ioctl(fd, UI_SET_RELBIT, REL_Y);
        ioctl(fd, UI_SET_RELBIT, REL_Z);
 
-       for (code = BTN_LEFT; code <= BTN_TASK; code++)
-               ioctl(fd, UI_SET_KEYBIT, code);
-
-       for (code = BTN_0; code <= BTN_9; code++)
-               ioctl(fd, UI_SET_KEYBIT, code);
+       ioctl(fd, UI_SET_KEYBIT, BTN_LEFT);
+       ioctl(fd, UI_SET_KEYBIT, BTN_RIGHT);
+       ioctl(fd, UI_SET_KEYBIT, BTN_MIDDLE);
+       ioctl(fd, UI_SET_KEYBIT, BTN_SIDE);
+       ioctl(fd, UI_SET_KEYBIT, BTN_EXTRA);
 
        udev.id.bustype = BUS_USB;
        udev.id.vendor = 0x0FAC;