system cursor shows over pointer constrained apps

[fxzzi]

Hyprland version

v0.56.0

Is this a regression?

idk at this point, the problem is a little nuanced and am having trouble bisecting

What's the bug?

When focussing a game with its own pointer (like osu!lazer), and without requiring the confine_pointer rule, if it was being focussed from an empty workspace, the system cursor will remain on screen until you move it.

If you instead let the system cursor remain on screen then use a different input device like a tablet, that will leave the system cursor on screen, following and flickering behind the in game cursor.

For some reason, the behaviour seems to change depending on whether or not hyprpaper is running or not.
When hyprpaper is running, the bug occurs, and you have to move your mouse to hide the system cursor.
When hyprpaper is not running, the bug does not occur, and your in game cursor position updates to the new position as expected and the system cursor isn't visible either.

Reproduction steps

1. launch osu, OSU_SDL3=1 SDL_VIDEO_DRIVER=wayland osu\! for me
2. confirm cursor is working in game
3. move to empty workspace and make sure your cursor is visible (not hidden by inactivity)
4. move back to osu workspace without moving your mouse
5. see system cursor and osu cursor on screen at the same time

(6). without moving your mouse, use your drawing tablet to move the cursor, and notice the system cursor not disappearing

Attachments

xdph captures the cursor differently, through it you cannot see the system cursor following your game cursor, and instead it is just locked in place where it was before focussing osu

out.mp4

I acknowledge that:

• I have searched all (open and closed) discussions and issues and my problem is not there.

• I have read the Code of Conduct and agree to follow it.

• I wrote this request with my own hands, my own keyboard, and in my own words.

[fxzzi]

pinging @iciclejj you broke my things

[iciclejj]

How did you determine that it was caused by d29916a?

I did manage to reproduce it one single time on that commit using your reproduction steps (out of a lot of attempts), and not yet on the parent commit. But if I have opentabletdriver running while also having tablet enabled in osu then I'm able to reproduce the same behavior very easily on both d29916a and its parent, by switching workspaces while using the tablet. Is it possible something similar was the issue for you?

I'll take a closer look at it later, maybe tomorrow.

[fxzzi]

hmm, maybe it's two different issues then. I am using opentabletdriver, with the "artist mode", not "absolute". But I still have an issue without the tablet involved at all, and only a mouse.

When I change focus from osu to an empty workspace, then move the cursor, then move back, the in game cursor position should update to the new position. Instead what happens is that the in game cursor is visible at the old position AND the system cursor is visible at the new position.

Then, after all of this, moving the mouse hides your system cursor and you get jumped to the ingame cursor position. In previous commits this was a lot more seamless. I'm relatively sure this one atleast is caused by this commit

[fxzzi]

nevermind wadafaq its random i see

i got lucky and it stopped reproing for some reason before your pr, my apolocheese
i will try and bisect it again and see what happens

[iciclejj]

np :)

[fxzzi]

right ok, i figured out why my bisect went wrong.

it seems the behaviour changes when hyprpaper isn't running, and at some point in -git we added the graphical target stuff - before that my hyprpaper wouldn't launch and so the issue wouldn't occur.

i guess i need to go through and start graphical session target in every bisect to figure out where it's coming from

and i haven't even started bisect on the tablet cursor thing i talked about it, gg

[fxzzi]

yeah so for some reason the behaviour is dependent on whether or not you are hovering over an app in the background layer.

with hyprpaper running the system cursor remains on screen and the game cursor does not jump to the system cursor position.

when hyprpaper is shutdown the system cursor is hidden immediately, and the game cursor follows the system cursor position.

pinging @vaxerski because it's somehow some weird layer shenanigans involving hyprpaper and not a regression.

below video is simply with a mouse and with no sort of tablet plugged in. you can see cursor is working correctly whilst hyprpaper isn't running, then borken when it is

26.10.15-10.48.44.mp4

[fxzzi]

can also be reproduced with RetroArch.

hovering over hyprpaper, then moving to retroarch doesn't hide the cursor.
shutting down hyprpaper, or hovering over an app instead of an empty workspace, makes the cursor immediately hide when refocussing retroarch

[fxzzi]

and after some quick check it can be repro'd on foot term even.

[mouse]
hide-when-typing=yes

put this in your foot config.
you'd expect that when you begin typing, your cursor will disappear.
but it only does that if you move your cursor on top of foot. if you keep your cursor still when you open it, then start typing, it doesn't hide itself

the problem at work below. you can see foot is supposed to change the cursor shape too. but it only does that once i move over the window.
closing hyprpaper, as above, fixes it lol

out.mp4