wlroots
wlroots copied to clipboard
swaywm
Disappearing cursor
wlroots: https://github.com/swaywm/wlroots/commit/20d404a091fa92f56be4825bbaf89e4027fda154 sway: ec5da0ca5bad6a433f727499d68ac1352397f5aa
Not sure if this is a problem with sway, or not. But, I'm now using sway on multiple systems, with the same software setup.
Everything is working fine on all systems except one, where the mouse cursor keeps disappearing. It seems its position is correctly tracked though, because e.g. hover effects in browsers work, focus-follows-mouse works and so, on. It's just that the cursor itself is invisible.
The cursor is usually invisible right from the start, before any windows have been opened. Then, if I move it between windows, it will sometimes become visible. When that happens, it stays visible until moved to another window.
The sway debug log doesn't contain any obvious error messages, except these:
2019-02-07 21:22:15 - [sway/input/cursor.c:1132] denying request to set cursor from unfocused client
and
glamor: No eglstream capable devices found
The main difference between the non-working system and the working ones are that the non-working one has an nvidia card (with nouveau), while the others are intel based.
wlroots has migrated to gitlab.freedesktop.org. This issue has been moved to:
https://gitlab.freedesktop.org/wlroots/wlroots/-/issues/3189
I have the same issue: Quadro 3000m + nouveau, Linux 4.20, Manjaro.
Edit: setting WLR_NO_HARDWARE_CURSORS=1 env. variable fixes it
This is probably established fact by now, but just wanted to let you know that https://github.com/swaywm/wlroots/pull/1537 (which I found through https://github.com/swaywm/wlroots/pull/1526) works for me too.
I have an Ivy Bridge CPU/gpu and I also had this problem, with cursor disappearing.
WLR_NO_HARDWARE_CURSORS fixed it but surely this bug ought to have a better fix?
I'm also having this problem, but on intel graphics and WLR_NO_HARDWARE_CURSORS=1 does not fix it.
For example, the cursor would suddenly become invisible over firefox, but become visible when moved over an edge or to another window. I noticed that changing the scale factor will sometimes bring back to the cursor over one application, but it would still become invisible over some other application. Normally, I use scale 3, and sometimes, the cursor would be tiny (as in scale 1) instead of being invisible.
I also noticed that if I start typing in gnome-terminal, the cursor will disappear (as intended), but if I move it again, it does not reappear. It stays invisible in all instances of gnome-terminal.
Also, if I change the cursor theme, e.g. gsettings set org.gnome.desktop.interface cursor-theme "Adwaita", the application which the cursor was on during the change will crash as soon as the cursor is moved after the theme change. I can reproduce this with firefox and gnome-terminal.
I guess these could all be separate bugs but they might be connected...
I'm having this issue on my laptop (radeon graphics) but it only occurs after I unlock from swaylock and only if the mouse pointer was on my external monitor when I locked the desktop.
Then it will only appear on the external monitor. Sway is definitely aware of where the cursor is as I can see the focus change on the laptop monitor when I move the mouse around but the cursor is invisible.
@emersion I have a debug log but it contains some personal information do you have some script to anonymize it a bit, like removing window titiles?
Nope. Can you at least share the [ERROR] lines to confirm this is the bug I think this is?
Alright, that's what I've been hitting as well:
41237:02:14:10.874 [ERROR] [wlr] [render/swapchain.c:97] No free output buffer slot
41238:02:14:10.874 [ERROR] [wlr] [types/wlr_output.c:1342] Failed to render cursor buffer
Same here, I'm on arch linux. I'm using 2 monitors one using display port the other hdmi. The cursor disappears only in the display port monitor, when i move the cursor to the other one i can see the cursor just fine.
I'm running into this issue as well with a HDMI monitor. And I'm having the opposite problem, the cursors is invisible only in the main display, being visible on the external one. going into a terminal and opening wdisplays seems to fix it until I close everything else and firefox resizes over the cursor in that workspace. After moving stuff around in workspaces sometimes it stays visible
It seems that resizing firefox over the cursor, or closing the last window of the workspace renders the cursor invisible, and having wdisplays open prevents any dissapearing
Extra info: this is happening on Arch linux, with intel graphics
Unplugging and replugging in the monitor fixed it for me when I unlocked my desktop system this morning and the right monitor cursor was invisible. It was DP-1
$ lspci | grep VGA
03:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 [Radeon RX 5600 OEM/5600 XT / 5700/5700 XT] (rev ca)
$ swaymsg -t get_outputs
Output HDMI-A-1 'Dell Inc. DELL U2414H FN8GP68S1JML'
Current mode: 1920x1080 @ 60.000 Hz
Position: 0,0
Scale factor: 1.000000
Scale filter: nearest
Subpixel hinting: unknown
Transform: normal
Workspace: 5
Max render time: off
Adaptive sync: disabled
Available modes:
720x400 @ 70.082 Hz
640x480 @ 59.940 Hz
640x480 @ 59.940 Hz
640x480 @ 60.000 Hz
640x480 @ 75.000 Hz
720x480 @ 59.940 Hz
720x480 @ 59.940 Hz
720x480 @ 59.940 Hz
720x480 @ 60.000 Hz
720x480 @ 60.000 Hz
720x576 @ 50.000 Hz
800x600 @ 60.317 Hz
800x600 @ 75.000 Hz
1024x768 @ 60.004 Hz
1024x768 @ 75.029 Hz
1280x720 @ 50.000 Hz
1280x720 @ 59.940 Hz
1280x720 @ 60.000 Hz
1280x720 @ 60.000 Hz
1152x864 @ 75.000 Hz
1280x800 @ 60.000 Hz
1440x900 @ 60.000 Hz
1280x1024 @ 60.020 Hz
1280x1024 @ 75.025 Hz
1600x900 @ 60.000 Hz
1680x1050 @ 60.000 Hz
1600x1200 @ 60.000 Hz
1920x1080 @ 50.000 Hz
1920x1080 @ 59.940 Hz
1920x1080 @ 60.000 Hz
1920x1080 @ 60.000 Hz
Output DP-1 'Dell Inc. DELL U2414H FN8GP68S1HEL' (focused)
Current mode: 1920x1080 @ 60.000 Hz
Position: 1920,0
Scale factor: 1.000000
Scale filter: nearest
Subpixel hinting: unknown
Transform: normal
Workspace: 2
Max render time: off
Adaptive sync: disabled
Available modes:
720x400 @ 70.082 Hz
640x480 @ 59.940 Hz
640x480 @ 59.940 Hz
640x480 @ 60.000 Hz
640x480 @ 75.000 Hz
720x480 @ 59.940 Hz
720x480 @ 59.940 Hz
720x480 @ 59.940 Hz
720x480 @ 60.000 Hz
720x480 @ 60.000 Hz
720x576 @ 50.000 Hz
800x600 @ 60.317 Hz
800x600 @ 75.000 Hz
1024x768 @ 60.004 Hz
1024x768 @ 75.029 Hz
1280x720 @ 50.000 Hz
1280x720 @ 59.940 Hz
1280x720 @ 60.000 Hz
1280x720 @ 60.000 Hz
1152x864 @ 75.000 Hz
1280x800 @ 60.000 Hz
1440x900 @ 60.000 Hz
1280x1024 @ 60.020 Hz
1280x1024 @ 75.025 Hz
1600x900 @ 60.000 Hz
1680x1050 @ 60.000 Hz
1600x1200 @ 60.000 Hz
1920x1080 @ 50.000 Hz
1920x1080 @ 59.940 Hz
1920x1080 @ 60.000 Hz
1920x1080 @ 60.000 Hz
Had the same thing happen this morning on a DP monitor (dual-monitor setup, Polaris GPU).
Have the same issue consistently after swayidle takes both the laptop monitor LVDS-1 and the external monitor HDMI-A-1 off and turns them on again. Mouse cursor is missing on LVDS-1, but is present on HDMI-A-1, afterwards. As observed by others, effects of the mouse are there (mouse overs are highlighted, clicks are possible), but there is just no pointer/cursor.
Messages by sway are lots of:
01:15:54.682 [ERROR] [wlr] [render/swapchain.c:97] No free output buffer slot 01:15:54.682 [ERROR] [wlr] [types/wlr_output.c:1342] Failed to render cursor buffer
Edit: Might be related that, when DPMS is turned off, the display with the mouse pointer still working says
2021-08-18 22:53:14 - [swaybg-1.1/main.c:373] Destroying output HDMI-A-1 (Hewlett Packard HP 27es 3CM9340NTF )
while the other display goes off, but the message is
2021-08-18 23:26:04 - [swaybg-1.1/main.c:373] Destroying output (null) ((null))
and appears only at powering DPMS on again.
Moreover, quite some time before that, there is:
00:17:45.884 [ERROR] [wlr] [backend/drm/atomic.c:35] connector LVDS-1: Atomic commit failed (pageflip): Device or resource busy
Is this a regression? I don't remember having this problem on 1.5.x, and I already had an HDMI monitor.
Is this a regression? I don't remember having this problem on 1.5.x, and I already had an HDMI monitor.
It certainly is for me.
Also seems to be a regression for me. I lost my external monitor for a couple months, and now that I have it back, I am seeing this issue. The only changes have been software updates: it was working just fine a few months ago, and now when I plug in the external monitor, have the monitors go to sleep and then wake back up, the cursor is invisible on the laptop's builtin display but visible on the external display. Single (internal) display: no problems.
Arch Linux, AMD Renoir integrated GPU, sway 1.6.1.
I don't believe it is a regression from 1.5.x to 1.6.x, since I was running 1.6.0 without problems. I would suspect the wlroots change from v0.13.0 to v0.14.0.
Indeed: in my case, downgrading to the 1.6.0 sway package (which uses wlroots v0.13.0) seems to prevent the problem.
This condition is triggered for me mid-session after launching certain programs (eg, the Arch AUR obsidian-appimage package).
One interesting bit I've noticed is that keeping the wdisplays program running keeps cursors visible, although the failed to render cursor buffer errors keep scrolling by. I wonder why this is.
Judging by the timing of when this issue appeared for me, and that some cursor-related items appeared in the changelog, could the issue have appeared in 0.14.1 and not 0.14.0?
I've experiencing something as well. I can keep the mouse visible when wdisplays is open but otherwise it is hidden. Oddly it's only on one display. The primary one DP-1 while DP-2 will display the mouse fine regardless. I've only noticed the behavior after waking from swaylock and maybe twice in the last 7 days.
The workaround I've come up with is to pull out and put back in the monitor cable.
The workaround I've come up with is to pull out and put back in the monitor cable.
Interestingly, that does nothing for me.
Have other people tried opening wdisplays? That seemed to do the trick for me, at least temporarily
Does wdisplay disable hardware mouse cursor? It solves the problem for me but mouse movement becomes noticeably more sluggish.
I don't know about wdisplay disabling the hardware cursor but i do know if wdisplay is open while some videos are playing on screen it can cause a performance drop. I'm guessing because of the video preview aspect of wdisplay, which is pretty slick. Though you can disable the preview once wdisplays is open there is a hamburger menu which allows you to toggle "Show Screen Contents" and another one "Overlay Screen Names". Disabling both and throwing wdisplays to a hidden workspace or scratchpad might be a viable temporary workaround. Granted I haven't tested to see if disabling "Show Screen Contents" actually improves performance, only an assumption.
I've had this issue, switching to a TTY without sway on it and back seems to bring the cursor back, nixos 21.05, sway 1.6.1, sandy bridge with integrated graphics.
Oddly it's only on one display. The primary one DP-1 while DP-2 will display the mouse fine regardless. I've only noticed the behavior after waking from swaylock and maybe twice in the last 7 days.
Exactly the same behavior here as well. Only on the first monitor and after resume from standby... (only happens sporadically though. In about 70% of the cases for me)
switching to a TTY without sway on it and back seems to bring the cursor back
Thx. Will try that next time.
