You are not logged in.
Since the upgrade to Firefox 137 I'm consistently seeing rendering artifacts while scrolling web pages. I'm under a Plasma Wayland session with an Intel integrated GPU. Downgrading to the 136 series fixes the problem.
I searched the web for such an issue but I didn't find anything. Does anyone have suggestions? Thanks.
Offline
Offline
Same here, but I'm on a AMD Radeon 780M (RADV PHOENIX). Kernel 6.14.3, mesa 25.0.4 and firefox 137.0.2. If I scroll quickly, I get lots of rectangles showing what's behind firefox as artifacts. Very annoying.
Offline
Me too, Same horrible Flickering using firefox 137.0.2 with Wayland on Intel® Iris® Xe Graphics, 6.14.3-1 but I'm on Opensuse-Tumbleweed.
Offline
Same with Firefox 138. Maybe the fact that it's running on a 120 Hz 4k monitor triggers the bug?
Offline
Did you try to run the output at 60Hz and/or FullHD only?
We recently ran into https://gitlab.freedesktop.org/drm/amd/-/issues/3708 - amdgpu has *some* problem w/ higher pclck demands
Offline
@seth I have an integrated Intel GPU so I think your suggestion does not apply to my case. Or should I try anyway to do some tests changing the resolution and the refresh rate?
Offline
No, that was a suggestion for Mads, the suggestion in https://bbs.archlinux.org/viewtopic.php?id=304733 was to disable the HW acceleration in FF
You could also "LIBGL_DRI3_DISABLE=1 firefox" except if you're using xf86-video-intel and/or xwayland - in which case the suggestion would be to stop doing that, https://wiki.archlinux.org/title/Firefox#Xwayland
Offline
I haven't used any X11-related rendering in years, so I don't think LIBGL_DRI3_DISABLE and the likes applies...
It's a weird thing, because everything worked nicely in Firefox 136. Although I think you're right that the problem goes away if you disable hardware acceleration - but that is not a workaround we want to end up with.
Last edited by Mads (2025-04-30 17:53:14)
Offline
Did you try this solution
https://bbs.archlinux.org/viewtopic.php?id=304962 ?
Offline
LIBGL_DRI3_DISABLE isn't X11 related at all - did you make sure that FF doesn't end up being an xwayland client?
However, there's way too much crosstalk here, I actually asked you whether you tried the behavior on lower refresh & resolution and suggested snack the stuff in #8 revolving around DRI3 and xf86-video-intel
Offline
Did you try this solution
https://bbs.archlinux.org/viewtopic.php?id=304962 ?
Yes, it was one of the first things I checked. I haven't got a single "force-enable" toggled in about:config at all.
In about:support it says "wayland" under "Window Protocol" so it hasn't toggled X11 without me knowing.
My understanding was that anything DRI3-related has to do with xorg ddx drivers. But anyway, I tested it now like you said - unfortunately it did not make any difference.
Mesa just got a new version out, I'll try updating and see if I see anything.
Offline
Yesno, DRI* is X11 specific but the environment affects the mesa code and idk whether there're overlapping code paths.
However, again: that wasn't meant for you tbw.
==> Did you test the impact of the refresh rate and resolution?
Offline
Yesno, DRI* is X11 specific but the environment affects the mesa code and idk whether there're overlapping code paths.
However, again: that wasn't meant for you tbw.
==> Did you test the impact of the refresh rate and resolution?
i tried 1920x1080 100% scaling both 120 and 60 hz. It's harder to render the artifacts (you really have to excessively scroll), but the bug is still there in both cases.
Offline
Does "export vblank_mode=0" have any impact on the sitation?
Also: do yo actally have the problem on X11 or xwayland?
Does it still work properly w/ 136? And what wayland compositor? Did you try any other one?
"what's behind firefox" can only be caused by the compositor (induced by FF or not) - is *everyone* here usign plasma?
Offline
No X11 has been involved in any of this.
vblank_mode=0 didn't affect anything.
It works with Firefox 136. I also use plasma 6.3.4.
Offline
Test a different compositor (can be weston) and/or X11 - it's obviously triggered by some change in FF but the symptom sounds like a compositor bug.
Where did you export vblank_mode=0 ? To the session or to firefox only?
Offline
I just tried 138.0.1-1 and it's affected as well.
Offline
Test a different compositor (can be weston) and/or X11 - it's obviously triggered by some change in FF but the symptom sounds like a compositor bug.
and in that case I'd not expect any FF update to "fix" this.
Offline
seth wrote:Test a different compositor (can be weston) and/or X11 - it's obviously triggered by some change in FF but the symptom sounds like a compositor bug.
and in that case I'd not expect any FF update to "fix" this.
Well it could have happened that FF reverted this supposed change in 138, but I understand your point.
Offline
compare "about:support" between 136 and 137, maybe you've been blessed w/ HD acceleration?
Offline
Still there with Firefox 138.0.4 and Mesa 25.1.0. Annoying as hell, because it's not an easy bug to report. Haven't found trace of anything like it on Mozilla's bugzilla either.
Offline
Here's a small recording of how it looks like if I just scroll Firefox with the scroll wheel quickly:
On the video the articating area gets black, but in real life it gets filled with junk (bits of random graphics/glitches)
Offline
I reported the issue here: https://bugzilla.mozilla.org/show_bug.cgi?id=1967250
Offline
Gut feeling from the video, have you tried to disable kinetic scrolling?
https://wiki.archlinux.org/title/Firefo … _scrolling
Offline