[kwin] [Bug 524117] Night Color toggle and brightness adjustment cause temporary graphical artifacts on AMD Wayland

https://bugs.kde.org/show_bug.cgi?id=524117

--- Comment #3 from Vlad Sizykh <[email protected]> ---
Thanks for the detailed investigation! I can reproduce the same glitches
using
the Night Light preview commands. Let me know if you need anything from my
side.

On Tue, Aug 11, 2026 at 5:00 PM bedo <[email protected]> wrote:

> https://bugs.kde.org/show_bug.cgi?id=524117
>
> bedo <[email protected]> changed:
>
>            What    |Removed                     |Added
>
> ----------------------------------------------------------------------------
>                  CC|                            |[email protected]
>
> --- Comment #2 from bedo <[email protected]> ---
> I'm getting this too and have the same GPU, but I'm running two DisplayPort
> monitors.
>
> Arch Linux
> AMD Radeon RX 6600 (Navi 23)
> Two DisplayPort monitors, both 1920x1080 @ 60Hz
> Plasma 6.7.4, Frameworks 6.28.0, Qt 6.11.1, Wayland
> Mesa 26.1.6
>
> The symptoms look the same as yours. Horizontal strips of glitchy pixels
> that
> only show up while the transition is happening, and they vanish the moment
> it
> finishes. A couple of extra details from my end that might help:
>
> - The strips turn up at random vertical positions and the widths vary a
> lot.
> It's not a fixed region of the screen.
> - With two monitors each screen glitches independently. One can glitch
> near the
> top while the other glitches near the bottom at the same time.
> - Nothing gets logged at all. No kernel messages, no KWin output, no GPU
> resets.
>
> It also happens at login a lot too.
>
> Simple way I've found to trigger it is a night light preview, which doesn't
> involve DDC or brightness at all:
>
>   qdbus6 org.kde.KWin /org/kde/KWin/NightLight
> org.kde.KWin.NightLight.preview
> 2000
>   sleep 3
>   qdbus6 org.kde.KWin /org/kde/KWin/NightLight
> org.kde.KWin.NightLight.stopPreview
>
> That glitches reliably here. Since there's no DDC/AUX traffic and no
> brightness
> change involved, it seems the colour temperature transition on its own is
> enough to set it off. So it might not be specific to brightness or monitor
> controls.
>
> I spent a while ruling out the kernel, and I'm fairly confident it isn't
> involved. It reproduces identically on:
> - linux 6.18.43-lts (stock Arch)
> - linux 7.1.6.arch1, locally patched (reverted ac11060c6d49 for an
> unrelated
> regression)
>
> Things I tried that made no major difference:
>   - amdgpu.damageclips=0
>   - KWIN_DRM_NO_DIRECT_SCANOUT=1
>   - KWIN_DRM_USE_MODIFIERS=0
>   - KWIN_DRM_DISABLE_TRIPLE_BUFFERING=1
>
> I made a script to reproduce before I found the this night light one. It
> runs
> two things at once:
> 1. A repaint load - KDE OSD popups via org.kde.osdService.volumeChanged at
> roughly 12/sec, plus notifications
> 2. Brightness changes through
> org.kde.ScreenBrightness.Display.SetBrightness,
> the same call the taskbar widget makes
>
> Glitched around 50-85% of brightness events, slightly worse when there are
> windows open on the screen.
> OSD repaints and DDC brightness writes alone gave me nothing. Only the two
> together.
> Happy to share the script, test patches, or grab whatever debug output
> would be
> useful.
>
> --
> You are receiving this mail because:
> You reported the bug.

-- 
You are receiving this mail because:
You are watching all bug changes.