https://bugs.kde.org/show_bug.cgi?id=526072
--- Comment #4 from [email protected] --- (In reply to Zamundaaa from comment #1) > Please do a git bisect to find out which commit caused the regression. I > don't see anything between 6.7.4 and 6.7.5 that I think would be relevant. Bisect update, plus a significant change in my repro situation. I can no longer reproduce the bug, and two things about my setup complicate the "regression in 6.7.5" framing in ways I don't think are visible from the report alone. 1) My A/B was confounded. I'm on Bazzite (Fedora Atomic/immutable), where the Plasma version ships inside the OS image. Every image swap changes Plasma, the kernel, and the NVIDIA driver together. The "innocent" 6.7.4 image runs kernel 7.2.3-ogc3.1; the "guilty" 6.7.5 image ran 7.2.4-ogc. So my weeks-clean-then-instant-onset evidence points equally at a kernel regression as at KWin. Given comment #3 (same symptom on 7.2.11-arch, gone on LTS 6.18), I'd take seriously that this bug is partly or wholly kernel/DRM-side, and I'd support re-triaging in that direction or at least recording the kernel as a co-suspect. 2) My repro disappeared when an unrelated USB fault was fixed. Hardware: 2x NVIDIA RTX PRO 6000 Blackwell (Workstation + Max-Q) central 5120x2160@180 over DisplayPort two portrait monitors on the same GPU. Since roughly the day this bug started, my MT7927 Bluetooth adapter had been in a firmware-missing reset loop, the kernel was resetting that USB device ~4 times per second whenever Bluetooth was powered on. Checking my journals afterwards, the blackout incidents clustered exactly in the windows when that loop was running (active mid-September, off 22 Sep–6 Oct, active again from 6 Oct). On 7 Oct I installed the missing firmware blob and the loop stopped. Zero blackout episodes since, every DPMS wake has been clean for days. I have no theory for the mechanism, but a constant hotplug/USB-churn environment may well have been part of the trigger. Worth noting: two episodes did happen on the *good* image back on 6-7 Oct, on a day the machine had been opened and the DP cable physically disturbed, so I can't rule out marginal link as a second, independent trigger at 5K@180. 3) On the bisect: I can't bisect what won't reproduce, a 300-iteration test where every build passes proves nothing. If the symptom returns I'll absolutely run a bisect, and I'm happy to retest specific builds or a kernel-side patch candidate on this box at short notice. If this report would be more useful against the kernel bug, I'd rather it move there than linger here mislabelled. Happy to provide journal excerpts from the reset-loop window if useful. Thanks for looking at it. -- You are receiving this mail because: You are watching all bug changes.
