https://bugs.kde.org/show_bug.cgi?id=517499

Aaron Tinio <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[email protected]

--- Comment #7 from Aaron Tinio <[email protected]> ---
I can add a week of systematic data on this. Same symptom on different
hardware: AMD iGPU (amdgpu), eDP-1 plus DP-2 external via a USB-C dock, DP-2 at
priority 1, Arch, Plasma 6.6.5 on Wayland.

I run a user service that wakes on drm and backlight uevents (brightness bursts
coalesced into one check), reads the priority-1 output from kscreen-doctor, and
re-asserts the panel onto it through the plasmashell scripting API, logging
whether the panel was already placed or had to be moved. Over the past week,
622 checks:

- 9 dock-state changes (8 while running, 1 across a reboot), and the panel
needed to be moved to the priority-1 output every single time. The built-in
relocation never placed it correctly once. The reboot case matches the original
report here: logged in while docked and the panel came up on the internal
display.
- One capture supports comment 6 that this is not tied to connector changes.
The panel had been verified on the external monitor at 09:21:01, and at
09:21:35 the same check found plasmashell's screen indices had swapped (the
index that resolved to DP-2 now resolved to eDP-1), with no drm uevent in
between (drm wakes are never debounced, so the log would show one) and no
suspend or monitor change. The panel was visibly on the external display
immediately before a brightness key press and jumped to the internal display as
I pressed it, so the swap coincided with the brightness change itself rather
than being something earlier that I only noticed then. So the index a panel is
pinned to can be re-resolved to a different output at runtime, while docked,
without any connector event.

For anyone needing a workaround in the meantime: deriving the correct screen
from the priority list (the priority-1 output from kscreen-doctor) and then
locating it in plasmashell by position rather than by screen index has been
reliable, since positions survive whatever re-enumeration flips the indices.

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

Reply via email to