| Summary: | turning one monitor off kills the panel configuration of the second monitor | ||
|---|---|---|---|
| Product: | [Plasma] plasmashell | Reporter: | Kai Krakow <kai> |
| Component: | generic-multiscreen | Assignee: | Plasma Bugs List <plasma-bugs> |
| Status: | RESOLVED DUPLICATE | ||
| Severity: | normal | CC: | aleixpol, gronslet, nate, notmart |
| Priority: | NOR | ||
| Version: | 5.24.5 | ||
| Target Milestone: | 1.0 | ||
| Platform: | Gentoo Packages | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed In: | ||
| Sentry Crash Report: | |||
| Attachments: | configuration completely doesn't match the output | ||
|
Description
Kai Krakow
2022-05-08 20:52:13 UTC
are all outputs connected to the internal videocard or there are external DP dongles involved? in plasmashell itself it doesn't seem that anything have happened between 5.23 and 5.25, wonder in what part the breaking change was. also, the interested part has been redone completely for 5.25, would be interesting to see how it behaves there (In reply to Marco Martin from comment #1) > are all outputs connected to the internal videocard or there are external DP > dongles involved? No dongles but there is one DP-to-HDMI cable. Not sure if that counts as a dongle. The iGPU card is not active, I'm only using the PCI-e card. > in plasmashell itself it doesn't seem that anything have happened between 5.23 and 5.25, wonder in what part the breaking change was. This morning when I turned the monitors back on, all was fine. So it may be some timing issue or something is racing. Or it just created "enough" panels in the qbus javascript dump that it no longer creates new configurations but just chooses from the existing ones. I'll observe it for some time and report back. Also, Plasma version is probably not the only thing that changed. The nvidia driver was also updated at least once, and Xorg components were also updated at least once. But I cannot really pin-point that to one of those changes, Plasma version seems to be the strongest indicator. It worked for a few days in a row. But yesterday evening and today morning, this happened again and the behavior is quite random: This time, only the top panel moved to the wrong screen while the bottom panel stood in place. Or maybe it's rather the other way around: Turning my main monitor off probably moves the panels to the other monitor, and when turning it back on, the panels do not properly move back. The interesting part this time is that the bottom panel behaved properly while the top panel behaved wrong. Latest 5.24 doesn't fix anything. Let's see if 5.25 solves something. Okay, so I'm now on Plasma 5.25 and this original issue seems fixed. But now, when I turn the monitor back on, the background image of that monitor is gone and completely black. I also cannot right click on the background: no context menu would appear. The panels are still there, and switching the panels to editing mode doesn't show the additional button bar to change background images, global design etc on that monitor (but it's shown on the others). Only restarting plasmashell will bring back the background image and context menu function. Cool, thanks! Can you file a new bug report for the new issue? Out of curiosity, does the wallpaper show up after 10 seconds or so? (In reply to Nate Graham from comment #7) > Cool, thanks! Can you file a new bug report for the new issue? > > Out of curiosity, does the wallpaper show up after 10 seconds or so? Re-opened, because unfortunately, the bug is still there: With the new panel management tool (available from the edit mode), I can now clearly see how panels move to the wrong screen. This usually happens when connected monitors swap positions: The last few days, I'm pretty sure I had the following order in the "Drag panels and desktops around" dialog: DP-4, DP-1, HDMI-0 (deactivated). Today, when I turned my monitors on, the order changed to "DP-4, HDMI-0 (deactivated), DP-1", and the panels have moved to the deactivated screen (but not the background image, that moved to a different screen a few days before already). Please take note that the background image shows the same bug but moves independently of the panels, so there are probably race conditions. I've set different wallpapers to track this: Both the wallpapers and the panels move to different screens, and they do NOT move the same way: Sometimes only the wallpaper moves to a different screen, sometimes only one panel of a screen moves, sometimes two panels of the same screen move. At least the panel management tools allows me to simply move panels and wallpapers back to their correct screen. The system hasn't been rebooted. All I do is turning off the monitors, and one monitor completely disconnects from the graphics card when turned off. So to reproduce it, in the worst case you'll have to physically disconnect the monitor because many monitors keep their connector online when turned off - but one of my monitors doesn't. Created attachment 150180 [details]
configuration completely doesn't match the output
Here's an example of messed up content:
As you can see, the panels overlap on the right screen although two of them are supposed to be on the left screen. Also, the left screen shows the background image of the deactivated screen. Dragging the panels back and forth in the editor fixes the panel placement. But dragging the background image will simply leave a black background that is not clickable or interactable anymore until I restart the plasmashell service.
So, no - it's not fixed. I'd even say it shows the exact same behavior in 5.25 as it did in 5.24. Configuration and outputs do not match, panels are in very different locations than what Plasma thinks they should be.
Rarely, I see how all panels disappear when I turn a monitor back on, and then re-appear around 5-10s later if that is what you were asking for. This seems to have the effect that plasma doesn't become confused about the monitor layout. I've last seen that in 5.24 but not yet in 5.25.
*** This bug has been marked as a duplicate of bug 450068 *** |