Today I tried debugging the strangest problem I've had since moving to Linux. I'm not sure how replicable it is, although someone on IRC was able to reproduce it as well. If you place a Memory widget to the upper-right of a System Load Viewer widget on the desktop, the RAM plasmoid will no longer have its settings and it will see no sensors, when plasma-desktop starts. Only the widget background is visible, and if you try to change its settings they are not persisted. I looked closely in ~/.kde4/share/config/plasma-desktop-appletsrc at the two widgets involved. Their settings don't change when the bug happens or not (except position as you move them) so this happens at run time and is not due to a corrupt configuration. Inside that file, each plasmoid has a line of the form "geometry=1,2,3,4". The first parameter (1 in my example) is the one where the problem takes place in my test. If the memory's is higher than the load viewer's, the issue takes place. Here is an image showing the problem: http://i40.tinypic.com/1zguec6.png In the upper panel, the Memory widget is left of the System Load Viewer widget, and everything is fine. In the lower panel however, the Memory widget is right of it, and as you can see only its background shows. If you right click it and go to settings you can see the entire breakage. Reproducible: Always Steps to Reproduce: 1. Place a Memory widget and a System Load Viewer widget on your desktop. 2. Make sure that the memory widget is positioned above the load viewer and to its right. 3. Shut down the plasma-desktop process and start it up again Actual Results: The memory widget has no settings nor detects any memory sensors Expected Results: It should work as normal I also have a Temperature widget, CPU widget, Disk Activity widget, and Network widget to the right of my System Load Viewer widget, but none of them break (only the RAM widget). The RAM widget also doesn't have issues being right of other widgets, only System Load Viewer from what I could tell.
I can confirm this bug on Arch Linux, KDE 4.10.2.
I can also confirm this bug on Arch Linux, KDE 4.11.0.
*** Bug 315119 has been marked as a duplicate of this bug. ***
In addition, I see this behaviour when putting these widgets on a panel at the bottom of my screen (as opposed to the dashboard). If I move the "Memory Status" widget to the left of the "System Load Viewer" widget, it now works fine. Kubuntu 13.04, KDE 4.10.5
Maybe the problem is in the "corrupted" activity settings (or file). I had the same problem and resolved deleting the "bad activity" and creating the new one. openSUSE 13.1. & KDE 4.11.5
I tried deleting the entire panel and replacing the widgets. I still see this bug.
But have you tried to create completly new activity and start from scratch? For example, right-mouse-click on the desktop, select "Activities" (or Alt+D,Alt+A), then "Create activity...", etc...
No, I haven't. In my situation (as detailed in the comments, and slightly different to the OP), I have the memory widget to the right of the system load widget, with both in a horizontal panel. It's currently impossible to create a different panel for each activity [1]. [1] https://bugs.kde.org/show_bug.cgi?id=292094
Can anyone test if this also happens in Plasma 5? Since the plasmoids were rewritten in QML, I imagine its variant of the System Load Viewer no longer has this issue. KDE4 development is pretty much abandoned now, so this can probably be closed if it works in plasmashell.
Yes, the original plasmoids have been replaced, so this bug is moot. I tested with the newer plasmoids, i.e. "CPU Load Monitor" and "Memory Status" and they work fine in this position. FWIW the new monitors are not appropriate for use in a panel, since they are ridiculously wide, with no way to specify a smaller width. Hence, I've been avoiding them anyway. They also suck up a lot of CPU, which is another reason to avoid their usage, even on the desktop. (There's a bug report floating around for this somewhere.)
(In reply to sparhawk from comment #10) Alright, thanks for confirming that. Closing this in that case... theoretically the problem was fixed.