Version: (using KDE 4.2.1) Compiler: gcc version 4.4.0 20090307 (Red Hat 4.4.0-0.23) (GCC) qt-4.5, fedora rawhide OS: Linux Installed from: Fedora RPMs While experiencing bug 187251, I noticed that at each plasma crash the size of my folder view on desktop is not preserved. After plasma is restarted, it shows up minimal sized. This can be reproduced quite easily: 1. Put a folder-view on desktop 2. open konsole, killall plasma 3. start plasma again Expected behaviour: Folder view should occupy the same size on desktop as before the plasma crash happend
Created attachment 32157 [details] plasma before crashed/killed
Created attachment 32158 [details] plasma after restart
Still there using svn trunk 1066188 (KDESC 4.4 beta 2). And I've waited 1 minute before killing plasma-desktop. The size, position and rotation are lost. But also, I've added a new folder view before killing plasma, and the new plasmoid is still there (in its original position, size and rotation).
Can you reproduce using KDE 4.4.4 or 4.5beta?
Still there in trunk svn revision 1134984. The size and rotation are lost.
I think this is actually a problem with the config saving in libplasma. there are some funny things that can happen when a plasmoid inherits Containment but isContainment() returns false... suspicious things happen in Applet, iirc, with config getting copied to [Configuration] when maybe it shouldn't.
Is this still an issue in kde 4.7x?
Yep same thing with KDE 4.7.3 I've put a folder view widget, rotated it and then killed plasma-desktop. When restarting plasma, the rotation is gone and the widget back to its previous position. I guess the last rotation wasn't saved (yet?) when I killed plasma.
Still there in trunk (from yesterday night). The folder view widget size and rotation are lost. I waited 5 minutes before I killed plasma-desktop.
First of all, crashes shouldn't occur, so if something's lost on a crash, the issue is in the crash itself, not in the fact that the data is lost as a result of the crash. So "widget size not preserver on plasma crash" is an invalid bugreport per se. Now the kquitapp plasma-desktop configuration loss is important. I couldn't reproduce this issue in 4.9 trunk. I guess it's been fixed as a result of the recent configuration code changes by Aaron Seigo, so I'm closing this as
fixed.
So, if a filesystem is very likely to loose data because of power failures, its clearly a hardware-bug - but wouldn't it be nice to have the filesystem handle such failures gracefully, instead of dooming the user?!