After I updated my Firefox to version 29, I noticed a pattern of plasma-desktop disappearing periodically. This happened on 3 systems already (all running current Debian testing with KDE 4.11.8 ), which all just got Firefox update. I'll perform more tests to exclude other reasons, but Firefox is the main suspect right now. Did anyone encounter similar issues? If I'll confirm Firefox, I'll file a bug.

have you tried renaming ~/.mozilla to test if it's a plugin/extention/theme issue ?

I'll run a test with all add-ons disabled to compare. Right now the crash seems to happen periodically, approximately once in 10 minutes.

When running plasma-desktop from the terminal, I see this error just before it disappeared:

Code: Select all

plasma-desktop: symbol lookup error: /usr/lib/apper/libapper.so: undefined symbol: appstream_database_new

Sorry for confusion, I just run a test without Firefox, and it still crashes. Apper appears to be broken. When I run apt-get update it crashes right away, and plasma-desktop produces the error like above. So Firefox isn't the reason, Apper icon with updates in the system tray is. I updated the title to clarify the case.

That's clearly a bug, may be caused by incompatible Apper version (0.8.1).

Filed a bug here: https://bugs.kde.org/show_bug.cgi?id=334188

this usually means that there are packages (programs, libraries) out of sync causing an incompatibility and not a KDE bug. Probably something occurred during an upgrade and you should make sure all the apper & appstream packages are correct, especially the appstream ones

That may be. Could be it's Debian testing specific, and they messed up dependencies upgrades for apper. I can file a bug against Debian as well, but it looks like apper 0.8.2 is already in the pipeline: http://packages.qa.debian.org/a/apper.html
so I'll wait until then and if it won't fix the issue, I'll file a Debian bug.

Running SolydK KDE based on Debian testing and use apper 0.8.1 without issue.
Also latest update was Kernel to 3.13.1 and KDE to 4.12.4 so looks like you are out of sync? Just guessing here.

As you stated

running current Debian testing with KDE 4.11.8

But haven't had any issues with running apper all this year.
.

http://packages.qa.debian.org/k/kde-workspace.html see version it testing (4.11.8 ). That's what I have. Kernel is 3.13. Apper is 0.8.1 for quite a while and it worked in the past. It can be something else in KDE that was updated and now is causing the issue because apper isn't compatible.

It appears that libappstream has been updated, and has broken it's ABI - which is why Apper is now failing to run. As it uses a Plasma Applet for some functions it causes Plasma Desktop to crash when this occurs. Can you please verify that all repositories are up to date and that libappstream and Apper are coming from either the primary distribution repositories, or the same repository? I suspect it may have been upgraded by the Firefox upgrade.

It's not related to Firefox, since I use straight Mozilla builds which don't manage any pacakges.

My repositories are up to date, and libappstream was indeed updated on Aprlil 27 in Debian testing: http://packages.qa.debian.org/a/appstream.html (from version 0.5 to 0.6.1), so if the ABI changed, it must have broken apper. Does apper 0.8.2 use that newer ABI?

Hard to tell. However Debian's tools should have detected the ABI breakage - please try refreshing the repositories and upgrading your system, as it is possible an updated Apper package was in the process of being built at the time in question.

If that doesn't fix the issue, please report this as a bug against Debian Testing to allow the packagers to take appropriate action.

I reported the bug to Debian, and maintainers closed it, since newer version of apper is coming soon to Debian testing. What I'm not sure about, is whether it's normal for Debian testing to push packages out which break ABI, expecting that the fix will arrive later. I'd expect that related packages have to be pushed out together. I closed the KDE bug, since it's not a KDE issue.

Okay - thanks for the update. It is rather interesting that ABI breakages would be allowed to be distributed even in a testing distribution though.