https://bugs.kde.org/show_bug.cgi?id=301453 It causes segfaults for at least kmail, gwenview and dolphin. If 4.8.4 hits the tree, it should stay hard masked.
Are you just reporting that here, or have you been hit yourself by the bug?
I've been hit by it myself, with kmail, gwenview and dolphin. kmail and gwenview crash directly on start, dolphin often runs and works, but crashes on shutdown. reports for gwenview: https://bugs.kde.org/show_bug.cgi?id=301468 https://bugs.kde.org/show_bug.cgi?id=301474 for kmail: https://bugs.kde.org/show_bug.cgi?id=277705 for dolphin I found 1 report, that was marked as duplicate. Nepomuk must be turned ON in order to reproduce those crashes.
This is a soprano issue. I upgraded to 2.7.56 and the crashes are gone.
Created attachment 314857 [details, diff] extracted new Soprano::Socket code from the complete diff between 2.7.6 and 2.7.56 In earlier discussions in bug reports several times was mentioned, that Qt Sockets can't live in more then one thread and causing problems. After 2.7.6, the new "Soprano::Socket" class was introduced, which addresses this problem. The attached patch replaces the old sockets with the new class. It applies fine to soprano-2.7.6, and now the crashes are gone. But I am sure, that soon there will be a soprano-release that addresses this issue, as of now the current stable stack (kde-4.8.4+soprano-2.7.6) is not usable.
Ugh. Awesome. Means we have to depend on a soprano beta version with unreleased patches in a KDE minor "bugfix" point release. All fixed in the overlay, thanks. We now depend on >=2.7.56-r1.
Created attachment 315243 [details, diff] Upstream fix Upstream now reverted two commits: http://quickgit.kde.org/index.php?p=kdelibs.git&a=commit&h=bde5aad1e9f7bfb6f8d5c60197797de4a315158d http://quickgit.kde.org/index.php?p=kdelibs.git&a=commit&h=76e0376d7582cfe5b898c56bf3779ca2279dace8 Looks now as in kdelibs-4.8.3. Although understandable (4.8.3 works, yes) IMHO this just hides a problem in another part of the code. The backtrace is completely unrelated to soprano, so we most likely face undefined behaviour (someone is writing to memory he does not own). Even worse, it changes the logic behind the forced init (no reconnect, anymore). Nevertheless, this change does not introduce new possible bugs (as they could be in Soprano::Socket), is straight forward and (most important) those crashes are gone now.
Forgot to mention: These changes mainly prevent kde depend on soprano-beta - it works with unpatched soprano-2.7.6.
