SIGPIPE with nepomukservicestub

[Daniel Franke]

Since a few days nepomukservices/virtuoso-t went haywire, used 100% CPU time, hour-long(!) KDE startup time, kontact not starting due to strigi being "Not available" (although it was supposed to be started according to kcm), the whole shebang.

After renaming the folder .kde/share/apps/nepomuk and restarting things worked for a while. Now I realized that it went bad again; `strace -p <nepomukservicespid>` shows gazillion lines of:

write(2, "\"00:00:50 Reference to page with"..., 75) = -1 EPIPE (Broken pipe)
--- SIGPIPE (Broken pipe) @ 0 (0) ---

written out as fast as it can, resulting in:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
 1969 daniel 39 19 79008 15m 12m R 98 0.5 1549:28 nepomukservices
 1985 daniel 39 19 44180 29m 4264 S 73 0.9 1138:31 virtuoso-t

Fun fact: I customized my folders to index to _none at all_ in the Nepomuk System Settings (see also #584341).

This KDE bug may or may not be related: https://bugs.kde.org/show_bug.cgi?id=291001 (nepomuk broken with soprano 2.7.4)
My file system contains multiple NFS and one fusefs/ssh mount.

I have installed:
  libsoprano4 2.7.4+dfsg.1-0ubuntu0.1
  kde-runtime (nepomukservicestub) 4:4.7.4-0ubuntu0.1

[Daniel Franke (franke-daniel)]

Btw, the quoted virtuoso-t process does:

write(3, "00:48:32 Reference to page with "..., 71) = -1 EFBIG (File too large)
write(2, "00:48:32 Reference to page with "..., 71) = -1 EPIPE (Broken pipe)
--- SIGPIPE (Broken pipe) @ 0 (0) ---

I found an exactly 2GB (2147483647) .log file in the renamed nepomuk folder. Deleted it all now. Let's see if it makes a difference ...