Ok! I did it… let’s see…
“Better” is relative, and highly subjective… 
My solution basically “deletes” systemd’s shipped 50-coredump.conf and therefore leaves the “kernel.core_pattern” setting at the default before systemd changed it to call systemd-coredump.
Which means, coredumps will still be generated unless ulimit’s “core file size” is 0, which it is by default though.
In other words, you can still set the ulimit differently (even for single processes) if you do want to have coredumps.
Setting kernel.core_pattern=/dev/null will cause coredumps being sent to /dev/null, i.e. discarded in general, regardless of ulimit.
The default (without systemd) is kernel.core_pattern=core btw, so you could also explicitly set this if you want (or use any other file name/pattern you prefer).
The point is that tracker-extract uses it though, and crashes while doing so.
If I understand the bug report correctly, it’s caused by seccomp (a security feature).
In any case, it’s a bug that should be fixed.
Forgot to mention that you should run “mkinitrd” after the change.
These settings are applied very early during boot when the root partition is not mounted yet.
I have tried the wolfi323 tip but when I turned on my computer again there was some error. I don’t know if was related to that, but I removed this solution.
Today there was an update in systemd in the system.
And what error?
Unlikely related to the change though.
Today there was an update in systemd in the system.
Yes, that fixes e.g. USB sticks being misidentified as CDs when you put an ISO on them and repartition it afterwards.
http://bugzilla.opensuse.org/show_bug.cgi?id=1046268
It won’t change anything regarding your problem though.
Again, your problem is that tracker-extract is crashing, it has nothing to do with systemd, so it won’t be fixed by a systemd update either.
It’s just that systemd-coredump generates coredumps from the crashes (because that’s how it’s configured by 50-coredump.conf), and that causes CPU time as well (because it compresses the data I think).
And a modified 50-coredump.conf will prevent systemd-coredump from being started, but it won’t fix the crashes (though you probably won’t notice them any more then).
I don’t know exactly but I needed to restart and choose a boot in recovery mode.
But anyway, using this tip http://forums.fedoraforum.org/showthread.php?t=303659 seems to be ok for now.
Really thank you.
wolfi323 wrote:
> “Better” is relative, and highly subjective… 
Thanks for the information’s and i have no problem to say that your solution is
better because you can still use ulimit. But nice to know that my solution
disable coredump completely 