On Thu, 11 Nov 2010 20:45:24 +0000, Henrik "Størner" wrote:
In <018528ADB17A124D84DD0D081A4C82D22B5C61FB at exchange.geosoft.com> Jason Chambers <Jason.Chambers at geosoft.com> writes:
- Program crashed
Fatal signal caught!
Anyone know how I can resolve this issue? The tail end of rrd-status.log lo= oks like this:
2010-11-11 14:43:34 RRD error updating /home/hobbit/data/rrd/xxx.yyy.com/di= sk,H.rrd from 192.168.0.0: This RRD was created on another architecture
The funny thing is, the architecture never changed. I just upgraded Fedora = Core from version 13 to 14.
Could it be an upgrade from a 32-bit to 64-bit version of the RRD library?
I think the problem has been found, but it might be difficult to fix. Look at the Bugzilla discussion in the Fedora mailing list: https://bugzilla.redhat.com/show_bug.cgi?id=638477#c29
Apparently the glibc folks implemented a new version of the glibc routine that does memory-copies fin Fedora 14. And unlike the old version, this one corrupts data when copying between memory areas that overlap.
Stricly speaking, doing that is not valid. But it is apparently quite common - the Bugzilla entry lists quite a few apps that break because of this change - and it has worked until the change in glibc. So the glibc maintainers is taking the "not a bug - won't fix" stance.
In which case the problem has to be fixed in all of the applications and libraries that do the invalid memcpy, i.e. the rrdtool library in this case.
Regards, Henrik