a bit of disaster - anybody knows what is going on?

[John Ellson]

Michal Jaegermann wrote:

>After today series of updates my x86_64 test box started seriously
>misbehave.  For example:
>
># iptables-restore /etc/sysconfig/iptables
>*** glibc detected *** iptables-restore: free(): invalid pointer:
>0x00002aaaaaac7150 ***
>Aborted
>
># runuser -s /bin/bash - nobody -c mDNSResponder
>*** glibc detected *** runuser: double free or corruption (out):
>0x00002aaaaaac7650 ***
>
># service xfs status
>xfs dead but pid file exists
>
>and many other things.  Some fail silently, other will complain
>like above.  Pointers indeed look a bit strange.
>
>Interestingly enough /sbin/iptables-restore was installed
>on 2004-Nov-12 and 'ldd /sbin/iptables-restore' shows only
>
>	libdl.so.2 => /lib64/libdl.so.2 (0x00000033f4b00000)
>	libc.so.6 => /lib64/tls/libc.so.6 (0x00000033f4600000)
>	/lib64/ld-linux-x86-64.so.2 (0x00000033f4400000)
>
>wher all three pieces were installed yesterday afternoon and a
>machine was running, and rebooting, a few times without any
>ill-effects until new rawhide packages were retrieved. 'rpm -V' does
>not report any troubles neither for 'iptables' nor for 'glibc'.
>
>A sudden hardware failure does not seem to be likely as another
>Linux installation on the same box runs just fine.  A swap space is
>common to all installations.  Booting older kernels also does not
>change anything.  Today packages look like the most probably cause
>but I do not see on that list anything obvious.  Oh, and x86 test
>system, updated to the same level, is not doing anything strange.
>
>I would file a bug report but I am somewhat at loss about what.
>Anybody seeing something similar?
>
>   Michal
>
>  
>
I had very similar symptoms today.    I spent half the day tring to fix 
it on my x86_64 box.

I _think_ I ultimately fixed it by reinstalling all of glibc but I can't 
be certain.

I spent a lot of time poking around the xfs problem, but I don't think I 
changed
anything related to fonts in the end.

On a possibly related note:  I'm getting all kinds of %postun errors 
when trying to uninstall rpms
on various i386 boxes.    And I'm having to "rpm --rebuilddb" on one box 
because it rpm claims that
some unknown device is running out of space when writing 
/var/lib/rpm/Packages, even though I have 3G available.

I wonder if there is something in rpm or in rpm %pre/%post scrpts that 
is broken?

John