occasional segmentation fault from rpm -V

occasional segmentation fault from rpm -V

Post by Robert M. Riches J » Mon, 30 Jun 2003 14:08:28



To quote Paul Harvey, "Here is a strange:"

After updating to XFree86 4.1.1-20 and the other updates
from the past week, two of my systems hung while doing the
'rpm -Uvh ...' command issued by autorpm.  I did the usual
'rpm --rebuilddb' thing and used 'rpm -e ...' to remove the
old packages that should have been removed as part of the
update.  Then, I tried my shell script that does 'rpm -V' on
each package on the system (one package at a time).  After a
total of two rounds of 'rpm --rebuilddb', the first system
passed.  The other system still gets segmentation fault on
10 of the roughly 1700 packages, and it's the same 10
packages every time.  However, if I manually do 'rpm -V ...'
on those 10 packages or on each one by itself, they all
work.  I made the script detect the segmentation faults,
sleep 10 seconds and try again, and 8 of the 10 work the
second time around.

I have done the 'rpm --rebuilddb' routine several times, and
the same packages keep getting Segmentation faults.  The
fact that it's always the same 10 indicates it's not random
hardware flakiness.  However, I'm puzzled why the same
command a second time would work when the first one failed,
unless there's a time-sensitive race condition in the
program.

I filed a bug reoprt in Redhat's bugzilla (98248), complete
with a core file.

Any other suggestions?  (... other than digging into the
source myself, of course, which would be fun if I could
afford the time...)

Thanks.

Robert Riches

(Yes, that is one of my email addresses.)

 
 
 

occasional segmentation fault from rpm -V

Post by dnoy » Mon, 30 Jun 2003 22:52:42


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


| To quote Paul Harvey, "Here is a strange:"
|
| After updating to XFree86 4.1.1-20 and the other updates
| from the past week, two of my systems hung while doing the
| 'rpm -Uvh ...' command issued by autorpm.  I did the usual
| 'rpm --rebuilddb' thing and used 'rpm -e ...' to remove the
| old packages that should have been removed as part of the
| update.  Then, I tried my shell script that does 'rpm -V' on
| each package on the system (one package at a time).  After a
| total of two rounds of 'rpm --rebuilddb', the first system
| passed.  The other system still gets segmentation fault on
| 10 of the roughly 1700 packages, and it's the same 10
| packages every time.  However, if I manually do 'rpm -V ...'
| on those 10 packages or on each one by itself, they all
| work.  I made the script detect the segmentation faults,
| sleep 10 seconds and try again, and 8 of the 10 work the
| second time around.
|
| I have done the 'rpm --rebuilddb' routine several times, and
| the same packages keep getting Segmentation faults.  The
| fact that it's always the same 10 indicates it's not random
| hardware flakiness.  However, I'm puzzled why the same
| command a second time would work when the first one failed,
| unless there's a time-sensitive race condition in the
| program.
|
| I filed a bug reoprt in Redhat's bugzilla (98248), complete
| with a core file.
|
| Any other suggestions?  (... other than digging into the
| source myself, of course, which would be fun if I could
| afford the time...)
|
| Thanks.
|
| Robert Riches

| (Yes, that is one of my email addresses.)

Perhaps your script is not releasing something that is causing the
eventual seg fault!?  What if you made the script run in reverse or skip
a few of the non-seg-fault packages here and there, would it still seg
on the same packages?

- --
L8r,

C.L. Gilbert
For a free Java interface to Freechess.org see
http://www.rigidsoftware.com/Chess/chess.html

"Verily, verily, I say unto you, He that entereth not by the door() into
the sheepfold{}, but climbeth up some other *way, the same is a thief
and a robber."  John 10:1

GnuPG Key Fingerprint:
82A6 8893 C2A1 F64E A9AD  19AE 55B2 4CD7 80D2 0A2D
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQE+/u8qVbJM14DSCi0RAmPvAKC1tREWNoM4C9+dsnF/Mj3dPSRDjQCg6QkR
JuSFHol8dpj5jeqbMoT5F7c=
=imtg
-----END PGP SIGNATURE-----