[CentOS-devel] Yum futex lockup upgrading from 4.0 to 4.1

Sat Jun 18 03:20:29 UTC 2005
Rodrigo Barbosa <rodrigob at suespammers.org>

Hash: SHA1

Okey, this problem seems to be related to yum.

After installing a new 4.0 machine (desktop, so lots of packages, custom
install), I pointed CentOS.repos to 4.1.

After that, I executed "yum update".

Everything started fine, downloads finished without a problem. During
the transation, tho, I have a hard locked of the yum process,
when finishing updating redhat-lsb. Yum output showed the packted 100%
done, but on /proc the process still had the redhat-lsb rpm packged
opened (filedescriptor). The python process that was running yum
was locked hard on a futex call. Interesting fact to notice that
there were no other yum related processes running, and that 
the running python instance was upgraded earlier during the transation
(so I have python.#prelink# running).

No signals made any different, and I had to issue a -9 to kill the
process (-15 was no good).

I had to do a bit of fixing after that, since it seems python doesn't
handle broken dependencies very well. Nothing a couple of "rpm -e --justdb"
didn't so. Another "yum upgrade" finished upgrading the machine.

All my previous tests were from a fully upgraded 4.0 machine, and for
those I had no trouble. With this one, a stock 4.0 machine with
no upgrades, yum locked. Might be worth trying to reproduce it.
Another interesting fact is that this was my only machine with SELinux
active, but I'm still betting on a python/yum related problem.

All hardware on my machine was fully tested and no problems were
detected. No processor overheat, no bad memory, perfect disks.

Best regards,

- -- 
Rodrigo Barbosa <rodrigob at suespammers.org>
"Quid quid Latine dictum sit, altum viditur"
"Be excellent to each other ..." - Bill & Ted (Wyld Stallyns)

Version: GnuPG v1.4.0 (GNU/Linux)