On Wed, 25 Jun 2008, Morten Werner Forsbring wrote:

==12662== Invalid read of size 1
==12662==    at 0x80778C8: (within /usr/sbin/cfagent)
==12662==    by 0x8077A8A: (within /usr/sbin/cfagent)
==12662==    by 0x807A48E: (within /usr/sbin/cfagent)
==12662==    by 0x8053AC0: (within /usr/sbin/cfagent)
==12662==    by 0x8053ECC: (within /usr/sbin/cfagent)
==12662==    by 0x431A44F: (below main) (libc-start.c:222)
==12662==  Address 0x0 is not stack'd, malloc'd or (recently) free'd
==12662==
==12662== Process terminating with default action of signal 11 (SIGSEGV)
==12662==  Access not within mapped region at address 0x0

Hi,

and sorry for this late reply. Have you tested if this is improved
with 2.2.7-1 of cfengine?

I don't see it on the servers (amd64), but it persists on the
client (x86-32) boxen - now with about the same regularity (ie, not
every single run, but a couple times a day) - so much better than
before :)

The work clients get this:
Jun 26 00:31:37 corkscrew kernel: cfagent[18939]: segfault at 0 ip
00000000 sp bff455bc error 4 in cfagent[8048000+b3000]
Jun 26 01:01:46 corkscrew kernel: cfagent[24563]: segfault at 0 ip
00000000 sp bfd19b8c error 4 in cfagent[8048000+b3000]
and
Jun 25 17:31:21 gothic-ave kernel: cfagent[8895]: segfault at 1f201 ip
0001f201 sp bfe524ac error 4 in cfagent[8048000+b3000]
Jun 25 18:01:37 gothic-ave kernel: cfagent[15938]: segfault at 1f201 ip
0001f201 sp bfb0a16c error 4 in cfagent[8048000+b3000]

At home, I see this instead:
Jun 26 00:31:21 sparks-ave kernel: cfagent[24499]: segfault at dbdbdbdb
ip dbdbdbdb sp bfa5c0cc error 15
Jun 26 01:01:28 sparks-ave kernel: cfagent[27125]: segfault at dbdbdbdb
ip dbdbdbdb sp bf830eac error 15

The shining light, if there is one is that they problems look to be
always the same failure (address/ip) - looks to be a bad branch ?

I can run a bit via valgrind if it'll help

--
Rick Nelson
* |Rain| prepares for polygon soup
<|Rain|> sweet merciful crap, it works?
* |Rain| faints



--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Reply via email to