Scheduler suddenly crashes #1417
Comments
|
CUPS.org User: mike From the logs, both clients are running some version of CUPS 1.2. I'll do some testing here to see if I can duplicate it... |
|
CUPS.org User: till.kamppeter The poster of the Mandriva bug report has posted lots of logs and config files now. It seems that the crash of the client is not caused by queues on the serevr being removed or shut down. The client times out the remote queues and crashes by itself, without any action coming from the server. The scheduler and the queues on the server are still living in peace. I have also shut down my server (unfortunately only Mandriva Corp Desktop 3.0 with CUPS 1.1.23) intendedly to try to crash my client (Cooker with CUPS 1.2 rev 5083) but the queus on the client disappeared normally after 5 minutes and the client stayed running. |
|
CUPS.org User: mike I wasn't able to duplicate it on my 1.2 system, either. The best thing would be for them to run cupsd on the client with valgrind: and then post the results. |
|
CUPS.org User: woojinlee till@mandriva.com asked me to post the result of valgrind /usr/sbin/cupsd -f from the cups client machine. Here it is. ==4129== Memcheck, a memory error detector. ... (I missed whole bunch of text scrolled by) ==4129== by 0x4126F82: cupsArrayFind (in /usr/lib/libcups.so.2) and this is the end of the /var/log/cups/error_log right after the crash. ... Here, woojin3 is the client machine, and the above above line repeats over and over until when it crashed. Woojin |
|
CUPS.org User: till.kamppeter I have also captured a scheduler crash with valgrind. This one happened directly after starting CUPS. ==406== Memcheck, a memory error detector. |
|
CUPS.org User: till.kamppeter I forgot to say that my valgrind run was on SVN rev 5102. |
|
CUPS.org User: till.kamppeter Here is another segfault (still on rev 5102), caused by a job error: ==4603== Memcheck, a memory error detector. Here the output in error_log (LogLevel debug): E [15/Feb/2006:15:45:24 +0100] [Job 9] Starting renderer |
|
CUPS.org User: till.kamppeter It happened now several times to me that CUPS segfaults shortly after the start. Here I have captured a gdb backtrace with CUPS rev 5106: [root@majax c]# gdb -c core.5101 /usr/sbin/cupsd Reading symbols from shared object read from target memory...done. warning: svr4_current_sos: Can't read pathname for load map: Input/output error Reading symbols from /usr/lib/libz.so.1...done. |
|
CUPS.org User: mike I think there were a few unchecked accesses of job->state which caused things to break - please check the current trunk code and report on any problems you have... |
|
CUPS.org User: till.kamppeter CUPS still segfaulting, but at another place: ==5574== Memcheck, a memory error detector. |
|
CUPS.org User: till.kamppeter This and the previous posting is rev 5108, here a segfault shortly after I had sent a job: [root@majax c]# gdb -c core.5547 /usr/sbin/cupsd Reading symbols from shared object read from target memory...done. warning: svr4_current_sos: Can't read pathname for load map: Input/output error Reading symbols from /usr/lib/libz.so.1...done. |
|
CUPS.org User: mike See if the latest update (r5120) fixes things; if not, please remove the remote.cache and job.cache files from the /var/cache/cups directory and restart cupsd. Thanks! |
|
CUPS.org User: till.kamppeter It seems to work now, thanks. In the case that I find other crashes, I will open a new STR. |
Version: 1.2-current
CUPS.org User: till.kamppeter
See
http://qa.mandriva.com/show_bug.cgi?id=21094
and especially the error_log snippet in the initial posting there. I have checked my test box and observed exactly the same result.
Short time after a broadcasting remote CUPS server is stopped (or a queue removed) the queue disappears from the local CUPS server and in exactly that moment where the remote queue disappears/times out on the local CUPS server, the scheduler of the local CUPS server crashes.
I am not sure which CUPS version the remote boxes are running, can be 1.2.x (on Cooker) or 1.1.x (on any released Mandriva, 2006 or older).
Local OS: Mandriva Cooker, both 32 and 64 bit x86
Local CUPS version: last weekly snapshot, rev 5083
Remote CUPS version: 1.1.x (perhaps also 1.2.x)
The text was updated successfully, but these errors were encountered: