Hi Mark,
Thanks for your reply.
After attaching to imapd with tusc (system call tracer) it appears that
child (slave) process dies with SIGBUS.
Further troubleshooting led me to line 979 of src/osdep/unix.c.
if (!tstream->rdonly) ((UNIXLOCAL *) tstream->local)->dirty = T;
Slave process seems to die when it hits this line.
Any idea why this happens?
Regards,
Ermin
Mark Crispin wrote:
On Fri, 10 Nov 2006, Ermin Borovac wrote:
The current command did not succeed. The mail server responded: Execution
process terminated abnormally (a).
This is not good. This indicates that the slave process created to do the
operation exited due to signal 0xa which is ECHILD on most systems. Put
another way, the master process did a waitpid() on the slave, and got 0xa
back in the status. The expected value is 0x400; the slave process
returns an exit code of 1 if all is well.
I don't have a clue as to how/why this happened; nor why other SVR4 sites
have not reported it (at least, not yet).
What's bizarre is that the master/slave protocol negotiation on the pipe
was successful; and apparently the message was copied successfully. The
only thing that seems to have failed was the slave passing back its exit
status (0 if failure, 1 if success) back to the master, and the master
reaping the slave.
This is in the code to work around fcntl() locking which SVR4 systems
(such as Solaris, HP-UX, and AIX) force you to use since they don't have
(true) flock(). fcntl() locking releases a lock whenever any file
designator open on the file is closed, regardless of whether the lock was
applied on a designator that is still opened.
If you want to try debugging, see if you can get a protocol transcript
from Thunderbird so you can see what it is doing. Try to repeat the same
in a manual run of imapd. The idea is to see if you can reliably
reproduce the problem.
Note that if you run imapd from the shell, you can do IMAP protocol
commands without having to authenticate (you'll be preauthenticated as
yourself). The routines in question are master() in flocksim.c which does
the entire master/slave interaction, and grim_pid_reap_status() in
gr_waitp.c which does the waitpid() call.
Alternatively, you can give up on SVR4 and switch to a Linux or BSD system
for your IMAP service. Such a system would be less expensive than your
HP-UX system, and would not have any of this flocksim nonsense. The
flocksim code is specific to the SVR4 port.
-- Mark --
http://panda.com/mrc
Democracy is two wolves and a sheep deciding what to eat for lunch.
Liberty is a well-armed sheep contesting the vote.
_______________________________________________
Imap-uw mailing list
[email protected]
https://mailman1.u.washington.edu/mailman/listinfo/imap-uw
