Re: Hangs [ was Re: Patch 411 ]

[Dan Nelson]

In the last episode (Jun 04), Jeff Jirsa said:
> I'm still seeing serious instability in 0.1.3a. I'm new to this list,
> but having checked most of the archives on Savannah, I still don't
> see any solution to the hang listed in various bugs. I've changed the
> close(fd++) call in closeall() to close-on-exec (suggestion
> http://mail.gnu.org/archive/html/spamass-milt-list/2002-11/msg00005.html
> ) and the crashes are down to twice a day,
> 
> I believe it's similar to
> http://savannah.nongnu.org/bugs/?func=detailbug&bug_id=2403&group_id=1083
> http://savannah.nongnu.org/bugs/?func=detailbug&bug_id=1099&group_id=1083,
> http://savannah.nongnu.org/support/?func=detailsupport&support_id=101558&group_id=1083

I think those those are all crash bugs, not hangs.  Bug 1099 does
provide a gdb backtrace but there's no useful data in it (i.e. none of
the stacks include any milter functions).

I think the list archive thread you referenced was due to a hang when
multiple forked processes still had libmilter socket connections open,
so that really shouldn't affect you if your process is really crashing.

If you can get a coredump and a good stack trace, that'd be great. 
Also if it crashes a lot, try runing it in the foreground under
valgrind and see if it sees anything.

-- 
        Dan Nelson
        address@hidden