Re: trace of scheduler core dumps

[Julio Polo <julio@hawaii.edu>]

I used gcc 2.7.2 for Solaris 2.4 recompiled under Solaris 2.5.  I have
encountered several weird problems while configuring and compiling zmailer (see
one of them below) and got it to compile after trying a couple of different
options.  I'll try to recompile gcc for Solaris 2.5 (if there's such a
version) and if the problem persists, I'll provide the debugger info.

Thanks.

-Julio Polo, University of Hawaii

  gcc -g -O -DHAVE_CONFIG_H -I.. -I../include -I../libresolv -Ino  -c  smtpserver.c
  smtpserver.c: In function `smtpserver':
  smtpserver.c:1744: `LOG_WARNING' undeclared (first use this function)
  smtpserver.c:1744: (Each undeclared identifier is reported only once
  smtpserver.c:1744: for each function it appears in.)
  *** Error code 1
  make: Fatal error: Command failed for target `smtpserver.o'
  Current working directory /usr/uh/spkg/zmailer-2.99.38-src/smtpserver


> > The scheduler core dumping problem I reported earlier might have
> > something to do with how the scheduler handles a high number of
> > mail/threads/child processes (?) (sorry, I haven't caught up with how
> > the scheduler works now).  I cut down the number of messages from 179
> > to about 19 and the scheduler has not died on me yet.
> > 
> > I'm including a trace of the scheduler process.  Can someone provide me
> > with some insight as to what might be happening?  I'm running 2.99.38
> > on a Sun Ultra under Solaris 2.5.1.  There are 179 messages in the
> > transport directory, all of them are supposed to smtp to
> > mac.its.hawaii.edu which should be down so I can watch the scheduler
> > do its work.    I am using the default scheduler.conf file.
> 
> 	A bus-error ?   That means access of unaligned data, which likely
> 	means access of invalid data.
> 	What compiler did you use ?
> 
> > If someone can confirm it's a bug and not something of my doing, then
> > I can stop worrying about it.  Please reply to julio@hawaii.edu as I'm
> > not in this mailing list.  Thanks.
> 
> 	Could you run this same with the scheduler on some debugger ?
> 	(So that we could get a backtrace to see where it crashes.)
> 
> > -Julio Polo, University of Hawaii
> 
> 	/Matti Aarnio <mea@nic.funet.fi> <matti.aarnio@tele.fi>
>