that's cool.  Really what I was suggesting was that if the server crashes
and normally would zombify, put an error in the syslog or messages file that
says how to setup for debug information that you're after.

Thanks Jan!

--
David J. Graff
BroadVision/Atlanta - ISG Consultant
mailto: dgraff_at_broadvision.com <mailto:dgraff_at_broadvision.com>
Phone: (770) 392-3337 x5551


> -----Original Message-----
> From: Jan Harkes [mailto:jaharkes_at_cs.cmu.edu]
> Sent: Tuesday, December 14, 1999 1:06 PM
> To: codalist_at_telemann.coda.cs.cmu.edu
> Subject: Re: Error while creating rootvolume
>
>
> On Tue, Dec 14, 1999 at 11:22:06AM -0500, David Graff wrote:
> > Jan is there someway that you could change this so that instead of just
> > putting it into an endless sleep could you gracefully exit the
> program with
> > a message on how to set it into the endless sleep (i.e. start with
> > commandline flag or zero length file)?
>
> Hi David,
>
> Well, yes, but it would be rather useless for my purposes ;) A coredump
> would either be too large as needs to dump the mmapped RVM, or would be
> useless whenever it doesn't dump RVM data. And `silently exiting servers'
> are almost as bad as sleeping ones, although a quick `ps' would reveal
> that the server isn't running anymore
>
> The coda_assert function is really quite flexible,
>
>     int coda_assert_action = CODA_ASSERT_SLEEP;
>     #define CODA_ASSERT_SLEEP       1
>     #define CODA_ASSERT_EXIT        2
>     #define CODA_ASSERT_ABORT       3
>     #define CODA_ASSERT_CORE        3
>
> Ok, done, set the default for codasrv to ASSERT_EXIT, and added a
> -zombify flag. I've also placed a test in the startserver script for the
> /vice/srv/ZOMBIFY file. Any suggestions for a better word to describe
> this?
>
> I really liked the couple of good tracebacks I got over the past year.
> Oh well, such is life. Our `sleeping beauty' servers are one of the most
> unexpected things for new users and maybe this will ease the
> transitioning pains a bit.
>
> Jan
>