[lug] Process dying without core dump

[Alan Robertson]

Sean Reifschneider wrote:
> 
> On Wed, Sep 15, 1999 at 01:03:16PM -0600, Alan Robertson wrote:
> >The symptoms are that one process of 4 (all forked w/o exec) dies
> >without leaving a core file.
> 
> Are you sure that you have the limits set up to LEAVE a core file?
> 
>         sylvia:nospam$ ulimit -a
>         core file size (blocks)  1000000
>         [...]

Looks like I do:
	core file size (blocks)  1000000

Or, running from sudo (like I usually do):

>	[alanr at just 6538]$ sudo sh -c ulimit -a
>	unlimited


I tried looking in /proc, but didn't see anything there that obviously
told me what the actual running process' ulimit was...

> Some people like to set that to 0 as a default to prevent leaving core
> files laying around.
> 
> You might want to try making a version that doesn't fork, so you can
> run from the command-line and see what's happening, and/or start
> catching signals and report that (the latter probably being the
> better idea).

It has to fork.  It needs 4 processes.  The one that's dying isn't the
parent of the others, it's one of the children.  But, catching signals
sounds better.  Is there a way to *force* a core dump?  For example,
would abort(3) work?  I suppose I could just block all signals and do a
pause, and wait for some human to come look me over too...

	Thanks!

___________________________________________________
Alan Robertson		mailto:alanr at bell-labs.com
            http://www.henge.com/~alanr/
___________________________________________________