Jason Thorpe wrote:
> On Mon, 8 Sep 1997 22:11:28 +0200 (MET DST) 
>  Christoph Badura <bad@ora.de> wrote:

>  > On the other hand, a process' enviroment includes more then what
>  > environ points to.  E.g. file descriptors and signal settings.
>  > 
>  > I'm just trying to push this idea to its (reasonable) limits.

> I think a reasonable solution might be:

> 	- New sysctl level: kern.coredumpdir

How about kern.coredump or kern.coredumps?  That way we don't get the
ugly kern.coredumpdir.dir variable.  Besides, it doesn't describe the
coredump directory only.  Hmm, then kern.coredumps.format could be
mistaken for the format of the core file itself.  Maybe .spec?

> 		- .enabled

This seems redundant.  I'd say it would be always enabled and the
defaults for .dir and .format are "." and "%n.core" respectively.
If you want to separate the directory from the format, that is.
Since this isn't really settable on a pre process basis with sysctl,
we could as well collapse .dir and .format into the a single entry.

Or did I misinterpret your message completely?

> 		- .dir
> 		- .format

> 	- Kernel compile options to set the defaults.
Yes.


-- 
Christoph Badura

Now available in print: Lion's Commentary on UNIX 6th Edition, with Source Code
			http://www.peer-to-peer.com/