Re: [OT] Bonehead question about coredumps

["Dmitri Nikulin" <dnikulin@xxxxxxxxx>]

On 5/2/06, walt <wa1ter@xxxxxxxxxxxxx> wrote:
> When a userland program segfaults, what determines whether
> it also dumps core?  Is there something either the programmer
> or the user can do to force a cordump on segfault?
>
> Thanks for any clues!

Some security-conscious programs will setrlimit() themselves to
prevent dumping core by default. This prevents cryptographic keys and
other sensitive data hitting a hard disk platter or possibly
world-readable directories, and so it's a nice partner to mlock().

So even if it seems like all the cases in core(5) should check out
okay, the program itself has free reign to change them in a way that
is largely transparent to the administrator. Check their source if you
just don't see what's happening (and if you're doing this for
debugging, you may as well just debug the whole program itself and
preclude the need to dump core automatically).

Hope this helps.

 -- Dmitri Nikulin