Subject: Re: DIAGASSERT in cryptographic functions.
To: None <tech-userlevel@netbsd.org>
From: Frank van der Linden <frank@wins.uva.nl>
List: tech-userlevel
Date: 09/17/1999 09:58:58
On Fri, Sep 17, 1999 at 12:02:11AM +0200, Christoph Badura wrote:
> ISTR Luke saying something about not wanting to have critical programs dump
> core on him.  One possible solution is to fork() and let the child dump core.

Well, I hope Luke will speak up about this himself, but... If you have
a critical program, and you link it with the DIAG library, that would
imply to me that you suspect something is wrong. So it's better to
have it bail out at that point, leaving evidence, then having it
continue in an unknown state, which is certainly not what you want
from a "critical program".

In any case, continuing where it isn't expected, leading to undefined
behaviour, is never a good thing.

- Frank