Re: Avoid C stack overflow

["Stephen J. Turnbull" <stephen@xemacs.org>]

Stefan writes:

 > > Emacs already has a stack-overflow prevention mechanism, right?
 > 
 > Yes.
 > 
 > > And we're talking about what to do when the user deliberately disables
 > > it? So the simplest answer is "don't do that".
 > 
 > Actually, when we bump into something like max-lisp-eval-depth, we
 > could look at the actual C stack size to compute the average "C stack

Surely you want a worst-case-experienced-in-practice estimate here.
So you should collect data on *variation* of the ratio of C stack size
to max-lisp-eval-depth as well.

N.B. Unless the stack is extremely well-behaved across Lisp programs,
you're computing max-lisp-eval-depth to stack size for the case where
you hit max-lisp-eval-depth.  It might very well be that this is quite
different from cases where you don't.  I don't know if this *matters*
in any practical application to Emacs development, probably not.  But
I don't know what other uses people might find for this statistic.

 > usage per Lisp eval depth" and from that compute a approximate upper
 > bound on the maximum safe value of max-lisp-eval-depth.