[kaffe] stack size problems

[Adam Heath]

On Tue, 30 Nov 2004, Riccardo wrote:

> Hey,
>
> Since I always got SOF on my darwin box during KJC run for class library
> build, I decided to rise it gradually.
> I was able to determine that for PPC, Interpreter and Jthreads, 800k is
> enough (and up of course) while 700K and below crashes.
> I wonder how these requirements would compare for Interpreter, Jthreads
> on other platforms. Up to not long ago, say 1-2 months ago, the 500K
> standard from kaffe where enough.
>
> On Pthread, intrp, I also needed to increase stack size and 800k work.
>
> what strikes me is that on UltraSparc using JIT and Pthreads, the
> standard stack size of 500K is ENOUGH and build completes perfectly! So
> now, what?!
>
> Is interpreter broken ? Very inefficient?

The stack requirements have grown.  This is standard.

dpkg had a bug filed against it; perl 5.6 was causing a regex to fail in
install-info, while perl 5.005 let it succeed.  5.6 was using more stack
space, and so hit the limit sooner.

The problem was the code, even tho it was technically correctly.

Of course, I don't know if kcj could be rewritten to not be so recursive, or
kaffe could be tweaked to use less stack space.