On Tue, Aug 16, 2011 at 08:18:48PM +0400, Dmitry Bely wrote:
> On Tue, Aug 16, 2011 at 8:10 PM, Richard W.M. Jones <rich@annexia.org> wrote:
> 
> > Well it would certainly help if we had a piece of runnable code to
> > look at.  The code supplied in the original email contains an infinite
> > loop.
> 
> Yes, that was my mistake that I corrected later. The loop should be
> 
>  for (; *s != NULL; s++) {
>     list = wrp_ml_cons(caml_copy_string(*s), list); /* bug! */
>  }

Well, contrary to what I said earlier, it really does seem
like this is undefined C behaviour.

Try the attached program, which distills the problem down to something
simple.  gcc with -O0 and -O2 produces completely different answers,
and the assembler output confirms it.

Rich.

-- 
Richard Jones
Red Hat