mcdonald@uxe.cso.uiuc.edu (09/05/88)
>The example everybody was citing when it came to proving that >Numerical Recipes was non-portable code (and yes, you have convinced >me, never worry about that) was the Intel 80*86. >Am I missing something, or was the 80*86 an invalid counterexample? The 8086 as implemented by the Microsoft and Turbo C compilers is an invalid counterexample for the exact problem being discussed. But, given the 80x86 family, there are cases at the bounds of the 1 meg address space where various members of the family (i.e. 8086 vs 80386) will give differing results. But it is hard to think of an example where on would use arrays wrapping around (as opposed to pointers pointing to specific things near the ends). Doug McDonald