Re: [gnugo-devel] stack overflow

[Arend Bayer]

On Sat, 22 Feb 2003 address@hidden wrote:

> 
> > FYI, the new new_semeai() function allocates 1,5 Mb on the stack. The VC++
> > build doesn't seem to be up to this, and crashes as soon as the function
> > is called.
> > 
> > /nando
> 
> Sorry. We need to allocate an array based on how many dragons 
> there are or guess a safe upper bound. I guessed 361. This
> makes the array size 361^2 and I guess that's too much.

> Index: engine/semeai.c
> ===================================================================
> RCS file: /cvsroot/gnugo/gnugo/engine/semeai.c,v
> retrieving revision 1.50
> diff -u -r1.50 semeai.c
> --- engine/semeai.c   21 Feb 2003 23:39:50 -0000      1.50
> +++ engine/semeai.c   22 Feb 2003 17:16:58 -0000
> @@ -94,7 +94,7 @@
>   * owl_analyze_semeai() instead of relying on static analysis.
>   */
>  
> -#define MAX_DRAGONS MAX_BOARD*MAX_BOARD
> +#define MAX_DRAGONS 50
>  
>  void
>  new_semeai(int color)
> @@ -105,8 +105,10 @@
>    int d1, d2;
>    int k;
>    int other = OTHER_COLOR(color);
> +  int num_dragons = number_of_dragons;
>  
> -  gg_assert (number_of_dragons <= MAX_DRAGONS);
> +  if (num_dragons > MAX_DRAGONS)
> +    num_dragons = MAX_DRAGONS;
>  
>    for (d1 = 0; d1 < number_of_dragons; d1++)
>      for (d2 = 0; d2 < number_of_dragons; d2++) {

I suppose you want to change all occurences of number_of_dragons below to
num_dragons? The way it is, it looks like a (pretty unlikely) array bound
violation to me if number_of_dragons > MAX_DRAGONS.

Could we, instead, maybe just use a single array for all semeai fights?

Arend