Re: Unstable pre-release version of enhanced ai_plan_research

[Eric McDonald <mcdonald at phy dot cmich dot edu>]

Lincoln Peters wrote:

> 	uWeight = malloc( 2 * numAdvances );


I would suggest:
  uweight = (int *)xmalloc(num_advances * sizeof(int));
Xconq's 'xmalloc' checks to see if the 'malloc' succeeded. If it failed, 
then 'run_error' is invoked. 'xmalloc' also scrubs the allocated memory 
to be 0 everywhere.
Furthermore, the problem could be here, because you allocate an array of 
2-byte slots, but you declared the storage type as an 'int' which is 
likely 4 bytes on your system. So, if 'numAdvances' is 4, then you 
allocate 8 bytes, but when you do 'uWeight[3]', you are accessing the 
12th through 15th bytes! Ooops.

> 	for_all_advance_types(count1) {
> 		uWeight[count1] = 0;

The problem could be here. Your 'count1' variable iterates from 0 to 
'numatypes', but the array is only allocated with 'numAdvances' 
positions. 'numAdvances' <= 'numatypes'. You have to worry about the < 
case, because that is a potential source of your crash.

> 				} else if ( ua_multiply_production( count3, count1 ) < 1 ) {
> 					uWeight[count2] -= 1.0 / ua_multiply_production( count3, count1 );
> 				}

The above snippet is a potential divide by zero case. Also, the 
righthand side evaluates to a floating point value, whereas the storage 
on the lefthand side is of an integer type. Furthermore, a factor of 1.0 
is internally 100 in Xconq, so to get a reciprocal, you can do:
  uweight[count2] -= 10000 / ua_multiply_production(count3, count1);
				
Eric