[GRASS5] PATCH: alloc_cell.c

[Glynn Clements]

Brad Douglas wrote:

> Any objections to me applying this patch to keep consistent with size_t
> *alloc() types?

Have you verified that nothing which calls G_raster_size() is relying
upon the return value being signed?

The issue is that if either operand to an arithmetic operation is
unsigned, the result is unsigned. This could be a problem if something
is multiplying a potentially-negative index by the result of
G_raster_size(), particularly if the result is promoted (explicitly or
implicitly) to a long on platforms where long is 64 bits:

	$ cat foo.c

	#include <stdio.h>
	
	int main(void)
	{
		int x = -5;
		int k1 = 2;
		unsigned int k2 = 2;
		/* my longs are only 32 bits, so use long long instead */
		long long y1 = x * k1;
		long long y2 = x * k2;
	
		printf("y1 = %lld\ny2 = %lld\n", y1, y2);
	
		return 0;
	}
	
	$ cc foo.c
	$ ./a.out
	y1 = -10
	y2 = 4294967286

-- 
Glynn Clements <glynn at gclements.plus.com>