IEEE floating point has a special zero value. See http://stevehollasch.com/cgindex/coding/ieeefloat.html in the "Special Values" section. You can compare against this value, and (a - a) = 0.0f. You do lose that precision when you introduce multiplication and division...so you might end up with ((b * a) / b - a) != 0...which is why everyone warns against comparing with zero. Nevertheless, 5.0f / 0.0f will still trigger a divide by zero error, and that can be avoided by checking the denominator against 0 (which is what this patch does).
On Apr 11, 2005 1:06 PM, Marcus Meissner marcus@jet.franken.de wrote:
On Mon, Apr 11, 2005 at 12:47:22PM -0700, Daniel Remenak wrote:
Changelog: Return DISP_E_DIVBYZERO instead of crashing when asked to divide a variant by zero.
You cannot really compare floats against 0 I think.
Ciao, Marcus