From: Brad K. <bra...@ki...> - 2004-01-12 19:09:20
|
Hello, I've run into problems recently with vnl_math_isnan, vnl_math_isinf, and vnl_math_isfinite on several platforms. These functions use the isnan, isinf, and isfinite macros defined by <math.h> in only some cases. The hand-written implementations are working on fewer and fewer platforms as long double gets bigger. I would like to use VXL_C_MATH_HAS_FINITE to switch on/off using the isnan, isinf, and isfinite macros without other checks. Does anyone know if this is safe everywhere? There seems to be alot of preprocessor logic to support older platforms. I'm curious as to the history of these implementations because I do not want to blow away years of accumulated knowledge in trying to fix these functions. Can anyone shed light on this? Thanks, -Brad |