Hi, I've been developing my own collision detection algorithm and have finally managed to pinpoint where it is going wrong. When running the following function, MSVC++ 6.0 returns the answer 10.3 which is clearly wrong, as any of you can check using a calculator:

  
#include <iostream.h> 
#include <math.h> 
  
float a, b, c, d, e, f, tempa, tempb, tempc;
float BadFunction (void)
{
  a = 9.0f;
  b = 4.0f;
  c = 5.0f;
  d = 4.0f;
  e = 7.0f;
  f = 5.0f;
  tempa = a-d;
  if (tempa==0.0f) tempa = 0.000001f; // without this, undefined result

  tempb = b-e;
  if (tempb==0.0f) tempa = 0.000001f; // without this, undefined result

  tempc = c-f;
  if (tempc==0.0f) tempa = 0.000001f; // without this, undefined result

  tempd = (tempa*tempa)+(tempb*tempb)+(tempc*tempc);
  return tempd
}
  

The function should return 25+9+0 which is... 34, right? Well it returns 10.3.... It was acually giving me an undefined result error before I put in the "if (tempb==0.0f) tempa = 0.000001f;" fix. And now it gives me a WRONG answer. How can MSVC++ be so bad at addition and subtraction? Am I using math.h wrongly? Is there another better math header file out there? Thanks for your comments. Keer Edited by - Keermalec on May 3, 2001 11:57:37 AM

Well, since 5-5 = 0 , wich is: tempc = c - f , and you then have:

if(tempc == 0.0f)
tempa = 0.000001f;

It''s not VC++ who can''t calculate, it''s your type-o

since, (0.000001*0.000001)+(-3*-3)+(0*0) = 9.something extreamly small

I don''t know how you get it to 10.3 though..





return 1;

OMG you are so right! My type-0!

Thank you so much for pointing this out.





Edited by - Keermalec on May 3, 2001 3:03:18 PM