Hi Piotr,

Please don't send unrelated changes in one patch, please split it into 2 patches.

Done. Thanks :)
 

  float CDECL MSVCRT__copysignf( float num, float sign )
  {
-    /* FIXME: Behaviour for Nan/Inf? */
-    if (sign < 0.0)
-        return num < 0.0 ? num : -num;
-    return num < 0.0 ? -num : num;
+    /* FIXME: Behaviour for signbit(NAN) is different in Linux and
+     *        Windows, where Windows gives a zero for -NAN
+     */
+    if (signbit(sign))
+        return signbit(num) ? num : -num;
+    return signbit(num) ? -num : num;
  }
 Why don't you add an isnan() check? Is there any reason for returning "signbit(num) ? -num : num" instead of returning num?

Linux and Windows treats NAN differently and I am not quite sure which I should follow here, so I just leave it for now...
I think the _copysignf() is supposed to return a (-num) when (num < 0 && sign >0), so I think this "signbit(num) ? -num : num" might be right.
 
It would be nice to also change signbit definition on systems that doesn't support it, so it at least works for normal numbers.

The signbit() function is actually part of C99, not my implementation, so I think it should be supported on most systems? Though I am not quite sure...

 
Thanks,
Kevin