These functions replace the sign bit of x with the sign bit of y.

double copysign(double x, double y);
float copysignf(float x, float y);

These functions treat all floating-point numbers as signed (including zeros and NaNs) and do not cause any errors or exceptions.