The fmod man page recommends this code to get the least positive residue:
z = fmod(x, y);
if (z < 0)
z += y;
That last line should be "z += fabs (y);" Otherwise, for x=-0.25 and
y=-1 you get z=-1.25 which isn't what anyone is looking for.
The man page could perhaps also state that the sign of y has no effect
on the output, other than possibly when the result is a NaN.
M.
