CGFloat is a float on 32-bit architectures and a double on 64-bit architectures. This makes the use of fabsf inappropriate on 64-bit arm. But changing the function to fabs would be inappropriate for 32-bit targets (making #434 incorrect).

I propose using the preprocessor macro ABS to avoid the problem altogether.

The reason for such a change is to help the use of this library on projects where strict static analysis is enforced.

A longer, more explicit architecture-dependent approach is also possible.

#if CGFLOAT_IS_DOUBLE
    CGFloat fullWidth = fabs(finalLeftEdge - initialLeftEdge);
#else
    CGFloat fullWidth = fabsf(finalLeftEdge - initialLeftEdge);
#endif

This may be preferable in cases where the distinct functions are known to have some advantage. However absolute value is a simple enough function that the macro's behavior will be identical.