Implicit conversion from float -39.46927374301677 to int loses precision