Implicit conversion from float -12.02205882352942 to int loses precision