Implicit conversion from float -12.312138728323703 to int loses precision