Implicit conversion from float -5.42334096109839 to int loses precision