Implicit conversion from float -1.0396039603960503 to int loses precision