Implicit conversion from float -5.105421686746993 to int loses precision