Implicit conversion from float -5.428571428571431 to int loses precision