Implicit conversion from float -7.445255474452566 to int loses precision