Implicit conversion from float -39.9625468164794 to int loses precision