Implicit conversion from float -40.35714285714286 to int loses precision