Implicit conversion from float -5.714285714285708 to int loses precision