Implicit conversion from float -0.8888888888888857 to int loses precision