Implicit conversion from float -2.941176470588232 to int loses precision