Implicit conversion from float -2.352941176470594 to int loses precision