Implicit conversion from float -8.360655737704917 to int loses precision