Implicit conversion from float -0.49881235154393266 to int loses precision