Implicit conversion from float -0.26946107784431206 to int loses precision