Implicit conversion from float -0.5999999999999943 to int loses precision