Implicit conversion from float -1.5999999999999943 to int loses precision