I think the point is that the integer arithmetic implementation your CPU provides is wrong in at least one jurisdiction, so (for example) the machine code in the article is wrong.