On 4 January 2015 at 13:41, Maxim Mikityanskiy wrote: > Hmm, it seems you are right. But if so, I'm writing object "global" of > type int using lvalue of type int, it's "a type compatible with the > effective type of the object", shouldn't the first bullet apply and 2 > be stored into "global"? But param does not point to a struct aggregate, so param->value is not valid.