David Laight wrote: > IIRC something in the standard requires the result to be a compile time > constant
You're right: "... which expands to an integer constant expression that has type size_t, the value of which is the offset in bytes ...". > I suspect uses like the above were not thought of when the standard > was written! If they allowed sizeof to be a run-time value for VLAs they would probably allowed such use of offsetof if they thought of it ... BTW, I read the standard in a hurry yesterday, member-designator is later used in the text: "... then the expression &(t.member-designator) evaluates to an address constant." Alex
