It's a change between C and C++, where Microsoft previously implemented the C rules. As noted in remyabel's answer, that has since been fixed.
In C, a cast to void
, or simply using an expression as a statement by itself (as in incomplete;
), still involves the lvalue-to-rvalue conversion. C calls it slightly differently, but it's the same conversion.
In C++, a cast to void
, or simply using an expression as a statement by itself doesn't involve the lvalue-to-rvalue conversion. This is needed because C++ makes assignment operators return lvalues, so if the lvalue-to-rvalue conversion were applied, then
volatile int i;
i = 1;
would not merely store, it would also immediately load afterwards.
The lvalue-to-rvalue conversion requires a complete type, even if the value is then discarded, since otherwise, it's impossible to know how many bytes should be read.