Question

Here is an example snippet:

_variant_t var;

var.vt = VT_UNKNOWN;

var.punkVal = unknownInterfaceSmartPointer;

unknownInterfaceSmartPointer->AddRef(); // Question Statement

// Setting unknownInterfaceSmartPointer to some other container

Whether manual AddRef is really needed here? Following will is better way than this:

_variant_t var(unknownInterfaceSmartPointer.GetInterfacePointer());

your thoughts?

Answer 1

If you directly assign to punkVal member (as in your first snippet) then you need an AddRef() – noone will do it for you. A much better way would be to use operator=() member of _variant_t:

_variant_t var;
var = unknownInterfaceSmartPointer.GetInterfacePointer();

or do as you suggest – use a conversion constructor:

_variant_t var(unknownInterfaceSmartPointer.GetInterfacePointer());

Both latter variants are good and don’t require an AddRef() in your code – it will be done inside _variant_t implementation. Furthermore, they are beneficial because if you reassign to the same variant the implementation will do proper cleanup for you:

_variant_t var;
// if you assign a BSTR or smart pointer here
// you want it to be properly released later
var = whatever1();
// this will do the proper release of the previously
// stored stuff and then assign the new value
var = whatever2();