[cfe-commits] More on templates and vtables

[Douglas Gregor]

On Mar 23, 2010, at 9:38 PM, Rafael Espindola wrote:

> This is becoming a snow ball, but I really think the patch has
> everything it needs now. What was missing from the previous one is
> that we were not producing the constructor for
> 
> template<class T> struct C {
>   virtual ~C();
> };
> template class C<int>;
> 
> The new patch fixes this, but got a bit larger since it has to handle
> some corner cases of when an implicit copy assignment is legal.


I'm not too thrilled about the code that checks whether implicit copy assignment is legal; we have very similar code in elsewhere (e.g., Sema::DefineImplicitOverloadedAssign), and it would be unfortunate to have to keep these in sync... also, if we have to do this for copy-assignment, wouldn't we have to do this for all of the implicit special member functions?

	- Doug