Re: Destructor not called with forward declaration instead of include

Vidar Hasfjord wrote:
> On Apr 22, 6:13 pm, Brendan <catph...@[EMAIL PROTECTED]
> wrote:
>> What I'm actually iffy on, is why you can compile code that deletes a
>> pointer to an incomplete type at all. This compiles with only a
>> warning about deleting a pointer to an incomplete type on VC9.

I'm also curious.

> I agree that the standard should at least require a diagnostic if
> doesn't already do so. I guess forbidding it is difficult because it
> actually works fine when the type has a trivial destructor.

You mean, when the type is incomplete but has a trivial destructor?
Does that help much? What do we lose by requiring the type to be
complete at a 'delete' expression with a pointer to that type?
(On the other hand, we gain a lot of correctness assurance!)

> An interesting question is what is the minimal type information
> required for the compiler to handle this correctly? Could a language
> extension allow this to be handled without requiring the full type to
> be revealed (and thus require the user to break encapsulation or
> redesign the class hierarchy). Could the following be enough?
> 
>    virtual class Clazz; // I have a virtual destructor, promise!

I guess the compiler may need the complete class definition even
just to invoke the virtual destructor; e.g. it has to know which
entry in the vtable represents the destructor.

-- 
Seungbeom Kim

      [ See http://www.gotw.ca/resources/clcm.htm
for info about ]
      [ comp.lang.c++.moderated.    First time posters: Do this! ]