Re: scoped_ptr: trivial dtor?

From: Vladimir Prus (ghost_at_[hidden])
Date: 2001-04-03 10:04:55

Beman Dawes wrote:

> >>> , and that one can assure that you never call delete on an object with
> >>> incomplete type.
> >
> >>How?
> >
> >As I have stated in previous post, with sizeof(T):
> >
> >       ~scoped_ptr() {
> >               sizeof(T);
> >               delete p;
> >       }

>If you did that, scoped_ptr could not be used with incomplete 
>types.  Although you have to be careful that the destructor is trivial, it 
>is highly desirable to be able to use scoped_ptr with incomplete types.  I 
>must be missing your point.

My first post was saying that point where pointer is deleted is near 
instantination point of ~scoped_ptr(), which allows to instantinate 
scoped_ptr on an incomplete type -- handle/body will work. It seems to be 
standard bahaviour, g++ 3.0 and bcc 5.5.1 work exactly this way.

sizeof(T) simply assures that the type is complete at point of deletion -- I 
really don't understand when this constraint will lead to anything but added 
safety.

Peter Dimov wrote:
> Are there any compilers that don't warn on this?

Yes, bcc.

-- 
Regards,
Vladimir