Anthony
Thu Jun 21 07:50:12 CDT 2007
"David Webber" <dave@musical-dot-demon-dot-co.uk> wrote in message
news:%23H5iJ5$sHHA.4796@TK2MSFTNGP04.phx.gbl...
>
> "Anthony Jones" <Ant@yadayadayada.com> wrote in message
> news:ukC%23bl$sHHA.4548@TK2MSFTNGP04.phx.gbl...
>
> >> And you need to understand that copying a pointer is not the same as
> > copying
> >> the thing it points to (and/or that the default copy constructor of
> > CMyClass
> >> copies a pointer.)
> >
> > Yep I do comprehend the difference between a pointer and the data it
> > points
> > to.
>
> In that case, I suspect you have some other misconception about the fate
of
> the data pointed to by all your char * pointers and/or what happens when
you
> include something like ...="Hello world" in the program.
>
> > What's a "default copy constructor" ?
>
> If you have a CMyClass a;
>
> The statements
>
> CMyClass b = a;
>
> or
>
> CMyClass b(a);
>
> invoke a "copy constructor" to construct b. If you don't provide one
> explicitly,
What does the signature of a copy constructor look like?
CMyClass(CMyClass src);
OR
is it:-
CMyClass& operator=(CMyClass& src);
> you are liable to get a default version of it supplied by the
> compiler. Its action is to copy the data members of the class, member by
> member. If (as in your case) one of those members is a pointer, then
> you'll create a copy of the pointer, and not a copy of the data it points
> to. When that data goes out of scope, you'll find yourself with a
pointer
> to junk.
Right I get that. What if the member was an instance of class that does
have it's own copy constructor would the default copy constructor call that
to copy the member or would it blindly copy the members of that object as
well? That to me is the critical question.
>
> So if a class contains a pointer to data which it *owns*, then the copy
> constructor must make a copy of the data and set its pointer to point at
the
> copy. [There are string classes which come complete with copy
constructors
> which manage all that. If you use standard nul-terminated strings as in
C,
> then you have to do it yourself.]
Ok got that.
>
> (NB Functions which return a class will also call the copy constructor,
> even though neither of the above statements occur explicitly in that form,
> as a copy of the class being returned is placed on the stack for the
benefit
> of the calling function. The same is true if you call a function with a
> class as one of its arguments.)
Surely that could be optimised away. The existing object is already on the
stack. Are you saying a second copy gets added to the stack, then the
original has it's destructor called, then after returning execution to the
caller another copy is performed, then presumably this temporary copy in the
stack has it's destructor called?
Perhaps I'm showing my ignorance again but since the callee knows its to
return an object it has on the stack it could defer its destruction to the
caller. The caller could perform the copy constructor once then destroy the
object from the callee function.
>
> Dave
>
> --
> David Webber
> Author of 'Mozart the Music Processor'
>
http://www.mozart.co.uk
> For discussion/support see
>
http://www.mozart.co.uk/mzusers/mailinglist.htm
>