Re: g++ 2.95.2 disallows partially specializated friends

[" " <sebor at my-Deja dot com>]

On Wed, 22 Mar 2000 01:16:30   Martin v. Loewis wrote:
>> g++ 2.95.s rejects friend declarations of partial class
>> specializations. I don't see anything in 14.5.3 that disallows this
>> (14.5.3, p1 says that "A friend of a class or class template can be
>> ... a specialization of a ... class template ...") - is this an
>> unimplemented feature, a bug, or have I missed something? (Other
>> conforming compilers such as edg 2.42 or HP aCC 3.14.10 accept the
>> code.)
>
>Thanks for your bug report. Without any deeper analysis: What exactly
>do you want this code to mean? If you want to make the global template
>a friend, then you can write
>
>template <class T, class U>
>struct A;
>
>
>template <class T>
>struct A<int, T>
>{
>    class B {
>        friend struct ::A;
>    };
>};

According to 14.6.1, p2, "when the name of the template is neither qualified nor followed by <, it is equivalent to the name of the template followed by the template-arguments enclosed in <>." Here, ::A is qualified so the paragraph doesn't apply and since ::A is a template, not an ordinary struct, I'd say it's illegal (other compilers concur). Allowing g++ to accept this syntax is IMO either a nonconforming extension or a bug.

A valid declaration could look like this

template <class T>
struct A<int, T>
{
    class B {
        template <class U, class V>
        friend struct A;
    };
};

but this isn't what's intended. The intent is to make only specialiazations of A<int> a friend of A<int>::B. The friend declaration above makes *any* specialization of A<> a friend of A<int>::B.

>
>Without that, 'A' identifies the injected class name, which is the
>partial specialization. g++ rejects it on the basis of
>
>      /* [temp.friend]
>	 
>	 Friend declarations shall not declare partial
>	 specializations.  */
>
>Now, I this is certainly a declaration. What does it declare? A
>specialization? It certainly refers to one. If you don't think this
>answer is satisfying, please discuss the problem in comp.std.c++
>first.

Will do.

Thanks
Martin

>
>Regards,
>Martin
>



--== Sent via Deja.com http://www.deja.com/ ==--
Share what you know. Learn what you don't.