Re: [boost] Re: method deduction (on vc?)

From: Aleksey Gurtovoy (agurtovoy_at_[hidden])
Date: 2004-07-14 04:00:43

Jonathan Turkanis writes:
> "David Abrahams" <dave_at_[hidden]> wrote in message
> news:upt6z5s2f.fsf_at_boost-consulting.com...
>> "Jonathan Turkanis" <technews_at_[hidden]> writes:
>
>> >
>> > I wouldn't count on any member detection with VC6. I little works
> with
>> > V7.0.
>>
>> You can detect type members on vc6.  See mpl/aux_/has_xxx.hpp.
>
> I'm familiar with the implementation of has_xxx, but I was pretty sure
> it didn't work on VC6

Among other things, "has_xxx.hpp" contains MSVC 6.0-specific
implementation.

>
> ....
>
> In fact, the test program at the end of this message compiles fine on
> VC6, but should static assert if has_xxx is supported. Am I doing
> something wrong?

Yep, please see below.

> By the way, I have an implementation of has_template_xxx, which
> detects member templates. It works on all my SFINAE-capable
> compilers -- seven, at last count -- but I'm afraid it's probably not
> standard-conforming. 

You don't imply that a conforming implementation is not possible, do you?

> Any interest?

After the release, sure.

[...]
>
>     #include <boost/mpl/aux_/has_xxx.hpp>
>     #include <boost/static_assert.hpp>
>
>
>     struct S { };
>
>     struct T {
             ^

That's a really unfortunate choice of name. One of the MSVC 6.0's
peculiar bugs is this:

struct T {};
template< typename T > struct her : T {}; // inherits from *global* T!

>         typedef int kids;
>     };
>
>
>     BOOST_MPL_HAS_XXX_TRAIT_NAMED_DEF(has_kids, kids, true)
>
>     int main()
>     {
>         BOOST_STATIC_ASSERT(has_kids<S>::value); // Should fail.

If you rename 'T' to something else, it will.

>         BOOST_STATIC_ASSERT(has_kids<T>::value);
>     }
>

-- 
Aleksey Gurtovoy
MetaCommunications Engineering