[Bug c++/17057] problems disambiguating std::allocator::rebind

["mark at codesourcery dot com" <gcc-bugzilla at gcc dot gnu dot org>]

------- Additional Comments From mark at codesourcery dot com  2004-08-17 16:22 -------
Subject: Re:  problems disambiguating std::allocator::rebind

bkoz at gcc dot gnu dot org wrote:

>------- Additional Comments From bkoz at gcc dot gnu dot org  2004-08-17 16:10 -------
>
>With all due respect, I think the preliminary analysis is not absolutely correct.
>
>First of all, it makes no difference what the EDG/icc flags are: both chug
>through  this code without error or warning. I'm using Version 8.1  Beta  Build
>20040602Z with gnu headers, but I'm not quite sure this really matters.
>
>Second, Andrew's summation is not what the code details. My apology: 
>Perhaps I attached the wrong pre-processed code the first time. I've tried
>again, see mt_allocator.ii.
>
>To recap:
>
>  template<typename _Tp, typename _PoolTp>
>    class __mt_alloc : public _PoolTp
>    {
>    public:
>
>      template<typename _Tp1, typename _PoolTp1 = _PoolTp>
>        struct rebind
>        { 
>#if 1
>          // ok with edg
>	  typedef _PoolTp1::_M_rebind<_Tp1>::other pool_type;
>#else
>          // as g++ suggests but cannot compile
>          typedef _PoolTp1::template _M_rebind<_Tp1>::other pool_type;
>#endif
>	  typedef __mt_alloc<_Tp1, pool_type> other; 
>	};
>    };
>
>None of these work with g++. If I just keep on adding junk, say
>
>	  typedef typename _PoolTp1::template _M_rebind<_Tp1>::other pool_type;
>
>
>this compiles with both g++ and EDG. Syntax only a mother could love (could this
>really get any uglier?), but hey, it works.
>
The last version is definitely valid.

I do not believe the case in "#if 1" is valid. 

There are DRs in the works that might make it valid, though; there has 
been some suggestion that in "type contexts" (like the first operand to 
"typedef") "typename" should be assumed.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17057