Re: [boost] RE: config.hpp problem

From: Beman Dawes (beman_at_[hidden])
Date: 2000-06-23 10:49:12

At 04:59 PM 6/22/2000 -0500, Ed Brey wrote:

 >From: <jsiek_at_[hidden]>
 >>
 >> I suppose this means that boost/config.hpp should be modified
 >> to only do the using ::wchar_t for non-conforming compilers.
 >>
 >> scleary_at_[hidden] writes:
 >>  > Under C++, wchar_t is a builtin type, like int, so it can't be in
 >a using
 >>  > declaration.
 >
 >wchar_t should not be in namespace std in any case.  The "using
 >::size_t", etc. is a workaround only for non-conforming compilers that
 >fail to put items into std.  In the case of wchar_t, the
 >non-conforming compilers have it right to begin with, by putting the
 >name in the global namespace.
 >
 >Well, they almost have it right.  The problem with wchar_t is that, as
 >a keyword, no include directive should be necessary to get it.  So the
 >real workaround is to make sure that the file that defines wchar_t as
 >a typedef  (albeit non-conformingly) is included.  It turns out that
 >including the file <cstddef> defines the typedef in question (at least
 >for VC6), and we happen include that file anyway to detect the
 >standard library implementation type.
 >
 >So, when everything is said and done, we can just remove, the "using
 >::wchar_t;" from config.hpp and make sure we don't have any references
 >to "std::wchar_t" in any of our code, and everything should work
 >hunky-dory.

Done, since the impact is minimal.

But I really wonder if there is any  hope for a compiler that is so 
out-of-date that it doesn't support wchar_t.

--Beman