Re: [boost] On boost.config

From: Peter Dimov (pdimov_at_[hidden])
Date: 2001-09-24 09:27:02

From: "John Maddock" <John_Maddock_at_[hidden]>
> >+ BOOST_NO_VOID_RETURNS
> >+ BOOST_NO_PARTIAL_ORDERING
> >- BOOST_WEAK_FUNCTION_TEMPLATE_ORDERING
>
> Agreed, however I'm going to be busy for the next few weeks - do you want
> to change this?

I can try. ;-) First I'll need to understand how the config system works,
though.

Other comments (result from initial understanding attempts):

* On MSVC 7, config_test fails in boost_no_exp_func_tem_arg.cxx, although
the individual test compiles. This is due to the fact that the test uses a
function template appropriately named 'foo', but earlier tests define a
class template 'foo' (in another namespace) which breaks the foo<...>(...)
syntax.

* On MSVC 7, std::allocator is nearly compliant when
_HAS_MEMBER_TEMPLATES_REBIND is defined to nonzero. The reason that the test
fails is that it uses the syntax

a.allocate(p, 0);

(note explicit 0 as a second argument)

whereas the Dinkumware STL (3.06) has

 template<class _Other>
  pointer allocate(size_type _Count, const _Other *)

that doesn't accept zeros. Is the literal zero syntax important for
something in boost?

* Related to the above: suffix.hpp automatically defines
BOOST_NO_STD_ALLOCATOR when BOOST_NO_MEMBER_TEMPLATES is defined; this is
not the case on MSVC.

* Shouldn't !defined(BOOST_DISABLE_THREADS) below

#if (defined(__MT__) || defined(_MT) || defined(_REENTRANT) \
    || defined(_PTHREADS)) && !defined(BOOST_DISABLE_THREADS)
#  define BOOST_HAS_THREADS
#endif

be !defined(BOOST_HAS_THREADS)?

I can (try to) fix the above if there are no objections.

--
Peter Dimov
Multi Media Ltd.