Boost logo

Boost :

From: Beman Dawes (beman_at_[hidden])
Date: 1999-07-13 15:57:11


At 12:16 PM 7/13/99 -0700, Reid Sweatman wrote:

>Looks good to me, although I suspect there are more things not
supported by
>VC++ in any version than just member templates. Just have a look at
the MS
>section in SGI's STL config header

If you remove the portions that deal with old versions of the
compiler, and the #defines to identify features which are present
(rather than missing), and the thread related defines, then what is
left is much smaller. But as far as boost/config.h[pp] is concerned,
only the template member function case has come up.

>(BTW, do you have to explicitly include
>that header somewhere, or is it implicitly included by the other
headers?

Included by other boost headers that need it. Users should not have
to know about it.

> I
>can find no docs that mention it, and the code is Byzantine, to say
the
>least).

No docs yet, cause it isn't a done deal yet. It is really for use by
boost library implementors rather than users, so it shouldn't be
documented like a normal boost library.

>In that vein, can anyone here explain exactly *what* you do to work
around a
>compiler (like VC++) that doesn't support pointer operators on
iterators? I
>get inconsistent behavior from SGI STL, in that I can't explicitly
reference
>the innards of a compound object given an iterator to it, but the [
]
>operator works just fine, although internally it's implemented with
an
>iterator pointer (this is with the hash_map class, but it would
probably
>come up with *any* iterator object).

That question is probably better asked on microsoft.public.vc.stl on
the msnews.microsoft.com news server.

--Beman

------------------------------------------------------------------------

eGroups.com home: http://www.egroups.com/group/boost
http://www.egroups.com - Simplifying group communications


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk