RE: [Boost-users] configuring boost on MS Visual Studio 6 where t he installation drive isn't C:

From: Administrator (Administrator_at_[hidden])
Date: 2004-02-06 12:57:17

Wayne,

There might still be conflicts if the compiler can't place everything into
it's proper namespace.    In addition, some sdk's force everything into std
namespace or into the global namespace.  If you use "using namespace
boost;", you're basically doing the same thing.

There are a couple of options that could be used separately.

1.  One is to not use "using namespace".  Preface everything used from there
with boost:: or std::.  More typing, less pain, less chance for conflict.

2.  The second is wrap the offending sdk in it's own namespace.  Here is
what I have come up with.  I have three files here.  c.h is the original
function.  b.h wraps it in it's own namespace, and main.cpp can only use it
with the new namespace.  Simple example, but hopefully it illustrates what
I'm trying to get across.

--------------------------------------------------
c.h
-------------------------------------------------
#ifndef _C_H
#define _C_H

inline int garbage_method() { return 0; }

#endif //

--------------------------------------------------
b.h
--------------------------------------------------
#ifndef _B_H
#define _B_H

namespace garbage_namespace {
#include "c.h"
}
#endif //

--------------------------------------------------
main.cpp
--------------------------------------------------
#include "b.h"

int main()
{
        return garbage_namespace::garbage_method();
}

-----Original Message-----
From: Dernoncourt, Wayne CTR NAVAIR 3184, ,10
[mailto:wayne.dernoncourt_at_[hidden]]
Sent: Friday, February 06, 2004 5:43 AM
To: Boost Users mailing list
Subject: RE: [Boost-users] configuring boost on MS Visual Studio 6 where
the installation drive isn't C:

Well that worked much better as far as it went (747 updated files out of 747
targets wth no skipped files and no failures).  The version of the bjam
binary was a little larger (~5k) than the one I  compiled<scratches head>,
but using this produced no errors.

Unfortunately we still have at least a few problems.  As I described
originally we are using a highly specialized third party SDK.  Part of this
SDK redefines reverse_iterator (as well as min & max) evidently the MS
Visual C++ reverse iterator is broken in some way.  From a cursory
examination of the two implementations, it appears that both the two are
very similiar in functionality if not implementation.  The issue with
min/max could be eliminated by commenting out line 59 of
boost/config/stllib/dinkumware.hpp (line defines BOOST_NO_STD_MIN_MAX).
There doesn't appear to be a similiar line for the reverse iterator.

One of the issues appears to be that this third party SDK doesn't seem to
use namespaces for the SDK.  From my limited understanding, for namespaces
to be used effectively, the different libraries/SDK's would need to be in
their own namespaces.  If Boost had theirs and the third party used the
global namespaces, there could still be conflicts.  Is my understanding
correct?