Borland C++ 5.6 for Win32 Copyright (c) 1993, 2002 Borland c:\BoostMainCVS\libs\serialization\test\test_shared_ptr_132.cpp: Error E2238 C:\BoostMainCVS\boost/serialization/detail/shared_count_132.hpp 75: Multiple declaration for 'bad_weak_ptr' Error E2344 C:\BoostMainCVS\boost/detail/bad_weak_ptr.hpp 39: Earlier declaration of 'bad_weak_ptr' Error E2238 C:\BoostMainCVS\boost/serialization/detail/shared_count_132.hpp 95: Multiple declaration for 'sp_counted_base' Error E2344 C:\BoostMainCVS\boost/detail/sp_counted_base_w32.hpp 36: Earlier declaration of 'sp_counted_base' Error E2238 C:\BoostMainCVS\boost/serialization/detail/shared_count_132.hpp 320: Multiple declaration for 'shared_count' Error E2228 C:\BoostMainCVS\boost/serialization/detail/shared_count_132.hpp 320: Too many error or warning messages *** 6 errors in Compile *** AlisdairM wrote:

Robert Ramey wrote:

...

Hmm I patched these in but now my test_shrare_ptr_132 doesn't compile. I'm can't see what we're doing differently.

Can you post the compiler errors you are seeing?

[I presume it is only the Borland compiler failing still, as I protected all updates with BOOST_WORKAROUND tests]

correct Robert Ramey

I used patch in merger your patch into my copy. I've also checked that I have on line 44: #if !BOOST_WORKAROUND( _BORLANDC_, BOOST_TESTED_AT( 0x564) ) using namespace boost; #endif Note the ! I got my free borland 5.6? from a borland website. I'm really sure whether I should have 0x564 or ox560 above. Hmm - a mystery to me. Robert Ramey AlisdairM wrote:

Robert Ramey wrote:

...

Borland C++ 5.6 for Win32 Copyright (c) 1993, 2002 Borland

c:\BoostMainCVS\libs\serialization\test\test_shared_ptr_132.cpp:

Error E2238 C:\BoostMainCVS\boost/serialization/detail/shared_count_132.hpp 75: Multiple declaration for 'bad_weak_ptr'

Error E2344 C:\BoostMainCVS\boost/detail/bad_weak_ptr.hpp 39: Earlier declaration of 'bad_weak_ptr'

Error E2238 C:\BoostMainCVS\boost/serialization/detail/shared_count_132.hpp 95: Multiple declaration for 'sp_counted_base'

Error E2344 C:\BoostMainCVS\boost/detail/sp_counted_base_w32.hpp 36: Earlier declaration of 'sp_counted_base'

Error E2238 C:\BoostMainCVS\boost/serialization/detail/shared_count_132.hpp 320: Multiple declaration for 'shared_count'

Error E2228 C:\BoostMainCVS\boost/serialization/detail/shared_count_132.hpp 320: Too many error or warning messages

*** 6 errors in Compile ***

Strange, those are the exact same errors I was getting *before* I applied the patch.

Can you confirm that the first workaround in shared_count_132.hpp looks like this:

namespace boost_132 {

#if !BOOST_WORKAROUND( __BORLANDC__, BOOST_TESTED_AT( 0x564) ) using namespace boost; #endif

Note the unsual '!' in the BOOST_WORKAROUND test.

I am compiling against the latest CVS repositoy, and it appears fine to me.

See my reply to Tobias for the precise cause of the problem. But basically we have to remove all 'using namespace boost' directives from the Borland view of the files. If you have any more usings on your local copy, they might be the cause an need similar wrapping.

"AlisdairM" <alisdair.meredith@uk.renaultf1.com> writes:

Robert Ramey wrote:

...

I used patch in merger your patch into my copy. I've also checked that I have on line 44:

#if !BOOST_WORKAROUND( BORLANDC, BOOST_TESTED_AT( 0x564) ) using namespace boost; #endif

Note the !

I got my free borland 5.6? from a borland website. I'm really sure whether I should have 0x564 or ox560 above.

Hmm - a mystery to me.

OK, now you have committed to CVS I can see the problem <g>

The __BORLANDC__ macro that is tested for should have two underscores either side, i.e. _ _ B O R L A N D C _ _

The version in CVS uses singe underscores, so will not be detected.

I suspect this will be an artefact of newsreaders or email clients. I know mine 'helpfully' translates underscores into requests to underline instead, so I guess we need to be extra careful when handling these identifiers in mailed code.

Generally speaking, mailing code as an attachment prevents these sorts of 'helpful' substitutions. -- Dave Abrahams Boost Consulting www.boost-consulting.com

Tobias Schwinger wrote:

Feel free to ask me if you need some testing -- I currently have a little BCC farm installed (5.5.0, 5.5.1, 5.6.4, 5.6.5).

Thanks. The Borland bug we are trying to work around can be seen in the following example: namespace A { struct duplicate {}; } namespace B { using namespace A; struct duplicate {}; } The compiler will erroneously complain that B::duplicate is a redefinition of A::duplicate. In this case, the using namesspace boost; inside namespace boost_132 is causing duplicates of the shared_ptr classes, which are #included inside Test library. The patch I gave yesterday solved the problem for me, I am not sure why it does not work for Robert. I am now looking for more places I might have missed namespace using directives. If you can apply my patch and try the shared_ptr_132 tests, I'd be interested in the results. -- AlisdairM