Shibboleth Developers

[Jim Fox <>]

Without the definition of HAVE_GOOD_STL I get this:
(3.2.3 compiler)

-----------------------------------------------------
...
Making all in saml
make[2]: Entering directory `/data/local/src/opensaml-2.0/saml'
if /bin/sh ../libtool --silent --mode=compile --tag=CXX g++ 
-DHAVE_CONFIG_H -I. -I. -I..   -I/data/local/include -I/data/local/include 
-I/data/local/include -I/data/local/include -pthread -Wall 
-I/data/local/include -O2 -DNDEBUG -I/data/local/include -g 
-I/data/local/include  -pthread -g -I/data/local/include -Wall -O2 
-DNDEBUG -MT MetadataImpl.lo -MD -MP -MF ".deps/MetadataImpl.Tpo" -c -o 
MetadataImpl.lo `test -f 'saml2/metadata/impl/MetadataImpl.cpp' || echo 
'./'`saml2/metadata/impl/MetadataImpl.cpp; \
then mv -f ".deps/MetadataImpl.Tpo" ".deps/MetadataImpl.Plo"; else rm -f 
".deps/MetadataImpl.Tpo"; exit 1; fi
saml2/metadata/impl/MetadataImpl.cpp: In member function `virtual void
   opensaml::saml2md::RoleDescriptorImpl::addSupport(const XMLCh*)':
saml2/metadata/impl/MetadataImpl.cpp:1010: no match for `
   std::basic_string<char, std::char_traits<char>, std::allocator<char> > 
+
   const XMLCh*&' operator
/usr/include/c++/3.2.3/bits/stl_bvector.h:209: candidates are:
   std::_Bit_iterator std::operator+(int, const std::_Bit_iterator&)
/usr/include/c++/3.2.3/bits/stl_bvector.h:268:
   std::_Bit_const_iterator std::operator+(int, const
   std::_Bit_const_iterator&)

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

I'll try another compiler as well.

Jim


On Fri, 7 Dec 2007, Scott Cantor wrote:

> Date: Fri, 07 Dec 2007 01:15:02 -0500
> From: Scott Cantor 
> <>
> To: Jim Fox 
> <>
> Cc: 
>
> Subject: Re: error in opensaml make
>
> Jim Fox wrote:
> > I see that when I configured xmltooling I also got the
> > missing "std::char_traits.." and that may have resulted
> > in the missing definition of HAVE_GOOD_STL.  It was lost in
> > the output of configure.  Didn't cause any errors there
> > except the missing definition.
>
> Well, that part is ok. On that compiler, HAVE_GOOD_STL *should* be undefined 
> because gcc intentionally screwed up the STL in a non-productive way, and the 
> test should fail. It works again on gcc 3.3 or so.
>
> What I don't understand is why that's a problem. If there were compile-time 
> problems, that's not too unexpected, I don't test those branches enough. But 
> link time makes no sense to me, and that link error means it's compiling 
> something with that macro on. It shouldn't be doing that.
>
> I'd first make sure that xmltooling's internal config.h and the public 
> xmltooling/config_pub.h header agree that it's undefined. That should make 
> things build. If not, there's a header ordering problem or something. Or 
> maybe I left a stale reference to HAVE_GOOD_STL someplace in another config.h 
> header...
>
> Anyway, I'll see about installing an old CentOS that has the broken compiler 
> so I can test it, it's on my list anyway.
>
> -- Scott