[Bug c++/29525] New: boost 1.34.0 (testing release) mpl multiset test fails to compile

["gcc at derived-software dot ltd dot uk" <gcc-bugzilla at gcc dot gnu dot org>]

Attached is a preprocessed file that generates the following error (just the
first one of a group shown):

% g++  -ftemplate-depth-128 -O0 -fno-inline -Wall -g -fPIC -D_GLIBCXX_DEBUG=1 
-DBOOST_ALL_NO_LIB=1 -I. -c -o multiset.o libs/mpl/test/multiset.cpp

libs/mpl/test/multiset.cpp: In function 'void count_test() [with T =
test_data<boost::mpl::multiset0<0> >]':
libs/mpl/test/multiset.cpp:75:   instantiated from here
libs/mpl/test/multiset.cpp:61: error:
'(((int)boost::mpl::count_impl<boost::mpl::aux::multiset_tag>::apply<boost::mpl::multiset0<0>,
int>::value) == 0)' is not a valid template argument for type 'bool' because it
is a non-constant expression
libs/mpl/test/multiset.cpp:61: error: no matching function for call to
'assertion_failed(mpl_::failed************ mpl_::assert_relation<equal, 0l,
0l>::************)'
...etc...

The (non-processed) line that causes this is:
{
MPL_ASSERT_RELATION( ( count<BOOST_DEDUCED_TYPENAME T::s0,int>::value ), ==, 0
);
//...
}

This problem also appears to exist on 4.0.3 and 4.1.0.  (See, for example, the
reports on
http://engineering.meta-comm.com/boost-regression/CVS-RC_1_34_0/developer/mpl_release.html)

What is interesting(?) is that if the following line is inserted before the
MPL_ASSERT_RELATION line, then the error goes away:

enum { v1 = count<typename T::s0,int>::value };


-- 
           Summary: boost 1.34.0 (testing release) mpl multiset test fails
                    to compile
           Product: gcc
           Version: 4.1.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: gcc at derived-software dot ltd dot uk


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29525