Re: [boost] Volunteers needed: clearing the last regressions for1.34

From: Phil Richards (news_at_[hidden])
Date: 2006-10-19 15:21:39

On 2006-10-18, John Maddock <john_at_[hidden]> wrote:
>  Markus Schöpflin wrote:
> > John Maddock wrote:
> > Hmm, I think this is completely beyond me. I can't even find where
> > that cast to int comes from. Here is the preprocessed source code for
> > reference:
>  Hmm, I managed to find a copy of gcc-4.1 to test with, and this completely 
>  perplexes me too.  In fact I can't find *anything* in the mpl sources or 
>  tests that look like the code in the error message!

I'm pretty sure it's a compiler bug.  The code in question is below
(with my little bit added):

template< typename T >
void count_test()
{
#if DEFINE_THIS_TO_MAKE_THE_TEST_COMPILE
    enum { v1 = count<typename T::s0,int>::value };
    enum { v2 = count<typename T::s1,int>::value };
    enum { v3 = count<typename T::s2,int>::value };
    enum { v4 = count<typename T::s2,char&>::value };
    enum { v5 = count<typename T::s3,int>::value };
    enum { v6 = count<typename T::s3,char&>::value };
    enum { v7 = count<typename T::s4,abstract>::value };
#endif

    MPL_ASSERT_RELATION( ( count<BOOST_DEDUCED_TYPENAME T::s0,int>::value ), ==, 0 );
    MPL_ASSERT_RELATION( ( count<BOOST_DEDUCED_TYPENAME T::s1,int>::value ), ==, 1 );
    MPL_ASSERT_RELATION( ( count<BOOST_DEDUCED_TYPENAME T::s2,int>::value ), ==, 1 );
    MPL_ASSERT_RELATION( ( count<BOOST_DEDUCED_TYPENAME T::s2,char&>::value ), ==, 1 );
    MPL_ASSERT_RELATION( ( count<BOOST_DEDUCED_TYPENAME T::s3,int>::value ), ==, 2 );
    MPL_ASSERT_RELATION( ( count<BOOST_DEDUCED_TYPENAME T::s3,char&>::value ), ==, 1 );
    MPL_ASSERT_RELATION( ( count<BOOST_DEDUCED_TYPENAME T::s4,abstract>::value ), ==, 1 );
}

If you enable the #if'd code, then the test compiles fine; if
you don't then you get that strange error.  It shouldn't, of course,
make any difference...

phil
-- 
change name before "@" to "phil" for email