[GCC-XML] last commit

[Roman Yakovenko]

On Tue, Jun 17, 2008 at 11:28 PM, Brad King <brad.king at kitware.com> wrote:
> Roman Yakovenko wrote:
>> Sorry for bringing this back, but it seems that I cannot work with the
>> current version, without commenting out some test cases.
>
> Can you point me at your code and these tests please?

No need I can describe it here. It is very similar to what you found.

I have such code:

namespace is_noncopyable{
namespace detail{
    struct const_item{ const int values[10]; };
namespace yes{
}
namespace no{
    typedef detail::const_item const_item_t;
}
}

and in test, I verify, that my type traits returns "the right" answers.

In this specific case if I place const_item_t in "no" namespace GCCXML
doesn't generate artificial declarations, and does this if I placed it
in "yes" namespace.

As I said, I could comment it out, but it proved itself - it found
"strange" behaviour, it deserve to stay :-)

>> <Class ... is_default_constructable="1" is_assignable="1"
>> is_copy_constructable="1">
>
> I was also thinking of this solution when the bug this commit fixed was
> originally reported.  The question is whether the information needed to
> extract these traits is available and correct at the time of the dump.
> Hopefully whatever is causing the strange missing-declaration behavior
> will not break this solution too.
>
> I'll add this to my to-do list but I'm about to go on vacation so I
> won't get to it for a couple weeks.

Enjoy. These things definitely can wait.

Thanks.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/