Re: [boost] [intrusive][container] boost/intrusive/detail/has_member_function_callable_with.hpp: inheriting from a const class

Subject: Re: [boost] [intrusive][container] boost/intrusive/detail/has_member_function_callable_with.hpp: inheriting from a const class
From: Jeffrey Lee Hellrung, Jr. (jeffrey.hellrung_at_[hidden])
Date: 2012-12-13 12:33:57

On Thu, Dec 13, 2012 at 9:26 AM, Krzysztof Czainski <1czajnik_at_[hidden]>wrote:

> 2012/12/2 Krzysztof Czainski <1czajnik_at_[hidden]>
>
> > Hello,
> >
> > I tried to use a boost::container::vector with a Texas Instruments
> > compiler, and I got an error pointing to line 88 of
> > <boost/intrusive/detail/has_member_function_callable_with.hpp> "Invalid
> > base class", with Type=const std::allocator<something>"
> >
> > Here's the context:
> >
> >       template <typename Type>
> >       class BOOST_PP_CAT(has_member_function_named_,
> > BOOST_INTRUSIVE_HAS_MEMBER_FUNCTION_CALLABLE_WITH_FUNCNAME)
> >       {
> >          struct BaseMixin
> >          {
> >             void
> > BOOST_INTRUSIVE_HAS_MEMBER_FUNCTION_CALLABLE_WITH_FUNCNAME();
> >          };
> >
> >          struct Base : public Type, public BaseMixin { Base(); }; // <--
> > line 88
> >
> > The problem seems to be the constness of Type. Here's a simplified
> example:
> >
> > struct X {};
> >
> > template < class T >
> > struct Y : T { Y() {} };
> >
> > int main()
> > {
> >     Y<const X> y;
> > }
> >
> > The above example compiles fine with Linux gcc version 4.5.4, and MinGW
> > 4.5.0, but fails with my TI compiler.
> >
> > My question is: is this a bug of the TI compiler, or is that compiler
> > actually right about this, and should we fix this in Boost? For now, my
> > workaround is:
> >
> > struct Base : public ::boost::remove_cv<Type>::type, public BaseMixin {
> > Base(); }; // <-- line 88
> >
> > Regards,
> > Kris
> >
>
> Hi,
>
> Since there is no response, and this is used by Boost.Container, I added
> the [container] tag to the subject.
>
> Any opinions about which compiler is right about this: is GCC too
> permissive, or is TI buggy in this respect?
>

FWIW, I believe the same compiler error pops up with at least some version
of the IBM compiler. I haven't received verification yet on whether the
remove_cv addition addresses that compiler error as well.

Also, since your asked for an opinion, my opinion is to side with GCC over
TI :)

- Jeff