Re: [C++ PATCH] Fix friend class name lookup part 3/n (PR3332)

[Nathan Sidwell <nathan at codesourcery dot com>]

Kriang Lerdsuwanakij wrote:
> Hi
>
> This is a revised version of the fix for friend name injection part 2:
>
>  http://gcc.gnu.org/ml/gcc-patches/2004-10/msg02056.html
>
> Now it is submitted here as part 3 for ease of understanding (but should
> be applied together with part 2 and some formatting changes suggested
> earlier).  This patch also fixes PR3332.
>
>
> This patch introduces new functions push_inner_scope and pop_inner_scope
> that setup chain of binding levels directly for declarations like below
>
>  class N::C::D { ... };
>
> where N and C are previously declared class or namespace.  They are used
> in cp_parser_class_specifier.  The function push_scope/pop_scope formerly
> used there do not work well in this situation since they set up 
> information suitable for doing scoped name lookup but not for pushing name.
> Names may be pushed to outer scopes via friend declaration (although they
> are hidden until actually declared).
>
>
> Note that we have to treat template parameter scope specially in
> push/pop_inner_scope.  For example:
>
>  template <class T> class N::C::D { ... };
>
> we already have
>  {template parm}
> but after push_inner_scope, the binding chain should be  {namespace N} 
> {template parm} {class C}
> so that name lookup of 'T' find the template parameter rather than a name
> in the namespace N (according to 14.6.1/6).
>
>
> Tested on i686-pc-linux-gnu.  OK for mainline?

This looks ok. I presume that we could migtate push_scope & pop_scope to
use this new interface, and remove the OUTER param from push_inner_scope
(it's always current_scope (), right?).

I'd like to run this by Mark, just to make sure.

nathan

--
Nathan Sidwell    ::   http://www.codesourcery.com   ::     CodeSourcery LLC
nathan@codesourcery.com    ::     http://www.planetfall.pwp.blueyonder.co.uk