2010/7/30 Fabien Chêne <fabien.chene@gmail.com>:
> Hi Jason, Hi All,
>
> 2010/6/18 Jason Merrill <jason@redhat.com>:
>> On 06/16/2010 04:30 PM, Fabien Chêne wrote:
>>>
>>> @@ -2703,6 +2703,8 @@ count_fields (tree fields)
>>>   int n_fields = 0;
>>>   for (x = fields; x; x = TREE_CHAIN (x))
>>>     {
>>> +      if (TREE_CODE (x) == USING_DECL)
>>> +       continue;
>>> @@ -2720,6 +2722,9 @@ add_fields_to_record_type (tree fields,
>>>   tree x;
>>>   for (x = fields; x; x = TREE_CHAIN (x))
>>>     {
>>> +      if (TREE_CODE (x) == USING_DECL)
>>> +       continue;
>>
>> I don't think we want to skip USING_DECLs here; we aren't only counting
>> FIELD_DECLs, we count TYPE_DECLs and CONST_DECLs, too.
>
> I will try to change that, but for now, it ICEs if we do not skip USING_DECLs.
>
> I'm close to the result, but I'm not sure about how should behave the
> compiler with the following examples.
> I think that in both cases, the using declaration is invalid, can you
> confirm that ?

I've treated all of them as invalid.

Here is an updated patch. The fight against the testsuite makes me go
a little bit beyond the scope of  PR 26256.

Tested x86_64-unknown-linux-gnu with no regressions. Is it ok for trunk ?


gcc/ChangeLog

2010-08-18  Fabien Chêne  <fabien@gcc.gnu.org>

	PR c++/26256
	* dbxout.c (dbxout_type_fields): Ignore using declarations.


gcc/testsuite/ChangeLog

2010-08-18  Fabien Chêne  <fabien@gcc.gnu.org>

	PR c++/26256
	* g++.dg/lookup/using23.C: New.
	* g++.dg/lookup/using24.C: New.
	* g++.dg/lookup/using25.C: New.
	* g++.dg/lookup/using26.C: New.
	* g++.dg/lookup/using27.C: New.
	* g++.dg/lookup/using28.C: New.
	* g++.dg/lookup/using29.C: New.
	* g++.dg/debug/using4.C: New.
	* g++.dg/debug/using5.C: New.
	* g++.old-deja/g++.other/using1.C: Adjust.
	* g++.old-deja/g++.other/using8.C: Likewise.
	* g++.dg/template/using2.C: Likewise.

gcc/cp/ChangeLog

2010-08-18  Fabien Chêne  <fabien@gcc.gnu.org>

	PR c++/26256
	* search.c (lookup_field_1): Get rid of the comment saying that
	USING_DECL should not be returned, and actually return USING_DECL
	if appropriate.
	* semantics.c (finish_member_declaration): Store USING_DECLs with
	TYPE_DECLs.
	* typeck.c (build_class_member_access_expr): Handle USING_DECLs.
	* class.c (check_field_decls): Keep using declarations and
	diagnose conflicting declarations involving USING_DECLs. Call
	diagnose_conflicting_using_decls.
	(diagnose_conflicting_using_decls): New function.
	(handle_using_decl): Tweak to make things work for
	g++.old-deja/g++.other/using8.C.


-- 
Fabien