[PATCH] Fix PR c++/47172

[Jason Merrill]

On 02/10/2011 10:26 AM, Dodji Seketeli wrote:
> Gabriel Dos Reis<gdr@integrable-solutions.net>  writes:
>
>> Note that there is a strong disagreement/debate in the CWG about this.
>> So, putting it in the C++ front-end source code does not necssarily
>> help :-)
>
> OK. I just added the comment b/c it didn't seem obvious to me at first
> sight. I didn't mean to add to the controversy. I can remove it.

Just add a reference to issues 515 and 1005.  For 4.6 we just want to 
restore the previous behavior.

> +	  || (DECL_NONSTATIC_MEMBER_FUNCTION_P (fn)
> +	      && current_class_ref
> +	      && type_dependent_expression_p (current_class_ref)))

I'm uncomfortable with checking DECL_NONSTATIC_MEMBER_FUNCTION_P before 
we've established that fn is a FUNCTION_DECL.  Let's check its TREE_CODE 
first.

>  /* Returns TRUE if the EXPRESSION is type-dependent, in the sense of
> -   [temp.dep.expr].  */
> +   [temp.dep.expr]. Note that an expression with no type is
> +   considered dependent.  */

Let's also note that other parts of the compiler arrange for an 
expression with type-dependent subexpressions to have no type, so we 
don't need to recurse in type_dependent_expression_p.

Jason