C++ PATCH for c++/91230 - wrong error with __PRETTY_FUNCTION__ and generic lambda

[Jason Merrill]

On 8/1/19 5:53 PM, Marek Polacek wrote:
> On Thu, Aug 01, 2019 at 05:11:47PM -0400, Jason Merrill wrote:
>> On 8/1/19 3:50 PM, Marek Polacek wrote:
>>> We started rejecting this test with r268321, whereby we process the
>>> compound_literal in finish_compound_literal normally even in a template
>>> if it's not instantiation-dependent.  This broke with __PRETTY_FUNCTION__
>>> in a template function, because instantiation_dependent_expression_p didn't
>>> consider it dependent, which resulted in a bogus error later.
>>>
>>> cp_make_fname_decl has TYPE_DEP indicating that the fname is dependent, so
>>> I think value_dependent_expression_p needs to be fixed as below, and actually
>>> treat such a fname as dependent.
>>>
>>> Bootstrapped/regtested on x86_64-linux, ok for trunk and 9?
>>>
>>> 2019-08-01  Marek Polacek  <polacek@redhat.com>
>>>
>>> 	PR c++/91230 - wrong error with __PRETTY_FUNCTION__ and generic lambda.
>>> 	* pt.c (value_dependent_expression_p): Consider __PRETTY_FUNCTION__
>>> 	inside a template function value-dependent.
>>>
>>> 	* g++.dg/cpp1y/lambda-generic-pretty1.C: New test.
>>>
>>> diff --git gcc/cp/pt.c gcc/cp/pt.c
>>> index 91a46745447..94706bc5ad1 100644
>>> --- gcc/cp/pt.c
>>> +++ gcc/cp/pt.c
>>> @@ -25553,7 +25553,11 @@ value_dependent_expression_p (tree expression)
>>>          if (DECL_HAS_VALUE_EXPR_P (expression))
>>>    	{
>>>    	  tree value_expr = DECL_VALUE_EXPR (expression);
>>> -	  if (value_dependent_expression_p (value_expr))
>>> +	  if (value_dependent_expression_p (value_expr)
>>> +	      /* __PRETTY_FUNCTION__ inside a template function is dependent
>>> +		 on the name of the function.  */
>>> +	      || (DECL_PRETTY_FUNCTION_P (expression)
>>> +		  && value_expr == error_mark_node))
>>
>> Hmm, when would value_expr not be error_mark_node for __PRETTY_FUNCTION__ in
>> a template?
> 
> E.g.
> 
> template<typename T>
> struct S {
>    T *t{__PRETTY_FUNCTION__};
> };
> 
> S<const char> s;
> 
> here cp_make_fname_decl is called in a template, but in_template_function() is
> false, so we create a decl whose DECL_VALUE_EXPR is "top level".

Ah, sure.  The patch is OK, but you might mention that in the comment.

Jason