Re: [PATCH] PR c++/26693

[Jason Merrill <jason at redhat dot com>]

Dodji Seketeli wrote:
> The test case in the attached patch, in file 
> gcc/testsuite/g++.dg/template/typedef14.C. The assert is hit when 
> calling fold_non_dependent_expr on "~static_cast<mytype> (0)", which 
> eventually calls tsubst on the mytype typedef.
>
> When we tsubst the mytype typedef, we hit the line:
> tree gen_args = tsubst (DECL_TI_ARGS (decl), args, complain, in_decl);

Ah, OK.  So in this case we're still in the template, so what we want is 
the template version of the typedef, so just adding

if (args == NULL_TREE)
  return t;

to the top of tsubst, or the top of the typedef reuse bit, would give 
the right answer.

> I changed this to actually put the code to append the type the list of 
> access checks for the template into check_accessibility_of_qualified_id, 

I thought about that possibility, but then you don't diagnose

class A
{
 typedef int foo;
};

template <class T>
class B: public A
{
  foo f;
};

> > > +append_type_to_template_for_access_check (tree type,

> > Don't we need to do the same thing in function templates?

> I am not sure to understand why this wouldn't work for typedef types 
> referenced in function templates. Could you please tell me what I am 
> missing ?

I guess I was just confused by the parameter named "type".  Please 
rename it to "tmpl" or some such.  :)

Jason