Re: Unreviewed C++ patches

[Mark Mitchell <mark at codesourcery dot com>]

Jakub Jelinek wrote:
> Hi!
>
> PR c++/21454
> TREE_READONLY for const array sized from initializer:
> http://gcc.gnu.org/ml/gcc-patches/2005-05/msg01066.html

OK.

> PR c++/21495
> fix for ICE on invalid in error reporting
> http://gcc.gnu.org/ml/gcc-patches/2005-05/msg00977.html

I don't think this is correct -- or, if it is, we should be simplifying 
things further.  In particular, I think id_declarator can be NULL here 
with invalid input like:

  static int : 3;

On the other hand, I have no idea what the current code is trying to do; 
the comment:

  /* Avoid trying to get an operand off an identifier node.  */

doesn't look like something I would have written, though you never know. 
 How about just replacing the entire contents of the "if (decl_context 
== FIELD)" block with:

  if (declarator && declarator->kind == cdk_id)
    error ("invalid storage class specified for data member %qD",
           declarator->u.id.unqualified_name);
  else
    error ("invalid storage class specified for data member");
  return error_mark_node;

?

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com
(916) 791-8304