Re: [patch c++]: Fix PR/64127 ICE on invalid: tree check: exprected identifier_node, have template_id_expr in cp_parser_diagnose_invalid_type_name

[Paolo Carlini]

Hi,

On 12/04/2014 07:17 PM, Kai Tietz wrote:
So added testcase for this pr (its c++98 only)
So:

ChangeLog testsuite

2014-12-04  Kai Tietz  <kti...@redhat.com>

       PR c++/64127
       * g++.dg/cpp/pr64127.C: New file.

Tested on x86_64-unknown-linux-gnu.

Ok to apply prior posted patch plus this new testcase?
The testcase is now in and is spuriously failing for everybody because 
in the testsuite, thus with -pedantic-errors, in one case an error 
instead of a warning is emitted plus the dg- directives are 
unnecessarily complicated and slightly wrong. Barring objections, I mean 
to commit the below pretty soon.

Thanks,
Paolo.

/////////////////////////////
Index: g++.dg/cpp/pr64127.C
===================================================================
--- g++.dg/cpp/pr64127.C        (revision 218586)
+++ g++.dg/cpp/pr64127.C        (working copy)
@@ -1,9 +1,4 @@
 /* { dg-do compile { target c++98_only } } */
 
-template <0> int __copy_streambufs_eof; // { dg-error "" }
-// { dg-error "numeric constant" "" { target *-*-* } 3 }
-// { dg-warning "variable templates" "" { target *-*-* } 3 }
-__copy_streambufs_eof < // { dg-error "" }
-// { dg-error "parse error" "" { target *-*-* } 6 }
-// { dg-error "not name a type" "" { target *-*-* } 6 }
-
+template <0> int __copy_streambufs_eof; // { dg-error "expected 
identifier|numeric constant|variable templates" }
+__copy_streambufs_eof < // { dg-error "template argument|parse error|not name 
a type" }