[backport gcc-4.7/trunk r179236 <mikpe@it.uu.se>]

From: "Marcus Shawcroft" <marcus dot shawcroft at arm dot com>
Subject: [PATCH, LIBSTDC++] Fix type_info equivalence test.
Date: Tue, 9 Nov 2010 10:43:21 -0000
List-Archive: <http://gcc.gnu.org/ml/gcc-patches/>

Hi,

The attached patch fixes the failure of g++.dg/abi/local1.C observed on ARM
linux-gnueabi. The type_info::operator== implementation obtains the type
names for comparison by calling name(). The implementation of name() strips
the leading  '*' prefix if present. The '*' prefix marker is emitted by gcc
to force a pointer comparison. This patch replaces relevant name() call with
a direct reference to __name to ensure the prefix if any is visible.

Ok?

libstdc++-v3/

2011-09-27  Marcus Shawcroft  <marcus.shawcroft@arm.com>

	* libsupc++/tinfo.cc (type_info::operator==): Test __name instead
	of name().

--- gcc-4.6.2/libstdc++-v3/libsupc++/tinfo.cc.~1~	2009-10-30 23:23:59.000000000 +0100
+++ gcc-4.6.2/libstdc++-v3/libsupc++/tinfo.cc	2011-12-26 23:54:55.000000000 +0100
@@ -41,8 +41,11 @@ operator== (const std::type_info& arg) c
 #if __GXX_MERGED_TYPEINFO_NAMES
   return name () == arg.name ();
 #else
+  /* The name() method will strip any leading '*' prefix. Therefore
+     take care to look at __name rather than name() when looking for
+     the "pointer" prefix.  */
   return (&arg == this)
-    || (name ()[0] != '*' && (__builtin_strcmp (name (), arg.name ()) == 0));
+    || (__name[0] != '*' && (__builtin_strcmp (name (), arg.name ()) == 0));
 #endif
 }