From patchwork Fri Dec 17 09:55:39 2010
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: Fix PR46916/46902
From: IainS <developer@sandoe-acoustics.co.uk>
X-Patchwork-Id: 75852
Message-Id: <E00DD7AD-F729-425D-BFDE-950828D1C6D8@sandoe-acoustics.co.uk>
To: GCC Patches <gcc-patches@gcc.gnu.org>
Cc: Mike Stump <mrs@gcc.gnu.org>
Date: Fri, 17 Dec 2010 09:55:39 +0000

The two PRs turned out to be manifestations of the same issue.

Basically,
gcc/system.h redefines bool => unsigned char.

this causes a conflict with any ABI where bool/_Bool is defined to  
something different (e.g. int).

(I don't think PPC/Darwin is the only case where this was done...  but  
it might be the only modern case).

anyway the "gotcha" is that all {external} system headers _must_ be  
included before gcc/system.h (noted in that header).

OK for trunk?
Iain

gcc:

	PR gcc/46902
	PR testsuite/46912
	* plugin.c: Ensure system headers are included before
	gcc/system.h

Index: gcc/plugin.c
===================================================================
--- gcc/plugin.c	(revision 167973)
+++ gcc/plugin.c	(working copy)
@@ -21,16 +21,19 @@ along with GCC; see the file COPYING3.  If not see
     APIs described in doc/plugin.texi.  */

  #include "config.h"
-#include "system.h"

  /* If plugin support is not enabled, do not try to execute any code
     that may reference libdl.  The generic code is still compiled in to
     avoid including too many conditional compilation paths in the rest
-   of the compiler.  */
+   of the compiler.
+
+   We must include system headers before "system.h" or the override
+   for bool might be upset. */
  #ifdef ENABLE_PLUGIN
  #include <dlfcn.h>
  #endif

+#include "system.h"
  #include "coretypes.h"
  #include "diagnostic-core.h"
  #include "tree.h"