From patchwork Sun Jun 20 17:18:58 2010
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: [4.4,libffi] fix support for FreeBSD
Date: Sun, 20 Jun 2010 07:18:58 -0000
From: Gerald Pfeifer <gerald@pfeifer.com>
X-Patchwork-Id: 56267
Message-Id: <alpine.LSU.1.99.1006201857170.9343@acrux.dbai.tuwien.ac.at>
To: gcc-patches@gcc.gnu.org
Cc: Loren James Rittle <rittle@labs.mot.com>

While mostly looking for something else, I noticed this patch Loren had
contributed to GCC 4.5 and later that we are missing on GCC 4.4 (which
is now the primary non-system compiler on FreeBSD).

Tested on i386-unknown-freebsd9.0, the expected passes for libjava grow
from 2475 to 2477, expected failures and unsupported testcases down by
one each: http://gcc.gnu.org/ml/gcc-testresults/2010-06/msg02062.html

Still, results on the 4.4 branch are a lot worse than on HEAD, but then
there have been a lot of improvements there in the last year.

Installed.

Gerald


2010-06-20  Gerald Pfeifer  <gerald@pfeifer.com>

	Backport from mainline:
	2009-09-17  Loren J. Rittle  <ljrittle@acm.org>

	PR testsuite/32843 (strikes again)
	src/x86/ffi.c (ffi_prep_cif_machdep): Add X86_FREEBSD to
	enable proper extension on char and short.


Index: src/x86/ffi.c
===================================================================
--- src/x86/ffi.c	(revision 161046)
+++ src/x86/ffi.c	(working copy)
@@ -123,7 +123,7 @@
 #ifdef X86
     case FFI_TYPE_STRUCT:
 #endif
-#if defined(X86) || defined(X86_DARWIN)
+#if defined(X86) || defined(X86_DARWIN) || defined(X86_FREEBSD)
     case FFI_TYPE_UINT8:
     case FFI_TYPE_UINT16:
     case FFI_TYPE_SINT8: