Re: possible bug?

["dou wen" <orbkernel hotmail com>]

Hi, Michael


>From: Michael Meeks <michael@ximian.com>
>To: dou wen <orbkernel@hotmail.com>
>CC: orbit <orbit-list@gnome.org>
>Subject: Re: possible bug?
>Date: 08 Mar 2002 22:44:49 +0000
>
>	Sorry it took you a while to find that - I hit this before and it's a
>pain.

      So pain :-( 
      It's my pleasure :-)

>	The problem with this is that it's incredibly slower than a simple
>ALIGN_VALUE. 

 oh,I see. 

>The best solution is (I think) to have a wrapper for your
>malloc ( specific to the GIOP allocation code ) that ensures that the
>memory is 8 byte aligned - but we do need that assertion, could you
>knock up a simple patch (and comment) for that vs HEAD ORBit2 ?
>

Sure, here is my patch, but do not know if it is suitable.:-)

best regards

                                                 dou wen



_________________________________________________________________
与联机的朋友进行交流，请使用 MSN Messenger：
http://messenger.microsoft.com/cn
diff -Naur ORBit2-2.3.103/src/orb/GIOP/giop-recv-buffer.c 
ORBit2-2.3.103-fix/src/orb/GIOP/giop-recv-buffer.c
--- ORBit2-2.3.103/src/orb/GIOP/giop-recv-buffer.c	Mon Jan 14 23:59:51 2002
+++ ORBit2-2.3.103-fix/src/orb/GIOP/giop-recv-buffer.c	Sat Mar  9 13:38:10 
2002
@@ -828,8 +828,15 @@

	if (!is_auth && (buf->msg.header.message_size > 
GIOP_INITIAL_MSG_SIZE_LIMIT))
		return TRUE;
-
-	buf->message_body = g_malloc (buf->msg.header.message_size + 12);
+	/*
+	 * FIXME: we assume the alignment is always 8 bytes, surely this will 
speed up
+	 * the demarshal operations,but in some platforms and situations, the 
"malloc"
+	 * result is not 8 bytes aligned,this will result in mismatch in offset 
when
+	 * decoding, so here we make sure the buf->message_body is always 8 bytes 
aligned.
+ 	*/
+
+	buf->message_body = (gpointer)ALIGN_VALUE(g_malloc 
(buf->msg.header.message_size + 12 + 8), 8 );
+
	buf->free_body = TRUE;
	buf->cur = buf->message_body + 12;
	buf->end = buf->cur + buf->msg.header.message_size;