Em Fri, Oct 31, 2003 at 01:50:04PM -0800, Joe Perches escreveu:

On Fri, 2003-10-31 at 13:23, David S. Miller wrote:

quoted

we're as mystified as you are as to why it is that:
	if (sum & 0x10000) {
		sum++;
		sum &= 0xffff;
	}
works while:
	sum = ((sym >> 16) + sum) & 0xffff;
does not.  The theory was that it might be some x86 gcc bug,
but looking at the assembler diff Arnaldo Carvalho de Melo
(the appletalk maintainer) showed me between the before and
after:
 	xorl	%eax, %eax
-	decl	%ecx
 	movb	(%ebx), %al
-	incl	%ebx
 	addl	%eax, %edx
 	addl	%edx, %edx
-	movl	%edx, %eax
-	shrl	$16, %eax
-	addl	%edx, %eax
-	movzwl	%ax,%edx
+	testl	$65536, %edx
+	je	.L982
+	incl	%edx
+	andl	$65535, %edx
+.L982:
+	decl	%ecx
+	incl	%ebx
 	cmpl	$-1, %ecx

we still can't see what's wrong.
He did confirm that the change in question makes IPX compute checksums
correctly.

If so, something major is WRONG.

I guess so. If people with an Appletalk testbed could try with and without
the patch, that would be great.

Code all over the place would need this "fix".

Haven't noticed any other place with this problem, but yes, this is something
that may well happen in other code.

Was an old NG Sniffer being used to verify this?

Nope, MacOS 7.5.5 and 9 just hang when communicating with the machine without
this patch, gcc 3.3.1/2 was used.

Sniffer had a long term problem with IPX checksums.

I added back the old atalk_checksum from 2.4 and compared the results, not
all were wrong, but some were.
 

Has the gcc team been contacted?

Nope, feel free to forward this message.

- Arnaldo