From mboxrd@z Thu Jan  1 00:00:00 1970
From: Eli Zaretskii <eliz@is.elta.co.il>
To: Jiri Smid <smid@suse.cz>
Cc: gdb-patches@sources.redhat.com, Mark Kettenis <kettenis@science.uva.nl>
Subject: Re: [RFA]: x86_64 target - multiarch
Date: Tue, 11 Sep 2001 05:43:00 -0000
Message-id: <Pine.SUN.3.91.1010911144046.4452D-100000@is>
References: <s8vsnduj589.fsf@naga.suse.cz>
X-SW-Source: 2001-09/msg00156.html

On 11 Sep 2001, Jiri Smid wrote:

> *** 466,472 ****
>   {
>     int retval;
>   
> !   if (len == 3 || len > 4 || addr % len != 0)
>       retval = i386_handle_nonaligned_watchpoint (WP_INSERT, addr, len, type);
>     else
>       {
> --- 481,488 ----
>   {
>     int retval;
>   
> !   if ((len != 1 && len !=2 && len !=4 && (TARGET_HAS_DR_LEN_8 && len !=8))
> !       || addr % len != 0)
>       retval = i386_handle_nonaligned_watchpoint (WP_INSERT, addr, len, type);
>     else
>       {

Unless I'm missing something, the new code loses if addr and len both 
equal to 3, for a target which does not define TARGET_HAS_DR_LEN_8.  I 
think this is better:

 if ((len != 1 && len !=2 && len !=4 && !(TARGET_HAS_DR_LEN_8 && len ==8))

(There's one more case like this.)

Other than that, I have no comments.