Re: [PATCH] gas/x86-64: properly distinguish low and high register ranges

["H.J. Lu" <hjl dot tools at gmail dot com>]

On Tue, Jul 31, 2012 at 9:28 AM, Jan Beulich <JBeulich@suse.com> wrote:
>>>> On 31.07.12 at 17:48, "H.J. Lu" <hjl.tools@gmail.com> wrote:
>> On Mon, Jul 30, 2012 at 11:25 PM, Jan Beulich <JBeulich@suse.com> wrote:
>>>>>> On 30.07.12 at 18:04, "H.J. Lu" <hjl.tools@gmail.com> wrote:
>>>> On Tue, Jul 24, 2012 at 11:56 PM, Jan Beulich <JBeulich@suse.com> wrote:
>>>>>>>> On 24.07.12 at 16:16, "H.J. Lu" <hjl.tools@gmail.com> wrote:
>>>>>> Can you add some testcases?
>>>>>
>>>>> I knew you would ask this, but sorry, this makes no sense - if test
>>>>> cases would are desirable here, they shouldn't be testing just the
>>>>> things that this patch fixes, but also any other invalid operand
>>>>> combinations. As an example - why would testing that "xlat [r11]"
>>>>> isn't accepted be needed, but not e.g. "xlat [ecx]"?
>>>>>
>>>>> Furthermore, this fixes actually broken behavior, so accepting
>>>>> the change shouldn't be dependent upon test case availability.
>>>>
>>>> What broken behavior does this change fix?
>>>
>>> I gave an example above - xlat [r11]. Other similar examples
>>> involve other string instructions requiring fixed registers as
>>> well as the one or two instructions requiring xmm0/ymm0 as
>>> their first/last operand.
>>>
>>
>> Please open a bug report for broken behaviors.
>
> That's bureaucracy that doesn't get us anywhere. I'd really
> like to know whether the patch is okay; entering bug reports
> is meaningful if one _can't_ fix a problem right away.
>

I'd like to see the impact of broken behaviors and verify that
the change fixes them.


-- 
H.J.