Re: [PATCH] x86: allow suffix-less movzw and 64-bit movzb

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

On Thu, Jun 30, 2016 at 4:47 AM, Jan Beulich <JBeulich@suse.com> wrote:
>>>> On 30.06.16 at 13:38, <hjl.tools@gmail.com> wrote:
>> On Thu, Jun 30, 2016 at 4:33 AM, Jan Beulich <JBeulich@suse.com> wrote:
>>>>>> On 30.06.16 at 13:26, <hjl.tools@gmail.com> wrote:
>>>> On Thu, Jun 30, 2016 at 3:42 AM, Jan Beulich <JBeulich@suse.com> wrote:
>>>>> --- 2016-06-30/gas/testsuite/gas/i386/movz32.d  1970-01-01
>> 01:00:00.000000000 +0100
>>>>> +++ 2016-06-30/gas/testsuite/gas/i386/movz32.d  2016-06-30
>> 12:04:26.000000000 +0200
>>>>> @@ -0,0 +1,22 @@
>>>>> +#objdump: -d
>>>>> +#source: movz.s
>>>>> +#name: x86 mov with zero-extend (32-bit object)
>>>>> +
>>>>> +.*: +file format .*
>>>>> +
>>>>> +Disassembly of section .text:
>>>>> +
>>>>> +0+ <movz>:
>>>>> +[      ]*[a-f0-9]+:    66 0f b6 c0 *   movzbw? %al,%ax
>>>>                                                                  ^
>>>>
>>>> What is `?' for?
>>>
>>> Now that we accept the suffix-less mnemonic I don't think it would
>>> be appropriate to demand the suffix to be issues by the disassembler
>>> (unless in suffix-always mode, which isn't the case here).
>>
>> Assembler output is controlled by "#objdump: -d".  Under what condition
>> will `?' be needed?
>
> This is to prevent having to touch this again when making the
> disassembler obey the absence of -Msuffix here. I'm of the
> opinion that test cases should check for valid output, not for what
> the tools currently produce.

objdump is the part of binutils.  I'd like to know if its behavior changes
by accident.  Please drop ?.

>>  Also please use "#objdump: -dw" so that there is
>> always one line for each insn.
>
> Easily done, though meaningless here afaict. Do I need to re-submit
> with that adjustment?

OK with "objdump -dw" and without ?

Thanks.

-- 
H.J.