Re: [Qemu-devel] [PATCH V5 1/7] monitor: avoid direct use of global *cur_mon in completion functions

[Wenchao Xia <xiawenc@linux.vnet.ibm.com>]

于 2013-7-9 22:14, Luiz Capitulino 写道:
> On Tue, 09 Jul 2013 22:03:42 +0800
> Wenchao Xia <xiawenc@linux.vnet.ibm.com> wrote:
>
>>
>>>> You could pass cur_mon to readline_completion() in readline_handle_byte()
>>>> to avoid all this, but it would be preferable to clarify the matter.
>>>>
>>>> This is also another benefit of having readline_completion() in a
>>>> different patch, you can (and should!) clarify this point in the commit
>>>> log along with testing results.
>>>>
>>>     OK, I'll test multiple monitors case.
>>>
>>>
>>>
>>     Do you have an example about how invoke qemu with multiple monitor?
>> Do you have recommendation about the way to connect to qemu's moniotr?
>> I tried pty with minicom but it seems not good, so wonder your method.
>
> You can open telnet sessions:
>
>   -monitor tcp:0:4444,server,nowait,telnet \
>   -monitor tcp:0:4445,server,nowait,telnet
>
> But note that, besides testing, I'd like to see an explanation on
> the commitlog as to why dropping cur_mon doesn't brake the current
> code.
>

   Thanks for the info, it works now.
   After re-check the code in readline_completion(), I think current
code with cur_mon, is redundant:

     mon = cur_mon;
     ...

     } else {
         monitor_printf(mon, "\n");
         ...
          ...
         for(i = 0; i < rs->nb_completions; i++) {
             monitor_printf(rs->mon, "%-*s", max_width, rs->completions[i]);
             if (++j == nb_cols || i == (rs->nb_completions - 1)) {
                 monitor_printf(rs->mon, "\n");
                 j = 0;
             }
         }
         readline_show_prompt(rs);
     }

   You can see later, it uses rs->mon to printf the auto completion
result, so the first 'monitor_printf(mon, "\n");', is used to start
a new line for the completion result, so mon should always be equal
to rs->mon, I can't think out a case that one monitor instance does
not print tips in new line while another monitor instance does. I'll
add a few line for it in commit message.

   About pass cur_mon to readline_completion(), I think parameter *rs
already have member rs->mon, allowing specify *mon would specify
the encapsulation a bit.

-- 
Best Regards

Wenchao Xia