[PATCH 2/4] staging/serqt_usb2: refactor qt_read_bulk_callback() in serqt_usb2.c

YAMANE Toshiaki yamanetoshi at gmail.com
Thu Nov 15 20:32:00 UTC 2012


On Fri, Nov 16, 2012 at 5:18 AM, Dan Carpenter <dan.carpenter at oracle.com> wrote:
> On Fri, Nov 16, 2012 at 05:01:55AM +0900, YAMANE Toshiaki wrote:
>> On Wed, Nov 14, 2012 at 9:41 PM, Dan Carpenter <dan.carpenter at oracle.com> wrote:
>> >
>> > Why can't we test whether i == (RxCount - 3) earlier and handle
>> > the errors there?  That way we wouldn't need to pass the limit
>> > variable.
>> >
>> > In fact, this whole function is sort of nasty.  We start by doing
>> > a switch (data[i + 2]) { then we combine the 0x00 and 0x01 and call
>> > this function which separates them out and sets a function pointer
>> > and then calls the function point?  Get rid of this whole function.
>> >
>> > You shouldn't need to use function pointers to do this; that's too
>> > many levels of abstraction.
>>
>> I feel it so diffcult to consider the fixing this patch more.
>>
>> There are some reasons why I have become such a description.
>> - The purpose of this patch is the resolution of the
>>   line over 80 characters issue
>> - I Wrote the code to be aware of the following:
>> -- Do not change the procedure
>> -- The shallow nest
>> -- To avoid the redundancy
>>
>> If I do not use a function pointer, which take the form below.
>>
>>         if (0x00 == data[i + 2])
>>                 dev_dbg(&port->dev, "Line status status.\n");
>>         else
>>                 dev_dbg(&port->dev, "Modem status status.\n");
>>
>>         if (i > limit) {
>>                 dev_dbg(&port->dev,
>>                         "Illegal escape seuences in received data\n");
>>                 return 0;
>>         }
>>
>>         if (0x00 == data[i + 2])
>>                 ProcessLineStatus(qt_port, data[i + 3]);
>>         else
>>                 ProcessModemStatus(qt_port, data[i + 3]);
>>
>>       return 1;
>>
>> I also feel it may be...
>>
>> And I am against to move the dev_dbg procedure call to
>> qt_status_change_check procedure because the nesting will be so deep.
>
> In the end, the new version is more confusing than the original
> code.  Checkpatch.pl is not a king which must be obeyed.  The only
> thing which matters is how easy it is for a human to understand the
> code.

Yes. I understand it.
I wil condider the improvement this.

>> >>       if (urb->status) {
>> >>               qt_port->ReadBulkStopped = 1;
>> >> -             dev_dbg(&urb->dev->dev, "%s - nonzero write bulk status received: %d\n",
>> >> +             dev_dbg(&urb->dev->dev,
>> >> +                     "%s - nonzero write bulk status received: %d\n",
>> >>                       __func__, urb->status);
>> >
>> > Don't mix in these unrelated 80 character limit changes.
>>
>> I think the purpose of refactoring is the resolution of the line over 80
>> characters issue. I think that the separation of the patch should stop taking
>> because they are already applied in the linux-next tree.
>>
>
> Yes, once it is merged into linux-next then it is too late to send a
> version 2 patch.
>
> I'm explaining that as a reviewer it is confusing for me to figure
> out when you do unrelated things in the same patch and mix
> everything up.

I understand it.
Thanks for your comments.


-- 

Regards,

YAMANE Toshiaki



More information about the devel mailing list