Re: [PATCH] scripts/kconfig: replace single character strcat() appends

[Ulf Magnusson]

On Sat, Mar 3, 2018 at 7:14 PM, Ulf Magnusson <ulfalizer@xxxxxxxxx> wrote:
> On Sat, Mar 3, 2018 at 12:12 AM, Joey Pabalinas <joeypabalinas@xxxxxxxxx> wrote:
>> On Fri, Mar 02, 2018 at 02:44:53PM +0100, Ulf Magnusson wrote:
>>> Not sure this is an improvement. Zeroing the bytes after the initial
>>> null terminator is redundant, and the explicit '\0' makes it clearer to
>>> me what's going on.
>>
>> Yes, I agree with you, that is definitely quite true. This along with
>> the other comments you made me want to rethink this a little bit.
>>
>> On Fri, Mar 02, 2018 at 02:44:53PM +0100, Ulf Magnusson wrote:
>>> I like the approach, but I wonder if we can take it a bit further.
>>> Here's what I'd do:
>>>
>>>       1. Rename the 'in' parameter to 's'.
>>>       2. Rename 'p' to 'in'.
>>>       3. Rename 'end' to 'out'
>>>
>>> At that point, you're reading from 'in' and writing to 'out', which
>>> seems pretty nice and readable.
>>>
>>> This code is pretty cold by the way, so it wouldn't matter for
>>> performance. GCC knows how functions like strcat() work too, and uses
>>> that to optimize (see
>>> https://gcc.gnu.org/onlinedocs/gcc/Other-Builtins.html).
>>>
>>> I'm all for trying to make Kconfig's code neater though.
>>
>> Since this code is pretty cold (completely agree with you there), I
>> think it would actually be much more useful to rework my patch to
>> have a more style-centric approach rather than an optimization-centric
>> one; this code would definitely benefit from being neater.
>
> I actually prefer the memcpy() version for style reasons, even though
> it might've looked like an optimization:
>
> With strncat(), the result string is written via both 'res' and 'end'.
> With memcpy(), it's only written via the 'end'. That seems less
> twisty.
>
> Maybe this is outside the scope of the original patch, but while we're here. :)
>
>>
>> Some useful changes would be to rename of the _atrociously_ short
>> identifiers like p and l.
>
> Yeah, 'l' in particular isn't the best name, IMO ('len' is both short
> and explicit, and won't be confused for 1). 'p' can be fine if it's
> obvious in context (bit dubious here), but 'in' and 'out' (for 'end')
> would be more informative.

Another alternative: src/dst

>
> 's' is clear from convention to me. In general, I fully agree that you
> should avoid hard-to-guess names though.
>
>>
>> Anyway I'll give that link a read over and try and make a V2 later
>> on today.
>>
>> Appreciate the feedback, thanks for the comments!
>>
>> --
>> Cheers,
>> Joey Pabalinas
>
> Cheers,
> Ulf

Cheers,
Ulf