Re: Clarification for acceptance statistics?

From: SF Markus Elfring
Date: Mon Dec 12 2016 - 17:11:36 EST


> The question was: have you ever had a patch changing code in the form
>
> {
> a = kmalloc(...);
> b = kmalloc(...);
>
> if (!a || !b)
> goto out;
>
> ...
>
> out:
> kfree(a);
> kfree(b);
> }
>
> to something else, accepted?

It seems that this case did not happen for me so far if you are looking
for this exact source code search pattern.

Variants of the current pattern were occasionally discussed a bit.


> I went checking and I haven't found such a patch.

A few similar update suggestions are still in development waiting queues.


> Did you understand my question?

Partly. - My interpretation of similar changes was eventually too broad
in my previous answer.


>> It is really needed to clarify the corresponding software development
>> history any further?
>
> It is relevant because you are submitting a patch and your changelog
> implies that it makes the code follow some code structure rule that
> needs to be applied to the kernel.

I am proposing a change which was described also around various other
functions in some software already.


> As the above is a recurring pattern in kernel code, it is legitimate
> to ask if such a rule exist, and has been enforced before, or you are
> making it up.

I got the impression that special software development habits can also
evolve over time.


> As a proposer of a new pattern, what is the evidence you can bring to
> the discussion that supports that your solution is better?

I am trying to increase the software development attention on error
detection and corresponding exception handling at various places.


> What is the metric you are using to define "better"?

Do response times for system failures matter here?

Regards,
Markus