Re: [PATCH linux-kselftest/test v6] lib/list-test: add a test for the 'list' doubly linked list

[shuah]

On 10/30/19 1:23 PM, Joe Perches wrote:
> On Wed, 2019-10-30 at 22:12 +0300, Dan Carpenter wrote:
> > On Wed, Oct 30, 2019 at 10:27:12AM -0600, shuah wrote:
> > > > It's better to ignore checkpatch and other scripts when they are wrong.
> > > > (unless the warning message inspires you to make the code more readable
> > > > for humans).
> > >
> > > It gets confusing when to ignore and when not to. It takes work to
> > > figure out and it is subjective.
> >
> > In this case, it's not subjective because checkpatch is clearly not
> > working as intended.
>
> checkpatch _is_ working as intended.
> It was never intended to be perfect.
>
> checkpatch _always_ depended on a reviewer deciding
> whether its output was appropriate.
>
> > I don't feel like "checkpatch clean" is a useful criteria for applying
> > patches.
>
> Nor do I.
>
> > The other things about warnings is that I always encourage people to
> > just ignore old warnings.  If you're running Smatch and you see a
> > warning in ancient code that means I saw it five years ago and didn't
> > fix it so it's a false positive.  Old warnings are always 100% false
> > positives.
>
> That'd be not absolute either because it depended on your
> historical judgment as to whether an old warning was in fact
> a defect or not.
>
> People make mistakes.
> Regex based scripts are by design stupid and untrustworthy.
>
> Mistakes will be made.
> Just fix the actual defects in code as soon as possible.


Thanks all for chiming in. I am taking v6 as is and adding an update
to commit log capture the spurious errors from checkpath.pl for this
specific case.

thanks,
-- Shuah