[Linux-kernel-mentees] checkpatch.pl investigation: NO_AUTHOR_SIGN_OFF issues

[Lukas Bulwahn]

On Sat, 26 Sep 2020, Dwaipayan Ray wrote:

> > > Yes, that should help I think. Currently checkpatch is very vague
> > > about Author sign offs, so subclassing it will really be helpful.
> > > But at the same point I don't think it should become
> > > too complex either.
> >
> > Agree, that is where the evaluation will help (what is really happening,
> > what do we need to consider) and the severity comes into play (if we
> > consider all subclasses of same severity, we do not need to differentiate
> > between them.)
> >
> > >
> > > I have already prepared a patch for the basic two classes:
> > > 1) same name, different address
> > > 2) same address, different name
> > >
> > > It would be great to have your feedback on this.
> >
> > Okay, I suggest to run this script on a larger set of commits and see.
> >
> > Probably, you can speed it your evaluation with checkpatch.pl by
> > restricting it to the AUTHOR_SIGN_OFF type. I will try that over the
> > weekend as well.
> >
> > Lukas
> 
> Hi,
> So I did run my script to evaluate the type of errors that were generated.
> My results are from commits between v5.7 and v5.8
>

Okay, good so far. How about running this on a larger set of commits?
 
> 1) Same address, different name
> 
> There were 32 such instances. In general, there was no regularity in the
> specific error.
> 
> There were some cases with letter case mismatch. Like there was one
> committer who used the names "Arend Van Spriel", and "Arend van Spriel".
> I believe this error should not be generated and matching be made case
> neutral.
> 
> Apart from this, the rest of the cases in this category were some with missing
> first or last names, using names in other languages or having commas in
> one name, using initials etc. These should be warned about.
>

Agree. Let us making this WARNING, but not ERROR.

> 2) Same name, different address
> 
> There were 204 such instances. Most of them were of the form in which they
> used a completely different email address.
>

I guess we would need to understand in more detail why developers do that 
and if .mailmap helps here.
  
> Only one case was found with an email extension.
> <hpeter at gmail.com>
> <hpeter+linux_kernel at gmail.com>
> in commit 423d9118c624
> 
> So in general I think most email mismatches should be specified by a WARN
> and for email extension there should be a --strict CHECK.
> 
> What do you think of this?
>

Basically, I agree. How about proposing a patch to Joe and lkml and then 
we see...

Lukas