Hi Andy,
On Sun, Mar 24, 2019 at 11:19:32PM +0200, Andy Shevchenko wrote:
On Sun, Mar 24, 2019 at 11:10:08PM +0200, Sakari Ailus wrote:
> On Fri, Mar 22, 2019 at 07:05:50PM +0200, Andy Shevchenko wrote:
> > On Fri, Mar 22, 2019 at 03:53:50PM +0200, Sakari Ailus wrote:
> >
> > > Porting a patch
> > > forward should have no issues either as checkpatch.pl has been
complaining
> > > of the use of %pf and %pF for a while now.
> >
> > And that's exactly the reason why I think instead of removing warning on
> > checkpatch, it makes sense to convert to an error for a while. People are
> > tending read documentation on internet and thus might have outdated one. And
> > yes, the compiler doesn't tell a thing about it.
> >
> > P.S. Though, if majority of people will tell that I'm wrong, then it's
okay to
> > remove.
>
> I wonder if you wrote this before seeing my other patchset.
Yes, I wrote it before seeing another series.
> What I think could be done is to warn of plain %pf (without following
"w")
> in checkpatch.pl, and %pf that is not followed by "w" in the kernel.
> Although we didn't have such checks to begin with. The case is still a
> little bit different as %pf used to be a valid conversion specifier whereas
> %pO likely has never existed.
>
> So, how about adding such checks in the other set? I can retain %p[fF] check
> here, too, if you like.
Consistency tells me that the warning->error transformation in checkpatch.pl
belongs this series.
All other invalid pointer conversion specifiers currently result into a
warning only. I see that as an orthogonal change to this set. I found
another issue in checkpatch.pl that may require some discussion; would you
be ok with addressing this in another set?
--
Regards,
Sakari Ailus
sakari.ailus(a)linux.intel.com