Re: [PATCH] checkpatch: use patch subject when reading from stdin

[Geert Uytterhoeven]

Hi Joe,

On Tue, Oct 8, 2019 at 8:10 PM Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> On Tue, Oct 8, 2019 at 7:02 PM Joe Perches <joe@xxxxxxxxxxx> wrote:
> > On Tue, 2019-10-08 at 17:28 +0200, Geert Uytterhoeven wrote:
> > > On Tue, Oct 8, 2019 at 5:20 PM Joe Perches <joe@xxxxxxxxxxx> wrote:
> > > > On Tue, 2019-10-08 at 11:40 +0200, Geert Uytterhoeven wrote:
> > > > > When reading a patch file from standard input, checkpatch calls it "Your
> > > > > patch", and reports its state as:
> > > > >
> > > > >     Your patch has style problems, please review.
> > > > >
> > > > > or:
> > > > >
> > > > >     Your patch has no obvious style problems and is ready for submission.
> > > > >
> > > > > Hence when checking multiple patches by piping them to checkpatch, e.g.
> > > > > when checking patchwork bundles using:
> > > > >
> > > > >     formail -s scripts/checkpatch.pl < bundle-foo.mbox
> > > > >
> > > > > it is difficult to identify which patches need to be reviewed and
> > > > > improved.
> > > > >
> > > > > Fix this by replacing "Your patch" by the patch subject, if present.
> > > > []
> > > > > diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
> > > > []
> > > > > @@ -1047,6 +1047,10 @@ for my $filename (@ARGV) {
> > > > >       }
> > > > >       while (<$FILE>) {
> > > > >               chomp;
> > > > > +             if ($vname eq 'Your patch') {
> > > > > +                     my ($subject) = $_ =~ /^Subject:\s*(.*)/;
> > > > > +                     $vname = '"' . $subject . '"' if $subject;
> > > >
> > > > Hi again Geert.
> > > >
> > > > Just some stylistic nits:
> > > >
> > > > $filename is not quoted so I think adding quotes
> > > > before and after $subject may not be useful.
> > >
> > > Filename is indeed not quoted, but $git_commits{$filename} is.
> >
> > If I understand your use case, this will only show the last
> > patch $subject of a bundle?
>
> False.
> "formail -s scripts/checkpatch.pl < bundle-foo.mbox" splits
> "bundle-foo.mbox" in separate patches, and invokes
> "scripts/checkpatch.pl" for each of them.
>
> > Also, it'll show things like "duplicate signature" when multiple
> > patches are tested in a single bundle.
>
> False, due to the splitting by formail.
>
> > For instance, if I have a git format-patch series in an output
> > directory and do
> >
> > $ cat <output_dir>/*.patch | ./scripts/checkpatch.pl
> >
> > Bad output happen.
>
> Yeah, because you're concatenating all patches.
> Currently it works for single patches only.
>
> > Maybe this might be better:
>
> > --- a/scripts/checkpatch.pl
> > +++ b/scripts/checkpatch.pl
> > @@ -2444,6 +2444,15 @@ sub process {
> >
> >                 my $rawline = $rawlines[$linenr - 1];
> >
> > +# if input from stdin, report the subject lines if they exist
> > +               if ($filename eq '-' && !$quiet &&
> > +                   $rawline =~ /^Subject:\s*(.*)/) {
> > +                       report("stdin", "STDIN", '-' x length($1));
> > +                       report("stdin", "STDIN", $1);
> > +                       report("stdin", "STDIN", '-' x length($1));
> > +                       %signatures = ();       # avoid duplicate signatures
> > +               }
> > +
> >  # check if it's a mode change, rename or start of a patch
> >                 if (!$in_commit_log &&
> >                     ($line =~ /^ mode change [0-7]+ => [0-7]+ \S+\s*$/ ||
>
> Perhaps.  Just passing the patchwork bundle to checkpatch, and fixing
> checkpatch to handle multiple patches in a single file was my first idea.
> But it looked fragile, with too much state that needs to be reset.
> I.e. the state is not limited to %signatures.  You also have to reset
> $author inside process(), and probably a dozen other variables.
> And make sure that future changes don't forget resetting all newly
> introduced variables.
>
> Hence I settled for the solution using formail.

I gave your solution a try.
It only enables the reset-on-next-patch feature when using stdin.
Thanks to the printed subject, it's now obvious to which patch a
message applies to.
However, the output is significantly different than when passing
a split patch series.  Can this be improved upon?

Note that the only reason I'm using stdin is that I use formail to split
a bundle in individual patches.  Once checkpatch supports bundles (or
mboxes) containing multiple patches, there's no longer a need for
using formail, and the reset-on-next-patch feature should be
enabled unconditionally.

Thanks!

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds