Re: [PATCH] checkpatch: Fix false positive check for dt-binding docs and include

[Alex Tran]

On Sat, Jan 3, 2026 at 1:24 AM Bagas Sanjaya <bagasdotme@xxxxxxxxx> wrote:
>
> [CC'ing devicetree people]
>
> On Wed, Dec 24, 2025 at 05:41:17PM -0800, Alex Tran wrote:
> > The check for when to display a DT_SPLIT_BINDING_PATCH warning
> > is performed on all files that are not MAINTAINERS files.
> > This causes issues, specifically when checking against a
> > patch that only touches docs/dt-bindings or a patch that
> > only touches include/dt-bindings. The warning is still
> > displayed when it should not be, giving false positives.
> >
> > The check for when to issue a DT_SPLIT_BINDING_PATCH warning
> > is currently too pervasive and should only be performed when
> > a change under docs/dt-bindings is detected or include/dt-bindings.
> > If the current changed file is under include/dt-bindings or
> > docs/dt-bindings then it is compared to the last file detected
> > under include/dt-bindings or docs/dt-bindings. If there is a
> > difference then a warning is issued.
>
> Can you elaborate on example of where this issue occurs?
> I'm still confused...

I recently submitted a patch that converted a dt-binding from txt to
yaml format. The patch included the deletion of a txt file and the
addition of a yaml file under Documentation/devicetree. No files were
edited under the include/dt-bindings directory and yet the warning:
"DT binding docs and includes should be a separate patch.", was still
issued when running the checkpatch script.

> >
> > Signed-off-by: Alex Tran <alex.t.tran@xxxxxxxxx>
> > ---
> >  scripts/checkpatch.pl | 12 ++++++------
> >  1 file changed, 6 insertions(+), 6 deletions(-)
> >
> > diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
> > index c0250244cf7a3cb5e289b04fa4a9f644692dc278..fe3499bc9adda158c88231b48b3b0100f3af0ee1 100755
> > --- a/scripts/checkpatch.pl
> > +++ b/scripts/checkpatch.pl
> > @@ -2668,7 +2668,7 @@ sub process {
> >       my $authorsignoff = 0;
> >       my $author_sob = '';
> >       my $is_patch = 0;
> > -     my $is_binding_patch = -1;
> > +     my $is_doc_binding_patch = -1;
> >       my $in_header_lines = $file ? 0 : 1;
> >       my $in_commit_log = 0;          #Scanning lines before patch
> >       my $has_patch_separator = 0;    #Found a --- line
> > @@ -2924,13 +2924,13 @@ sub process {
> >                       }
> >                     > > +                             my $last_binding_patch_is_doc = $is_doc_binding_patch;  $checklicenseline = 1;
> >
> > -                     if ($realfile !~ /^MAINTAINERS/) {
> > -                             my $last_binding_patch = $is_binding_patch;
> > +                     if ($realfile =~ m@^(?:Documentation/devicetree/|include/dt-bindings/)@) {
> > +                             my $last_binding_patch_is_doc = $is_doc_binding_patch;
> >
> > -                             $is_binding_patch = () = $realfile =~ m@^(?:Documentation/devicetree/|include/dt-bindings/)@;
> > +                             $is_doc_binding_patch = () = $realfile =~ m@^(?:Documentation/devicetree/)@;
> >
> > -                             if (($last_binding_patch != -1) &&
> > -                                 ($last_binding_patch ^ $is_binding_patch)) {
> > +                             if (($last_binding_patch_is_doc != -1) &&
> > +                                 ($last_binding_patch_is_doc ^ $is_doc_binding_patch)) {
> >                                       WARN("DT_SPLIT_BINDING_PATCH",
> >                                            "DT binding docs and includes should be a separate patch. See: Documentation/devicetree/bindings/submitting-patches.rst\n");
> >                               }
> >
>
> I can't say for the actuall diff, though.
The diff is changing when the warning should be issued; If the current
file is under docs/dt-bindings or include/dt-bindings then it checks
if the last bindings file was a docs dt-binding file. And compares it
with the current file. If there is a mismatch then the current file
must be under include/dt-bindings and the warning should be issued,
otherwise not.

--
Yours,
Alex Tran