Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.31 as permitted sender) client-ip=23.128.96.31;
MIME-Version: 1.0
References: <20231004-get_maintainer_change_k-v1-1-ac7ced18306a@google.com>
 <3dca40b677dd2fef979a5a581a2db91df2c21801.camel@perches.com>
 <CAFhGd8oTHFDQ05M++E3ggAvs0567w5fSxovumX+vs8YXT8VXTA@mail.gmail.com> <6e13b9b1a964b49079a2f7814c0d65e767cd010a.camel@perches.com>
In-Reply-To: <6e13b9b1a964b49079a2f7814c0d65e767cd010a.camel@perches.com>
From:   Justin Stitt <justinstitt@google.com>
Date:   Thu, 5 Oct 2023 11:30:48 -0700
Message-ID: <CAFhGd8rGr3fm-U3XCjRkJQRymvjqGFYzsPu61zbMZCebuN5Rww@mail.gmail.com>
Subject: Re: [PATCH] get_maintainer/MAINTAINERS: confine K content matching to patches
To:     Joe Perches <joe@perches.com>
Cc:     linux-kernel@vger.kernel.org, Kees Cook <keescook@chromium.org>,
        Nick Desaulniers <ndesaulniers@google.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Oct 5, 2023 at 11:15=E2=80=AFAM Joe Perches <joe@perches.com> wrote=
:
>
> On Thu, 2023-10-05 at 11:06 -0700, Justin Stitt wrote:
> > On Wed, Oct 4, 2023 at 7:40=E2=80=AFPM Joe Perches <joe@perches.com> wr=
ote:
> > >
> > > On Wed, 2023-10-04 at 21:21 +0000, Justin Stitt wrote:
> > > > The current behavior of K: is a tad bit noisy. It matches against t=
he
> > > > entire contents of files instead of just against the contents of a
> > > > patch.
> > > >
> > > > This means that a patch with a single character change (fixing a ty=
po or
> > > > whitespace or something) would still to/cc maintainers and lists if=
 the
> > > > affected file matched against the regex pattern given in K:. For
> > > > example, if a file has the word "clang" in it then every single pat=
ch
> > > > touching that file will to/cc Nick, Nathan and some lists.
> > > >
> > > > Let's change this behavior to only content match against patches
> > > > (subjects, message, diff) as this is what most people expect the
> > > > behavior already is. Most users of "K:" would prefer patch-only con=
tent
> > > > matching. If this is not the case let's add a new matching type as
> > > > proposed in [1].
> > >
> > > I'm glad to know  you are coming around to my suggestion.
> > :)
> >
> > >
> > > I believe the file-based keyword matching should _not_ be
> > > removed and the option should be added for it like I suggested.
> >
> > Having a command line flag allowing get_maintainer.pl
> > users to decide the behavior of K: is weird to me. If I'm a maintainer =
setting
> > my K: in MAINTAINERS I want some sort of consistent behavior. Some
> > patches will start hitting mailing list that DO have keywords in the pa=
tch
> > and others, confusingly, not.
>
> Not true.
>
> If a patch contains a keyword match, get_maintainers will _always_
> show the K: keyword maintainers unless --nokeywords is specified
> on the command line.

...

>
> If a file contains a keyword match, it'll only show the K:
> keyword  if --keywords-in-file is set.

Right, what I'm saying is a patch can arrive in a maintainer's inbox
wherein the patch itself has no mention of the keyword (if
get_maintainer user opted for --keywords-in-file). Just trying to
avoid some cases of the question: "Why is this in my inbox?"

>
> > To note, we get some speed-up here as pattern matching a patch that
> > touches lots of files would result in searching all of them in their
> > entirety. Just removing this behavior _might_ have a measurable
> > speed-up for patch series touching dozens of files.
>
> Again, not true.
>
> Patches do _not_ scan the original modified files for keyword matches.
> Only the patch itself is scanned.  That's the current behavior as well.
>

Feel like I'm missing something here. How is K: matching keywords in
files without reading them.

If my patch touches 10 files then all 10 of those files are scanned for
K: matches right?