Re: [Ksummit-discuss] [RFC PATCH 2/3] MAINTAINERS, Handbook: Subsystem Profile

[Julia Lawall]

On Sat, 17 Nov 2018, Rob Herring wrote:

> On Fri, Nov 16, 2018 at 11:57 AM Joe Perches <joe@xxxxxxxxxxx> wrote:
> >
> > On Fri, 2018-11-16 at 14:44 +0200, Jani Nikula wrote:
> > > I quickly cooked up this script to produce the top-5 commit prefixes for
> > > the given files over the arbitrary last 200 commits. It'll give you a
> > > pretty good idea if you're even close.
> > >
> > > ---
> > > #!/bin/sh
> > > # usage: subject-prefix FILE [...]
> > > # show top 5 subject prefixes for FILEs
> > >
> > > git log --format=%s -n 200 -- "$@" |\
> > >       grep -v "^Merge " |\
>
> --no-merges in git log can replace this line.
>
> > >       sed 's/\(.*\):.*/\1/' |\
> > >       sort | uniq -c | sort -nr | sed 's/ *[0-9]\+ //' |\
> > >       head -n 5
> > > ---
> > >
> > > Someone who knows perl could turn that into a checkpatch check: See if
> > > the patch subject prefix is one of the top-5 for all files changed by
> > > the patch, and ask the user to double check if it isn't. Or some
> > > heuristics thereof.
> >
> > This won't work when a patch contains multiple files
> > from different paths, or even multiple files from a
> > single driver.
>
> Different paths is often, but not always a sign that patches may need
> to be split up. Maybe that is something checkpatch should point out.

Between v4.0 and v4.19, 18% of commits touch multiple .c files.  35% of
commits touch multiple files in general.

julia

>
> > Perhaps it's better to use a generic mechanism like
> >
> >         basename $(dirname $filename):
> >
> > with some exceptions and add an override patch subject
> > grammar to appropriate various sections of MAINTAINERS.
>
> Perhaps just use the script as a starting point to populate
> MAINTAINERS as it may never be that accurate.
>
> > I also think it's better to use a separate script like
> > scripts/spdxcheck.py and tie any necessary checkpatch
> > use to that script.
>
> Yes, checkpatch is getting pretty unwieldy.
>
> Rob
>