2013-06-11 04:49:17

by Rob Landley

[permalink] [raw]
Subject: [RFC] is it time to split up the MAINTAINERS file?

Quite possibly the answer is "no", but the MAINTAINERS file is
approaching 10,000 lines. Getting a bit unwieldy.

Most of the entries look like:

ARM/SAMSUNG MOBILE MACHINE SUPPORT
M: Kyungmin Park <[email protected]>
L: [email protected] (moderated for
non-subscribers)
S: Maintained
F: arch/arm/mach-s5pv210/mach-aquila.c
F: arch/arm/mach-s5pv210/mach-goni.c
F: arch/arm/mach-exynos/mach-universal_c210.c
F: arch/arm/mach-exynos/mach-nuri.c

Which could be moved to an arch/arm/MAINTAINERS file, and the relevant
paths trimmed.

The question is: would this be an improvement? (And worth the changes
to checkpatch.pl and such required to make it work?)

One potential _advantage_ of this is we could make the reporting
hierarchy more explicit. The first entry in arch/arm/MAINTAINERS would
be the arm maintainer and everybody else _under_ there goes through
him. (Also, that guy could handle updates to the local MAINTAINERS file
itself, so we're not always spamming Andrew. Such updates could even
post to the architecture-specific list rather than linux-kernel.)

Yeah, reality isn't neatly nested. Lots of things refer to include
files and Documentation files, but there's generally a main area of
focus (where's the actual _code_?), and when you do have something like:

ARM/SHMOBILE ARM ARCHITECTURE
M: Simon Horman <[email protected]>
M: Magnus Damm <[email protected]>
L: [email protected]
W: http://oss.renesas.com
Q: http://patchwork.kernel.org/project/linux-sh/list/
T: git
git://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas.git next
S: Supported
F: arch/arm/mach-shmobile/
F: drivers/sh/

You could either have the same entry in more than one MAINTAINERS file
or keep it at a higher level. (This wouldn't eliminate the top level
MAINTAINERS, merely trim it down a bit.)

Just throwing it out there. Seems like it might be a thing, someday
anyway...

Rob-


2013-06-11 07:13:47

by Joe Perches

[permalink] [raw]
Subject: Re: [RFC] is it time to split up the MAINTAINERS file?

On Mon, 2013-06-10 at 23:49 -0500, Rob Landley wrote:
> Quite possibly the answer is "no", but the MAINTAINERS file is
> approaching 10,000 lines. Getting a bit unwieldy.

I think it hasn't been much of a bottleneck problem.

> The question is: would this be an improvement? (And worth the changes
> to checkpatch.pl and such required to make it work?)

shrug.

MAINTAINERS is still the most frequently updated file.
This division eliminates those changes.

> One potential _advantage_ of this is we could make the reporting
> hierarchy more explicit. The first entry in arch/arm/MAINTAINERS would
> be the arm maintainer and everybody else _under_ there goes through
> him. (Also, that guy could handle updates to the local MAINTAINERS file
> itself, so we're not always spamming Andrew. Such updates could even
> post to the architecture-specific list rather than linux-kernel.)

MAINTAINERS updates aren't centralized.
There are lots of MAINTAINERS updates from sub-maintainers.

> Yeah, reality isn't neatly nested. Lots of things refer to include
> files and Documentation files, but there's generally a main area of
> focus (where's the actual _code_?), and when you do have something like:
>
> ARM/SHMOBILE ARM ARCHITECTURE
> M: Simon Horman <[email protected]>
> M: Magnus Damm <[email protected]>
> L: [email protected]
> W: http://oss.renesas.com
> Q: http://patchwork.kernel.org/project/linux-sh/list/
> T: git
> git://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas.git next
> S: Supported
> F: arch/arm/mach-shmobile/
> F: drivers/sh/
>
> You could either have the same entry in more than one MAINTAINERS file
> or keep it at a higher level. (This wouldn't eliminate the top level
> MAINTAINERS, merely trim it down a bit.)
>
> Just throwing it out there. Seems like it might be a thing, someday
> anyway...

Patches talk...

Sprinkling a few hundred MAINTAINER styles files
around the tree would be the biggest negative.

But paths and files could be relative and absolute

F: ./ (everything at this directory and lower)
F: ./foo.c (single file)
F: Documentation/foo.txt (absolute single file)

or add an initial / for the absolutes

F: */ (everything at this directory and lower)
F: foo.c (single file)
F: /Documentation/foo.txt (absolute single file)

make could be taught to create an overall integrated
MAINTAINERS, which would not be part of the files
managed by git/cvs from these submaintainer files.

Still, I think the "best" approach would be to enhance
git to manage this additional information instead.

Something akin to:
https://lkml.org/lkml/2007/8/14/256

Maybe some standardization of "git notes" or
"git annotate" might work.

A script could be written to create something like the
existing MAINTAINERS file from that too.

2013-06-11 17:10:25

by Rob Landley

[permalink] [raw]
Subject: Re: [RFC] is it time to split up the MAINTAINERS file?

On 06/11/2013 02:13:45 AM, Joe Perches wrote:
> On Mon, 2013-06-10 at 23:49 -0500, Rob Landley wrote:
> > You could either have the same entry in more than one MAINTAINERS
> file
> > or keep it at a higher level. (This wouldn't eliminate the top level
> > MAINTAINERS, merely trim it down a bit.)
> >
> > Just throwing it out there. Seems like it might be a thing, someday
> > anyway...
>
> Patches talk...

So you suggest sending a patch series to break out arch directories and
go "here, a whole new task for you the architecture maintainer to take
on!" and that's the _polite_ way to ask whether or not it's a good idea?

> or add an initial / for the absolutes
>
> F: */ (everything at this directory and lower)
> F: foo.c (single file)
> F: /Documentation/foo.txt (absolute single file)

That one, obviously. (Optimize for the common case.)

> make could be taught to create an overall integrated
> MAINTAINERS, which would not be part of the files
> managed by git/cvs from these submaintainer files.

Why? (What's the point? Does it make finding who is in charge of $THING
easier?)

> Still, I think the "best" approach would be to enhance
> git to manage this additional information instead.

Oh please no.

> Something akin to:
> https://lkml.org/lkml/2007/8/14/256

I'm sorry I brought it up.

> Maybe some standardization of "git notes" or
> "git annotate" might work.

The horror! The horror!

> A script could be written to create something like the
> existing MAINTAINERS file from that too.

I won't mention it again.

*shudder*

Rob-

2013-06-11 17:33:16

by Joe Perches

[permalink] [raw]
Subject: Re: [RFC] is it time to split up the MAINTAINERS file?

On Tue, 2013-06-11 at 12:10 -0500, Rob Landley wrote:
> On 06/11/2013 02:13:45 AM, Joe Perches wrote:
> > On Mon, 2013-06-10 at 23:49 -0500, Rob Landley wrote:
> > > You could either have the same entry in more than one MAINTAINERS
> > file
> > > or keep it at a higher level. (This wouldn't eliminate the top level
> > > MAINTAINERS, merely trim it down a bit.)
> > >
> > > Just throwing it out there. Seems like it might be a thing, someday
> > > anyway...
> >
> > Patches talk...
>
> So you suggest sending a patch series to break out arch directories and
> go "here, a whole new task for you the architecture maintainer to take
> on!" and that's the _polite_ way to ask whether or not it's a good idea?

I replied with suggestions, but I'm not about to
bother with implementations.

Throwing it out there without doing the work to
implement it doesn't generally inspire others to do
very much about it.

It's your itch, scratch away.

> > or add an initial / for the absolutes
> >
> > F: */ (everything at this directory and lower)
> > F: foo.c (single file)
> > F: /Documentation/foo.txt (absolute single file)
>
> That one, obviously. (Optimize for the common case.)

Not really as get_maintainers or another tool would
have to snarf all these files together before searching
once or every time it was run.

> > make could be taught to create an overall integrated
> > MAINTAINERS, which would not be part of the files
> > managed by git/cvs from these submaintainer files.
>
> Why? (What's the point? Does it make finding who is in charge of $THING
> easier?)

To make it easier for those that don't use git to find
maintainers and lists to send bug reports and patches.

> > Still, I think the "best" approach would be to enhance
> > git to manage this additional information instead.
> Oh please no.
> > Something akin to:
> > https://lkml.org/lkml/2007/8/14/256
>
> I'm sorry I brought it up.

<smile> What makes it horrific?
I think it'd be useful for more than lkml.