Re: [PATCH] MAINTAINERS: Start using the 'reviewer' (R) tag

[Javier Martinez Canillas]

Hello Lee,

On Wed, Oct 28, 2015 at 2:34 PM, Lee Jones <lee.jones@xxxxxxxxxx> wrote:
> On Wed, 28 Oct 2015, Javier Martinez Canillas wrote:
>> On Wed, Oct 28, 2015 at 1:14 PM, Lee Jones <lee.jones@xxxxxxxxxx> wrote:
>> > On Wed, 28 Oct 2015, Lee Jones wrote:
>> >
>> >> On Wed, 28 Oct 2015, Javier Martinez Canillas wrote:
>> >>
>> >> > Hello Joe,
>> >> >
>> >> > On Wed, Oct 28, 2015 at 12:06 PM, Joe Perches <joe@xxxxxxxxxxx> wrote:
>> >> > > On Wed, 2015-10-28 at 11:53 +0100, Javier Martinez Canillas wrote:
>> >> > >> (Lee) think(s) that the difference between a maintainer and
>> >> > >> a reviewer is if a branch with fixes / new features are kept and pull
>> >> > >> requests sent while I think that the difference is the level of
>> >> > >> involvement someone has with a driver regardless of how patches ends
>> >> > >> in the subsystem tree (picked directly by subsystem maintainers or
>> >> > >> sent through pull requests).
>> >> > >>
>> >> > >> Is the first time I heard your definition but maybe I'm the one that
>> >> > >> is wrong so it would be great to get a consensus on that and get it
>> >> > >> documented somewhere.
>> >> > >
>> >> > > I think Lee is over-analyzing.
>> >> > >
>> >> > > From MAINTAINERS:
>> >> > >         M: Mail patches to: FullName <address@domain>
>> >> > >         R: Designated reviewer: FullName <address@domain>
>> >> > >            These reviewers should be CCed on patches.
>> >> > >         S: Status, one of the following:
>> >> > >            Supported:   Someone is actually paid to look after this.
>> >> > >            Maintained:  Someone actually looks after it.
>> >> > >
>> >> > > "looking after" doesn't mean upstreaming.
>> >> > >
>> >> >
>> >> > Agreed and upstreaming doesn't mean sending pull request, you can for
>> >> > example upstream the downstream changes for a driver you maintain by
>> >> > posting patches or ack patches others post and let the subsystem
>> >> > maintainer to pick those (even if you are listed as the driver
>> >> > maintainer in MAINTAINERS).
>> >> >
>> >> > So by following Lee's definition, then most drivers' maintainers
>> >> > should not be called maintainers since keeping a tree with patches for
>> >> > both fixes and new features, sending pull requests, etc is only
>> >> > justified for drivers that have a lot of changes per release. Is not
>> >> > worth it for drivers that are in "maintenance mode" where only bugs
>> >> > are fixed every once in a while or features are seldom added.
>> >>
>> >> Exactly right.
>> >>
>> >> Although, it looks like M: doesn't even mean Maintainer.  If it did, I
>> >> would have made these points over and over until death (or until I got
>> >> bored).  However, as M: actually means "Mail patches to", there seems
>> >> to be very little difference between that and "Designated reviewer"
>> >> and makes me wonder why the R: tag was ever even introduced.  I guess
>> >> all of the other guys in the threads below also thought M: meant
>> >> Maintainer, or else they would have just added poor old Josh as a
>> >> "Mail patches to" recipient and been done with it.
>> >
>> > Ah, but wait.  get_maintainer.pl *does* assume M means Maintainer
>> > doesn't it?  Which is why this came about.  So if we have a "Mail
>> > patches to" entry, get_maintainer.pl tells the user that this is a
>>
>> Joe already answered but I'll elaborate a little bit:
>>
>> "M:" means "Mail patches to" and "S:" means "Status" so what
>> get_maintainers.pl is reports the person in "M:" printing the status
>> of the file(s).
>>
>> So for example if the file has "S: Supported" then the person listed
>> in "M:" is shown as "supporter" while if the status is "S:
>> Maintained", the person is listed as "maintainer".
>>
>> There isn't a "Reviewed" status to specify files that are looked by
>> "Designated reviewers", it can be added but I don't see the reason of
>> it.
>
> Not sure why you wrote all of this.  We know *why* get_maintainer.pl
> does what it does.  What I'm saying is, that I personally believe this
> is the wrong behaviour in what I'm *guessing* is the majority of the
> time.
>

Ok sorry if it was clear to you already.

>> > Maintainer, which (given that there are over 1000 unique M: entries
>> > and I know that there are no where near that many actual Maintainers)
>> > means that it's printing out incorrect information most of the time.
>>
>> Well, that's correct according to your definition of maintainer
>> (people with a tree that sends pull requests) but I can believe that
>> there are 1K unique maintainers for different small components
>> (without taking into account what components may be obsolete / not
>> used anymore) even if these maintainers don't send pull request
>> because the patch load is low and rely on subsystem maintainers to
>> pick directly from the list with their Acks.
>
> Then they are not Maintainers.  They are Reviewers who rely on
> Maintainers.  In your example above it's the Maintainers that are
> Maintainers and the people who review the code are Reviewers.  The
> clues are in the words. ;)
>

They are not maintainers according to your definition of maintainer
that doesn't seem what most people agree with.

>> For me the maintainer is that a) cares about the file and makes sure
>> that things remain working, fix issues, reviews patches from others,
>> etc b) is someone that actually understand the code in the files. A
>> subsystem maintainer has the last word of what gets merged into the
>> subsystem but may not be familiar with code under the subsystem and
>> relies on the file / driver maintainer to Ack the patch as correct
>> since that person is the one that knows the code.
>
> Then what is a Reviewer?
>

As I mentioned before, I think a reviewer is someone who is interested
in a given file / driver / subsystem and reviews patches to the list
but does not have an authoritative answer on that file. I gave as an
example that I review patches posted for Exynos SoC support although
I'm not mentioned in MAINTAINERS and I don't plan it to be. People
don't know that I'm interested in that and don't cc me, I just review
what is posted to the linux-samsung-soc ML.

A designated reviewer is someone that is mentioned in the MAINTAINERS
file so people know that this person is interested and will add
him/her as cc. So designated reviewers are expected to review more
patches, be more available, etc and it seems the idea is that they can
become maintainers in the future if there is some need for it.

>> > So back to my original thought then, what can we do to rectify this
>> > situation and make the information printed more meaningful.  Again,
>> > I'm back to using the R: tag appropriately.
>> >
>> > Any (technical, which aren't based on "but I really want to be listed
>> > as a Maintainer") thoughts?
>> >
>>
>> I haven't read a "but I really want to be listed as a maintainer"
>> thought in this thread.
>
> If you had no vested interest in this, you would be able to see the
> logic in what I'm saying.  Again I'm *guess*, but I think there is
> some emotional reasons for you pushing back so hard.  I could always
> be wrong though.
>

Oh, I have no personal interest. I'm not even listed as a maintainer /
reviewer for anything that is MFD related and I do understand your
point is just that I don't agree. The fact that we disagree doesn't
mean that you are right and I'm wrong.

So if anything I think this thread is more about "but I really don't
want people without a tree to be listed as a Maintainer".

What I do care is about consistency across different subystems since
the kernel development process is already complicated and there are
already differences in the workflow of the maintainers so people
contributing to many subsystems have to first do some research to
learn all the unwritten rules and conventions of a given susbystem and
subsystem maintainer before posting patches.

That's why adding the semantics of maintainers and reviewers to the
mix of things to keep in mind is not something that thrills me.

>> I think the problem is the definition of what a maintainer in Linux
>> really means. For you is someone that keeps a tree and sends pull
>> request (for me that is what I call a subsystem maintainer)
>
> Right.  A Level 3 Maintainer sends pull-requests to a Level 2
> Maintainer, who sends pull-requests to a Level 1 Maintainer, who sends
> pull-requests to Linus.  Maintainers at *all* levels collect patches
> and Maintain them before sending on.
>

Here is where we disagree (and it seems is not only me), someone who
acks patches posted to some files even if those are later picked for
someone higher in the maintenance hierarchy could be named maintainers
as well since they "own" those files and their answer should be
authoritative (no matter how the patches end into mainline).

At the end of the day is to make easier for developers to understand
who are the interested parties for a given change. It is much easier
to think in terms of different levels of maintainers that should be
copied so they can review and possibly people outside of that group if
they are subscribed to the mailing lists. Naming the first levels
reviewers makes no sense for me since it will only confuse new comers
IMHO.

> reiterate:
>
> Reviewers care about particular driver(s)/domain(s) that they know
> well.  They provide solid reviews and possible testing (because they
> are likely to have the h/w).  They then provide Acked-by:s so that a
> Maintainer can collect the patch.
>
>> while for
>> others it seems that maintainer means someone who care about a set of
>> files, knows the code, makes sure that things keep working and Ack
>> patches to those files when posted.
>
>         goto reiterate;   ;)
>

No need for that, I already explained my point of view several times
and you just think I'm wrong because I don't agree with you. So let's
just agree on disagree ;-)

Best regards,
Javier
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/