A brief discussion on the lkml began whan it was asked why Andrew Morton [interview] isn't using git [story] to manage his -mm Linux kernel tree [story]. Andrew summarized, "because everthing would take me 100x longer?" He went on to explain that this is because he and Linus have a different way of working, "he reverts about one patch a month. I drop tens a day. He never _alters_ patches. 2.6.15-mm1 had about 200 patches which modify earlier patches and which get rolled up into the patch-which-they-modify before going upstream."

Andrew went on to note that even if he took the time to manage the -mm tree with git, it wouldn't be all that useful to people, "my tree at any random point in time is a random piece of doesn't-even-compile-let-alone-run crap, believe me. Often not all the patches even apply. I don't think there's much point in exposing people to something like that." That said, Andrew is exploring offering each -mm release under git, or even managing the individual patches themeselves with git.

From: Coywolf Qi Hunt [email blocked]
To:  akpm
Subject: why no -mm git tree?
Date:	Wed, 11 Jan 2006 13:56:16 +0800

hello,

Why don't use a -mm git tree? Maybe it was time for it.
With a -mm git tree, we can help -mm test much earlier and quicker,
and no more need of the mm-commits ML.

Also an option, to use git, and still gernerate broken-out from git.
-- 
Coywolf Qi Hunt



From: Andrew Morton [email blocked]
Subject: Re: why no -mm git tree?
Date:	Tue, 10 Jan 2006 22:44:51 -0800

Coywolf Qi Hunt <qiyong@fc-cn.com> wrote:
>
> Why don't use a -mm git tree?
>

Because everthing would take me 100x longer?

I'm looking into generating a pullable git tree for each -mm.  Just as a
convenience for people who can't type "ftp".

That'll just be a dump of the whole -mm lineup into git.  I don't know how
workable it'll be - we'll see.



From: Coywolf Qi Hunt <qiyong@fc-cn.com>
Subject: Re: why no -mm git tree?
Date:	Wed, 11 Jan 2006 15:00:43 +0800

On Tue, Jan 10, 2006 at 10:44:51PM -0800, Andrew Morton wrote:
> Coywolf Qi Hunt <qiyong@fc-cn.com> wrote:
> >
> > Why don't use a -mm git tree?
> >
> 
> Because everthing would take me 100x longer?

Really? So does Linus?

> 
> I'm looking into generating a pullable git tree for each -mm.  Just as a
> convenience for people who can't type "ftp".

That doesn't help much if it's only for each -mm.
If you make git commits for each each patch merged in, then
we can always run the `current' -mm git tree.

Test the -mm patches, not leave them sleeping for most of the time.

> 
> That'll just be a dump of the whole -mm lineup into git.  I don't know how
> workable it'll be - we'll see.
> 

-- 
Coywolf Qi Hunt



From: Andrew Morton [email blocked]
Subject: Re: why no -mm git tree?
Date:	Tue, 10 Jan 2006 23:18:18 -0800

Coywolf Qi Hunt <qiyong@fc-cn.com> wrote:
>
> On Tue, Jan 10, 2006 at 10:44:51PM -0800, Andrew Morton wrote:
> > Coywolf Qi Hunt <qiyong@fc-cn.com> wrote:
> > >
> > > Why don't use a -mm git tree?
> > >
> > 
> > Because everthing would take me 100x longer?
> 
> Really? So does Linus?
> 

Linus does a totally different thing from me.

He reverts about one patch a month.  I drop tens a day.

He never _alters_ patches.  2.6.15-mm1 had about 200 patches which modify
earlier patches and which get rolled up into the patch-which-they-modify
before going upstream.

He never alters the order of patches.

etc.

> > 
> > I'm looking into generating a pullable git tree for each -mm.  Just as a
> > convenience for people who can't type "ftp".
> 
> That doesn't help much if it's only for each -mm.
> If you make git commits for each each patch merged in, then
> we can always run the `current' -mm git tree.

Ah.  If you're suggesting that the -mm git tree have _patches_ under git,
and the way of grabbing the -mm tree is to pull everything and to then
apply all the patches under the patches/ directory then yeah, that would
work.

But my tree at any random point in time is a random piece of
doesn't-even-compile-let-alone-run crap, believe me.  Often not all the
patches even apply.  I don't think there's much point in exposing people to
something like that.

• Archive of above thread
• KernelTrap interview with Andrew Morton