Re: Linux 2.6.11.2

[Matt Mackall <mpm@selenic.com>]

On Fri, Mar 11, 2005 at 01:45:46PM -0500, Bill Davidsen wrote:
> Matt Mackall wrote:
> >On Wed, Mar 09, 2005 at 12:11:02PM +0100, Pavel Machek wrote:
> >
> >>On St 09-03-05 09:52:46, Marcos D. Marado Torres wrote:
> >>
> >>>-----BEGIN PGP SIGNED MESSAGE-----
> >>>Hash: SHA1
> >>>
> >>>On Wed, 9 Mar 2005, Greg KH wrote:
> >>>
> >>>
> >>>>which is a patch against the 2.6.11.1 release.  If consensus arrives
> >>>>that this patch should be against the 2.6.11 tree, it will be done that
> >>>>way in the future.
> >>>
> >>>IMHO it sould be against 2.6.11 and not 2.6.11.1, like -rc's that are'nt 
> >>>againt
> >>>the last -rc but against 2.6.x.
> >>
> >>You expect people to go through all 2.6.11.1, 2.6.11.2, ... . That
> >>means .11.2 should be relative to .11.1, because otherwise people will
> >>have to revert (ugly). And you want people to track -stable kernels as
> >>fast as possible.
> >
> >
> >There are three ways we can do this:
> >
> >a) all 2.6.x.y are diffs against 2.6.x
> >b) interdiffs for .1, .2, etc. with 2.6.x+1 diffed against 2.6.x
> >c) interdiffs and 2.6.12 is a diff against 2.6.11.last
> >
> >Imagine we want to go from 2.6.11.3 to 2.6.12
> >
> >case a)
> >revert patch 2.6.11.3
> >get and apply 2.6.12
> 
> Would anyone actually do that? About the time of the first patch usually 
> do something like:
>   cd linux-2.6.11
>   cp -rl . ../linux-2.6.11.1
>   cd $_
>   bzcat ../Patches/patch-2.6.11.1.bz2 | patch -p1
>   make oldconfig

In your world, do you want to do:

cp -rl linux-2.6.11 linux-2.6.11.5
cd linux-2.6.11.5
bzcat ../Patches/patch-2.6.11.1.bz2 | patch -p1
bzcat ../Patches/patch-2.6.11.2.bz2 | patch -p1
bzcat ../Patches/patch-2.6.11.3.bz2 | patch -p1
bzcat ../Patches/patch-2.6.11.4.bz2 | patch -p1
bzcat ../Patches/patch-2.6.11.5.bz2 | patch -p1

I suspect you might find that tedious, especially if only the last one
addressed a bug that affected you.

Or do you want to do it the same way you do for every other branch? I
don't want to special-case it in my code and I don't think users want
to special-case it in their brains. Have separate interdiffs on the
side, please, and then people can choose, but do it the standard way.

Dear ${SUCKER}s, can we have a decision on this? My ketchup tool is
broken for 2.6.11.2 and I don't want to cut a new release until a firm
decision is made. Obviously I have a strong preference for all 2.6.x.y
diffs being against 2.6.x, it means that .y can be treated the same as
-rc, -bk, -mm, ... (and I already coded it that way when 2.6.8.1 came
out).

-- 
Mathematics is the supreme nostalgia of our time.