Re: [GIT PULL] KVM updates for the 3.4 merge window

[Avi Kivity <avi@redhat.com>]

On 03/25/2012 10:51 PM, Benjamin Herrenschmidt wrote:
> On Sun, 2012-03-25 at 12:09 +0200, Avi Kivity wrote:
>
> > Well I've been doing this ever since I moved to git.  The motivation was
> > actually to make things easier for patch authors by allowing them to
> > work against a tree of all applied patches, while the update for the
> > next merge window is just a subset, with more fixes going into the merge
> > window even late in the cycle, and features being deferred to the next
> > one.  I also fold fixes or reverts into their parent patches to improve
> > bisectability.
> > 
> > I can switch to fast-forward-only in the future, but I'm afraid that
> > this particular tree is broken for good.  The un-rebased
> > fast-forward-only source for this is kvm.git master, which I don't think
> > you want to pull.  It will cause every kvm commit to appear twice and
> > confuse everyone. 
>
> The problem is that it makes it very hard if not impossible to work
> with a combination of your tree & other trees (for example at some point
> I had to work on a merge of alex'tree, powerpc-next and pci-next).
>
> I don't see the problem with using the standard way and having
> sub-maintainers/developers.... Most of my sub-maintainers work on top of
> some upstream or they branch off my -next branch (which is known to
> never be rebased, so it's resync'ed as soon as Linux pulls it)

Say a fix comes in which needs to be mainlined during -rc.  So I put it
in some other branch, to be sent off to Linus in a few days after
maturing a little.  Meanwhile developers see an incomplete tree, since
that patch is not in it.

Once Linus pulls, I can merge it back (or even before, if I'm reasonably
certain it's not going to change), but it leaves a history of unneeded
merges.  Or we can do throwaway merges like tip.git.


> . Dealing
> with branches & merges in git is trivial and easier than dealing with
> the clashes caused by the rebases :-)
>
> One thing I do, is to also sometimes put out a powerpc-test branch that
> people know can and will be rebased, it's purely there if I want some
> folks to test a bunch of stuff but without basing their own work on top
> of it.
>
> And yes, there's a drawback vs. bisectability. You can still fold-in if
> you pickup patches from the list (vs pulling from sub-maintainers) as
> long as you haven't committed them to a "non-rebase" branch (ie, you can
> let things stage in a test branch for example for a couple of weeks to
> flush out those issues).

Right, we'll probably do something along these lines.

-- 
error compiling committee.c: too many arguments to function