linux-next: manual merge of the xen tree with the swiotlb-xen tree

linux-next: manual merge of the xen tree with the swiotlb-xen tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 708 bytes --]

Hi all,

Today's linux-next merge of the xen tree got conplex conflicts in
drivers/xen/events.c between several commits from the swiotlb-xen tree
and several commits from the xen tree.

I am unqualified to fix this mess, sorry.  I am dropping the xen tree
for today (and only not dropping the swiotlb-xen tree because that would
be too much work).

I wonder, of course, why I have seen nothing in either of these trees
until today (less than a week before the merge window opens).

Please sort this mess out.  I will drop the swiotlb-xen tree as well
tomorrow if nothing has changed.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 490 bytes --]

Re: linux-next: manual merge of the xen tree with the swiotlb-xen tree

[Jeremy Fitzhardinge]

 On 10/17/2010 09:52 PM, Stephen Rothwell wrote:
> Hi all,
>
> Today's linux-next merge of the xen tree got conplex conflicts in
> drivers/xen/events.c between several commits from the swiotlb-xen tree
> and several commits from the xen tree.
>
> I am unqualified to fix this mess, sorry.  I am dropping the xen tree
> for today (and only not dropping the swiotlb-xen tree because that would
> be too much work).

Argh, sorry, I should have coordinated with Konrad.  I'd merged my tree
into linux-next to make sure it was OK, but I guess I overlooked something.

> I wonder, of course, why I have seen nothing in either of these trees
> until today (less than a week before the merge window opens).

I had found Xen regressions that had made it into linux-next and wanted
to sort those out before adding anything else to the mix.

> Please sort this mess out.  I will drop the swiotlb-xen tree as well
> tomorrow if nothing has changed.

Will do.

Thanks,
    J

Re: linux-next: manual merge of the xen tree with the swiotlb-xen tree

[Konrad Rzeszutek Wilk]

On Mon, Oct 18, 2010 at 01:11:57AM -0700, Jeremy Fitzhardinge wrote:
>  On 10/17/2010 09:52 PM, Stephen Rothwell wrote:
> > Hi all,
> >
> > Today's linux-next merge of the xen tree got conplex conflicts in
> > drivers/xen/events.c between several commits from the swiotlb-xen tree
> > and several commits from the xen tree.
> >
> > I am unqualified to fix this mess, sorry.  I am dropping the xen tree
> > for today (and only not dropping the swiotlb-xen tree because that would
> > be too much work).

Ugh. Let me revert the branch to pull this from to what it was a week ago.
That way Jeremy's tree should not have any trouble.

> 
> Argh, sorry, I should have coordinated with Konrad.  I'd merged my tree
> into linux-next to make sure it was OK, but I guess I overlooked something.
> 
> > I wonder, of course, why I have seen nothing in either of these trees
> > until today (less than a week before the merge window opens).

That is my fault. I thought the merge window was this _week_ so I happily
pushed the trigger button.

Since I seem to have a knack for making mistakes, let me get some clarification
so that I won't do it again.

When is it Ok for me to put in the #linux-next, new stable features (so
reviewed, acked, etc)? a) Is it post rc7? b) Or is it when Linus releases
the kernel and Linus's merge window opens?

I was thinking it is a), b/c if it would be b), then there is just two weeks
to fix any fallout that might happen due to merging of vast new features.

> 
> I had found Xen regressions that had made it into linux-next and wanted
> to sort those out before adding anything else to the mix.
> 
> > Please sort this mess out.  I will drop the swiotlb-xen tree as well
> > tomorrow if nothing has changed.
> 
> Will do.

Sorry about this. Will definitly work this out.

> 
> Thanks,
>     J

Re: linux-next: manual merge of the xen tree with the swiotlb-xen tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 2056 bytes --]

Hi Konrad,

On Mon, 18 Oct 2010 09:55:12 -0400 Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> wrote:
>
> Since I seem to have a knack for making mistakes, let me get some clarification
> so that I won't do it again.
> 
> When is it Ok for me to put in the #linux-next, new stable features (so
> reviewed, acked, etc)? a) Is it post rc7? b) Or is it when Linus releases
> the kernel and Linus's merge window opens?

Stuff destined for 2.6.n should be enter linux-next (after being posted,
reviewed and tested) between 2.6.(n-1)-rc1 and about 2 weeks before 2.6.
(n-1) (usually -rc6 or 7).  That way, hopefully we will have most
integration problems sorted out before Linus releases 2.6.(n-1) and the
code can go into Linus' tree during the merge window.  We can then sort
out any bugs etc in Linus's tree during the -rc releases for 2.6.n (at
the same time putting features for 2.6.(n+1) into linux-next.

So the answer is c) pre -rc7 :-)

i.e. all the new features for 2.6.37 should have been in linux-next at
least a week ago ...

In an idea world, the week before the merge window should be spent
settling stuff down ready to go into Linus' tree.  In the real world,
everyone seems to madly shove their new features into linux-next, rebase
their trees against some recent version of Linus' tree or do merges with
his tree (sometimes all three).  A lot of this latter is completely
unnecessary and just means that all the previous testing is thrown away
with the introduction of more code from Linus' tree.

Have a look at the graphs at http://neuling.org/linux-next-size.html .
In the last week, we have had ~930 new commits enter linux-next (that is
over 12% of the total in linux-next).  Since -rc7 came out, that number
is 1580 commits (over 20%). So instead of flattening off (and
stabilising) as we approach the merge window, the rate of change tends to
accelerate ...

/me gets off his soap box :-)

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 490 bytes --]