Which patches are NOT for 4.0?

[Jeremy Allison]

On Tue, Sep 25, 2012 at 09:25:38AM +0200, Stefan (metze) Metzmacher wrote:
> Hi Andrew,
> 
> > I just wanted to raise an interesting observation from watching the GIT
> > tree since rc1.  That is, when selecting the patches we want to get in
> > the rc2 release, we might want to work not just off the bugs in
> > bugzilla, but instead look over *all* the patches in master.
> 
> We should look over all patches and create bugreports for all logic
> subunits.
> E.g. one bug report for all sorts of quota related fixes.
> Or one bug report for smb2.durable-v2-open.app-instance related tests
> and the implementation.
> 
> > I don't see this as us 'releasing from master', but instead just noting
> > that there isn't any feature work going on yet, and that everyone is and
> > has been working exclusively on release-critical bugs.  
> > 
> > Naturally, this gets harder as soon as someone starts on features for
> > 4.1, but for now, I just wanted to raise the idea as either:
> >  - a benchmark to compare with (so we don't miss something)
> >  - or a way perhaps two of us could work over the patch list and prepare
> > a single, larger bug with most/all of rc1..master changes in it. 
> > 
> > To be totally clear: I'm not wanting to change our policy on what can be
> > put into RC at this point, just trying to make the mechanics simpler for
> > all involved.
> 
> I'm maintaining a branch locally where I'll rebase origin/master on
> origin/v4-0-test
> in order to see which things are missing, I think that's enough.
> 
> I think for documentation changes we could use a bulk bugreport/patchset,
> but the rest should be based on isolated bug reports.
> 
> And there're already things in master which we don't need in 4.0.
> E.g. the "libtorture: factor out simple ui backend" change and some
> pysmb changes
> doesn't need to be in 4.0
> If it turns out just before some release, that we only have a few
> patches between
> between master and v4-0-test, we can think about picking them for
> consistency,
> but I wouldn't decide that now.

I'm really happy with the current system of creating bug
reports for *every* change that goes into 4.0.0 now we've
reached rc.

No blanket backports or cherry-picks allowed, we must
think about *every* change we want to make, and it must
be reviewed appropriately.

We need to have this kind of discipline if we're going to
create a quality 4.0.0 release.

Jeremy.