ctdb in autobuild broken -- ctdb depends on winbind now????

[Martin Schwenke]

On Fri, 06 Jan 2017 21:05:44 +1300, Andrew Bartlett
<abartlet at samba.org> wrote:

> On Fri, 2017-01-06 at 18:25 +1100, Martin Schwenke wrote:
> > On Fri, 6 Jan 2017 07:59:18 +0100, Volker Lendecke <vl at samba.org>
> > wrote:
> >   
> > > 
> > > 
> > > Sure, but as I said: This is a winbind failure, and I don't have
> > > the
> > > chance to take a serious look today. Leaving it broken is also no
> > > option. So please push the reverts just posted.  
> > 
> > As discussed on IRC...
> > 
> > Can you please merge in my autobuild branch with the reverts and run
> > a
> > (or several) private autobuild?  If it passes reliably then:
> > 
> > Reviewed-by: Martin Schwenke <martin at meltin.net>
> > 
> > If it doesn't fix the flaky tests then there's no use reverting.
> > 
> > This also gives someone else time to look into the failures.  I've
> > copied the logs for the last failed autobuild to:  
> 
> Revert or not, I think we should remember that GIT is a distributed
> version control system, and while we use a centralised repo, we are
> well past the days of CVS or SVN were everyone really did stop on a
> breakage.  

Sure, but with autobuild we have a single point of failure.  We need to
push code often so we don't end up with divergent development trees.

> It is well worth taking the time to understand and fix things, just as
> you have with ctdb, rather than the rushed revert of a whole series of
> patches (in the hope of accidentally including the problem patch).
> 
> We are under no artificial pressure here: 4.6.0rc1 was (wisely)
> branched directly, and is already released.
> 
> I say this because rushed reverts (and in general other changes made in
> haste) have not turned out to be such a great thing in hindsight. 

Well, the question is always: how long do we wait?  :-)

If someone has the time to debug this then they should say so and offer
that as an alternative solution...

peace & happiness,
martin