[PATCH] Solve issues with flapping patches

[Martin Schwenke]

On Fri, 30 Jun 2017 16:30:34 +1200, Andrew Bartlett
<abartlet at samba.org> wrote:

> On Fri, 2017-06-30 at 14:28 +1000, Martin Schwenke wrote:
> > On Fri, 30 Jun 2017 16:20:06 +1200, Andrew Bartlett
> > <abartlet at samba.org> wrote:
> >   
> > > On Fri, 2017-06-30 at 14:09 +1000, Martin Schwenke via samba-
> > > technical
> > > wrote:  
>  [...]  
> > > 
> > > That would leave the most critical components of the AD DC
> > > untested,
> > > which isn't acceptable either.  
> > 
> > As I said, the tests can be switched back on by the developers who
> > want
> > to test that code.  So, all of your testing can run those tests.
> > 
> > The autobuild process is used as a gatekeeper for getting good
> > code/tests into the tree.  It doesn't help you when I test critical
> > components of the AC DC and flakiness causes my autobuilds to
> > fail! You don't even notice it... but I do...  :-)  
> 
> How about this to go with it then?
> 
> diff --git a/script/autobuild.py b/script/autobuild.py
> index e4a52ee7ff5..1b265173ed0 100755
> --- a/script/autobuild.py
> +++ b/script/autobuild.py
> @@ -69,7 +69,6 @@ tasks = {
>                 ("configure", "./configure " + ctdb_configure_params,
> "text/plain"),
>                 ("make", "make all", "text/plain"),
>                 ("install", "make install", "text/plain"),
> -               ("test", "make autotest", "text/plain"),
>                 ("check-clean-tree", "../script/clean-source-tree.sh",
> "text/plain"),
>                 ("clean", "make clean", "text/plain") ],
>  
> I think we can both agree this is absurd, but removing tests is an
> absurd answer to this problem.

Well, that would be absurd because the CTDB test failures happen 90%
less than the others, so they're not playing a major role in keeping
people's work out of the tree. This isn't black and white, it is a
matter of degree: clear the most significant blockers first.

Also, there is a known work-around for the flakiness in the CTDB tests.
There's no workaround for the others.

Apart from that, I'm out. This is a waste of time. I should have spent
today working on the new CTDB connection tracking code, which will now
only make it into 4.8.

Sorry for raising the issue...

peace & happiness,
martin