Killing obsolete Jenkins builds

[Ashod Nakashian]

On Fri, Nov 20, 2015 at 5:43 AM, David Tardon <dtardon at redhat.com> wrote:

> Hi,
>
> On Thu, Nov 19, 2015 at 05:55:01PM -0500, Ashod Nakashian wrote:
> > On Thu, Nov 19, 2015 at 5:13 PM, Norbert Thiebaud <nthiebaud at gmail.com>
> > wrote:
> >
> > > On Thu, Nov 19, 2015 at 10:39 AM, Stephan Bergmann <
> sbergman at redhat.com>
> > > wrote:
> > > > By the way, one situation where it is debatable whether all the
> triggered
> > > > builds are useful is if you push a series of changes to gerrit, and
> > > Jenkins
> > > > does builds for each of the changes in the series. For me at least,
> in
> > > such
> > > > a situation it would suffice if Jenkins just only did a build for the
> > > > topmost change.
> > >
> > > Evey commit should build.. bibisect will build each and every one of
> > > them eventually
> > >
> > >
> > But isn't that the bibisect build instance, which is unrelated to the
> > gerrit builds?
>
> You did not understand what Norbert was trying to say... Every commit
> should _be buildable_. The bibisect builder pick the commits to build at
> random, not only the last commit in a series (because there is no such
> thing as a commit series in git).
>
>
Maybe I'm confusing things, but in my mind gerrit commits are distinctly
different from upstream commits.

I'm assuming bibisect needs to build commits made into upstream, not the
gerrit submissions that may or may not make it into upstream.

Now, when a patch is sent to gerrit, we want to validate whether or not
it's buildable. That's good. But if the submitter amends the patch, then it
matters not whether the original builds or fails, because it has been
amended by a new change and the last one is all that matters.

This thread is about cancelling builds on the first version of the patch,
and building only the latest amended version.

In other words, it has the same strong guarantees that bibisect would
expect, but with less waste on unnecessary builds on outdated patches.


(I have made note-to-self to be more explicit about the scenario I'm
talking about when starting similar discussions. Clearly there are too many
closely-related concepts that it's hard to expect everyone has the same
frame of mind.)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/libreoffice/attachments/20151120/8c756259/attachment.html>