[Buildbot-devel] RFC: more flexible web status display

[Neal Norwitz]

On 4/12/06, Brian Warner <warner-buildbot at lothar.com> wrote:
> > And while I'm here, I would like if the changes were grouped together
> > in the next batch.  Since the tests take an hour + to run, some slaves
> > get really far behind.  It would be nice to aggregate outstanding
> > changes into a batch even if they didn't arrive within the same five
> > minutes.
>
> They should be doing that already. Any changes that arrive after the current
> build has started but before the next one starts will be lumped together into
> the second build. In the case of SVN, this means that the second build will
> be run with something like "-r %d" % max([c.revision for c in changes]) . If
> you're seeing behavior that doesn't match this, point me at your Waterfall
> and I'll take a look.

It was in the previous mail:  http://python.org/dev/buildbot/all/

I don't think there was a problem today, but I think yesterday was
pretty bad.  Not sure if you can get the old data.  Also, some of the
changes were on the branch and we lock one branch while the other is
running.

> If you've got developers who are in the habit of spreading related commits
> out over 10 or 15 minutes, and want to improve the chances that their changes
> all get put into the same build, you can always crank up the treeStableTimer
> (an argument to the Scheduler class). But this comes at the expense of timely
> notification of build errors, of course.

No, it's not the developers fault.  It's just that sometimes we are
checking in from different continents. :-)

n