[Numpy-discussion] Continuous Integration
ceball at gmail.com
Tue May 1 18:22:48 EDT 2012
Pauli Virtanen <pav <at> iki.fi> writes:
> 01.05.2012 11:14, David Froger kirjoitti:
> > Excerpts from Travis Oliphant's message of mar. mai 01 01:39:26 +0200 2012:
> > > If you have particular reasons why we should choose a particular CI
> > > please speak up and let your voice be heard. There is still time to
> > > a difference in what we are setting up.
> > Hi all,
> > What about buildbot? (http://trac.buildbot.net/)
> > I'm using it currently, and like it because is GPL 2, configuration
> > powerful Python scripts, and development team is active and dynamic.
> We're currently using it: http://buildbot.scipy.org
> Although, it hasn't been building automatically for some time now.
I've been working on setting up a new buildbot for NumPy. Unfortunately, I
don't have much time to work on it, so it's slow going! Right now I am still at
the stage of getting NumPy to pass all its tests on the machines I'm using as
test slaves. After that, I plan to transfer existing slaves to the new setup,
and then maybe ask for new volunteer slave machines (if people think the
buildbot setup is useful).
One nice feature of the new buildbot setup is that it can build and test a pull
request, reporting the results back via a comment on the pull request.
> It has the following shortcomings:
> - Apparently, for one master instance, only one project.
As others have said, this is no longer true.
> - Last time I looked, Git integration was poor. Maybe this has
> improved since...
Again, as others have said, I think this has been improved since last time you
looked. Buildbot can receive github notifications or can poll a repository
On the other hand, I've also been trying ShiningPanda (https://
jenkins.shiningpanda.com/scipy/) because I like the idea of a hosted CI
solution. Again, though, I haven't yet got all the tests to pass on
ShiningPanda's default Debian 6 setup (and again, that's because of a lack of
So, I'm working on Buildbot and ShiningPanda from the community side, but am
always ready to step aside if someone else has time :)
More information about the NumPy-Discussion