It would be nice to have something that makes sure every patch/PR builds
correctly and no unit tests fail. That would save the reviewer some time
when that happens but that tends to be a rare case. Even with that, I would
either have to build locally or download the build artifact to test the
fix, so I'm not sure there's much to gain there. I think our
PR-to-build-failure rate is < 5% but I don't have the numbers to back that
(yet ;) ).


On Fri, Jun 22, 2012 at 12:48 PM, Lance Speelmon <[email protected]> wrote:

> Good points - I certainly do not want to dilute the code review practices
> already in place.  Just trying to reduce the amount of time it takes to do
> so… Any ideas on how to get the best of both worlds?  i.e. letting CI do
> what it is good at but not compromising code review?
>
>
> On Jun 22, 2012, at 11:26 AM, Chris Tweney <[email protected]>
> wrote:
>
> > For the most part, I think committers need to build every PR they
> > review, for a couple of reasons:
> >
> > 1) Smoke testing and edge case testing is part of the review process. If
> > you don't build it, you obviously aren't running it and making sure it
> > works. Code reviews that only involve reading code tend to drift toward
> > mere enforcement of style standards, which is necessary, but not
> > sufficient. We adopted code reviews because there was a time when a lot
> > of unfinished, untested work was getting committed.
> >
> > 2) We have a lot of commits going in, and most builds have multiple
> > commits. If a build breaks, everyone who committed in that build needs
> > to stop their work and check the build. And everyone unlucky enough to
> > pull from master during the broken build time will have their work
> > interrupted. That amounts to more people and more time lost than if the
> > committer runs a build.
> >
> > Now, there are occasional exceptions -- I just merged a pull from Zach
> > that I didn't build. But it was a 2-line change that just introduced the
> > use of a constant in place of a numeric literal. Not the kind of thing
> > that really needed a second compile. (Of course, if the constant was
> > misspelled, I'll have egg on my face in a little while... but then, so
> > will Zach, for submitted an uncompiled change!)
> >
> > One thing that would save a lot of my time is if I had a second box to
> > do builds on. I should look into getting an EC2 instance to do builds
> > on. I think my usage might fit into the free usage tier.
> >
> > -chris
> >
> >
> > On 6/22/12 9:39 AM, Lance Speelmon wrote:
> >> Is it correct in assuming we have CI builds for every commit in
> projects?  If so, could I suggest that people's time is too valuable to
> spend building every PR manually?
> >>
> >> WDYT?  Thanks, L
> >> _______________________________________________
> >> oae-dev mailing list
> >> [email protected]
> >> http://collab.sakaiproject.org/mailman/listinfo/oae-dev
> > _______________________________________________
> > oae-dev mailing list
> > [email protected]
> > http://collab.sakaiproject.org/mailman/listinfo/oae-dev
>
> _______________________________________________
> oae-dev mailing list
> [email protected]
> http://collab.sakaiproject.org/mailman/listinfo/oae-dev
>
_______________________________________________
oae-dev mailing list
[email protected]
http://collab.sakaiproject.org/mailman/listinfo/oae-dev

Reply via email to