[Buildbot-devel] adhoc code reviews

Dustin J. Mitchell dustin at zmanda.com
Mon Nov 2 18:49:11 UTC 2009


On Mon, Nov 2, 2009 at 12:28 PM, Marc-Antoine Ruel <maruel at chromium.org> wrote:
> Just wondering, from what I've seen, when I want to push something to
> djmitche/buildbot/master, I just make a push request to Dustin and he
> commits it if it looks good to him. Rinse and repeat. And when he
> feels like to, he make a new release out of master.

That's been the plan so far, but I'm not wedded to it :)

> As for most(?) open source projects, there's no real formal code
> reviews. No big deal. The only code review tools I know are rietveld
> and gerrit, both fairly new, both made by google (first by Guido,
> second by Android team). I looked at Rietveld (which is mostly
> svn-based and is what we use for Chromium) but it just behaves just
> "fine enough" with git. See http://codereview.appspot.com/144075/show
> for an example. Note that git-cl makes it bearable with git. I haven't
> really played with Gerrit but at least it's 100% git based and is more
> logical to use with a git back-end since it understand branches and
> can merge automatically. It seems it's quite involved to install.
> There's probably other better git-friendly, web-based, open source
> code review tools that I don't know of. (?)

GitHub supports some similar things.

We have a mandatory review process at Zmanda, too, and it works quite
well but that requires the buy-in of all of the developers.  When
folks are away or not doing reviews for some reason, the works gum up
quite quickly..

> - Any opinion on code reviews? Just an overhead? Useful to follow
> developments? Don't care?

In general, I try to "just commit" things that do not adversely affect
other operations, or that are clearly correct.  For example, adding a
new BuildStep subclass won't hurt any other parts of buildbot.  For
more significant changes, I usually recommend that the author post to
the list.  In many cases, though, those posts go unanswered.

If we can all make a commitment to review patches posted to the list,
then I can start to push more patches that way.

> - If so, would people appreciate a basic setup to try it out? Wouldn't
> use it anyway?

I don't think we would overcome the inertia required to start using a
new tool for such a thing -- at least in the short term, it should be
on the mailing list.

Dustin

-- 
Open Source Storage Engineer
http://www.zmanda.com




More information about the devel mailing list