[Buildbot-devel] How to implement some sort of "speculative builds"?

Giovanni Gherdovich g.gherdovich at gmail.com
Wed Mar 4 08:43:33 UTC 2015

Hello Pierre,

thanks for sharing your workflow. It's really helpful to see what
other people in the industry are doing (FWIW, here is electronic payment
at Amadeus. Opinions are my own etc etc).

On Mon, Mar 2, 2015 at 4:05 PM, Pierre Tardy <tardyp at gmail.com> wrote:
> After doing this for a while, we got some metrics about regression found
> in various stages of the process, and we found out that the serialized
> *never* found any regression, because those have already been found
> in the parrallel pre-integration builders.
> So we changed to: [...]

I am not completely sold on "serialized tests don't add much value".
There is a detail I didn't mention in my last message:
at the moment we are facing some nasty heisenbugs (mostly due
to synchronization problems) that make our tests *sometimes* fail even
on correct code. This is the worst thing you can have,
because people are tempted to say "not my fault" in front of a test failure.
And while we have devs working on said bugs, my role as an integrator
is to keep everybody in line and assure the best possible quality.

This is to say: it is very important for me to have all devs
see black on white how their code change is behaving when merged to master.
If I allow those "multi-pull-request merges" I will have failures
that nobody feels responsible for.

For the moment I'll stick to my plan to dive into buildbot internals
and come up with a patch that implement optimistic builds and
all the needed bookkeeping.

As a side note, I've been shown that the OpenStack developers use
this very same integration strategy for their software:

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://buildbot.net/pipermail/devel/attachments/20150304/6cb87ea6/attachment.html>

More information about the devel mailing list