launchpad-reviewers team mailing list archive

[Raphaël Badin <raphael.badin@xxxxxxxxxxxxx>]

> I don't get the impression that test_ensure_master_returns_oldest_nodegroup is
> quite enough to cover the "return the oldest one" functionality.  Whether
> ensure_master does the right thing doesn't really become obvious until you
> test with two nodegroups present, with different timestamps.
> 
> It takes a bit of tweaking to get different timestamps with certainty, because
> the database transaction clock doesn't advance during the test (and even if it
> did, because things may happen within the clock's granularity).  What I
> usually do when faced with this problem is create two objects as normal, but
> subtract a fixed timedelta from the first one's timestamp.

Arg, you're right! I forgot about that.

This is fixed now, thanks.
-- 
https://code.launchpad.net/~rvb/maas/ensure-master-takes-first/+merge/126498
Your team Launchpad code reviewers is requested to review the proposed merge of lp:~rvb/maas/ensure-master-takes-first into lp:maas.