launchpad-dev team mailing list archive

[Julian Edwards <julian.edwards@xxxxxxxxxxxxx>]

On Tuesday 09 November 2010 14:31:35 Francis J. Lacoste wrote:
> We used to run the windmill tests in a separate buildbot and they did
> bitrot and we only got to fixing them when it became a blocker for
> landing. Why would it be different this time?
> 
> That's why it should be a blocker for deployment. Either these tests have
> value and are preventing regressions in production, or they don't and we
> should just kill them.
> 
> The whole point of Deryck's experiment is to assess this cost/benefit, and
> while doing the experiment, its important to ensure reversibility which
> means no bitrot.

Right, that's exactly why I've suggested doing it this way.  I think the 
difference though as that I expect that not everything will be a deployment 
blocker and we can choose what to fix or leave/delete - I don't have any 
evidence for that though, it's just a gut feel.

Thinking about it, I guess we're actually suggesting the same thing but in my 
jetlagged haze, clear thinking is competing with closing eyelids!