Re: [boost] Regression logs

From: David Abrahams (dave_at_[hidden])
Date: 2003-03-29 09:46:56

Beman Dawes <bdawes_at_[hidden]> writes:

> I think Alisdair's raised some good points. While I'm not sure
> regression testers will want to put a lot of effort into back tests, I
> think it would be good if from here on out we segregated current from
> past tests.
>
> So I'd say give it a bit more thought and see what you can come up
> with. It will probably be easier to leave everything in one directory,
> but have the script display each release separately. Maybe default to
> current state, but with an option to look at past tests.
>
> Keep in mind the number of test platforms is likely to increase as
> time goes on. That also argues for better separation between releases
> and the current CVS state.

I just returned from PyConDC 2003, where a system was described which
will watch a CVS repository for checkins, launch new remote tests on
any number of platforms, collect the results, and notify people of
failures (e.g. via mail or even IRC chat).  It can be set up to bother
the person at fault first, and then bother a different group, for
example.  http://buildbot.sourceforge.net/.  It looks like it could
be really useful for us.

Brian Warner expressed interest in helping us get that BuildBot up for Boost
if we want to use it.  

Thoughts?

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com