timing test run on buildbot

[Vincent Ladeuil]

>>>>> "robert" == Robert Collins <robertc at robertcollins.net> writes:

    robert> On Wed, 2009-09-16 at 08:44 +0200, Vincent Ladeuil wrote:
    robert> It needs to not use --parallel, because I haven't taught
    >> the
    robert> concurrency-folding code to preserve intervals properly.
    >> 
    >> Hmm, how about you fix that then ? I'm a bit tired to push back
    >> --parallel use because it tends to regress as soon as I stop
    >> using it...

    robert> Well keep using it! This is just one specific use; and yes its on my
    robert> todo to fix... ETIME.

Sorry, that turned out to sound more grumpy that I wanted :-/

    robert> This would be useful for making some ongoing reports
    robert> of test timing data.
    >> 
    >> Hmm, isn't it something that hudson will be far better at ? And
    >> able to gather them for all runs ? 

    robert> I don't see that hudson would be intrinsically better
    robert> or worse at it.

I thought it already proposed some graphs for that kind of
things, I may be wrong but that was what prompted my remark.

    robert> We don't really need it from all runs (though getting
    robert> the failures in introspectable format would be great
    robert> :P).

Switching to hudson means using subunit for all runs in my head.

          Vincent