Hyrum K. Wright wrote:
> On Oct 22, 2009, at 1:17 PM, Bhuvaneswaran A wrote:
>> On Thu, 2009-10-22 at 13:49 -0400, Mark Phippard wrote:
>>> Before we backport it, let's be sure we have finished fine-tuning it
>>> on trunk. For example, I do not think the times it is reporting is
>>> right. 12 minutes for all the tests? I do not think so. Also see
>>> this page that has times:
> Kinda hopin' not to put that URL in the email archive just yet...
>>> Being able to see trends in the test times is one of the things I am
>>> hoping to get out of this so it would be good to be sure it is right.
>> It's the time taken by "make check" process, yes! It seem to be a
>> with 4GB RAM having *4* processors, each running "Intel(R) Core(TM)2
>> Quad CPU @ 2.66GHz". Hyrum, correct me if i'm wrong.
> This is indeed the configuration of the machine.
>> The sum of all >1 seconds test cases sum up to 12.6 minutes.
> That does seem rather quick. I just ran 'make check' on a separate
> working copy on the same machine, and got something different:
> real 32m51.756s
> user 6m40.321s
> sys 6m27.504s
> I don't know that this is a problem with the junit converter script,
> or perhaps the test running reporting bogus ELAPSED times.
I do believe that the elapsed times added in the patch did not take
account of the test setup and especially cleanup times. The latter could
easily account for half of the test runtime.
Upon further consideration, I think that we should be recording
milliseconds for elapsed times, too.
Received on 2009-10-22 22:09:57 CEST