|
|
Boost Testing : |
From: Aleksey Gurtovoy (agurtovoy_at_[hidden])
Date: 2007-11-07 22:00:03
John Pavel wrote:
>>>> Beman, thanks. The current (16:07) summary is from 02:41, which is
>>>> some way off 3.5h, so I am still a bit mystified, Best, John
>>> I'm suspecting some recent changes to the reports has made the
>> interval
>>> jump up significantly. Which is why I've been thinking of putting in
>>> some log of the interval. And in case anyone is curious I suspect
>> adding
>>> the rev number to all the pages has caused things to slow down.
>> I would be very surprised if that turned out to be the case. The change
>> in questions does not add neither any significant computations nor
>> more than a couple of HTML lines to an output page:
>>
>> http://svn.boost.org/trac/boost/changeset/40599
>
> Nonetheless, the current (11:28:09) results on beta.boost.org seem 12h
> behind (rather than 3.5h), unless I am misinterpreting something.
Report generation times are greatly affected by:
1) the overall number of failing test cases (an output page is generated
for each of these).
2) the number of updated toolsets since the last report generation cycle
(some of the intermediate report data is cached).
3) used version of xsltproc/libxml/libxslt.
-- not to mention general factors such as processor load.
IMHO it's much more likely that the increase in the cycle's length we
are experiencing is due to one of these than the changes in question.
-- Aleksey Gurtovoy MetaCommunications Engineering