|
|
Boost Testing : |
From: Beman Dawes (bdawes_at_[hidden])
Date: 2005-07-15 08:15:23
"Carl Daniel" <cpdaniel_at_[hidden]> wrote in message
news:db783l$b6$1_at_sea.gmane.org...
> Rene Rivera wrote:
>> Carl Daniel wrote:
>>> Aleksey Gurtovoy wrote:
>>>> Is somebody else able/willing to pick up regular VC 7.1/8.0
>>>> regression testing?
>>>
>>> I now have access to weekly builds of VC8 (currently testing the
>>> 7/11/2005 build!). At the moment, it's not feasible to do boost
>>> regression testing on those builds (it's a machine at MSFT), but I
>>> could look into getting it set up.
>>>
>>> I now (once again) have the infrastructure in place locally to do
>>> boost regression testing on VC8 beta 2, and/or VC8 June CTP. Unless
>>> someone else volunteers, I can try to get my regression testing up
>>> and running again on either or both of these builds.
>>
>> Is that a "willing to run vc8 tests but not vc71" ?
>
> I'm willing to do both, but I don't have both on any one machine, so I'd
> be setting up and administrating fewer regression test rigs if I stick to
> VC8.
>
>> For those volunteering it would be good to mention what resources you
>> have available; CPU, memory, disk space, and machine time. As it
>> takes a considerable amount of resources for testing, especially disk
>> space. And having the faster machine and large ram will speed up the
>> testing.
>
> I have VC7.1 on a P4 3Ghz with 512Mb of RAM and lots of disk space, VC8
> beta 2 on a P4 2.8Ghz with 2Gb of RAM and lots of disk space, VC8 June CTP
> on a virtual server running on a P4 2.8Ghz with 3Gb of RAM and lots of
> space, and if I'm able to test the weekly builds of VC8, that's on a
> quad-proc Xeon 2.4Ghz with 2Gb of RAM and enough disk space.
>
>> I'm not trying to discourage people from volunteering though :-) Just
>> want to make sure we keep up a reasonable testing pace.
>
> When I was testing before, I was never able to get automation set up
> because of popup message boxes that stalled the tests (and weren't caught
> by the debug monitor program), and because of python.exe simply crashing
> out mid-run. Have those issues gotten any better? I'm certainly willing
> to test more configurations if I can get an automation set up that works
> reliably.
I had similar troubles until Victor suggested running the Boost tests as a
different user. So I set up a user named "Boost", and used the Windows
scheduler to start the tests as that user. Nothing else is ever done in that
account.
That seemed to be a big help. The monitor program is now running reliably,
and I'm no longer having to spend any time straightening out testing
problems.
Thanks again to Victor for his suggestion!
--Beman