Boost logo

Boost Testing :

Subject: Re: [Boost-testing] Trunk testing (general) and MSVC11
From: Christoph Macheiner (christoph.macheiner_at_[hidden])
Date: 2012-08-23 13:29:31

"Tom Kent" <lists_at_[hidden]> wrote in message
> On Thu, Aug 23, 2012 at 6:05 AM, Christoph Macheiner
> <christoph.macheiner_at_[hidden]> wrote:
>> I am trying to setup a trunk test runner for MSVC11, and I have a few
>> issues:
> I've also got one of these coming up...its hacking through the first run.

Saw (when the site worked again) that there already is one up, so for now
stick to 64-bit tests only (which not seem to be overly prominent anyway).
time permits, I'll run the 32-bit tests as well for comparison (a little
difficult otherwise
because the other "nightly" runs seem to catch slightly different svn

>> - How are the MSVC testers handling crashes (of which I had several
>> during
>> this first run)? By default, the JIT-debugger kicks in with a message box
>> which is obviously not good for an unattended run.
> Its annoying that they popup, but it doesn't seem to affect the
> tests...I'm assuming that the tester gets the errror code or just
> times out when this happens.

I have temporarily registered cdb as jit debugger (I want to create dumps
the tests crash). But good to know that there is a timeout after which the
test is
killed (I overlooked this at first).

>> - Even on a quite modern machine (quad-core processor, enough RAM),
>> "performance" is rather poor. I know this is quite a general statement,
>> but
>> without further ado, did anybody look into this? The above-mentioned
>> out-of-disk occured after about 7 hours, with the tests obviously still
>> not
>> finished. Point is, I can spare the night on my dev rigs, but if the
>> tests
>> take longer than say, 8 hours, I am out of luck. Especially when I want
>> to
>> test 32-bit and 64-bit.
> My VMs usually take ~12hrs to do the first build. After that, I set
> the --incremental flag and I get runs in 1.5-4hrs.

Yes, I read about the incremental option but to be on the safe side I'd
run the full test. As mentioned in my other post, we'll see how long it
with the -j bjam option.


Boost-testing list run by mbergal at