Boost logo

Boost :

From: Markus Schöpflin (markus.schoepflin_at_[hidden])
Date: 2008-07-29 04:39:05


Ruediger Berlich wrote:

> Hi there,
>
> I was wondering whether there would be a need for additional systems for
> automated, nightly builds. My employer would agree to make a few (3-4)
> machines available, with Linux-based operating systems. This would be a
> VMWare ESX-based setup, although on high-performance hardware (8-core
> blades).

I'm sure there is. Linux doesn't have a good test coverage on the release
branch, and OpenSolaris isn't tested at all.

> I would personally be most interested in a test environment with the
> OpenSUSE 11 64 bit / g++ 4.3.1 combo. It might be possible to also have an
> OpenSolaris/Intel installation there, or an installation with Intel's C++
> compiler.
>
> A pre-condition would be that no external access to the machines is needed,
> i.e. it would not be possible for Boost members to log in. Instead, I would
> set up the test environment as needed, and of course it would be possible
> for these machines to contact the outside world and transmit test results.

That's how the regression testing machines usually are set up. The machine
needs to be able to do http(s) downloads and FTP uploads.

> Thus my question, what needs to be done to set up such a test environment.

Have a look at http://www.boost.org/development/running_regression_tests.html.

> Note that I cannot give a guaranty for long-term availability, as a) I do
> not know how much work is involved in running such an environment and b) it
> might become necessary for the systems to be used in other contexts. If
> this turns out to be pretty much self-hosting (i.e. the build-environment
> is self-sustained on a normally configured Linux system), I would aim to
> ensure that the systems are available for a longer period.

Once it is set up, the testing runs pretty autonomous. It is demanding on
the hardware though, you need enough RAM (>1GB) and disk space, and a fast
CPU certainly helps.

HTH, Markus

PS: Since we are both located in the same city, don't hesitate to contact
me personally if you need some help getting the regression tests going. I
might be able to answer some things on the phone, if it helps.


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk