On Wed, 22 Mar 2006, Matthew Orgass wrote:

> On 2006-03-22 msporleder@gmail.com wrote:
>
> > As I started to think about it more, however, I wondered what testing
> > could actually be performed on an operating system.  It's such a
> > complex mix of features.  Testing all of the binaries/libraries should
> > be pretty easy (ls works, compress works, libxyz works, etc), but
> > testing kernel features and network stuff seemed like it could be very
> > difficult.
>
>   With unit tests, serial consoles, xen, and hardware emulators it is
> *possible* to automate testing of anything.  Of course, then you need to
> write and fix bugs in the tests too :).  See also "fuzzing" for a test
> method that is often very successful with little effort.  pkgsrc is good
> for this too... general use, the test target, bulk builds, bulk builds
> over a network...

The FreeBSD kernel stress test project appears to get interesting
results -- http://people.freebsd.org/~pho/stress/log/index.html

-- 
Kirk Russell            <kirk@ba23.org>            http://www.ba23.org/
Bridlewood Software Testers Guild                  Ottawa Ontario Canada