"Environment tests"

Andreas Raab andreas.raab at gmx.de
Fri Nov 7 10:14:50 UTC 2003


Hi Lex,

> If a program passes SUnit tests when run within a certain environment,
> then hasn't there been a decent "environment test" happening 
> here?  The idea being: a correct environment is one in which
> the program works according to the tests.

Yes, absolutely. If a series of (well-written) tests passes in a given
environment the environment is probably "compatible". What's more
interesting to me is what it means if the tests fail and what (if anything)
we can tell from that about our environment. This is why I am interested in
modelling the environment a little more explicitily, e.g., one could
rightfully argue that even unit tests should contain a decent amount of
tests about assumptions in the environment.

> This approach gives a straightforward and very useful way to approach
> the problem Andreas describes.  If every package has an associated set
> of tests, then the tests can be executed automatically every once in a
> while.  The results of these tests can be fed back into the SqueakMap
> entry for the package.

Yes. 

> Incidentally, it is important to think about what versions of 
> Squeak we *care* about a package working with.  In general, there are
> just two interesting versions for most people: the current stable version,
> and the current unstable version.  I could see maybe a third version
> emerging over time, some sort of super-stable release.  This should be
> food for thought before tilting at the windmill of accurate version
> dependencies.

This may or may not be true. For various reasons, people may want to try out
packages in older Squeak versions (or older packages in newer Squeak
versions) and it would be incredibly helpful if the assumptions about our
environment would be available in a somewhat more explicit manner.

Cheers,
  - Andreas




More information about the Squeak-dev mailing list