[Axiom-developer] AxiomUnit

[daly]

Christian,

I'd encourage you to discuss AldorUnit/AxiomUnit on this list.
There are a lot of people who are interested.

The current regression testing machinery uses string comparison of
new output against previous output. It has the advantage of being
simple but handling most kinds of results, including failure messages.

It also steps around the zero-equivalence problem. How can we test
Axiom integration by using differentiation? They may be inverse 
operations in some cases but Axiom can't simplify the results to 0.

However AldorUnit may be useful for testing various aspects of the
system which are not handled well with the current simplistic scheme.
For instance, the current scheme does not handle randomly generated
input gracefully.

Any and all forms of testing are vital and are to be strongly encouraged.

Tim