Mitch wrote:

For instance, while some developers love such a thing, other test engineers ask "How do you test such a thing?" If the answer comes back "with great difficulty", moving forward with an XML strategy suddenly gets a whole lot harder.

If the result of evaluating the generated code is predictable,
ie. if the generated code is consistent, then it is absolutely
subject to testing.  You can always generate the tests too...

Yes, that's were I've had the best successes. Test engineers really appreciate the ability to characterize hardware and software
modules and parameterize test suites, with a consistent template system merging the information and generating tests.
(I wonder if predictability of the transformation can lead one to
make a stronger statement, something like a proof by induction...)

I don't think testing is a show-stopper, but circumstances may
vary - I'd have to consider different cases.

Cheers,
Steve Ball