Klaus Karcher wrote:
The more I think a about it, the more it becomes clear to me that I have to make additional test. Maybe the problem is too device specific to be generalized.
I think I addressed most of the issues in the new code. I've just picked 0,0,0 and 1,1,1, but it's designed to be graceful if the real test set has those values or ones close to them in it. Graeme.