Debugging by Diagnosing Assumptions

We present a novel and uniform technique for normal logic program declarative error diagnosis. We lay down the foundations on a general approach to diagnosis using logic programming, and bring out the close relationship between debugging and fault{{nding. Diagnostic debugging can be enacted by contradiction removal methods. It relies on a simple program transformation to provide a contradiction removal approach to debugging, based on revising the assumptions about predicates' correctness and completeness. The contradiction removal method is justiied in turn in terms of well{ founded semantics. The loop detection properties of well{founded semantics will allow in the future for a declarative treatment of otherwise endless derivations. The debugging of programs under well{founded semantics with explict negation is also foreseen. Here, we apply our techniques to nite SLDNF derivations, whose meaning coincides with the well{founded model, for which our contradiction removal method and algorithm is sound and complete. Several examples illustrate the algorithm at work.