We present a generic scheme for the abstract debugging of functional logic programs. We associate to programs a semantics based on a (continuous) immediate consequence operator, T[P], which models correctly the powerful features of modern functional logic languages (non-deterministic, non-strict functions defined by non-confluent programs and call-time choice behaviour). Then, we develop an effective debugging methodology which is based on abstract interpretation: by approximating the intended specification of the semantics of P we derive a finitely terminating bottom-up diagnosis method, which can be used statically. Our debugging framework does not require the user to provide error symptoms in advance and is applicable with partial specifications and even partial programs.
