Testing Concurrent Systems Through Event Structures

Complex systems are everywhere and are part of our daily life. As a consequence, their failures can range from being inconvenient to being life-threatening. Testing is one of the most widely accepted techniques (especially in industry) to detect errors in a system. When the requirements of the system are described by a formal specification, conformance testing is used to guarantee a certain degree of confidence in the correctness of an implementation; in this setting a conformance relation formalizes the notion of correctness. This thesis focuses on conformance testing for concurrent systems. Conformance testing for concurrent system has mainly focused on models that interpret concurrency by interleavings. This approach does not only suffer from the state space explosion problem, but also lacks the ability to test properties of the specification such as independence between actions. For such reasons, we focus not only on partial order semantics for concurrency, but also propose a new semantics that allows to interleave some actions while forcing others to be implemented as independent. We propose a generalization of the ioco conformance relation, based on Petri nets specifications and their partial order semantics given by their unfoldings, preserving thus independence of actions from the specification. A complete testing framework for this conformance relation is presented. We introduce the notion of global test cases which handle concurrency, reducing not only the size of the test case, but also the number of tests in the test suite. We show how global test cases can be constructed from the unfolding of the specification based on a SAT encoding and we reduce the test selection problem to select a finite prefix of such unfolding: different testing criteria are defined based on the notion of cut-off events. Finally, we show that assuming each process of a distributed system has a local clock, global conformance can be tested in a distributed testing architecture using only local testers without any communication.