Actor- and Task-Selection Strategies for Pruning Redundant State-Exploration in Testing

Abstract

Abstract. Testing concurrent systems requires exploring all possible non-deterministic interleavings that the concurrent execution may have. This is because any of the interleavings may reveal the erroneous behaviour. In testing of actor systems, we can distinguish two sources of non-determinism: (1) actor-selection, the order in which actors are explored and (2) task-selection, the order in which the tasks within each actor are explored. This paper provides new strategies and heuristics for pruning redundant state-exploration when testing actor systems by reducing the amount of unnecessary non-determinism. First, we propose a method and heuristics for actor-selection based on tracking the amount and the type of interactions among actors. Second, we can avoid further redundant interleavings in task-selection by taking into account the access to the shared-memory that the tasks make.

Test Case Generation of Actor Systems

“…As usual in the actor model [2,14,22], we suppose that a method does not return a value, but rather that its computation modifies the actor state. The language is simple to explain the contributions of the paper in a clear way, as done in [3,22].…”

“…Partial-order reduction (POR) [9,11] is a general theory that helps mitigate the problem by exploring the subset of all possible interleavings, which lead to a different configuration. Concrete algorithms have been proposed in [3,10,22] for dynamic testing.…”

“…In this section, we adapt to the context of symbolic execution and improve the notion of temporal stability of an actor introduced in [3] to avoid redundant exploration. This notion states that, at a given state, if we first select a temporarily stable actor, i.e., an actor to which no other actors will post tasks, unless it executes, it is guaranteed that it is not necessary to try executions in which the other actors in the state execute before this one, thus, avoiding such redundant explorations.…”

“…Briefly, the TCG framework that we propose has three main components, which are the contributions of this work: (1) in Sec. 3, we leverage the semantics used for testing in [3] to the more general setting of symbolic execution; (2) in Sec. 4, we extend and improve the techniques to avoid redundant computation of [3] to eliminate redundancies in symbolic execution and; (3) in Sec.…”

“…3, we leverage the semantics used for testing in [3] to the more general setting of symbolic execution; (2) in Sec. 4, we extend and improve the techniques to avoid redundant computation of [3] to eliminate redundancies in symbolic execution and; (3) in Sec. 5, we propose novel termination and coverage criteria, which guarantee termination of the process.…”