The interlocutory tool box: techniques for curtailing coincidental correctness

Abstract

Eliminating faults in software systems is important, because they can have catastrophic consequences. This can be achieved by testing and debugging. Testing involves executing the system with a test case to obtain an output. The output is evaluated against the tester’s expectations; deviation from these expectations indicates that a fault has been detected. Debugging involves using information about the fault, that was gleaned during testing, to isolate the fault in the system. Coincidental correctness is a widespread phenomenon in which a fault corrupts a program state, and despite this, the system produces an output that satisfies the tester’s expectations. Coincidental correctness can compromise the effectiveness of testing and debugging techniques. This thesis investigated methods for alleviating coincidental correctness in testing and debugging. The investigation culminated in four techniques. The first technique is called Interlocutory Testing. Interlocutory Testing is a framework for the development of test oracles that are referred to as Interlocutory Relations. Interlocutory Relations are the first type of oracle that has been specifically designed to operate effectively in the presence of coincidental correctness. Metamorphic Testing was pioneered for testing non-testable systems. However, the effectiveness of this technique can be compromised by coincidental correctness. The second technique, Interlocutory Metamorphic Testing, is a version of Metamorphic Testing that has been integrated with Interlocutory Testing, to alleviate the impact of coincidental correctness on Metamorphic Testing. Interlocutory Mutation Testing is the third technique. This technique uses similar principles to Interlocutory Testing to alleviate the Equivalent Mutant Problem in the presence of coincidental correctness and non-determinism. Finally, the fourth technique is Interlocutory Spectrum-based Fault Localisation. This technique uses Interlocutory Relations to ameliorate the effects of coincidental correctness on fault localisation. Each technique was empirically evaluated. The results were promising, and indicated that these techniques were capable of mitigating the impact of coincidental correctness.