Model Checking and Delta Debugging for Automated Fault Localization; A Tutorial

Debugging programs is a complicated and cumbersome task. It is an iterative process which includes hypothesizing the cause of a failure and validating the hypothesis. Since programs comprise of many components with complex interactions this chain often involves several trials. Fault localization (FL) is the task of identifying components that are likely to cause the failure. Automated fault localization techniques aim to accelerate the debugging process by guiding developers to a small portion of the program that is likely to be culpable for the program failures. In essence, they reduce the search space programmers need to inspect manually to locate the fault. There are several different approaches to fault localization. In this paper, we study two approaches: delta debugging, and model checking. Delta debugging framework offers techniques for simplification of test cases and isolation of failure-inducing parts from test cases, which help in debugging and can be adapted for fault localization. Delta debugging requires minimal knowledge about the structure of the program and specification. In this paper, we explain delta debugging algorithms and how it could be adapted for fault localization. Model checking is the problem of checking whether a model (e.g. program) satisfies a property. Model checking needs to know the internal structure of the programs that it works on. Since it can answer different queries about the execution path of the program, it can be employed for fault localization. In this paper, we study a couple of techniques that have been implemented in model checkers for fault localization.