Fault Localization Using a Weighted Function Dependency Graph

We present a hierarchical scheme of fault localization (FL). First, we prioritize the functions based on suspiciousness. Subsequently, the fault is localized to specific statements within the suspected function. Our approach includes the proposal of a novel function dependency graph, based on which prioritization of a function is performed. We use function complexity metrics to break ties among the functions with same suspiciousness score. We proposed two scoring schemes to assign weights to the dependency edges viz., Highest Weight Normalization (HWN) Method and Distribution Specified Normalization (DSN) Method. Our experimental results show that on an average there is a reduction of 38.48% and 42.58% statements to be examined using HWN and DSN scoring schemes respectively as compared to the well-accepted spectrum based fault localization technique Tarantula.