Event-Based Relative Debugging

Identifying the root causes of failures is one of the most time-consuming and tedious components of program debugging. Many automated approaches have been proposed to facilitate program debugging. Relative debugging is one of them. In relative debugging, a faulty program is debugged by comparing it to a reference implementation, which is assumed to be correct. Both programs are executed in parallel. The executions are interrupted at certain points during executions. The states of the programs at the point of interruption are then compared and the differences (if any) are visualized as a debugging aid. One downside of existing relative debugging approaches is that they only compare program states, without taking the history of events causing those states into account. In this work we present an event-based relative debugging approach. In this approach, we infer finite state machine models from sequences of events occurring in executions, compute the structural differences between these models, and report them as a debugging aid. A case study conducted at a small-scale shows promise