Automated Breakpoint Generation for Debugging

In debugging processes breakpoints are frequently used to inspect and understand the run-time program behavior. Although modern development environments provide convenient breakpoint utilities, it mostly requires considerable human effort to create useful breakpoints. Before setting a breakpoint or typing a breakpoint condition, developers usually have to make some judgements and hypotheses based on their observations and experiences. To reduce such kind of effort, we propose an approach to automatically generating breakpoints for debugging. We combine the nearest neighbor queries method, dynamic program slicing, and memory graph comparison to identify suspicious program statements and states. Based on this information, breakpoints are generated and divided into two groups, where the primary group contains conditional breakpoints and the secondary group contains unconditional ones. We implement the presented approach on top of the Eclipse JDT platform. Our objective experiments and user study indicate that the generated breakpoints can be effective in aiding debugging work.