Analysis of Depth-First Search Based Mousavian Technique’s Scalability for Software Bug Localization

Debugging in software development is one of the most time-consuming processes. To that end, various bug localization techniques have been developed. One approach showing great result is the graph mining-based bug localization techniques. Unfortunately, the graph mining algorithm tends to scale poorly when dealing with large graph input. Mousavian et al. developed a new bug localization technique that replaces the graph mining algorithm with a new, more scalable graph analyzing technique. However, the mousavian technique has not been tested using highly complex graphs as input. In addition, its suspicious paths generation algorithm was not explained in detail to enable complete implementation. The test result shows a suspicious paths generation algorithm can be implemented using a modified depth- first search algorithm. Furthermore, the mousavian technique can be categorized as a scalable bug localization technique because it has a polynomial growth pattern for both processing time and maximum memory usage. However, the proposed depth-first search algorithm shows an exponential growth pattern for its process time. The processing time of the depth-first search algorithm can be significantly reduced by limiting the depth-first search depth limit at the cost of information completeness.