Architectural-Based Speculative Analysis to Predict Bugs in a Software System

Over time, a software system's code and its underlying design tend to decay steadily and, in turn, to complicate the system's maintenance. In order to address that phenomenon, many researchers tried to help engineers predict parts of a system that are most likely to create problems while or even before they are modifying the system. Problems that creep into a system may manifest themselves as bugs, in which case engineers have no choice but to fix them or develop workarounds. However, these problems may also be more subtle, such as code clones, circular dependencies among system elements, very large APIs, individual elements that implement multiple diffuse concerns, etc. Even though such architectural and code 'smells' may not crash a system outright, they impose real costs in terms of engineers' time and effort, as well as system correctness and performance. Along the time, implicit problems may be revealed as explicit problems. However, most current techniques predict explicit problems of a system only based on explicit problems themselves. Our research takes a further step by using implicit problems, e.g., architectural- and code-smells, in combination with explicit problems to provide an accurate, systematic and in depth approach to predict potential system problems, particularly bugs.