Debugging as a science, that too, when your program is changing

Abstract

Program debugging is an extremely time-consuming process, and it takes up a large portion of software development time. In practice, debugging is still very much of an art, with the developer painstakingly going through volumes of execution traces to locate the actual cause of an observable error. In this work, we discuss recent advances in debugging which makes it systematic scientific activity in its own right. We explore the delicate connections between debugging and formal methods (such as model checking) in the overall task of validating software. Moreover, since any deployed software undergoes changes in its lifetime, we need debugging methods which can take the software evolution into account. We show how symbolic execution and Satisfiability Modulo Theories (SMT) solvers can be gainfully employed to greatly automate software debugging of evolving programs. © 2010 Elsevier B.V.