Detecting Infeasible Branches Based on Code Patterns

Infeasible branches are program branches that can never be exercised regardless of the inputs of the program. Detecting infeasible branches is important to many software engineering tasks such as test case generation and test coverage measurement. Applying full-scale symbolic evaluation to infeasible branch detection could be very costly, especially for a large software system. In this work, we propose a code pattern based method for detecting infeasible branches. We first introduce two general patterns that can characterize the source code containing infeasible branches. We then develop a tool, called Pattern-based method for Infeasible branch Detection (PIND), to detect infeasible branches based on the discovered code patterns. PIND only performs symbolic evaluation for the branches that exhibit the identified code patterns, therefore significantly reduce the number of symbolic evaluations required. We evaluate PIND from two aspects: accuracy and efficiency. The experimental results show that PIND can effectively and efficiently detect infeasible branches in real-world Java and Android programs. We also explore the application of PIND in measuring test case coverage,