Analisando suítes de teste manuais e automáticas para identificar faltas de refatoramento.

Abstract

Developing systems with high quality involves activities that allow easy maintenance and provide confidence about the code produced. Software tests are related to reliability, already refactoring, with maintainability. By definition, refactoring editions aim to improve the code's structure, while preserving its behavior. However, bad refactoring can change the behavior of the system, they are called refactoring faults. Such faults may not be detected by unreliable test suites. An alternative for the systematic creation of test suites is the use of automatic generation tools. This work aims to evaluate the effectiveness of test suites generated manually and automatically to detect refactoring faults of the Extract Method type. For this, projects written in Java were selected, with test suites generated manually, new test suites were created automatically with the tools Randoop and EvoSuite, a set of faults were injected into the systems. Manual suites detected 61.9% of injected faults, while the Randoop suite detected only 46.7% and EvoSuite 55.8%. Randoop obtained a low detection rate, EvoSuite, however, obtained a result significantly comparable to that of manual suites.