Instrumentation and transformation of Java source code for automated testing with search-based testing algorithms

Abstract

Search-based Software Test Data Generation is a field of research treating test input generation as a search problem. Search algorithms require that an objective function computes the quality of solution candidates. When the goal for testing is to achieve code coverage, one type of objective function uses branch distance, a heuristic describing how “close” the condition in a control flow statement is from being true.

An attempt is made to define procedures to allow the calculation of branch distance for both predicates in conjunction and disjunction predicates with short-circuiting operators && and ||, by identifying and avoiding situations causing side effects and exceptions.

This thesis also attempts to give examples of instrumenting and transforming control flow statements, examples which are limited or lacking in the research literature.

A program implementing the defined rules for instrumentation and transformation of control flow statements was developed to validate these rules.