An Exploratory Study on Exception Handling Bugs in Java Programs

• Ano de defesa: 2013
• Autor(a) principal: Ebert, Felipe
• Orientador(a): Castor, Fernando
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Pernambuco
• Palavras-chave em Português: Exception handling; Bugs; Survey; Repository mining; Tratamento de exceções; Questionário; Mineração de repositórios
• Link de acesso: https://repositorio.ufpe.br/handle/123456789/11984

Several studies argue that exception handling code is usually of poor quality and that it is commonly neglected by developers. Moreover, it is said to be the least understood, documented, and tested part of the implementation of a system. In spite of this scenario, there are very few studies that analyze the actual exception handling bugs that occur in real software systems and no study that attempts to understand developers’ perceptions about these bugs. In this work we present an exploratory study on exception handling bugs that employs two complementary approaches: a survey of 154 developers and an analysis of 220 bugs from the repositories of Eclipse and Tomcat. Respondents of our survey believe that exception handling bugs are more easily fixed than other kinds of bugs. There is also a significant difference in the opinion of the respondents pertaining to the quality of the exception handling code: more experienced developers tend to believe that it is worse. Analysis of the repositories of Eclipse and Tomcat revealed conflicting results. The fix time for exception handling bugs in Eclipse is significantly shorter than for other bugs. However, exception handling bugs have a significantly greater number of discussion messages than non-exception handling bugs. On the other hand, for Tomcat, we could not find a significant difference for fix time and exception handling bugs have significantly less discussion messages than other bugs. Moreover, we discovered that bug reports describing bugs stemming from overly general catch blocks, a well-known bad smell in programs that use exceptions, are rare, even though there are many opportunities for them to occur. In addition, empty catch blocks are not only prevalent, as previously reported in literature, but they are also commonly used as part of bug fixes, which includes fixes for exception handling bugs. Furthermore, we found very few bug reports whose causes are empty catch blocks, although developers often mention them as causes of bugs they have fixed in the past. And lastly, we present a proposal of the classification of exception handling bugs based on the data we collected.