A Multilevel Quantitative Analysis of Distribution of Defects in Open Source Software

摘要

Organizations encounter numerous problems that impede the growth of quality software systems. Defects can be detected and prevented by efficient study of their distribution in the systems. The Pareto principle is studied and applied in many fields of management and research, and it is widely used to study the distribution of defects in software systems as well. However, there is no study that analyzes defect distribution from the very first release until the end for open source software (OSS). This study provides a multilevel quantitative analysis of distribution of defects in OSS. The study explores the existence of the Pareto principle for 137 releases spread across seven open-source Java data sets. The study found: 1) changes due to defects are more prevalent than changes due to any other issues in the OSS repository; 2) defect distribution in OSS follows the Pareto principle at the combined release level, at the project level, and at the individual release level; 3) defect persistence exists for all subsequent releases and components, with the maximum number of defects having the maximum number of blocker and critical defects in the system.