Analisando as relações entre mudanças no código fonte e bugs no software

Abstract

Considering the evolution of software systems, some services that were performed manually have migrated to computer applications such as complex software for financial control or systems that control aircraft flights. Although these services have been facilitated by the help of the technology, keeping them available requires care, either during the development phase of the program that provides these services or by the users who make use of these application. A still prevalent problem in systems is the occurrence of bugs in the software. These are commonly associated with anomalies in elementary units in systems, such as files, classes, functions, or methods. These parts of source code can often be too long or complex, which can compromise the bugs inspections and fixes. In order to do so, this work describes the use of a collection and analysis mechanism, about the changes in source codes that, commonly, can be associated to the appearance of bugs in the software. Such a mechanism is based on modifications made over time in atomic structures, which are often changed within the software, such as functions, methods and classes. As a result of the collection and analysis engine applications, we obtained a data set containing information for the change history of fourteen open-source projects, three of which were developed in C and C ++ programming languages, and eleven projects in JAVA. Two analysis were also carried out to evaluate the use of the collected data, of which the first was an analysis of the history of changes associated with software vulnerabilities (security bugs) for C, C ++ projects, and the second about the relationships between changes and bugs in JAVA projects. From the analysis made, it is possible to affirm that history of changes can serve as reference to several strategies that seek to improve the process of software development, be it in the creation of models for prediction or detection of bugs, or in the generation automatically of source codes.