Developer´s assumptions on identifying bug-introducing changes

Abstract

During the code review, developers can make assumptions that may guide their review process (e. g., developers may consider that code with low test coverage is more likely to introduce bugs. Thus, they can focus their attention on this code). Although studies have investigated developers’ assumptions involving different concerns in software engineering, none of those studies focus on analyzing developers’ assumptions when reviewing code to identify bug-introducing changes. Our study investigates those assumptions, how relevant they are for developers, why developers put more/less relevance for some assumptions, and which assumptions are consensus/dissensus among groups of developers. We apply the Qmethodology to conduct our study with 41 developers analyzing 41 assumptions extracted from different sources. The results indicate: (i) the developers’ assumptions involve concerns related to the programming language, developers’ experience and habits, organizational practices as well as code maintenance, ownership and testing; (ii) five common viewpoints among developers regarding the relevance of assumptions; (iii) code complexity and ownership, short delivery time, developers practices and familiarity with the code are the main reasons for developers to put more/less relevance to some assumptions; and (iv) while assumptions involving programming languages, features invocation and pull-request acceptance present more consensus among developers, assumptions involving merge commits, software history, familiarity with the programming language, code reuse and ownership present more dissensus. These findings are valuable knowledge for both practitioners and researchers towards improving state-of-the-art code review tools/techniques to identifying bug-introducing changes.