Aspect of Code Cloning Towards Software Bug and Imminent Maintenance: A Perspective on Open-source and Industrial Mobile Applications

Abstract

As a part of the digital era of microtechnology, mobile application (app) development is evolving with lightning speed to enrich our lives and bring new challenges and risks. In particular, software bugs and failures cost trillions of dollars every year, including fatalities such as a software bug in a self-driving car that resulted in a pedestrian fatality in March 2018 and the recent Boeing-737 Max tragedies that resulted in hundreds of deaths. Software clones (duplicated fragments of code) are also found to be one of the crucial factors for having bugs or failures in software systems. There have been many significant studies on software clones and their relationships to software bugs for desktop-based applications. Unfortunately, while mobile apps have become an integral part of today’s era, there is a marked lack of such studies for mobile apps. In order to explore this important aspect, in this thesis, first, we studied the characteristics of software bugs in the context of mobile apps, which might not be prevalent for desktop-based apps such as energy-related (battery drain while using apps) and compatibility-related (different behaviors of same app in different devices) bugs/issues. Using Support Vector Machine (SVM), we classified about 3K mobile app bug reports of different open-source development sites into four categories: crash, energy, functionality and security bug. We then manually examined a subset of those bugs and found that over 50% of the bug-fixing code-changes occurred in clone code. There have been a number of studies with desktop-based software systems that clearly show the harmful impacts of code clones and their relationships to software bugs. Given that there is a marked lack of such studies for mobile apps, in our second study, we examined 11 open-source and industrial mobile apps written in two different languages (Java and Swift) and noticed that clone code is more bug-prone than non-clone code and that industrial mobile apps have a higher code clone ratio than open-source mobile apps. Furthermore, we correlated our study outcomes with those of existing desktop based studies and surveyed 23 mobile app developers to validate our findings. Along with validating our findings from the survey, we noticed that around 95% of the developers usually copy/paste (code cloning) code fragments from the popular Crowd-sourcing platform, Stack Overflow (SO) to their projects and that over 75% of such developers experience bugs after such activities (the code cloning from SO). Existing studies with desktop-based systems also showed that while SO is one of the most popular online platforms for code reuse (and code cloning), SO code fragments are usually toxic in terms of software maintenance perspective. Thus, in the third study of this thesis, we studied the consequences of code cloning from SO in different open source and industrial mobile apps. We observed that closed-source industrial apps even reused more SO code fragments than open-source mobile apps and that SO code fragments were more change-prone (such as bug) than non-SO code fragments. We also experienced that SO code fragments were related to more bugs in industrial projects than open-source ones. Our studies show how we could efficiently and effectively manage clone related software bugs for mobile apps by utilizing the positive sides of code cloning while overcoming (or at least minimizing) the negative consequences of clone fragments.