Date Awarded

Fall 2016

Document Type


Degree Name

Doctor of Philosophy (Ph.D.)


Computer Science


Denys Poshyvanyk

Committee Member

Denys Poshyvanyk

Committee Member

Peter Kemper

Committee Member

andreas Stathopoulos

Committee Member

Xu Liu


Mobile developers and testers face a number of emerging challenges. These include rapid platform evolution and API instability; issues in bug reporting and reproduction involving complex multitouch gestures; platform fragmentation; the impact of reviews and ratings on the success of their apps; management of crowd-sourced requirements; continuous pressure from the market for frequent releases; lack of effective and usable testing tools; and limited computational resources for handheld devices. Traditional and contemporary methods in software evolution and maintenance were not designed for these types of challenges; therefore, a set of studies and a new toolbox of techniques for mobile development are required to analyze current challenges and propose new solutions. This dissertation presents a set of empirical studies, as well as solutions for some of the key challenges when evolving and maintaining android apps. In particular, we analyzed key challenges experienced by practitioners and open issues in the mobile development community such as (i) android API instability, (ii) performance optimizations, (iii) automatic GUI testing, and (iv) energy consumption. When carrying out the studies, we relied on qualitative and quantitative analyses to understand the phenomena on a large scale by considering evidence extracted from software repositories and the opinions of open-source mobile developers. From the empirical studies, we identified that dynamic analysis is a relevant method for several evolution and maintenance tasks, in particular, because of the need of practitioners to execute/validate the apps on a diverse set of platforms (i.e., device and OS) and under pressure for continuous delivery. Therefore, we designed and implemented an extensible infrastructure that enables large-scale automatic execution of android apps to support different evolution and maintenance tasks (e.g., testing and energy optimization). In addition to the infrastructure we present a taxonomy of issues, single solutions to the issues, and guidelines to enable large execution of android apps. Finally, we devised novel approaches aimed at supporting testing and energy optimization of mobile apps (two key challenges in evolution and maintenance of android apps). First, we propose a novel hybrid approach for automatic GUI-based testing of apps that is able to generate (un)natural test sequences by mining real applications usages and learning statistical models that represent the GUI interactions. In addition, we propose a multi-objective approach for optimizing the energy consumption of GUIs in android apps that is able to generate visually appealing color compositions, while reducing the energy consumption and keeping a design concept close to the original.



© The Author