CRII: SHF: RUI: Leroid: Bug Oracle and Environment Configuration Automation for Android Bug Report Reproduction

Developers of mobile apps rely heavily on bug reports in issue-tracking systems to reproduce failures. However, the process of failure reproduction is often manually done by developers, making the resolution of bugs inefficient and costly, especially since bug reports are often written in natural language. For example, bug reports may miss necessary steps-to-reproduce information or include unclear discussions.

However, existing approaches are far from solving all problems in bug report reproduction. The first open problem is the bug oracle generation problem. Current approaches can only support reproducing crash bug reports while other types of bugs still need automatic-generated oracles to determine whether the bug has been reproduced or not.

Another open problem is that reproducing environment configurations (e.g., Android SDK version) still requires human effort to manually collect information from bug reports and configure them. This project will develop a novel approach that can support reproducing more bug types of Android bug reports by automatically generating oracles and configuring environments.

This research allows software developers to more quickly reproduce bug reports and to fix bugs, which will increase software quality. This, in turn, will benefit our modern society, which greatly relies on software. Additionally, this project contributes toward a diverse workforce, through integration with course curricula, training of students from underrepresented minority groups, and organizing summer workshops aimed at increasing K-12 girls’ interest in STEM-related careers.

To substantially improve automated bug report reproduction by supporting oracle generation and environment configuration, novel techniques and tools will be developed that address three important challenges: (1) extract observation/expectation and environment information sentences in a bug report, (2) generate oracle assertion codes that can determine whether the targeted bug has been triggered, (3) configure reproducing environment. In this project, techniques and tools based on a combination of techniques include static/dynamic program analyses, deep learning, natural language processing, and web crawling.

The approach will be embedded in a user interface and evaluated extensively on public bug reports and through industrial collaborations.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.