[Paper Review] FUSION: A Tool for Facilitating and Augmenting Android Bug Reporting

Kevin Moran, Mario Linares‐Vásquez+2 more|arXiv (Cornell University)|Jan 18, 2018

Software Engineering Research9 references4 citations

TL;DR

FUSION is a tool that enhances Android bug reporting by automatically extracting GUI components and event flows via static and dynamic analysis, then guiding users through structured, auto-completed reproduction steps using contextual app data. It significantly improves bug report reproducibility—107 out of 120 Fusion reports were reproducible versus 97 out of 120 using traditional systems like GCIT.

ABSTRACT

As the popularity of mobile smart devices continues to climb the complexity of "apps" continues to increase, making the development and maintenance process challenging. Current bug tracking systems lack key features to effectively support construction of reports with actionable information that directly lead to a bug's resolution. In this demo we present the implementation of a novel bug reporting system, called Fusion, that facilitates users including reproduction steps in bug reports for mobile apps. Fusion links user-provided information to program artifacts extracted through static and dynamic analysis performed before testing or release. Results of preliminary studies demonstrate that Fusion both effectively facilitates reporting and allows for more reliable reproduction of bugs from reports compared to traditional issue tracking systems by presenting more detailed contextual app information. Tool website: www.fusion-android. com Video url: https://youtu.be/AND9h0ElxRg

Motivation & Objective

• Address the critical gap in traditional bug tracking systems, which rely on unstructured, incomplete user reports that are often non-reproducible.
• Reduce the lexical gap between non-developer reporters and developers by linking functional descriptions to low-level code and GUI artifacts.
• Improve the quality and actionability of bug reports by integrating automated program analysis with a guided, web-based reporting interface.
• Enable developers to reproduce bugs more reliably by providing structured, traceable, and context-rich reports with screenshots and component-level metadata.
• Facilitate not only bug reporting but also future support for feature requests through the same structured reporting pipeline.

Proposed method

• FUSION performs static and dynamic analysis on Android APKs to extract GUI components, event flows, and screen transitions before app release.
• It uses a depth-first search (DFS) heuristic in the dynamic analysis engine to explore possible user interaction paths through the app’s UI components.
• The auto-completion engine in the web-based report generator suggests valid UI actions based on prior steps and known app states, reducing user input errors.
• Reporters construct bug reports using guided combo boxes that suggest components by type, text, and screen location, with fallback for unsupported gestures.
• Each step in the report is linked to the source code class where the component is instantiated, enabling traceability to implementation.
• Final reports include a unique ID, step-by-step actions with screenshots, component metadata (type, location, class), and full-screen captures for visual traceability.

Experimental results

Research questions

• RQ1 Can automated static and dynamic analysis improve the completeness and structure of user-reported bug reproduction steps?
• RQ2 To what extent does Fusion reduce the lexical gap between non-developer reporters and developers compared to traditional issue trackers?
• RQ3 How does the reproducibility of bug reports generated with Fusion compare to those created using conventional systems like Google Code Issue Tracker?
• RQ4 Can the auto-completion mechanism in Fusion reduce user effort and error in reporting while maintaining high fidelity to actual app behavior?
• RQ5 What is the impact of traceability links to source code components on developer debugging efficiency?

Key findings

• Fusion reports achieved a 107 out of 120 success rate in reproducibility on physical devices, significantly outperforming traditional systems like GCIT, which achieved 97 out of 120 reproducibility.
• The guided, auto-completed input interface in Fusion reduced the cognitive load on reporters, enabling more consistent and detailed reproduction steps.
• The integration of static and dynamic analysis enabled accurate suggestion of UI components and transitions, even when reporters lacked technical knowledge.
• Reporters preferred Fusion over traditional systems due to its structured, context-aware input, which reduced ambiguity and increased confidence in report quality.
• The tool’s ability to link each step to the source code class and provide visual screenshots improved developer understanding and debugging speed.
• The system demonstrated robustness in handling edge cases through a manual input fallback mode for unsupported gestures or components.