Page 106 - 《软件学报》2025年第5期

软件学报 ISSN 1000-9825, CODEN RUXUEW E-mail: jos@iscas.ac.cn
2025,36(5):2006−2025 [doi: 10.13328/j.cnki.jos.007203] [CSTR: 32375.14.jos.007203] http://www.jos.org.cn
©中国科学院软件研究所版权所有. Tel: +86-10-62562563

*
基于事件标记的多粒度结合安卓测试序列约减

郝 蕊, 冯 洋, 李玉莹, 陈振宇

(计算机软件新技术国家重点实验室 (南京大学), 江苏 南京 210093)
通信作者: 冯洋, E-mail: fengyang@nju.edu.cn

摘 要: 针对安卓自动化测试工具生成的崩溃测试序列包含过多冗余事件, 造成测试回放、缺陷理解与修复困难
的现状, 很多测试序列约减工作被提出. 但目前工作仅关注应用界面状态变化而忽略了程序执行过程中内部状态
变化, 此外, 目前工作仅在单一抽象粒度上对应用状态进行建模, 例如控件粒度或活动粒度, 导致约减后测试序列
过长或约减效率低下. 针对以上问题, 提出基于事件标记的多粒度结合的安卓测试序列约减方法, 结合安卓生命周
期管理机制、程序静态数据流分析等对触发程序崩溃的关键事件进行标记, 缩小序列约减空间, 并设计了低粒度
粗筛选、高粒度细约减的策略. 最后, 收集包含程序间交互、用户输入等复杂场景的崩溃测试序列集, 在此数据集
上与其他测试序列约减工作的对比评估结果也验证了所提方法的有效性.
关键词: 安卓测试; 崩溃回放; 测试序列约减
中图法分类号: TP311

中文引用格式: 郝蕊, 冯洋, 李玉莹, 陈振宇. 基于事件标记的多粒度结合安卓测试序列约减. 软件学报, 2025, 36(5): 2006–2025.
http://www.jos.org.cn/1000-9825/7203.htm
英文引用格式: Hao R, Feng Y, Li YY, Chen ZY. Multi-granularity Fusion Android Test Sequence Reduction Based on Event Labeling.
Ruan Jian Xue Bao/Journal of Software, 2025, 36(5): 2006–2025 (in Chinese). http://www.jos.org.cn/1000-9825/7203.htm

Multi-granularity Fusion Android Test Sequence Reduction Based on Event Labeling
HAO Rui, FENG Yang, LI Yu-Ying, CHEN Zhen-Yu
(State Key Laboratory for Novel Software Technology (Nanjing University), Nanjing 210093, China)
Abstract: As too many redundant events included in crash test sequences generated by Android automated test tools may result in test
replay, defect comprehension, and repairing difficulty, a great number of test sequence reduction works have been proposed. While current
works only focus on the application interface changes and ignore the internal state changes during program execution. Moreover, current
works only model application states at a single and abstract granularity, such as control widget granularity or activity granularity, resulting
in long test sequences after reduction or inefficient reduction. This study proposes an Android test sequence reduction method combined
with multi-granularity based on event labeling. By taking into account the Android lifecycle management mechanism and data flow
analysis to label critical events that trigger crashes, this method can narrow the sequence reduction space and design a strategy of rough
selection under low granularity and detailed reduction under high granularity. At last, a crash test sequence set containing complex
scenarios such as inter-application interaction and user input is collected, and the comparison with other test sequence reduction works on
this set verifies the effectiveness of the method proposed in this study.
Key words: Android test; crash replay; test sequence reduction

截至 2022 年第 3 季度, Google Play Store 拥有超过 350 万个安卓应用 [1] , 下载次数超过 270 亿次 [2] , 占据了市
场份额的 80%, 与此同时, 复杂的交互环境、网络状态、多种系统兼容要求等也使得安卓应用相比传统的软件更
易发生崩溃问题, 严重影响用户体验, 这也是导致应用程序用户评分低的最重要的问题之一 [3−5] . 为了保障安卓应

* 部分工作为郝蕊在武汉理工大学做博士后期间完成
收稿时间: 2023-10-01; 修改时间: 2024-01-08; 采用时间: 2024-04-08; jos 在线出版时间: 2024-06-20
CNKI 网络首发时间: 2024-06-21