针对能耗和并发缺陷的安卓应用测试用例生成研究

发布时间：2018-03-28 20:38

本文选题：安卓应用测试　切入点：测试用例生成　出处：《南京大学》2017年硕士论文

【摘要】：智能手机在人们的生活中成为了不可或缺的一部分。智能手机的计算能力变得越来越强,搭载的传感器越来越繁多。智能手机应用会采用并发编程技术以提高CPU的利用效率,利用传感器获取用户的物理环境的一些参数。但是,这里相应的出现了两个严重的问题——能耗和并发缺陷,从而降低了应用的质量,最终导致较差的用户体验。感知操作是一个十分消耗能源的操作,从中获取到的传感器数据应当被有效地利用。但是,已有的研究表明应用市场中的智能手机应用的传感器数据利用率通常较低,从而导致了严重的能耗浪费。为了解决这个问题,最新的技术GreenDroid能够自动地生成传感器数据,记录它们的传播以及分析它们在应用中的使用情况。然而GreenDroid的传感器数据生成方式是随机的,而这会使诊断结果不稳定和降低其有效性。众所周知,并发程序比较难以编写、测试和调试。安卓官方提供了一系列约束以避免出现并发缺陷,比如只能在Main Thread进行GUI更新。随着应用的复杂度不断增加,开发者越来越难以正确地理解应用的行为,从而导致出现难以察觉的并发缺陷。部分已有的工作通过利用已有的输入事件序列构建happens-before图,从而检测潜在的数据竞争。但是,它们严重地依赖于这些输入事件序列,从而造成较高的假阳性。针对能耗和并发缺陷,本文采用了测试用例生成技术,即利用应用的历史执行和当前状态生成新的测试用例。为了实现更加有效的能耗低效性诊断,本文提出了一个创新的方法以系统地生成多维的传感器数据,同时用更细粒度的方式定义了应用的状态。本文实现了一个原型工具CyanDroid,并且用了真实的应用以及它们的变异体对CyanDroid进行了评估。结果表明CyanDroid在诊断由传感器数据低利用率所导致的能耗缺陷时更加稳定、有效。另一方面,基于能够同时生成输入事件及其调度这样的事实,本文提出了一个触发安卓应用中并发缺陷的方法。本文实现了动静结合的混合分析以找到应用中潜在的冲突资源访问,然后通过导向型的事件和调度生成自动地运行被测应用。本文实现了相应的原型工具AATT,并用真实的应用对其进行了评估,结果表明了AATT的有效性和高效性。
[Abstract]:Smartphones have become an integral part of people's lives. The computing power of smartphones is getting stronger and more and more sensors are being carried. Smartphone applications will use concurrent programming technology to improve the efficiency of CPU. The sensor is used to obtain some parameters of the user's physical environment. However, there are two serious problems here-energy consumption and concurrency defects, which reduces the quality of the application. The end result is a poor user experience. Perception is a very energy-consuming operation from which sensor data should be used efficiently. Existing studies have shown that sensor data utilization in smartphone applications in the application market is often low, resulting in a serious waste of energy. To solve this problem, the latest technology, GreenDroid, can automatically generate sensor data. Record their propagation and analyze their use in applications. However, the way GreenDroid sensor data is generated is random, which makes diagnostics unstable and reduces its effectiveness. Concurrency programs are notoriously difficult to write. Testing and debugging. Android officials provide a series of constraints to avoid concurrency defects, such as GUI updates on Main Thread. As the complexity of applications increases, developers are finding it increasingly difficult to understand the behavior of apps correctly. Part of the existing work detects potential data competition by building happens-before diagrams using existing input event sequences. However, they rely heavily on these input event sequences. In order to achieve more efficient energy consumption and low efficiency diagnosis, this paper adopts test case generation technology, that is, using the historical execution and current status of the application to generate new test cases. This paper presents an innovative method to systematically generate multidimensional sensor data. At the same time, the state of the application is defined in a more granular way. This paper implements a prototype tool, Cyan Droid, and evaluates CyanDroid using real applications and their variants. The results show that CyanDroid is used in diagnosis by sensor data. The energy deficit caused by low utilization rate is more stable. Valid. On the other hand, based on the fact that input events and their scheduling can be generated at the same time, This paper proposes a method to trigger concurrency defects in Android applications. Then the test application is automatically run through the oriented event and scheduling generation. This paper implements the prototype tool AATT and evaluates it with real applications. The results show that AATT is effective and efficient.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN929.53;TP316;TP212

【相似文献】