Performance Testing On Android Application Based On GUI Testing

The rapid development of mobile devices has enabled mobile application software to have a large user base.The mobile operating system market is dominated by Android and iOS.Due to Android's open source,free and convenient,Android system has gradually occupied most of the market share in recent years.According to the latest statistics,the Android system already has 87.7% of the market,becoming the most popular mobile device.Android application software(App)has the characteristics of rapid development,continuous iteration,and agile development.The Android application is based on event-driven,frequently using the callback function.So the traditional testing techniques are difficult to apply to such applications.The features of these Android applications make it difficult for developers to guarantee the quality of the program.As a result,problems such as crash,and abnormal power consumption are often encountered in the use of the App.Current work focuses mostly on crash detection and performance issues.In addition,although the existing performance test work can find problems such as CPU problems,memory leak,and unreleased resources,it usually relies on manually test the app.Therefore,in order to solve the above problems and challenges,we propose a GUI-based Android automation performance testing framework,PerDroid.PerDroid can effectively combine automated testing and performance testing technique,focusing on the latency issues appearing during runtime.We propose a feedback-based random traversal method based on the similarity of pages,so as to automate the traversal of App to find the root cause of the latency issues.PerDroid will automatically discover new states of GUI pages during the execution.In the process of automated exploration,PerDroid records the number of each GUI page and the execution time of some callback methods.PerDroid analyzes the log,collecting during the execution,and tries to find the root cause of the performance issues.PerDroid can automatically execute test on the App,collect run-time data and find the root cause of performance issues.This tool can not only improve the reliability and effectiveness of the App's testing,but also reduce the cost of performance testing.We experimented our approach at randomly-selected open source projects.Experiments result shows that PerDroid is able to explore more execution path,compared to the state of art tools.Not only explorer to the point where the app experiencing with latency issues,but also report detected information regarding its root cause.