在安卓系統下黑箱測試之來源碼層次涵蓋率技術

Abstract

在新的行動資訊世代,行動裝置應用程式已經逐漸成為軟體界的主流,而 Android 則是當中市佔率最高的平台,同時關於 Android 行動裝置的軟體測試案例 產生方法的論文或是工具的數量一直在快速增加,因此如何評估這些 test case 的品 質顯得日益重要。Android 行動裝置的軟體測試案例主要分為白盒測試以及黑盒測 試,目前在黑盒測試方法下主要採取是否找到錯誤來當作判斷測試案例的準則,然 而我們需要另一種準則當沒有找到錯誤或是找到錯誤的數量相近時,此時程式碼 的涵蓋率是一種很直觀的判斷準則,但在黑盒測試下程式碼的涵蓋率資訊是很難 取得的。因此本篇論文提出了一個技術夠在黑盒的情況下得到來源碼層次的資訊, 可以用此技術以該應用程式的行涵蓋率去評估測試案例的品質。另外也依據此架 構提出了一個演算法能在測試案例產生當中動態地選擇要按壓的按鈕,以使用更 少的步數但獲得相同行涵蓋率。

In the mobile generation, mobile applications have become the first choice in the software industry. Many large enterprises have claimed their future strategies are mobile first. Android has the highest market share in mobile platform. Recently, many researches about test case generation algorithm on Android have been done. Therefore, a criterion used to evaluate the quality of generated test cases becomes more and more important. There are two types of testing techniques, one is black-box and the other one is white-box. In black-box, the main criterion depends on the number of bugs found. However, we need a secondary criterion to tell the confidence level of a set of test cases when there are no bugs found or the number of bugs found by each test case are close. Code coverage is an intuitive criterion to meet the requirement. But in the case of black- box testing, it is hard to get code coverage information. This paper introduces a framework that can get source code level coverage information in black-box and then use this information to calculate line coverage in order to evaluate the quality of test cases. Furthermore, with the framework, we propose a test case generation algorithm that can dynamically choose a button to click in order to reach the same coverage rate with less steps.