統計測試流程應用於缺陷測試之研究

Abstract

無瑕法可靠度模型(Cleanroom Reliability Model)之統計測試流程為一系統化、嚴謹之測試方法，可客觀的呈現軟體品質狀態，其高品質的代價就是高成本；缺陷測試則是成本較低、經驗導向、具有彈性的之流程，但相對的缺點就是流程鬆散，缺乏系統化程序。以往缺陷測試之相關研究，並不著重在測試規劃，但規劃的結果卻是決定測試成功與否的最重要依據。本文修改統計測試流程，使其可適用於缺陷測試中，透過完整的測試規劃與流程改善，有限度的增加測試成本以使缺陷測試品質獲得提升。主要的修改包含：將統計測試之使用模式(以機率為基礎)改為缺陷測試之狀態模式(以圖論為基礎)、導入分割測試觀念、提出新的測試停止點、完全自動產生測試路徑(軟體工具支援)。本文主要貢獻有1)提出一改善的缺陷流程，2)完成支援統計測試與改善之缺陷測試流程之軟體工具，並實際運用此流程於”網路連線傳輸”案例之測試與分析，證實此流程之可行性。

Statistical testing of a Cleanroom Reliability Model is a systematical and conscientious method. It aims for the software quality control objectively. The Defect testing is a low cost, experience oriented, and flexibility process. Unlike, Statistical testing, it has no enforced systematical process. In this thesis study, we modify the statistical testing process to fit into the defect testing. Specifically, the modification included 1) The usage model of statistical testing is based on probability model and now is modified as the state model (based on graph theory) for defect testing does not require the use of Markov Chain model for the computation of Reliability, 2) The concept of partition testing is employed to generate the “nodes (states)” of the “graph” in the analysis , and 3) A new stopping criterion based on the search of the Euler graph is proposed to reduce the testing cases. To design and implement the testing tool for the modified defect testing, we define an auto test scripts based on MatLab environment. The proposed prototype system was implemented and used for the testing of a real application case “Network Link and Transfer” and the result confirms the applicability of the proposed approach.