Using Tasks to Automate Regression Testing of GUIs

摘要

Graphical User Interfaces (GUIs) present unique challenges for software testing. In this paper, we demonstrate that a test suite originally used to test a GUI contains a large number of unusable test cases for the modified GUI. We present a novel technique to recreate unusable test cases by associating meta-information (called a task) with each test case. Tasks represent activities that can be performed by using the software. The sequence of events in a test case represents the actions needed to complete its associated task. Even when changes to the GUI make test cases unusable, many tasks remain valid across successive GUI versions. We experimentally show that our technique is able to automatically and efficiently regenerate a large number of test cases.