Please use this identifier to cite or link to this item: http://repositorio.unicamp.br/jspui/handle/REPOSIP/93358
Type: Artigo de evento
Title: Regression Test Selection For Testable Classes
Author: Martins E.
Vieira V.G.
Abstract: A reusable class must be tested many times: each time modifications are applied to it or its base classes; when a subclass is created, in which case the inherited and redefined features must be retested in the new context. Therefore, a class should be easy to test, specifically for test execution and results analysis, since these activities must be repeated often. Inspired by R. Binder's self-testing class concept [4] we defined, in a previous work, a testable class as a 3-tuple: class implementation, class behavior model and built-in test (BIT) mechanisms. In this work we present how to use this information when a class is changed. Regression testing is necessary each time a software is changed, to assure that the modifications do not adversely affect the unchanged parts. It is assumed that the test suite applied when testing the old version is available for reuse. However, test suites can be large and require too much effort to be reapplied in their totality. In such cases, a subset of the tests must be selected. This selection usually requires extra information besides the source code. This work aims at answering the following question: how to use test information contained in a testable class to do regression testing? The answer involves, among other aspects, the definition of an approach to select tests for reuse. The approach can be fully automated and does not need the source code for regression-test selection. © Springer-Verlag Berlin Heidelberg 2005.
Editor: 
Rights: fechado
Identifier DOI: 
Address: http://www.scopus.com/inward/record.url?eid=2-s2.0-24944550097&partnerID=40&md5=e708a8c120960f8bd732b794e596c8f1
Date Issue: 2005
Appears in Collections:Unicamp - Artigos e Outros Documentos

Files in This Item:
File Description SizeFormat 
2-s2.0-24944550097.pdf153 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.