Coverage based debugging visualization

• Ano de defesa: 2014
• Autor(a) principal: Mutti, Danilo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Cobertura de código; Localização de defeitos; Visualização de informação de depuração; Visualização de software
• Link de acesso: http://www.teses.usp.br/teses/disponiveis/100/100131/tde-15122014-230109/

Fault localization is a costly task in the debugging process. Usually, developers analyze failing test cases to search for faults in the programs code. Visualization techniques have been proposed to help developers grasp the source code and focus their attention onto locations more likely to contain bugs. In general, these techniques utilize two-dimensional visualization approaches. We introduce a three-dimentional visual metaphor, called CodeForest, which represents a software as a cacti forest. In the CodeForest, nodes (sets of statements executed in sequence) are thorns, methods are branches, and classes are cacti. Heuristicsbased on the frequency that lines of codes are executed in successful and failing test cases are used to assign suspiciousness values to the elements (thorns, branches, and cacti) of the forest. The new metaphor was implemented as a plug-in targeted to the Eclipse Platform. This plug-in includes the mapping of suspiciousness values to elements of a virtual forest, a parameterized trimmer, which filters elements based on their score or text, and a list of most suspicious methods (also known as \"roadmap\"), to guide the developer on his/her debugging session. An exploratory experiment was conducted; the results indicates that the tool supports developers with and without experience. Users with low or no experience utilized the roadmap and the virtual 3D environment to investigate the defect. More experienced users prefer to use the roadmap as a guide to narrow which parts of the source code should be explored.