Reparação automática de programas: Exploração do potencial em contexto industrial

• Autor(a) principal: Queirós, André Emanuel Gomes
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.22/19237
• topic: Automatic Program Repair (APR); Continuous Integration (CI); Continuous Delivery (CD); APR tools; Real world

Today the automatic repair of programs is an issue increasingly discussed. The discussion of it comes from the evolution and advancement of technology, namely the environments of continuous integration and continuous delivery (CI/CD). These environments imply that the delivery of functional software is increasingly accelerated, and sometimes defects delay this process, implying the traditional debugging by a programmer, with the aim of detecting the defect and repairing it with quality. Automatic programming has been showing an increasing evolution, namely through techniques of automatic repair of APR programs. These techniques focus on the identification, localization, generation of correction and verification of the repaired version of the defective program, avoiding the programmer's effort in the identification and debugging process. This work aims at experimenting with two automatic repair tools, Arja and jGenProg, implanted in a pipeline, in an industrial context. The tools evaluation was done through the process of comparing them between themselves and with manual repairing process, and with the application of two questionnaires. In conclusion, both tools were able to be implanted in a pipeline, despite the necessary time required. The jGenProg tool revels a better performance, because it could detect a suspicious case and presented an even more completed report. However, the inquired developers choose the manual repairing process, against any of the evaluated tools.