Submitted by Automa??o e Estat?stica (sst@bczm.ufrn.br) on 2017-11-01T21:17:48Z No. of bitstreams: 1 TeresaDoCarmoBarretoFernandes_DISSERT.pdf: 5261298 bytes, checksum: 1a7e32ec8483e6e7e31101df7f8675f9 (MD5) / Approved for entry into archive by Arlan Eloi Leite Silva (eloihistoriador@yahoo.com.br) on 2017-11-07T21:08:03Z (GMT) No. of bitstreams: 1 TeresaDoCarmoBarretoFernandes_DISSERT.pdf: 5261298 bytes, checksum: 1a7e32ec8483e6e7e31101df7f8675f9 (MD5) / Made available in DSpace on 2017-11-07T21:08:03Z (GMT). No. of bitstreams: 1 TeresaDoCarmoBarretoFernandes_DISSERT.pdf: 5261298 bytes, checksum: 1a7e32ec8483e6e7e31101df7f8675f9 (MD5) Previous issue date: 2017-06-30 / Exce??es n?o capturadas (do ingl?s: uncaught) n?o s?o cen?rios excepcionais nas aplica??es Java atuais. Eles s?o, na verdade, uma das principais causas de falha das aplica??es Java - que podem originar-se de erros de programa??o (e.g., acesso a refer?ncias nulas); falhas no hardware ou em APIs utilizadas. Essas exce??es uncaught resultam em stack traces que s?o frequentemente usados pelos desenvolvedores como fonte de informa??es para a depura??o. Atualmente, essa informa??o ? frequentemente usada pelos desenvolvedores em mecanismos de busca ou sites de perguntas e respostas (do ingl?s: Question and Answer - Q&A) para tentar compreender melhor a causa do crash e assim poder resolv?lo. Este estudo fez a minera??o de stack traces inclu?das nas perguntas e respostas do StackOverflow (SOF). O objetivo deste estudo foi: (i) identificar caracter?sticas das stack traces mineradas do SOF e (ii) investigar como tais informa??es podem ser usadas para evitar exce??es uncaught durante o desenvolvimento de software. Neste estudo, 121.253 stack traces foram extra?das e analisadas em combina??o com inspe??es de postagens do SOF. Tamb?m ? proposta a ferramenta ExMinerSOF, que alerta o desenvolvedor sobre as exce??es que podem ser potencialmente sinalizadas por um m?todo de API. Essas informa??es s?o descobertas aplicando uma estrat?gia de minera??o apresentada neste trabalho. Ao faz?-lo, a ferramenta permite que o desenvolvedor evite falhas com base em falhas relatadas por outros desenvolvedores. / Uncaught exceptions are not an exceptional sce- nario in current Java applications. They are actually one of the main causes of applications crashes, which can originate from programming errors on the application itself (null pointer dereferences); faults in underlying hardware or re-used APIs. Such uncaught exceptions result in exception stack traces that are often used by developers as a source of information for debugging. Currently, this information is ofttimes used by developers on search engines or Question and Answer sites while the developer tries to: better understand the cause of the crash and solve it. This study mined the exception stack traces embedded on StackOverflow (SOF) questions and answers. The goal of this work was to two-fold: to identify characteristics of stack traces mined from SOF and to investigate how such information can be used to prevent uncaught exceptions during software development. Overall 121.253 exception stack traces were extracted and analyzed in combination with Q&A inspections. Hence, this study proposes ExMinerSOF tool, which alerts the developer about the exceptions that can be potentially signaled by an API method but are not part of the API documentation - and was discovered by applying a mining strategy in SOF repository. Doing so, the tool enable the developer to prevent faults based on failures reported by the crowd.