Submitted by Automa??o e Estat?stica (sst@bczm.ufrn.br) on 2017-10-18T19:35:37Z No. of bitstreams: 1 DemostenesSantosDeSena_TESE.pdf: 4593790 bytes, checksum: 3e0845d816f16e3a8f7659744e28f8ad (MD5) / Approved for entry into archive by Arlan Eloi Leite Silva (eloihistoriador@yahoo.com.br) on 2017-10-19T19:11:29Z (GMT) No. of bitstreams: 1 DemostenesSantosDeSena_TESE.pdf: 4593790 bytes, checksum: 3e0845d816f16e3a8f7659744e28f8ad (MD5) / Made available in DSpace on 2017-10-19T19:11:30Z (GMT). No. of bitstreams: 1 DemostenesSantosDeSena_TESE.pdf: 4593790 bytes, checksum: 3e0845d816f16e3a8f7659744e28f8ad (MD5) Previous issue date: 2017-02-13 / Os mecanismos de tratamento de exce??es s?o recursos fornecidos pelas principais linguagens de programa??o para auxiliar no desenvolvimento de sistemas robustos. A pol?tica de tratamento de exce??es corresponde ao conjunto de regras de design do tratamento excepcional e definem os elementos de c?digo (m?todos, classes ou pacotes) respons?veis pela sinaliza??o, propaga??o, captura das exce??es e as respectivas a??es de tratamento. Alguns estudos emp?ricos demonstraram que o tratamento inadequado de exce??es, consequ?ncia da falta da pol?tica documentada, ? uma poss?vel fonte de defeitos. Por outro lado, devido ? natureza impl?cita dos fluxos de exce??es, a identifica??o e corre??o dos tratamentos de exce??es tornam-se tarefas complexas. Para amenizar os problemas decorrentes do tratamento inadequado devido ? falta de documenta??o do tratamento de exce??es, algumas abordagens definiram linguagens de especifica??o das regras de tratamento com suporte ferramental para auxiliar na defini??o e checagem das regras. Entretanto, historicamente, as pol?ticas de tratamento de exce??es dos sistemas s?o postergadas ou ignoradas no processo de desenvolvimento. Adicionalmente, nenhuma das abordagens propostas oferece suporte ? defini??o das regras, de forma a auxiliar o arquiteto a extrair as regras a partir da an?lise de c?digo fonte pr?-existente, e este ? o objetivo da abordagem apresentada neste trabalho. Para apoiar a execu??o da abordagem proposta, foi desenvolvida uma ferramenta de an?lise est?tica que permite: (i) a coleta dos fluxos excepcionas e das respectivas a??es de tratamentos; (ii) a identifica??o e defini??o dos agrupamentos, que s?o os elementos de c?digo que possuem os mesmos comportamentos em rela??o ao tratamento de exce??es; (iii) a extra??o das regras; e, (iv) a checagem das regras e identifica??o das causas das viola??es ? pol?tica. A abordagem ? demonstrada em dois estudos emp?ricos. No primeiro estudo emp?rico foram analisadas 656 bibliotecas (libs) Java do reposit?rio central Maven com objetivo de extrair e caracterizar a pol?tica de tratamento de exce??es destas libs. Este estudo revelou que 80,9% das bibliotecas possuem fluxos excepcionais que implementam pelo menos um anti-pattern do tratamento excepcional. O segundo estudo emp?rico teve como objetivo investigar os benef?cios da extra??o das regras excepcionais a partir do c?digo pr?-existente no processo de defini??o e checagem da pol?tica de tratamento de exce??es. Dois sistemas de informa??o Web (i.e., IProject e SIGAA) foram utilizados neste segundo estudo. Neste estudo pudemos observar que todas as regras reportadas pelos arquitetos foram extra?das pelo suporte ferramental, e que os resultados do processo de extra??o permitiram que novas regras fossem adicionadas pelos arquitetos. Essas regras adicionadas foram as regras n?o definidas pelos arquitetos e corresponderam ? 57,1% (IProject) e 52,8% (SIGAA/Gradua??o) das regras da pol?tica dos sistemas analisados. O processo de checagem das regras definidas com o apoio da abordagem mostrou que 35,6% e 45,7% dos fluxos excepcionais do IProject e SIGAA/Gradua??o, respectivamente, violavam alguma das regras de tratamento de exce??es. / The Exception handling (EH) mechanism is a technique embedded in most of the mainstream programming languages to support the development of robust systems. The exception handling policy is composed by the set of exception handling design rules and which specify the elements (methods, classes and packages) or that contains the elements responsible for raising, propagating and catching of exceptions as well as the handling actions. Empirical studies have demonstrated that an inappropriate exception handling as consequence of undocumented exception handling policy is a source of bug hazards. On the other hand, due to the implicit nature of exception flows, the identification of exception handling code is a complex task. To address the problems resulting from the not-understood or inadequate exception handling, some approaches have been proposed languages to specify exception handling rules as well as a set of support tool to verify the constraints and checking the rules. However, historically, the exception handling policies are postponed or ignored in the software process development. Additionally, none of the proposed approaches provide support to the phase of exception policy definition. This work proposes an approach that helps the architect to extract the EH rules by performing an analysis on the existing code. Doing so, this approach fills the previous gap the EH policy definition. To support the proposed approach, a static tool suite was developed, which performs: (i) the discovery of exception flows and its handling actions; (ii) the definition of compartments; (iii) the semi-automatic rule extraction process; and (iv) the rule checking and identification of rule violation causes. This approach was assessed in two empirical studies. In the first study, 656 libraries from Maven central repository were analyzed. The main goal of this study was to reveal and to characterize the exception handling policy of the analyzed libraries. This study revealed that 80.9% of the analyzed libraries have exception flows that implement at least one exception handling anti-pattern. In the second study, we investigated the benefits of rule extraction process in the definition and checking of exception handling rules. Two web information systems (i.e., IProject and SIGAA) were analyzed in this second study. We found that all set of rules reported by the architects were extracted by our tool and the result of extraction process allowed that new rules were added to the policy. These added rules were not defined by architects and they corresponded to 57.1% (IProject) and 52.8% (SIGAA/Gradua??o) of the rules of analyzed systems. The checking process of defined rules supported by our approach verified that 35.6% (IProject) and 45.7% (SIGAA/Gradua??o) of exception flows violated some defined rule.