A crescente complexidade das aplicações, a contínua evolução tecnológica e o uso cada vez mais disseminado de redes de computadores têm impulsionado os estudos referentes ao desenvolvimento de sistemas distribuídos. Como estes sistemas não podem ser facilmente desenvolvidos com tecnologias de software tradicionais por causa dos limites destas em lidar com aspectos relacionados, por exemplo, à distribuição e interoperabilidade, a tecnologia baseada em agentes parece ser uma resposta promissora para facilitar o desenvolvimento desses sistemas, pois ela foi planejada para suportar estes aspectos, dentre outros. Portanto, é necessário também que a arquitetura dos ambientes de desenvolvimento de software (ADS) evolua para suportar novas metodologias de desenvolvimento que ofereçam o suporte necessário à construção de softwares complexos, podendo também estar integrada a outras tecnologias como a de agentes. Baseada nesse contexto, essa dissertação tem por objetivo apresentar a especificação de uma arquitetura de um ADS distribuído baseada em agentes (DiSEN – Distributed Software Engineering Environment). Esse ambiente deverá fornecer suporte ao desenvolvimento de software distribuído, podendo estar em locais geograficamente distintos e também os desenvolvedores envolvidos poderão estar trabalhando de forma cooperativa. Na arquitetura proposta podem ser identificadas as seguintes camadas: dinâmica, que será responsável pelo gerenciamento da (re)configuração do ambiente em tempo de execução; aplicação, que terá, entre os elementos constituintes, a MDSODI (Metodologia para Desenvolvimento de Software Distribuído), que leva em consideração algumas características identificadas em sistemas distribuídos, já nas fases iniciais do projeto e o repositório para armazenamento dos dados necessários ao ambiente; e, infra-estrutura, que proverá suporte às tarefas de nomeação, persistência e concorrência e incorporará o canal de comunicação. Para validar o ambiente será realizada uma simulação da comunicação que pode ser necessária entre as partes constituintes do DiSEN, por meio da elaboração de diagramas de use case e de seqüência, conforme a notação MDSODI. Assim, as principais contribuições desse trabalho são: (i) especificação da arquitetura de um ADS distribuído que poderá estar distribuído geograficamente; incorporará a MDSODI; proporcionará desenvolvimento distribuído; possuirá atividades executadas por agentes; (ii) os agentes identificados para o DiSEN deverão ser desenvolvidos obedecendo ao padrão FIPA (Foundation for Intelligent Physical Agents); (iii) a identificação de um elemento que irá oferecer apoio ao trabalho cooperativo, permitindo a integração de profissionais, agentes e artefatos.

Técnicas que partem de modelos de alto nível de abstração são requeridas para lidar com a complexidade encontrada nas novas gerações de sistemas embarcados, sendo cruciais para o sucesso do projeto. Uma grande redução do esforço pode ser obtida com o uso de modelos quando código em uma linguagem de programação pode ser gerado automaticamente a partir desses. Porém, ferramentas disponíveis para modelagem e geração de código normalmente são dependentes de domínio e o software embarcado normalmente possui comportamento heterogêneo, requerendo suporte a múltiplos modelos de computação. Nesta tese, estratégias para desenvolvimento de software embarcado baseado em modelos de alto nível usando UML e Simulink são analisadas. A partir desta análise, observaram-se as principais limitações das abordagens para geração de código baseadas em UML e Simulink. Esta tese, então, propõe estratégias para melhorar a automação provida por estas ferramentas, como por exemplo, propondo uma abordagem para geração de código multithread a partir de modelos Simulink. A comparação feita entre UML e Simulink mostra que, embora UML seja a linguagem mais usada no domínio de engenharia de software, UML é baseada em eventos e não é adequada para modelar sistemas dataflow. Por outro lado, Simulink é largamente usado por engenheiros de hardware e de controle, além de suportar dataflow e geração de código. Porém, Simulink provê abstrações de mais baixo nível, quando comparado a UML. Conclui-se que tanto UML como Simulink possuem prós e contras, o que motiva a integração de ambas linguagens em um único fluxo de projeto. Neste contexto, esta tese propõe também uma abordagem integradora para desenvolvimento de software embarcado que inicia com uma especificação de alto nível descrita usando diagramas UML, a partir da qual modelos dataflow e control-flow podem ser gerados. Desta maneira, o modelo UML pode ser usado como front-end para diferentes abordagens de geração de código, incluindo UML e a proposta geração de código multithread a partir de modelos Simulink. / The use of techniques starting from higher abstraction levels is required to cope with the complexity that is found in the new generations of embedded systems, being crucial to the design success. A large reduction of design effort when using models in the development can be achieved when there is a possibility to automatically generate code from them. Using these techniques, the designer specifies the system model using some abstraction and code in a programming language is generated from that. However, available tools for modeling and code generation are domain-specific and embedded software usually shows heterogeneous behavior, which pushes the need for supporting software automation under different models of computation. In this thesis, strategies for embedded software development based on high-level models using UML and Simulink were analyzed. We observed that the embedded software generation approaches based on UML and Simulink have limitations, and hence this thesis proposes strategies to improve the automation provided on those approaches, for example, proposing a Simulink-based multithread code generation. UML is a well used language in the software engineering domain, and we consider that it has several advantages. However, UML is event-based and not suitable to model dataflow systems. On the other side, Simulink is widely used by control and hardware engineers and supports dataflow, and time-continuous models. Moreover, tools are available to generate code from a Simulink model. However, Simulink models represent lower abstraction level compared to UML ones. This comparison shows that UML and Simulink have pros and cons, which motivates the integration of both languages in a single design process. As the main contribution, we propose in this thesis an integrated approach to embedded software design, which starts from a high-level specification using UML diagrams. Both dataflow and control-flow models can be generated from that. In this way, an UML model can be used as front-end for different code generation approaches, including UML-based one and the proposed Simulink-based multithread code generation.

Tese de mestrado integrado. Engenharia Informática e Computação. Faculdade de Engenharia. Universidade do Porto. 2009

Estágio realizado na EFACEC e orientado pelo Eng.º Miguel Ferreira Pereira Gomes / Tese de mestrado integrado. Engenharia Informática e Computação. Faculdade de Engenharia. Universidade do Porto. 2009

Software reuse via component libraries suffers from the twin problems of code location and comprehension. The Intelligent Code Object Planner (ICOP) is a cognitively motivated system that facilitates code reuse by answering queries about how to produce an effect with the library. It can plan for effects which are not primitive with respect to the library by building a plan that incorporates multiple components. The primary subsystems of ICOP are a knowledge base which describes the ontology of the library, a natural language interface which translates user queries into a formal effect language (predicates), a planner which accepts the effect and produces a plan utilizing the library components, and an explanation generator which accepts the plan and produces example code illustrating the plan. ICOP is currently implemented in Prolog and supports a subset of the Windows 3.0 APL

Random testing is a standard software testing method. It is a popular method for reli-ability assessment, but its use for debug testing has been opposed by some authorities. Random testing does not use any information to guide test case selection, and so, it is argued, testing is less likely to be effective than other methods. Based on the observation that failures often cluster in contiguous regions, Adaptive Random Testing (ART) is a more effective random testing method. While retaining random selection of test cases, selection is guided by the idea that tests should be widely spread throughout the input domain. A simple way to implement this concept, FSCS-ART, involves randomly generating a number of candidates, and choosing the candidate most widely spread from any already-executed test. This method has already shown to be up to 50% more effective than random testing. This thesis examines a number of theoretical and practical issues related to ART. Firstly, an theoretical examination of the scope of adaptive methods to improve testing effectiveness is conducted. Our results show that the maximum improvement in failure detection effectiveness possible is only 50% - so ART performs close to this limit on many occasions. Secondly, the statistical validity of the previous empirical results is examined. A mathematical analysis of the sampling distribution of the various failure-detection effectiveness methods shows that the measure preferred in previous studies has a slightly unusual distribution known as the geometric distribution, and that that it and other measures are likely to show high variance, requiring very large sample sizes for accurate comparisons. A potential limitation of current ART methods is the relatively high selection overhead. A number of methods to obtain lower overheads are proposed and evaluated, involving a less-strict randomness or wide-spreading criterion. Two methods use dynamic, as-needed partitioning to divide the input domain, spreading test cases throughout the partitions as required. Another involves using a class of numeric sequences called quasi-random sequences. Finally, a more efficient implementation of the existing FSCS-ART method is proposed using the mathematical structure known as the Voronoi diagram. Finally, the use of ART on programs whose input is non-numeric is examined. While existing techniques can be used to generate random non-numeric candidates, a criterion for 'wide spread' is required to perform ART effectively. It is proposed to use the notion of category-partition as such a criterion.

Thesis (Ph. D.)--West Virginia University, 2006. / Title from document title page. Document formatted into pages; contains xiii, 166 p. : ill. (some col.). Includes abstract. Includes bibliographical references (p. 130-140).

Relationships between classes and objects in object-oriented software are necessary in order for the parts of the systems to provide dynamic behavior. These inherent relationships also create dependencies which can give rise to problems for software evolution of object-oriented software systems. Dependencies in software make systems difficult to understand, reuse, change and verify. This thesis presents analytical and empirical investigations of dependency-related problems in software evolution of object-oriented software and on how such problems can be handled with dependency focused techniques, methods and processes. The research presented in this thesis includes: Development of a programming language construct for controlling dependencies; formal experiments on code inspection techniques; exploring change strategies' effects on test suites; an industrial case study of regression test selection techniques for object-oriented software; proving the efficiency and defect detection capabilities of a novel regression test selection technique. The thesis contributes to increased knowledge on the role of dependencies in software evolution of object-oriented software. Specific contributions are a programming language construct that can control access to dependencies in software. Other main contributions are insights on the efficiency of dependency focused code inspection techniques and contribution to the knowledge on dependency-based regression test selection techniques for large scale software. Another contribution is a novel change-based regression test selection technique.

Software designers make decisions between alternate approaches early in the development of a software application and these decisions can be difficult to change later. Designers make these decisions based on estimates of how alternatives affect software qualities. One software quality that can be difficult to predict is performance, that is, the efficient use of resources in the system. It is particularly challenging to estimate the performance of large, interconnected software systems composed of components. With the proliferation of class libraries, middle-ware systems, web services, and third party components, many software projects rely on third party services to meet their requirements. Often choosing between services involves considering both the functionality and performance of the services. To help software developers compare their designs and third-party services, I propose using performance prototypes of alternatives and test suites to estimate performance trade-offs early in the development cycle, a process called Design-Time Performance Testing (DTPT).<p> Providing software designers with performance evidence based on prototypes will allow designers to make informed decisions regarding performance trade-offs. To show how DTPT can help inform real design decisions. In particular: a process for DTPT, a framework implementation written in Java, and experiments to verify and validate the process and implementation. The implemented framework assists when designing, running, and documenting performance test suites, allowing designers to make accurate comparisons between alternate approaches. Performance metrics are captured by instrumenting and running prototypes.<p> This thesis describes the process and framework for gathering software performance estimates at design-time using prototypes and test suites.

Feature-oriented requirements decompose a system's requirements into individual bundles of functionality called features, where each feature's behaviour can be expressed as a state-machine model. However, state machines are difficult to write; determining how to decompose behaviour into states is not obvious, different stakeholders will have different opinions on how to structure the state machine, and the state machines can easily become too complex. This thesis proposes a pattern for decomposing and structuring the model of a feature's behavioural requirements, based on modes of operation (e.g., Active, Inactive, Failed) that are common to features in multiple domains. Interestingly, the highest-level modes of the pattern can serve as a generic behavioural interface for all features that adhere to the pattern. The thesis proposes also several pattern extensions that provide guidance on how to structure the Active and Inactive behaviour of the feature. The pattern was applied to model the behavioural requirements of 21 automotive features that were specified in 7 production-grade requirements documents. The pattern was applicable to all 21 features, and the proposed generic feature interface was applicable to 50 out of 58 inter-feature references. A user study with 18 participants evaluated whether use of the pattern made it easier than otherwise to write state machines for features and whether feature state machines written with the help of the pattern are more readable than those written without the help of the pattern. The results of the study indicate that use of the pattern facilitates writing of feature state machines.