Support for correct evolution of software product lines

今天的软件通常是作为一组相关的产品开发的，这些产品来自一个公共的基础设施，即所谓的软件产品线。产品线工程中的一个主要问题是产品线的不断发展，因为所有产品都是紧密相连的，产品线的总寿命比任何衍生产品都长。在这个项目中，我们将研究长期存在的软件产品线及其持续发展，特别关注嵌入式系统，特别是工业自动化系统。在这些领域中，可变性通常通过预处理器指令静态实现，或者通过在初始化时或稍后在运行时设置配置变量动态实现。尽管它们的相关性，这些变异性技术仍然是不够的研究。我们将开发技术来检查产品线实现的完整性，无论它在演变过程中何时发生变化。也就是说，与大多数其他分析产品线的工作相反，重点不是分析产品线的一个特定版本，而是修改产品线时引入的差异。该项目将结合联合收割机逆向工程、程序分析和正式的产品线分析，通过渐进式更改来发现缺陷的引入。这些缺陷通常来自于与功能相关的进化活动的意外副作用。该项目将采取全面的方法，考虑到与可变性模型相关的维护行动以及实施变更，并考虑到两者的组合。

Today software is often developed as a set of related products that are derived from a common infrastructure, a so-called software product line. A major issue in product line engineering is the continuous evolution of the product line as all products are intimately connected and the total lifetime of a product line is longer than that of any of its derived products. In this project, we will study long-living software product lines and their continuous evolution with a specific focus on embedded systems and in particular industry automation systems. In these domains, variability is often implemented statically through preprocessor directives or dynamically through setting configuration variables at initialization time or later at runtime. Despite their relevance, these variability techniques are still only insufficiently researched. We will develop techniques to check the integrity of a product line implementationwhenever it is changed during its evolution. That is, as opposed to most other work on the analysis of product lines, the focus will not be on the analysis of one particular version of a product line, but on the difference introduced when modifying a product line. The project will combine reverse engineering, program analysis, and formal product line analysis to discover the introduction of flaws through evolutionary changes. These flaws often arise from unintended side effects of evolution activities related to a feature. The project will take a comprehensive approach, taking into account maintenance actions related to the variability model as well as implementation changes and also from combinations of both.