2011-11-21 Medverkande: Richard T Robert F Thomas T Kristina L Lars-Ola D Per R * Nya doktorander (mdh): Joe, Kristian, Kivanc Research direction - Jiale “Joe” Zhou (MDH): Complex software systems are still developed using a cascading processes of artifact generation and software refinement that occurs over time until a set of acceptance criteria have been met. The artifacts and the code together define the entire system at different levels of abstraction, resulting in a complex and often unmanageable environment wherein the dependencies are not clear, and V&V activities are extremely labor intensive and error prone. The aim of the research is to develop a holistic framework supporting various specification, design and implementation languages used in the architecture design phase all the way to the implementation phase, where each corresponding artifact is automatically verified against its specified constraints. As a result, the framework can be used to automatically check the consistency and completeness among artifacts generated from different phases in the development process. The research goal faces two overarching challenges, 1) automated verification demands formal traceability among the various artifacts, and 2) consistency and completeness checking among diverse artifacts demands a versatile language to implement the verification techniques. We propose traceability through a common formal language – Timed Abstract State Machines (TASM) – to formally specify the different languages’ semantics, and thereby having a common underpinning making traceability possible. - Joe antagen till SWELL 21/11. Kristian Wiklund 1 Industry Relevance Successful test automation is one of the primary enablers for efficiency in a modern software design organization. Apart from reductions in time-to-market by decreasing the time spent waiting for feedback, several other benefits can be reaped, for example, the amount of people involved in manual testing can be decreased and used for other activities, the general confidence in the product is increased by allowing for a higher test coverage in the same or lower lead time. Also, testing can be performed hands-off during otherwise idle times, such as breaks, week-ends, and nights, which contributes greatly to the over-all efficiency. However, test automation is traditionally performed ad hoc, typically begin- ning with a small set of test cases are executed together, a report generator is added, then everything is wrapped in an off-hours execution system, and so on. Clearly, the results from an activity like that will vary, from stable and reliable, to only providing a false sense of security to the user and actually not performing any reliable testing at all. Typically, one is interested in several types of deliverables from the test activity. The primary deliverable is usually feed-back to the design organization that will get the opportunity to correct problems and defects before they reach the customer. Speed and reliability in the test verdict is essential to handle corrections in a good way. Another deliverable, much less tangible and not as often discussed, is the confidence in the testing, that allows the product owner to take an informed release decision after which the product will distributed to the customers, where any defects will be very costly to handle. Both these deliverables are influenced by the quality of the test automation, such as how the automated test cases are designed, and how the test harness itself is designed. 2 Research Area We believe that there are common properties to the majority of good software test automation efforts, as well as common properties to the majority of bad software test automation efforts. To guide practitioners in designing and im- plementing good test automation, these properties need to be identified and ultimately used to construct a design- and improvement model that can be used not only to achieve as good results as possible, but also to evaluate existing automation implementations. - Kristian (industridoktorand, ///) antagen till SWELL 21/11 Kivanc Doganay Search-based software testing is based on translating testing tasks into optimization problems, and using meta-heuristic search algorithms to achieve these tasks. There are quite many studies showing that the search-based methods in testing are useful (e.g., superior to random testing), but most of these studies are focused on small to medium sized open-source software. This leads to doubts on scalability and applicability of search-based testing to industrial software. Moreover, we are interested in applying these methods on embedded software, which raises further applicability concerns. In this project we will focus on applying search-based testing methods in two different complex embedded systems: Ericsson’s telecoms platform, and Bombardier’s rail vehicle control system. Our initial aim is to use search-based methods for test data generation for structural testing. We will also look into subsystems where functional, temporal or other non-functional testing is relevant – and use search-based testing for those purposes. Component level, integration, and system level testing will be investigated. Subsystems with different characteristics, and different levels of testing serve as case studies, where we will apply search-based methods. The goal of the project is to identify and overcome the applicability problems of search-based testing at different testing levels, and for different types of testing purposes, in the context of the aforementioned complex industrial embedded software. Furthermore, we will investigate ways to combine other approaches with search-based testing, such as model-based methods, in order to create hybrid methods that overcome limitations of pure methods. - Kivanc (industridoktorand, /// & Bombardier) antagen till SWELL 21/11 * Next SWELL industry day? - 2012 okt i Gbg prel. * Information regarding the Automated V&V course. Skall slutföras senast Jan 2012. * Next course?? What about "Evidence-based decision support for V&V"? I gave a tutorial at ESEM in Banff about strategic and operational decisions, which may be turned into a PhD student course. http://esem.cpsc.ucalgary.ca/esem2011/iasese/Downloads/VV_decision_support_handouts.pdf /Per If we gear that towards org and mgmt issues etc. then I can see a match :) /Richard - SWELL (Lund) startar kursen under våren 2012. Slutar sommaren 2012. Därefter är det MdH:s tud förmodligen Nov 2012? * Info om IRSES http://ec.europa.eu/research/participants/portal/page/people?callIdentifier=FP7-PEOPLE-2012-IRSES Richard rekommenderar att vi inte jobbar mot denna ansökan. - Next meeting 23/1, 2012, 14.00 - Info: Kurser bättre anpassade mot näringslivet hade varit önskvärt. Samtidigt har vi samhandledning, industridag, och en mängd företag som redan vill jobba med SWELL - som vi inte har tid med i nuläget... - SWELL doktorander håller seminarium mot andra företag om så önskas. Richard ska försöka påminna ftg representanter att detta är möjligt vid varje möte.