Presented at DVCon Europe 2021
Session T4.1
We come from a history where a design is being developed and specified and then thrown over the fence to verify it. Verification was seen as a completely independent team. The standard method is random biased, the only metric is coverage, and it results in the wellknown design/verification gap.
New features or engineering change orders challenge this process since they may break the originally intended behavior. In the worst case, this new issue is unveiled by a full regression run. We think, it is time to change this game!
You know the feeling and the pain: You want to bring a new and essential feature online. When is it done? When do you reach full coverage? When do you exercise the new feature with all the other features? If you are lucky, the coverage is sufficient at the first run and you are done. In more usual case, you are frustrated at the end of a sprint that the new feature is still not in full regression. During the following process, you might uncover holes in the specification, window conditions or complex interactions with other features causing unexpected side effects.
Presenters:
Bodo Hoppe, Georg Gläser, Tudor Timisecu, Matthew Balance, Philipp Wagner, Holger Horbach
https://dvcon.org
https://dvcon-proceedings.org