Model-based Testing

Functional criteria:

• Specific domain or application; requirements

• Methodological support

Structural criteria:

• Independent of domain

• data flow, control flow, data

Stochastic criteria:

• Uniform distributions: “purely random”

• User proile

• In general: not worse than structural criteria

Defect-based criteria

encode test cases for the SUT

input (sequence) and expected output (sequence)

necesarrily more abstract than the SUT

-> model as precise as SUT — directily validate SUT

-> model-based training only efficient if model is more abstract than SUT

models encapsulates details

One model for both testing and development

• no reduncancy -> no verification

• test models different from development models (normally)

Two seperate models

• too expensive

• different level of abstractions

• both tests + code profit from advantages of model-based development

Model for TC Generation

• expensive

• changing requirements -> interleaving model and code development

• model fairly complex to generate code from it

• specification doesnt profit from benefits of model-based development

Model extraction from code

• automatic generation -> redunancy questionable

• ex-post development of test cases only

Techniques:

• Dedicated algorithms for dedicated criteria

• (bounded) model checking

• symbolic execution

• heuristic search