Recovering dynamic models from run time information is an important aspect of software system maintenance and reverse engineering.; This thesis presents an approach to detect and decipher dynamic information of C++ programs and recover the corresponding scenario diagrams. The approach is composed of two parts: code instrumentation and model recovery. The methodology of code instrumentation, which is platform and compiler independent, is provided to collect dynamic information for sequential, non-distributed C++ program under the form of a trace file. A generic methodology of model recovery is introduced to analyze a trace file. The transformation of a trace file into a scenario diagram is formally defined using OCL expressions and UML class diagrams (metamodels to structure the information in trace files and scenario diagrams).; Prototype tools, for both code instrumentation and model recovery, are developed and used to illustrate the approach on a realistic case study. (Abstract shortened by UMI.)... |