Fig. 1. Example source code.

Fig. 2. Example object process graph for *s1.

Fig. 3. Normalized and instrumented code example; see Fig. 1 for the original source code of reverse. Begin-/end-of-lifetime and call site logging have been omitted for better readability.

Fig. 4. From trace to DOPG.

Fig. 5. Example trace excerpt for the call of routine reverse from Fig. 1 and intermediate object process graphs for object *s1.

Fig. 6. Raw graph construction; get_or_create_node(e) yields a graph node for event e which is based on e’s location identifier. The corresponding node is returned if it exists, or a new node for e is created otherwise; start is the start node of the graph (i.e., the root node).

Fig. 7. Example for irrelevant control constructs: Nodes 2 and 3 can be removed. Control dependencies are shown by dashed arrows.

Fig. 8. Final dynamic object process graph for the example. Only when *s2 contains at least one element, the result is identical to the statically extracted object process graph. Otherwise, the call of reverse is not relevant.

Fig. 9. Exception handling. Exceptional control flow is denoted by dashed lines. The right method is left by an exceptional return edge, the exception is then caught by an exception handler within the calling method.

Fig. 10. Size measures of the subject systems.

Fig. 11. sdcc object process graph.

Fig. 12. Example protocol for C files.

Fig. 13. ircII object process graph.

Fig. 14. Rhapsody IRC object process graph.

Fig. 15. Simplified object process graph for SQLite (abstracting from details).

Fig. 16. Size measures and statement coverage of test suite for subject systems.

Fig. 17. Static/dynamic object counts.

Fig. 18. Object process graph node and edge counts.

Fig. 19. Local variables: Percentage of nodes in dynamic object process graph compared to the corresponding static one. The X axis is an equidistant enumeration of all OPGs in ascending order of percentages.

Fig. 20. Heap objects: Percentage of nodes present in DOPG. ftp is missing because it only contains three heap objects.

Fig. 21. Local variables: Number of nodes in statically extracted object process graphs (X) compared to the number of nodes in their dynamic counterparts (Y).

Fig. 22. Heap objects: Number of nodes in statically extracted object process graphs (X) compared to the number of nodes in their dynamic counterparts (Y).