Graph based verification of software and identification of defects by binary search on graph

In this paper a technique is proposed to simplify testing of software using dependency graphs. Dependency graphs help to highlight dependencies in the software and isolate pieces of code that would run independent of each other. By doing graph based tests, we could ensure full path coverage. Also, if defects are found, dependency graphs would help to isolate the defects to different independent parts of the graph through the use of graph partitioning techniques. A binary search for the defect over the graph would then help to isolate the source of the defect. Graph partitioning would partition the graph into 2 independent halves or 2 dependent halves with the input and output dependencies clearly mapped. A test on each half would help to isolate the defect to one or the other half. Each identified part of the graph would further be tested using the same technique till the defect is identified.