2009 Mexican International Conference on Computer Science
DOI: 10.1109/enc.2009.11
Full text: Download
A group of 17 students applied 5 unit verification techniques in a simple Java program as training for a formal experiment. The verification techniques applied are desktop inspection, equivalence partitioning and boundary-value anal-ysis, decision table, linearly independent path, and multiple condition coverage. The first one is a static technique, while the others are dynamic. JUnit test cases are generated when dynamic techniques are applied. Both the defects and the execution time are registered. Execution time is considered as a cost measure for the techniques. Preliminary results yield three relevant conclusions. As a first conclusion, performance defects are not easily found. Secondly, unit verification is rather costly and the percentage of defects it detects is low. Finally desktop inspection detects a greater variety of defects than the other techniques.