Software Safety Checks Using Stored Tested Paths

Abstract The paper describes an investigation on a technique which idea is to store all program paths executed during the testing of a safety critical computer program. During on-line execution of this program a check can be made on whether an actually executed path is found in the file of tested paths. If an untested path is followed, a special action can be taken. Various techniques to store the tested paths, and to check executed paths against the paths in the data base of tested paths were investigated, with respect to storage requirement, time requirement and accuracy of approximation. Some of the methods were applicable. A large number of test runs were made, with different test data profiles, different storing techniques and different faults seeded into the program. The objective was to investigate the two questions: “how reliable is the method to trap failures due to residual faults”, and “how often are paths in correct executions not found among the tested paths”, i.e. “how often are spurious actions taken”.