Abstract: A server computer acts as a central repository for tests performed by any number of connected client computers, as well the results of these tests returned by the client computers. A test manager executed by the server computer analyzes the results of the tests performed by the client computers, and determines which tests have passed or failed during execution, whether the tests failed because of an application or other error, and which tests should be re-executed by the same or different client computer. A test grouper analyzes the error messages produced by the application being tested, so that tests that reveal the same defect can be grouped together. A test reducer is iteratively applied to selected tests to reduce the test to the smallest subset of the original test that still reveals the defect. In this manner, the present invention maximizes the testing efficiency of the resources used and minimizes the amount of time required of the operator to confirm failures.

Abstract: A test generator creates a cyclic directed graph representation of the interface of a program being tested and then generates tests from this representation. In generating the tests, the test generator iteratively selects traversal paths through the cyclic directed graph that result in traversal of every edge in the graph in a random order with a minimum number of iterations. The resulting tests contain randomly selected actions and randomly generated data, and thus when executed, these tests randomly manipulate the program being tested.