Patent US 5,692,122

A method of generating pairs of random interprogram conversation testcases is provided in which conversation statements are selected with which to construct the testcases in dependence on what statements are permitted for the predicted current conversation states of the testcases. The effect of each statement that is added to the testcases on the states of the testcases at eventual execution is predicted from predefined state tables for particular conversation protocols. As each conversation-type statement is added to a testcase, it is determined whether matching statements must be added to the other testcase and, if so, these matching statements are randomly generated in the same way. Self-checking code is added to the testcases to check correct execution. The method is particularly suitable for testing the support of asynchronous conversations between transaction processing programs running in different CICS systems.