When the server encounters an error during the execution of a program, it calls the program again, indicating the error condition in the control block field flag_value. The program then does one of the following, indicating its response in the action_value field:

• Free any memory allocated during program execution, and exit, issuing an action_value of 9 (exit). The program must allocate and free its own dynamic storage. Make sure that the program frees any allocated resources (especially memory) before issuing an action_value of 9. Not freeing memory may cause the server to fail at a later point in time.
• Return a message to the server to explain the error, issuing an action_value of 4. The server then attempts to return the message to the client application and call the program again, which must free its resources and end, as described above.

  Messages are retrieved by the client application before the processing of an answer set, or after the completion of answer set processing.

Only the action_values for returning a message, or for exiting, are valid after the server has reported an error. Any other action_value returned by the program causes the server to end without further calls to the program. If the program does return another action_value, the server attempts to report the program's incorrect behavior to the client application using a server-initiated message.

The following figure illustrates the correct error handling sequence. In the figure, the following flag_values are shown:

1 Indicates the first call to the program.

0 Indicates a subsequent call to the program, without an error.

606 Indicates an error.

The program's choices when it receives the error code 606 are also illustrated.