My article "Easy and Powerful Message Presentation" (April 1992) described the usage of message subfiles to present messages in application programs typically written in high-level languages like RPG. Everything was peachy until we came across a bug-not in the message subfile technique, but rather in its implementation in RPG. This is what happens:
You code a CALL operation to execute a CL program that sends messages to the RPG program, including an indicator in columns 56-57 as shown in 2a. This indicator turns on if the CL program ends abnormally, so your RPG program can sense this abnormal termination and take corrective action.
You code a CALL operation to execute a CL program that sends messages to the RPG program, including an indicator in columns 56-57 as shown in Figure 2a. This indicator turns on if the CL program ends abnormally, so your RPG program can sense this abnormal termination and take corrective action.
The problem occurs when you use this technique with the message subfile: the system insists on issuing a function check message (CPF9999, which is *ESCAPE) whenever the CL program ends abnormally, even if you monitor for this condition with the RPG indicator in columns 56-57. So when your message subfile is presented, it will have one message at the very bottom that you didn't send and that looks serious enough to confuse (and scare off) the user.
Fortunately, the solution is rather simple. AS/400 messages are powerful; all you need to do is CALL a new CL program if the RPG indicator is on. The CL program (RMVLMSG) removes the last message from the caller's program message queue (which is the CPF9999 message issued by the system), thus cleaning up the message subfile. The RMVLMSG program is shown in 2b.
Fortunately, the solution is rather simple. AS/400 messages are powerful; all you need to do is CALL a new CL program if the RPG indicator is on. The CL program (RMVLMSG) removes the last message from the caller's program message queue (which is the CPF9999 message issued by the system), thus cleaning up the message subfile. The RMVLMSG program is shown in Figure 2b.
We thought that this problem should be fixed by IBM, but (quoting) "The development at this time has no plan to incorporate such enhancement." Unquote. Hooray for IBM.
TechTalk: Message Subfile Bug
Figure 2A Typical CALL statement
Figure 2a: Typical CALL Statement ... 1 ...+... 2 ...+... 3 ...+... 4 ...+... 5 ...+... 6 C CALL 'CLPGM' 99
TechTalk: Message Subfile Bug
Figure 2B CL program RMVLMSG
RMVLMSG: + PGM RCVMSG PGMQ(*PRV) MSGTYPE(*LAST) RMV(*YES) ENDPGM
LATEST COMMENTS
MC Press Online