Re: CPF5026 in ILE RPG trigger program eventually not getting caught?

James H. H. Lampert wrote:

Kurt Anderson wrote:

Did you check the job log to see if the abend is as you
suspected, the message queue filling up and not extending?
That's what I think would have happened as well.

Not from the time it blew up, but today, the user tried the same thing, and it didn't abend, but he did end up with a joblog full
of CPF5026 messages. Which is exactly why both of the
possibilities I mentioned occurred to me (in the order in which I
mentioned them).

At any rate, it was trivial to add a check for the existence of the record before attempting to write it. Just did a SETLL on FOO, and checked NOT %EQUAL(FOO) before writing to FOO.

I have seen similar for non-trigger programs at least. This case seems just like one of many cases of poor coding, for having ignored an error condition. In some similar cases that I have seen, the ignored error condition causes the job to terminate for filling the job msgq or wrap the job msgq many times where the number of ignored incidents was as many objects or records that the program would attempt to process.

Perhaps no reason to merely speculate.... Even if the joblog is not available from the past job, the request to DSPLOG QHST JOB(named) PERIOD((appropriate)) would show messages about joblog wrapping; i.e. it is the job itself which sends the CPI2417 messages, not the subsystem job or some other system job. If similarly, as I expect, the message indicating that the job message queue full effected the job termination would appear in that same DSPLOG request. Otherwise knowing the specific message identifier to search irrespective of JOB(), or instead searching the job name or job number in a spool of all available history data [over a period] might show a message with more details about if\why the job ended.

Regards, Chuck