Re: Rare CPF3C17 Error

On 13-May-2016 10:52 -0500, Rich Loeber wrote:

On 13-May-2016 10:22 -0500, Rich Loeber wrote:

I have a CGI application running on multiple customer systems.
One customer reported today getting a CPF3C17 error once on their
system running IBM i OS 7.1. The error is on the very initial CGI
call that everyone has to run through to log in and use the
application.

Any ideas on what might cause this to happen once when there are no
other customers reporting this issue where literally hundreds of
users log in using this process at many, many sites?

Could a special character in the input stream cause something to
happen? The input here is simply a 10 character user profile.

Since this is on a customer's machine, I don't have direct access and
I have not been able to duplicate the problem, so no debugging.

This is happening on a call to "QtmhCvtDb".

So do they know if the error will recur for the specific user? Unpredictable errors often have an origin in initialization, set\reset, or variable corruption [e.g. per parameter size mismatches] problems.

Is the joblog available? How the error is manifest\percolated might reveal ways to collect more docs. If the program monitors for the error or generically, then the program might best dump the parameters for review have written them as part of logging; or if there is already a log file detailing the invocations, then be sure to check that log.

FWiW the message has two Message Data Fields Formats (FMT), though neither are manifest in the [neither first-level nor second-level] text. Based on other messages in that range, the value of (&1) could possibly identify the input parameter for which the "error occurred"; on that system, consider issuing CHGMSGD CPF3C1F MSG('Error occurred with input data parameter. (&1) (&2)') /* for whichever MSGF might get used */ so those message data elements are revealed, if there is ever a recurrence. Doubtful to assist [assuming the problem occurs again], else one would have expected the Recovery to have been written to suggest something like these; but currently, there are no (&#) in the recovery text of that message:
-- Verify the input data parameter &1 is correctly specified.
-- Verify the value for the length &2 of data parameter &1 is valid.