I think you're doing it the right way. Sub-procedures invoked in the
call cause entries on the stack too. Ecuyer's code may have been simply
a matter of what he saw every time he tried it. You're walking up the
stack looking for the entry immediately prior to the piece of database
code that called the trigger program. Be careful, that may itself be a
sub-procedure name and not a program name. If your logic is dependent on
comparing the program name, you may want to test your results to see if
it is a sub-procedure name and, if so, continue to walk up the stack. Do
you suppose Ecuyer's test program had two sub-procedure calls wrapping
the insert?
-----Original Message-----
From: midrange-l-bounces@xxxxxxxxxxxx
[mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of James Lampert
Sent: Thursday, September 04, 2008 2:14 PM
To: Midrange Systems Technical Discussion
Subject: Re: Trigger question
Modifying Richard Ecuyer's example from February 2006:
http://archive.midrange.com/midrange-l/200602/msg00666.html
with
c Eval pm_counter=1
c movel 'QDBUDR ' Pm_CSEntry
and then, if the first QMHSNDPM call fails with CPF2479, calling it
again with
c movel 'QDBPUT ' Pm_CSEntry
I get the name of the program that immediately tripped the trigger.
I noticed a lag in the response time, so to determine whether it was in
the API calls or in the DSPLY statement, I added a couple of timestamps,
one before the first QMHSNDPM call, and one after the QMHRCVPM call,
DSPLYing them both just before the final DSPLY with my result, and found
that the API calls only took 2-4 milliseconds on a "slim" 170 running
V4R4, so I guess I'm OK on time.
But can anybody think of a case (or an OS release) where climbing one
call stack level beyond the QDBUDR/QDBPUT call might fail to return the
right program? Or explain why Ecuyer's example just climbed a straight 5
call stack levels, overshooting the target by 2 levels?
midrange.com
