Hi Lizette,
Thanks for your clarification.
How about posting out the whole kit & caboodle of what you've got so we can all
see directly?
Shouldn't take up too much space.
>Dave
>This is all native commands, no rexx
>Each step is separate JCL step
>Lizette
Paul,
>I think DanD hit it on the head with his suggestion about using infile &
>outfile dd statements instead of ids & ods.
>Ids & Ods use a disp of "old" during dynamic allocation.
>
>> IDS, even? I'd expect SHR to suffice for IDS.
Nope
>I have a sneaky feeling that your job is contending with itself because grs or
>mim can't always see quickly enough that the dataset is free'd after the
>del/def.
>As the message is an IKJ its implying that its all done within a TSO rexx
>routine and rexx is quite well know for erratic behaviour.
>
>> I believe Rexx messages are IRX; TSO are IKJ.
Quite so, hence implying.
My expectation was that it would be contained within some rexx.
Different strokes for different folks. Does any one really use native tso these
days for file manipulation? how retro :-)
>>The news of Rexx's "erratic behavior" hadn't reached me. Can you
>> cite an example?
Yep, practically anything that's poorly coded. Ever forget to initialise a
variable? or delstack? results not always as designed! :-)
>Do you have explicit close & free's coded in your rexx for the dataset prior
>to the final repro?
>
> Can contention occur within a single address space?
With poorly written code or implementation? been there so i wouldn't be
suprised, hence "sneaky feeling" :-)
>In any event, I'd suggest changing your repro to hardcoded infile outfile dd's
>with a disp of shr.
>
> Wouldn't EXC be safer for OUTFILE?
Yes, but you have that already by the dynamic ids and its failing to work.
I've suggested that Lizette posts out what she's got; which would help zero in
the issue.
If we can get to the root cause then we can get the correct dispositions all
the way though.
TTFN
Its Kipper tie time!
Dave
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
