On 05/19/2015 07:16 PM, Shmuel Metz (Seymour J.) wrote:
> In <[email protected]>, on 05/14/2015
>    at 09:17 AM, Joel Ewing <[email protected]> said:
> 
>> I think we resolved the issue by converting the CLIST to REXX so >another 
>> TCB was involved.
> 
> How is that relevant? What is relevant is how you dealt with ERROR and
> FAILURE in the REXX code.
>  
> 

My recollection is that the immediate bailout from the CLIST and from
batch TSO on a non-zero TSO command processor RC only occurred for
commands running "directly" under the batch TMP TCB (and running a CLIST
command directly from SYSTSIN was actually causing all TSO commands
within the CLIST to run at that level as well!).  When executing TSO
commands from a REXX exec invoked from SYSTSIN that was not the case,
and premature termination of the REXX exec and batch TSO did not occur
when TSO SEND was invoked within the EXEC and gave a non-zero return
code.  The replacement REXX EXEC produced a zero return code for the
entire EXEC based on its own internal criteria, which did not involve
checking the RC from SEND.

We found the premature bailout difference in behavior between CLIST and
REXX environments unexpected and counter-intuitive, but it seemed to
make perfect sense to those at IBM TSO support familiar with TSO
internals.  I accepted IBM closing our problem as a DOC change: the TSO
manuals at the time did document this behavior, but in a way that
required a TSO internals specialist for correct interpretation.

The root cause of our problem was that TSO SEND (working as designed
with user log files) was returning a non-zero return code when it
actually did NOT have an error and when message delivery eventually
succeeded. An actual "failure" of the requested SEND command action was
not an issue, and it was unnecessary to do anything special to handle an
ERROR or FAILURE of SEND in the REXX Exec.

The nature of the original CLIST in question was such that a failure to
execute statements in the CLIST following the SEND was a much more
serious problem for us than if the message had just failed to reach its
intended destination, which did not occur.   In our case the end user
would eventually have detected loss of an expected confirmation message
on his own and followed up if that had occurred.

-- 
Joel C. Ewing,    Bentonville, AR       [email protected] 

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to