IBM Mainframe Forum

Hello,

Here is fun JCL question, take a dig

<Disclaimer: This is not a query; it is not pointed to moderators/senior members, though you're welcome to take pot-shots at it for laughs only ; this is entirely to test a beginner's own understanding of COND parameter in JCLs; you can use this question to play around with your JCL working on various scenarios; in case a moderator feels that this topic can lead to helpful info, it can be moved to JCL part of the forum with edits. To those attempting the question - ensure that you test your solutions prior to posting them.>

Requirement: Skip the first step in a multi-step JCL using COND parameter; you're allowed to use only and only COND parameter to achieve this.

Cheerio.

OK. The answer is undocumented and somewhat counter-intuitive, and should not be used in Prodiction, Trying to find it could get people finding out more about COND on the way, we'll see.

If allowed to insert an additional step as the 1st step, the following would work:

What was the 1st step now becomes step S02 (the original 1st step).

Yes, but that is doing more than just using COND and still what is now the first step in the JOB will be executed.

The idea is a COND on the first step in the JOB which will prevent that first step executing. You'll need to "suspend disbelief" and look at what is available for COND. Don't say to yourself "that won't work because..." because one of them will

These users thanked the author BillyBoyo for the post:

Aki88 (Sat Nov 22, 2014 12:10 pm)

Okay, after a bit of trial and error, I found the following worked:

I HATE COND. I much prefer IF/THEN/ELSE/ENDIF.

Yep. Only if the previous step abended, the first step runs. No previous step, so didn't abend, so step doesn't run. Someone showed it to me a couple of months ago, and I didn't believe it till I ran it myself

It is not documented (that I can find) so behaviour can change without warning in a future z/OS release, so don't use it. COND is, shall we say, esoteric.

The JCL Reference manual explicitly states that COND on the first EXEC of a job evaluates to false to allow the step to execute. ONLY extends this documented behavior.

Except it doesn't run. And the documentation doesn't say "except for COND=ONLY". It's exactly what I thought when I saw it first Took me running a JOB to realise what I was being shown. Despite what is documented for COND on the first step, COND=ONLY makes the step not run. Yes, according to the documentation the first step should always run, irrespective of COND.

These users thanked the author BillyBoyo for the post:

Aki88 (Sat Nov 22, 2014 12:13 pm)

I think Figure 16-2 in the Reference Manual documents this situation:
"16.5.7 Summary of COND Parameters"
The 1st of the 4 ONLY scenarios is the one that applies, I believe. So it IS documented.
© Copyright IBM Corp. 1988, 2001
I STILL hate COND.

I don't think so. It is logically impossible for a previous step to have abended (there is no previous step). That can't just be lumped together with no destinction between it and the case where no previous actual step has abended.

It does mean if your COND=ONLY step fails, you have to override the COND (to anything other than ONLY) to get it to to be the target of a RESTART and actually execute. I think that if the actual behaviour of COND=ONLY on the first step was the intended behaviour, that would need to be documented.

I wonder if it always worked like that. Cue someone with Hercules running MVS 3.8J...