On Thu, 19 Sep 2013 14:06:25 -0500, Doug Henry wrote:

>On Thu, 19 Sep 2013 13:50:22 -0400, Gerhard Postpischil wrote:
>
>>In any case, IBM should be persuaded to either produce a JCL error or
>>modify the directory build to write an EOF.
>
>Of course their are methods to default the number of directory blocks. One 
>that easily comes to mind is dataclass. But isn't all of this something that 
>occurs during allocation and open is to late. Unless you create at least one 
>directory block for a pds , what good is it?
>
I can think of a scenario, far-fetched but still plausible, where this
might be normal operation.  Suppose I have a program that generates
JCL.  The submitted job creates and processes a variable number of
members, but the number is known a priori.  Compulsively, I calculate
the precise number of directory blocks needed.  The number may
sometimes have been zero, in which case the job allocates a PDS
with no directory blocks; processes no members; deletes the PDS
and terminates.

I would be shocked and dismayed to find my job, which has been working
for years, to start failing on conditions which it has previously handled
successfully.  One might argue that for economy I should bypass the
allocate and delete when there are no members.  I don't care; it's a
boundary condition that should be handled smoothly.

All hypothetical, of course.

It was pointed out here that when allocating a PDS the number of
directory blocks must be specified.  Barbara specified a number.
The documentation, as cited, did not state that it must be specified
as nonzero.  Not writing the EOF is contrary to the doc, and should
be APARed.

-- gil

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

Reply via email to