On Fri, 6 May 2016 09:16:34 +1000, Andrew Rowley wrote:
>On 6/05/2016 8:11, Mark Zelden wrote:
>> 1) There is a DDDEF called AAAA. It points to a DSN defined with DISP=SHR
>> (no volser / unit).
>> 2) The SYSLIB concatenation includes DDNAME AAAA (last in the concatenation,
>> I didn't
>> test a different order).
>> 3) At execution time DD AAAA is overridden in the JCL to point to a
>> different VOLSER than
>> the cataloged version.
>> 4) The data set used in the SYSLIB concatenation is the cataloged version,
>> not the override.
>
>I think it is the behaviour I would expect.
>I think the SYSLIB concatenation includes DDDEFs which specify datasets,
>not DDNAMEs.
>
It's probably WAD, although it's equally reasonable to expect that SMP/E
would allocate the DDNAMES as explicitly coded or overridden, then perform
a DYNALLOC CONCAT.
A syntax less likely to mislead would eliminate the bound variable names
and use lists such as:
DDDEF SYSLIB( ( DSN( SYS1.MACLIB ) SHR )
( DSN( ISV.MACLIB ) SHR ) ) .
>Overriding SMP/E DDDEFs via JCL always seemed to me like it was a bad
>thing to do... I always felt that the SMP/E zones should contain all the
>definitions for the installation. Overriding DDs seemed like a recipe
>for an inconsistent installation if e.g. some people did the override
>and others didn't, or if some overrides became out of date.
>
A related hazard is splitting SMP/E operations beteen systems which
might have different levels of SYS1.MACLIB in various master catalogs.
I'd be pretty comfortable overriding SYSPRINT.
-- gil
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
