JCL can submit further jobs to the internal reader to get around the 255 steps limit. Loops can be implemented in JCL by submitting a job that then resubmits itself n times, based on a counter passed in a dataset (and decremented at each resubmission) and on a same LPAR. A jobstep that creates a new GDG entry would create it as *.GDG(+1) and all subsequent steps would then refer to it as *.GDG(0) - hence there would be no problem with restarting a same job from a subsequent step onwards. A job's Boolean logic processing can be implemented through COND= testing alone, using IEFBR14 'dummy' steps. The only useful purpose REXX could serve would be to submit one or more jobs *as JCL* to the internal reader, but in a same way that CONTROL-M already does. REXX could not replace JCL: it would be unmaintainable. Chris Poncelet
On 14/07/2018 00:46, Robert Prins wrote: > On 2018-07-13 01:29, CM Poncelet wrote: >> What 'problem' with JCL is fixed by replacing it with REXX? > > REXX doesn't have a limit of 255 steps in a job, and that limit may be > reached when doing regression tests. It also allows loops. Example I > have is too long to post here (probably). > > Robert ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
