On Wed, 9 Feb 2011 16:26:44 -0800, Charles Mills wrote: >It appears to me that the 8-character module name(s) passed to >CSVDYLPA must >be unique, is that correct? That is, unlike LOAD, it is not possible for two >different jobs or STCs to have access to different versions of a given >module, both of which have the same name. Am I missing anything? > >I am thinking specifically of a test version running in parallel with a >production version. Both want to load their subsidiary module 'FOO', but the >test version has a newer FOO than the one production already has in storage. >Would the ADD fail? Or would the new FOO cause the old FOO to be deleted, >leaving users who referenced it by address pointing at orphaned storage? > >Or would the new FOO replace the old FOO logically, but leave the old FOO in >memory such that users who referenced it by address would continue to use >the old copy?
To add to Binyamin's comments, if you want to delete a module from LPA that you had added with CSVDYLPA, you must do it explicitly. When you add a module to LPA using CSVDYLPA and that name is already in LPA, you end up with two modules with the same name. When a subsequent request is made to LOAD (or LINK, etc.) to that module, the last one to be added to LPA will be found, assuming no overriding LNKLST, JOBLIB, STEPLIB, TASKLIB or Job Pack area module is found. If you do add two modules of the same name to LPA, you will need some other mechanism to obtain the correct copy. If you subsequently delete the newest copy from LPA, the older one will be found in a normal LPA search. If your test module is located in a STEPLIB, it simplifies your task. LPA can always be overridden by STEPLIB. -- Tom Marchant ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
