RE: Activation groups for beginners

[Buck Calabro <Buck.Calabro@xxxxxxxxxxxx>]

Nelson Smith wrote:

>Now, I've got to find the best
>solution to fix the 500 or so
>driver CL's.

I don't think you can address only the CL programs...

>1.)  I can't remove the Rclrsc without
>     restructuring the many programs that
>     depend on it to close them down.

Remember that RCLRSC won't close down ILE programs unless they were compiled
DFTACTGRP(*YES), so the action you think is happening may in fact not be
occurring.

>2.)  I can't recompile the CL's without
>     a through investigation of the
>     override scopes, etc., throughout
>     the entire job.

Recompile them to make them ILE?

>3.)  I can see no justification for having
>     the RclActGrp in the driver CL's
>     at all, can you?

Certainly not RCLACTGRP *ELIGIBLE, but why did they get put there in the
first place?  I think this is the key to how to address the situation.  If
it's a simple misunderstanding of how AGs work, then perhaps you really want
to put every ILE program in *NEW except the service programs, which should
be *CALLER.  Then the AGs would clean themselves up:

CL *DAG
  RPG *NEW (AG0001)
    SRVPGM *CALLER (AG0001)
  seton LR
  /* AG0001 is gone */
Endpgm

If you really aren't sharing, then perhaps the client programs should be in
a single AG instead of *NEW, but this doesn't sound like your scenario
because the CLs are doing RCLRSC to close things down.  Does this mean your
RPG programs return without LR?

>My plan now is to remove the RclActGrp from
>the CL's, but leave the service programs in
>their own activation group.  I think this
>should solve the immediate problems.

Difficult to say, but reading between the lines makes me think you have RPG
programs that don't set LR on and currently (OPM version) rely on RCLRSC to
do the cleanup.  I'd be leery of a design where ILE client programs run in
*CALLER if that means they will run in *DAG.  You can't ever really close
them properly because RCLRSC has no effect on them.  I think making the
clients *NEW fits your scenario closest, but like I said, I'm reading
between the lines and guessing.
  --buck