Re: Identifying where *modules are used

["Charles Wilt" <charles.wilt@xxxxxxxxx>]

On Wed, Oct 8, 2008 at 2:35 PM,  <GKern@xxxxxxxxxxxxxxxx> wrote:
<snip>
> The only problem (and I know this is why you and your peers recommend
> avoiding this) is that when a modification is required to the module then
> all the programs using it need to be recompiled (and I know that these
> should be in a service program and accessed via prototyped calls). But we
> can't just convert everything to service programs overnight for practical
> reasons, so I'm left with trying to deal with the existing code.

As Scott mentions, you should have never used CALLB in the first
place.  CALLP should have been used instead of CALLB or for that
matter just the plain CALL.

Had you been using CALLP,  moving a given module to a *SRVPGM  would
simply be a matter of recreating the program that was making use of
it, no other code changes needed.  (Well you might add an h-spec with
the binding directory )

One additional problem with having a module bound into multiple
programs is that you end up with multiple copies of that module's
executable code in memory.  The OS shares executable code at the
program level.  So if job A is running PGM1 and job B is running PGM1,
there is only one copy of PGM1 executable code in memory.  On the
other hand, if PGM1 and PGM2 both have MODA bound into them, and job A
is running PGM1 and job B is running PGM2, then the executable code
that makes up MODA is duplicated.  If MODA was in a service program,
then both jobs would be using the single copy of the executable code
from the service program.

Not a huge deal I grant you, but could be significant if large modules
are bound into 100's of programs.


Charles Wilt