Re: ILE Concepts

Two comments Dieter ...

" (it's only using procedure names, not parms)" - At run time (which is what you were talking about) it is using the slot number compiled into the program at creation time. That is why it is critical when managing multiple signatures to realize that the _only_ one that really matters is *Current.

This is why I disagree with your later comment "@one signature for all versions: - a bad idea at all (comparable to LVLCHK *NONE for all files)". Since you have to control the slot sequence via the *Current signature anyway then to me using a single signature makes perfect sense. The LVLCHK is really a false equivalency - the approach IBM have taken (and to which I subscribe) is to use a fixed signature so that you _can_ change it later if you really need to face a re-bind.


Jon P

On Nov 2, 2022, at 5:08 AM, dieter@xxxxxxxxxxxx <dieter.bender@xxxxxxxxxxxx> wrote:

... interisting discussion, some corections and additional thoughts:

@export *all:
- all procedures ar exported (using binder language only a subset)
- sequence is aplphabetically (not source sequence)
- errors will be forced to activation time as far as the compiler could do (it's only using procedure names, not parms)

@one module - one SRVPGM:
- if the exported procedures are prefixed by modulename all external references are unique

These two guidelines together will provide a simple and foolproof strategy to avoid strange effects at runtime. But you would have to pay some price (you'll never get anything for nothing!).
- you would pay with some additional binding steps.
- you would pay with some additional deployed objects.
- you would need some check what signatures are broken (rather easy to compare imports and exports automatically)

To illustrate why and when you will need a more sophisticated approach, a short discussion about publishing a public library of functions:
- deploying a complete application in one SRVPGM (and a corresponding BNDDIR) makes it easier for other applications to use the library (=> Java would deploy one jar file)
- Binder Source is used to ensure, that only the public procedures are exported, signatures are used for versioning.
- possible libl conflicts are minimized by a one for all aproach
- additional possibilities to controll multiple activations of the same module (thinking of OA handlers)
But you would have to pay some price (you'll never get anything for nothing!).
- errors in handling your Binder Source could transfer errors from activationtime to runtime, wich is harder to detect.

(ArdGate is using an alternative aproach. Deployment unit is a library and all internal references are resolved by dynamic binding at runtime - but this is another discussion)

@one signature for all versions:
- a bad idea at all (comparable to LVLCHK *NONE for all files)
- has no influence to activation

D*B





--
This is the RPG programming on IBM i (RPG400-L) mailing list
To post a message email: RPG400-L@xxxxxxxxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: https://lists.midrange.com/mailman/listinfo/rpg400-l
or email: RPG400-L-request@xxxxxxxxxxxxxxxxxx
Before posting, please take a moment to review the archives
at https://archive.midrange.com/rpg400-l.

Please contact support@xxxxxxxxxxxxxxxxxxxx for any subscription related questions.

Help support midrange.com by shopping at amazon.com with our affiliate link: https://amazon.midrange.com