Re: passing parameters longer than expected

[Vern Hamberg <vhamberg@xxxxxxxxxxx>]

Joep, Dave, or anyone else

This is a perfect example of a requirement - you can submit one at the 
COMMON web site - www.common.org - then go to the Community tab and look 
for IBM i requirements, as I recall.

You don't need to join COMMON - although that's a good idea. You do need 
a user ID and password, there should be instructions there.

Once you do that, it will be transferred to an IBM database and the CAAC 
will work with IBM on your behalf. CAAC is COMMON Americas Advisory 
Council - I'm a member, along with just over a dozen others.

This path is good, because IBM listen to what CAAC says - because it is 
a group of small/medium business people that vet the requirements. Not 
just one person. The DCR process is also important, so that is an option 
- Rob has a lot of experience with that, also leading to benefit for us all.

Regards
Vern

On 3/28/2012 6:22 AM, J.Beckeringh@xxxxxxxxxxxxxxxxxxxxxxxxxx wrote:
> Barbara,
>
> Are you contemplating extending the parameter definition with
> OPTIONS(*EXACT)? If so, would it not be more appropriate to add a control
> specification keyword, e.g. PARMSIZE(*EXACT), to force the proper
> behaviour?
>
> Joep Beckeringh
>
>
> rpg400-l-bounces@xxxxxxxxxxxx wrote on 27-03-2012 23:44:09:
>
> > Barbara Morris
> >
> > For a prototyped parameter passed by reference (no CONST or VALUE), the
> > compiler doesn't complain about a fixed length passed parameter being
> > longer than the fixed length prototyped parameter. It only cares about
> > the passed parameter being shorter.
> >
> > For varying-length parameters passed by reference, the compiler does
> > require that the declared length of the passed parameter match the
> > declared length of the prototyped parameter.
> >
> > When we added support for prototyped parameters, we were mainly
> > concerned with data corruption, which isn't possible with too-long
> > parameters. After 15+ years of reflection, I think we were perhaps a bit
> > short-sighted not to also worry about data loss, and maybe should have
> > required *VARSIZE for any reference-parameter mismatch.
> >
> > Maybe OPTIONS(*EXACT) is needed.
> >
> > --
> > Barbara
> > --