Re: Service programs and QSRVSRC

[Vern Hamberg <vhamberg@xxxxxxxxxxx>]

Exactly, Rory. Kurt, keeping a hard-coded signature was a choice I made 
- it seemed the easier way to deal with the issue of treating customers 
right. So long as I am very careful to keep the original order of 
exports the same, and add only to the end, as Simon said, there is no 
need to recompile anything that does not use new exports.

Vern

Rory Hewitt wrote:
> Kurt,
>
> I think the point is that if you had a hard-coded signature, you wouldn't
> have to recompile anything anyway. You just leave the signature as it was
> before you added the new procedure. No need to add the *PRV block or
> anything else - just add the new procedure.
>
> If you're using generated signatures, then yes, you need to use *PRV in
> order not to have to recompile all 40 programs. So *PRV isn't useless - it's
> a requirement. But if you use hard-coded signatures in the first place, then
> your life would be much easier.
>
> Rory
>
> On Tue, Oct 13, 2009 at 2:23 PM, Kurt Anderson <kurt.anderson@xxxxxxxxxxxxxx
>   
> > wrote:
> >     
>
>   
> > PGMLVL(*PRV) is useless?  Ok, I have to comment.  Maybe there's something
> > I'm missing which would result in me not having to use *PRV, but as it
> > stands, I am a fan.
> >
> > Let's say I have 40 programs using one service program.  I now add an
> > exported procedure (no CMS here), and that new exported procedure is only
> > used by 5 of those 40 programs.  Why would I not have a *PRV Export List
> > signature so I can safely leave those other 35 programs be?
> >
> > If I ever do modify any of those 35 programs, then they will have to be
> > compiled, and they will pick up the latest signature.
> >
> > And in regard to Vern as an ISP - I really don't see why *PRV wouldn't
> > benefit you.  Your requirement was to not require the customer to recompile.
> >  Well, there you go.  Unless I'm not understanding the exact change you had
> > in mind (I did come into this thread late).
> >
> > I still think Scott's idea of embedding the version number into a signature
> > is neat, although nothing more useful than readability (and one should never
> > underestimate readability).  Although as someone stated, a
> > programmer-defined signature is prone to error (i.e. forgetting to change
> > it).
> >
> > -Kurt
> >
> >