Re: CRTSRVPGM

On Wed, Jan 12, 2011 at 12:24 PM, David FOXWELL <David.FOXWELL@xxxxxxxxx> wrote:

Thanks Mark, Bryce, Charles

By lying to the system, I think you mean that in this case I told it that the current signature used by the program was still
supported but it was in fact not. The program trusted me thinking it would find PROC2 in slot2 but found PROC3. So, if I now
recompile the program, it will now know that it must use slot3 for PROC2. I hope I have that right.

Yep, that's correct.

Still, we don't have a single srvpgm in the shop, and when a module containing exported procedures gets modified, then all
programs using it get recompiled. I'm imagining that recompiling everything is a bigger issue elsewhere than it is for us.

Recompiling everything isn't the issue really, assuming you've got a decent CMS.

Retesting everything you've recompiled is, unless you've got automated
regression testing.

When you just recompile the service program, you've only got one
changed object to test.

I feel that it's a little scary though. Imagine someone made a mistake in the binder source? I suppose it's difficult to make,
but it must be possible.

Sure, but if somebody can't be trusted to add lines to the
end...perhaps they are in the wrong line of work?

And it's easy enough to double check. Lot's easier than double
checking that you didn't miss recompiling something that used a bound
by copy module.

Finally, at this stage I still don't know which technique to prefer with the binder source.

I agree with Barbara's conclusions from the message you linked to:
1. Don't use EXPORT(*ALL).
2. You can't change the order of the exports with either style of binder
source, so you might as well use the simple form with the single
list of exports.


Charles