Re: CL pgm and pointer variables

[Hans Boldt <boldt@xxxxxxxxxx>]

Booth wrote:
> Wouldn't an even easier solution to these inherent
> problems with pointers be to just not use pointers?

To which Rob answered:
> Wouldn't an easy solution to divide-by-zero errors
> be to just not use division?

Yes, that would in fact be acceptable. As those who've used division
in RPG expressions have learned (sometimes the hard way), it is
sometimes best to avoid division in expressions if at all possible.

;-)

I do see your point, though. Why avoid something just because it is
potentially troublesome? The difference here is that when you have a
problem with division, you can normally easily locate the bug to the
appropriate expression.

On the other hand, pointer bugs can often be very difficult to trace
down. For example, let's say you DEALLOC some piece of storage, but
somewhere else you still have a pointer pointing at that piece of
storage. Since that storage may then be returned on subsequent
%ALLOC operations, unpredictable results can happen when accessing
data through the original pointer, or worse, when the storage is
changed using the original pointer. Even worse, the problems can
often *seem* to disappear just by changing one seemingly unrelated
thing.

I can attest to the problems you can have with pointers from my own
experience with the ILE RPG compiler, and that's using a language
with strong type checking of pointers! With untyped pointers (as in
RPG), the potential problems are much worse. At first glance,
debugging pointer problems can seem unsurmountable. Even at second
or third glance!

Cheers!  Hans