Re: Passing By Reference

On Tue, 26 Nov 2002, JJW wrote:
>
> Dow (%scan(C1:F1) > *Zeros)  AND  (%scan(C1:F1) <= InLen);
>     %subst(F1:1:InLen) = %replace('':F1:%scan(C1:F1):1);
>
>     F1 = %replace(' ':F1:InLen:1);
>
> endDO;
>


Any time you have a 'F1 = <anything>' in your subprocedure, and F1 is a
*VARSIZE parameter, you will have problems.

Why is that?  Okay, I'll explain.

The 'eval' op-code (Which is what you're using here, it's implied in
a free form statement that has no op-code) always sets every character
when you do an assignment.

For example, if you have a 25A variable, and you do:

              My_25A = 'X';

The eval op-code not only sets the first character to an X, it also sets
the other 24 characters to spaces.

That's a problem with a *VARSIZE variable, because the compiler doesn't
actually know the length of the variable.   If you have a 32766A which
is *VARSIZE, then the compiler ALWAYS treats it as if it's 32766 bytes
long.   It doesn't matter that you only passed 10 bytes, it still thinks
it's 32766 bytes long.

So, if you've passed a 10A as a 32766A *VARSIZE, and do a

        F1 = 'X';

It sets the first byte to an X.  It also sets the other 32765 bytes to
blanks.   Since you only passed 10 bytes, you are overwriting 32755 bytes
with blanks which don't belong to you.   Those bytes could be ANYTHING.
It might end up having no effect, but then again, it might end up
causing mass chaos, including program crashes that are extremely difficult
to track down.

Therefore, I'll say it again.   If you use *VARSIZE, don't EVER do an
    F1 = <anything>

That's why you need to use
    %subst(F1:1:InLen) = <anything>

That way, you're telling the compiler 'ONLY work with InLen bytes of
my variable!'