Re: Infinit loop in /free

[Hans Boldt <boldt@xxxxxxxxxx>]

Scott Klement wrote:

> > But to the question at hand, which is, does "DO *HIVAL" loop an
> > indeterminate number of times? Well, technically, yes. At least, if you
> > compile using TRUNCNBR(*YES), which is the default. By default, decimal
> > arithmetic on fixed-form calcs (but not the expression calcs) truncates
> > on overflow. That means that when the implicit loop index reaches the
> > value of *HIVAL, rather than overflowing when it's incremented again, it
> > gets truncated. Thus, the implicit loop index never exceeds the limit,
> > and the loop keeps churning away. (At least until the operator notices
> > and cancels the job!)
>
>
> Wow... that's obscure.   I can't believe people would write code like
> this!  I mean, if we didn't have the ability to get answers from the
> compiler team, how would I have EVER known that "DO *HIVAL" was going to
> loop infinitely?
>
> It sure isn't obvious.

You may well have run into this issue in another guise. For example, how 
many times does the DO loop iterate in the following program?

C     1             do        9             x                 1 0
C     x             dsply                   a                 1
C                   enddo
C                   seton                                        lr

9 times? Uh-uh. Since the loop index X can never exceed the maximum 
value for that variable, the loop never ends.

If this hasn't bitten you, it has almost certainly caught some readers 
of this list off guard. It's yet another reason to compile with 
TRUNCNBR(*NO). Or to always use integer or unsigned variables for loop 
indexes, and not decimal.

Cheers! Hans