Floating point and pic 9.99

[Alain Reymond]

Hello,

here is a small program :

       1 mydouble usage float-long.
       1 wss-test pic -99.99.
       procedure division.
       move -21.42 to mydouble
       display "double  : " mydouble
       move -21.42 to wss-test
       move wss-test to mydouble
       display "wss-test: " wss-test
       display "double  : " mydouble

Here is the result using Gnucobol:

double  : -21.42
wss-test: -21.42
double  : 13224.42

and using MF:

double  : -.214200000000000015E 002
wss-test: -21.42
double  : -.214200000000000015E 002

The result of the move from a pic -99.99 is "correct" under MF - at least with 12 zeroes precision - but not with Gnucobol. Is there a reason or is it a bug ?
Regards.

[Alain Reymond]

As follow-up, the solution is ti use the numval function :
move numval(wss-test) to mydouble
But the difference of behaviour of the two compilers is still astonishing.

[Ralph Linkletter]

I believe this to be the proper assignment.
1 wss-test pic S99V99.
move wss-test to mydouble

1 wss-test pic -99.99.
Defined as above is a numeric edited field.
As such the intent of usage is a receiving field only .

Behavior in using a numeric edited field in the manner of your example would produce unpredictable results regardless the compiler vendor.

I would suspect that a warning diagnostic was produced ?

[Alain Reymond]

You are absolutely right.
No warning issued.
I was just confused because we are porting (or trying to!) an app from MF to Gnucobol and it works without any problem under MF.
Now, we know!