>Hi Tom...
>
>I have a question based on a comment you made to Itzchak. You said:
>
>"None specifically for the VS COBOL II modules. OTOH, when recompiling
>> with COBOL V5 we recommend a lot of testing, because it can reveal hidden=20
>> invalid COBOL programs. These programs could have been invalid for 20=20
>> or 30 years, but each new release generated the same instructions so=20
>> the behavior was the same. With COBOL V5, these bad programs might=20
>> start getting the wrong answers. So, for every program recompiled=20
>> with COBOL V5 we recommend thorough regression testing."
>
>I'm aware of the issue with the SEARCH ALL verb and the issue (described in
> Appendix G) about the wrong length record being flagged now in COBOL 5.
Neither of these have been problems for customer migrating to COBOL V5.
>Are there other issues that would force significant regression testing over
>and above normal testing? Is your comment primarily for programs last
>compiled with VS COBOL II?
Programs last compiled with VS COBOL II or any other compiler are also not
a problem. What is a problem is illegal programs that for some reason have
not been failing, but the code generation was the same from VS COBOL II
through Enterprise COBOL V4.2, so they got the same behavior. You
should have been in my SHARE session yesterday!
Some of the illegal/invalid programs we have seen that behave differently
with COBOL V5 compared to any previous compilers:
1) Mis-matched parameters: Caller passes 100 bytes, callee uses 500 bytes
and overwrites IBM control blocks
2) Over-used tables: Table defined as OCCURS 0 to 1 TIMES DEPENDING ON X,
then X is set to 250.
3) Spaces in numeric DISPLAY data items (zoned decimal)
4) Programs that assume a specific layout of WORKING-STORAGE and use data
across data item boundaries
Cheers,
TomR >> COBOL is the Language of the Future! <<
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
