Re: MCH1210 Error (digression)

[Robert Donovan <rdonovan@xxxxxxxxxxxx>]

> Excerpts from midrange: 8-Aug-98 MCH1210 Error in QEXCAFI John
> Martz@us.ibm.com (2632*)

> --------> Start rambling digression <------------
> ... The default behavior is that the hardware will detect an
> attempt
> to store a signed negative value into an unsigned variable. If
> this happens,
> exception MCH1210 is signalled.

> For example, the following C code would probably run just fine on
> a PC, but
> will abend with MCH1210 on AS/400.
>  signed int a = -1;
>  unsigned int b;
>  b = a;   /* MCH1210 -- ka-booom! *./
> --------> End of  rambling digression <------------


Run-time detection of numeric overflow is a policy of a programming
language.  For example, the C sample code shown above will not abend
with MCH1210 because the C language doesn't allow it.  RPG ignores
numeric overflow, too, but the compiler provides an option to detect it.
COBOL ignores overflow unless ON SIZE is specified. ... 

All hardware platforms I am aware of provide some form of overflow
detection.  The usual technique is to set a condition code/register to
reflect carry into and out the high order bit of the numeric result.  If
a run-time exception is required, extra instructions must be generated
(by a compiler/translator) to test for overflow conditions and to post
an exception.  The PowerPC hardware works this way.  The IMPI hardware
is a little different -- some fixed-point instructions actually signal
exceptions when overflow is detected.

Bob Donovan, IBM Rochester, rjd@us.ibm.com

+---
| This is the Midrange System Mailing List!
| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
| To unsubscribe from this list send email to MIDRANGE-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: david@midrange.com
+---