home *** CD-ROM | disk | FTP | other *** search
- From: wayne@hpfcso.FC.HP.COM (Wayne Covington)
- Date: Thu, 19 Nov 1992 16:43:20 GMT
- Subject: Re: Denormal Exception, Turbo Pascal
- Message-ID: <9110035@hpfcso.FC.HP.COM>
- Organization: Hewlett-Packard, Fort Collins, CO, USA
- Path: sparky!uunet!wupost!sdd.hp.com!hpscit.sc.hp.com!scd.hp.com!hpscdm!hplextra!hpfcso!wayne
- Newsgroups: comp.lang.pascal
- References: <1992Nov1.164049.906@newsserver.rrzn.uni-hannover.de>
- Lines: 24
-
- korte@alhena.itp.uni-hannover.de (Gerd Kortemeyer) writes:
-
- > ...
- > if you unmask an exception of your math-copro you usually get a
- > run-time error if it occures. For example, if 'zero divide' is
- > ...
- > This works fine with any exception except for the denormal exception:
- > Turbo-Pascal continues to work on the program even though you had
- > a denormal operand in an arithmetic expression. When you look at things
- > more closely you will find out the copro raises an interrupt, but the
- > Turbo-Pascal exception handler doesn't react with a run-time error -
- > it rather produces TRASH because after coming back to the program the
- > copro-stack is left in a mess.
-
- Sounds to me like a disastrous bug.
-
- > Does anybody know what this behaviour of Turbo-Pascal is good for?
- > What SHOULD the exception handler do if it didn't have some bug?
- > ...
-
- If I unmask the denorm exception, I would expect a "Runtime error xxx ...
- denormalized number" indication.
-
- Wayne
-