| john.collins@simcon.uk.com 2007-11-27, 7:18 pm |
| On Nov 26, 9:24 pm, "Craig Dedo" <cd...@wi.rr.com> wrote:
>
> You remember what I said on October 18 very well. For the benefit of other
> readers, here is a recap of what I said then.
>
> [Begin Excerpt]
> There is nothing in the Fortran standard that requires a Fortran processor
> to trap exceptions. Among other exclusions, Section 1.4, "Exclusions", says,
> "1.4 Exclusions
> This standard does not specify ...
Yes, I do remember what you said. It upset a lot of people.
The issue is that Fortran programs are used to design commercial
aircraft,
nuclear power plant and all sorts of potentially exciting things. The
loss
of Sleipner A, (About #700,000,000) was not the most expensive
software
error. It was simply (one of) the most well published (e.g.
http://www.ima.umn.edu/~arnold/disasters/sleipner.html )
What my colleagues and I would like to see is a requirement in the
standard
for exceptions and errors to be trapped. We would also like a
requirement
for errors to be trapped in widely used extensions to Fortran, like
STRUCTURE, MAP and UNION or constructs like INTEGER*4. We write
compilers,
so I realise the can of worms this could open, but I would like to go
on
flying in airplanes, using electricity and driving automobiles with
some
confidence in the design process. A program which crashes is a
nuisance.
A program which silently reports the wrong answers is, to us, a
problem.
I would like to know what other users feel about this, and what, if
anything, could be done in the standard.
Best wishes,
John.
|