(j3.2006) (SC22WG5.4899) [ukfortran] Comment on a comment on the WG5 letterballot on N1947

[Keith Bierman]

Also, if memory serves, one of the Patriot missile failures (but again, not
Fortran code, and indeed, a choice to use fixed point rather than floating
point due to perceived hw limitations).

I do not doubt that there are environments where catching integer overflow
is a GoodThing. However, I concur with Bill in that the number of such
applications that I have seen in Fortran applications (and where the code
made sense to write as provided) is so close to zero as to be strongly
suggestive that this isn't where the committee (or implementors) should
lavish much attention.

Obviously, there is a potential chicken and egg situation ... people who
really, really, really demand such support will pick a language such as Ada
which has it.

Unclear from the press clippings if the Ariane code should have been using
an integer data type anyway. But certainly since it was written in a
language with such support AND the coders chose to ignore it doesn't make
the case very compelling. Programmers can find ways to make serious
blunders no matter how many safety features we provide ;>

Keith Bierman
khbkhb at gmail.com
kbiermank AIM
303 997 2749


On Wed, Jan 16, 2013 at 12:04 PM, Van Snyder <Van.Snyder at jpl.nasa.gov>wrote:

> On Wed, 2013-01-16 at 11:07 -0600, Bill Long wrote:
> > In recent memory, I've only encountered one case of harmful integer
> > overflow,
>
> There was the Ariane 5 crash, which was rather expensive.
>
> True, the code was written in Ada, not Fortran, and the integer overflow
> was caught, but ignored.
>
>
> _______________________________________________
> J3 mailing list
> J3 at mailman.j3-fortran.org
> http://mailman.j3-fortran.org/mailman/listinfo/j3
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.j3-fortran.org/pipermail/j3/attachments/20130116/6116c5a8/attachment.html