Re: Gfortran (mis)information in the Polyhedron site.

[John Appleyard <john dot appleyard at polyhedron dot com>]

FX Coudert wrote:
> Hello,
>
> > Yes, gfortran's documentation is fairly weak in describing all of
> > the above.  So, it is understandable why the chart has the inaccuracies.
> > Hopefully, you can correct the chart at your convenience.
>
> I would add another point:
>
> (9) underflow, overflow and divide by zero are marked as "silent", while 
> we do have an option (-ffpe-trap=) that ensures that a floating-point 
> exception is raised on these cases (and more: -ffpe-trap= accepts a 
> comma-separated list composed of underflow, overflow, zero, precision 
> and denormal).
>
>
> Thanks,
> FX
Hi Guys

I've changed these entries - and the ones Steve mentioned.  I 'm not 
sure how an ordinary user would discover some of them, and I did 
consider insisting on documentation - but that's not a stated 
requirement, and other compilers don't document everything (though 
generally a bit more than this!)

One thing I did notice is that I got a bounds check error from tfft when 
attempting to time it with checking on.  I think that must be spurious, 
though it's hard to be sure without a traceback.

(I'm away until Friday - so excuse me if my responses are delayed)

--
John Appleyard                         John.Appleyard@polyhedron.com
Polyhedron Software Ltd.
Programs for Programmers - QA, Compilers, Graphics, Consultancy
********* Visit our Web site on http://www.polyhedron.com/ *********