NAG F95 v5 on Mac Os X

I have a demo license for the new NAGWare Fortran 95 ver 5.0
for Mac Os X. It installs and compiles the suggested supplied
test program. All with no fuss like other commercial Mac products.
However execution fails instantly with an arithmetic exception
before the first write statement. Tech support for free demo
licenses seems to take a back burner.

The vendor tested configuration is listed as Mac Os X 10.3.2. I have
passed that point and have Mac Os X 10.3.5. As well I have Xcode 1.5
which I believe installs a newer GCC (did not pay attention to that
detail at the time).

The suggested invocations are

> $ f95 -o f90_util f90_util.f90
> $ ./f90_util
> *** Arithmetic exception: Floating underflow - aborting
> Abort trap

but not with the exception as f90_util is intended to show of
the various arithmetic model intrinsics. The executable is 39072
bytes. There is a f90_util.crash.log with the usual inscrutable
processor dump information. Some recognizable text includes

Binary Images Description:
f90_util              ./f90_util
dyld                  /usr/lib/dyld
libSystem.B.dylib     /usr/lib/libSystem.B.dylib
libmathCommon.A.dylib /usr/lib/system/libmathCommon.A.dylib

which suggests that execution had started.

Any suggestions on what to check for installation/configuration
blunders? Anyone running on 10.3.5? With Xcode 1.5? With both?

Are there any canned recipes on how to "integrate" NAG's f95
with Xcode?

Gordon Sande <g.sande@worldnet.att.net> writes:

> Any suggestions on what to check for installation/configuration
> blunders? Anyone running on 10.3.5? With Xcode 1.5? With both?

Yep.  Both. NAG has 2 versions for the Mac.  Try the other one.
I forget the details of what each version was claimed to support,
and I think I recall that I found the description of that misleading.
But it sounds like you got the wrong one, so just try the other
one instead of reasoning it out the hard way.

Yes, there are picky differences.. including a dependence on whether
you are on a G5 or a G4.  IIRC, one of the versions is "safe" in that
it will work in all cases, but has poorer performance for sqrt or some
such thing that I just don't care that much about. I just choose the
"safe" one because having the same install work on multiple machines
is more important to me than the peak performance on sqrt (which
just doesn't tend to be a performance driver for me).  My current
machine is a G4, but my wife is getting a G5 next w, and I have
to support people on a mix of systems.

> Are there any canned recipes on how to "integrate" NAG's f95
> with Xcode?

Afraid I never bothered to look. When I got my first OS X box as an
experiment, one of the things that sold me was that I didn't have to
throw out all the tools I was used to using.  Xemacs, make, and the
other tools I'm used to still work just fine.  Perhaps I'll also
learn some new tricks eventually...though I'm awfully cautious
about getting too tied into vendor-specific stuff.

--
Richard Maine                       |  Good judgment comes from experience;
email: my first.last at org.domain  |  experience comes from bad judgment.
org: nasa, domain: gov              |        -- Mark Twain

Richard E Maine wrote:
> Gordon Sande <g.sande@worldnet.att.net> writes:
>
> 
>
>
> Yep.  Both. NAG has 2 versions for the Mac.  Try the other one.
> I forget the details of what each version was claimed to support,
> and I think I recall that I found the description of that misleading.
> But it sounds like you got the wrong one, so just try the other
> one instead of reasoning it out the hard way.
>

I guess the versions are 4.2 and 5.0. I am particularly interested
in 5.0 with it's undefined variable checking. Yet another no checking
compiler it not all that interesting although not having to switch
to another machine across my network is a bonus. Then I just have to
switch between the Mac and the mostly hidden underlying Unix command
line until I get the hang of the Xcode IDE and things are actually
in it.

Tech support was having a day off and eventually got back to me.

The bother is that the shell supplied by Apple has picked up
a bug which leaves faulty floating point states somewhere. One
presumes it will be fixed shortly. The workaround is to set
a flag for "-ieee=all" and ignore the marvelous

"Warning: Floating invalid operand occurred"

that the shell issues after the resulting program has run.

Or maybe NAG needs to modify their pre-execution initialization
sequence. Must make for interesting discussions between NAG and
Apple.

Gordon Sande <g.sande@worldnet.att.net> writes:

> Richard E Maine wrote:
 

>
> I guess the versions are 4.2 and 5.0.

No.  There are 2 version 5.0s.  Just go look at their download
site.  Unless it changed real recently.  They are for different
Mac configurations, not different compiler release levels.

--
Richard Maine                       |  Good judgment comes from experience;
email: my first.last at org.domain  |  experience comes from bad judgment.
org: nasa, domain: gov              |        -- Mark Twain