[Wrf-users] Problem with WRFPOST on Linux

[Jaakko Hyvätti]

Hi,

I solved the same problem with pgroup compiler in some code that 
included similar timef() function definition.  Replace the return value 
definition of real*8 with just real in TIMEF.f.  That is, remove the *8 .
Hope this helps.

Jaakko Hyvätti
Foreca Ltd 
Jaakko.Hyvatti at foreca.com
Tammasaarenkatu 5, FI-00180 Helsinki, Finland
http://www.foreca.com


On Tue, 14 Nov 2006, Eric L. Altshuler wrote:
> I have been able to find out a bit more about my problem with WRFPOST on
> Linux. Following a suggestion from Rorik Peterson, I used the debugger
> pgdbg to analyze the program. The floating point exception is happening
> at line 407 in the WRFPOST.f routine (main program) and the problem
> seems to be with the result being returned from the timef() function
> (TIMEF.f, which in turn calls an intrinsic function 'etime'). I inserted
> a line immediately above line 407:
>
>      print *, timef()
>
> and the values printed seemed crazy. They had magnitudes of 1.E+35 to
> 1.E+37, although the exact values differed in successive executions of
> the program. I suppose, as Laurie Carson reported, that it's possible
> for some values to cause an FPE and others not, so the program may not
> always crash. I have no idea why the timef() routine is behaving this
> way.