[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20020319: netCDF Fortran-90 on UXP/V using "ftr": SIGFPE in nf_test



Frank,

>Date: Tue, 19 Mar 2002 12:04:30 +0100
>From: "Schmitz, Frank" <address@hidden>
>Organization: ?
>To: "Steve Emmerson" <address@hidden>
>Subject: WG: 20020228: netCDF Fortran-90 on UXP/V using "ftr": typeSizes 
>unavailable
>Keywords: 200202220850.g1M8oax11501

The above message contained the following:

> Waiting for an answere!
> The problem is again the same, the C-test is running, the Fortran-test will 
> fail on VPP5000.
> Best Regards,
> Frank

Did you receive the enclosed email from me?

Regards,
Steve Emmerson   <http://www.unidata.ucar.edu>

--- Begin Message ---
Frank,

>Date: Fri, 1 Mar 2002 08:34:41 +0100
>From: "Schmitz, Frank" <address@hidden>
>Organization: ?
>To: "Steve Emmerson" <address@hidden>
>Subject: AW: 20020228: netCDF Fortran-90 on UXP/V using "ftr": typeSizes 
>unavailable
>Keywords: 200202220850.g1M8oax11501

The above message contained the following:

> Thank's,
> the option -Am for the modul definition file should be add.
> Now, the library was created and I've run the C interface tests
> successful, but I've got a floating exception after calling
> ./nf_test -c. After calling 'err = nf_put_att_double(ncid,
> varid(i),ATT_NAME(j,i),ATT_TYPE(j,i),ATT_LEN(j,i), att) in the
> function put_atts the error occurs in the routine ncx_put_int_double.
> Now I'm debugging the problem, but do you have any ideas?

> Best Regards,
> Frank

Some of the more extreme tests of the netCDF package can cause
floating-point exceptions on some systems.  Your system might be one of
them.

I'm surprised that you're seeing a SIGFPE, however, because the program
"nf_test" calls the subroutine "ignorefpe", which is defined in file
"nf_test/fortlib.c", which invokes the signal(3) function to cause
SIGFPE to be ignored. It could be that your signal-handling system
resets to the default behavior upon receiving a signal. This would
be very non-standard behavior and I would complain to the vendor if
it turned out to be true. The work-around would be to modify the
"ignorefpe" function so that a signal-handling function receives the
SIGFPE and reregisters itself for that signal every time.

More on this can be found in the netCDF support-email archive at

    http://www.unidata.ucar.edu/glimpsedocs/ghnetcdf.html

Try entering the search string "SIGFPE;signal".

Regards,
Steve Emmerson   <http://www.unidata.ucar.edu>

--- End Message ---