970220: NETCDF2_4R2 and R3 OSF problems

Steve,

> To: address@hidden
> cc: address@hidden
> From: Steve Penton <address@hidden>
> Subject: NETCDF2_4R2 and R3 OSF problems
> Organization: Research Systems, Inc.
> Keywords: 199702200132.SAA04501

In the above message, you wrote:

>       I am trying to resolve a few errors that I get on the OSF1.40
> platform with both the 2_4r2 and 2_4r3 Netcdf libraries. I can build 
> these libraries, and they pass nctest. However, when I call ncvarputg or
> ncvargetg I get a fatal error (presumably from NCgenio) of the form:
> 
>       NCDF_VARPUT: Non-positive stride not allowed
> 
> 
>       Have you or any other users encountered this problem ? The
> code I am using works on the following other platforms,
> Mac, Windows, HPUX, OpenVMS, VMS, IRIX, SUNOS, SOLARIS, SOLARISX86, and
> a few others.
> 
> 
>       Unfortunately, this is an urgent matter as I need to get this
> resolved before we can ship the next release of our applications (IDL,
> ENVI).

netCDF 2.4.3 supports positive strides only.  The functions ncvarputg()
and ncvargetg() check to ensure that the stride values are positive.
Apparently, you're passing in non-positive strides on the OSF1.40
system.  The fact that this happens on the (presumably) Alpha system and
not the others suggests that the problem lies with the `long' type of
the `stride' parameter.  Could it be that the vector that you're passing
in is an `int' vector rather that a `long' one?

Incidentally, the restriction to positive strides is relaxed in
netCDF-3.

Please keep me apprised.

--------
Steve Emmerson   <address@hidden>