Build Pnet-CDF for Windows and link against MSMPI

[Wei-keng Liao]

Hi, Kunal

You can use ncmpi_strerror to print the error message. For example,
   printf("Error message : %s\n", ncmpi_strerror(status));

This error means the open mode NC_64BIT_OFFSET is not allowed on a 32bit machine. 
Try remove it and see if it works.

Wei-keng

On Oct 20, 2010, at 5:38 AM, Kunal Rao wrote:

> Hi
> 
>   I had planned to do this in June but lost track of it in between. I came back to doing this few days back. 
> 
>   Now, I have been able to successfully build the parallel netCDF on Windows using mingw compiler with cygwin and link against 
>   MSMPI . ("configure", "make" and "make install" steps were completed successfully).
> 
>   I tried linking this library and the header file with my application which supports parallel netcdf. 
>   The compilation of the application went fine, but while running the application, I am getting problem with ncmpi_create api.
> 
>   there is a call to this function as follows:
> 
>  --------
>   status = ncmpi_create(MPI_COMM_WORLD, local_filename, NC_CLOBBER | NC_64BIT_OFFSET, FILE_INFO_TEMPLATE, file_identifier);
> 
>  --------
>   This returns value as -25 and gives error. ( If proper, it should have returned 0 I guess..)
> 
>   Can you please let me know what this return value means and what could be going wrong ?
>   Also, how do I verify if the library built is proper ?
> 
> Thanks & Regards,
> Kunal
> 
>   
> 
> On Fri, Jun 4, 2010 at 3:14 PM, Rob Latham <robl at mcs.anl.gov> wrote:
> On Fri, Jun 04, 2010 at 03:10:53PM -0400, kunal rao wrote:
> > Hi,
> >
> > I would like to build Pnet-CDF for Windows and link against MSMPI. Can you
> > please tell me the steps on how to do that?
> 
> You are in for something of an adventure.  I don't have any windows
> experience, but maybe someone on the list does.
> 
> I think if you used cygwin you might get pretty far, but I don't know
> how cygwin libraries interact with something like MSMPI.
> 
> I'll be more than happy to add a README.windows to the distribution if
> you figure it out :>
> 
> ==rob
> 
> --
> Rob Latham
> Mathematics and Computer Science Division
> Argonne National Lab, IL USA
>