[Met_help] known problem but not solution for pb2nc on a 64-bit machine (UNCLASSIFIED)

[John Halley Gotway]

Barbara,

I'm glad we're making progress.

The 64-bit issue is actually within the BUFRLIB to which MET links, not the PB2NC tool itself.  That error message is coming from BUFRLIB, which is developed and maintained by NCEP.  And we've been in
touch with them regarding this issue, but haven't made much progress.

However, if you were able to build MET using the 32-bit compilation options, we should still be able to make it work.  There's a tool call the "cwordsh" tool which blocks and unblocks PREPBUFR files.
 I'll try not to go into too much detail here, but basically FORTRAN-blocking on a file is a rather machine-dependent thing.  The sample PREPBUFR files we distributed with METv1.1 were already
FORTRAN-blocked on a machine here.  So if we use the "cwordsh" tool to unblock them on your system and then reblock them, hopefully PB2NC will run fine.

These pages of the MET online tutorial discuss FORTRAN-blocking and how you'd acquire and run the cwordsh tool:
http://www.dtcenter.org/met/users/support/online_tutorial/compilation/rec_software.php
http://www.dtcenter.org/met/users/support/online_tutorial/pb2nc/index.php

If you're able, please try unblocking and reblocking the sample PREPBUFR observation files in the METv1.1/data/sample_obs/prepbufr directory.  And then re-run PB2NC to see if that did the trick.

We realize that running the "cwordsh" tool is a rather tedious step.  So for the next version of MET in February 2009, we have modified the PB2NC tool to do the FORTRAN blocking itself.  So hopefully
you won't need to run the "cwordsh" tool when using METv2.0.

Thanks and please let me know how it goes.

John

Sauter, Barbara (Civ, ARL/CISD) wrote:
> Classification:  UNCLASSIFIED 
> Caveats: NONE
> 
> It's me again.  I understand Bob Flanigan used the suggestions you
> provided and he was able to successfully compile pb2nc and the
> point_stat tool on my 64-bit system, specifying 32-bit options.  I ran
> point_stat using the netcdf output you provided for me, and I was very
> excited to finally have that working. 
> 
> However, when I tried to run pb2nc myself I got the exact same error
> message that I had received before:
> 
>      ERROR READING INPUT FILE CONNECTED TO UNIT 11 WHEN CHECKING FOR
> 'BUFR' IN FIRST 4 BYTES OF RECORD
> 
> Bob Flanigan is on vacation until January 5, and as you may have figured
> out I'm not a computer person, so I don't expect any more assistance for
> the time being, but I wanted to let you know that your efforts have
> produced some progress for us but not a total solution yet.  Is there a
> plan to rewrite pb2nc to work more seamlessly on 64-bit systems?
> 
> Thanks for your help.  Have a Merry Christmas and Happy 2009.
> 
> Barb Sauter, meteorologist
> US Army Research Laboratory
> White Sands Missile Range, NM
> (575) 678-2840
> bsauter at arl.army.mil
> 
> 
> Classification:  UNCLASSIFIED 
> Caveats: NONE
>