[Met_help] RE: MET FEEDBACK from John Huddleston

[John Huddleston]

Hi John

Good to work with you again on this; your question is very interesting.

I've upgraded my compiler to the latest gcc, g++, and gfortran 4.4.0 from
the gnu subversion site. The reason was to make everything binary
compatible. I easily built the GrADS 2.0a2 binaries with this new system.

This new compiler is very strict. I have about 20 files from MET that need
additional include files, the pb2nc Fortran files have been modified to
conform to the warnings, and I also had to rewrite the bufrlib for the same
reason.

I'm still having issues and have used gdb to resolve many of the issues.
However, I believe the problem goes deeper than MET or BUFR.  I sent in one
issue to the GCC GNU folks and they confirmed it matched another bug to be
fixed.  I have another one to send to them. This second one directly impacts
the bufrlib software.

Let's say, I am waiting on the GNU GCC folks to fix them. The current 3.4.4
GCC compiler that comes with Cygwin is totally useless with MET.  Plus, it
uses the older g77 and not the newer gfortran.  

All the Cygwin binaries except pb2nc.exe work fine. I think that I may try
the melbufr C library and remove the dependency on the fortran compiler.
The alternative is rewrite bufrlib, move it out of the 1970s, and add true
64 bit precision.

If you have a 64 bit Linux box that you can arrange for a user Id, I will
attempt to rewrite bufrlib.  I currently use a 32 bit Windows XP box.

John Huddleston


-----Original Message-----
From: John Halley Gotway [mailto:johnhg at rap.ucar.edu] 
Sent: Monday, July 21, 2008 1:03 PM
To: jhuddleston at hughes.net; met_help
Subject: Re: MET FEEDBACK from John Huddleston


John,

Thanks for letting us know about that.  I'll change the FILEMXSTRL value to
be 512 to be consistent with the other definition.  So once you set those
values the same, were you able to successfully run 
MET through cygwin?

John Halley-Gotway

MET.FEEDBACK at rap.ucar.edu wrote:
> First Name = John
> Last Name = Huddleston
> Institution = Easy Computer Solutions
> Email = jhuddleston at hughes.net
> MET Feedback = John,
> 
> I finally got around to figuring out why the Windows CYGWIN PB2NC.EXE was
doing a core dump. There is a character boundary issue.
> 
> In your METv1.1 the size of the FORTRAN character file names are
FILEMXSTRL which is defined in src/pb2nc/readpb.prm to be 256. The CC
calling functions use the max_str_len which is defined in
lib/vx_met_util/constants.h as 512.
> 
> John Huddleston, PhD, PE
> formerly with NOAA NCEP