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

Re: 970729: netcdf 3.3 on Digital Unix

>To: address@hidden
>From: "Robert Shectman" <address@hidden>
>Subject: netcdf 3.3 on Digital Unix
>Organization: .
>Keywords: 199707291836.MAA12623

Hi Robert,

> We are having problems with netcdf3.3 and 3.3.1 at least on Digital Unix.
> At its simplest, ncdump cannot read a file produced by ncgen. It claims
> its not a netcdf file.  Things seem to work much better on sgi and solaris
> versions.
> 
> I have spent a small amount of time trying to track down the error through
> the C++ interface, but must admit, mainly what I've done is get lost in
> following the open.  But I've only spent about 20 minutes on it so far.
> 
> Here is the relevant info about compilers and OS rev levels, and
> attached is a short cdl that triggers the problem.
> 
> (frosty) src >uname -a
> OSF1 frosty.llnl.gov V4.0 564 alpha
> 
> (frosty) src >cxx -V
> cxx  (cxx)
> DEC C++ V5.5-004 on Digital UNIX (Alpha)
> 
> (frosty) src >c89 -V
> c89  (c89)
>         Digital UNIX Compiler Driver 3.11
> DEC C V5.2-036 on Digital UNIX V4.0 (Rev. 564)

We have exactly the same versions of the OS and all the compilers here.
I'm not sure the version of the C++ compiler or anything in the C++
interface is relevant, since ncdump and ncgen both use the C interface
exclusively.

> netcdf avn{           // 126 Wave, 18 Layer Spectral Model Aviation Run
>                       // on international exchange grids 21, 22, 23, 24
> 
> dimensions:
> 
>       record =        UNLIMITED ;     // (reference time, forecast time)
> 
>       // Vertical Coordinate Dimensions
>       vert =  16 ;           // AVN has 16 pressure levels
> 
>       y =     181 ;           // latitude
>       x =     360 ;           // longitude
> 
> variables:
> 
> 
> 
> 
> ////////////////////////////////////////////////////////////////
> // Values based on the Primary Vertical Coordinate (Pressure)
> ////////////////////////////////////////////////////////////////
> 
> 
>       float   z(vert, y, x) ;
>               z:long_name =                           "Height
>                                  " ;
>               z:units =                               "m
>                                       " ;
>               z:x_grid_geometry =                     "grid_point
>                              " ;
>               z:y_grid_geometry =                     "grid_point
>                              " ;
>               z:vertical_grid_geometry =              "grid_point
>                              " ;
>               z:invalid_data =                        "true
>                                    " ;
>               z:_FillValue =                          -99999.f ;
> 
> 
> }

I've just stored the above in a file named "avn.cdl" and then tried the
netCDF 3.3.1 versions of ncgen and ncdump on this file, as follows:

  $ ncgen -b avn.cdl
  $ ls -l avn.nc
  -rw-rw-r--   1 russ     usystem  4170796 Jul 29 13:51 avn.nc
  $ ncdump avn.nc > avn+.cdl
  $ ls -l avn+.cdl
  -rw-rw-r--   1 russ     usystem  3336812 Jul 29 13:52 avn+.cdl
  $ ncgen -b -o avn+.nc avn+.cdl
  $ ls -l avn+.nc
  -rw-rw-r--   1 russ     usystem  4170796 Jul 29 13:53 avn+.nc
  $ cmp avn.nc avn+.nc

and the latter comparison of the two netCDF files created by ncgen
showed they were identical byte-for-byte.

So I can't reproduce the problem.  Before running the "configure" script
to build ncdump and ncgen from the netCDF 3.3.1 sources, the only
environment variables I set were CXX=g++ and CFLAGS=-g, but I don't
think these are relevant.  The "make all test" worked fine and all tests
were successful.

Can you try the above sequence of ncgen and ncdump commands and let us
know if ncdump still doesn't recognize avn.nc as a netCDF file?

If this doesn't demonstrate the problem, do you have have some other
symptom that we might be able to use here to diagnose the problem?  If
you can't come up with another way to demonstrate the bug, it's possible
we could spot a difference by getting from you

 - the complete output of running the "configure" script
 - config.status
 - the output from running "make all"
 - the output from running "make test"

Thanks.

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu