[Met_help] Problem with Sample Test
John Halley Gotway
johnhg at rap.ucar.edu
Tue Apr 8 10:43:02 MDT 2008
Segayle,
I apologize for not getting back to you sooner. I had been working on a response to your email but got pulled away to work on some urgent things.
You last sent me the NetCDF output files from the pcp-combine tool. When I look at them on my machine, the contents of them look completely fine. So there must be some issue with how NetCDF is
installed on your machine.
Throughout your installation process, I'd suggest using only the GNU compilers: gcc (C compiler), g++ (C++ compiler), and g77 (fortran-77 compiler). This may be frustrating, but perhaps we could try
starting from scratch and rebuilding BUFRLIB, NetCDF, and MET.
To compile BUFRLIB:
/usr/bin/gcc -c -DUNDERSCORE *.c
/usr/bin/g77 -c -DUNDERSCORE -fno-second-underscore *.f
ar crv libbufr.a *.o
To compile NetCDF (since we don't actually use an Fortran things in NetCDF, don't set FC):
set CC = /usr/bin/gcc
set CXX = /usr/bin/g++
./configure --prefix=/home/swalford/netcdf-3.6.2
make check
make install
To compile MET, in the Makefile set:
CPP_COMPILER = /usr/bin/g++
F77_COMPILER = /usr/bin/g77
Then in the top-level MET directory, do a:
make clean
make install
And try rerunning the test scripts.
If, after doing this, you're still getting the runtime NetCDF error, then I'm completely stumped! Please let me know how it goes. And I'm sorry again for the delay in getting back to you.
John
walford02 at hotmail.com wrote:
> John,
>
> The only other thing I think it could be is that I use
>
> gcc -c -DUNSERSCORE *.c
> f77 -c -DUNSERSCORE -fno-second-underscore *.f
> ar crv libbufr.a *.o
>
> to compile BURFLIB
>
> but in the Makefile I use
>
> # Path to the C++ Compiler
> CPP_COMPILER=/usr/bin/g++
>
> # Path to the Fortran-77 Compiler
> F77_COMPILER=/usr/bin/g77
>
> should I use
> # Path to the C++ Compiler
> CPP_COMPILER=/usr/bin/gcc
>
> # Path to the Fortran-77 Compiler
> F77_COMPILER=/usr/bin/f77
>
> My NetCDF configure file looks like this:
>
> #!/bin/sh
> CC=/usr/bin/gcc; export CC
> CPP='gcc -E'; export CPP
> CPPFLAGS='-DNDEBUG -DpgiFortran'; export CPPFLAGS
> CFLAGS='-O -w'; export CFLAGS
> FC=/usr/local/pgi/linux86/6.1/bin/pgf90; export FC
> FFLAGS=''; export FFLAGS
> CXX=/usr/bin/g++; export CXX
> CXXCPPFLAGS=$CPPFLAGS; export CXXCPPFLAGS
> ./configure\
> --prefix=/home/swalford/netcdf-3.6.2
>
> should I make my
> FC=/usr/local/pgi/linux86/6.1/bin/pgf77; export FC
>
> Attached are the files you requested.
> Segayle
>
> --------------------------------------------------
> From: "John Halley Gotway" <johnhg at rap.ucar.edu>
> Sent: Tuesday, April 01, 2008 3:56 PM
> To: <walford02 at hotmail.com>
> Cc: "met_help" <met_help at ucar.edu>
> Subject: Re: [Met_help] problems compiling MET
>
>> Segayle,
>>
>> I'd have to agree with you - something definitely doesn't look right.
>>
>> Could you please send me the following two files:
>> METv1.0/out/pcp_combine/sample_fcst_12L_2005080712V_12A.nc
>> METv1.0/out/pcp_combine/sample_obs_2005080712V_12A.nc
>>
>> Those were built by the PCP-Combine tool in the test test script and
>> are used as input into MODE.
>>
>> I'll take a look at their contents and see if there's a problem in
>> them. Could be a NetCDF-related issue.
>>
>> Thanks,
>> John
>>
>> walford02 at hotmail.com wrote:
>>>
>>> I forgot to attach the test_all.log file... Here it is ..
>>>
>>> Segayle
>>> --------------------------------------------------
>>> From: <walford02 at hotmail.com>
>>> Sent: Tuesday, April 01, 2008 3:29 PM
>>> To: "John Halley Gotway" <johnhg at rap.ucar.edu>
>>> Subject: Re: [Met_help] problems compiling MET
>>>
>>>> John,
>>>>
>>>> Thank you for the reply. I have tried the PGI compilers and I seem
>>>> to get many errors so I am going to compile using GNU.
>>>>
>>>> I did get a successful compilation after I recompiled the BUFR and
>>>> NetCDF using GNU compilers. However, I still believe that something
>>>> is not correct because when I tried to run the sample test case using
>>>>
>>>> test_all.sh >& test_all.log&
>>>>
>>>> When testing the Mode application I get an error that looks like:
>>>>
>>>> ERROR: read_pcp_combine_netcdf() -> Projection type ^Xû~^Hection
>>>> not currently supported.
>>>>
>>>> ERROR: read_pcp_combine_netcdf() -> Projection type X^Q}^Hection not
>>>> currently supported.
>>>>
>>>> I believe either the Mode application did not build correctly or
>>>> there was a problem in the sample test data. Are there any thoughts
>>>> as to why I get this error?
>>>>
>>>>
>>>> Thank you
>>>> Segayle
>>>>
>>>> --------------------------------------------------
>>>> From: "John Halley Gotway" <johnhg at rap.ucar.edu>
>>>> Sent: Tuesday, April 01, 2008 11:45 AM
>>>> To: "Segayle Thompson" <walford02 at hotmail.com>; "met_help"
>>>> <met_help at ucar.edu>
>>>> Subject: Re: [Met_help] problems compiling MET
>>>>
>>>>> Segayle,
>>>>>
>>>>> I'd suggest just trying to get MET to compile with a single family
>>>>> of compilers... either, the GNU compilers or the PGI ones.
>>>>>
>>>>> The issue is that BUFRLIB and the NETCDF library need to be
>>>>> compiled with whichever you choose. Do you know how those were
>>>>> compiled? With the PGI compilers or the GNU compilers?
>>>>>
>>>>> Here are comments relating to PGI - if you decide that's what you'd
>>>>> like to use:
>>>>>
>>>>> First, looking at your PGI Makefile and Logfile, the problem here
>>>>> is that the linker can't find the NetCDF library archive files it's
>>>>> looking for. Please look for the files:
>>>>> /home/swalford/netcdf-3.6.2/lib/libnetcdf.a and
>>>>> /home/swalford/netcdf-3.6.2/lib/libnetcdf_c++.a. If they don't
>>>>> exist, then that's problem, and I'd suggest rebuilding NetCDF. But
>>>>> when you do, make sure to pass the --prefix option to the configure
>>>>> script. And let's set some environment variables to ensure that
>>>>> NetCDF will be compiled with the PGI compilers.
>>>>>
>>>>> For example...
>>>>> setenv CC /usr/local/pgi/linux86/6.1/bin/pgcc
>>>>> setenv FC /usr/local/pgi/linux86/6.1/bin/pgf77
>>>>> setenv F77 /usr/local/pgi/linux86/6.1/bin/pgf77
>>>>> setenv CXX /usr/local/pgi/linux86/6.1/bin/pgCC
>>>>> cd /home/swalford/netcdf-3.6.2
>>>>> configure --prefix=/home/swalford/netcdf-3.6.2
>>>>> make check
>>>>> make install
>>>>>
>>>>> That indicates that NetCDF should be compiled with the PGI
>>>>> compilers and "installed" in the /home/swalford/netcdf-3.6.2
>>>>> directory. Once you do that, those library files should be placed
>>>>> in the "lib" directory. So check again, and make sure
>>>>> /home/swalford/netcdf-3.6.2/lib contains the files libnetcdf.a and
>>>>> libnetcdf_c++.a.
>>>>>
>>>>> If you're going to use PGI, you'll also need to make sure that
>>>>> BUFRLIB was compiled with the PGI compilers.
>>>>>
>>>>> Here are comments relating to GNU - if you decide that's what you'd
>>>>> like to use:
>>>>>
>>>>> The problem here is that you're using the wrong compilers. In your
>>>>> Makefile for GNU, try setting the CPP_COMPILER and F77_COMPILER as
>>>>> follows:
>>>>>
>>>>> # Path to the C++ Compiler
>>>>> CPP_COMPILER=/usr/bin/g++
>>>>>
>>>>> # Path to the Fortran-77 Compiler
>>>>> F77_COMPILER=/usr/bin/g77
>>>>>
>>>>> Please make sure that /usr/bin/g77 exists. If not, try
>>>>> /usr/bin/gfortran instead. But let me know if you're using
>>>>> gfortran because it'll require a couple of minor changes to get it
>>>>> to work.
>>>>>
>>>>> And again, if using the GNU compilers, you'll need to make sure
>>>>> that NetCDF and BUFRLIB are compiled with them as well.
>>>>>
>>>>> Thanks and good luck,
>>>>> John
>>>>>
>>>>> Segayle Thompson wrote:
>>>>>> Hello John,
>>>>>>
>>>>>> I have tried a number of things. I tried compiling with the
>>>>>> regular Makefile changing the paths make >& make.log I have
>>>>>> attached the makefile and the output ... (make.log and the Makefile )
>>>>>>
>>>>>> Not sure what the problems are here. Then I tried the Makefile_pgi
>>>>>> changing it to the Makefile (as suggested) and running make >&
>>>>>> make_pgi.log I have attached the makefile and the output ...
>>>>>> (make_pgi.log and Makefile_pgi) I am not sure what these messages
>>>>>> mean. I do notice that when I run the pgi version I get
>>>>>> significantly more warnings and some 'breaks' (which I am not sure
>>>>>> what they mean). Segayle
>>>>>> ----- Original Message ----- From: John Halley
>>>>>> Gotway<mailto:johnhg at rap.ucar.edu> To:
>>>>>> walford02 at hotmail.com<mailto:walford02 at hotmail.com> ;
>>>>>> met_help<mailto:met_help at ucar.edu> Sent: Monday, March 31, 2008
>>>>>> 4:55 PM
>>>>>> Subject: Re: [Met_help] problems compiling MET
>>>>>>
>>>>>>
>>>>>> Thomas,
>>>>>>
>>>>>> I just saw your second email.
>>>>>>
>>>>>> I'm guessing that you have a couple of errors in your Makefile
>>>>>> that are throwing things off. In particular, please look in your
>>>>>> Makefile for the following...
>>>>>> (1) Remove any trailing spaces, especially in the lines for
>>>>>> NETCDF_BASE and BUFR_BASE.
>>>>>> (2) Make sure that the line for GSL_INCS begins with a -I, for
>>>>>> example: GSL_INCS = /my/GSL/directory/include
>>>>>> (3) Make sure that the line for GSL_LIBS begins with a -L, for
>>>>>> example: GSL_LIBS = /my/GSL/directory/lib
>>>>>>
>>>>>> Please let me know if that does the trick or if you encounter
>>>>>> additional problems.
>>>>>>
>>>>>> Thanks,
>>>>>> John Halley-Gotway
>>>>>> johnhg at ucar.edu<mailto:johnhg at ucar.edu>
>>>>>>
>>>>>> walford02 at hotmail.com<mailto:walford02 at hotmail.com> wrote:
>>>>>> > Hello,
>>>>>> > > I have been having problems compiling the MET v1.0 code. I
>>>>>> have implemented all of the 'known fixes" but I continue to get
>>>>>> this error when compiling that is related to the
>>>>>> vx_contable_stats.o file. I have enclosed the make.log file.
>>>>>> Please let me know if there is something wrong with any of these
>>>>>> files. > > Thank you > Segayle Walford
>>>>>> > > > >
>>>>>> ------------------------------------------------------------------------
>>>>>>
>>>>>>
>>>>>> > > _______________________________________________
>>>>>> > Met_help mailing list
>>>>>> > Met_help at mailman.ucar.edu<mailto:Met_help at mailman.ucar.edu>
>>>>>> >
>>>>>> http://mailman.ucar.edu/mailman/listinfo/met_help<http://mailman.ucar.edu/mailman/listinfo/met_help>
>>>>>>
>>>>>>
>>>>>
>>
More information about the Met_help
mailing list