[Met_help] [rt.rap.ucar.edu #44711] History for Met building problems: compilation aborted for vx_contable_stats.cc

RAL HelpDesk {for Paul Oldenburg} met_help at ucar.edu
Wed Feb 23 14:45:59 MST 2011


----------------------------------------------------------------
  Initial Request
----------------------------------------------------------------

Hi,

I open another post, but it could be a continuation about
my previous ticket #44675.

In order to build MET on my Linux Intel machine (64bit),
into ../netcdf../include I did as following:
cp netcdf.h netcdf.hh.
In this way I resolved my ".hh" problem.

I also built succesfully GSL as well.

*Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so on...) is built by
Intel compiler.*

But when I "make", I get the errors in attachment.
In attachment the makefile too.

Where do I wrong, now?

Thank you veru much,
Fabio.


----------------------------------------------------------------
  Complete Ticket History
----------------------------------------------------------------

Subject: Re: [rt.rap.ucar.edu #44711] Met building problems: compilation aborted for vx_contable_stats.cc
From: Paul Oldenburg
Time: Tue Feb 22 12:34:38 2011

Fabio,

Your solution of copying the NetCDF header file is causing the
problem.  The file include/netcdf.hh is generated when
the c/c++ interface of NetCDF is built.  The instance you are trying
to use may not have that interface built.

We recommend that you download and build your own instances of the
libraries BUFRLIB, NetCDF and GSL.  This will ensure
that they are built in the way that MET expects.  The process should
be pretty easy if you carefully read and follow the
instructions at the online tutorial I mentioned earlier, using the
intel compilers for each step:

http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php

You may see some error messages when building BUFRLIB when executing
this step:

ifort -c -DUNDERSCORE *.f *.F

Ignore these error messages and continue with the process.  MET should
work with the partially built version of BUFRLIB.
 Please let us know if you have any other questions.

Paul


On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>
> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
> Transaction: Ticket created by fabio.gervasi70 at gmail.com
>        Queue: met_help
>      Subject: Met building problems: compilation aborted for
vx_contable_stats.cc
>        Owner: Nobody
>   Requestors: fabio.gervasi70 at gmail.com
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>
>
> Hi,
>
> I open another post, but it could be a continuation about
> my previous ticket #44675.
>
> In order to build MET on my Linux Intel machine (64bit),
> into ../netcdf../include I did as following:
> cp netcdf.h netcdf.hh.
> In this way I resolved my ".hh" problem.
>
> I also built succesfully GSL as well.
>
> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so on...) is
built by
> Intel compiler.*
>
> But when I "make", I get the errors in attachment.
> In attachment the makefile too.
>
> Where do I wrong, now?
>
> Thank you veru much,
> Fabio.
>


------------------------------------------------
Subject: Met building problems: compilation aborted for vx_contable_stats.cc
From: Fabio F.Gervasi
Time: Tue Feb 22 13:13:49 2011

Ok Paul,

I removed my netcdf.hh copy.

I'm following the tutorial and I understand about BUFRLIB, NetCDF and
GSL.
Regarding BUFRLIB I'll ignore the ifort errors. GSL seems built ok.

*About NetCDF, I understand that MET wants ".hh" but I think (but
maybe I
wrong) I compiled it correctly. My env settings are:*
export CC=icc
export FC=ifort
export F9X=ifort
export F90=ifort
export F77=ifort
export CXX=icc

*Then I do:*
./configure --prefix=/home/master/netcdf-3.6.2
make
make install

But *my building process doesn't produce any .hh file into /include
folder,
but only a .h file*.
I built NetCDF several times in the past, and I never had any problems
so
far with it, because WPS, WRF, WRFDA and so on are okay.

Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx folder, so I
tried
to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
on my makefile, but now I get:
/home/master/netcdf-3.6.2/cxx/ncvalues.h(16): catastrophic error:
could not
open source file "netcdf.h"
  #include "netcdf.h"

Sorry Paul, maybe I'm doing some stupid errors, but really I don't
understand what's wrong on my NetCDF building!

Thank you, Paul, for your patience,
Fabio.



2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>

> Fabio,
>
> Your solution of copying the NetCDF header file is causing the
problem.
>  The file include/netcdf.hh is generated when
> the c/c++ interface of NetCDF is built.  The instance you are trying
to use
> may not have that interface built.
>
> We recommend that you download and build your own instances of the
> libraries BUFRLIB, NetCDF and GSL.  This will ensure
> that they are built in the way that MET expects.  The process should
be
> pretty easy if you carefully read and follow the
> instructions at the online tutorial I mentioned earlier, using the
intel
> compilers for each step:
>
>
>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
>
> You may see some error messages when building BUFRLIB when executing
this
> step:
>
> ifort -c -DUNDERSCORE *.f *.F
>
> Ignore these error messages and continue with the process.  MET
should work
> with the partially built version of BUFRLIB.
>  Please let us know if you have any other questions.
>
> Paul
>
>
> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
> >
> > Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
> > Transaction: Ticket created by fabio.gervasi70 at gmail.com
> >        Queue: met_help
> >      Subject: Met building problems: compilation aborted for
> vx_contable_stats.cc
> >        Owner: Nobody
> >   Requestors: fabio.gervasi70 at gmail.com
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >
> >
> > Hi,
> >
> > I open another post, but it could be a continuation about
> > my previous ticket #44675.
> >
> > In order to build MET on my Linux Intel machine (64bit),
> > into ../netcdf../include I did as following:
> > cp netcdf.h netcdf.hh.
> > In this way I resolved my ".hh" problem.
> >
> > I also built succesfully GSL as well.
> >
> > *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so on...) is
built by
> > Intel compiler.*
> >
> > But when I "make", I get the errors in attachment.
> > In attachment the makefile too.
> >
> > Where do I wrong, now?
> >
> > Thank you veru much,
> > Fabio.
> >
>
>
>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #44711] Met building problems: compilation aborted for vx_contable_stats.cc
From: Paul Oldenburg
Time: Tue Feb 22 13:40:28 2011

Fabio,

The final step in the NetCDF should basically copy cxx/netcdf.hh to
include/netcdf.hh.  We suspect that there is
something funny going on with your configure and build process.  Can
you please send us log files of your configuration
and build using the following commands:

$ make clean
$ ./configure --prefix=/netcdf/install/path &> nc_configure.log
$ make &> nc_make.log
$ make install &> nc_make_install.log

Thanks,

Paul


On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>
> Ok Paul,
>
> I removed my netcdf.hh copy.
>
> I'm following the tutorial and I understand about BUFRLIB, NetCDF
and GSL.
> Regarding BUFRLIB I'll ignore the ifort errors. GSL seems built ok.
>
> *About NetCDF, I understand that MET wants ".hh" but I think (but
maybe I
> wrong) I compiled it correctly. My env settings are:*
> export CC=icc
> export FC=ifort
> export F9X=ifort
> export F90=ifort
> export F77=ifort
> export CXX=icc
>
> *Then I do:*
> ./configure --prefix=/home/master/netcdf-3.6.2
> make
> make install
>
> But *my building process doesn't produce any .hh file into /include
folder,
> but only a .h file*.
> I built NetCDF several times in the past, and I never had any
problems so
> far with it, because WPS, WRF, WRFDA and so on are okay.
>
> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx folder, so I
tried
> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
> on my makefile, but now I get:
> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16): catastrophic error:
could not
> open source file "netcdf.h"
>   #include "netcdf.h"
>
> Sorry Paul, maybe I'm doing some stupid errors, but really I don't
> understand what's wrong on my NetCDF building!
>
> Thank you, Paul, for your patience,
> Fabio.
>
>
>
> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>
>> Fabio,
>>
>> Your solution of copying the NetCDF header file is causing the
problem.
>>  The file include/netcdf.hh is generated when
>> the c/c++ interface of NetCDF is built.  The instance you are
trying to use
>> may not have that interface built.
>>
>> We recommend that you download and build your own instances of the
>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
>> that they are built in the way that MET expects.  The process
should be
>> pretty easy if you carefully read and follow the
>> instructions at the online tutorial I mentioned earlier, using the
intel
>> compilers for each step:
>>
>>
>>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
>>
>> You may see some error messages when building BUFRLIB when
executing this
>> step:
>>
>> ifort -c -DUNDERSCORE *.f *.F
>>
>> Ignore these error messages and continue with the process.  MET
should work
>> with the partially built version of BUFRLIB.
>>  Please let us know if you have any other questions.
>>
>> Paul
>>
>>
>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>>>
>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
>>> Transaction: Ticket created by fabio.gervasi70 at gmail.com
>>>        Queue: met_help
>>>      Subject: Met building problems: compilation aborted for
>> vx_contable_stats.cc
>>>        Owner: Nobody
>>>   Requestors: fabio.gervasi70 at gmail.com
>>>       Status: new
>>>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>
>>>
>>> Hi,
>>>
>>> I open another post, but it could be a continuation about
>>> my previous ticket #44675.
>>>
>>> In order to build MET on my Linux Intel machine (64bit),
>>> into ../netcdf../include I did as following:
>>> cp netcdf.h netcdf.hh.
>>> In this way I resolved my ".hh" problem.
>>>
>>> I also built succesfully GSL as well.
>>>
>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so on...) is
built by
>>> Intel compiler.*
>>>
>>> But when I "make", I get the errors in attachment.
>>> In attachment the makefile too.
>>>
>>> Where do I wrong, now?
>>>
>>> Thank you veru much,
>>> Fabio.
>>>
>>
>>
>>


------------------------------------------------
Subject: Met building problems: compilation aborted for vx_contable_stats.cc
From: Fabio F.Gervasi
Time: Tue Feb 22 14:00:07 2011

I'm sorry Paul, I made a mistake:

I remembered that when I compile netcdf, *I usually set CXX='',
otherwhise I
obtain an
abort compilation, as you can see in attachment.*
Again, now I remember had a similar problem with gcc (about 2 years
ago!)

Apparently, for all the other programs do not set CXX is not a
problem.
I never solved that issue that now back to me forcefully!

How can I fix it, in your opinion?

Thanks,
Fabio.


2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>

> Fabio,
>
> The final step in the NetCDF should basically copy cxx/netcdf.hh to
> include/netcdf.hh.  We suspect that there is
> something funny going on with your configure and build process.  Can
you
> please send us log files of your configuration
> and build using the following commands:
>
> $ make clean
> $ ./configure --prefix=/netcdf/install/path &> nc_configure.log
> $ make &> nc_make.log
> $ make install &> nc_make_install.log
>
> Thanks,
>
> Paul
>
>
> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >
> > Ok Paul,
> >
> > I removed my netcdf.hh copy.
> >
> > I'm following the tutorial and I understand about BUFRLIB, NetCDF
and
> GSL.
> > Regarding BUFRLIB I'll ignore the ifort errors. GSL seems built
ok.
> >
> > *About NetCDF, I understand that MET wants ".hh" but I think (but
maybe I
> > wrong) I compiled it correctly. My env settings are:*
> > export CC=icc
> > export FC=ifort
> > export F9X=ifort
> > export F90=ifort
> > export F77=ifort
> > export CXX=icc
> >
> > *Then I do:*
> > ./configure --prefix=/home/master/netcdf-3.6.2
> > make
> > make install
> >
> > But *my building process doesn't produce any .hh file into
/include
> folder,
> > but only a .h file*.
> > I built NetCDF several times in the past, and I never had any
problems so
> > far with it, because WPS, WRF, WRFDA and so on are okay.
> >
> > Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx folder, so
I
> tried
> > to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
> > on my makefile, but now I get:
> > /home/master/netcdf-3.6.2/cxx/ncvalues.h(16): catastrophic error:
could
> not
> > open source file "netcdf.h"
> >   #include "netcdf.h"
> >
> > Sorry Paul, maybe I'm doing some stupid errors, but really I don't
> > understand what's wrong on my NetCDF building!
> >
> > Thank you, Paul, for your patience,
> > Fabio.
> >
> >
> >
> > 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
> >
> >> Fabio,
> >>
> >> Your solution of copying the NetCDF header file is causing the
problem.
> >>  The file include/netcdf.hh is generated when
> >> the c/c++ interface of NetCDF is built.  The instance you are
trying to
> use
> >> may not have that interface built.
> >>
> >> We recommend that you download and build your own instances of
the
> >> libraries BUFRLIB, NetCDF and GSL.  This will ensure
> >> that they are built in the way that MET expects.  The process
should be
> >> pretty easy if you carefully read and follow the
> >> instructions at the online tutorial I mentioned earlier, using
the intel
> >> compilers for each step:
> >>
> >>
> >>
>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
> >>
> >> You may see some error messages when building BUFRLIB when
executing
> this
> >> step:
> >>
> >> ifort -c -DUNDERSCORE *.f *.F
> >>
> >> Ignore these error messages and continue with the process.  MET
should
> work
> >> with the partially built version of BUFRLIB.
> >>  Please let us know if you have any other questions.
> >>
> >> Paul
> >>
> >>
> >> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
> >>>
> >>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
> >>> Transaction: Ticket created by fabio.gervasi70 at gmail.com
> >>>        Queue: met_help
> >>>      Subject: Met building problems: compilation aborted for
> >> vx_contable_stats.cc
> >>>        Owner: Nobody
> >>>   Requestors: fabio.gervasi70 at gmail.com
> >>>       Status: new
> >>>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
> >>>
> >>>
> >>> Hi,
> >>>
> >>> I open another post, but it could be a continuation about
> >>> my previous ticket #44675.
> >>>
> >>> In order to build MET on my Linux Intel machine (64bit),
> >>> into ../netcdf../include I did as following:
> >>> cp netcdf.h netcdf.hh.
> >>> In this way I resolved my ".hh" problem.
> >>>
> >>> I also built succesfully GSL as well.
> >>>
> >>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so on...) is
built
> by
> >>> Intel compiler.*
> >>>
> >>> But when I "make", I get the errors in attachment.
> >>> In attachment the makefile too.
> >>>
> >>> Where do I wrong, now?
> >>>
> >>> Thank you veru much,
> >>> Fabio.
> >>>
> >>
> >>
> >>
>
>
>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #44711] Met building problems: compilation aborted for vx_contable_stats.cc
From: Paul Oldenburg
Time: Tue Feb 22 14:11:17 2011

Fabio,

It looks like your c++ compiler is not configured correctly.  The
error message indicates that your compiler cannot find
a standard c++ library function (strncpy).  I think you may want to
talk to your administrator about the compiler
configuration.  This will be a problem when you try to compile MET,
too.

ncvalues.cpp(238): error: identifier "strncpy" is undefined
      strncpy(s, (const char*)the_values + n, (int)the_number);

We would like to provide in-person support to you in Rome.  Would your
operations budget support that?

Paul


On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>
> I'm sorry Paul, I made a mistake:
>
> I remembered that when I compile netcdf, *I usually set CXX='',
otherwhise I
> obtain an
> abort compilation, as you can see in attachment.*
> Again, now I remember had a similar problem with gcc (about 2 years
ago!)
>
> Apparently, for all the other programs do not set CXX is not a
problem.
> I never solved that issue that now back to me forcefully!
>
> How can I fix it, in your opinion?
>
> Thanks,
> Fabio.
>
>
> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>
>> Fabio,
>>
>> The final step in the NetCDF should basically copy cxx/netcdf.hh to
>> include/netcdf.hh.  We suspect that there is
>> something funny going on with your configure and build process.
Can you
>> please send us log files of your configuration
>> and build using the following commands:
>>
>> $ make clean
>> $ ./configure --prefix=/netcdf/install/path &> nc_configure.log
>> $ make &> nc_make.log
>> $ make install &> nc_make_install.log
>>
>> Thanks,
>>
>> Paul
>>
>>
>> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>
>>> Ok Paul,
>>>
>>> I removed my netcdf.hh copy.
>>>
>>> I'm following the tutorial and I understand about BUFRLIB, NetCDF
and
>> GSL.
>>> Regarding BUFRLIB I'll ignore the ifort errors. GSL seems built
ok.
>>>
>>> *About NetCDF, I understand that MET wants ".hh" but I think (but
maybe I
>>> wrong) I compiled it correctly. My env settings are:*
>>> export CC=icc
>>> export FC=ifort
>>> export F9X=ifort
>>> export F90=ifort
>>> export F77=ifort
>>> export CXX=icc
>>>
>>> *Then I do:*
>>> ./configure --prefix=/home/master/netcdf-3.6.2
>>> make
>>> make install
>>>
>>> But *my building process doesn't produce any .hh file into
/include
>> folder,
>>> but only a .h file*.
>>> I built NetCDF several times in the past, and I never had any
problems so
>>> far with it, because WPS, WRF, WRFDA and so on are okay.
>>>
>>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx folder, so
I
>> tried
>>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
>>> on my makefile, but now I get:
>>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16): catastrophic error:
could
>> not
>>> open source file "netcdf.h"
>>>   #include "netcdf.h"
>>>
>>> Sorry Paul, maybe I'm doing some stupid errors, but really I don't
>>> understand what's wrong on my NetCDF building!
>>>
>>> Thank you, Paul, for your patience,
>>> Fabio.
>>>
>>>
>>>
>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>>>
>>>> Fabio,
>>>>
>>>> Your solution of copying the NetCDF header file is causing the
problem.
>>>>  The file include/netcdf.hh is generated when
>>>> the c/c++ interface of NetCDF is built.  The instance you are
trying to
>> use
>>>> may not have that interface built.
>>>>
>>>> We recommend that you download and build your own instances of
the
>>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
>>>> that they are built in the way that MET expects.  The process
should be
>>>> pretty easy if you carefully read and follow the
>>>> instructions at the online tutorial I mentioned earlier, using
the intel
>>>> compilers for each step:
>>>>
>>>>
>>>>
>>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
>>>>
>>>> You may see some error messages when building BUFRLIB when
executing
>> this
>>>> step:
>>>>
>>>> ifort -c -DUNDERSCORE *.f *.F
>>>>
>>>> Ignore these error messages and continue with the process.  MET
should
>> work
>>>> with the partially built version of BUFRLIB.
>>>>  Please let us know if you have any other questions.
>>>>
>>>> Paul
>>>>
>>>>
>>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>>>>>
>>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
>>>>> Transaction: Ticket created by fabio.gervasi70 at gmail.com
>>>>>        Queue: met_help
>>>>>      Subject: Met building problems: compilation aborted for
>>>> vx_contable_stats.cc
>>>>>        Owner: Nobody
>>>>>   Requestors: fabio.gervasi70 at gmail.com
>>>>>       Status: new
>>>>>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
>>>>>
>>>>>
>>>>> Hi,
>>>>>
>>>>> I open another post, but it could be a continuation about
>>>>> my previous ticket #44675.
>>>>>
>>>>> In order to build MET on my Linux Intel machine (64bit),
>>>>> into ../netcdf../include I did as following:
>>>>> cp netcdf.h netcdf.hh.
>>>>> In this way I resolved my ".hh" problem.
>>>>>
>>>>> I also built succesfully GSL as well.
>>>>>
>>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so on...) is
built
>> by
>>>>> Intel compiler.*
>>>>>
>>>>> But when I "make", I get the errors in attachment.
>>>>> In attachment the makefile too.
>>>>>
>>>>> Where do I wrong, now?
>>>>>
>>>>> Thank you veru much,
>>>>> Fabio.
>>>>>
>>>>
>>>>
>>>>
>>
>>
>>
>


------------------------------------------------
Subject: Met building problems: compilation aborted for vx_contable_stats.cc
From: Fabio F.Gervasi
Time: Tue Feb 22 14:40:15 2011

It is really bad news, Paul!
I am the administrator. I live near Rome, but I run a small computer
center
in a company that is located in Udine. Before upgrading the computers
in
Udine,
I usually do experiments on my laptop (like right now).

What about in-person support? Is it really necessary?
Unfortunatley we no have budget to support it, at the moment.

So the bad news is.. maybe I have to stop before I start!

Fabio.


2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>

> Fabio,
>
> It looks like your c++ compiler is not configured correctly.  The
error
> message indicates that your compiler cannot find
> a standard c++ library function (strncpy).  I think you may want to
talk to
> your administrator about the compiler
> configuration.  This will be a problem when you try to compile MET,
too.
>
> ncvalues.cpp(238): error: identifier "strncpy" is undefined
>      strncpy(s, (const char*)the_values + n, (int)the_number);
>
> We would like to provide in-person support to you in Rome.  Would
your
> operations budget support that?
>
> Paul
>
>
> On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >
> > I'm sorry Paul, I made a mistake:
> >
> > I remembered that when I compile netcdf, *I usually set CXX='',
> otherwhise I
> > obtain an
> > abort compilation, as you can see in attachment.*
> > Again, now I remember had a similar problem with gcc (about 2
years ago!)
> >
> > Apparently, for all the other programs do not set CXX is not a
problem.
> > I never solved that issue that now back to me forcefully!
> >
> > How can I fix it, in your opinion?
> >
> > Thanks,
> > Fabio.
> >
> >
> > 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
> >
> >> Fabio,
> >>
> >> The final step in the NetCDF should basically copy cxx/netcdf.hh
to
> >> include/netcdf.hh.  We suspect that there is
> >> something funny going on with your configure and build process.
Can you
> >> please send us log files of your configuration
> >> and build using the following commands:
> >>
> >> $ make clean
> >> $ ./configure --prefix=/netcdf/install/path &> nc_configure.log
> >> $ make &> nc_make.log
> >> $ make install &> nc_make_install.log
> >>
> >> Thanks,
> >>
> >> Paul
> >>
> >>
> >> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
> >>>
> >>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >>>
> >>> Ok Paul,
> >>>
> >>> I removed my netcdf.hh copy.
> >>>
> >>> I'm following the tutorial and I understand about BUFRLIB,
NetCDF and
> >> GSL.
> >>> Regarding BUFRLIB I'll ignore the ifort errors. GSL seems built
ok.
> >>>
> >>> *About NetCDF, I understand that MET wants ".hh" but I think
(but maybe
> I
> >>> wrong) I compiled it correctly. My env settings are:*
> >>> export CC=icc
> >>> export FC=ifort
> >>> export F9X=ifort
> >>> export F90=ifort
> >>> export F77=ifort
> >>> export CXX=icc
> >>>
> >>> *Then I do:*
> >>> ./configure --prefix=/home/master/netcdf-3.6.2
> >>> make
> >>> make install
> >>>
> >>> But *my building process doesn't produce any .hh file into
/include
> >> folder,
> >>> but only a .h file*.
> >>> I built NetCDF several times in the past, and I never had any
problems
> so
> >>> far with it, because WPS, WRF, WRFDA and so on are okay.
> >>>
> >>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx folder,
so I
> >> tried
> >>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
> >>> on my makefile, but now I get:
> >>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16): catastrophic
error: could
> >> not
> >>> open source file "netcdf.h"
> >>>   #include "netcdf.h"
> >>>
> >>> Sorry Paul, maybe I'm doing some stupid errors, but really I
don't
> >>> understand what's wrong on my NetCDF building!
> >>>
> >>> Thank you, Paul, for your patience,
> >>> Fabio.
> >>>
> >>>
> >>>
> >>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
> >>>
> >>>> Fabio,
> >>>>
> >>>> Your solution of copying the NetCDF header file is causing the
> problem.
> >>>>  The file include/netcdf.hh is generated when
> >>>> the c/c++ interface of NetCDF is built.  The instance you are
trying
> to
> >> use
> >>>> may not have that interface built.
> >>>>
> >>>> We recommend that you download and build your own instances of
the
> >>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
> >>>> that they are built in the way that MET expects.  The process
should
> be
> >>>> pretty easy if you carefully read and follow the
> >>>> instructions at the online tutorial I mentioned earlier, using
the
> intel
> >>>> compilers for each step:
> >>>>
> >>>>
> >>>>
> >>
>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
> >>>>
> >>>> You may see some error messages when building BUFRLIB when
executing
> >> this
> >>>> step:
> >>>>
> >>>> ifort -c -DUNDERSCORE *.f *.F
> >>>>
> >>>> Ignore these error messages and continue with the process.  MET
should
> >> work
> >>>> with the partially built version of BUFRLIB.
> >>>>  Please let us know if you have any other questions.
> >>>>
> >>>> Paul
> >>>>
> >>>>
> >>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
> >>>>>
> >>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
> >>>>> Transaction: Ticket created by fabio.gervasi70 at gmail.com
> >>>>>        Queue: met_help
> >>>>>      Subject: Met building problems: compilation aborted for
> >>>> vx_contable_stats.cc
> >>>>>        Owner: Nobody
> >>>>>   Requestors: fabio.gervasi70 at gmail.com
> >>>>>       Status: new
> >>>>>  Ticket <URL:
> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
> >>>>>
> >>>>>
> >>>>> Hi,
> >>>>>
> >>>>> I open another post, but it could be a continuation about
> >>>>> my previous ticket #44675.
> >>>>>
> >>>>> In order to build MET on my Linux Intel machine (64bit),
> >>>>> into ../netcdf../include I did as following:
> >>>>> cp netcdf.h netcdf.hh.
> >>>>> In this way I resolved my ".hh" problem.
> >>>>>
> >>>>> I also built succesfully GSL as well.
> >>>>>
> >>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so on...)
is built
> >> by
> >>>>> Intel compiler.*
> >>>>>
> >>>>> But when I "make", I get the errors in attachment.
> >>>>> In attachment the makefile too.
> >>>>>
> >>>>> Where do I wrong, now?
> >>>>>
> >>>>> Thank you veru much,
> >>>>> Fabio.
> >>>>>
> >>>>
> >>>>
> >>>>
> >>
> >>
> >>
> >
>
>
>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #44711] Met building problems: compilation aborted for vx_contable_stats.cc
From: Paul Oldenburg
Time: Tue Feb 22 15:00:48 2011

Fabio,

I was just kidding about in-person support.  We were just joking that
we would like to visit Rome.

If you are the administrator, perhaps you have GNU compilers installed
on your machine?  You might want to try using
those to build the libraries and MET.  It is what we used to develop
MET, so we have the most experience with them.  You
may already have GNU compilers installed.

Don't give up yet!  Once your compilers are configured correctly,
building MET should just be a matter of following the
instructions on the online tutorial.

Good luck.

Paul


On 02/22/2011 02:40 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>
> It is really bad news, Paul!
> I am the administrator. I live near Rome, but I run a small computer
center
> in a company that is located in Udine. Before upgrading the
computers in
> Udine,
> I usually do experiments on my laptop (like right now).
>
> What about in-person support? Is it really necessary?
> Unfortunatley we no have budget to support it, at the moment.
>
> So the bad news is.. maybe I have to stop before I start!
>
> Fabio.
>
>
> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>
>> Fabio,
>>
>> It looks like your c++ compiler is not configured correctly.  The
error
>> message indicates that your compiler cannot find
>> a standard c++ library function (strncpy).  I think you may want to
talk to
>> your administrator about the compiler
>> configuration.  This will be a problem when you try to compile MET,
too.
>>
>> ncvalues.cpp(238): error: identifier "strncpy" is undefined
>>      strncpy(s, (const char*)the_values + n, (int)the_number);
>>
>> We would like to provide in-person support to you in Rome.  Would
your
>> operations budget support that?
>>
>> Paul
>>
>>
>> On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>
>>> I'm sorry Paul, I made a mistake:
>>>
>>> I remembered that when I compile netcdf, *I usually set CXX='',
>> otherwhise I
>>> obtain an
>>> abort compilation, as you can see in attachment.*
>>> Again, now I remember had a similar problem with gcc (about 2
years ago!)
>>>
>>> Apparently, for all the other programs do not set CXX is not a
problem.
>>> I never solved that issue that now back to me forcefully!
>>>
>>> How can I fix it, in your opinion?
>>>
>>> Thanks,
>>> Fabio.
>>>
>>>
>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>>>
>>>> Fabio,
>>>>
>>>> The final step in the NetCDF should basically copy cxx/netcdf.hh
to
>>>> include/netcdf.hh.  We suspect that there is
>>>> something funny going on with your configure and build process.
Can you
>>>> please send us log files of your configuration
>>>> and build using the following commands:
>>>>
>>>> $ make clean
>>>> $ ./configure --prefix=/netcdf/install/path &> nc_configure.log
>>>> $ make &> nc_make.log
>>>> $ make install &> nc_make_install.log
>>>>
>>>> Thanks,
>>>>
>>>> Paul
>>>>
>>>>
>>>> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>>>>>
>>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>>>
>>>>> Ok Paul,
>>>>>
>>>>> I removed my netcdf.hh copy.
>>>>>
>>>>> I'm following the tutorial and I understand about BUFRLIB,
NetCDF and
>>>> GSL.
>>>>> Regarding BUFRLIB I'll ignore the ifort errors. GSL seems built
ok.
>>>>>
>>>>> *About NetCDF, I understand that MET wants ".hh" but I think
(but maybe
>> I
>>>>> wrong) I compiled it correctly. My env settings are:*
>>>>> export CC=icc
>>>>> export FC=ifort
>>>>> export F9X=ifort
>>>>> export F90=ifort
>>>>> export F77=ifort
>>>>> export CXX=icc
>>>>>
>>>>> *Then I do:*
>>>>> ./configure --prefix=/home/master/netcdf-3.6.2
>>>>> make
>>>>> make install
>>>>>
>>>>> But *my building process doesn't produce any .hh file into
/include
>>>> folder,
>>>>> but only a .h file*.
>>>>> I built NetCDF several times in the past, and I never had any
problems
>> so
>>>>> far with it, because WPS, WRF, WRFDA and so on are okay.
>>>>>
>>>>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx folder,
so I
>>>> tried
>>>>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
>>>>> on my makefile, but now I get:
>>>>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16): catastrophic
error: could
>>>> not
>>>>> open source file "netcdf.h"
>>>>>   #include "netcdf.h"
>>>>>
>>>>> Sorry Paul, maybe I'm doing some stupid errors, but really I
don't
>>>>> understand what's wrong on my NetCDF building!
>>>>>
>>>>> Thank you, Paul, for your patience,
>>>>> Fabio.
>>>>>
>>>>>
>>>>>
>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>>>>>
>>>>>> Fabio,
>>>>>>
>>>>>> Your solution of copying the NetCDF header file is causing the
>> problem.
>>>>>>  The file include/netcdf.hh is generated when
>>>>>> the c/c++ interface of NetCDF is built.  The instance you are
trying
>> to
>>>> use
>>>>>> may not have that interface built.
>>>>>>
>>>>>> We recommend that you download and build your own instances of
the
>>>>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
>>>>>> that they are built in the way that MET expects.  The process
should
>> be
>>>>>> pretty easy if you carefully read and follow the
>>>>>> instructions at the online tutorial I mentioned earlier, using
the
>> intel
>>>>>> compilers for each step:
>>>>>>
>>>>>>
>>>>>>
>>>>
>>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
>>>>>>
>>>>>> You may see some error messages when building BUFRLIB when
executing
>>>> this
>>>>>> step:
>>>>>>
>>>>>> ifort -c -DUNDERSCORE *.f *.F
>>>>>>
>>>>>> Ignore these error messages and continue with the process.  MET
should
>>>> work
>>>>>> with the partially built version of BUFRLIB.
>>>>>>  Please let us know if you have any other questions.
>>>>>>
>>>>>> Paul
>>>>>>
>>>>>>
>>>>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>>>>>>
>>>>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
>>>>>>> Transaction: Ticket created by fabio.gervasi70 at gmail.com
>>>>>>>        Queue: met_help
>>>>>>>      Subject: Met building problems: compilation aborted for
>>>>>> vx_contable_stats.cc
>>>>>>>        Owner: Nobody
>>>>>>>   Requestors: fabio.gervasi70 at gmail.com
>>>>>>>       Status: new
>>>>>>>  Ticket <URL:
>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
>>>>>>>
>>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> I open another post, but it could be a continuation about
>>>>>>> my previous ticket #44675.
>>>>>>>
>>>>>>> In order to build MET on my Linux Intel machine (64bit),
>>>>>>> into ../netcdf../include I did as following:
>>>>>>> cp netcdf.h netcdf.hh.
>>>>>>> In this way I resolved my ".hh" problem.
>>>>>>>
>>>>>>> I also built succesfully GSL as well.
>>>>>>>
>>>>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so on...)
is built
>>>> by
>>>>>>> Intel compiler.*
>>>>>>>
>>>>>>> But when I "make", I get the errors in attachment.
>>>>>>> In attachment the makefile too.
>>>>>>>
>>>>>>> Where do I wrong, now?
>>>>>>>
>>>>>>> Thank you veru much,
>>>>>>> Fabio.
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>
>>>>
>>>>
>>>
>>
>>
>>


------------------------------------------------
Subject: Met building problems: compilation aborted for vx_contable_stats.cc
From: Fabio F.Gervasi
Time: Wed Feb 23 00:26:44 2011

Paul, if you want to visit Rome and be our guest (I live in Orvieto,
Umbria) you
are welcome!
I speak English just a little bit, I graduated in the Netherlands
although now
I have lost a bit of my English ..
But It doesn't matter...! :)

About GNU compiler. Sure!
I started using Intel compiler because I had some problems with WRF
model
compiling, but maybe it's time to get back to GNU.

So, in attachment the *netcdf-3.6.2* building by GNU. But as you can
see I
have similar problem!
What do you think about?

In the meantime I also tried* netcdf-4.1.1* (with the option
--disable-netcdf-4 in the configure). Following my results:
1) GNU --> see attachments
2) INTEL --> see attachments

Similar warning/errors both make and make check.

About compilers the problem is (also) that usually I have to use the
same
compiler for each programs.
My final target is to use the "chain": WRFDA (or GSI) --> WPS/WRF-
NMM/WPP
--> MET.
But GSI, for example, seems not "GNU" supported (only Intel or PGI).
With MET, instead, you suggest GNU but I have problems with both
compilers.

What must I do??? I don't give up, but it is very hard! :)

Fabio.

PS: I use Linux Fedora14 and GNU version is 4.5.1.



2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>

> Fabio,
>
> I was just kidding about in-person support.  We were just joking
that we
> would like to visit Rome.
>
> If you are the administrator, perhaps you have GNU compilers
installed on
> your machine?  You might want to try using
> those to build the libraries and MET.  It is what we used to develop
MET,
> so we have the most experience with them.  You
> may already have GNU compilers installed.
>
> Don't give up yet!  Once your compilers are configured correctly,
building
> MET should just be a matter of following the
> instructions on the online tutorial.
>
> Good luck.
>
> Paul
>
>
> On 02/22/2011 02:40 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >
> > It is really bad news, Paul!
> > I am the administrator. I live near Rome, but I run a small
computer
> center
> > in a company that is located in Udine. Before upgrading the
computers in
> > Udine,
> > I usually do experiments on my laptop (like right now).
> >
> > What about in-person support? Is it really necessary?
> > Unfortunatley we no have budget to support it, at the moment.
> >
> > So the bad news is.. maybe I have to stop before I start!
> >
> > Fabio.
> >
> >
> > 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
> >
> >> Fabio,
> >>
> >> It looks like your c++ compiler is not configured correctly.  The
error
> >> message indicates that your compiler cannot find
> >> a standard c++ library function (strncpy).  I think you may want
to talk
> to
> >> your administrator about the compiler
> >> configuration.  This will be a problem when you try to compile
MET, too.
> >>
> >> ncvalues.cpp(238): error: identifier "strncpy" is undefined
> >>      strncpy(s, (const char*)the_values + n, (int)the_number);
> >>
> >> We would like to provide in-person support to you in Rome.  Would
your
> >> operations budget support that?
> >>
> >> Paul
> >>
> >>
> >> On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
> >>>
> >>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >>>
> >>> I'm sorry Paul, I made a mistake:
> >>>
> >>> I remembered that when I compile netcdf, *I usually set CXX='',
> >> otherwhise I
> >>> obtain an
> >>> abort compilation, as you can see in attachment.*
> >>> Again, now I remember had a similar problem with gcc (about 2
years
> ago!)
> >>>
> >>> Apparently, for all the other programs do not set CXX is not a
problem.
> >>> I never solved that issue that now back to me forcefully!
> >>>
> >>> How can I fix it, in your opinion?
> >>>
> >>> Thanks,
> >>> Fabio.
> >>>
> >>>
> >>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
> >>>
> >>>> Fabio,
> >>>>
> >>>> The final step in the NetCDF should basically copy
cxx/netcdf.hh to
> >>>> include/netcdf.hh.  We suspect that there is
> >>>> something funny going on with your configure and build process.
Can
> you
> >>>> please send us log files of your configuration
> >>>> and build using the following commands:
> >>>>
> >>>> $ make clean
> >>>> $ ./configure --prefix=/netcdf/install/path &> nc_configure.log
> >>>> $ make &> nc_make.log
> >>>> $ make install &> nc_make_install.log
> >>>>
> >>>> Thanks,
> >>>>
> >>>> Paul
> >>>>
> >>>>
> >>>> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
> >>>>>
> >>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711
>
> >>>>>
> >>>>> Ok Paul,
> >>>>>
> >>>>> I removed my netcdf.hh copy.
> >>>>>
> >>>>> I'm following the tutorial and I understand about BUFRLIB,
NetCDF and
> >>>> GSL.
> >>>>> Regarding BUFRLIB I'll ignore the ifort errors. GSL seems
built ok.
> >>>>>
> >>>>> *About NetCDF, I understand that MET wants ".hh" but I think
(but
> maybe
> >> I
> >>>>> wrong) I compiled it correctly. My env settings are:*
> >>>>> export CC=icc
> >>>>> export FC=ifort
> >>>>> export F9X=ifort
> >>>>> export F90=ifort
> >>>>> export F77=ifort
> >>>>> export CXX=icc
> >>>>>
> >>>>> *Then I do:*
> >>>>> ./configure --prefix=/home/master/netcdf-3.6.2
> >>>>> make
> >>>>> make install
> >>>>>
> >>>>> But *my building process doesn't produce any .hh file into
/include
> >>>> folder,
> >>>>> but only a .h file*.
> >>>>> I built NetCDF several times in the past, and I never had any
> problems
> >> so
> >>>>> far with it, because WPS, WRF, WRFDA and so on are okay.
> >>>>>
> >>>>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx
folder, so I
> >>>> tried
> >>>>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
> >>>>> on my makefile, but now I get:
> >>>>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16): catastrophic
error:
> could
> >>>> not
> >>>>> open source file "netcdf.h"
> >>>>>   #include "netcdf.h"
> >>>>>
> >>>>> Sorry Paul, maybe I'm doing some stupid errors, but really I
don't
> >>>>> understand what's wrong on my NetCDF building!
> >>>>>
> >>>>> Thank you, Paul, for your patience,
> >>>>> Fabio.
> >>>>>
> >>>>>
> >>>>>
> >>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
> >>>>>
> >>>>>> Fabio,
> >>>>>>
> >>>>>> Your solution of copying the NetCDF header file is causing
the
> >> problem.
> >>>>>>  The file include/netcdf.hh is generated when
> >>>>>> the c/c++ interface of NetCDF is built.  The instance you are
trying
> >> to
> >>>> use
> >>>>>> may not have that interface built.
> >>>>>>
> >>>>>> We recommend that you download and build your own instances
of the
> >>>>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
> >>>>>> that they are built in the way that MET expects.  The process
should
> >> be
> >>>>>> pretty easy if you carefully read and follow the
> >>>>>> instructions at the online tutorial I mentioned earlier,
using the
> >> intel
> >>>>>> compilers for each step:
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>
> >>
>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
> >>>>>>
> >>>>>> You may see some error messages when building BUFRLIB when
executing
> >>>> this
> >>>>>> step:
> >>>>>>
> >>>>>> ifort -c -DUNDERSCORE *.f *.F
> >>>>>>
> >>>>>> Ignore these error messages and continue with the process.
MET
> should
> >>>> work
> >>>>>> with the partially built version of BUFRLIB.
> >>>>>>  Please let us know if you have any other questions.
> >>>>>>
> >>>>>> Paul
> >>>>>>
> >>>>>>
> >>>>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
> >>>>>>>
> >>>>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
> >>>>>>> Transaction: Ticket created by fabio.gervasi70 at gmail.com
> >>>>>>>        Queue: met_help
> >>>>>>>      Subject: Met building problems: compilation aborted for
> >>>>>> vx_contable_stats.cc
> >>>>>>>        Owner: Nobody
> >>>>>>>   Requestors: fabio.gervasi70 at gmail.com
> >>>>>>>       Status: new
> >>>>>>>  Ticket <URL:
> >> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
> >>>>>>>
> >>>>>>>
> >>>>>>> Hi,
> >>>>>>>
> >>>>>>> I open another post, but it could be a continuation about
> >>>>>>> my previous ticket #44675.
> >>>>>>>
> >>>>>>> In order to build MET on my Linux Intel machine (64bit),
> >>>>>>> into ../netcdf../include I did as following:
> >>>>>>> cp netcdf.h netcdf.hh.
> >>>>>>> In this way I resolved my ".hh" problem.
> >>>>>>>
> >>>>>>> I also built succesfully GSL as well.
> >>>>>>>
> >>>>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so on...)
is
> built
> >>>> by
> >>>>>>> Intel compiler.*
> >>>>>>>
> >>>>>>> But when I "make", I get the errors in attachment.
> >>>>>>> In attachment the makefile too.
> >>>>>>>
> >>>>>>> Where do I wrong, now?
> >>>>>>>
> >>>>>>> Thank you veru much,
> >>>>>>> Fabio.
> >>>>>>>
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>
> >>>>
> >>>>
> >>>
> >>
> >>
> >>
>
>
>

------------------------------------------------
Subject: Met building problems: compilation aborted for vx_contable_stats.cc
From: Fabio F.Gervasi
Time: Wed Feb 23 00:54:40 2011

Errata corrige, Paul:
about GNU compiler, maybe I have to set CXX=g++ (not gcc).
In this way, netcdf-4.1.1 seems work fine. I confirm, instead, netcdf-
3.6.2
has still problems.

But now, I have to check if I can use netcdf version 4 with WRF and
the
other my software!

Fabio.


2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>

> Paul, if you want to visit Rome and be our guest (I live in Orvieto,
> Umbria) you are welcome!
> I speak English just a little bit, I graduated in the Netherlands
although now
> I have lost a bit of my English ..
> But It doesn't matter...! :)
>
> About GNU compiler. Sure!
> I started using Intel compiler because I had some problems with WRF
model
> compiling, but maybe it's time to get back to GNU.
>
> So, in attachment the *netcdf-3.6.2* building by GNU. But as you can
see I
> have similar problem!
> What do you think about?
>
> In the meantime I also tried* netcdf-4.1.1* (with the option
> --disable-netcdf-4 in the configure). Following my results:
> 1) GNU --> see attachments
> 2) INTEL --> see attachments
>
> Similar warning/errors both make and make check.
>
> About compilers the problem is (also) that usually I have to use the
same
> compiler for each programs.
> My final target is to use the "chain": WRFDA (or GSI) --> WPS/WRF-
NMM/WPP
> --> MET.
> But GSI, for example, seems not "GNU" supported (only Intel or PGI).
> With MET, instead, you suggest GNU but I have problems with both
compilers.
>
> What must I do??? I don't give up, but it is very hard! :)
>
> Fabio.
>
> PS: I use Linux Fedora14 and GNU version is 4.5.1.
>
>
>
> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>
>> Fabio,
>>
>> I was just kidding about in-person support.  We were just joking
that we
>> would like to visit Rome.
>>
>> If you are the administrator, perhaps you have GNU compilers
installed on
>> your machine?  You might want to try using
>> those to build the libraries and MET.  It is what we used to
develop MET,
>> so we have the most experience with them.  You
>> may already have GNU compilers installed.
>>
>> Don't give up yet!  Once your compilers are configured correctly,
building
>> MET should just be a matter of following the
>> instructions on the online tutorial.
>>
>> Good luck.
>>
>> Paul
>>
>>
>> On 02/22/2011 02:40 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>> >
>> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>> >
>> > It is really bad news, Paul!
>> > I am the administrator. I live near Rome, but I run a small
computer
>> center
>> > in a company that is located in Udine. Before upgrading the
computers in
>> > Udine,
>> > I usually do experiments on my laptop (like right now).
>> >
>> > What about in-person support? Is it really necessary?
>> > Unfortunatley we no have budget to support it, at the moment.
>> >
>> > So the bad news is.. maybe I have to stop before I start!
>> >
>> > Fabio.
>> >
>> >
>> > 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>> >
>> >> Fabio,
>> >>
>> >> It looks like your c++ compiler is not configured correctly.
The error
>> >> message indicates that your compiler cannot find
>> >> a standard c++ library function (strncpy).  I think you may want
to
>> talk to
>> >> your administrator about the compiler
>> >> configuration.  This will be a problem when you try to compile
MET,
>> too.
>> >>
>> >> ncvalues.cpp(238): error: identifier "strncpy" is undefined
>> >>      strncpy(s, (const char*)the_values + n, (int)the_number);
>> >>
>> >> We would like to provide in-person support to you in Rome.
Would your
>> >> operations budget support that?
>> >>
>> >> Paul
>> >>
>> >>
>> >> On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>> >>>
>> >>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>> >>>
>> >>> I'm sorry Paul, I made a mistake:
>> >>>
>> >>> I remembered that when I compile netcdf, *I usually set CXX='',
>> >> otherwhise I
>> >>> obtain an
>> >>> abort compilation, as you can see in attachment.*
>> >>> Again, now I remember had a similar problem with gcc (about 2
years
>> ago!)
>> >>>
>> >>> Apparently, for all the other programs do not set CXX is not a
>> problem.
>> >>> I never solved that issue that now back to me forcefully!
>> >>>
>> >>> How can I fix it, in your opinion?
>> >>>
>> >>> Thanks,
>> >>> Fabio.
>> >>>
>> >>>
>> >>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>> >>>
>> >>>> Fabio,
>> >>>>
>> >>>> The final step in the NetCDF should basically copy
cxx/netcdf.hh to
>> >>>> include/netcdf.hh.  We suspect that there is
>> >>>> something funny going on with your configure and build
process.  Can
>> you
>> >>>> please send us log files of your configuration
>> >>>> and build using the following commands:
>> >>>>
>> >>>> $ make clean
>> >>>> $ ./configure --prefix=/netcdf/install/path &>
nc_configure.log
>> >>>> $ make &> nc_make.log
>> >>>> $ make install &> nc_make_install.log
>> >>>>
>> >>>> Thanks,
>> >>>>
>> >>>> Paul
>> >>>>
>> >>>>
>> >>>> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>> >>>>>
>> >>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711
>
>> >>>>>
>> >>>>> Ok Paul,
>> >>>>>
>> >>>>> I removed my netcdf.hh copy.
>> >>>>>
>> >>>>> I'm following the tutorial and I understand about BUFRLIB,
NetCDF
>> and
>> >>>> GSL.
>> >>>>> Regarding BUFRLIB I'll ignore the ifort errors. GSL seems
built ok.
>> >>>>>
>> >>>>> *About NetCDF, I understand that MET wants ".hh" but I think
(but
>> maybe
>> >> I
>> >>>>> wrong) I compiled it correctly. My env settings are:*
>> >>>>> export CC=icc
>> >>>>> export FC=ifort
>> >>>>> export F9X=ifort
>> >>>>> export F90=ifort
>> >>>>> export F77=ifort
>> >>>>> export CXX=icc
>> >>>>>
>> >>>>> *Then I do:*
>> >>>>> ./configure --prefix=/home/master/netcdf-3.6.2
>> >>>>> make
>> >>>>> make install
>> >>>>>
>> >>>>> But *my building process doesn't produce any .hh file into
/include
>> >>>> folder,
>> >>>>> but only a .h file*.
>> >>>>> I built NetCDF several times in the past, and I never had any
>> problems
>> >> so
>> >>>>> far with it, because WPS, WRF, WRFDA and so on are okay.
>> >>>>>
>> >>>>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx
folder, so I
>> >>>> tried
>> >>>>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
>> >>>>> on my makefile, but now I get:
>> >>>>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16): catastrophic
error:
>> could
>> >>>> not
>> >>>>> open source file "netcdf.h"
>> >>>>>   #include "netcdf.h"
>> >>>>>
>> >>>>> Sorry Paul, maybe I'm doing some stupid errors, but really I
don't
>> >>>>> understand what's wrong on my NetCDF building!
>> >>>>>
>> >>>>> Thank you, Paul, for your patience,
>> >>>>> Fabio.
>> >>>>>
>> >>>>>
>> >>>>>
>> >>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
>> >>>>>
>> >>>>>> Fabio,
>> >>>>>>
>> >>>>>> Your solution of copying the NetCDF header file is causing
the
>> >> problem.
>> >>>>>>  The file include/netcdf.hh is generated when
>> >>>>>> the c/c++ interface of NetCDF is built.  The instance you
are
>> trying
>> >> to
>> >>>> use
>> >>>>>> may not have that interface built.
>> >>>>>>
>> >>>>>> We recommend that you download and build your own instances
of the
>> >>>>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
>> >>>>>> that they are built in the way that MET expects.  The
process
>> should
>> >> be
>> >>>>>> pretty easy if you carefully read and follow the
>> >>>>>> instructions at the online tutorial I mentioned earlier,
using the
>> >> intel
>> >>>>>> compilers for each step:
>> >>>>>>
>> >>>>>>
>> >>>>>>
>> >>>>
>> >>
>>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
>> >>>>>>
>> >>>>>> You may see some error messages when building BUFRLIB when
>> executing
>> >>>> this
>> >>>>>> step:
>> >>>>>>
>> >>>>>> ifort -c -DUNDERSCORE *.f *.F
>> >>>>>>
>> >>>>>> Ignore these error messages and continue with the process.
MET
>> should
>> >>>> work
>> >>>>>> with the partially built version of BUFRLIB.
>> >>>>>>  Please let us know if you have any other questions.
>> >>>>>>
>> >>>>>> Paul
>> >>>>>>
>> >>>>>>
>> >>>>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>> >>>>>>>
>> >>>>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
>> >>>>>>> Transaction: Ticket created by fabio.gervasi70 at gmail.com
>> >>>>>>>        Queue: met_help
>> >>>>>>>      Subject: Met building problems: compilation aborted
for
>> >>>>>> vx_contable_stats.cc
>> >>>>>>>        Owner: Nobody
>> >>>>>>>   Requestors: fabio.gervasi70 at gmail.com
>> >>>>>>>       Status: new
>> >>>>>>>  Ticket <URL:
>> >> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
>> >>>>>>>
>> >>>>>>>
>> >>>>>>> Hi,
>> >>>>>>>
>> >>>>>>> I open another post, but it could be a continuation about
>> >>>>>>> my previous ticket #44675.
>> >>>>>>>
>> >>>>>>> In order to build MET on my Linux Intel machine (64bit),
>> >>>>>>> into ../netcdf../include I did as following:
>> >>>>>>> cp netcdf.h netcdf.hh.
>> >>>>>>> In this way I resolved my ".hh" problem.
>> >>>>>>>
>> >>>>>>> I also built succesfully GSL as well.
>> >>>>>>>
>> >>>>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so
on...) is
>> built
>> >>>> by
>> >>>>>>> Intel compiler.*
>> >>>>>>>
>> >>>>>>> But when I "make", I get the errors in attachment.
>> >>>>>>> In attachment the makefile too.
>> >>>>>>>
>> >>>>>>> Where do I wrong, now?
>> >>>>>>>
>> >>>>>>> Thank you veru much,
>> >>>>>>> Fabio.
>> >>>>>>>
>> >>>>>>
>> >>>>>>
>> >>>>>>
>> >>>>
>> >>>>
>> >>>>
>> >>>
>> >>
>> >>
>> >>
>>
>>
>>
>

------------------------------------------------
Subject: Met building problems: compilation aborted for vx_contable_stats.cc
From: Fabio F.Gervasi
Time: Wed Feb 23 01:58:44 2011

An updating, Paul:
As I thought: WRF wnats NetCDF-3.6.2, not 4. Although I compiled
NetCDF with the
--disable-netcdf4 configure option.
Some times ago, wrf helpdesk told me that WRF wants 3.6.2...


2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>

> Errata corrige, Paul:
> about GNU compiler, maybe I have to set CXX=g++ (not gcc).
> In this way, netcdf-4.1.1 seems work fine. I confirm, instead,
netcdf-3.6.2
> has still problems.
>
> But now, I have to check if I can use netcdf version 4 with WRF and
the
> other my software!
>
> Fabio.
>
>
> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
>
> Paul, if you want to visit Rome and be our guest (I live in Orvieto,
>> Umbria) you are welcome!
>> I speak English just a little bit, I graduated in the Netherlands
>> although now I have lost a bit of my English ..
>> But It doesn't matter...! :)
>>
>> About GNU compiler. Sure!
>> I started using Intel compiler because I had some problems with WRF
model
>> compiling, but maybe it's time to get back to GNU.
>>
>> So, in attachment the *netcdf-3.6.2* building by GNU. But as you
can see
>> I have similar problem!
>> What do you think about?
>>
>> In the meantime I also tried* netcdf-4.1.1* (with the option
>> --disable-netcdf-4 in the configure). Following my results:
>> 1) GNU --> see attachments
>> 2) INTEL --> see attachments
>>
>> Similar warning/errors both make and make check.
>>
>> About compilers the problem is (also) that usually I have to use
the same
>> compiler for each programs.
>> My final target is to use the "chain": WRFDA (or GSI) --> WPS/WRF-
NMM/WPP
>> --> MET.
>> But GSI, for example, seems not "GNU" supported (only Intel or
PGI).
>> With MET, instead, you suggest GNU but I have problems with both
>> compilers.
>>
>> What must I do??? I don't give up, but it is very hard! :)
>>
>> Fabio.
>>
>> PS: I use Linux Fedora14 and GNU version is 4.5.1.
>>
>>
>>
>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>>
>>> Fabio,
>>>
>>> I was just kidding about in-person support.  We were just joking
that we
>>> would like to visit Rome.
>>>
>>> If you are the administrator, perhaps you have GNU compilers
installed on
>>> your machine?  You might want to try using
>>> those to build the libraries and MET.  It is what we used to
develop MET,
>>> so we have the most experience with them.  You
>>> may already have GNU compilers installed.
>>>
>>> Don't give up yet!  Once your compilers are configured correctly,
>>> building MET should just be a matter of following the
>>> instructions on the online tutorial.
>>>
>>> Good luck.
>>>
>>> Paul
>>>
>>>
>>> On 02/22/2011 02:40 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>>> >
>>> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>> >
>>> > It is really bad news, Paul!
>>> > I am the administrator. I live near Rome, but I run a small
computer
>>> center
>>> > in a company that is located in Udine. Before upgrading the
computers
>>> in
>>> > Udine,
>>> > I usually do experiments on my laptop (like right now).
>>> >
>>> > What about in-person support? Is it really necessary?
>>> > Unfortunatley we no have budget to support it, at the moment.
>>> >
>>> > So the bad news is.. maybe I have to stop before I start!
>>> >
>>> > Fabio.
>>> >
>>> >
>>> > 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>>> >
>>> >> Fabio,
>>> >>
>>> >> It looks like your c++ compiler is not configured correctly.
The
>>> error
>>> >> message indicates that your compiler cannot find
>>> >> a standard c++ library function (strncpy).  I think you may
want to
>>> talk to
>>> >> your administrator about the compiler
>>> >> configuration.  This will be a problem when you try to compile
MET,
>>> too.
>>> >>
>>> >> ncvalues.cpp(238): error: identifier "strncpy" is undefined
>>> >>      strncpy(s, (const char*)the_values + n, (int)the_number);
>>> >>
>>> >> We would like to provide in-person support to you in Rome.
Would your
>>> >> operations budget support that?
>>> >>
>>> >> Paul
>>> >>
>>> >>
>>> >> On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>> >>>
>>> >>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711
>
>>> >>>
>>> >>> I'm sorry Paul, I made a mistake:
>>> >>>
>>> >>> I remembered that when I compile netcdf, *I usually set
CXX='',
>>> >> otherwhise I
>>> >>> obtain an
>>> >>> abort compilation, as you can see in attachment.*
>>> >>> Again, now I remember had a similar problem with gcc (about 2
years
>>> ago!)
>>> >>>
>>> >>> Apparently, for all the other programs do not set CXX is not a
>>> problem.
>>> >>> I never solved that issue that now back to me forcefully!
>>> >>>
>>> >>> How can I fix it, in your opinion?
>>> >>>
>>> >>> Thanks,
>>> >>> Fabio.
>>> >>>
>>> >>>
>>> >>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
>>> >>>
>>> >>>> Fabio,
>>> >>>>
>>> >>>> The final step in the NetCDF should basically copy
cxx/netcdf.hh to
>>> >>>> include/netcdf.hh.  We suspect that there is
>>> >>>> something funny going on with your configure and build
process.  Can
>>> you
>>> >>>> please send us log files of your configuration
>>> >>>> and build using the following commands:
>>> >>>>
>>> >>>> $ make clean
>>> >>>> $ ./configure --prefix=/netcdf/install/path &>
nc_configure.log
>>> >>>> $ make &> nc_make.log
>>> >>>> $ make install &> nc_make_install.log
>>> >>>>
>>> >>>> Thanks,
>>> >>>>
>>> >>>> Paul
>>> >>>>
>>> >>>>
>>> >>>> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>> >>>>>
>>> >>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>> >>>>>
>>> >>>>> Ok Paul,
>>> >>>>>
>>> >>>>> I removed my netcdf.hh copy.
>>> >>>>>
>>> >>>>> I'm following the tutorial and I understand about BUFRLIB,
NetCDF
>>> and
>>> >>>> GSL.
>>> >>>>> Regarding BUFRLIB I'll ignore the ifort errors. GSL seems
built ok.
>>> >>>>>
>>> >>>>> *About NetCDF, I understand that MET wants ".hh" but I think
(but
>>> maybe
>>> >> I
>>> >>>>> wrong) I compiled it correctly. My env settings are:*
>>> >>>>> export CC=icc
>>> >>>>> export FC=ifort
>>> >>>>> export F9X=ifort
>>> >>>>> export F90=ifort
>>> >>>>> export F77=ifort
>>> >>>>> export CXX=icc
>>> >>>>>
>>> >>>>> *Then I do:*
>>> >>>>> ./configure --prefix=/home/master/netcdf-3.6.2
>>> >>>>> make
>>> >>>>> make install
>>> >>>>>
>>> >>>>> But *my building process doesn't produce any .hh file into
/include
>>> >>>> folder,
>>> >>>>> but only a .h file*.
>>> >>>>> I built NetCDF several times in the past, and I never had
any
>>> problems
>>> >> so
>>> >>>>> far with it, because WPS, WRF, WRFDA and so on are okay.
>>> >>>>>
>>> >>>>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx
folder, so
>>> I
>>> >>>> tried
>>> >>>>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
>>> >>>>> on my makefile, but now I get:
>>> >>>>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16): catastrophic
error:
>>> could
>>> >>>> not
>>> >>>>> open source file "netcdf.h"
>>> >>>>>   #include "netcdf.h"
>>> >>>>>
>>> >>>>> Sorry Paul, maybe I'm doing some stupid errors, but really I
don't
>>> >>>>> understand what's wrong on my NetCDF building!
>>> >>>>>
>>> >>>>> Thank you, Paul, for your patience,
>>> >>>>> Fabio.
>>> >>>>>
>>> >>>>>
>>> >>>>>
>>> >>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
>>> >>>>>
>>> >>>>>> Fabio,
>>> >>>>>>
>>> >>>>>> Your solution of copying the NetCDF header file is causing
the
>>> >> problem.
>>> >>>>>>  The file include/netcdf.hh is generated when
>>> >>>>>> the c/c++ interface of NetCDF is built.  The instance you
are
>>> trying
>>> >> to
>>> >>>> use
>>> >>>>>> may not have that interface built.
>>> >>>>>>
>>> >>>>>> We recommend that you download and build your own instances
of the
>>> >>>>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
>>> >>>>>> that they are built in the way that MET expects.  The
process
>>> should
>>> >> be
>>> >>>>>> pretty easy if you carefully read and follow the
>>> >>>>>> instructions at the online tutorial I mentioned earlier,
using the
>>> >> intel
>>> >>>>>> compilers for each step:
>>> >>>>>>
>>> >>>>>>
>>> >>>>>>
>>> >>>>
>>> >>
>>>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
>>> >>>>>>
>>> >>>>>> You may see some error messages when building BUFRLIB when
>>> executing
>>> >>>> this
>>> >>>>>> step:
>>> >>>>>>
>>> >>>>>> ifort -c -DUNDERSCORE *.f *.F
>>> >>>>>>
>>> >>>>>> Ignore these error messages and continue with the process.
MET
>>> should
>>> >>>> work
>>> >>>>>> with the partially built version of BUFRLIB.
>>> >>>>>>  Please let us know if you have any other questions.
>>> >>>>>>
>>> >>>>>> Paul
>>> >>>>>>
>>> >>>>>>
>>> >>>>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>> >>>>>>>
>>> >>>>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
>>> >>>>>>> Transaction: Ticket created by fabio.gervasi70 at gmail.com
>>> >>>>>>>        Queue: met_help
>>> >>>>>>>      Subject: Met building problems: compilation aborted
for
>>> >>>>>> vx_contable_stats.cc
>>> >>>>>>>        Owner: Nobody
>>> >>>>>>>   Requestors: fabio.gervasi70 at gmail.com
>>> >>>>>>>       Status: new
>>> >>>>>>>  Ticket <URL:
>>> >> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
>>> >>>>>>>
>>> >>>>>>>
>>> >>>>>>> Hi,
>>> >>>>>>>
>>> >>>>>>> I open another post, but it could be a continuation about
>>> >>>>>>> my previous ticket #44675.
>>> >>>>>>>
>>> >>>>>>> In order to build MET on my Linux Intel machine (64bit),
>>> >>>>>>> into ../netcdf../include I did as following:
>>> >>>>>>> cp netcdf.h netcdf.hh.
>>> >>>>>>> In this way I resolved my ".hh" problem.
>>> >>>>>>>
>>> >>>>>>> I also built succesfully GSL as well.
>>> >>>>>>>
>>> >>>>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so
on...) is
>>> built
>>> >>>> by
>>> >>>>>>> Intel compiler.*
>>> >>>>>>>
>>> >>>>>>> But when I "make", I get the errors in attachment.
>>> >>>>>>> In attachment the makefile too.
>>> >>>>>>>
>>> >>>>>>> Where do I wrong, now?
>>> >>>>>>>
>>> >>>>>>> Thank you veru much,
>>> >>>>>>> Fabio.
>>> >>>>>>>
>>> >>>>>>
>>> >>>>>>
>>> >>>>>>
>>> >>>>
>>> >>>>
>>> >>>>
>>> >>>
>>> >>
>>> >>
>>> >>
>>>
>>>
>>>
>>
>

------------------------------------------------
Subject: Met building problems: compilation aborted for vx_contable_stats.cc
From: Fabio F.Gervasi
Time: Wed Feb 23 06:10:59 2011

Hi Paul, since yesterday I wrote to you a lot of emails, so I'd like
to
summarize the situation, after some tests.

*When I use** GNU, env are:*
export CC=gcc
export FC=gfortran
export F9X=gfortran
export F90=gfortran
export F77=gfortran
export CXX=g++

*When I use Intel, env are:*
export CC=icc
export FC=ifort
export F9X=ifort
export F90=ifort
export F77=ifort
export CXX=icpc

*1)* Both GNU and Intel compilers, I'm able to build GSL and BUFRLIB
(for
bufrlib I ignored ifort errors, about gnu I replaced nvnwin.f file)
*2)* Both GNU and Intel compilers, I'm able to build NetCDF-4.1.1 with
CXX
*3)* About MET now I get the logs in attachments for both compilers
(errors,
but gnu seems better then intel). In attachment the gnu and intel
makefile
that I'm using, too.
*4)* Anyway, unfortunately I'm not able to compile my WRF-NMM model
with NetCDF-4.1.1
(also using --disable-netcdf4 option). I'm able to build WRF-NMM only
using
NetCDF-3.6.2. Can I, in case, to use Net3 with WRF and Net4 with MET
at the
same time?
*5)* About the choosing between GNU or Intel, I have advantages and
disvantages. For example at the moment I have some problems with GNU
building WRF model (but I think helpdesk can help me). But at the same
time,
GSI seems supports better Intel.

As I can told you previously, I need to do the "best choice" between
GNU or
INTEL, taking into account the chain:
*GSI (or WRFDA) --> WPS/WRF-NMM/WPP --> MET*, and and related
library...

I hope that my situation is more clear now.
Let me know, if possible, what you think about it.

Ciao!
Fabio.



2011/2/23 RAL HelpDesk {for Fabio F.Gervasi} <met_help at ucar.edu>
>
> An updating, Paul:
> As I thought: WRF wnats NetCDF-3.6.2, not 4. Although I compiled
> NetCDF with the
> --disable-netcdf4 configure option.
> Some times ago, wrf helpdesk told me that WRF wants 3.6.2...
>
>
> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
>
> > Errata corrige, Paul:
> > about GNU compiler, maybe I have to set CXX=g++ (not gcc).
> > In this way, netcdf-4.1.1 seems work fine. I confirm, instead,
netcdf-3.6.2
> > has still problems.
> >
> > But now, I have to check if I can use netcdf version 4 with WRF
and the
> > other my software!
> >
> > Fabio.
> >
> >
> > 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
> >
> > Paul, if you want to visit Rome and be our guest (I live in
Orvieto,
> >> Umbria) you are welcome!
> >> I speak English just a little bit, I graduated in the Netherlands
> >> although now I have lost a bit of my English ..
> >> But It doesn't matter...! :)
> >>
> >> About GNU compiler. Sure!
> >> I started using Intel compiler because I had some problems with
WRF
model
> >> compiling, but maybe it's time to get back to GNU.
> >>
> >> So, in attachment the *netcdf-3.6.2* building by GNU. But as you
can
see
> >> I have similar problem!
> >> What do you think about?
> >>
> >> In the meantime I also tried* netcdf-4.1.1* (with the option
> >> --disable-netcdf-4 in the configure). Following my results:
> >> 1) GNU --> see attachments
> >> 2) INTEL --> see attachments
> >>
> >> Similar warning/errors both make and make check.
> >>
> >> About compilers the problem is (also) that usually I have to use
the
same
> >> compiler for each programs.
> >> My final target is to use the "chain": WRFDA (or GSI) -->
WPS/WRF-NMM/WPP
> >> --> MET.
> >> But GSI, for example, seems not "GNU" supported (only Intel or
PGI).
> >> With MET, instead, you suggest GNU but I have problems with both
> >> compilers.
> >>
> >> What must I do??? I don't give up, but it is very hard! :)
> >>
> >> Fabio.
> >>
> >> PS: I use Linux Fedora14 and GNU version is 4.5.1.
> >>
> >>
> >>
> >> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
> >>
> >>> Fabio,
> >>>
> >>> I was just kidding about in-person support.  We were just joking
that
we
> >>> would like to visit Rome.
> >>>
> >>> If you are the administrator, perhaps you have GNU compilers
installed
on
> >>> your machine?  You might want to try using
> >>> those to build the libraries and MET.  It is what we used to
develop
MET,
> >>> so we have the most experience with them.  You
> >>> may already have GNU compilers installed.
> >>>
> >>> Don't give up yet!  Once your compilers are configured
correctly,
> >>> building MET should just be a matter of following the
> >>> instructions on the online tutorial.
> >>>
> >>> Good luck.
> >>>
> >>> Paul
> >>>
> >>>
> >>> On 02/22/2011 02:40 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
> >>> >
> >>> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711
>
> >>> >
> >>> > It is really bad news, Paul!
> >>> > I am the administrator. I live near Rome, but I run a small
computer
> >>> center
> >>> > in a company that is located in Udine. Before upgrading the
computers
> >>> in
> >>> > Udine,
> >>> > I usually do experiments on my laptop (like right now).
> >>> >
> >>> > What about in-person support? Is it really necessary?
> >>> > Unfortunatley we no have budget to support it, at the moment.
> >>> >
> >>> > So the bad news is.. maybe I have to stop before I start!
> >>> >
> >>> > Fabio.
> >>> >
> >>> >
> >>> > 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
> >>> >
> >>> >> Fabio,
> >>> >>
> >>> >> It looks like your c++ compiler is not configured correctly.
The
> >>> error
> >>> >> message indicates that your compiler cannot find
> >>> >> a standard c++ library function (strncpy).  I think you may
want to
> >>> talk to
> >>> >> your administrator about the compiler
> >>> >> configuration.  This will be a problem when you try to
compile MET,
> >>> too.
> >>> >>
> >>> >> ncvalues.cpp(238): error: identifier "strncpy" is undefined
> >>> >>      strncpy(s, (const char*)the_values + n,
(int)the_number);
> >>> >>
> >>> >> We would like to provide in-person support to you in Rome.
Would
your
> >>> >> operations budget support that?
> >>> >>
> >>> >> Paul
> >>> >>
> >>> >>
> >>> >> On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
> >>> >>>
> >>> >>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >>> >>>
> >>> >>> I'm sorry Paul, I made a mistake:
> >>> >>>
> >>> >>> I remembered that when I compile netcdf, *I usually set
CXX='',
> >>> >> otherwhise I
> >>> >>> obtain an
> >>> >>> abort compilation, as you can see in attachment.*
> >>> >>> Again, now I remember had a similar problem with gcc (about
2
years
> >>> ago!)
> >>> >>>
> >>> >>> Apparently, for all the other programs do not set CXX is not
a
> >>> problem.
> >>> >>> I never solved that issue that now back to me forcefully!
> >>> >>>
> >>> >>> How can I fix it, in your opinion?
> >>> >>>
> >>> >>> Thanks,
> >>> >>> Fabio.
> >>> >>>
> >>> >>>
> >>> >>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
> >>> >>>
> >>> >>>> Fabio,
> >>> >>>>
> >>> >>>> The final step in the NetCDF should basically copy
cxx/netcdf.hh
to
> >>> >>>> include/netcdf.hh.  We suspect that there is
> >>> >>>> something funny going on with your configure and build
process.
 Can
> >>> you
> >>> >>>> please send us log files of your configuration
> >>> >>>> and build using the following commands:
> >>> >>>>
> >>> >>>> $ make clean
> >>> >>>> $ ./configure --prefix=/netcdf/install/path &>
nc_configure.log
> >>> >>>> $ make &> nc_make.log
> >>> >>>> $ make install &> nc_make_install.log
> >>> >>>>
> >>> >>>> Thanks,
> >>> >>>>
> >>> >>>> Paul
> >>> >>>>
> >>> >>>>
> >>> >>>> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
> >>> >>>>>
> >>> >>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >>> >>>>>
> >>> >>>>> Ok Paul,
> >>> >>>>>
> >>> >>>>> I removed my netcdf.hh copy.
> >>> >>>>>
> >>> >>>>> I'm following the tutorial and I understand about BUFRLIB,
NetCDF
> >>> and
> >>> >>>> GSL.
> >>> >>>>> Regarding BUFRLIB I'll ignore the ifort errors. GSL seems
built
ok.
> >>> >>>>>
> >>> >>>>> *About NetCDF, I understand that MET wants ".hh" but I
think
(but
> >>> maybe
> >>> >> I
> >>> >>>>> wrong) I compiled it correctly. My env settings are:*
> >>> >>>>> export CC=icc
> >>> >>>>> export FC=ifort
> >>> >>>>> export F9X=ifort
> >>> >>>>> export F90=ifort
> >>> >>>>> export F77=ifort
> >>> >>>>> export CXX=icc
> >>> >>>>>
> >>> >>>>> *Then I do:*
> >>> >>>>> ./configure --prefix=/home/master/netcdf-3.6.2
> >>> >>>>> make
> >>> >>>>> make install
> >>> >>>>>
> >>> >>>>> But *my building process doesn't produce any .hh file into
/include
> >>> >>>> folder,
> >>> >>>>> but only a .h file*.
> >>> >>>>> I built NetCDF several times in the past, and I never had
any
> >>> problems
> >>> >> so
> >>> >>>>> far with it, because WPS, WRF, WRFDA and so on are okay.
> >>> >>>>>
> >>> >>>>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx
folder,
so
> >>> I
> >>> >>>> tried
> >>> >>>>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
> >>> >>>>> on my makefile, but now I get:
> >>> >>>>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16): catastrophic
error:
> >>> could
> >>> >>>> not
> >>> >>>>> open source file "netcdf.h"
> >>> >>>>>   #include "netcdf.h"
> >>> >>>>>
> >>> >>>>> Sorry Paul, maybe I'm doing some stupid errors, but really
I
don't
> >>> >>>>> understand what's wrong on my NetCDF building!
> >>> >>>>>
> >>> >>>>> Thank you, Paul, for your patience,
> >>> >>>>> Fabio.
> >>> >>>>>
> >>> >>>>>
> >>> >>>>>
> >>> >>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
> >>> >>>>>
> >>> >>>>>> Fabio,
> >>> >>>>>>
> >>> >>>>>> Your solution of copying the NetCDF header file is
causing the
> >>> >> problem.
> >>> >>>>>>  The file include/netcdf.hh is generated when
> >>> >>>>>> the c/c++ interface of NetCDF is built.  The instance you
are
> >>> trying
> >>> >> to
> >>> >>>> use
> >>> >>>>>> may not have that interface built.
> >>> >>>>>>
> >>> >>>>>> We recommend that you download and build your own
instances of
the
> >>> >>>>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
> >>> >>>>>> that they are built in the way that MET expects.  The
process
> >>> should
> >>> >> be
> >>> >>>>>> pretty easy if you carefully read and follow the
> >>> >>>>>> instructions at the online tutorial I mentioned earlier,
using
the
> >>> >> intel
> >>> >>>>>> compilers for each step:
> >>> >>>>>>
> >>> >>>>>>
> >>> >>>>>>
> >>> >>>>
> >>> >>
> >>>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
> >>> >>>>>>
> >>> >>>>>> You may see some error messages when building BUFRLIB
when
> >>> executing
> >>> >>>> this
> >>> >>>>>> step:
> >>> >>>>>>
> >>> >>>>>> ifort -c -DUNDERSCORE *.f *.F
> >>> >>>>>>
> >>> >>>>>> Ignore these error messages and continue with the
process.  MET
> >>> should
> >>> >>>> work
> >>> >>>>>> with the partially built version of BUFRLIB.
> >>> >>>>>>  Please let us know if you have any other questions.
> >>> >>>>>>
> >>> >>>>>> Paul
> >>> >>>>>>
> >>> >>>>>>
> >>> >>>>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio
F.Gervasi}
wrote:
> >>> >>>>>>>
> >>> >>>>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
> >>> >>>>>>> Transaction: Ticket created by fabio.gervasi70 at gmail.com
> >>> >>>>>>>        Queue: met_help
> >>> >>>>>>>      Subject: Met building problems: compilation aborted
for
> >>> >>>>>> vx_contable_stats.cc
> >>> >>>>>>>        Owner: Nobody
> >>> >>>>>>>   Requestors: fabio.gervasi70 at gmail.com
> >>> >>>>>>>       Status: new
> >>> >>>>>>>  Ticket <URL:
> >>> >> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
> >>> >>>>>>>
> >>> >>>>>>>
> >>> >>>>>>> Hi,
> >>> >>>>>>>
> >>> >>>>>>> I open another post, but it could be a continuation
about
> >>> >>>>>>> my previous ticket #44675.
> >>> >>>>>>>
> >>> >>>>>>> In order to build MET on my Linux Intel machine (64bit),
> >>> >>>>>>> into ../netcdf../include I did as following:
> >>> >>>>>>> cp netcdf.h netcdf.hh.
> >>> >>>>>>> In this way I resolved my ".hh" problem.
> >>> >>>>>>>
> >>> >>>>>>> I also built succesfully GSL as well.
> >>> >>>>>>>
> >>> >>>>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so
on...)
is
> >>> built
> >>> >>>> by
> >>> >>>>>>> Intel compiler.*
> >>> >>>>>>>
> >>> >>>>>>> But when I "make", I get the errors in attachment.
> >>> >>>>>>> In attachment the makefile too.
> >>> >>>>>>>
> >>> >>>>>>> Where do I wrong, now?
> >>> >>>>>>>
> >>> >>>>>>> Thank you veru much,
> >>> >>>>>>> Fabio.
> >>> >>>>>>>
> >>> >>>>>>
> >>> >>>>>>
> >>> >>>>>>
> >>> >>>>
> >>> >>>>
> >>> >>>>
> >>> >>>
> >>> >>
> >>> >>
> >>> >>
> >>>
> >>>
> >>>
> >>
> >
>

------------------------------------------------
Subject: Met building problems: compilation aborted for vx_contable_stats.cc
From: Paul Oldenburg
Time: Wed Feb 23 09:53:53 2011

Fabio,

In response to your questions, I would like to make three points:

1.  It does not matter if you use intel compilers to compile
GSI/WRF/WPP and GNU compilers to compile MET.  MET only
reads the output of WPP, and it does not interact with any of the
other components of your "weather chain" in any way.
If you use different compilers for MET than for the other components,
it will not matter.

2.  If you have two separate instances of NetCDF - one for GSI/WRF/WPP
and one for MET, that is ok too.  MET will still
be able to read WPP output if you use different NetCDF instances for
each of them.  Actually, the output of WPP is GRIB,
so you should not have to worry about how MET uses NetCDF, which is
internal to MET.  MET will accept the GRIB output of
WPP.

3.  In make_gnu_met.log, the error message "/usr/bin/ld: cannot find
-lstdc++" indicates that your GNU g++ compiler
cannot find the stdc++ library, which should be part of the compiler
installation.  As an experiment, please try to
compile and run the attached test c++ program using the following
command:

$ g++ test.cpp
$ ./a.out

If this does not work, then you need to fix the GNU compiler
configuration so that it does.  This test.cpp is the
simplest possible c++ program.

Let me know if you have any questions.

Good luck,

Paul


On 02/23/2011 06:10 AM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>
> Hi Paul, since yesterday I wrote to you a lot of emails, so I'd like
to
> summarize the situation, after some tests.
>
> *When I use** GNU, env are:*
> export CC=gcc
> export FC=gfortran
> export F9X=gfortran
> export F90=gfortran
> export F77=gfortran
> export CXX=g++
>
> *When I use Intel, env are:*
> export CC=icc
> export FC=ifort
> export F9X=ifort
> export F90=ifort
> export F77=ifort
> export CXX=icpc
>
> *1)* Both GNU and Intel compilers, I'm able to build GSL and BUFRLIB
(for
> bufrlib I ignored ifort errors, about gnu I replaced nvnwin.f file)
> *2)* Both GNU and Intel compilers, I'm able to build NetCDF-4.1.1
with CXX
> *3)* About MET now I get the logs in attachments for both compilers
(errors,
> but gnu seems better then intel). In attachment the gnu and intel
makefile
> that I'm using, too.
> *4)* Anyway, unfortunately I'm not able to compile my WRF-NMM model
> with NetCDF-4.1.1
> (also using --disable-netcdf4 option). I'm able to build WRF-NMM
only using
> NetCDF-3.6.2. Can I, in case, to use Net3 with WRF and Net4 with MET
at the
> same time?
> *5)* About the choosing between GNU or Intel, I have advantages and
> disvantages. For example at the moment I have some problems with GNU
> building WRF model (but I think helpdesk can help me). But at the
same time,
> GSI seems supports better Intel.
>
> As I can told you previously, I need to do the "best choice" between
GNU or
> INTEL, taking into account the chain:
> *GSI (or WRFDA) --> WPS/WRF-NMM/WPP --> MET*, and and related
library...
>
> I hope that my situation is more clear now.
> Let me know, if possible, what you think about it.
>
> Ciao!
> Fabio.
>
>
>
> 2011/2/23 RAL HelpDesk {for Fabio F.Gervasi} <met_help at ucar.edu>
>>
>> An updating, Paul:
>> As I thought: WRF wnats NetCDF-3.6.2, not 4. Although I compiled
>> NetCDF with the
>> --disable-netcdf4 configure option.
>> Some times ago, wrf helpdesk told me that WRF wants 3.6.2...
>>
>>
>> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
>>
>>> Errata corrige, Paul:
>>> about GNU compiler, maybe I have to set CXX=g++ (not gcc).
>>> In this way, netcdf-4.1.1 seems work fine. I confirm, instead,
> netcdf-3.6.2
>>> has still problems.
>>>
>>> But now, I have to check if I can use netcdf version 4 with WRF
and the
>>> other my software!
>>>
>>> Fabio.
>>>
>>>
>>> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
>>>
>>> Paul, if you want to visit Rome and be our guest (I live in
Orvieto,
>>>> Umbria) you are welcome!
>>>> I speak English just a little bit, I graduated in the Netherlands
>>>> although now I have lost a bit of my English ..
>>>> But It doesn't matter...! :)
>>>>
>>>> About GNU compiler. Sure!
>>>> I started using Intel compiler because I had some problems with
WRF
> model
>>>> compiling, but maybe it's time to get back to GNU.
>>>>
>>>> So, in attachment the *netcdf-3.6.2* building by GNU. But as you
can
> see
>>>> I have similar problem!
>>>> What do you think about?
>>>>
>>>> In the meantime I also tried* netcdf-4.1.1* (with the option
>>>> --disable-netcdf-4 in the configure). Following my results:
>>>> 1) GNU --> see attachments
>>>> 2) INTEL --> see attachments
>>>>
>>>> Similar warning/errors both make and make check.
>>>>
>>>> About compilers the problem is (also) that usually I have to use
the
> same
>>>> compiler for each programs.
>>>> My final target is to use the "chain": WRFDA (or GSI) -->
> WPS/WRF-NMM/WPP
>>>> --> MET.
>>>> But GSI, for example, seems not "GNU" supported (only Intel or
PGI).
>>>> With MET, instead, you suggest GNU but I have problems with both
>>>> compilers.
>>>>
>>>> What must I do??? I don't give up, but it is very hard! :)
>>>>
>>>> Fabio.
>>>>
>>>> PS: I use Linux Fedora14 and GNU version is 4.5.1.
>>>>
>>>>
>>>>
>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>>>>
>>>>> Fabio,
>>>>>
>>>>> I was just kidding about in-person support.  We were just joking
that
> we
>>>>> would like to visit Rome.
>>>>>
>>>>> If you are the administrator, perhaps you have GNU compilers
installed
> on
>>>>> your machine?  You might want to try using
>>>>> those to build the libraries and MET.  It is what we used to
develop
> MET,
>>>>> so we have the most experience with them.  You
>>>>> may already have GNU compilers installed.
>>>>>
>>>>> Don't give up yet!  Once your compilers are configured
correctly,
>>>>> building MET should just be a matter of following the
>>>>> instructions on the online tutorial.
>>>>>
>>>>> Good luck.
>>>>>
>>>>> Paul
>>>>>
>>>>>
>>>>> On 02/22/2011 02:40 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>>>>>
>>>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>>>>
>>>>>> It is really bad news, Paul!
>>>>>> I am the administrator. I live near Rome, but I run a small
computer
>>>>> center
>>>>>> in a company that is located in Udine. Before upgrading the
> computers
>>>>> in
>>>>>> Udine,
>>>>>> I usually do experiments on my laptop (like right now).
>>>>>>
>>>>>> What about in-person support? Is it really necessary?
>>>>>> Unfortunatley we no have budget to support it, at the moment.
>>>>>>
>>>>>> So the bad news is.. maybe I have to stop before I start!
>>>>>>
>>>>>> Fabio.
>>>>>>
>>>>>>
>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>>>>>>
>>>>>>> Fabio,
>>>>>>>
>>>>>>> It looks like your c++ compiler is not configured correctly.
The
>>>>> error
>>>>>>> message indicates that your compiler cannot find
>>>>>>> a standard c++ library function (strncpy).  I think you may
want to
>>>>> talk to
>>>>>>> your administrator about the compiler
>>>>>>> configuration.  This will be a problem when you try to compile
MET,
>>>>> too.
>>>>>>>
>>>>>>> ncvalues.cpp(238): error: identifier "strncpy" is undefined
>>>>>>>      strncpy(s, (const char*)the_values + n, (int)the_number);
>>>>>>>
>>>>>>> We would like to provide in-person support to you in Rome.
Would
> your
>>>>>>> operations budget support that?
>>>>>>>
>>>>>>> Paul
>>>>>>>
>>>>>>>
>>>>>>> On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>>>>>>>
>>>>>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711
>
>>>>>>>>
>>>>>>>> I'm sorry Paul, I made a mistake:
>>>>>>>>
>>>>>>>> I remembered that when I compile netcdf, *I usually set
CXX='',
>>>>>>> otherwhise I
>>>>>>>> obtain an
>>>>>>>> abort compilation, as you can see in attachment.*
>>>>>>>> Again, now I remember had a similar problem with gcc (about 2
> years
>>>>> ago!)
>>>>>>>>
>>>>>>>> Apparently, for all the other programs do not set CXX is not
a
>>>>> problem.
>>>>>>>> I never solved that issue that now back to me forcefully!
>>>>>>>>
>>>>>>>> How can I fix it, in your opinion?
>>>>>>>>
>>>>>>>> Thanks,
>>>>>>>> Fabio.
>>>>>>>>
>>>>>>>>
>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
>>>>>>>>
>>>>>>>>> Fabio,
>>>>>>>>>
>>>>>>>>> The final step in the NetCDF should basically copy
cxx/netcdf.hh
> to
>>>>>>>>> include/netcdf.hh.  We suspect that there is
>>>>>>>>> something funny going on with your configure and build
process.
>  Can
>>>>> you
>>>>>>>>> please send us log files of your configuration
>>>>>>>>> and build using the following commands:
>>>>>>>>>
>>>>>>>>> $ make clean
>>>>>>>>> $ ./configure --prefix=/netcdf/install/path &>
nc_configure.log
>>>>>>>>> $ make &> nc_make.log
>>>>>>>>> $ make install &> nc_make_install.log
>>>>>>>>>
>>>>>>>>> Thanks,
>>>>>>>>>
>>>>>>>>> Paul
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>>>>>>>>>
>>>>>>>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>>>>>>>>
>>>>>>>>>> Ok Paul,
>>>>>>>>>>
>>>>>>>>>> I removed my netcdf.hh copy.
>>>>>>>>>>
>>>>>>>>>> I'm following the tutorial and I understand about BUFRLIB,
> NetCDF
>>>>> and
>>>>>>>>> GSL.
>>>>>>>>>> Regarding BUFRLIB I'll ignore the ifort errors. GSL seems
built
> ok.
>>>>>>>>>>
>>>>>>>>>> *About NetCDF, I understand that MET wants ".hh" but I
think
> (but
>>>>> maybe
>>>>>>> I
>>>>>>>>>> wrong) I compiled it correctly. My env settings are:*
>>>>>>>>>> export CC=icc
>>>>>>>>>> export FC=ifort
>>>>>>>>>> export F9X=ifort
>>>>>>>>>> export F90=ifort
>>>>>>>>>> export F77=ifort
>>>>>>>>>> export CXX=icc
>>>>>>>>>>
>>>>>>>>>> *Then I do:*
>>>>>>>>>> ./configure --prefix=/home/master/netcdf-3.6.2
>>>>>>>>>> make
>>>>>>>>>> make install
>>>>>>>>>>
>>>>>>>>>> But *my building process doesn't produce any .hh file into
> /include
>>>>>>>>> folder,
>>>>>>>>>> but only a .h file*.
>>>>>>>>>> I built NetCDF several times in the past, and I never had
any
>>>>> problems
>>>>>>> so
>>>>>>>>>> far with it, because WPS, WRF, WRFDA and so on are okay.
>>>>>>>>>>
>>>>>>>>>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx
folder,
> so
>>>>> I
>>>>>>>>> tried
>>>>>>>>>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
>>>>>>>>>> on my makefile, but now I get:
>>>>>>>>>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16): catastrophic
> error:
>>>>> could
>>>>>>>>> not
>>>>>>>>>> open source file "netcdf.h"
>>>>>>>>>>   #include "netcdf.h"
>>>>>>>>>>
>>>>>>>>>> Sorry Paul, maybe I'm doing some stupid errors, but really
I
> don't
>>>>>>>>>> understand what's wrong on my NetCDF building!
>>>>>>>>>>
>>>>>>>>>> Thank you, Paul, for your patience,
>>>>>>>>>> Fabio.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
>>>>>>>>>>
>>>>>>>>>>> Fabio,
>>>>>>>>>>>
>>>>>>>>>>> Your solution of copying the NetCDF header file is causing
the
>>>>>>> problem.
>>>>>>>>>>>  The file include/netcdf.hh is generated when
>>>>>>>>>>> the c/c++ interface of NetCDF is built.  The instance you
are
>>>>> trying
>>>>>>> to
>>>>>>>>> use
>>>>>>>>>>> may not have that interface built.
>>>>>>>>>>>
>>>>>>>>>>> We recommend that you download and build your own
instances of
> the
>>>>>>>>>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
>>>>>>>>>>> that they are built in the way that MET expects.  The
process
>>>>> should
>>>>>>> be
>>>>>>>>>>> pretty easy if you carefully read and follow the
>>>>>>>>>>> instructions at the online tutorial I mentioned earlier,
using
> the
>>>>>>> intel
>>>>>>>>>>> compilers for each step:
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>
>>>>>>>
>>>>>
>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
>>>>>>>>>>>
>>>>>>>>>>> You may see some error messages when building BUFRLIB when
>>>>> executing
>>>>>>>>> this
>>>>>>>>>>> step:
>>>>>>>>>>>
>>>>>>>>>>> ifort -c -DUNDERSCORE *.f *.F
>>>>>>>>>>>
>>>>>>>>>>> Ignore these error messages and continue with the process.
MET
>>>>> should
>>>>>>>>> work
>>>>>>>>>>> with the partially built version of BUFRLIB.
>>>>>>>>>>>  Please let us know if you have any other questions.
>>>>>>>>>>>
>>>>>>>>>>> Paul
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio F.Gervasi}
> wrote:
>>>>>>>>>>>>
>>>>>>>>>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
>>>>>>>>>>>> Transaction: Ticket created by fabio.gervasi70 at gmail.com
>>>>>>>>>>>>        Queue: met_help
>>>>>>>>>>>>      Subject: Met building problems: compilation aborted
for
>>>>>>>>>>> vx_contable_stats.cc
>>>>>>>>>>>>        Owner: Nobody
>>>>>>>>>>>>   Requestors: fabio.gervasi70 at gmail.com
>>>>>>>>>>>>       Status: new
>>>>>>>>>>>>  Ticket <URL:
>>>>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> Hi,
>>>>>>>>>>>>
>>>>>>>>>>>> I open another post, but it could be a continuation about
>>>>>>>>>>>> my previous ticket #44675.
>>>>>>>>>>>>
>>>>>>>>>>>> In order to build MET on my Linux Intel machine (64bit),
>>>>>>>>>>>> into ../netcdf../include I did as following:
>>>>>>>>>>>> cp netcdf.h netcdf.hh.
>>>>>>>>>>>> In this way I resolved my ".hh" problem.
>>>>>>>>>>>>
>>>>>>>>>>>> I also built succesfully GSL as well.
>>>>>>>>>>>>
>>>>>>>>>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so
on...)
> is
>>>>> built
>>>>>>>>> by
>>>>>>>>>>>> Intel compiler.*
>>>>>>>>>>>>
>>>>>>>>>>>> But when I "make", I get the errors in attachment.
>>>>>>>>>>>> In attachment the makefile too.
>>>>>>>>>>>>
>>>>>>>>>>>> Where do I wrong, now?
>>>>>>>>>>>>
>>>>>>>>>>>> Thank you veru much,
>>>>>>>>>>>> Fabio.
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>
>>
>


------------------------------------------------
Subject: Met building problems: compilation aborted for vx_contable_stats.cc
From: Fabio F.Gervasi
Time: Wed Feb 23 14:20:44 2011

Great Paul!

I'm glad to hear that at least MET is independent from other
programsregarding
NetCDF and compilers.

So, I'm going to proceed with GNU. Anyway, in the meantime, I
discovered
that the NetCDF-3.6.3  (instead of 3.6.2) *builds succesfully about
CXX and
also it is right for WPS/WRF/WPP!*

Regarding stdc++ library, I installed "libstdc++-static.x86_64"
package,
then I did:
[root at aspire5920g master]# g++ test.cpp
[root at aspire5920g master]# ./a.out
hello world

So your test works fine. Now:

[root at aspire5920g MET]# make >& make_met.log --> *EUREKA! 14
executable
created.
*
So now I'll start the case study in order to learn about MET.

Thank you very much Paul for your support.
I'll write again about other problems, but we can close this topic.
Anyway you are welcome in Italy, and I'm not kidding. :)

Fabio.


2011/2/23 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>

> Fabio,
>
> In response to your questions, I would like to make three points:
>
> 1.  It does not matter if you use intel compilers to compile
GSI/WRF/WPP
> and GNU compilers to compile MET.  MET only
> reads the output of WPP, and it does not interact with any of the
other
> components of your "weather chain" in any way.
> If you use different compilers for MET than for the other
components, it
> will not matter.
>
> 2.  If you have two separate instances of NetCDF - one for
GSI/WRF/WPP and
> one for MET, that is ok too.  MET will still
> be able to read WPP output if you use different NetCDF instances for
each
> of them.  Actually, the output of WPP is GRIB,
> so you should not have to worry about how MET uses NetCDF, which is
> internal to MET.  MET will accept the GRIB output of
> WPP.
>
> 3.  In make_gnu_met.log, the error message "/usr/bin/ld: cannot find
> -lstdc++" indicates that your GNU g++ compiler
> cannot find the stdc++ library, which should be part of the compiler
> installation.  As an experiment, please try to
> compile and run the attached test c++ program using the following
command:
>
> $ g++ test.cpp
> $ ./a.out
>
> If this does not work, then you need to fix the GNU compiler
configuration
> so that it does.  This test.cpp is the
> simplest possible c++ program.
>
> Let me know if you have any questions.
>
> Good luck,
>
> Paul
>
>
> On 02/23/2011 06:10 AM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >
> > Hi Paul, since yesterday I wrote to you a lot of emails, so I'd
like to
> > summarize the situation, after some tests.
> >
> > *When I use** GNU, env are:*
> > export CC=gcc
> > export FC=gfortran
> > export F9X=gfortran
> > export F90=gfortran
> > export F77=gfortran
> > export CXX=g++
> >
> > *When I use Intel, env are:*
> > export CC=icc
> > export FC=ifort
> > export F9X=ifort
> > export F90=ifort
> > export F77=ifort
> > export CXX=icpc
> >
> > *1)* Both GNU and Intel compilers, I'm able to build GSL and
BUFRLIB (for
> > bufrlib I ignored ifort errors, about gnu I replaced nvnwin.f
file)
> > *2)* Both GNU and Intel compilers, I'm able to build NetCDF-4.1.1
with
> CXX
> > *3)* About MET now I get the logs in attachments for both
compilers
> (errors,
> > but gnu seems better then intel). In attachment the gnu and intel
> makefile
> > that I'm using, too.
> > *4)* Anyway, unfortunately I'm not able to compile my WRF-NMM
model
> > with NetCDF-4.1.1
> > (also using --disable-netcdf4 option). I'm able to build WRF-NMM
only
> using
> > NetCDF-3.6.2. Can I, in case, to use Net3 with WRF and Net4 with
MET at
> the
> > same time?
> > *5)* About the choosing between GNU or Intel, I have advantages
and
> > disvantages. For example at the moment I have some problems with
GNU
> > building WRF model (but I think helpdesk can help me). But at the
same
> time,
> > GSI seems supports better Intel.
> >
> > As I can told you previously, I need to do the "best choice"
between GNU
> or
> > INTEL, taking into account the chain:
> > *GSI (or WRFDA) --> WPS/WRF-NMM/WPP --> MET*, and and related
library...
> >
> > I hope that my situation is more clear now.
> > Let me know, if possible, what you think about it.
> >
> > Ciao!
> > Fabio.
> >
> >
> >
> > 2011/2/23 RAL HelpDesk {for Fabio F.Gervasi} <met_help at ucar.edu>
> >>
> >> An updating, Paul:
> >> As I thought: WRF wnats NetCDF-3.6.2, not 4. Although I compiled
> >> NetCDF with the
> >> --disable-netcdf4 configure option.
> >> Some times ago, wrf helpdesk told me that WRF wants 3.6.2...
> >>
> >>
> >> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
> >>
> >>> Errata corrige, Paul:
> >>> about GNU compiler, maybe I have to set CXX=g++ (not gcc).
> >>> In this way, netcdf-4.1.1 seems work fine. I confirm, instead,
> > netcdf-3.6.2
> >>> has still problems.
> >>>
> >>> But now, I have to check if I can use netcdf version 4 with WRF
and the
> >>> other my software!
> >>>
> >>> Fabio.
> >>>
> >>>
> >>> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
> >>>
> >>> Paul, if you want to visit Rome and be our guest (I live in
Orvieto,
> >>>> Umbria) you are welcome!
> >>>> I speak English just a little bit, I graduated in the
Netherlands
> >>>> although now I have lost a bit of my English ..
> >>>> But It doesn't matter...! :)
> >>>>
> >>>> About GNU compiler. Sure!
> >>>> I started using Intel compiler because I had some problems with
WRF
> > model
> >>>> compiling, but maybe it's time to get back to GNU.
> >>>>
> >>>> So, in attachment the *netcdf-3.6.2* building by GNU. But as
you can
> > see
> >>>> I have similar problem!
> >>>> What do you think about?
> >>>>
> >>>> In the meantime I also tried* netcdf-4.1.1* (with the option
> >>>> --disable-netcdf-4 in the configure). Following my results:
> >>>> 1) GNU --> see attachments
> >>>> 2) INTEL --> see attachments
> >>>>
> >>>> Similar warning/errors both make and make check.
> >>>>
> >>>> About compilers the problem is (also) that usually I have to
use the
> > same
> >>>> compiler for each programs.
> >>>> My final target is to use the "chain": WRFDA (or GSI) -->
> > WPS/WRF-NMM/WPP
> >>>> --> MET.
> >>>> But GSI, for example, seems not "GNU" supported (only Intel or
PGI).
> >>>> With MET, instead, you suggest GNU but I have problems with
both
> >>>> compilers.
> >>>>
> >>>> What must I do??? I don't give up, but it is very hard! :)
> >>>>
> >>>> Fabio.
> >>>>
> >>>> PS: I use Linux Fedora14 and GNU version is 4.5.1.
> >>>>
> >>>>
> >>>>
> >>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
> >>>>
> >>>>> Fabio,
> >>>>>
> >>>>> I was just kidding about in-person support.  We were just
joking that
> > we
> >>>>> would like to visit Rome.
> >>>>>
> >>>>> If you are the administrator, perhaps you have GNU compilers
> installed
> > on
> >>>>> your machine?  You might want to try using
> >>>>> those to build the libraries and MET.  It is what we used to
develop
> > MET,
> >>>>> so we have the most experience with them.  You
> >>>>> may already have GNU compilers installed.
> >>>>>
> >>>>> Don't give up yet!  Once your compilers are configured
correctly,
> >>>>> building MET should just be a matter of following the
> >>>>> instructions on the online tutorial.
> >>>>>
> >>>>> Good luck.
> >>>>>
> >>>>> Paul
> >>>>>
> >>>>>
> >>>>> On 02/22/2011 02:40 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
> >>>>>>
> >>>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711
>
> >>>>>>
> >>>>>> It is really bad news, Paul!
> >>>>>> I am the administrator. I live near Rome, but I run a small
computer
> >>>>> center
> >>>>>> in a company that is located in Udine. Before upgrading the
> > computers
> >>>>> in
> >>>>>> Udine,
> >>>>>> I usually do experiments on my laptop (like right now).
> >>>>>>
> >>>>>> What about in-person support? Is it really necessary?
> >>>>>> Unfortunatley we no have budget to support it, at the moment.
> >>>>>>
> >>>>>> So the bad news is.. maybe I have to stop before I start!
> >>>>>>
> >>>>>> Fabio.
> >>>>>>
> >>>>>>
> >>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
> >>>>>>
> >>>>>>> Fabio,
> >>>>>>>
> >>>>>>> It looks like your c++ compiler is not configured correctly.
The
> >>>>> error
> >>>>>>> message indicates that your compiler cannot find
> >>>>>>> a standard c++ library function (strncpy).  I think you may
want to
> >>>>> talk to
> >>>>>>> your administrator about the compiler
> >>>>>>> configuration.  This will be a problem when you try to
compile MET,
> >>>>> too.
> >>>>>>>
> >>>>>>> ncvalues.cpp(238): error: identifier "strncpy" is undefined
> >>>>>>>      strncpy(s, (const char*)the_values + n,
(int)the_number);
> >>>>>>>
> >>>>>>> We would like to provide in-person support to you in Rome.
Would
> > your
> >>>>>>> operations budget support that?
> >>>>>>>
> >>>>>>> Paul
> >>>>>>>
> >>>>>>>
> >>>>>>> On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
> >>>>>>>>
> >>>>>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >>>>>>>>
> >>>>>>>> I'm sorry Paul, I made a mistake:
> >>>>>>>>
> >>>>>>>> I remembered that when I compile netcdf, *I usually set
CXX='',
> >>>>>>> otherwhise I
> >>>>>>>> obtain an
> >>>>>>>> abort compilation, as you can see in attachment.*
> >>>>>>>> Again, now I remember had a similar problem with gcc (about
2
> > years
> >>>>> ago!)
> >>>>>>>>
> >>>>>>>> Apparently, for all the other programs do not set CXX is
not a
> >>>>> problem.
> >>>>>>>> I never solved that issue that now back to me forcefully!
> >>>>>>>>
> >>>>>>>> How can I fix it, in your opinion?
> >>>>>>>>
> >>>>>>>> Thanks,
> >>>>>>>> Fabio.
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
> >>>>>>>>
> >>>>>>>>> Fabio,
> >>>>>>>>>
> >>>>>>>>> The final step in the NetCDF should basically copy
cxx/netcdf.hh
> > to
> >>>>>>>>> include/netcdf.hh.  We suspect that there is
> >>>>>>>>> something funny going on with your configure and build
process.
> >  Can
> >>>>> you
> >>>>>>>>> please send us log files of your configuration
> >>>>>>>>> and build using the following commands:
> >>>>>>>>>
> >>>>>>>>> $ make clean
> >>>>>>>>> $ ./configure --prefix=/netcdf/install/path &>
nc_configure.log
> >>>>>>>>> $ make &> nc_make.log
> >>>>>>>>> $ make install &> nc_make_install.log
> >>>>>>>>>
> >>>>>>>>> Thanks,
> >>>>>>>>>
> >>>>>>>>> Paul
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
> >>>>>>>>>>
> >>>>>>>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >>>>>>>>>>
> >>>>>>>>>> Ok Paul,
> >>>>>>>>>>
> >>>>>>>>>> I removed my netcdf.hh copy.
> >>>>>>>>>>
> >>>>>>>>>> I'm following the tutorial and I understand about
BUFRLIB,
> > NetCDF
> >>>>> and
> >>>>>>>>> GSL.
> >>>>>>>>>> Regarding BUFRLIB I'll ignore the ifort errors. GSL seems
built
> > ok.
> >>>>>>>>>>
> >>>>>>>>>> *About NetCDF, I understand that MET wants ".hh" but I
think
> > (but
> >>>>> maybe
> >>>>>>> I
> >>>>>>>>>> wrong) I compiled it correctly. My env settings are:*
> >>>>>>>>>> export CC=icc
> >>>>>>>>>> export FC=ifort
> >>>>>>>>>> export F9X=ifort
> >>>>>>>>>> export F90=ifort
> >>>>>>>>>> export F77=ifort
> >>>>>>>>>> export CXX=icc
> >>>>>>>>>>
> >>>>>>>>>> *Then I do:*
> >>>>>>>>>> ./configure --prefix=/home/master/netcdf-3.6.2
> >>>>>>>>>> make
> >>>>>>>>>> make install
> >>>>>>>>>>
> >>>>>>>>>> But *my building process doesn't produce any .hh file
into
> > /include
> >>>>>>>>> folder,
> >>>>>>>>>> but only a .h file*.
> >>>>>>>>>> I built NetCDF several times in the past, and I never had
any
> >>>>> problems
> >>>>>>> so
> >>>>>>>>>> far with it, because WPS, WRF, WRFDA and so on are okay.
> >>>>>>>>>>
> >>>>>>>>>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx
folder,
> > so
> >>>>> I
> >>>>>>>>> tried
> >>>>>>>>>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
> >>>>>>>>>> on my makefile, but now I get:
> >>>>>>>>>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16):
catastrophic
> > error:
> >>>>> could
> >>>>>>>>> not
> >>>>>>>>>> open source file "netcdf.h"
> >>>>>>>>>>   #include "netcdf.h"
> >>>>>>>>>>
> >>>>>>>>>> Sorry Paul, maybe I'm doing some stupid errors, but
really I
> > don't
> >>>>>>>>>> understand what's wrong on my NetCDF building!
> >>>>>>>>>>
> >>>>>>>>>> Thank you, Paul, for your patience,
> >>>>>>>>>> Fabio.
> >>>>>>>>>>
> >>>>>>>>>>
> >>>>>>>>>>
> >>>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
> >>>>>>>>>>
> >>>>>>>>>>> Fabio,
> >>>>>>>>>>>
> >>>>>>>>>>> Your solution of copying the NetCDF header file is
causing the
> >>>>>>> problem.
> >>>>>>>>>>>  The file include/netcdf.hh is generated when
> >>>>>>>>>>> the c/c++ interface of NetCDF is built.  The instance
you are
> >>>>> trying
> >>>>>>> to
> >>>>>>>>> use
> >>>>>>>>>>> may not have that interface built.
> >>>>>>>>>>>
> >>>>>>>>>>> We recommend that you download and build your own
instances of
> > the
> >>>>>>>>>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
> >>>>>>>>>>> that they are built in the way that MET expects.  The
process
> >>>>> should
> >>>>>>> be
> >>>>>>>>>>> pretty easy if you carefully read and follow the
> >>>>>>>>>>> instructions at the online tutorial I mentioned earlier,
using
> > the
> >>>>>>> intel
> >>>>>>>>>>> compilers for each step:
> >>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>
> >>>>>>>
> >>>>>
> >
>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
> >>>>>>>>>>>
> >>>>>>>>>>> You may see some error messages when building BUFRLIB
when
> >>>>> executing
> >>>>>>>>> this
> >>>>>>>>>>> step:
> >>>>>>>>>>>
> >>>>>>>>>>> ifort -c -DUNDERSCORE *.f *.F
> >>>>>>>>>>>
> >>>>>>>>>>> Ignore these error messages and continue with the
process.  MET
> >>>>> should
> >>>>>>>>> work
> >>>>>>>>>>> with the partially built version of BUFRLIB.
> >>>>>>>>>>>  Please let us know if you have any other questions.
> >>>>>>>>>>>
> >>>>>>>>>>> Paul
> >>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio
F.Gervasi}
> > wrote:
> >>>>>>>>>>>>
> >>>>>>>>>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
> >>>>>>>>>>>> Transaction: Ticket created by
fabio.gervasi70 at gmail.com
> >>>>>>>>>>>>        Queue: met_help
> >>>>>>>>>>>>      Subject: Met building problems: compilation
aborted for
> >>>>>>>>>>> vx_contable_stats.cc
> >>>>>>>>>>>>        Owner: Nobody
> >>>>>>>>>>>>   Requestors: fabio.gervasi70 at gmail.com
> >>>>>>>>>>>>       Status: new
> >>>>>>>>>>>>  Ticket <URL:
> >>>>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
> >>>>>>>>>>>>
> >>>>>>>>>>>>
> >>>>>>>>>>>> Hi,
> >>>>>>>>>>>>
> >>>>>>>>>>>> I open another post, but it could be a continuation
about
> >>>>>>>>>>>> my previous ticket #44675.
> >>>>>>>>>>>>
> >>>>>>>>>>>> In order to build MET on my Linux Intel machine
(64bit),
> >>>>>>>>>>>> into ../netcdf../include I did as following:
> >>>>>>>>>>>> cp netcdf.h netcdf.hh.
> >>>>>>>>>>>> In this way I resolved my ".hh" problem.
> >>>>>>>>>>>>
> >>>>>>>>>>>> I also built succesfully GSL as well.
> >>>>>>>>>>>>
> >>>>>>>>>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so
on...)
> > is
> >>>>> built
> >>>>>>>>> by
> >>>>>>>>>>>> Intel compiler.*
> >>>>>>>>>>>>
> >>>>>>>>>>>> But when I "make", I get the errors in attachment.
> >>>>>>>>>>>> In attachment the makefile too.
> >>>>>>>>>>>>
> >>>>>>>>>>>> Where do I wrong, now?
> >>>>>>>>>>>>
> >>>>>>>>>>>> Thank you veru much,
> >>>>>>>>>>>> Fabio.
> >>>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>
> >>>
> >>
> >
>
>
>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #44711] Met building problems: compilation aborted for vx_contable_stats.cc
From: Paul Oldenburg
Time: Wed Feb 23 14:25:44 2011

Fabio,

Please make sure to run the test scripts in the scripts directory.  In
particular, test_all.sh will exercise all of the
MET executables to make sure they run.

Good luck,

Paul


On 02/23/2011 02:20 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>
> Great Paul!
>
> I'm glad to hear that at least MET is independent from other
programsregarding
> NetCDF and compilers.
>
> So, I'm going to proceed with GNU. Anyway, in the meantime, I
discovered
> that the NetCDF-3.6.3  (instead of 3.6.2) *builds succesfully about
CXX and
> also it is right for WPS/WRF/WPP!*
>
> Regarding stdc++ library, I installed "libstdc++-static.x86_64"
package,
> then I did:
> [root at aspire5920g master]# g++ test.cpp
> [root at aspire5920g master]# ./a.out
> hello world
>
> So your test works fine. Now:
>
> [root at aspire5920g MET]# make >& make_met.log --> *EUREKA! 14
executable
> created.
> *
> So now I'll start the case study in order to learn about MET.
>
> Thank you very much Paul for your support.
> I'll write again about other problems, but we can close this topic.
> Anyway you are welcome in Italy, and I'm not kidding. :)
>
> Fabio.
>
>
> 2011/2/23 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>
>> Fabio,
>>
>> In response to your questions, I would like to make three points:
>>
>> 1.  It does not matter if you use intel compilers to compile
GSI/WRF/WPP
>> and GNU compilers to compile MET.  MET only
>> reads the output of WPP, and it does not interact with any of the
other
>> components of your "weather chain" in any way.
>> If you use different compilers for MET than for the other
components, it
>> will not matter.
>>
>> 2.  If you have two separate instances of NetCDF - one for
GSI/WRF/WPP and
>> one for MET, that is ok too.  MET will still
>> be able to read WPP output if you use different NetCDF instances
for each
>> of them.  Actually, the output of WPP is GRIB,
>> so you should not have to worry about how MET uses NetCDF, which is
>> internal to MET.  MET will accept the GRIB output of
>> WPP.
>>
>> 3.  In make_gnu_met.log, the error message "/usr/bin/ld: cannot
find
>> -lstdc++" indicates that your GNU g++ compiler
>> cannot find the stdc++ library, which should be part of the
compiler
>> installation.  As an experiment, please try to
>> compile and run the attached test c++ program using the following
command:
>>
>> $ g++ test.cpp
>> $ ./a.out
>>
>> If this does not work, then you need to fix the GNU compiler
configuration
>> so that it does.  This test.cpp is the
>> simplest possible c++ program.
>>
>> Let me know if you have any questions.
>>
>> Good luck,
>>
>> Paul
>>
>>
>> On 02/23/2011 06:10 AM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>
>>> Hi Paul, since yesterday I wrote to you a lot of emails, so I'd
like to
>>> summarize the situation, after some tests.
>>>
>>> *When I use** GNU, env are:*
>>> export CC=gcc
>>> export FC=gfortran
>>> export F9X=gfortran
>>> export F90=gfortran
>>> export F77=gfortran
>>> export CXX=g++
>>>
>>> *When I use Intel, env are:*
>>> export CC=icc
>>> export FC=ifort
>>> export F9X=ifort
>>> export F90=ifort
>>> export F77=ifort
>>> export CXX=icpc
>>>
>>> *1)* Both GNU and Intel compilers, I'm able to build GSL and
BUFRLIB (for
>>> bufrlib I ignored ifort errors, about gnu I replaced nvnwin.f
file)
>>> *2)* Both GNU and Intel compilers, I'm able to build NetCDF-4.1.1
with
>> CXX
>>> *3)* About MET now I get the logs in attachments for both
compilers
>> (errors,
>>> but gnu seems better then intel). In attachment the gnu and intel
>> makefile
>>> that I'm using, too.
>>> *4)* Anyway, unfortunately I'm not able to compile my WRF-NMM
model
>>> with NetCDF-4.1.1
>>> (also using --disable-netcdf4 option). I'm able to build WRF-NMM
only
>> using
>>> NetCDF-3.6.2. Can I, in case, to use Net3 with WRF and Net4 with
MET at
>> the
>>> same time?
>>> *5)* About the choosing between GNU or Intel, I have advantages
and
>>> disvantages. For example at the moment I have some problems with
GNU
>>> building WRF model (but I think helpdesk can help me). But at the
same
>> time,
>>> GSI seems supports better Intel.
>>>
>>> As I can told you previously, I need to do the "best choice"
between GNU
>> or
>>> INTEL, taking into account the chain:
>>> *GSI (or WRFDA) --> WPS/WRF-NMM/WPP --> MET*, and and related
library...
>>>
>>> I hope that my situation is more clear now.
>>> Let me know, if possible, what you think about it.
>>>
>>> Ciao!
>>> Fabio.
>>>
>>>
>>>
>>> 2011/2/23 RAL HelpDesk {for Fabio F.Gervasi} <met_help at ucar.edu>
>>>>
>>>> An updating, Paul:
>>>> As I thought: WRF wnats NetCDF-3.6.2, not 4. Although I compiled
>>>> NetCDF with the
>>>> --disable-netcdf4 configure option.
>>>> Some times ago, wrf helpdesk told me that WRF wants 3.6.2...
>>>>
>>>>
>>>> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
>>>>
>>>>> Errata corrige, Paul:
>>>>> about GNU compiler, maybe I have to set CXX=g++ (not gcc).
>>>>> In this way, netcdf-4.1.1 seems work fine. I confirm, instead,
>>> netcdf-3.6.2
>>>>> has still problems.
>>>>>
>>>>> But now, I have to check if I can use netcdf version 4 with WRF
and the
>>>>> other my software!
>>>>>
>>>>> Fabio.
>>>>>
>>>>>
>>>>> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
>>>>>
>>>>> Paul, if you want to visit Rome and be our guest (I live in
Orvieto,
>>>>>> Umbria) you are welcome!
>>>>>> I speak English just a little bit, I graduated in the
Netherlands
>>>>>> although now I have lost a bit of my English ..
>>>>>> But It doesn't matter...! :)
>>>>>>
>>>>>> About GNU compiler. Sure!
>>>>>> I started using Intel compiler because I had some problems with
WRF
>>> model
>>>>>> compiling, but maybe it's time to get back to GNU.
>>>>>>
>>>>>> So, in attachment the *netcdf-3.6.2* building by GNU. But as
you can
>>> see
>>>>>> I have similar problem!
>>>>>> What do you think about?
>>>>>>
>>>>>> In the meantime I also tried* netcdf-4.1.1* (with the option
>>>>>> --disable-netcdf-4 in the configure). Following my results:
>>>>>> 1) GNU --> see attachments
>>>>>> 2) INTEL --> see attachments
>>>>>>
>>>>>> Similar warning/errors both make and make check.
>>>>>>
>>>>>> About compilers the problem is (also) that usually I have to
use the
>>> same
>>>>>> compiler for each programs.
>>>>>> My final target is to use the "chain": WRFDA (or GSI) -->
>>> WPS/WRF-NMM/WPP
>>>>>> --> MET.
>>>>>> But GSI, for example, seems not "GNU" supported (only Intel or
PGI).
>>>>>> With MET, instead, you suggest GNU but I have problems with
both
>>>>>> compilers.
>>>>>>
>>>>>> What must I do??? I don't give up, but it is very hard! :)
>>>>>>
>>>>>> Fabio.
>>>>>>
>>>>>> PS: I use Linux Fedora14 and GNU version is 4.5.1.
>>>>>>
>>>>>>
>>>>>>
>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>>>>>>
>>>>>>> Fabio,
>>>>>>>
>>>>>>> I was just kidding about in-person support.  We were just
joking that
>>> we
>>>>>>> would like to visit Rome.
>>>>>>>
>>>>>>> If you are the administrator, perhaps you have GNU compilers
>> installed
>>> on
>>>>>>> your machine?  You might want to try using
>>>>>>> those to build the libraries and MET.  It is what we used to
develop
>>> MET,
>>>>>>> so we have the most experience with them.  You
>>>>>>> may already have GNU compilers installed.
>>>>>>>
>>>>>>> Don't give up yet!  Once your compilers are configured
correctly,
>>>>>>> building MET should just be a matter of following the
>>>>>>> instructions on the online tutorial.
>>>>>>>
>>>>>>> Good luck.
>>>>>>>
>>>>>>> Paul
>>>>>>>
>>>>>>>
>>>>>>> On 02/22/2011 02:40 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>>>>>>>
>>>>>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711
>
>>>>>>>>
>>>>>>>> It is really bad news, Paul!
>>>>>>>> I am the administrator. I live near Rome, but I run a small
computer
>>>>>>> center
>>>>>>>> in a company that is located in Udine. Before upgrading the
>>> computers
>>>>>>> in
>>>>>>>> Udine,
>>>>>>>> I usually do experiments on my laptop (like right now).
>>>>>>>>
>>>>>>>> What about in-person support? Is it really necessary?
>>>>>>>> Unfortunatley we no have budget to support it, at the moment.
>>>>>>>>
>>>>>>>> So the bad news is.. maybe I have to stop before I start!
>>>>>>>>
>>>>>>>> Fabio.
>>>>>>>>
>>>>>>>>
>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
>>>>>>>>
>>>>>>>>> Fabio,
>>>>>>>>>
>>>>>>>>> It looks like your c++ compiler is not configured correctly.
The
>>>>>>> error
>>>>>>>>> message indicates that your compiler cannot find
>>>>>>>>> a standard c++ library function (strncpy).  I think you may
want to
>>>>>>> talk to
>>>>>>>>> your administrator about the compiler
>>>>>>>>> configuration.  This will be a problem when you try to
compile MET,
>>>>>>> too.
>>>>>>>>>
>>>>>>>>> ncvalues.cpp(238): error: identifier "strncpy" is undefined
>>>>>>>>>      strncpy(s, (const char*)the_values + n,
(int)the_number);
>>>>>>>>>
>>>>>>>>> We would like to provide in-person support to you in Rome.
Would
>>> your
>>>>>>>>> operations budget support that?
>>>>>>>>>
>>>>>>>>> Paul
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>>>>>>>>>
>>>>>>>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>>>>>>>>
>>>>>>>>>> I'm sorry Paul, I made a mistake:
>>>>>>>>>>
>>>>>>>>>> I remembered that when I compile netcdf, *I usually set
CXX='',
>>>>>>>>> otherwhise I
>>>>>>>>>> obtain an
>>>>>>>>>> abort compilation, as you can see in attachment.*
>>>>>>>>>> Again, now I remember had a similar problem with gcc (about
2
>>> years
>>>>>>> ago!)
>>>>>>>>>>
>>>>>>>>>> Apparently, for all the other programs do not set CXX is
not a
>>>>>>> problem.
>>>>>>>>>> I never solved that issue that now back to me forcefully!
>>>>>>>>>>
>>>>>>>>>> How can I fix it, in your opinion?
>>>>>>>>>>
>>>>>>>>>> Thanks,
>>>>>>>>>> Fabio.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
>>>>>>>>>>
>>>>>>>>>>> Fabio,
>>>>>>>>>>>
>>>>>>>>>>> The final step in the NetCDF should basically copy
cxx/netcdf.hh
>>> to
>>>>>>>>>>> include/netcdf.hh.  We suspect that there is
>>>>>>>>>>> something funny going on with your configure and build
process.
>>>  Can
>>>>>>> you
>>>>>>>>>>> please send us log files of your configuration
>>>>>>>>>>> and build using the following commands:
>>>>>>>>>>>
>>>>>>>>>>> $ make clean
>>>>>>>>>>> $ ./configure --prefix=/netcdf/install/path &>
nc_configure.log
>>>>>>>>>>> $ make &> nc_make.log
>>>>>>>>>>> $ make install &> nc_make_install.log
>>>>>>>>>>>
>>>>>>>>>>> Thanks,
>>>>>>>>>>>
>>>>>>>>>>> Paul
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>>>>>>>>>>>
>>>>>>>>>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>>>>>>>>>>
>>>>>>>>>>>> Ok Paul,
>>>>>>>>>>>>
>>>>>>>>>>>> I removed my netcdf.hh copy.
>>>>>>>>>>>>
>>>>>>>>>>>> I'm following the tutorial and I understand about
BUFRLIB,
>>> NetCDF
>>>>>>> and
>>>>>>>>>>> GSL.
>>>>>>>>>>>> Regarding BUFRLIB I'll ignore the ifort errors. GSL seems
built
>>> ok.
>>>>>>>>>>>>
>>>>>>>>>>>> *About NetCDF, I understand that MET wants ".hh" but I
think
>>> (but
>>>>>>> maybe
>>>>>>>>> I
>>>>>>>>>>>> wrong) I compiled it correctly. My env settings are:*
>>>>>>>>>>>> export CC=icc
>>>>>>>>>>>> export FC=ifort
>>>>>>>>>>>> export F9X=ifort
>>>>>>>>>>>> export F90=ifort
>>>>>>>>>>>> export F77=ifort
>>>>>>>>>>>> export CXX=icc
>>>>>>>>>>>>
>>>>>>>>>>>> *Then I do:*
>>>>>>>>>>>> ./configure --prefix=/home/master/netcdf-3.6.2
>>>>>>>>>>>> make
>>>>>>>>>>>> make install
>>>>>>>>>>>>
>>>>>>>>>>>> But *my building process doesn't produce any .hh file
into
>>> /include
>>>>>>>>>>> folder,
>>>>>>>>>>>> but only a .h file*.
>>>>>>>>>>>> I built NetCDF several times in the past, and I never had
any
>>>>>>> problems
>>>>>>>>> so
>>>>>>>>>>>> far with it, because WPS, WRF, WRFDA and so on are okay.
>>>>>>>>>>>>
>>>>>>>>>>>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx
folder,
>>> so
>>>>>>> I
>>>>>>>>>>> tried
>>>>>>>>>>>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
>>>>>>>>>>>> on my makefile, but now I get:
>>>>>>>>>>>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16):
catastrophic
>>> error:
>>>>>>> could
>>>>>>>>>>> not
>>>>>>>>>>>> open source file "netcdf.h"
>>>>>>>>>>>>   #include "netcdf.h"
>>>>>>>>>>>>
>>>>>>>>>>>> Sorry Paul, maybe I'm doing some stupid errors, but
really I
>>> don't
>>>>>>>>>>>> understand what's wrong on my NetCDF building!
>>>>>>>>>>>>
>>>>>>>>>>>> Thank you, Paul, for your patience,
>>>>>>>>>>>> Fabio.
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
>>>>>>>>>>>>
>>>>>>>>>>>>> Fabio,
>>>>>>>>>>>>>
>>>>>>>>>>>>> Your solution of copying the NetCDF header file is
causing the
>>>>>>>>> problem.
>>>>>>>>>>>>>  The file include/netcdf.hh is generated when
>>>>>>>>>>>>> the c/c++ interface of NetCDF is built.  The instance
you are
>>>>>>> trying
>>>>>>>>> to
>>>>>>>>>>> use
>>>>>>>>>>>>> may not have that interface built.
>>>>>>>>>>>>>
>>>>>>>>>>>>> We recommend that you download and build your own
instances of
>>> the
>>>>>>>>>>>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
>>>>>>>>>>>>> that they are built in the way that MET expects.  The
process
>>>>>>> should
>>>>>>>>> be
>>>>>>>>>>>>> pretty easy if you carefully read and follow the
>>>>>>>>>>>>> instructions at the online tutorial I mentioned earlier,
using
>>> the
>>>>>>>>> intel
>>>>>>>>>>>>> compilers for each step:
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>
>>>>>>>
>>>
>>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
>>>>>>>>>>>>>
>>>>>>>>>>>>> You may see some error messages when building BUFRLIB
when
>>>>>>> executing
>>>>>>>>>>> this
>>>>>>>>>>>>> step:
>>>>>>>>>>>>>
>>>>>>>>>>>>> ifort -c -DUNDERSCORE *.f *.F
>>>>>>>>>>>>>
>>>>>>>>>>>>> Ignore these error messages and continue with the
process.  MET
>>>>>>> should
>>>>>>>>>>> work
>>>>>>>>>>>>> with the partially built version of BUFRLIB.
>>>>>>>>>>>>>  Please let us know if you have any other questions.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Paul
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio
F.Gervasi}
>>> wrote:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted upon.
>>>>>>>>>>>>>> Transaction: Ticket created by
fabio.gervasi70 at gmail.com
>>>>>>>>>>>>>>        Queue: met_help
>>>>>>>>>>>>>>      Subject: Met building problems: compilation
aborted for
>>>>>>>>>>>>> vx_contable_stats.cc
>>>>>>>>>>>>>>        Owner: Nobody
>>>>>>>>>>>>>>   Requestors: fabio.gervasi70 at gmail.com
>>>>>>>>>>>>>>       Status: new
>>>>>>>>>>>>>>  Ticket <URL:
>>>>>>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I open another post, but it could be a continuation
about
>>>>>>>>>>>>>> my previous ticket #44675.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> In order to build MET on my Linux Intel machine
(64bit),
>>>>>>>>>>>>>> into ../netcdf../include I did as following:
>>>>>>>>>>>>>> cp netcdf.h netcdf.hh.
>>>>>>>>>>>>>> In this way I resolved my ".hh" problem.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I also built succesfully GSL as well.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so
on...)
>>> is
>>>>>>> built
>>>>>>>>>>> by
>>>>>>>>>>>>>> Intel compiler.*
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> But when I "make", I get the errors in attachment.
>>>>>>>>>>>>>> In attachment the makefile too.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Where do I wrong, now?
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Thank you veru much,
>>>>>>>>>>>>>> Fabio.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>>
>>


------------------------------------------------
Subject: Met building problems: compilation aborted for vx_contable_stats.cc
From: Fabio F.Gervasi
Time: Wed Feb 23 14:34:50 2011

Paul,

in attachment the "test_all" script results.
I seem ok, are you agree?

Fabio.


2011/2/23 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>

> Fabio,
>
> Please make sure to run the test scripts in the scripts directory.
In
> particular, test_all.sh will exercise all of the
> MET executables to make sure they run.
>
> Good luck,
>
> Paul
>
>
> On 02/23/2011 02:20 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >
> > Great Paul!
> >
> > I'm glad to hear that at least MET is independent from other
> programsregarding
> > NetCDF and compilers.
> >
> > So, I'm going to proceed with GNU. Anyway, in the meantime, I
discovered
> > that the NetCDF-3.6.3  (instead of 3.6.2) *builds succesfully
about CXX
> and
> > also it is right for WPS/WRF/WPP!*
> >
> > Regarding stdc++ library, I installed "libstdc++-static.x86_64"
package,
> > then I did:
> > [root at aspire5920g master]# g++ test.cpp
> > [root at aspire5920g master]# ./a.out
> > hello world
> >
> > So your test works fine. Now:
> >
> > [root at aspire5920g MET]# make >& make_met.log --> *EUREKA! 14
executable
> > created.
> > *
> > So now I'll start the case study in order to learn about MET.
> >
> > Thank you very much Paul for your support.
> > I'll write again about other problems, but we can close this
topic.
> > Anyway you are welcome in Italy, and I'm not kidding. :)
> >
> > Fabio.
> >
> >
> > 2011/2/23 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
> >
> >> Fabio,
> >>
> >> In response to your questions, I would like to make three points:
> >>
> >> 1.  It does not matter if you use intel compilers to compile
GSI/WRF/WPP
> >> and GNU compilers to compile MET.  MET only
> >> reads the output of WPP, and it does not interact with any of the
other
> >> components of your "weather chain" in any way.
> >> If you use different compilers for MET than for the other
components, it
> >> will not matter.
> >>
> >> 2.  If you have two separate instances of NetCDF - one for
GSI/WRF/WPP
> and
> >> one for MET, that is ok too.  MET will still
> >> be able to read WPP output if you use different NetCDF instances
for
> each
> >> of them.  Actually, the output of WPP is GRIB,
> >> so you should not have to worry about how MET uses NetCDF, which
is
> >> internal to MET.  MET will accept the GRIB output of
> >> WPP.
> >>
> >> 3.  In make_gnu_met.log, the error message "/usr/bin/ld: cannot
find
> >> -lstdc++" indicates that your GNU g++ compiler
> >> cannot find the stdc++ library, which should be part of the
compiler
> >> installation.  As an experiment, please try to
> >> compile and run the attached test c++ program using the following
> command:
> >>
> >> $ g++ test.cpp
> >> $ ./a.out
> >>
> >> If this does not work, then you need to fix the GNU compiler
> configuration
> >> so that it does.  This test.cpp is the
> >> simplest possible c++ program.
> >>
> >> Let me know if you have any questions.
> >>
> >> Good luck,
> >>
> >> Paul
> >>
> >>
> >> On 02/23/2011 06:10 AM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
> >>>
> >>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >>>
> >>> Hi Paul, since yesterday I wrote to you a lot of emails, so I'd
like to
> >>> summarize the situation, after some tests.
> >>>
> >>> *When I use** GNU, env are:*
> >>> export CC=gcc
> >>> export FC=gfortran
> >>> export F9X=gfortran
> >>> export F90=gfortran
> >>> export F77=gfortran
> >>> export CXX=g++
> >>>
> >>> *When I use Intel, env are:*
> >>> export CC=icc
> >>> export FC=ifort
> >>> export F9X=ifort
> >>> export F90=ifort
> >>> export F77=ifort
> >>> export CXX=icpc
> >>>
> >>> *1)* Both GNU and Intel compilers, I'm able to build GSL and
BUFRLIB
> (for
> >>> bufrlib I ignored ifort errors, about gnu I replaced nvnwin.f
file)
> >>> *2)* Both GNU and Intel compilers, I'm able to build NetCDF-
4.1.1 with
> >> CXX
> >>> *3)* About MET now I get the logs in attachments for both
compilers
> >> (errors,
> >>> but gnu seems better then intel). In attachment the gnu and
intel
> >> makefile
> >>> that I'm using, too.
> >>> *4)* Anyway, unfortunately I'm not able to compile my WRF-NMM
model
> >>> with NetCDF-4.1.1
> >>> (also using --disable-netcdf4 option). I'm able to build WRF-NMM
only
> >> using
> >>> NetCDF-3.6.2. Can I, in case, to use Net3 with WRF and Net4 with
MET at
> >> the
> >>> same time?
> >>> *5)* About the choosing between GNU or Intel, I have advantages
and
> >>> disvantages. For example at the moment I have some problems with
GNU
> >>> building WRF model (but I think helpdesk can help me). But at
the same
> >> time,
> >>> GSI seems supports better Intel.
> >>>
> >>> As I can told you previously, I need to do the "best choice"
between
> GNU
> >> or
> >>> INTEL, taking into account the chain:
> >>> *GSI (or WRFDA) --> WPS/WRF-NMM/WPP --> MET*, and and related
> library...
> >>>
> >>> I hope that my situation is more clear now.
> >>> Let me know, if possible, what you think about it.
> >>>
> >>> Ciao!
> >>> Fabio.
> >>>
> >>>
> >>>
> >>> 2011/2/23 RAL HelpDesk {for Fabio F.Gervasi} <met_help at ucar.edu>
> >>>>
> >>>> An updating, Paul:
> >>>> As I thought: WRF wnats NetCDF-3.6.2, not 4. Although I
compiled
> >>>> NetCDF with the
> >>>> --disable-netcdf4 configure option.
> >>>> Some times ago, wrf helpdesk told me that WRF wants 3.6.2...
> >>>>
> >>>>
> >>>> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
> >>>>
> >>>>> Errata corrige, Paul:
> >>>>> about GNU compiler, maybe I have to set CXX=g++ (not gcc).
> >>>>> In this way, netcdf-4.1.1 seems work fine. I confirm, instead,
> >>> netcdf-3.6.2
> >>>>> has still problems.
> >>>>>
> >>>>> But now, I have to check if I can use netcdf version 4 with
WRF and
> the
> >>>>> other my software!
> >>>>>
> >>>>> Fabio.
> >>>>>
> >>>>>
> >>>>> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
> >>>>>
> >>>>> Paul, if you want to visit Rome and be our guest (I live in
Orvieto,
> >>>>>> Umbria) you are welcome!
> >>>>>> I speak English just a little bit, I graduated in the
Netherlands
> >>>>>> although now I have lost a bit of my English ..
> >>>>>> But It doesn't matter...! :)
> >>>>>>
> >>>>>> About GNU compiler. Sure!
> >>>>>> I started using Intel compiler because I had some problems
with WRF
> >>> model
> >>>>>> compiling, but maybe it's time to get back to GNU.
> >>>>>>
> >>>>>> So, in attachment the *netcdf-3.6.2* building by GNU. But as
you can
> >>> see
> >>>>>> I have similar problem!
> >>>>>> What do you think about?
> >>>>>>
> >>>>>> In the meantime I also tried* netcdf-4.1.1* (with the option
> >>>>>> --disable-netcdf-4 in the configure). Following my results:
> >>>>>> 1) GNU --> see attachments
> >>>>>> 2) INTEL --> see attachments
> >>>>>>
> >>>>>> Similar warning/errors both make and make check.
> >>>>>>
> >>>>>> About compilers the problem is (also) that usually I have to
use the
> >>> same
> >>>>>> compiler for each programs.
> >>>>>> My final target is to use the "chain": WRFDA (or GSI) -->
> >>> WPS/WRF-NMM/WPP
> >>>>>> --> MET.
> >>>>>> But GSI, for example, seems not "GNU" supported (only Intel
or PGI).
> >>>>>> With MET, instead, you suggest GNU but I have problems with
both
> >>>>>> compilers.
> >>>>>>
> >>>>>> What must I do??? I don't give up, but it is very hard! :)
> >>>>>>
> >>>>>> Fabio.
> >>>>>>
> >>>>>> PS: I use Linux Fedora14 and GNU version is 4.5.1.
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
> >>>>>>
> >>>>>>> Fabio,
> >>>>>>>
> >>>>>>> I was just kidding about in-person support.  We were just
joking
> that
> >>> we
> >>>>>>> would like to visit Rome.
> >>>>>>>
> >>>>>>> If you are the administrator, perhaps you have GNU compilers
> >> installed
> >>> on
> >>>>>>> your machine?  You might want to try using
> >>>>>>> those to build the libraries and MET.  It is what we used to
> develop
> >>> MET,
> >>>>>>> so we have the most experience with them.  You
> >>>>>>> may already have GNU compilers installed.
> >>>>>>>
> >>>>>>> Don't give up yet!  Once your compilers are configured
correctly,
> >>>>>>> building MET should just be a matter of following the
> >>>>>>> instructions on the online tutorial.
> >>>>>>>
> >>>>>>> Good luck.
> >>>>>>>
> >>>>>>> Paul
> >>>>>>>
> >>>>>>>
> >>>>>>> On 02/22/2011 02:40 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
> >>>>>>>>
> >>>>>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >>>>>>>>
> >>>>>>>> It is really bad news, Paul!
> >>>>>>>> I am the administrator. I live near Rome, but I run a small
> computer
> >>>>>>> center
> >>>>>>>> in a company that is located in Udine. Before upgrading the
> >>> computers
> >>>>>>> in
> >>>>>>>> Udine,
> >>>>>>>> I usually do experiments on my laptop (like right now).
> >>>>>>>>
> >>>>>>>> What about in-person support? Is it really necessary?
> >>>>>>>> Unfortunatley we no have budget to support it, at the
moment.
> >>>>>>>>
> >>>>>>>> So the bad news is.. maybe I have to stop before I start!
> >>>>>>>>
> >>>>>>>> Fabio.
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
> >>>>>>>>
> >>>>>>>>> Fabio,
> >>>>>>>>>
> >>>>>>>>> It looks like your c++ compiler is not configured
correctly.  The
> >>>>>>> error
> >>>>>>>>> message indicates that your compiler cannot find
> >>>>>>>>> a standard c++ library function (strncpy).  I think you
may want
> to
> >>>>>>> talk to
> >>>>>>>>> your administrator about the compiler
> >>>>>>>>> configuration.  This will be a problem when you try to
compile
> MET,
> >>>>>>> too.
> >>>>>>>>>
> >>>>>>>>> ncvalues.cpp(238): error: identifier "strncpy" is
undefined
> >>>>>>>>>      strncpy(s, (const char*)the_values + n,
(int)the_number);
> >>>>>>>>>
> >>>>>>>>> We would like to provide in-person support to you in Rome.
Would
> >>> your
> >>>>>>>>> operations budget support that?
> >>>>>>>>>
> >>>>>>>>> Paul
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
> >>>>>>>>>>
> >>>>>>>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
> >>>>>>>>>>
> >>>>>>>>>> I'm sorry Paul, I made a mistake:
> >>>>>>>>>>
> >>>>>>>>>> I remembered that when I compile netcdf, *I usually set
CXX='',
> >>>>>>>>> otherwhise I
> >>>>>>>>>> obtain an
> >>>>>>>>>> abort compilation, as you can see in attachment.*
> >>>>>>>>>> Again, now I remember had a similar problem with gcc
(about 2
> >>> years
> >>>>>>> ago!)
> >>>>>>>>>>
> >>>>>>>>>> Apparently, for all the other programs do not set CXX is
not a
> >>>>>>> problem.
> >>>>>>>>>> I never solved that issue that now back to me forcefully!
> >>>>>>>>>>
> >>>>>>>>>> How can I fix it, in your opinion?
> >>>>>>>>>>
> >>>>>>>>>> Thanks,
> >>>>>>>>>> Fabio.
> >>>>>>>>>>
> >>>>>>>>>>
> >>>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
> >>>>>>>>>>
> >>>>>>>>>>> Fabio,
> >>>>>>>>>>>
> >>>>>>>>>>> The final step in the NetCDF should basically copy
> cxx/netcdf.hh
> >>> to
> >>>>>>>>>>> include/netcdf.hh.  We suspect that there is
> >>>>>>>>>>> something funny going on with your configure and build
process.
> >>>  Can
> >>>>>>> you
> >>>>>>>>>>> please send us log files of your configuration
> >>>>>>>>>>> and build using the following commands:
> >>>>>>>>>>>
> >>>>>>>>>>> $ make clean
> >>>>>>>>>>> $ ./configure --prefix=/netcdf/install/path &>
nc_configure.log
> >>>>>>>>>>> $ make &> nc_make.log
> >>>>>>>>>>> $ make install &> nc_make_install.log
> >>>>>>>>>>>
> >>>>>>>>>>> Thanks,
> >>>>>>>>>>>
> >>>>>>>>>>> Paul
> >>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>>> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio
F.Gervasi}
> wrote:
> >>>>>>>>>>>>
> >>>>>>>>>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
> >>>>>>>>>>>>
> >>>>>>>>>>>> Ok Paul,
> >>>>>>>>>>>>
> >>>>>>>>>>>> I removed my netcdf.hh copy.
> >>>>>>>>>>>>
> >>>>>>>>>>>> I'm following the tutorial and I understand about
BUFRLIB,
> >>> NetCDF
> >>>>>>> and
> >>>>>>>>>>> GSL.
> >>>>>>>>>>>> Regarding BUFRLIB I'll ignore the ifort errors. GSL
seems
> built
> >>> ok.
> >>>>>>>>>>>>
> >>>>>>>>>>>> *About NetCDF, I understand that MET wants ".hh" but I
think
> >>> (but
> >>>>>>> maybe
> >>>>>>>>> I
> >>>>>>>>>>>> wrong) I compiled it correctly. My env settings are:*
> >>>>>>>>>>>> export CC=icc
> >>>>>>>>>>>> export FC=ifort
> >>>>>>>>>>>> export F9X=ifort
> >>>>>>>>>>>> export F90=ifort
> >>>>>>>>>>>> export F77=ifort
> >>>>>>>>>>>> export CXX=icc
> >>>>>>>>>>>>
> >>>>>>>>>>>> *Then I do:*
> >>>>>>>>>>>> ./configure --prefix=/home/master/netcdf-3.6.2
> >>>>>>>>>>>> make
> >>>>>>>>>>>> make install
> >>>>>>>>>>>>
> >>>>>>>>>>>> But *my building process doesn't produce any .hh file
into
> >>> /include
> >>>>>>>>>>> folder,
> >>>>>>>>>>>> but only a .h file*.
> >>>>>>>>>>>> I built NetCDF several times in the past, and I never
had any
> >>>>>>> problems
> >>>>>>>>> so
> >>>>>>>>>>>> far with it, because WPS, WRF, WRFDA and so on are
okay.
> >>>>>>>>>>>>
> >>>>>>>>>>>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx
> folder,
> >>> so
> >>>>>>> I
> >>>>>>>>>>> tried
> >>>>>>>>>>>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
> >>>>>>>>>>>> on my makefile, but now I get:
> >>>>>>>>>>>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16):
catastrophic
> >>> error:
> >>>>>>> could
> >>>>>>>>>>> not
> >>>>>>>>>>>> open source file "netcdf.h"
> >>>>>>>>>>>>   #include "netcdf.h"
> >>>>>>>>>>>>
> >>>>>>>>>>>> Sorry Paul, maybe I'm doing some stupid errors, but
really I
> >>> don't
> >>>>>>>>>>>> understand what's wrong on my NetCDF building!
> >>>>>>>>>>>>
> >>>>>>>>>>>> Thank you, Paul, for your patience,
> >>>>>>>>>>>> Fabio.
> >>>>>>>>>>>>
> >>>>>>>>>>>>
> >>>>>>>>>>>>
> >>>>>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <
> met_help at ucar.edu>
> >>>>>>>>>>>>
> >>>>>>>>>>>>> Fabio,
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> Your solution of copying the NetCDF header file is
causing
> the
> >>>>>>>>> problem.
> >>>>>>>>>>>>>  The file include/netcdf.hh is generated when
> >>>>>>>>>>>>> the c/c++ interface of NetCDF is built.  The instance
you are
> >>>>>>> trying
> >>>>>>>>> to
> >>>>>>>>>>> use
> >>>>>>>>>>>>> may not have that interface built.
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> We recommend that you download and build your own
instances
> of
> >>> the
> >>>>>>>>>>>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
> >>>>>>>>>>>>> that they are built in the way that MET expects.  The
process
> >>>>>>> should
> >>>>>>>>> be
> >>>>>>>>>>>>> pretty easy if you carefully read and follow the
> >>>>>>>>>>>>> instructions at the online tutorial I mentioned
earlier,
> using
> >>> the
> >>>>>>>>> intel
> >>>>>>>>>>>>> compilers for each step:
> >>>>>>>>>>>>>
> >>>>>>>>>>>>>
> >>>>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>
> >>>>>>>
> >>>
> >>
>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> You may see some error messages when building BUFRLIB
when
> >>>>>>> executing
> >>>>>>>>>>> this
> >>>>>>>>>>>>> step:
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> ifort -c -DUNDERSCORE *.f *.F
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> Ignore these error messages and continue with the
process.
>  MET
> >>>>>>> should
> >>>>>>>>>>> work
> >>>>>>>>>>>>> with the partially built version of BUFRLIB.
> >>>>>>>>>>>>>  Please let us know if you have any other questions.
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> Paul
> >>>>>>>>>>>>>
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio
F.Gervasi}
> >>> wrote:
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted
upon.
> >>>>>>>>>>>>>> Transaction: Ticket created by
fabio.gervasi70 at gmail.com
> >>>>>>>>>>>>>>        Queue: met_help
> >>>>>>>>>>>>>>      Subject: Met building problems: compilation
aborted for
> >>>>>>>>>>>>> vx_contable_stats.cc
> >>>>>>>>>>>>>>        Owner: Nobody
> >>>>>>>>>>>>>>   Requestors: fabio.gervasi70 at gmail.com
> >>>>>>>>>>>>>>       Status: new
> >>>>>>>>>>>>>>  Ticket <URL:
> >>>>>>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> Hi,
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> I open another post, but it could be a continuation
about
> >>>>>>>>>>>>>> my previous ticket #44675.
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> In order to build MET on my Linux Intel machine
(64bit),
> >>>>>>>>>>>>>> into ../netcdf../include I did as following:
> >>>>>>>>>>>>>> cp netcdf.h netcdf.hh.
> >>>>>>>>>>>>>> In this way I resolved my ".hh" problem.
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> I also built succesfully GSL as well.
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so
on...)
> >>> is
> >>>>>>> built
> >>>>>>>>>>> by
> >>>>>>>>>>>>>> Intel compiler.*
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> But when I "make", I get the errors in attachment.
> >>>>>>>>>>>>>> In attachment the makefile too.
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> Where do I wrong, now?
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> Thank you veru much,
> >>>>>>>>>>>>>> Fabio.
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>
> >>>>>>>>>>>>>
> >>>>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>
> >>>>>
> >>>>
> >>>
> >>
> >>
> >>
>
>
>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #44711] Met building problems: compilation aborted for vx_contable_stats.cc
From: Paul Oldenburg
Time: Wed Feb 23 14:42:21 2011

Fabio,

You're looking good.  Remember that you can copy and rename those test
scripts and then modify them to suit your own
needs when you want to run a MET component.  That is a good way to
start with MET.  Also, it's helpful to do the same
thing with the config files in scripts/config.

Paul



On 02/23/2011 02:34 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>
> Paul,
>
> in attachment the "test_all" script results.
> I seem ok, are you agree?
>
> Fabio.
>
>
> 2011/2/23 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>
>> Fabio,
>>
>> Please make sure to run the test scripts in the scripts directory.
In
>> particular, test_all.sh will exercise all of the
>> MET executables to make sure they run.
>>
>> Good luck,
>>
>> Paul
>>
>>
>> On 02/23/2011 02:20 PM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>
>>> Great Paul!
>>>
>>> I'm glad to hear that at least MET is independent from other
>> programsregarding
>>> NetCDF and compilers.
>>>
>>> So, I'm going to proceed with GNU. Anyway, in the meantime, I
discovered
>>> that the NetCDF-3.6.3  (instead of 3.6.2) *builds succesfully
about CXX
>> and
>>> also it is right for WPS/WRF/WPP!*
>>>
>>> Regarding stdc++ library, I installed "libstdc++-static.x86_64"
package,
>>> then I did:
>>> [root at aspire5920g master]# g++ test.cpp
>>> [root at aspire5920g master]# ./a.out
>>> hello world
>>>
>>> So your test works fine. Now:
>>>
>>> [root at aspire5920g MET]# make >& make_met.log --> *EUREKA! 14
executable
>>> created.
>>> *
>>> So now I'll start the case study in order to learn about MET.
>>>
>>> Thank you very much Paul for your support.
>>> I'll write again about other problems, but we can close this
topic.
>>> Anyway you are welcome in Italy, and I'm not kidding. :)
>>>
>>> Fabio.
>>>
>>>
>>> 2011/2/23 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>>>
>>>> Fabio,
>>>>
>>>> In response to your questions, I would like to make three points:
>>>>
>>>> 1.  It does not matter if you use intel compilers to compile
GSI/WRF/WPP
>>>> and GNU compilers to compile MET.  MET only
>>>> reads the output of WPP, and it does not interact with any of the
other
>>>> components of your "weather chain" in any way.
>>>> If you use different compilers for MET than for the other
components, it
>>>> will not matter.
>>>>
>>>> 2.  If you have two separate instances of NetCDF - one for
GSI/WRF/WPP
>> and
>>>> one for MET, that is ok too.  MET will still
>>>> be able to read WPP output if you use different NetCDF instances
for
>> each
>>>> of them.  Actually, the output of WPP is GRIB,
>>>> so you should not have to worry about how MET uses NetCDF, which
is
>>>> internal to MET.  MET will accept the GRIB output of
>>>> WPP.
>>>>
>>>> 3.  In make_gnu_met.log, the error message "/usr/bin/ld: cannot
find
>>>> -lstdc++" indicates that your GNU g++ compiler
>>>> cannot find the stdc++ library, which should be part of the
compiler
>>>> installation.  As an experiment, please try to
>>>> compile and run the attached test c++ program using the following
>> command:
>>>>
>>>> $ g++ test.cpp
>>>> $ ./a.out
>>>>
>>>> If this does not work, then you need to fix the GNU compiler
>> configuration
>>>> so that it does.  This test.cpp is the
>>>> simplest possible c++ program.
>>>>
>>>> Let me know if you have any questions.
>>>>
>>>> Good luck,
>>>>
>>>> Paul
>>>>
>>>>
>>>> On 02/23/2011 06:10 AM, RAL HelpDesk {for Fabio F.Gervasi} wrote:
>>>>>
>>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>>>
>>>>> Hi Paul, since yesterday I wrote to you a lot of emails, so I'd
like to
>>>>> summarize the situation, after some tests.
>>>>>
>>>>> *When I use** GNU, env are:*
>>>>> export CC=gcc
>>>>> export FC=gfortran
>>>>> export F9X=gfortran
>>>>> export F90=gfortran
>>>>> export F77=gfortran
>>>>> export CXX=g++
>>>>>
>>>>> *When I use Intel, env are:*
>>>>> export CC=icc
>>>>> export FC=ifort
>>>>> export F9X=ifort
>>>>> export F90=ifort
>>>>> export F77=ifort
>>>>> export CXX=icpc
>>>>>
>>>>> *1)* Both GNU and Intel compilers, I'm able to build GSL and
BUFRLIB
>> (for
>>>>> bufrlib I ignored ifort errors, about gnu I replaced nvnwin.f
file)
>>>>> *2)* Both GNU and Intel compilers, I'm able to build NetCDF-
4.1.1 with
>>>> CXX
>>>>> *3)* About MET now I get the logs in attachments for both
compilers
>>>> (errors,
>>>>> but gnu seems better then intel). In attachment the gnu and
intel
>>>> makefile
>>>>> that I'm using, too.
>>>>> *4)* Anyway, unfortunately I'm not able to compile my WRF-NMM
model
>>>>> with NetCDF-4.1.1
>>>>> (also using --disable-netcdf4 option). I'm able to build WRF-NMM
only
>>>> using
>>>>> NetCDF-3.6.2. Can I, in case, to use Net3 with WRF and Net4 with
MET at
>>>> the
>>>>> same time?
>>>>> *5)* About the choosing between GNU or Intel, I have advantages
and
>>>>> disvantages. For example at the moment I have some problems with
GNU
>>>>> building WRF model (but I think helpdesk can help me). But at
the same
>>>> time,
>>>>> GSI seems supports better Intel.
>>>>>
>>>>> As I can told you previously, I need to do the "best choice"
between
>> GNU
>>>> or
>>>>> INTEL, taking into account the chain:
>>>>> *GSI (or WRFDA) --> WPS/WRF-NMM/WPP --> MET*, and and related
>> library...
>>>>>
>>>>> I hope that my situation is more clear now.
>>>>> Let me know, if possible, what you think about it.
>>>>>
>>>>> Ciao!
>>>>> Fabio.
>>>>>
>>>>>
>>>>>
>>>>> 2011/2/23 RAL HelpDesk {for Fabio F.Gervasi} <met_help at ucar.edu>
>>>>>>
>>>>>> An updating, Paul:
>>>>>> As I thought: WRF wnats NetCDF-3.6.2, not 4. Although I
compiled
>>>>>> NetCDF with the
>>>>>> --disable-netcdf4 configure option.
>>>>>> Some times ago, wrf helpdesk told me that WRF wants 3.6.2...
>>>>>>
>>>>>>
>>>>>> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
>>>>>>
>>>>>>> Errata corrige, Paul:
>>>>>>> about GNU compiler, maybe I have to set CXX=g++ (not gcc).
>>>>>>> In this way, netcdf-4.1.1 seems work fine. I confirm, instead,
>>>>> netcdf-3.6.2
>>>>>>> has still problems.
>>>>>>>
>>>>>>> But now, I have to check if I can use netcdf version 4 with
WRF and
>> the
>>>>>>> other my software!
>>>>>>>
>>>>>>> Fabio.
>>>>>>>
>>>>>>>
>>>>>>> 2011/2/23 Fabio F.Gervasi <fabio.gervasi70 at gmail.com>
>>>>>>>
>>>>>>> Paul, if you want to visit Rome and be our guest (I live in
Orvieto,
>>>>>>>> Umbria) you are welcome!
>>>>>>>> I speak English just a little bit, I graduated in the
Netherlands
>>>>>>>> although now I have lost a bit of my English ..
>>>>>>>> But It doesn't matter...! :)
>>>>>>>>
>>>>>>>> About GNU compiler. Sure!
>>>>>>>> I started using Intel compiler because I had some problems
with WRF
>>>>> model
>>>>>>>> compiling, but maybe it's time to get back to GNU.
>>>>>>>>
>>>>>>>> So, in attachment the *netcdf-3.6.2* building by GNU. But as
you can
>>>>> see
>>>>>>>> I have similar problem!
>>>>>>>> What do you think about?
>>>>>>>>
>>>>>>>> In the meantime I also tried* netcdf-4.1.1* (with the option
>>>>>>>> --disable-netcdf-4 in the configure). Following my results:
>>>>>>>> 1) GNU --> see attachments
>>>>>>>> 2) INTEL --> see attachments
>>>>>>>>
>>>>>>>> Similar warning/errors both make and make check.
>>>>>>>>
>>>>>>>> About compilers the problem is (also) that usually I have to
use the
>>>>> same
>>>>>>>> compiler for each programs.
>>>>>>>> My final target is to use the "chain": WRFDA (or GSI) -->
>>>>> WPS/WRF-NMM/WPP
>>>>>>>> --> MET.
>>>>>>>> But GSI, for example, seems not "GNU" supported (only Intel
or PGI).
>>>>>>>> With MET, instead, you suggest GNU but I have problems with
both
>>>>>>>> compilers.
>>>>>>>>
>>>>>>>> What must I do??? I don't give up, but it is very hard! :)
>>>>>>>>
>>>>>>>> Fabio.
>>>>>>>>
>>>>>>>> PS: I use Linux Fedora14 and GNU version is 4.5.1.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
>>>>>>>>
>>>>>>>>> Fabio,
>>>>>>>>>
>>>>>>>>> I was just kidding about in-person support.  We were just
joking
>> that
>>>>> we
>>>>>>>>> would like to visit Rome.
>>>>>>>>>
>>>>>>>>> If you are the administrator, perhaps you have GNU compilers
>>>> installed
>>>>> on
>>>>>>>>> your machine?  You might want to try using
>>>>>>>>> those to build the libraries and MET.  It is what we used to
>> develop
>>>>> MET,
>>>>>>>>> so we have the most experience with them.  You
>>>>>>>>> may already have GNU compilers installed.
>>>>>>>>>
>>>>>>>>> Don't give up yet!  Once your compilers are configured
correctly,
>>>>>>>>> building MET should just be a matter of following the
>>>>>>>>> instructions on the online tutorial.
>>>>>>>>>
>>>>>>>>> Good luck.
>>>>>>>>>
>>>>>>>>> Paul
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On 02/22/2011 02:40 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>>>>>>>>>
>>>>>>>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>>>>>>>>
>>>>>>>>>> It is really bad news, Paul!
>>>>>>>>>> I am the administrator. I live near Rome, but I run a small
>> computer
>>>>>>>>> center
>>>>>>>>>> in a company that is located in Udine. Before upgrading the
>>>>> computers
>>>>>>>>> in
>>>>>>>>>> Udine,
>>>>>>>>>> I usually do experiments on my laptop (like right now).
>>>>>>>>>>
>>>>>>>>>> What about in-person support? Is it really necessary?
>>>>>>>>>> Unfortunatley we no have budget to support it, at the
moment.
>>>>>>>>>>
>>>>>>>>>> So the bad news is.. maybe I have to stop before I start!
>>>>>>>>>>
>>>>>>>>>> Fabio.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
>>>>>>>>>>
>>>>>>>>>>> Fabio,
>>>>>>>>>>>
>>>>>>>>>>> It looks like your c++ compiler is not configured
correctly.  The
>>>>>>>>> error
>>>>>>>>>>> message indicates that your compiler cannot find
>>>>>>>>>>> a standard c++ library function (strncpy).  I think you
may want
>> to
>>>>>>>>> talk to
>>>>>>>>>>> your administrator about the compiler
>>>>>>>>>>> configuration.  This will be a problem when you try to
compile
>> MET,
>>>>>>>>> too.
>>>>>>>>>>>
>>>>>>>>>>> ncvalues.cpp(238): error: identifier "strncpy" is
undefined
>>>>>>>>>>>      strncpy(s, (const char*)the_values + n,
(int)the_number);
>>>>>>>>>>>
>>>>>>>>>>> We would like to provide in-person support to you in Rome.
Would
>>>>> your
>>>>>>>>>>> operations budget support that?
>>>>>>>>>>>
>>>>>>>>>>> Paul
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> On 02/22/2011 02:00 PM, RAL HelpDesk {for Fabio F.Gervasi}
wrote:
>>>>>>>>>>>>
>>>>>>>>>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711 >
>>>>>>>>>>>>
>>>>>>>>>>>> I'm sorry Paul, I made a mistake:
>>>>>>>>>>>>
>>>>>>>>>>>> I remembered that when I compile netcdf, *I usually set
CXX='',
>>>>>>>>>>> otherwhise I
>>>>>>>>>>>> obtain an
>>>>>>>>>>>> abort compilation, as you can see in attachment.*
>>>>>>>>>>>> Again, now I remember had a similar problem with gcc
(about 2
>>>>> years
>>>>>>>>> ago!)
>>>>>>>>>>>>
>>>>>>>>>>>> Apparently, for all the other programs do not set CXX is
not a
>>>>>>>>> problem.
>>>>>>>>>>>> I never solved that issue that now back to me forcefully!
>>>>>>>>>>>>
>>>>>>>>>>>> How can I fix it, in your opinion?
>>>>>>>>>>>>
>>>>>>>>>>>> Thanks,
>>>>>>>>>>>> Fabio.
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg}
<met_help at ucar.edu>
>>>>>>>>>>>>
>>>>>>>>>>>>> Fabio,
>>>>>>>>>>>>>
>>>>>>>>>>>>> The final step in the NetCDF should basically copy
>> cxx/netcdf.hh
>>>>> to
>>>>>>>>>>>>> include/netcdf.hh.  We suspect that there is
>>>>>>>>>>>>> something funny going on with your configure and build
process.
>>>>>  Can
>>>>>>>>> you
>>>>>>>>>>>>> please send us log files of your configuration
>>>>>>>>>>>>> and build using the following commands:
>>>>>>>>>>>>>
>>>>>>>>>>>>> $ make clean
>>>>>>>>>>>>> $ ./configure --prefix=/netcdf/install/path &>
nc_configure.log
>>>>>>>>>>>>> $ make &> nc_make.log
>>>>>>>>>>>>> $ make install &> nc_make_install.log
>>>>>>>>>>>>>
>>>>>>>>>>>>> Thanks,
>>>>>>>>>>>>>
>>>>>>>>>>>>> Paul
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 02/22/2011 01:13 PM, RAL HelpDesk {for Fabio
F.Gervasi}
>> wrote:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Ok Paul,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I removed my netcdf.hh copy.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I'm following the tutorial and I understand about
BUFRLIB,
>>>>> NetCDF
>>>>>>>>> and
>>>>>>>>>>>>> GSL.
>>>>>>>>>>>>>> Regarding BUFRLIB I'll ignore the ifort errors. GSL
seems
>> built
>>>>> ok.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> *About NetCDF, I understand that MET wants ".hh" but I
think
>>>>> (but
>>>>>>>>> maybe
>>>>>>>>>>> I
>>>>>>>>>>>>>> wrong) I compiled it correctly. My env settings are:*
>>>>>>>>>>>>>> export CC=icc
>>>>>>>>>>>>>> export FC=ifort
>>>>>>>>>>>>>> export F9X=ifort
>>>>>>>>>>>>>> export F90=ifort
>>>>>>>>>>>>>> export F77=ifort
>>>>>>>>>>>>>> export CXX=icc
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> *Then I do:*
>>>>>>>>>>>>>> ./configure --prefix=/home/master/netcdf-3.6.2
>>>>>>>>>>>>>> make
>>>>>>>>>>>>>> make install
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> But *my building process doesn't produce any .hh file
into
>>>>> /include
>>>>>>>>>>>>> folder,
>>>>>>>>>>>>>> but only a .h file*.
>>>>>>>>>>>>>> I built NetCDF several times in the past, and I never
had any
>>>>>>>>> problems
>>>>>>>>>>> so
>>>>>>>>>>>>>> far with it, because WPS, WRF, WRFDA and so on are
okay.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Anyway, I discovered a ".hh" file into netcdf-3.6.2/cxx
>> folder,
>>>>> so
>>>>>>>>> I
>>>>>>>>>>>>> tried
>>>>>>>>>>>>>> to set NETCDF_INCS  = -I$(NETCDF_BASE)/cxx
>>>>>>>>>>>>>> on my makefile, but now I get:
>>>>>>>>>>>>>> /home/master/netcdf-3.6.2/cxx/ncvalues.h(16):
catastrophic
>>>>> error:
>>>>>>>>> could
>>>>>>>>>>>>> not
>>>>>>>>>>>>>> open source file "netcdf.h"
>>>>>>>>>>>>>>   #include "netcdf.h"
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Sorry Paul, maybe I'm doing some stupid errors, but
really I
>>>>> don't
>>>>>>>>>>>>>> understand what's wrong on my NetCDF building!
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Thank you, Paul, for your patience,
>>>>>>>>>>>>>> Fabio.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> 2011/2/22 RAL HelpDesk {for Paul Oldenburg} <
>> met_help at ucar.edu>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Fabio,
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Your solution of copying the NetCDF header file is
causing
>> the
>>>>>>>>>>> problem.
>>>>>>>>>>>>>>>  The file include/netcdf.hh is generated when
>>>>>>>>>>>>>>> the c/c++ interface of NetCDF is built.  The instance
you are
>>>>>>>>> trying
>>>>>>>>>>> to
>>>>>>>>>>>>> use
>>>>>>>>>>>>>>> may not have that interface built.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> We recommend that you download and build your own
instances
>> of
>>>>> the
>>>>>>>>>>>>>>> libraries BUFRLIB, NetCDF and GSL.  This will ensure
>>>>>>>>>>>>>>> that they are built in the way that MET expects.  The
process
>>>>>>>>> should
>>>>>>>>>>> be
>>>>>>>>>>>>>>> pretty easy if you carefully read and follow the
>>>>>>>>>>>>>>> instructions at the online tutorial I mentioned
earlier,
>> using
>>>>> the
>>>>>>>>>>> intel
>>>>>>>>>>>>>>> compilers for each step:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>
>>>>>
>>>>
>>
http://www.dtcenter.org/met/users/support/online_tutorial/METv3.0/compilation/index.php
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> You may see some error messages when building BUFRLIB
when
>>>>>>>>> executing
>>>>>>>>>>>>> this
>>>>>>>>>>>>>>> step:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> ifort -c -DUNDERSCORE *.f *.F
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Ignore these error messages and continue with the
process.
>>  MET
>>>>>>>>> should
>>>>>>>>>>>>> work
>>>>>>>>>>>>>>> with the partially built version of BUFRLIB.
>>>>>>>>>>>>>>>  Please let us know if you have any other questions.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Paul
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> On 02/22/2011 10:20 AM, RAL HelpDesk {for Fabio
F.Gervasi}
>>>>> wrote:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Tue Feb 22 10:20:16 2011: Request 44711 was acted
upon.
>>>>>>>>>>>>>>>> Transaction: Ticket created by
fabio.gervasi70 at gmail.com
>>>>>>>>>>>>>>>>        Queue: met_help
>>>>>>>>>>>>>>>>      Subject: Met building problems: compilation
aborted for
>>>>>>>>>>>>>>> vx_contable_stats.cc
>>>>>>>>>>>>>>>>        Owner: Nobody
>>>>>>>>>>>>>>>>   Requestors: fabio.gervasi70 at gmail.com
>>>>>>>>>>>>>>>>       Status: new
>>>>>>>>>>>>>>>>  Ticket <URL:
>>>>>>>>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=44711>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> I open another post, but it could be a continuation
about
>>>>>>>>>>>>>>>> my previous ticket #44675.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> In order to build MET on my Linux Intel machine
(64bit),
>>>>>>>>>>>>>>>> into ../netcdf../include I did as following:
>>>>>>>>>>>>>>>> cp netcdf.h netcdf.hh.
>>>>>>>>>>>>>>>> In this way I resolved my ".hh" problem.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> I also built succesfully GSL as well.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> *Each program (WRF, WRFDA, Netcdf, mpich, GSL ando so
on...)
>>>>> is
>>>>>>>>> built
>>>>>>>>>>>>> by
>>>>>>>>>>>>>>>> Intel compiler.*
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> But when I "make", I get the errors in attachment.
>>>>>>>>>>>>>>>> In attachment the makefile too.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Where do I wrong, now?
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Thank you veru much,
>>>>>>>>>>>>>>>> Fabio.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>>
>>>>
>>
>>
>>
>


------------------------------------------------


More information about the Met_help mailing list