This archive contains answers to questions sent to Unidata support through mid-2025. Note that the archive is no longer being updated. We provide the archive for reference; many of the answers presented here remain technically correct, even if somewhat outdated. For the most up-to-date information on the use of NSF Unidata software and data services, please consult the Software Documentation first.
Hi Dan, Sorry it's taken us so long to respond to your question ... > I'm running into a problem building 3.6.3 on a x86_64 linux > platform (uname -a '2.6.8.1-suse92-osmp ...x86_64 GNU/Linux') > > with ifort --version ( ifort 9.1 20070215) > > I've followed the 'intel' building directions on your website, > searched the web for solutions but in the end here I am... By "our website", do you mean the Intel instructions we point to at this URL? http://software.intel.com/en-us/articles/performance-tools-for-software-developers-building-netcdf-with-the-intel-compilers/ > The problem appears to be that configure can't determine the > mapping for fortran types to C types, ( my laptop that I'm writing the > message to can't readily connect to the server, so cut/paste isn't so > simple). The configure results include: > > ... > checking if Fortran "byte"...yes > checking if Fortran "integer*2"... yes > checking if Fortran "byte" is C "signed char"... no > checking if Fortran "byte" is C "short"... no > ... for all the types > > with the result that fortran/nfconfig.inc does not define any of > the NF_INT_IS_C... > > resulting in the fortran code not compiling. That's usually a symptom of a basic problem with the Fortran compiler installation, such as when it's not configured right to use its run-time libraries or some compiler component is installed in the wrong place. Can you compile simple non-netCDF Fortran programs OK with ifort? > I can go in and modify the nfconfig.inc to define these (guessing > what the fortran types should map to), and can get netcdf to build. > > but 'make check' will fail when it goes into nf_test and performs > the tests. Yup, I've never had any luck trying manual workarounds to what configure is supposed to do, so I wouldn't recommend that. > Any workarounds for this version of the compiler? Any suggestions? We have successfully built netCDF-3 libraries from the latest 4.1.1 netCDF distribution, using the --disable-netcdf-4 flag. We used the settings: CPPFLAGS= CC=icc CFLAGS=-g -O2 CXX=icpc CXXFLAGS=-I/opt/intel/cce/10.1.023/include FC=ifort FCFLAGS=-g F77=ifort FFLAGS=-g LDFLAGS= LIBS= FLIBS= F90LIBS= and invoked the configure script as follows: configure --prefix=/machine/netcdf/nc_test_8059/in1 --disable-compiler-recover --disable-netcdf-4 make check install ftpbin The output of the configure and "make check install" steps are here: http://www.unidata.ucar.edu/netcdf/builds/4_1_1/out_spock_2010040807_19_0.txt This was with version 10.1 of ifort. We have binaries available here: ftp://ftp.unidata.ucar.edu/pub/binary/linux_2.6-x86_64/binary-netcdf-4.1.1_nc3_ifort_ifort_icpc.tar.gz This was on a Linux 64-bit platform. --Russ Russ Rew UCAR Unidata Program address@hidden http://www.unidata.ucar.edu Ticket Details =================== Ticket ID: GNF-292202 Department: Support netCDF Priority: High Status: Closed