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

[netCDF #ZPH-410959]: Problem with "make check" netcdf version 3.6.1

Hi Annette,

First, we have a newer version 3.6.2 that has a more advanced build system that 
might work better, although there are no fixes that are specifically for  
platforms using the Intel Fortran compiler.  But if you want to try it instead, 
it's available from the web page 

http://www.unidata.ucar.edu/software/netcdf/release-notes-3.6.2.html

Second, are you setting all the C and Fortran compiler option flags 
consistently to contain the "-i4" option specifying four-byte integers instead 
of the default eight-byte integers on this platform?  The kind of error you are 
seeing might occur if the C netCDF library is compiled assuming eight-byte 
integers but you specify "-i4" for the Fortran-77 and Fortran-90 interfaces, or 
if you specify FFLAGS to not contain "-i4" and F90FLAGS to contain "-i4" before 
running the configure script.

For the Intel Fortran compiler ifort, some users have reported that it's 
necessary to use the "-mp" flag when building netCDF to get a library that 
"should use default numerics and not optimized ones", by setting FFLAGS and 
F90FLAGS to contain "-mp" before running the configure script and setting 
FLIBS="/opt/intel/compiler70/ia32/lib/libPEPCF90.a" (or wherever libPEPCF90.a 
is stored for your installation).  

I don't have current access to a platform with mpif90 installed so I can't 
duplicate the problem, but if none of the above suggestions help, please send 
all the information requested here and we'll try to diagnose the cause:

http://www.unidata.ucar.edu/software/netcdf/docs/netcdf-install.html#Reporting-Problems

--Russ

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



Ticket Details
===================
Ticket ID: ZPH-410959
Department: Support netCDF
Priority: Normal
Status: Closed