NOTICE: This version of the NSF Unidata web site (archive.unidata.ucar.edu) is no longer being updated.
Current content can be found at unidata.ucar.edu.
To learn about what's going on, see About the Archive Site.
> From: jhudd@xxxxxxxxxxxxxxxxxxxxxxxxx (John Huddleston) > To: netcdfgroup@xxxxxxxxxxxxxxxx (Group NetCDF) Hi John, > I have been reading the mail on the g++ libraries for > netCDF. This morning I downloaded the newest patches > and have the following comments: > > Please add the c++/all to the Makefile and the configure files. Sorry, but this will have to wait until the next major netCDF release for at least two reasons. First, fixing the `configure' script to determine whether a C++ compiler is available and if so which features it supports requires a newer version of the `autoconf' package than was used to generate the `configure' script in the current release, so entails more work than we can justify for a patch. Second, the C++ interface is still labeled experimental, which means we still want to be free to change the interface in ways that may not be strictly backward-compatible. We want users of the experimental C++ interface to realize its status. Requiring users to manually change to the C++ directory and read the `README' file to install it helps inform them that they may have to change their programs in the future if they choose to use this interface. > When you do the patch, uprev the version number for the VERSION > variable to 2.3.3 or 2.3.4 as appropriate. The version of the package is still 2.3.2, but at patch level 4 (abbreviated `2.3.2pl4'). If we were to increment the version to 2.3.3, we would have to retest it on lots of platforms and change some documentation. I've made the the judgment release of a minor patch that only affects the experimental C++ interface doesn't justify the effort required to do this right now. > Please recompile your binaries on your unidata.ucar.edu host > with the most recent g++ libraries for the Sun (and other > computers). Sorry, but we don't have the resources to maintain binaries for multiple versions of different C++ compilers on a wide range of hardware and OS combinations. For example, we would need to maintain at least six versions of the netCDF C++ library just for Sun platforms (g++ or SunSoft C++ or ObjectCenter C++ for SunOS 4.x or Solaris 2.x) since each compiler uses different name-mangling algorithms for type-safe linkage. We make binary distributions available for a small set of platforms that are currently most used at Unidata university sites. If a sizable number of Unidata university sites began using g++, we would build and maintain g++ binaries. > I am using gcc version 2.5.8 and with the current Makefile > in the c++ directory, I get the following errors on a Sun > with 4.1.3 operating system: > > g++ -c -g -I../libsrc netcdf.cc > In file included from netcdf.hh:16, > from netcdf.cc:12: > ncvalues.hh:20: conflicting types for `unsigned char' > ../libsrc/netcdf.h:281: previous declaration as `char' This appears to be a problem that was fixed with patch 3 to netCDF 2.3.2, which should have been applied to the netcdf-2.3.2pl2 distribution before the patch 4 that was just made available. At least one other user has reported that the latest patch works with the latest release of g++ (2.6.3), but I need to rebuild g++ to verify this. If other users have encountered problems using g++, please let us know (preferably by email to support@xxxxxxxxxxxxxxxx). Also, if anyone figures out how to build a working Fortran interface on top of a gcc-compiled library, several other users would be interested. Thanks. ______________________________________________________________________________ Russ Rew UCAR Unidata Program russ@xxxxxxxxxxxxxxxx P.O. Box 3000 http://www.unidata.ucar.edu/ Boulder, CO 80307-3000 ______________________________________________________________________________ P.S. to the group: netCDF needs a graphical logo. Our WWW pages would look better with a recognizable logo, and we could also use it for the cover of the printed documentation. I have a few pedestrian ideas, but I'm unfortunately "logo design"-challenged, so if anyone out their has any good ideas, please reply. Thanks!
netcdfgroup
archives: