Hi Reto, > I've now built and tested both the 3.6.3 and 4.1.3 on OSX with gcc/gfortran > 4.5.3. It is correctly noted in the installation instruction that the > -DgFortran (or -DpgiFortran) cpp flag is automatically detected and set > during the configuration. This works only if the respective fortran compiler > is gfortran (pgf90/95). For NetCDF-4 compilation using MPI the fortran > compiler is a wrapper like mpif90 and thus -DgFortran (-DpgiFortran) have to > be manually set by the user. > > Would it be possible to either add this to the Installation instructions or > alternatively find a way to auto-detect the compiler type even though the > compiler is embedded in a MPI wrapper? Thanks for the suggestion, I've added it to our list of tasks. If you want to follow its progress, you can see the Jira entry here: https://www.unidata.ucar.edu/jira/browse/NCF-84 --Russ > On Jun 20, 2011, at 6:34 PM, Unidata netCDF Support wrote: > > > Hi Reto > > > >> I've upgraded to both NetCDF 3.6.3 and the fink install of gcc/gfortran > >> 4.5.2 on OSX 10.6 x86_64. NetCDF 3.6.3 builds fine and runs with both > >> the C and F90 interfaces. So far so good. > >> > >> What worries me is the following warning during build-time, that was > >> never there before in earlier NetCDF or gcc/gfortran versions. This may > >> be a hint that the array is initialized with 0 dimension size and may > >> cause problem under certain circumstances? > >> > >> libtool: compile: gfortran -I../libsrc -I. -g -O2 -c netcdf.f90 -o > >> netcdf.o > >> netcdf_text_variables.f90:284.93: > >> Included at netcdf.f90:51: > >> > >> localMap(:numDims ) = (/ 1, (product(localCount(:counter)), counter = 1, > >> numDims - 1) /) > >> 1 > >> Warning: DO loop at (1) will be executed zero times > >> netcdf_text_variables.f90:60.93: > >> Included at netcdf.f90:51: > >> > >> localMap(:numDims ) = (/ 1, (product(localCount(:counter)), counter = 1, > >> numDims - 1) /) > >> 1 > >> Warning: DO loop at (1) will be executed zero times > >> > >> > >> Please let me know what you think. numDims is initialized earlier on in > >> netcdf_text_variables.f90 with the value of 1, so numDims - 1 is 0, > >> which means that the localMap(:numDims) will be of size 0. Is that ok or > >> not? > > > > It's OK, and has never caused any reported problems other than this > > harmless warning that gfortran added in an upgrade. Getting rid of > > the warning actually makes the code less maintainable, but it's > > probably worth it for reassuring users who look at the compiler > > warnings that it's not a symptom of a bug. > > > > By the way, we just released netCDF-4.1.3 and it can be built with the > > --disable-netcdf-4 configure option that results in a netCDF-3 library > > completely compatible with the old netCDF-3.6.3 library, but with some > > bug fixes and enhancements that haven't been incorporated in the 3.6.3 > > release. > > > > --Russ > > > > Russ Rew UCAR Unidata Program > > address@hidden http://www.unidata.ucar.edu > > > > > > > > Ticket Details > > =================== > > Ticket ID: QVV-651237 > > Department: Support netCDF > > Priority: Normal > > Status: Closed > > > > Russ Rew UCAR Unidata Program address@hidden http://www.unidata.ucar.edu Ticket Details =================== Ticket ID: QVV-651237 Department: Support netCDF Priority: High Status: Closed
NOTE: All email exchanges with Unidata User Support are recorded in the Unidata inquiry tracking system and then made publicly available through the web. If you do not want to have your interactions made available in this way, you must let us know in each email you send to us.