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

[netCDF #JJE-781466]: NetCDF 3.6.2, gfortran, cygwin

> Support,
> 
> My colleague and I both have Windows XP notebooks with up-to-date
> CYGWIN installations.
> 
> We both followed this steps:
> 
> A. gfortran.
> 
> 1. Get the latest gfortran
> 
> http://gcc.gnu.org/wiki/GFortranBinaries
> 
> 2. Install:   tar -xjvf gfortran-4.3-Cygwin-i686.tar.bz2  -C  /
> (this will install the gfortran in /usr/local/gfortran/bin)

Sigh.

Sometimes computers can be such a trial!

I installed gfortran, and it gives me it's version, but will not compile a test 
program for me:

ed@DUDLEY ~
$ gfortran --version
GNU Fortran (GCC) 4.3.0 20070512 (experimental)
Copyright (C) 2007 Free Software Foundation, Inc.

GNU Fortran comes with NO WARRANTY, to the extent permitted by law.
You may redistribute copies of GNU Fortran
under the terms of the GNU General Public License.
For more information about these matters, see the file named COPYING


ed@DUDLEY ~
$ gfortran -o test test.f90

ed@DUDLEY ~
$ ls
%backup%~            local   n3_new2   netcdf-3.6.2-beta6     tmp1.sh
backups              mnt     n3_test   netcdf-daily-1.tar.gz  tmp_profile
emacs.exe.stackdump  n3_new  netcdf-3  test.f90               tmp_profile.sh

ed@DUDLEY ~
$ ./test
-bash: ./test: No such file or directory

ed@DUDLEY ~
$ ./test.exe
-bash: ./test.exe: No such file or directory

I tells me the compile succeeded, but there is no output executable!

I am going to keep trying to get gfortran to work.


> 
> B. Netcdf and netcdf-f90 interface with gfortran support
> 
> 1. Get netcdf3.6.2 (stable version) ..
> 
> 2. make distclean    (if it allows...we want to make sure we are
> really building from scratch)
> 
> 3. CC=gcc  FC=gfortran F90=gfortran F77=gfortran
> CPPFLAGS=-DpgiFortran  ./configure
> --prefix=/usr/local/netcdf/gfortran
> (Note: I have removed the F90FLAGS and forced the FC and F77 to
> use gfortran)

You only need to set FC, not F77 and F90.

> 
> 4. make
> 
> 5. make check
> 
> 
> On my colleagues system, all TESTS PASS.
> 
> On my system, I start out with all PASS, but then eventually,
> 
> ..
> *** All ncgen and ncdump test output for 64-bit offset format passed!
> PASS: tst_output.sh
> *** testing length of classic file
> FAIL: tst_lengths.sh
> =========================================
> 1 of 6 tests failed
> Please report to address@hidden
> =========================================
> make[2]: *** [check-TESTS] Error 1
> make[2]: Leaving directory `/usr/local/netcdf-3.6.2/ncdump'
> make[1]: *** [check-am] Error 2
> make[1]: Leaving directory `/usr/local/netcdf-3.6.2/ncdump'
> make: *** [check-recursive] Error 1
> 
> 
> Any ideas of what might cause this difference?

I will notify Russ, the programmer involved in this test, and try to reproduce 
it on my two cygwin machines here.

However, this has nothing to do with gfortran. This is a C library issue.

If all other tests pass (in directories nf_test, nc_test, etc.), then I would 
just go ahead and use it while we work on this problem.

This relates to the length of the netCDF file. Recently we discovered that some 
very small netCDF files would get zero-padded out to some slightly larger size. 
This does not affect the user of the file in any way, except the file takes up 
slightly more disk space than it should. And only files below a certain (pretty 
small) size are affected.

Recently Russ fixed this bug, but apparently windows is doing something 
different from the rest of the world. (As it often does). But only sometimes, 
because this test also passes on our cygwin testing machine:

http://www.unidata.ucar.edu/software/netcdf/builds/snapshot/

Have you tried updating your cygwin installation with all the latest versions 
of everything? My test machine is reasonably up to date...

Thanks,

Ed

Ticket Details
===================
Ticket ID: JJE-781466
Department: Support netCDF
Priority: Normal
Status: Open