[netCDF #JCT-681222]: Fwd: netcdf fortran configure fail on sizeof

Hi Ted,

I do have an idea; I will describe what I think is going on.  If you provide 
the full config.log, I can confirm or reject my idea.

Lately, when this has cropped up, it is a result of the `conftest` executable 
failing because of a "missing" .so, typically libnetcdf.  This is often the 
case on modern Linux systems when libnetcdf lives outside of '/usr'.  
Specifically, the location of the shared library needs to live in the 
'LD_LIBRARY_PATH' variable.  

The fix is to run configure as follows:

    $ LD_LIBRARY_PATH="/custom/location/lib" CPPFLAGS=<...> LDFLAGS=<...> 
./configure 

If you provide the config.log file, I can probably figure out the exact way you 
would need to invoke configure.  Although it may also be self-evident if you 
take a look yourself.  If config.log doesn't reveal any symptom of a missing 
shared library, I will admit I'm wrong and figure out a new cause from there!

I hope this helps,

-Ward

> 
> Hi,
> 
> I got no response on this from the netcdfgroup mailing list, and I've run 
> into it again on a system that I really need it get it working. Somehow the 
> configure script (or autoconf or something else) is putting extra parentheses 
> around the test variable in the conftest program: it should have (sizeof 
> (off_t)), not (sizeof ((off_t))). The compilation then fails:
> 
> configure:21173: mpicc -c -fPIC -I/usr/local/include 
> -I/home/emansell/opt/pnetcdf161/include 
> -I/home/emansell/local/netcdf4331mpi/include -fPIC conftest.c >&5
> conftest.c(101): error: expected an expression
> if (sizeof ((off_t)))
> ^
> 
> compilation aborted for conftest.c (code 2)
> configure:21173: $? = 2
> configure: failed program was:
> 
> Some more info from config.log is attached at the end. The netcdf-c works 
> fine, obviously, since I got to this step. The hdf5 (1.8.13) files are also 
> in /home/emansell/local/netcdf4331mpi. I have compile this successfully on 
> other systems, but I have no clue on this issue.
> 
> Any ideas?
> 
> Thanks,
> 
> -- Ted
> 
> __________________________________________________________
> | Edward Mansell <address@hidden>
> | National Severe Storms Laboratory
> |--------------------------------------------------------------
> | "The contents of this message are mine personally and
> | do not reflect any position of the U.S. Government or NOAA."
> |--------------------------------------------------------------
> 
> 
> Begin forwarded message:
> 
> > From: Ted Mansell <address@hidden>
> > Subject: netcdf fortran configure fail on sizeof
> > Date: August 8, 2015 at 4:03:48 PM CDT
> > To: "netcdfgroup (address@hidden)" <address@hidden>
> >
> > I'm trying to configure/compile the netcdf-fortran-4 on an AMD/Redhat Linux 
> > system, and 4.2, 4.4.0, 4.4.2 all are having a problem with configure. No 
> > problem with the netcdf-c (e.g., netcdf-c-4.3.1.1). It is failing on the 
> > "sizeof" tests, where an extra set of parentheses is getting put around the 
> > test variable (e.g., _Bool, off_t, etc.) so that the compile of conftest.c 
> > fails because it has
> >
> >  if (sizeof ((_Bool)))
> >
> > when it should be
> >
> >  if (sizeof (_Bool))
> >
> > The netcdf-c does exactly the same test, but apparently does not add the 
> > extraneous parentheses. I'm using Portland Group 10.8 (it's an older AMD 
> > system: Linux alpha 2.6.18-398.el5 #1 SMP Tue Aug 12 06:26:17 EDT 2014 
> > x86_64 x86_64 x86_64 GNU/Linux).
> >
> > I suppose this is some kind of autoconf issue?
> >
> > -- Ted
> >
> > __________________________________________________________
> > | Edward Mansell <address@hidden>
> > | National Severe Storms Laboratory
> > |--------------------------------------------------------------
> > | "The contents of this message are mine personally and
> > | do not reflect any position of the U.S. Government or NOAA."
> > |--------------------------------------------------------------
> >
> 
> This file contains any messages produced by compilers while
> running configure, to aid debugging if configure makes a mistake.
> 
> It was created by netCDF-Fortran configure 4.4.0, which was
> generated by GNU Autoconf 2.69.  Invocation command line was
> 
> $ ./configure --prefix=/home/emansell/local/netcdf4331mpi FC=mpif90 
> F90=mpif90 CC=mpicc F77=mpif90 LDFLAGS=-L/home/emansell/opt/pnetcdf161/lib 
> -L/home/emansell/local/netcdf4331mpi/lib -L/usr/local/lib 
> CPPFLAGS=-I/home/emansell/opt/pnetcdf161/include 
> -I/home/emansell/local/netcdf4331mpi/include -fPIC CFLAGS=-fPIC 
> -I/usr/local/include FCFLAGS=-fPIC --enable-parallel-tests
> 
> ## --------- ##
> ## Platform. ##
> ## --------- ##
> 
> hostname = hrothgar.hpcc.ttu.edu
> uname -m = x86_64
> uname -r = 2.6.32-431.5.1.el6.x86_64
> uname -s = Linux
> uname -v = #1 SMP Wed Feb 12 00:41:43 UTC 2014
> 
> /usr/bin/uname -p = unknown
> /bin/uname -X     = unknown
> 
> /bin/arch              = x86_64
> /usr/bin/arch -k       = unknown
> /usr/convex/getsysinfo = unknown
> /usr/bin/hostinfo      = unknown
> /bin/machine           = unknown
> /usr/bin/oslevel       = unknown
> /bin/universe          = unknown
> 
> PATH: /lustre/work/apps/intel/composer_xe_2013.1.117/bin/intel64
> PATH: /lustre/work/apps/intel/composer_xe_2013.1.117/mpirt/bin/intel64
> PATH: /lustre/work/apps/intel/composer_xe_2013.1.117/bin/intel64
> PATH: /lustre/work/apps/intel/composer_xe_2013.1.117/bin/intel64_mic
> PATH: /lustre/work/apps/intel/composer_xe_2013.1.117/debugger/gui/intel64
> PATH: /opt/gridengine/bin/lx26-amd64
> PATH: /lustre/work/apps/openmpi/bin
> PATH: /lustre/work/apps/svn/bin
> PATH: /sbin
> PATH: /usr/sbin
> PATH: /bin
> PATH: /usr/local/bin
> PATH: /usr/bin
> PATH: /usr/X11R6/bin
> PATH: /lustre/work/apps/bin
> PATH: /home/emansell/bin
> 
> 
> 
> 
> 


Ticket Details
===================
Ticket ID: JCT-681222
Department: Support netCDF
Priority: Normal
Status: Closed