Closed keflavich closed 11 years ago
@keflavich - it looks like the gfortran compiler you are using on the Linux machine is too old. Could you run:
gfortran -dumpversion
and let me know the output?
Regarding the Mac installation, I've pushed a commit (81091590d244b76f8a68655e1ff798d367c4bc3c) that should fix your issue with the HDF5 installation. Could you update to the latest git version and try again?
GNU Fortran (GCC) 4.1.2 20080704 (Red Hat 4.1.2-52)
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
Re: Mac - yes, I'll give it a shot (tomorrow, probably).
On my mac, I get the same error.
fortranlib/src/posix_default.f90:3.19:
use iso_c_binding
1
Fatal Error: Can't open module file 'iso_c_binding.mod' for reading at (1): No such file or directory
etc.
gfortran version is:
GNU Fortran (GCC) 4.2.3
Copyright (C) 2007 Free Software Foundation, Inc.
Ok - I need to check exactly what version of gfortran is required. If you have MacPorts installed, you can install e.g. gcc46 which gives you gfortran-mp-4.6 - then you just do:
./configure FC=gfortran-mp-4.6
I assume I need to do that for the deps install, since the compiler used for the make is h5fc
? Does the python install.py
in fortran/deps
respect the FC
environmental variable? (it seems like it does not; I installed gfortran-4.6 with no success)
Ah yes - sorry - for the dependencies, try:
python install.py --fc-compiler=gfortran-mp-4.6 --cc-compiler=gcc-mp-4.6 -cxx-compiler=g++-mp-4.6 <directory>
Any luck?
Luck indeed! I used the gfortran compilers hosted here: http://gcc.gnu.org/wiki/GFortranBinaries, specifically the gfortran-maintained mac compilers that install to /usr/local/gfortran/ (which is nice because there's a lot that doesn't play nice with gfortran-4.6)
Here are the commands I ended up using:
python install.py /usr/local/hyperion --fc-compiler=/usr/local/gfortran/bin/gfortran --cc-compiler=/usr/local/gfortran/bin/gcc --cxx-compiler=/usr/local/gfortran/bin/g++
FC=/usr/local/gfortran/bin/gfortran CC=/usr/local/gfortran/bin/gcc GCC=/usr/local/gfortran/bin/g++ ./configure --prefix=/usr/local/hyperion
FC=/usr/local/gfortran/bin/gfortran CC=/usr/local/gfortran/bin/gcc GCC=/usr/local/gfortran/bin/g++ make
Now it runs, and I see a bunch of WARNING: photon exceeded maximum number of interactions - killing [do_lucy]
but I assume that means it's working and my model is just dumb.
This solves the mac problem, not the original linux one in the issue header, but I think the answer is: gcc/gfortran >~ 4.5 (?) works.
I just checked my notes, and gfortran 4.3 and above should work. I'll try and clarify the docs for all this.
The note about the number of iterations usually indicates your optical depths are too large. If you continue to have problems with that, let me know!
gfortran test.f90 -o test.exe use modulename 1 Fatal Error: Can't open module file 'modulename.mod' for reading at (1): No such file or directory if file modulename.mod is at /usr/local/include and libraries at /usr/local/lib such error goes away using: gfortran test.f90 -o test.exe -I/usr/local/include -L/usr/local/lib
On a linux machine, I've successfully installed all of the dependencies as far as I can tell - h5py imports, mpif90 exists and looks healthy.
When I tried to ./configure & make the fortran binaries, I ran into a very long list of errors, leading with:
This only happened with the gfortran compiler. When I switched to the ifort compiler (which required re-installing the dependencies with ifort included in my path), there were no errors.
This is a non-urgent bug, but I figured it's worth reporting because of the apparent incompatibility with gfortran. I also wonder if general gfortran incompatibility issues might explain my mac install issues.