1.6. Troubleshooting¶
On a fresh Linux or Mac machine, the instructions given in the “Download and Install” guides should result in a working installation of HORTON, without using any of the suggestions below. In reality, however, the Unix system of an average researcher isn’t pristine but rahter ranges from cleverly customized to completely borked. Such customizations may interfere with the installation of HORTON. This section provides some guidance for the novice Unix user on how to get HORTON working on a not-so-well-maintained Unix system.
If you are still stuck after trying the suggestions in this section, do not hesitate to contact us on the the HORTON mailing list.
1.6.1. Introduction¶
Let us assume you have already built and installed all your dependencies. However, when you try to install HORTON, i.e.
./setup.py install --user
or when you run nosetests, you get an unexpected error message. The problem
is most likely related to finding and using the dependencies. You have to make
sure setup.py and the HORTON modules can find the right dependencies and are
able to use them. We have seen problems with several types of dependencies: missing Python
packages, executables, libraries and failing tests.
1.6.2. Python packages¶
If you have installed a python package (e.g. NumPy, SciPy, Cython, H5Py,
SymPy, MatPlotLib, Nosetests, Sphinx, Breathe, Docutils) and you get an error
saying your system cannot find that package, then you need to check the
directories in which Python searches for package. These are stored
in the attribute path of the sys module, which can be accessed by:
python -c "import sys; import pprint; pprint.pprint(sys.path)"
A typical output can be as follows (but is probably different in your case):
['',
'/usr/lib64/python27.zip',
'/usr/lib64/python2.7',
'/usr/lib64/python2.7/plat-linux2',
'/usr/lib64/python2.7/lib-tk',
'/usr/lib64/python2.7/lib-old',
'/usr/lib64/python2.7/lib-dynload',
'/home/foo/.local/lib/python2.7/site-packages',
'/usr/lib64/python2.7/site-packages',
'/usr/lib64/python2.7/site-packages/gtk-2.0',
'/usr/lib/python2.7/site-packages']
If you installed a Python package in another directory, Python will not be able
to load it. This can be fixed by adding your directory to the PYTHONPATH
variable in your ${HOME}/.bashrc (Linux) or ${HOME}/.bash_profile (Mac),
e.g.
export PYTHONPATH=/some/custom/path/lib/python2.7/site-packages:${PYTHONPATH}
The next time you start Python (or any program implemented in with Python), the
packages you installed in a non-standard location will become importable. If the
same Python module or package is installed in multiple directories, the one
found in the first directory in the sys.path list takes precedence.
A typical problem is that there are multiple lines like these in .bashrc or
.bash_profile of which the last is overwriting the former ones, e.g.:
export PYTHONPATH=/some/custom/path/lib/python2.7/site-packages:${PYTHONPATH}
# and several lines further ...
export PYTHONPATH=/some/other/path/lib/python2.7/site-packages
The second export line overrides the first one because it does not end with
:${PYTHONPATH}.
Some examples are given below. Note that, in principle, none of these should be necessary but they seem to have helped people with a broken installation of Python:
Some Mac users needed to set the
PYTHONPATHafter installing modules through PIP:export PYTHONPATH=${HOME}/Library/Python/2.7/lib/python/site-packages:${PYTHONPATH}
or their system site-packages:
export PYTHONPATH=/Library/Python/2.7/lib/python/site-packages:${PYTHONPATH}
Similarly, a few Linux users needed to set
PYTHONPATHafter installation through PIP:export PYTHONPATH=${HOME}/.local/lib/python2.7/site-packages:${PYTHONPATH}
or
export PYTHONPATH=/lib/python2.7/site-packages:${PYTHONPATH}
or
export PYTHONPATH=/lib64/python2.7/site-packages
1.6.3. Excecutables¶
During the installation (or when building the documentation) HORTON will use
some executables, e.g. a compiler, sphinx-build, etc. These executables must
be in one of the directories in the PATH environment variable. The essential
changes to the PATH variable were already discussed in the “Download and
install” guides but if your system is somehow broken, more changes may be
needed.
The contents of PATH can be accessed by:
echo $PATH
In unfavorable circumstances, some directories may be missing from the PATH,
e.g because it got carelessly overwritten in ${HOME}/.bashrc (Linux) or
${HOME}/.bash_profile (Mac). For example, the following should be avoided:
export PATH=/some/custom/path/bin
Instead, make sure the existing PATH variable is included as follows:
export PATH=/some/custom/path/bin:${PATH}
If the same executable name occurs in several directories in the PATH, the
one in the first directory takes precedence.
The following examples are in principle not needed but they seemed to be helpful for some:
Mac users that uses python scripts might do
# Already mentioned in "Download and install" guide: export PATH=${HOME}/Library/Python/2.7/bin:${PATH} # Should already be in the PATH anyway, unless your system is broken: export PATH=/Library/Python/2.7/bin:${PATH}
Similarly, Linux users may do
# Already mentioned in "Download and install" guide: export PATH=${HOME}/.local/bin:${PATH} # Should already be in the PATH anyway, unless your system is broken: export PATH=/usr/bin:${PATH}
When you forgot where you installed a dependency, the find command may help
you find the appropriate directory. The following example will search for
location of the sphinx-build executable:
find / | grep sphinx-build
1.6.4. Libraries¶
LibXC-3
When you get the following error message upon running
./setup.py, you are trying to compile HORTON with LibXC-3:gcc -pthread -fno-strict-aliasing -O2 -g -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -m64 -mtune=generic -D_GNU_SOURCE -fPIC -fwrapv -DNDEBUG -O2 -g -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -m64 -mtune=generic -D_GNU_SOURCE -fPIC -fwrapv -fPIC -I/usr/lib64/python2.7/site-packages/numpy/core/include -I. -I/usr/include/python2.7 -c horton/meanfield/cext.cpp -o build/temp.linux-x86_64-2.7/horton/meanfield/cext.o -std=c++11 In file included from /usr/lib64/python2.7/site-packages/numpy/core/include/numpy/ndarraytypes.h:1777:0, from /usr/lib64/python2.7/site-packages/numpy/core/include/numpy/ndarrayobject.h:18, from /usr/lib64/python2.7/site-packages/numpy/core/include/numpy/arrayobject.h:4, from horton/meanfield/cext.cpp:449: /usr/lib64/python2.7/site-packages/numpy/core/include/numpy/npy_1_7_deprecated_api.h:15:2: warning: #warning "Using deprecated NumPy API, disable it by " "#defining NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION" [-Wcpp] #warning "Using deprecated NumPy API, disable it by " \ ^~~~~~~ horton/meanfield/cext.cpp: In function ‘PyObject* __pyx_pf_6horton_9meanfield_4cext_12LibXCWrapper_4refs___get__(__pyx_obj_6horton_9meanfield_4cext_LibXCWrapper*)’: horton/meanfield/cext.cpp:2110:74: error: cannot convert ‘func_reference_type* const*’ to ‘const char*’ for argument ‘1’ to ‘PyObject* PyString_FromString(const char*)’ __pyx_t_1 = __Pyx_PyBytes_FromString((__pyx_v_self->_func.info[0]).refs); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 117, __pyx_L1_error) ^ error: command 'gcc' failed with exit status 1
The solution is to install LibXC-2.2.2. This can always be done by running
./tools/qa/install_libxc-2.2.2.shbefore running./setup.py.
1.6.5. Failing tests¶
The following failing tests are symptoms of specific problems:
horton.meanfield.test.test_libxc.test_dot_hessian_o3lyp_cs_polynomial. This is most likely caused by linking against a LibXC that has been compiled with too aggressive optimization flags. Use the script/toos/qa/install_libxc-2.2.2.shto build a more modest version of LibXC, which can then be used to compile HORTON.