Closed krivenko closed 12 years ago
Steps to reproduce the problem:
1) Compile and install TRIQS as described in the manual (with SHARED_LIBS=ON
).
2) Go to ${CMAKE_INSTALL_PREFIX}
and make clean
.
3) Invoke ${CMAKE_INSTALL_PREFIX}/INSTALL_DIR/bin/pytriqs
4) In the Python shell run an example about Green's functions. The import statement from pytriqs.Base.GF_Local import GFBloc_ImFreq
should fail.
@krivenko: Can you check that the above commit solves the problem and close the ticket if so?
@mferrero: Yes, now the installation is complete and works well. Thanks a lot!
Sorry, but there is another issue, perhaps caused by your commit fc11ffb
Hamiltonian with Eps0 term :
2.5 C^(up,1)C(up,1)C^(down,1)C(down,1) + C^(up,1)C(up,1) + C^(down,1)C(down,1)
Inv Fourier done
Traceback (most recent call last):
File "./Jernej.py", line 37, in <module>
S.Solve()
File "/usr/lib64/python2.7/site-packages/pytriqs/Solvers/HybridizationExpansion/__init__.py", line 296, in Solve
C_Module.MC_solve(self.__dict__ ) # C++ solver
TypeError: No registered converter was able to produce a C++ rvalue of type GF_C<GF_Bloc_ImTime> from this Python object of type GF
How can I reproduce this error? Is it during the make test?
How can I reproduce this error? Is it during the make test?
Try to run, for example, the script from #17.
I have compiled and installed TRIQS successfully (git master) following the user manual page. However there is an implicit issue with the installation. When the
SHARED_LIBS
option is enabled, python modules in the resulting installation are linked againstlibLocalGreenFunction_c.so
andlibtriqs.so
residing in the build directory. The problem gets partially solved by switching onInstall_dev
option, because thenlibtriqs.so
is installed into${CMAKE_INSTALL_PREFIX}/lib
. I guess this behavior was introduced with the static linking in mind, because it indeed makes sense to install the static library only to develop TRIQS itself. Nevertheless, both those shared libraries must be installed into${CMAKE_INSTALL_PREFIX}/lib
and properly linked to make it possible to create binary distributions of TRIQS.