Resolve "VariableRF, FFA, Offset and MultipoleT: unify units internally"

Unfortunately I get a fail at compile time, probably unrelated to the issue. It looks like a problem from Vektor, defined in ippl/src/AppTypes/Vektor.h, although diff tells me there are no differences between my IPPL AppTypes and the version in this source tree.

[ 20%] Building CXX object src/CMakeFiles/libOPALObj.dir/AbstractObjects/BeamSequence.cpp.o
In file included from /usr/include/c++/11/cassert:44,
                 from /home/cr67/Software/install/include/boost/numeric/ublas/detail/config.hpp:16,
                 from /home/cr67/Software/install/include/boost/numeric/ublas/exception.hpp:19,
                 from /home/cr67/Software/install/include/boost/numeric/ublas/storage.hpp:26,
                 from /home/cr67/Software/install/include/boost/numeric/ublas/vector.hpp:21,
                 from /home/cr67/Software/install/include/boost/numeric/ublas/matrix.hpp:18,
                 from /home/cr67/Software/OPAL/opal_src_3/src/Algorithms/BoostMatrix.h:21,
                 from /home/cr67/Software/OPAL/opal_src_3/src/Classic/Algorithms/CoordinateSystemTrafo.h:4,
                 from /home/cr67/Software/OPAL/opal_src_3/src/Classic/AbsBeamline/ElementBase.h:67,
                 from /home/cr67/Software/OPAL/opal_src_3/src/AbstractObjects/Element.h:34,
                 from /home/cr67/Software/OPAL/opal_src_3/src/AbstractObjects/BeamSequence.h:21,
                 from /home/cr67/Software/OPAL/opal_src_3/src/AbstractObjects/BeamSequence.cpp:19:
/home/cr67/Software/OPAL/opal_src_3/src/Algorithms/BoostMatrix.h: In instantiation of ‘T prod_boost_vector(const boost::numeric::ublas::matrix<double>&, const T&) [with T = Vektor<double, 3>]’:
/home/cr67/Software/OPAL/opal_src_3/src/Classic/Algorithms/CoordinateSystemTrafo.h:78:29:   required from here
/home/cr67/Software/OPAL/opal_src_3/src/Algorithms/BoostMatrix.h:28:19: error: ‘const class Vektor<double, 3>’ has no member named ‘size’; did you mean ‘Size’?
   28 |     assert(vector.size() == 3);
      |            ~~~~~~~^~~~
make[2]: *** [src/CMakeFiles/libOPALObj.dir/build.make:160: src/CMakeFiles/libOPALObj.dir/AbstractObjects/BeamSequence.cpp.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:2182: src/CMakeFiles/libOPALObj.dir/all] Error 2
make: *** [Makefile:136: all] Error 2

approved this merge request

I don't see how anything can build at all, the compile error is correctly flagging a non-existent function that is called by BoostMatrix, namely Vektor::size. Anyway, I agree it is not related to this issue and it can be worked around by commenting the relevant assert statement.

PyOpal tests are not run by default because it is still an experimental feature. I don't know how to make it all work because it depends on testing infrastructure that I do not have any sight of (like the nightly tests and things). It should be possible make some sort of C++ unit test that auto-runs the PyOpal tests, and even sees the relevant CMake compile flag, but I have not implemented that.

Upon spending a morning digging, it looks like a terrible evil has befallen the PyOpal tests. Some horrible confusion of Ippl and MPI that I do not find straight forwards to untangle. Sigh.

Anyway, in general the elements in src/PyOpal/PyElements wrap the OpalElement, so this should be correctly implemented by the changes here, I hope. However, there are direct calls to element::apply in order to directly access the field values (for user validation), and implemented as get_field_value on the python side, where unit conversions are necessary to maintain the correct user-side units.

I raised the build error as #789 (closed)

added 2 commits

7b9fcfc9 - fix PyOpal units
4ef05b2e - add license header

Compare with previous version

added 1 commit

09849174 - replace NULL by mullptr

Compare with previous version

Admin message

Resolve "VariableRF, FFA, Offset and MultipoleT: unify units internally"

Merge request reports

Activity