* Fixes#1454; still needs a bit more thinking though
* forgot the necessary version bump
* review related cleanup and a fix to allow the SWIG wrappers to be built
* apply (manually) #1807 from @bp-kelley to modern_cxx branch and fix (hopefully) a windows problem
* another windows problem
* another windows problem
* windows now builds locally
* update link targets
* New syntax to import Qt5 dependency in MolDraw2D
* In QTDemo, not necessary to import Qt explicitly because it is automatically imported through target MolDraw2D
* improve import of Cairo dependency
* Remove explicit linking of Cairo in C++ examples
* Use imported targets for Boost
* Using imported CMake targets for Eigen
* Use imported targets for Threads
* Simplify sample CMakeLists
* Remove duplication
* a couple small changes for travis
* typo
* update toolchain to trusty
* switch to requiring at least boost 1.56
* typo
* update boost dependency to pull in the build that has more libraries
* meaningless commit to trigger travis
* switch to ust libc++
* install libc++
* check abi
* try the other one; disable py27 for the moment to speed things up
* switch to c++11, that seems necessary with the conda boost builds at the moment
* re-enable python2.7
* update .gitignore
* foundation for 3D descriptors
move PBF into core
* cleanup work
* a bit more cleanup
* move the principal moments calc to MolTransforms
* cleanup
* cleanups
* add caching of the principal moments and values
* do not include the 3D descriptors in MolDescriptors.h
* the properties are computed
* add PMI descriptors and tests
* add tests for NPR descriptors
* return 0 when the largest PMI is zero
* PMI edge case tests
* NPR edge case tests
* PBF edge case tests
* PBF edge case tests
* more edge cases
* add a few more 3d descriptors
* add defns to docs
* tests for the new descriptors
* add versions to new descriptors
* add 3d descriptors to python wrapper
* add eigen support to the travis build
* try to get non-windows builds working
* remove computeCovarianceMatrix() from java wrapper
* make pmi property names "private"
* - this should fix the problem with "Inchi" being hardcoded into PostgreSQL
* - moved the InChI check code out of the main CMakeLists.txt
into FindInchi.cmake
* - made Code/GraphMol/DistGeomHelpers/testDgeomHelpers.cpp and Code/GraphMol/DistGeomHelpers/Wrap/testDistGeom.py
more robust against very small coordinate variations which may induce UFF energy variations in the 1-kcal/mol range.
This was observed in a single case using the MinGW 32-bit g++ compiler. The check for identity of conformations
between single- and multi-threaded runs is now complemented by a 3D coordinate MSD check
* - lowered the energy tolerance to 1.0 kcal/mol
* - added rdkitBuild string with compile-time information
- added runtime check in Code/GraphMol/DistGeomHelpers/testDgeomHelpers.cpp
and Code/GraphMol/DistGeomHelpers/Wrap/testDistGeom.py
to use more lenient tolerances with MinGW compilers
To use code coverage:
cmake -DCMAKE_BUILD_TYPE=Debug \
-DRDK_USE_FLEXBISON=ON \
-DRDK_TEST_COVERAGE=ON \
<path to rdkit source tree>
make install
make RDKit_coverage
Note, when making RDKit_coverage, your python paths, (DY)LD_LIBRARY paths must be
set up correctly. If any of the tests fail, the coverage generator will most
likely fail and certainly will not be complete.
Here is an example from my OSX Box (note, I set the install directory to rdkit_build via)
cmake -DCMAKE_INSTALL_PREFIX=`pwd`/rdkit_build \
...
RDBASE=<path to rdkit source> DYLD_LIBRARY_PATH=`pwd`/rdkit_build/lib \
PYTHONPATH=`pwd`/rdkit_build/lib/python2.7/site-packages \
make RDKit_coverage
open coverage/index.html in a browser
