- Feature #1152: Coords has been split in two classes: Coords (to be renamed soon probably) and SpatialPoint. The former is tightly related to a mesh, the second not (or else it's a Coords: Coords inherits from SpatialPoint). - Feature #1144: Fibers will now be built on a domain and not the entire mesh (this is a proper implementation of what was wrongfully attempted in #1117). - Feature #1124: Implement VariableTimeStep policy for TimeManager. - Support #1157 - #1143 - #1149: Integration tests have been cleant-up and completed: some tests that weren't there have been added, and for all models EnsightOutput is now checked both in sequential and parallel cases. Tests have been rewritten so that it could be more easily integrated in a continuous integration system (for instance some of them now yield an exception rather than just print a result on screen that must be interpreted). - Support #1129: Add the list of enclosing Domain in Coords. - Support #1156: Add a function to check whether two files share the same content. - Support #1154: Coords constructors are now private. - Support #1151: Coords copy and move constructors are now deactivated. - Support #1145: Implement wrappers over MPI_Gatherv and MPI_Allgatherv; add a test case to check the behaviour is fine. - Support #1142: Add a facility in VariationalFormulation to apply all boundary conditions to a vector or a matrix different to the system one. It's just a nice wrapper over GodOfDof namesake method. - Support #1141: Change print of time to Time: 0. -> 0.01. - Support #1138: Constant Time Step policy should not increment the time with time step as the numerical error propagates; it's better to recompute from initial time and time step index. - Support #1137: New function heaviside in Numeric. - Support #1135: New features for Petsc matrices and vectors. - Support #1134: Write a function to convert char array into a string. - Support #1133: Add error code support for Mpi. Previously there was a call to a function to trigger an abort whenever am operation fails, but it's clear Openmpi doesn't implement it properly, whereas error code works as expected. - Support #1132: Coords pointers have been changed from shared to unique or raw. - Support #1131: Extend PointerComparison so that it may work upon unique pointers. - Support #1130: Implement properly NcoordsInDomain. - Design #1136: EnsightCaseReader can't be in Core and depends on Geometry as it breaks the hierarchy. - Bug #1155: Underlying vector was not automatically updated after the destruction of AccessVectorContent<Utilities::Access::read_and_write>. - Bug #1153: Movemesh: ghost Coords are actually neglected - Bug #1148: Assert in Nunknown ElementaryDataImpl.hxx was outdated (the implementation itself was flawed, even if it was of no consequence for most of the operators as they seldom feature three unknowns or more). - Bug #1139: Sign() function was incorrect in a very specific case. - Bug #1127: Remove and fix feature #1117 (Replaced by #1129). - Bug #1121: A Finite Element Space dof_list_ contained dofs that are not related to the NumberingSubset and Unknown used to define it. WARNING: Fix is not complete; there is a hack for shape function labels that is to be addressed in #1146. Ongoing tickets: - Support #1081: Introduce some C++ 17 features supported by both clang and gcc (most notably if constexpr...). This truly enables full support of #1076: it's now possible to define a GlobalVariationalOperator that can't act on some RefGeomElt (for instance Elastic operator now can't work upon 0D and 1D geometric elements).