• Support #1544: Use StrongTypes in PrintContainer facility. Doing so lessen the risk of swapping the string separating items and the one opening the printed container for instance.
  • Support #1552: Enable std::string key in PrintContainer for associative containers in Lua format
  • Support #1554: StrongType: add operators <, == and <<
  • Support #1557 StrongType: add support for inheriting some of the underlying type functionality.
  • Support #1563 Extend IsString so that a class may be defined to register true when queried.
  • Support #1572 PrintContainer: slight modification for better behaviour in parallel (limit interspeding between ranks).
  • Support #1579 Enforce ordering in automatically generated SourceList.cmake files to avoid pollution from need
  • Support #1569 Add a wrapper to set the divergence tolerance for SNES (PETSc).


  • Support #1558: Complete test for Mpi Send/Receive.
  • Support #1574: Add a test checking Xtensor behaves properly.
  • Support #1576: Mpi: add a functionality to check a container is the same on all ranks.
  • Support #1578: Update the code to fit with new versions of third party libraries
  • Support #1580: Mpi: Secure usage of Mpi::Broadcast.


  • Support #1548: Check of format support should be accessible in RefGeomElt
  • Support #1555: Use StrongType to differentiate Ensight name and generic name in Geometry
  • Support #1556: Also use StrongType for the different Coords indexes
  • Support #1561: Make sure Ensight format is properly supported for runs from prepartitioned data.


  • Feature #1545 - Feature #1570: Refactor heavily the partition process, so that it is now much more straightforward, and also much closer for the vanilla parallel process and for a run from prepartitioned data. #1545 began as a much smaller-scale issue (investigate how to avoir storing pattern matrix) bbut end up doing many modifications, including a better definition of what are ghost elements for objects and a better track of them (for some objects their whereabouts were rather blurry - and it shows whern we try to reload from prepartitioned data. #1570 added a huge memory benefit: now Node and Dofs are created only after the partition is done, thus consuming much less memory in initialization phase when running a parallel run.
  • Support #1573: Do not put too many times the warning message about operator related to an empty FEltSpace
  • Feature #1575: Introduce NdofHolder (and thus easy access to information about number of dofs) in FEltSpace
  • Support #1590: Split GodOfDof.cpp into several smaller files.
  • Support #1591: Tidy a bit the internal directory in FiniteElementSpace folder.
  • Feature #1567: Implement Q0 elements.


  • Feature #1506 Separate time schemes for the penalization law used by the QuasiIncompressibleSecondPiolaKirchhoffStressTensor.
  • Feature #1550 Add a deviatoric version of the exponentialJ1J4 law.


  • Feature #1521 Add the option for fiber parameters to be defined at quadrature points instead of nodes.


  • Support #1559 Gitlab: use the new needs field (which enables checking the warnings without having to wait for all the build jobs to be done).
  • Feature #1543: Introduce a Python script which interprets the summary of a Valgrind output, and use it to loosen very slightly the br to pass Valgrind test (a tolerance of ONE leak with less than 10 bytes is introduced - because a leak of 8 bytes appeared with the update of third party libraries and Valgrind versions).
  • Bug #1546: Fix Sonarqube path.


  • Support #1582: LGPL license has been added.

Ongoing tickets

  • Feature #1577: Ticket #1570 broke the FromVertexMatching interpolator. As this one is used so far only in two models not currently active (FSI and Poromechanics) I temporarily decommissioned the operator, as I really wanted a release of all the other improvements in the library. This ticket will be handled with high priority for next release.
  • Feature #1560: Decommission OutputDeformedMesh from PostProcessing. This has been done but needs to be discussed in code meeting: if the functionality is needed we must refresh it and add some tests; if not the code must be removed entirely.
  • Support #1562: replace my macro HAS_MEMBER by new std::is_detected. However this new feature is still in std::experimental - I will replace entirely when the support is in std namespace directly for both clang and gcc. The test currently check both my macro and new feature works.