- Support #679: Add the possibility to inverse another matrix than system_matrix.
- Bug #701: Singleton definition assumed the singleton didn't include data related to MPI (that can't be used' after the call to Mpi::Finalize(). A manual method to destroy the singleton has been introduced to counter this issue; singleton implementation had to be amended slightly to allow this. Base Model now provides a Finalize() that destroy two such singletons.
- Support #702: Update SCons compilation (add new targets, fix issues).
- Feature #699 - #692: Introduce a FiberManager class to handle Ensight fiber files. A new Parameter policy has been introduced for that purpose.
- Bug #698: Geometry/mpi: Number of vertices wasn't correcly after reduction.
- Feature #694: In the input parameter file, modify the interface to choose the solver. Now it's more straightforward to choose Mumps and it's possible to choose UMFPACK, which was completely impossible before. Iterative solvers have been forbidden: I want to reintoduce them when they are needed and after checking they work as intended (the list was the one from FELiScE and none has ever been checked).
- Bug #690: Computation of dof list inside a finite element space was faulty: some dofs could be forgotten. It is now computed an entirely different way.
- Bug #688: Dirichlet boundary conditions were updated only for one numbering subset, not for all of them (this was only visible in FSI as it is the only advanced enough model to manipulate boundary conditions to update and several numbering subsets).
- Bug #686: Bulk term was missing in both hyperelastic laws.
- Feature #657 Conform interpolator. Therein:
. Feature #658: LocalNode now stores a brand-new LocalInterface object, which is able to tell the indexes of the vertices that delimit the interface.
. Bug #659: In GodOfDof, mpi barrier creation of dof information file was buggy.
. Bug #660: Choice of processor for a local_felt_space was made independantly for each finite element space, which could lead to inconsistencies. Now a finite element space must make the same decision of a previous one if a processor was already allotted.
. Feature #661: P1 -> P2 interpolator.
. Feature #668: P2 -> P1 interpolator.
. Feature #663 - Bug #683: Implement the current ConformInterpolator in terms of class newly introduced.
. Feature #664 - #672: Implement the possibility to do something like (v, p) -> p in any interpolators (done in both P2->P1 and P1->p2 so far).
- Support #666: Petsc sequential vectors may now be initialized from a sequential vector dumped by Print() method. Parallel case is entirely dismissed at the moment: this feature is only there for test purposes.
- Design #655 - Bug #678: FEltSpaceProgramWise index has been removed: the cases it handled should be managed solely by numbering subsets.
- Bug #651: Fix runtime errors of Ensight output, due to ill-defined input parameter tuple.
- Support #645: Compute local2global required in InitializeVectorSystemSolution() if they do not exist yet.
- Bug #693: All boundary conditions were always applied; interface has been modified so that each VariationalFormulation may cherry-pick the one it needs.
- Feature #608 [still in progress]: implement FSI model. Currently works with elastic formulation for solid, aitken convergence scheme, no Nitzsche operator (term is dismissed entirely) and one aggregated mesh decomposed in subdomains for solid and fluid.