Implement Barry's comments

MatISRestoreLocalMat: new (currently no-op) paired with MatISGetLocalMat

PCBDDCGRAPH: always print cc when verbosity_level > 2

PCBDDC: include self in csr connectivity out of DMPLEX

PCBDDC: fix warnings from nightly builds

Merge branch 'master' into stefano_zampini/feature-fetidp-ex69

fix compile error with mlib build
arch-linux-pkgs-cxx-mlib/obj/src/ksp/pc/impls/bddc/bddcschurs.o: In function `PCBDDCSubSchursSetUp(_PCBDDCSubSchurs*, _p_Mat*, _p_Mat*, PetscBool, int*, int*, int, _

fix compile error with mlib build
arch-linux-pkgs-cxx-mlib/obj/src/ksp/pc/impls/bddc/bddcschurs.o: In function `PCBDDCSubSchursSetUp(_PCBDDCSubSchurs*, _p_Mat*, _p_Mat*, PetscBool, int*, int*, int, _p_Vec*, PetscBool, PetscBool, PetscBool, int, int*, _p_IS**, _p_Mat*, _p_IS*)':
/sandbox/petsc/petsc.clone-3/src/ksp/pc/impls/bddc/bddcschurs.c:1302: undefined reference to `MatSeqDenseInvertFactors_Private'
/sandbox/petsc/petsc.clone-3/src/ksp/pc/impls/bddc/bddcschurs.c:1316: undefined reference to `MatSeqDenseInvertFactors_Private'
/usr/bin/ld: /sandbox/petsc/petsc.clone-3/arch-linux-pkgs-cxx-mlib/lib/libpetscksp.so.3.7.6: hidden symbol `MatSeqDenseInvertFactors_Private' isn't defined

show more ...

PCBDDC: fix reusage of solvers

If we are not reusing the solvers, remove them if they are present

PCBDDC: fix logic in reusing solver

PCBDDC: flag the fake matrix as assembled

Merge branch 'tisaac/feature-plex-deterministic-viewing' into tisaac/feature-plex-tests-new

* tisaac/feature-plex-deterministic-viewing:
DMLabel: sort strata after distributing
PetscError() did

Merge branch 'tisaac/feature-plex-deterministic-viewing' into tisaac/feature-plex-tests-new

* tisaac/feature-plex-deterministic-viewing:
DMLabel: sort strata after distributing
PetscError() did not properly copy over the specific error message when error type was PETSC_ERROR_INITIAL
Additional clarification for MatMeshToCellGraph() manual page
med: --with-hdf5, not --with-hdf5-dir
DMPlex_GMSH: revert to not inverting triangles
sys tests ex13f: fix type mismatch in composition test
PCBDDC: use a sparse rhs when the solver is MUMPS
Sys ex13: Added test for object composition in Fortran
Sys: Added Fortran binding for object composition
MatMUMPS: added support for sparse right-hand sides in sequential

show more ...

PCBDDC: correct order of printf arguments

Merge branch 'stefano_zampini/feature-allow-bddc-monolithic' into stefano_zampini/feature-fetidp-ex69

PCBDDC: fix bug with l2l

We cannot assume vins is a subvector of v. Using VecScatter is a more general solution

Merge branch 'tisaac/feature-fe-dual-apply' into tisaac/feature-plex-tests-new

There has unfortunately been a lot of interplay between our work to
change the petscspace/dualspace to accept vector co

Merge branch 'tisaac/feature-fe-dual-apply' into tisaac/feature-plex-tests-new

There has unfortunately been a lot of interplay between our work to
change the petscspace/dualspace to accept vector components, and the
work to transfer the tests in config/builder.py to the new test harness.
I'm merging them together to get them to play nice with each other, and
so that I only have one feature branch to handler when merging into
master/next.

* tisaac/feature-fe-dual-apply: (64 commits)
snes tutorials ex12: fix restart test arguments
nightly build fixes for arch-viennacl
fix run command to match the command in master/makefile. [the diff will be correct in master/next]
PetscFE: fix typo in for loop declaration
PetscPartitioner: use simple partitioning for regression tests
DT tests ex1: fix test arguments to match regression output
DT ex1: fix options
DMPlexCreateHexBoxMesh: no cells in z direction for 2D
PCBDDC: use %D for PetscInt printfs
Plex: helper function used in all cases now
Plex tests ex7: make partitions portable for testing
Plex: use MPIU_2INT instead of MPI_2INT
MatAIJ KLU: remove redefinition of SuiteSparse_long
DMPlexRefine_CTetgen() mimics DMPlexRefine_Tetgen()
DMPlexGenerate_Tetgen(): mimic ctetgen so it can be used equivalently
Plex: Cylinder fixes for complex
Plex ex1: Test for periodic cylinder
Plex: I think periodic cylinder is working
Plex ex1: Updated torus test
Plex: Torus can now be arbitrarily refined
...

show more ...

Merge remote-tracking branch 'origin/stefano_zampini/feature-sparserhs'

PCBDDC: fix memory leak

PCBDDC: don't factorize the Schur complement on dirichlet dofs

Mat: Added schur and schur_status to the private implementation

This permits to simplify a lot the code for handling Schur complements computations
from factorization packages.
Added factorschur.c t

Mat: Added schur and schur_status to the private implementation

This permits to simplify a lot the code for handling Schur complements computations
from factorization packages.
Added factorschur.c to store intermediate routines
Removed MatFactorSetSchurComplementSolver that was horrible

show more ...

Mat: added MatFactorSchurStatus to enable the user to progress factorization/inversion of the Schur complement

PCBDDC: use plex information (if available) to compute disconnected subdomains

If we have access to the mesh, we can also identify cross-vertices

PCFieldSplit: added customization for the scaling factor of the Schur complement

PCBDDC: get fields from the DM when available

KSPFETIDP/PCBDDC: allow using a single matrix for deluxe when the change of basis context is of type QR