Merge remote-tracking branch 'origin/release'

Merge branch 'PR--petsc4py-build-against-current' into 'release'

petsc4py: build against the PETSc being built

See merge request petsc/petsc!3748

Merge remote-tracking branch 'origin/release'

Delay trying empty PIC flag until end of shared library checks unless PIC flag...

petsc4py: build against the PETSc being built

Fixes Issue 862:

./configure --with-petsc4py=1 && make && make install DESTDIR=foo
<error>

Prior to this patch, when the Python extension is b

petsc4py: build against the PETSc being built

Fixes Issue 862:

./configure --with-petsc4py=1 && make && make install DESTDIR=foo
<error>

Prior to this patch, when the Python extension is built against
the PETSc installed in the system, i.e. PETSc needs to be fully
installed before the extension can be built. This is a problem
for packaging recipes that want to use `--with-petsc4py=1` but
won't have PETSc installed into the system at the time the
one single top-level build gets to building the Python extension.

This patch allows petsc4py to be built against PETSc located
in the staging directory (DESTDIR). This involves four actions:

1. set PETSC_DIR the DESTDIR/PREFIX

2. but still pass DESTDIR to petsc4py build code (for items 3, 4)

3. input paths read from PETSc config files point into the final
installation location of PETSc, so need to be prepended with DESTDIR

4. output paths to be used after installation need to be stripped
of DESTDIR (they contain DESTDIR because of item 1)

NOTE: DESTDIR passed to the top-level makefile must be an absolute path.
A relative path would not resolve intuitively at the time of building
petsc4py, since the working directory would be `src/binding/petsc4py`.

Signed-off-by: Alexei Colin <acolin@isi.edu>

show more ...

Run mpi* --show after building NPI so configure has a record of what is being run

Commit-type: configure, enhancement, debugability
/spend 15m

remove nonworking PETSCVIEWERADIOS2 impl

slepc: update to latest snapshot

Merge branch 'jczhang/fix-fortran-object-reinit' into 'main'

Fortran: re-init destroyed petsc objects so that they are different from null objects

See merge request petsc/petsc!3786

BuildSystem: update exodusII to v2021-01-20
- Fixes compilation issues on macOS 11.2 / gcc10
- Adds libexoIIv2for32.a to self.liblist [when using fortran]
- Installs exodiff and exotxt

Fortran: re-init destroyed petsc objects so that they are different from null objects

Reported-by: Jose E. Roman <jroman@dsic.upv.es>

Merge branch 'barry/2021-03-28/openmpi-libevent-internal/release' into 'main'

OpenMPI: build with --with-libevent=internal - needed for MacOS

See merge request petsc/petsc!3779

Merge remote-tracking branch 'origin/release'

Configure: Protect GMsh execution

OpenMPI: build with --with-libevent=internal

This fixes OpenMPI failures on Mac with latest OS and Xcode

Commit-type: portability-fix
/spend 45m

Add test for mpi4py and clearer error message on how to PYTHONPATH to use mpi4py

Commit-type: error-checking,testing-fix,feature,example
/spend 2h

Config: add OpenMP flag to CXXPPFLAGS

Merge remote-tracking branch 'origin/release'

Stop passing integers to function that that may take pointers depending on the MPI implementation.

Commit-type: portability-fix, configure
/spend 5m

Merge branch 'PR--config-petsc4py-with-arg-bool' into 'main'

config: petsc4py: fix type of --with arg to bool

See merge request petsc/petsc!3741

Merge branch 'balay/revert-mpi.mod-check' into 'main'

mpif.mod: revert requirement - as MS-MPI does not have it.

See merge request petsc/petsc!3759

Merge branch 'knepley/feature-plex-ds-traversals' into 'main'

DS: Moved management of pointwise functions into PetscWeakForm object - Added...

See merge request petsc/petsc!3726

Merge branch 'jolivet/feature-before-release-fixes' into 'main'

fixes and updates before new release

See merge request petsc/petsc!3712

mpif.mod: revert requirement - as MS-MPI does not have it.
Do not add back the complex code for corner cases

Sys+Configure: Added dladdr() support
- Turn on feature test macro for dladdr() support
- PetscDLAddr() should not fail if there is no support