Merge branch 'balay/fix_FC_DEPFLAGS' into 'master'

configure: fix FC_DEPFLAGS

See merge request petsc/petsc!2105

Fix configuration crash when cmake was not found/working as 'e' was not defined.

configure: fix FC_DEPFLAGS

Reported-by: Lisandro Dalcin <dalcinl@gmail.com>

mpiexec: fix shell escape of path-to-mpiexec only when using autodected-path.
Also spectrum MPI uses OMPI_MAJOR_VERSION etc - so check if mpiexec supports --oversubscribe - before using it.

Add chec

mpiexec: fix shell escape of path-to-mpiexec only when using autodected-path.
Also spectrum MPI uses OMPI_MAJOR_VERSION etc - so check if mpiexec supports --oversubscribe - before using it.

Add check for 'mpiexec -n 1' usage (required)

debugging usage: --with-mpiexec="jsrun -g 1"

Reported-by: Mark Adams <mfadams@lbl.gov>

show more ...

Merge branch 'jed/sorted-petscconf-defines' into 'master'

config: sort defines going into petscconf.h [Fix #383]

Closes #383

See merge request petsc/petsc!2099

config MPI.py: stricter detection of Open MPI environment

Reported-by: Jacob Faibussowitsch <jfaibussowitsch@anl.gov>

config: sort defines going into petscconf.h [Fix #383]

Merge branch 'barry/2019-08-05/features-cray-knl' into 'master'

Features for Cray KNL systems

See merge request petsc/petsc!2044

Features for Cray KNL systems

Detect Cray compilers and use them automatically, no longer need to use --with-cc=cc etc.
more installation.html information on using theta
Detect MKLROOT and use it if

Features for Cray KNL systems

Detect Cray compilers and use them automatically, no longer need to use --with-cc=cc etc.
more installation.html information on using theta
Detect MKLROOT and use it if no specific BLAS/LAPACK requested
Error if Cray environment detected but no Cray compiler modules set, prevents compiling with old sequential gnu compilers on the system
For Intel and GNU Cray environments on KNL systems allow compiling packages whose configures call executables by stripping the
KNL specific arguments from the compile libraries for the external libraries, thus --download-hdf5 etc work (note this does
not work for the Cray compilers because I could not figure out how to run the drive.cc programs that the cc compiler
lists when generating the complete compile/link list of compiler arguments, includes, and libraries)

Commit-type: ease-of-use, feature

Minor updates based on comments of the reviewers of the MR

Commit-type: style-fix
Reported-by: Satish Balay <balay@mcs.anl.gov>

Add self.installwithbatch = 0 for hypre since it attempts to run programs during configure
Note that with the PETSc automatic cross compiling on Theta hypre will configure and build successfully

Fix directory for checking default BLAS/LAPACK locations before checking MKLROOT variable

Fixed error in BlasLapack.py which resulted from bad merge and the variables were set too late, changed used of ls to Python function

Commit-type: bug-fix, style-fix
Reported-by: gitlab-ci

show more ...

Merge branch 'knepley/feature-dm-boundingbox' into 'master'

Knepley/feature dm boundingbox

See merge request petsc/petsc!2043

Merge branch 'barry/2019-07-27/error-check-mpiexec' into 'master'

Large number of fixes, optimizations for configure, speeds up the configure

See merge request petsc/petsc!2046

Merge branch 'master' into 'master'

Add --download-hpddm

See merge request petsc/petsc!1994

Put back test for 'dgeev' in BLAS/LAPACK, put back detection of Fortran calls for accessing command line arguments, added comments why shared library detector did not work

Commit-type: portability-f

Put back test for 'dgeev' in BLAS/LAPACK, put back detection of Fortran calls for accessing command line arguments, added comments why shared library detector did not work

Commit-type: portability-fix, style-fix, feature
Reported-by: Satish Balay <balay@mcs.anl.gov>
Reported-by: Matthew Knepley <knepley@gmail.com>

show more ...

--download-hpddm for PCHPDDM and KSPHPDDM

Fixes and improvements for PCHPDDM and KSPHPDDM

petsc4py: fix --prefix install [conflict with package-prefix-hash]

Reported-by: Pierre Jolivet <pierre.jolivet@enseeiht.fr>

Large number of fixes, optimizations for configure, speeds up the configure

1) turn off checking if MPI library is shared since that code has been broken for 3+ years

2) move pure Fortran configure

Large number of fixes, optimizations for configure, speeds up the configure

1) turn off checking if MPI library is shared since that code has been broken for 3+ years

2) move pure Fortran configure tests to compilersFortran.py

3) turned threads back on but only use them for performing timeout for *runtest() commands (since they slow everything down if used
everywhere)

4) changed parallel configure (which doesn't work anyways) to use --useParallel and not overload --useThreads

5) added runtime check of mpiexec that uses timeout (replaces previous test in MPI shared libraries)

6) removed unused tests in configure in the interest of speed

7) minor refactorization of configure for better support of parallel configure (which doesn't work)

8) added timing results in configure.log. These may be accessed with grep child configure.log | sort -k3 -g

9) add check that mpiexec given in configure test matches the MPI compilers and libraries used

Commit-type: optimization, bug-fix, portability-fix, feature
Reported-by: Carl Steefel <cisteefel@lbl.gov>

show more ...

mpi4py: fix for package-prefix-hash

petsc4py: update to snapshot with corresponding changes

pkg-hash: fix for --download-petsc4py

petsc4py: update to snapshot with VECCuda fixes for !2074

Merge branch 'barry/2019-09-10/jed/gitlab-ci' into 'master'

Barry/2019 09 10/jed/gitlab ci

See merge request petsc/petsc!2042

netcdf: add backup download url

Update to new commit of OpenBLAS that truly does fix the TARGET=GENERIC problem

Commit-type: package-update

Fix for Python3, string input must be converted to [] before processing because list and map behave differently.

Commit-type: bug-fix, portability-fix
Reported-by: gitlab-ci