GitLab

VecCUDA: fix a typo

See merge request !3864

DMStag: fix incorrect usage of PETSC_COMM_WORLD in errors

See merge request !3844

Build some "classic" docs with Sphinx docs

See merge request !3684

Turn back on filtering of linker output instead of just dropping all output...

See merge request !3855

Add linking filter for multiple libgfortran

Commit-type: error-checking, portability-fix
/spend 25m

Do not set higher bits in error codes since error codes should never be larger than 255

See merge request !3845

Add proper error return codes for cuSolver error checker. CHKERRCUSOLVER()

See merge request !3827

Fix some mispellings etc in install_documentation.rst and add dtails on ALCF GPU systems

See merge request !3810

fix build warning with !3851 - for some reason TaoGetGradientNorm is in include/petsc/private/taoimpl.h - so do not enable fortran interface

/home/balay/petsc/src/tao/interface/ftn-auto/taosolverf.c: In function ‘taogradientnorm_’:
/home/balay/petsc/src/tao/interface/ftn-auto/taosolverf.c:521:11: warning: implicit declaration of function ‘TaoGradientNorm’; did you mean ‘TaoGetGradientNorm’? [-Wimplicit-function-declaration]
  521 | *__ierr = TaoGradientNorm(
      |           ^~~~~~~~~~~~~~~
      |           TaoGetGradientNorm

Documentation fixes

See merge request !3851

Minor fixes

See merge request !3847

Turn back on filtering of linker output instead of just dropping all output (which may be error messages)

This is in response to !3703 (comment 548743488)



There is the possibility this is too fragile for all the compilers and compiler versions of user but it is
the systematic way to handle linker output that may trigger error conditions but is really only a warning.

Commit-type: error-checking, configure, CI
Reported-by: Satish Balay <balay@mcs.anl.gov>
/spend 2h

These higher bits were apparently causing the confusing error codes produced in the CI when GPU jobs
ran out of resources.

Do not trap SIGTERM with OpenMPI since OpenMPI uses it to shut down MPI jobs and this
produces identical confusing messages to the user

/spend 6h

PetscStackCallHDF5: success if return code nonnegative (not only 0)

See merge request !3842

Plex: Fortran wrapper for DMPlexExtrude

See merge request !3843

Fix 'pip install' for systems that do not have 'python' in PATH

See merge request !3841

Fix compiler warning

See merge request !3840

balay@pj02:~/petsc$ make test gmakesearch='ts_tutorials-ex11_adv_2d_quad_%' FFLAGS='-Werror -std=f2008' CXXFLAGS=-Werror CFLAGS='-Werror'
/usr/bin/make  --no-print-directory -f gmakefile.test PETSC_ARCH=arch-linux-c-opt PETSC_DIR=/home/balay/petsc test
Using MAKEFLAGS: --no-print-directory -- PETSC_DIR=/home/balay/petsc PETSC_ARCH=arch-linux-c-opt CFLAGS=-Werror CXXFLAGS=-Werror FFLAGS=-Werror -std=f2008 gmakesearch=ts_tutorials-ex11_adv_2d_quad_%
          CC arch-linux-c-opt/tests/ts/tutorials/ex11.o
/home/balay/petsc/src/ts/tutorials/ex11.c: In function ‘PhysicsRiemann_Euler_Godunov’:
/home/balay/petsc/src/ts/tutorials/ex11.c:2397:5: error: array subscript is below array bounds [-Werror=array-bounds]
     --nn;
     ^~~~
cc1: all warnings being treated as errors

Fixed by removing the f2c generated shift in nn[] that made the compiler think nn[] might be accessed out of range

Commit-type: portability-fix
/spend 10m
Reported-by: Satish Balay <balay@mcs.anl.gov>

fixed bug in ex1.c in tao constrained tutorials

See merge request !3830

Stop the configure process if the user provides incorrect compiler flags...

See merge request !3818

Minor improvements

See merge request !3829

Stop the configure process if the user provides incorrect compiler flags instead of silently ignoring them

Commit-type: error-checking, feature, configure
/spend 45m
Reported-by: Milan Pelletier <milan.pelletier@protonmail.com>

Configure: Fix ExodusII for the case of no Fortran

See merge request !3825

Declare MPIU_MAXINDEX_OP etc as PETSC_EXTERN

See merge request !3831

cygwin: disable --download-mpich - it does not work anymore. Also add similar...

See merge request !3823

fix 'make check' for kokkos-kernels

See merge request !3820

Co-Authored-by: Junchao Zhang <jczhang@mcs.anl.gov>

boost: fix build failure on macOS by updating to newer version - as old

See merge request !3832

bugfix for MatRestoreRow_xxx()

See merge request !3826