vtk: fix reset/reuse of 'dtype' variable - that caused memory leak

arch-linux-opt-cxx-quad src/ts/examples/tutorials runex26

> [ 0]3200 bytes PetscViewerVTKFWrite() line 265 in /sandbox/petsc/petsc

vtk: fix reset/reuse of 'dtype' variable - that caused memory leak

arch-linux-opt-cxx-quad src/ts/examples/tutorials runex26

> [ 0]3200 bytes PetscViewerVTKFWrite() line 265 in /sandbox/petsc/petsc.clone-3/src/sys/classes/viewer/impls/vtk/vtkv.c

show more ...

fix formatting of manual page

Commit-type: documentation
Reported-by: Jaroslaw Piwonski <jpi@informatik.uni-kiel.de>

Merge remote-tracking branch 'origin/jed/fix-stdout-attr-delete'

Merge remote-tracking branch 'origin/stefano_zampini/cherrypick-commits-for-maint-from-continuousadjoint' into maint

x: disable unused option warnings with -nox

PetscOptionsGetViewer: Support for option `draw:image:filename`

PetscOptionsGetViewer: Support for option `draw::format` (empty drawtype)

PetscViewerBinary: Allow using the same viewer object changing the filename

Remove use of PetscDataType from VTK code and other trivial location.

Reason: PetscDataType is duplicative to MPI_Datatype therefor we hope to limit its
use and perhaps eventually completely remove

Remove use of PetscDataType from VTK code and other trivial location.

Reason: PetscDataType is duplicative to MPI_Datatype therefor we hope to limit its
use and perhaps eventually completely remove it.

Commit-type: style-fix
Thanks-to: Jed Brown <jed@jedbrown.org>
Development Tools: Vim, Emacs, Eclipse

show more ...

PetscViewerDestroy_ASCII: clean up stdout and stderr attributes from MPI_Comm

Since -log_view is processed after PetscObjectRegisterDestroyAll() in
PetscFinalize(), it can attempt to access attribut

PetscViewerDestroy_ASCII: clean up stdout and stderr attributes from MPI_Comm

Since -log_view is processed after PetscObjectRegisterDestroyAll() in
PetscFinalize(), it can attempt to access attributes from
PetscViewerASCIIGetStdout(). When there are no references to the inner
communicator, it will be freed no later than
PetscObjectRegisterDestroyAll() and thus all attributes become unset.
But if there is any reference to the inner communicator, as in the
example below with the dummy PetscContainer (the type does not matter),
it remains after PetscObjectRegisterDestroyAll() and thus any stale
attributes remain accessible.

#include <petscviewer.h>

int main(int argc,char **args)
{
PetscErrorCode ierr;
PetscViewer view;
PetscContainer container;

ierr = PetscInitialize(&argc,&args,(char*)0,NULL);CHKERRQ(ierr);
ierr = PetscContainerCreate(PETSC_COMM_WORLD,&container);CHKERRQ(ierr);
ierr = PetscViewerASCIIGetStdout(PETSC_COMM_WORLD,&view);CHKERRQ(ierr);
ierr = PetscFinalize();
return ierr;
}

Reported-by: Stefano Zampini <stefano.zampini@gmail.com>

show more ...

fix: redundant CHKERRQ() call after SETERRQ()

Craycc complained about a stray CHKERRQ after SETERRQ in plextree.c
Reported-by: Patrick Sanan <patrick.sanan@gmail.com>

PetscRandom tests ex3: fix size calculation to PetscInt64

PetscRandomView(): use `%ul` for unsigned long

PetscRandom: change test ex3 to work on 32-bit systems

nightlybuild fixes

PetscRandom: added wrapper for Random123

PetscRandom: birthday spacing for failures of rand48, sprng

The birthday spacing test: if we divide up the unit hypercube into equal
size numbered bins, put random points in those bins, and count th

PetscRandom: birthday spacing for failures of rand48, sprng

The birthday spacing test: if we divide up the unit hypercube into equal
size numbered bins, put random points in those bins, and count the
number of times the same jump between successive bins occurs, then for a
large number of bins and points we expect that collision count to follow
a Poisson distribution. Poor quality PRNGs produce way too many
collisions to be random, which this test demonstrates

show more ...

Start adding a test to demonstrate weaknesses of LCGs

PETSCVIEWERGLVIS: fix improper use of PetscFunctionBeginUser

Free Petsc_Seq_Keyval if it has been created.

Commit-type: testing-fix,style-fix
Funded-by: ECP
Time: .1 hours
Reported-by: nightly tests

Development Tools: Vim, Emacs, Eclipse

Must free the MPI_Keyval's associated with viewers in PetscViewerFinalizePackage()

since PetscSysFinalizePackage() is not always called (if it never needed to be initialized)

Commit-type: bug-fix,

Must free the MPI_Keyval's associated with viewers in PetscViewerFinalizePackage()

since PetscSysFinalizePackage() is not always called (if it never needed to be initialized)

Commit-type: bug-fix, testing-fix,style-fix
Reported-by: nightly tests

show more ...

fix windows/c++ errors in d4c7638e66bfbab42594e49bfd8787d3c0f17499

libpetsc.lib(dlregispetsc.o) : error LNK2019: unresolved external symbol Petsc_Viewer_Draw_keyval referenced in function "int __cde

fix windows/c++ errors in d4c7638e66bfbab42594e49bfd8787d3c0f17499

libpetsc.lib(dlregispetsc.o) : error LNK2019: unresolved external symbol Petsc_Viewer_Draw_keyval referenced in function "int __cdecl PetscSysFinalizePackage(void)" (?PetscSysFinalizePackage@@YAHXZ)
libpetsc.lib(dlregispetsc.o) : error LNK2019: unresolved external symbol Petsc_Viewer_Binary_keyval referenced in function "int __cdecl PetscSysFinalizePackage(void)" (?PetscSysFinalizePackage@@YAHXZ)

show more ...

Free the MPI Keyval that may have been created during a run

Problem located by building MPICH with the option --enable-error-messages=all

Commit-type: style-fix
Funded-by: ECP
Time: .45 hours
Re

Free the MPI Keyval that may have been created during a run

Problem located by building MPICH with the option --enable-error-messages=all

Commit-type: style-fix
Funded-by: ECP
Time: .45 hours
Reported-by: Daniel Alejandro Ibanez <daibane@sandia.gov>

Development Tools: Vim, Emacs, Eclipse

show more ...

Merge remote-tracking branch 'origin/stefano_zampini/feature-glvis-viewer-fixdmdependency2'

fixup earlier (broken) fix