July 2013 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] moab nightlybuild failure
by Satish Balay 05 Jul '13

05 Jul '13

On Mon, 1 Jul 2013, Satish Balay wrote: > On Fri, 28 Jun 2013, Satish Balay wrote: > > > On Thu, 27 Jun 2013, Jed Brown wrote: > > > > > Satish Balay <balay(a)mcs.anl.gov> writes: > > > > > > > Ok - fixed in 'balay/fix-moab-git'. Any comments? > > > > > > > > https://bitbucket.org/petsc/petsc/commits/1f16e75d68ca70d003ea3ef4dde0f67dc… > > > > > > Looks okay. > > > > merged to master now [so that nightly builds can start working] > > > Error running autoreconf on MOAB: Could not execute "cd /usr/home/balay/petsc.clone-2/externalpackages/moab && /usr/local/bin/autoreconf -fi": > Can't exec "aclocal": No such file or directory at /usr/local/share/autoconf-2.69/Autom4te/FileUtils.pm line 326. > autoreconf-2.69: failed to run aclocal: No such file or directory > > > Looks like the existance of 'autoreconf' doesn't mean its functional :( > > [we need to add in a run test for autoreconf ..] Pushed a fix to 'balay/fix-moab-git' https://bitbucket.org/petsc/petsc/commits/e87d3dc4639f8572ae22d2fec3466958e… Satish

1 0

Re: [petsc-dev] Pushing to 'master'
by Jed Brown 04 Jul '13

04 Jul '13

Hong Zhang <hzhang(a)mcs.anl.gov> writes: > Jed, > Double checking, here is my configure options used on petsc machine: > > --download-cmake --download-hdf5 --download-metis --download-mpich=1 > --download-mumps --download-parmetis --download-scalapack > --download-sundials --download-superlu --download-superlu_dist > --with-cc=gcc --with-clanguage=cxx --with-cxx=g++ --with-debugging=1 > --with-fc=gfortran PETSC_ARCH=arch-linux-gcc-gfortran The --with-clanguage=cxx is why you didn't see an error. All plain C builds fail with a compilation error on this code.

1 0

Re: [petsc-dev] Pushing to 'master'
by Hong Zhang 04 Jul '13

04 Jul '13

Jed, Double checking, here is my configure options used on petsc machine: --download-cmake --download-hdf5 --download-metis --download-mpich=1 --download-mumps --download-parmetis --download-scalapack --download-sundials --download-superlu --download-superlu_dist --with-cc=gcc --with-clanguage=cxx --with-cxx=g++ --with-debugging=1 --with-fc=gfortran PETSC_ARCH=arch-linux-gcc-gfortran It tolerates missing {declaring variables}. Hong On Thu, Jul 4, 2013 at 11:14 AM, Hong Zhang <hzhang(a)mcs.anl.gov> wrote: > Jed : > >> > Jed, >> > Sorry, I checked examples_next_arch-xxx.log, not make_next_arch-xx.log >> :-( >> >> Hmm, everything fails here too because there is no libpetsc. >> >> >> http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/examples_next_arch-linux-gcc-i… >> >> > It built smoothly on petsc machine, thus I merged to master. >> >> C++ build? >> > Yes, I did not use this configure flag. I've added it to my configure now. > Hong > >

1 0

Re: [petsc-dev] Pushing to 'master'
by Hong Zhang 04 Jul '13

04 Jul '13

Jed : > > Jed, > > Sorry, I checked examples_next_arch-xxx.log, not make_next_arch-xx.log > :-( > > Hmm, everything fails here too because there is no libpetsc. > > > http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/examples_next_arch-linux-gcc-i… > > > It built smoothly on petsc machine, thus I merged to master. > > C++ build? > Yes, I did not use this configure flag. I've added it to my configure now. Hong

1 0

Re: [petsc-dev] Pushing to 'master'
by Jed Brown 03 Jul '13

03 Jul '13

Hong Zhang <hzhang(a)mcs.anl.gov> writes: > Jed, > Sorry, I checked examples_next_arch-xxx.log, not make_next_arch-xx.log :-( Hmm, everything fails here too because there is no libpetsc. http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/examples_next_arch-linux-gcc-i… > It built smoothly on petsc machine, thus I merged to master. C++ build?

1 0

Re: [petsc-dev] Pushing to 'master'
by Hong Zhang 03 Jul '13

03 Jul '13

Jed, Sorry, I checked examples_next_arch-xxx.log, not make_next_arch-xx.log :-( It built smoothly on petsc machine, thus I merged to master. I'm getting older, thinking about retiring ... Hong On Wed, Jul 3, 2013 at 5:48 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > Hong, you are doing well with the formal workflow, but in this last > round with 'hzhang/mat-cleanup', it looks like you didn't check the > nightly logs before merging (they all failed). > > > http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/make_next_arch-linux-gcc-ifc_c… > > Matt fixed the problem below, unbreaking 'master', but this should never > have been merged when (a) the nightlies failed, (b) the branch cannot be > compiled on any machine, and (c) the result of the merge cannot be > compiled on any machine. > > Remember that when integrating (e.g., to 'next', 'master', etc), you > should _always_ build the result of the merge before pushing. > > > commit 275476c694b7affb79a129990509c16cc25cadec > Author: Matthew G. Knepley <knepley(a)gmail.com> > Date: Wed Jul 3 17:23:45 2013 -0500 > > Mat: Need a scope when declaring variables in a switch block > > diff --git a/src/mat/impls/aij/mpi/mpimatmatmult.c > b/src/mat/impls/aij/mpi/mpimatmatmult.c > index bf6f620..f924000 100644 > --- a/src/mat/impls/aij/mpi/mpimatmatmult.c > +++ b/src/mat/impls/aij/mpi/mpimatmatmult.c > @@ -823,6 +823,7 @@ PetscErrorCode MatTransposeMatMult_MPIAIJ_MPIAIJ(Mat > P,Mat A,MatReuse scall,Pets > ierr = > MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable(P,A,fill,C);CHKERRQ(ierr); > break; > case 2: > + { > Mat Pt; > Mat_PtAPMPI *ptap; > Mat_MPIAIJ *c; > @@ -833,6 +834,7 @@ PetscErrorCode MatTransposeMatMult_MPIAIJ_MPIAIJ(Mat > P,Mat A,MatReuse scall,Pets > ptap->Pt = Pt; > (*C)->ops->mattransposemultnumeric = > MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_matmatmult; > PetscFunctionReturn(0); > + } > break; > default: > ierr = > MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ(P,A,fill,C);CHKERRQ(ierr); >

1 0

Pushing to 'master'
by Jed Brown 03 Jul '13

03 Jul '13

Hong, you are doing well with the formal workflow, but in this last round with 'hzhang/mat-cleanup', it looks like you didn't check the nightly logs before merging (they all failed). http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/make_next_arch-linux-gcc-ifc_c… Matt fixed the problem below, unbreaking 'master', but this should never have been merged when (a) the nightlies failed, (b) the branch cannot be compiled on any machine, and (c) the result of the merge cannot be compiled on any machine. Remember that when integrating (e.g., to 'next', 'master', etc), you should _always_ build the result of the merge before pushing. commit 275476c694b7affb79a129990509c16cc25cadec Author: Matthew G. Knepley <knepley(a)gmail.com> Date: Wed Jul 3 17:23:45 2013 -0500 Mat: Need a scope when declaring variables in a switch block diff --git a/src/mat/impls/aij/mpi/mpimatmatmult.c b/src/mat/impls/aij/mpi/mpimatmatmult.c index bf6f620..f924000 100644 --- a/src/mat/impls/aij/mpi/mpimatmatmult.c +++ b/src/mat/impls/aij/mpi/mpimatmatmult.c @@ -823,6 +823,7 @@ PetscErrorCode MatTransposeMatMult_MPIAIJ_MPIAIJ(Mat P,Mat A,MatReuse scall,Pets ierr = MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable(P,A,fill,C);CHKERRQ(ierr); break; case 2: + { Mat Pt; Mat_PtAPMPI *ptap; Mat_MPIAIJ *c; @@ -833,6 +834,7 @@ PetscErrorCode MatTransposeMatMult_MPIAIJ_MPIAIJ(Mat P,Mat A,MatReuse scall,Pets ptap->Pt = Pt; (*C)->ops->mattransposemultnumeric = MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_matmatmult; PetscFunctionReturn(0); + } break; default: ierr = MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ(P,A,fill,C);CHKERRQ(ierr);

1 0

Re: [petsc-dev] windows fixes.
by Satish Balay 03 Jul '13

03 Jul '13

On Wed, 3 Jul 2013, Jed Brown wrote: > Satish Balay <balay(a)mcs.anl.gov> writes: > > > $ ssh fusion.lcrc.anl.gov which python2 > > /usr/bin/python2 > > Was that added later? A vanilla install of Python-2.4.6 does contains > only bin/python and bin/python2.4. I remember digging through this two > years ago and finding that many systems did not have a python2. Yeah - its possible that some distributions are adding the python2 linkage - and not python vanilla install. Satish

1 0

Re: [petsc-dev] windows fixes.
by Jed Brown 03 Jul '13

03 Jul '13

Satish Balay <balay(a)mcs.anl.gov> writes: > $ ssh fusion.lcrc.anl.gov which python2 > /usr/bin/python2 Was that added later? A vanilla install of Python-2.4.6 does contains only bin/python and bin/python2.4. I remember digging through this two years ago and finding that many systems did not have a python2.

1 0

Re: [petsc-dev] windows fixes.
by Satish Balay 03 Jul '13

03 Jul '13

On Wed, 3 Jul 2013, Jed Brown wrote: > Satish Balay <balay(a)mcs.anl.gov> writes: > > > I expect all these distros to have python2 available aswell. [and most > > will be installed - due to other depencencies] > > No, the base install will not have python2. It _might_ be required due > to dependencies, but all the major packages work with python3 now. > > > And we might have to change '#!/usr/bin/env python' to '#!/usr/bin/env python2' > > in our scripts. > > This breaks old systems that don't have a binary called 'python2' > (including RHEL5, I think). $ ssh fusion.lcrc.anl.gov which python2 /usr/bin/python2 Satish > > >> One day we will switch from python 2* to python 3*, just not yet. > > > > I was hoping to use RHEL as a baseline for our minimum python version > > > > RHEL5 with python2.4 has EOL date March 2017 > > > > RHEL6 has python-2.7 [EOL November 2020] > > > > Presumably we can have code that works with both python3 & python2.7 when rhel5 is dead? > > Yes, it's reasonable to maintain code that works for python2.6 through > python3. >

1 0