[petsc-dev] Possible error running C/C++ src/snes/examples/tutorials/ex19 with 1 MPI process

[Satish Balay]

On Sat, 8 Oct 2016, Antonio Trande wrote:

> On 10/08/2016 11:30 PM, Satish Balay wrote:
> >>>>
> >>>
> >>> Previously you had a build where these errors didn't come up
> >>>
> >>> https://paste.fedoraproject.org/445643/
> >>
> >> I seen that 'csh' tool was not found.
> > 
> > Not sure what you mean. I don't see any 'csh' in either of the logs.
> > 
> 
> Same build that you see here https://paste.fedoraproject.org/445643/,
> using valgrind showed something like
> 
> ... 'bin/chs' not found

Sorry - somehow I don't see the mention of csh in the logs..

> 
> I have included 'csh' as needed package for testing and obtained latest
> build log:
> https://copr-be.cloud.fedoraproject.org/results/sagitter/petsc/fedora-rawhide-x86_64/00462512-petsc/build.log.gz
> 
> >>
> >>>
> >>> But now the build gives errors.
> >>>
> >>>>>>>>
> >>> Running test examples to verify correct installation
> >>> Using PETSC_DIR=/builddir/build/BUILD/petsc-3.7.3/petsc-3.7.3 and PETSC_ARCH=x86_64
> >>> Possible error running C/C++ src/snes/examples/tutorials/ex19 with 1 MPI process
> >>> See http://www.mcs.anl.gov/petsc/documentation/faq.html
> >>> [0]PETSC ERROR: #1 PetscLogInitialize() line 219 in /builddir/build/BUILD/petsc-3.7.3/petsc-3.7.3/src/sys/logging/plog.c
> >>> [0]PETSC ERROR: #2 PetscInitialize() line 891 in /builddir/build/BUILD/petsc-3.7.3/petsc-3.7.3/src/sys/objects/pinit.c
> >>> Possible error running C/C++ src/snes/examples/tutorials/ex19 with 2 MPI processes
> >>> <<<<<<<
> >>>
> >>> Looks like we discussed this issue before:
> >>>
> >>> http://lists.mcs.anl.gov/pipermail/petsc-users/2016-June/029634.html
> >>>
> >>> The key is to make sure the correct petsc.so and petscconf.h are used
> >>> in the test [with the way this build is being done, and stuff getting
> >>> copied arround - the makefiles that do 'make test' might have invalid
> >>> info..]
> >>>
> >>> Perhaps you can rerun this via valgraind again to see what it says.
> >>>
> >>>
> >>> Satish
> >>>
> >>
> >> I dont see any line related to openblas now.
> > 
> > In the valgrind output? What lines do you see?
> 
> None line about openblas now.
> 
> > 
> > Its crashing in PetscInitialize() - i.e its crashing at the very begining - and didn't get to the blas calls.
> > 
> > [and this somehow worked in  https://paste.fedoraproject.org/445643/ - so I don't really understand
> > whats going on...]
> > 
> 
> Do tests need internet by chance?

Nope - but you might want to disable petscnagupgrade.py in lib/petsc/conf/rules

I was previously looking at wrong sources..

> >>> [0]PETSC ERROR: #1 PetscLogInitialize() line 219 in /builddir/build/BUILD/petsc-3.7.3/petsc-3.7.3/src/sys/logging/plog.c

#if defined(PETSC_HAVE_PAPI)
  ierr = PAPI_library_init(PAPI_VER_CURRENT);
  if (ierr != PAPI_VER_CURRENT) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"Cannot initialize PAPI");
  ierr = PAPI_query_event(PAPI_FP_INS);CHKERRQ(ierr);
  ierr = PAPI_create_eventset(&PAPIEventSet);CHKERRQ(ierr);
  ierr = PAPI_add_event(PAPIEventSet,PAPI_FP_INS);CHKERRQ(ierr);
  ierr = PAPI_start(PAPIEventSet);CHKERRQ(ierr);
#endif

I don't know if this stuff is supporsed to work. Its best if you disable PAPI in the build.

Satish