July 2020 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] gcc 10
by Satish Balay 13 Jul '20

13 Jul '20

The c++ thingy is on a centos-7 box - i.e with gcc-4.8.5 as the system compiler. However the build was with [a manually built] clang-6.0 - which for some reason is using the includes from this system gcc. Satish On Mon, 13 Jul 2020, Jed Brown wrote: > I don't know what failure you're referring to. The C++ complex thing might have been a matter of the antique compiler. > > Matthew Knepley <knepley(a)gmail.com> writes: > > > Why did Mark have the gcc 10 failure on our example. Was it something with > > did in it? > > > > Matt > > > > On Mon, Jul 13, 2020 at 12:43 PM Jed Brown <jed(a)jedbrown.org> wrote: > > > >> You want this for PETSc or for external packages? We add it for a bunch > >> of external packages. I've been building PETSc with gcc/gfortran-10 for > >> months now. > >> > >> Matthew Knepley <knepley(a)gmail.com> writes: > >> > >> > Do we have gfortran from gcc 10 running anywhere? We need a check to > >> enable > >> > > >> > -fallow-argument-mismatch > >> > > >> > from configure. Should be as simple as compiling Fortran which has > >> > PetscObjectSetName() in it. > >> > > >> > Thanks, > >> > > >> > Matt > >> > > >> > -- > >> > What most experimenters take for granted before they begin their > >> > experiments is infinitely more interesting than any results to which > >> their > >> > experiments lead. > >> > -- Norbert Wiener > >> > > >> > https://www.cse.buffalo.edu/~knepley/ < > >> http://www.cse.buffalo.edu/~knepley/> > >> > > > > > > -- > > What most experimenters take for granted before they begin their > > experiments is infinitely more interesting than any results to which their > > experiments lead. > > -- Norbert Wiener > > > > https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/> >

1 0

Re: [petsc-dev] gcc 10
by Barry Smith 13 Jul '20

13 Jul '20

The C++ complex thing and the gfortran thing are totally unrelated except they both come from Mark and involve the Gnu community. > On Jul 13, 2020, at 12:40 PM, Jed Brown <jed(a)jedbrown.org> wrote: > > I don't know what failure you're referring to. The C++ complex thing might have been a matter of the antique compiler. > > Matthew Knepley <knepley(a)gmail.com> writes: > >> Why did Mark have the gcc 10 failure on our example. Was it something with >> did in it? >> >> Matt >> >> On Mon, Jul 13, 2020 at 12:43 PM Jed Brown <jed(a)jedbrown.org> wrote: >> >>> You want this for PETSc or for external packages? We add it for a bunch >>> of external packages. I've been building PETSc with gcc/gfortran-10 for >>> months now. >>> >>> Matthew Knepley <knepley(a)gmail.com> writes: >>> >>>> Do we have gfortran from gcc 10 running anywhere? We need a check to >>> enable >>>> >>>> -fallow-argument-mismatch >>>> >>>> from configure. Should be as simple as compiling Fortran which has >>>> PetscObjectSetName() in it. >>>> >>>> Thanks, >>>> >>>> Matt >>>> >>>> -- >>>> What most experimenters take for granted before they begin their >>>> experiments is infinitely more interesting than any results to which >>> their >>>> experiments lead. >>>> -- Norbert Wiener >>>> >>>> https://www.cse.buffalo.edu/~knepley/ < >>> http://www.cse.buffalo.edu/~knepley/> >>> >> >> >> -- >> What most experimenters take for granted before they begin their >> experiments is infinitely more interesting than any results to which their >> experiments lead. >> -- Norbert Wiener >> >> https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>

1 0

Re: [petsc-dev] gcc 10
by Barry Smith 13 Jul '20

13 Jul '20

I told everyone in that thread how it should be fixed. The fix for this is to add PetscObjectSetName() to src/sys/f90-mod/petscsys.h90 with the * notation to allow any type. You will see in that file PetscObjectReference() etc. PetsFE is an INTEGER(8) in Fortran because no one has updated it yet to be a Fortran type. You will see in include/petsc/finclude/petscdm.h that only DM has be properly updated to be a Fortran type, a bunch of others are still treated as integers and need to be fixed eventually. #define DM type(tDM) #define DMAdaptor PetscFortranAddr #define PetscQuadrature PetscFortranAddr #define PetscDS PetscFortranAddr #define PetscFE PetscFortranAddr #define PetscSpace PetscFortranAddr #define PetscDualSpace PetscFortranAddr #define PetscFV PetscFortranAddr #define PetscLimiter PetscFortranAddr #define PetscPartitioner PetscFortranAddr #define DMField PetscFortranAddr We do not need to configure check this, we need to fix the PETSc source so it follows F90 (yes, F90 standards, Gnu is still catching up :-). Barry A separate issue is external packages do not follow Fortran standards so generate such errors as well, but that is not our business, we just always force turn off the Fortran error checking for external packages. > On Jul 13, 2020, at 12:34 PM, Matthew Knepley <knepley(a)gmail.com> wrote: > > Why did Mark have the gcc 10 failure on our example. Was it something with did in it? > > Matt > > On Mon, Jul 13, 2020 at 12:43 PM Jed Brown <jed(a)jedbrown.org <mailto:[email protected]>> wrote: > You want this for PETSc or for external packages? We add it for a bunch of external packages. I've been building PETSc with gcc/gfortran-10 for months now. > > Matthew Knepley <knepley(a)gmail.com <mailto:[email protected]>> writes: > > > Do we have gfortran from gcc 10 running anywhere? We need a check to enable > > > > -fallow-argument-mismatch > > > > from configure. Should be as simple as compiling Fortran which has > > PetscObjectSetName() in it. > > > > Thanks, > > > > Matt > > > > -- > > What most experimenters take for granted before they begin their > > experiments is infinitely more interesting than any results to which their > > experiments lead. > > -- Norbert Wiener > > > > https://www.cse.buffalo.edu/~knepley/ <https://www.cse.buffalo.edu/~knepley/> <http://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>> > > > -- > What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. > -- Norbert Wiener > > https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>

1 0

Re: [petsc-dev] gcc 10
by Jed Brown 13 Jul '20

13 Jul '20

I don't know what failure you're referring to. The C++ complex thing might have been a matter of the antique compiler. Matthew Knepley <knepley(a)gmail.com> writes: > Why did Mark have the gcc 10 failure on our example. Was it something with > did in it? > > Matt > > On Mon, Jul 13, 2020 at 12:43 PM Jed Brown <jed(a)jedbrown.org> wrote: > >> You want this for PETSc or for external packages? We add it for a bunch >> of external packages. I've been building PETSc with gcc/gfortran-10 for >> months now. >> >> Matthew Knepley <knepley(a)gmail.com> writes: >> >> > Do we have gfortran from gcc 10 running anywhere? We need a check to >> enable >> > >> > -fallow-argument-mismatch >> > >> > from configure. Should be as simple as compiling Fortran which has >> > PetscObjectSetName() in it. >> > >> > Thanks, >> > >> > Matt >> > >> > -- >> > What most experimenters take for granted before they begin their >> > experiments is infinitely more interesting than any results to which >> their >> > experiments lead. >> > -- Norbert Wiener >> > >> > https://www.cse.buffalo.edu/~knepley/ < >> http://www.cse.buffalo.edu/~knepley/> >> > > > -- > What most experimenters take for granted before they begin their > experiments is infinitely more interesting than any results to which their > experiments lead. > -- Norbert Wiener > > https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>

1 0

Re: [petsc-dev] gcc 10
by Matthew Knepley 13 Jul '20

13 Jul '20

Why did Mark have the gcc 10 failure on our example. Was it something with did in it? Matt On Mon, Jul 13, 2020 at 12:43 PM Jed Brown <jed(a)jedbrown.org> wrote: > You want this for PETSc or for external packages? We add it for a bunch > of external packages. I've been building PETSc with gcc/gfortran-10 for > months now. > > Matthew Knepley <knepley(a)gmail.com> writes: > > > Do we have gfortran from gcc 10 running anywhere? We need a check to > enable > > > > -fallow-argument-mismatch > > > > from configure. Should be as simple as compiling Fortran which has > > PetscObjectSetName() in it. > > > > Thanks, > > > > Matt > > > > -- > > What most experimenters take for granted before they begin their > > experiments is infinitely more interesting than any results to which > their > > experiments lead. > > -- Norbert Wiener > > > > https://www.cse.buffalo.edu/~knepley/ < > http://www.cse.buffalo.edu/~knepley/> > -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>

1 0

Re: [petsc-dev] gcc 10
by Satish Balay 13 Jul '20

13 Jul '20

Here is the (ARM) build with gnu-10 compilers https://gitlab.com/petsc/petsc/-/jobs/636087963 Compilers: C Compiler: mpicc -Wall -Wwrite-strings -Wno-strict-aliasing -Wno-unknown-pragmas -fstack-protector -fvisibility=hidden -g -O Version: gcc (GCC) 10.1.1 20200507 (Red Hat 10.1.1-1) C++ Compiler: mpicxx -Wall -Wwrite-strings -Wno-strict-aliasing -Wno-unknown-pragmas -fstack-protector -fvisibility=hidden -g -O Version: g++ (GCC) 10.1.1 20200507 (Red Hat 10.1.1-1) Fortran Compiler: mpif90 -Wall -ffree-line-length-0 -Wno-unused-dummy-argument -g -O Version: GNU Fortran (GCC) 10.1.1 20200507 (Red Hat 10.1.1-1) Satish On Mon, 13 Jul 2020, Satish Balay via petsc-dev wrote: > Yeah - we fixed PETSc [maint] to not require -fallow-argument-mismatch - and use it only with externalpacakges that need it > > And we do have one test box with gfortran-10 - but the test coverage with it is minimal. [mostly no externalpacakges] > > Satish > > On Mon, 13 Jul 2020, Jed Brown wrote: > > > You want this for PETSc or for external packages? We add it for a bunch of external packages. I've been building PETSc with gcc/gfortran-10 for months now. > > > > Matthew Knepley <knepley(a)gmail.com> writes: > > > > > Do we have gfortran from gcc 10 running anywhere? We need a check to enable > > > > > > -fallow-argument-mismatch > > > > > > from configure. Should be as simple as compiling Fortran which has > > > PetscObjectSetName() in it. > > > > > > Thanks, > > > > > > Matt > > > > > > -- > > > What most experimenters take for granted before they begin their > > > experiments is infinitely more interesting than any results to which their > > > experiments lead. > > > -- Norbert Wiener > > > > > > https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/> > > >

1 0

Re: [petsc-dev] gcc 10
by Satish Balay 13 Jul '20

13 Jul '20

Yeah - we fixed PETSc [maint] to not require -fallow-argument-mismatch - and use it only with externalpacakges that need it And we do have one test box with gfortran-10 - but the test coverage with it is minimal. [mostly no externalpacakges] Satish On Mon, 13 Jul 2020, Jed Brown wrote: > You want this for PETSc or for external packages? We add it for a bunch of external packages. I've been building PETSc with gcc/gfortran-10 for months now. > > Matthew Knepley <knepley(a)gmail.com> writes: > > > Do we have gfortran from gcc 10 running anywhere? We need a check to enable > > > > -fallow-argument-mismatch > > > > from configure. Should be as simple as compiling Fortran which has > > PetscObjectSetName() in it. > > > > Thanks, > > > > Matt > > > > -- > > What most experimenters take for granted before they begin their > > experiments is infinitely more interesting than any results to which their > > experiments lead. > > -- Norbert Wiener > > > > https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/> >

1 0

Re: [petsc-dev] gcc 10
by Jed Brown 13 Jul '20

13 Jul '20

You want this for PETSc or for external packages? We add it for a bunch of external packages. I've been building PETSc with gcc/gfortran-10 for months now. Matthew Knepley <knepley(a)gmail.com> writes: > Do we have gfortran from gcc 10 running anywhere? We need a check to enable > > -fallow-argument-mismatch > > from configure. Should be as simple as compiling Fortran which has > PetscObjectSetName() in it. > > Thanks, > > Matt > > -- > What most experimenters take for granted before they begin their > experiments is infinitely more interesting than any results to which their > experiments lead. > -- Norbert Wiener > > https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>

1 0

gcc 10
by Matthew Knepley 13 Jul '20

13 Jul '20

Do we have gfortran from gcc 10 running anywhere? We need a check to enable -fallow-argument-mismatch from configure. Should be as simple as compiling Fortran which has PetscObjectSetName() in it. Thanks, Matt -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>

1 0

Re: [petsc-dev] C++ error with just PETSc includes
by Lisandro Dalcin 13 Jul '20

13 Jul '20

We should really try to find a workaround for this issue. The problem only triggers with a quite old GCC version. I cannot avoid feeling guilty about it. On Mon, 13 Jul 2020 at 18:24, Satish Balay via petsc-dev < petsc-dev(a)mcs.anl.gov> wrote: > Checking src/mat/tests/ex231.cxx you need: > > diff --git a/src/dm/impls/plex/landau/kokkos/tutorials/ex1.cxx > b/src/dm/impls/plex/landau/kokkos/tutorials/ex1.cxx > index d04eaf96e7..967efc3625 100644 > --- a/src/dm/impls/plex/landau/kokkos/tutorials/ex1.cxx > +++ b/src/dm/impls/plex/landau/kokkos/tutorials/ex1.cxx > @@ -1,5 +1,7 @@ > static char help[] = "Simple Kokkos test\n\n"; > > +#define PETSC_SKIP_CXX_COMPLEX_FIX > #include <petscdmplex.h> > > > > Also #requiresdefine is not meant for example makefile. You can use: > > build: requires: in /*TEST section > > Satish > > > On Mon, 13 Jul 2020, Mark Adams wrote: > > > I have a wip C++ test that just calls PetscInitialize and Finalize, and I > > am getting some sort of error in the include file: > > > > https://gitlab.com/petsc/petsc/-/jobs/635768809 > > > > All I do is: > > > > #include <petscdmplex.h> > > > > Any ideas? > > Thanks, > > Mark > > > > -- Lisandro Dalcin ============ Research Scientist Extreme Computing Research Center (ECRC) King Abdullah University of Science and Technology (KAUST) http://ecrc.kaust.edu.sa/

1 0