GCC works great on Linux, but IMHO it is always problematic on Mac. I had a lot 
of trouble when using GCC on Mac for netcdf, libmesh and some other stuff. And 
it seems not easy to fix the problems on the library side. For example, the 
netcdf problem I got has existed for several years, and there is no solution 
yet.

Hong (Mr.)

On Aug 2, 2017, at 11:48 AM, Richard Tran Mills 
<[email protected]<mailto:[email protected]>> wrote:

Things work fine when I use clang, but I'd really like to be able to use GCC as 
well.

I updated the ignored issue to reflect that the bug is still present when using 
GCC 7.1.0.

I suppose I will give OpenMPI a try when I get a chance.

--Richard

On Tue, Aug 1, 2017 at 5:15 PM, Matthew Knepley 
<[email protected]<mailto:[email protected]>> wrote:
On Tue, Aug 1, 2017 at 7:03 PM, Barry Smith 
<[email protected]<mailto:[email protected]>> wrote:

   Reported to MPICH folks https://github.com/pmodels/mpich/issues/2437  and 
ignored.

If you guys cannot walk down the hall and get them to consider these bugs, we 
should just stop using MPICH and recommending
it to unsuspecting people. This is development malpractice.

   Matt

   What about using Gnu compilers and OpenMPI (allegedly OpenMPI has fixed up 
many of their ongoing bugs).

    Barry

> On Aug 1, 2017, at 6:58 PM, Dave May 
> <[email protected]<mailto:[email protected]>> wrote:
>
>
>
> On 2 August 2017 at 00:39, Richard Tran Mills 
> <[email protected]<mailto:[email protected]>> wrote:
> Hi Folks,
>
> I am seeing a lot of errors when I run the full test suite (with latest 
> master) on my MacBook, related to SEGVs inside (or thereabouts) 
> PetscCommBuildTwoSided_Ibarrier. Do I have something mis-configured or ... ?
>
> Nope - I have exactly the same issue since moving to El Capitan. I posted a 
> message to maint some time earlier this year / late last year about 
> MatZeroRows being broken.
>
> The only way I know around this issue is to simply "avoid the fancy shit" 
> (Barry) via the option
> -build_twosided allreduce
>
>
>
> Sample error message below:
>
> > [2]PETSC ERROR: 
> > ------------------------------------------------------------------------
> > [2]PETSC ERROR: Caught signal number 11 SEGV: Segmentation Violation, 
> > probably memory access out of range
> > [2]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger
> > [2]PETSC ERROR: or see 
> > http://www.mcs.anl.gov/petsc/documentation/faq.html#valgrind
> > [2]PETSC ERROR: or try http://valgrind.org<http://valgrind.org/> on 
> > GNU/linux and Apple Mac OS X to find memory corruption errors
> > [2]PETSC ERROR: likely location of problem given in stack below
> > [2]PETSC ERROR: ---------------------  Stack Frames ---------------------> 
> > [3]PETSC ERROR: Note: The EXACT line numbers in the stack are not available,
> > [3]PETSC ERROR:       INSTEAD the line number of the start of the function
> > [3]PETSC ERROR:       is given.
> > [3]PETSC ERROR: [3] PetscCommBuildTwoSided_Ibarrier line 97 
> > /Users/rmills/proj/petsc/src/sys/utils/mpits.c
> > [3]PETSC ERROR: [3] PetscCommBuildTwoSided line 279 
> > /Users/rmills/proj/petsc/src/sys/utils/mpits.c
> > [3]PETSC ERROR: [3] PetscSFSetUp_Basic line 539 
> > /Users/rmills/proj/petsc/src/vec/is/sf/impls/basic/sfbasic.c
> > [3]PETSC ERROR: [3] PetscSFSetUp line 179 
> > /Users/rmills/proj/petsc/src/vec/is/sf/interface/sf.c
> > [3]PETSC ERROR: [3] PetscSFBcastBegin line 925 
> > /Users/rmills/proj/petsc/src/vec/is/sf/interface/sf.c
> > [3]PETSC ERROR: [3] MatISSetPreallocation_IS line 1018 
> > /Users/rmills/proj/petsc/src/mat/impls/is/matis.c
> > [3]PETSC ERROR: [3] MatISSetPreallocation line 1005 
> > /Users/rmills/proj/petsc/src/mat/impls/is/matis.c
>
> And I'm using MPICH downloaded/built by PETSc:
>
> rmills@Richards-MBP:mat(master=)arch-osx-gcc7-debug$ 
> /Users/rmills/proj/petsc/arch-osx-gcc7-debug/bin/mpirun --version
> HYDRA build details:
>     Version:                                 3.2
>     Release Date:                            Wed Nov 11 22:06:48 CST 2015
>     CC:                              gcc-7  -fstack-protector -g
>     CXX:                             g++-7  -fstack-protector -g
>     F77:                             gfortran -ffree-line-length-0 -g
>     F90:                             gfortran -ffree-line-length-0 -g
>     Configure options:                       '--disable-option-checking' 
> '--prefix=/Users/rmills/proj/petsc/arch-osx-gcc7-debug' 'MAKE=/usr/bin/make' 
> '--libdir=/Users/rmills/proj/petsc/arch-osx-gcc7-debug/lib' 'CC=gcc-7' 
> 'CFLAGS=-fstack-protector -g -O0' 'AR=/usr/bin/ar' 'ARFLAGS=cr' 'CXX=g++-7' 
> 'CXXFLAGS=-fstack-protector -g ' 'F77=gfortran' 'FFLAGS=-ffree-line-length-0 
> -g ' 'FC=gfortran' 'FCFLAGS=-ffree-line-length-0 -g ' '--enable-shared' 
> '--with-device=ch3:sock' '--with-pm=hydra' '--enable-fast=no' 
> '--enable-g=meminit' '--cache-file=/dev/null' '--srcdir=.' 'LDFLAGS=' 
> 'LIBS=-lpthread ' 'CPPFLAGS= 
> -I/Users/rmills/proj/petsc/arch-osx-gcc7-debug/externalpackages/mpich-3.2/src/mpl/include
>  
> -I/Users/rmills/proj/petsc/arch-osx-gcc7-debug/externalpackages/mpich-3.2/src/mpl/include
>  
> -I/Users/rmills/proj/petsc/arch-osx-gcc7-debug/externalpackages/mpich-3.2/src/openpa/src
>  
> -I/Users/rmills/proj/petsc/arch-osx-gcc7-debug/externalpackages/mpich-3.2/src/openpa/src
>  -D_REENTRANT 
> -I/Users/rmills/proj/petsc/arch-osx-gcc7-debug/externalpackages/mpich-3.2/src/mpi/romio/include'
>     Process Manager:                         pmi
>     Launchers available:                     ssh rsh fork slurm ll lsf sge 
> manual persist
>     Topology libraries available:            hwloc
>     Resource management kernels available:   user slurm ll lsf sge pbs cobalt
>     Checkpointing libraries available:
>     Demux engines available:                 poll select
>
> --Richard
>




--
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

http://www.caam.rice.edu/~mk51/

Reply via email to