Hello!
Hum, I think it’s a linker problem … otool -L (forgot about this tool …)
returns that the linked MPI libraries are inside the /usr/local/lib/ folder,
while the libraries created by the macports' openmpi are located at
/opt/local/lib/openmpi-mp …
I’ve copied the summary file below. What’s really weird now is that the
/opt/local/lib/openmpi-mp is given as part of the library path:
libmesh_optional_LIBS............ : [ … ] -L/opt/local/lib/openmpi-mp [ … ]
-Wl,-rpath,/opt/local/lib/openmpi-mp [ … ]
So, the system is choosing to prioritise the libraries found inside the
/usr/local/lib/ folder. How can I “convince” libMesh, during the configure
step, to use the MPI libraries found at /opt/local/lib/openmpi-mp ? I’m already
using the option "--with-mpi=/opt/local/libexec/openmpi-mp”, to indicate where
the proper compiler wrappers are.
Thanks in advance,
Thiago Milanetto Schlittler
—> Configure summary:
----------------------------------- SUMMARY -----------------------------------
Package version.................... : libmesh-1.0.0-pre
C++ compiler type.................. : clang
C++ compiler....................... : clang++
C compiler......................... : mpicc
Fortran compiler................... : mpif90
Build Methods...................... : dbg devel opt
CPPFLAGS...(dbg)................... : -DDEBUG
CXXFLAGS...(dbg)................... : -std=gnu++11 -O0 -felide-constructors -g
-pedantic -W -Wall -Wextra -Wno-long-long -Wunused -Wpointer-arith -Wformat
-Wparentheses -Qunused-arguments -std=gnu++11
CFLAGS.....(dbg)................... : -g -Wimplicit -Qunused-arguments
CPPFLAGS...(devel)................. :
CXXFLAGS...(devel)................. : -std=gnu++11 -O2 -felide-constructors -g
-pedantic -W -Wall -Wextra -Wno-long-long -Wunused -Wpointer-arith -Wformat
-Wparentheses -Wuninitialized -Qunused-arguments -std=gnu++11
CFLAGS.....(devel)................. : -O2 -Qunused-arguments -g -Wimplicit
CPPFLAGS...(opt)................... : -DNDEBUG
CXXFLAGS...(opt)................... : -std=gnu++11 -O2 -felide-constructors
-Qunused-arguments -std=gnu++11
CFLAGS.....(opt)................... : -O2 -Qunused-arguments
Install dir........................ : /Users/breubreubreu/Execs/libmesh/test
Build user......................... : breubreubreu
Build host......................... : Thiagos-Air.home
Build architecture................. : x86_64-apple-darwin14.5.0
Git revision....................... : 6fb1e5c4e2af4f8b79965e25d295311df544e543
Library Features:
library warnings................. : yes
adaptive mesh refinement......... : yes
blocked matrix/vector storage.... : no
complex variables................ : no
example suite.................... : yes
ghosted vectors.................. : yes
high-order shape functions....... : yes
unique-id support................ : no
id size (boundaries)............. : 2 bytes
id size (dofs)................... : 4 bytes
id size (processors)............. : 2 bytes
id size (subdomains)............. : 2 bytes
infinite elements................ : no
Dirichlet constraints............ : yes
node constraints................. : no
parallel mesh.................... : no
performance logging.............. : no
periodic boundary conditions..... : yes
reference counting............... : yes
shape function 2nd derivatives... : yes
stack trace files................ : no
track node valence............... : yes
variational smoother............. : yes
xdr binary I/O................... : yes
Optional Packages:
boost............................ : yes
capnproto........................ : no
cppunit.......................... : no
curl............................. : no
eigen............................ : yes
exodus........................... : yes
version....................... : v5.22
fparser.......................... : yes
build from version............ : release
glpk............................. : no
gmv.............................. : yes
gzstream......................... : yes
hdf5............................. : yes
laspack.......................... : no
libhilbert....................... : yes
metis............................ : yes
mpi.............................. : yes
nanoflann........................ : yes
nemesis.......................... : yes
version....................... : v5.22
netcdf........................... : yes
version....................... : 4
nlopt............................ : no
openmp........................... : no
parmetis......................... : yes
petsc............................ : yes
version....................... : 3.6.2
pthreads......................... : yes
qhull............................ : yes
sfcurves......................... : no
slepc............................ : yes
version....................... : 3.6.1
tbb.............................. : no
c++ threads...................... : no
c++ rtti ........................ : yes
tecio............................ : no
tecplot...(vendor binaries)...... : no
tetgen........................... : no
triangle......................... : no
trilinos......................... : no
vtk.............................. : yes
version....................... : 5.10.1
libmesh_optional_INCLUDES........ : -I/opt/local/include/vtk-5.10
-D_THREAD_SAFE -pthread -I/Users/breubreubreu/Execs/slepc-3.6.1/include
-I/Users/breubreubreu/Execs/slepc-3.6.1/arch-darwin-c/include
-I/Users/breubreubreu/Execs/petsc-3.6.2/include
-I/Users/breubreubreu/Execs/petsc-3.6.2/arch-darwin-c/include
-I/opt/local/include -I/opt/local/include/openmpi-mp
libmesh_optional_LIBS............ : -Wl,-rpath,/opt/local/lib/vtk-5.10
-L/opt/local/lib/vtk-5.10 -lvtkIO -lvtkCommon -lvtkFiltering -lvtkImaging -lz
-Wl,-rpath,/Users/breubreubreu/Execs/slepc-3.6.1/arch-darwin-c/lib
-L/Users/breubreubreu/Execs/slepc-3.6.1/arch-darwin-c/lib -lslepc
-L/Users/breubreubreu/Execs/petsc-3.6.2/arch-darwin-c/lib -L/opt/local/lib
-L/opt/local/lib/openmpi-mp
-L/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/lib/clang/7.0.0/lib/darwin
-L/opt/local/lib/gcc49/gcc/x86_64-apple-darwin14/4.9.3 -L/opt/local/lib/gcc49
-Wl,-rpath,/Users/breubreubreu/Execs/petsc-3.6.2/arch-darwin-c/lib -lpetsc
-llapack -lblas -Wl,-rpath,/opt/local/lib -lhwloc -lssl -lcrypto
-lmpi_usempif08 -lmpi_usempi_ignore_tkr -lmpi_mpifh
-Wl,-rpath,/opt/local/lib/gcc49/gcc/x86_64-apple-darwin14/4.9.3
-Wl,-rpath,/opt/local/lib/gcc49 -lgfortran -lgcc_ext.10.5 -lquadmath -lm
-lmpi_cxx -lc++ -Wl,-rpath,/opt/local/lib/openmpi-mp -lmpi -lSystem
-Wl,-rpath,/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/lib/clang/7.0.0/lib/darwin
-lclang_rt.osx -ldl
-------------------------------------------------------------------------------
> On 27 Oct 2015, at 14:16, Roy Stogner <[email protected]> wrote:
>
>
> On Tue, 27 Oct 2015, Thiago Milanetto Schlittler wrote:
>
>> I’m having some problems when running the libMesh examples. When I
>> try to do so using LIBMESH_RUN="mpirun -np 2”, the checks work
>> until the “introduction_ex1” check, where I get the following
>> following message:
>>
>> [Thiagos-Air.home:53055] [[49367,0],0] tcp_peer_recv_connect_ack: received
>> different version from [[49367,1],1]: 1.8.3 instead of 1.10.0
>> [Thiagos-Air.home:53055] [[49367,0],0] tcp_peer_recv_connect_ack: received
>> different version from [[49367,1],0]: 1.8.3 instead of 1.10.0
>>
>> and both processes stall. My computer has the 1.10.0 version of
>> OpenMPI, installed through macports (verified using ompi_info). I’ve
>> already purged and reinstalled my whole macports stack, so there are
>> no traces of a 1.8.3 OpenMPI version. I’ve also ran the configure
>> with the “with-mpi” option pointing towards the correct folder. I’m
>> using "libmesh-1.0.0-pre”, from the GitHub repository, but I got the
>> same problem with the stable release. Other libraries, such as PETSc
>> and SLEPc, run their tests with MPI with no problems. What could be
>> wrong? I’ve attached the configure summary.
>
> The summary didn't seem to survive our listserv. I'd normally ask for
> a copy via private email, but I'm not a Mac person so probably won't
> be much help here.
>
> The OSX equivalent of ldd is "otool -L"? You might try running that
> on your example-dbg and see whether there's still an old 1.8.3 library
> getting found somewhere by the linker.
>
> If it's not an old library, then it's either an old static object file
> (which you might be able to diagnose via "nm") or an old header (which
> I can't imagine how to diagnose - maybe use gdb and hope that
> whichever object file included the wrong header was built with
> debugging support too?).
> ---
> Roy
------------------------------------------------------------------------------
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users
