MPI configuration is somehow broken

[knepley]

I get this. What log do you need from me?

Error configuring PRAGMATIC with cmake Could not execute "['/PETSc3/petsc/petsc-pylith/arch-pylith-debug/bin/cmake .. -DCMAKE_INSTALL_PREFIX=/PETSc3/petsc/petsc-pylith/arch-pragmatic-debug -DCMAKE_VERBOSE_MAKEFILE=1 -DCMAKE_C_COMPILER="/PETSc3/petsc/bin/mpicc" -DCMAKE_AR=/usr/bin/ar -DCMAKE_RANLIB=/usr/bin/ranlib -DCMAKE_C_FLAGS:STRING="-Qunused-arguments -g3" -DCMAKE_C_FLAGS_DEBUG:STRING="-Qunused-arguments -g3" -DCMAKE_C_FLAGS_RELEASE:STRING="-Qunused-arguments -g3" -DCMAKE_CXX_COMPILER="/PETSc3/petsc/bin/mpicxx" -DCMAKE_CXX_FLAGS:STRING="-g" -DCMAKE_CXX_FLAGS_DEBUG:STRING="-g" -DCMAKE_CXX_FLAGS_RELEASE:STRING="-g" -DBUILD_SHARED_LIBS=on -DMETIS_DIR=/PETSc3/petsc/petsc-pylith/arch-pragmatic-debug -DENABLE_VTK=OFF -DENABLE_OPENMP=OFF -DEIGEN_INCLUDE_DIR=/PETSc3/petsc/petsc-pylith/arch-pragmatic-debug/include/eigen3 -DCMAKE_INSTALL_RPATH_USE_LINK_PATH:BOOL=ON']": -- The C compiler identification is AppleClang 6.0.0.6000057 -- The CXX compiler identification is AppleClang 6.0.0.6000057 -- Check for working C compiler: /PETSc3/petsc/bin/mpicc -- Check for working C compiler: /PETSc3/petsc/bin/mpicc -- works -- Detecting C compiler ABI info -- Detecting C compiler ABI info - done -- Detecting C compile features -- Detecting C compile features - done -- Check for working CXX compiler: /PETSc3/petsc/bin/mpicxx -- Check for working CXX compiler: /PETSc3/petsc/bin/mpicxx -- works -- Detecting CXX compiler ABI info -- Detecting CXX compiler ABI info - done -- Detecting CXX compile features -- Detecting CXX compile features - done -- LIBRT_LIBRARIES: -- Configured without VTK support. -- Performing Test COMPILER_SUPPORTS_CXX11 -- Performing Test COMPILER_SUPPORTS_CXX11 - Success -- Performing Test COMPILER_SUPPORTS_CXX0X -- Performing Test COMPILER_SUPPORTS_CXX0X - Success -- Found MPI_C: /PETSc3/petsc/bin/mpicc
-- Configuring incomplete, errors occurred! See also "/PETSc3/petsc/petsc-pylith/arch-pragmatic-debug/externalpackages/git.pragmatic/petsc-build/CMakeFiles/CMakeOutput.log".CMake Error at /PETSc3/petsc/petsc-pylith/arch-pylith-debug/share/cmake-3.9/Modules/FindPackageHandleStandardArgs.cmake:137 (message): Could NOT find MPI_CXX (missing: MPI_CXX_LIBRARIES MPI_CXX_INCLUDE_PATH) Call Stack (most recent call first): /PETSc3/petsc/petsc-pylith/arch-pylith-debug/share/cmake-3.9/Modules/FindPackageHandleStandardArgs.cmake:377 (_FPHSA_FAILURE_MESSAGE) /PETSc3/petsc/petsc-pylith/arch-pylith-debug/share/cmake-3.9/Modules/FindMPI.cmake:640 (find_package_handle_standard_args) CMakeLists.txt:65 (FIND_PACKAGE)

[taupalosaurus]

Damn. At least the error message is pretty clear, the line that fails is FIND_PACKAGE(MPI REQUIRED) from the main CMakeLists.txt

Now, if I understand what I just googled, find_package looks if it finds a FindMPI.cmake in the Cmake/Modules directory, and since there is no such file, tries to find it by itself.

So. Cmake can't find your MPI... is it in a very non standard place ?

[taupalosaurus]

We're currently using this: https://cmake.org/cmake/help/v3.0/module/FindMPI.html It should find MPI in standard locations.

I quote their doc if it fails:

If this fails, or if your MPI implementation does not come with a compiler wrapper, then set both MPI_LIBRARIES and MPI_INCLUDE_PATH. You may also set any other variables listed above, but these two are required. This will circumvent autodetection entirely.

I don't really know if your fancy build system lets you pass cmake command line options for external modules....

[taupalosaurus]

This line from your log tells us what is going wrong doesn't it ?

Could NOT find MPI_CXX (missing: MPI_CXX_LIBRARIES MPI_CXX_INCLUDE_PATH)

Where are these ?

[knepley]

The first line show what we executed the build with:

-DCMAKE_CXX_COMPILER="/PETSc3/petsc/bin/mpicxx"

Do we need to set another option?

[taupalosaurus]

Oh yes sorry. I can only guess from the scarce documentation/user forums I find online.

Error message suggests that you set -DMPI_CXX_LIBRARIES and -DMPI_CXX_INCLUDE_PATH Another forum suggests using -DMPI_CXX_COMPILER could be enough

If neither of these work, I would try inserting CMAKE in the variable names, and if neither work I would cry for help to @ggorman .

I'm also going to try to setup a clean virtual machine to experiment with Stephan's bug (issue #120), if you tell me exactly how you configure I can try to replicate and experiment myself

[taupalosaurus]

https://cmake.org/pipermail/cmake/2016-November/064634.html

Also suggests

-DMPI_C_COMPILER=/path/to/mpicc
-DMPI_CXX_COMPILER=/path/to/mpiCC
-DMPI_Fortran_COMPILER=/path/to/mpif90

is enough

[taupalosaurus]

@knepley was any of the above useful ? Can you pass specific configuration options for external modules to your Build System ?

[knepley]

On Sun, Apr 7, 2019 at 6:57 AM Nicolas Barral notifications@github.com wrote:

@knepley https://github.com/knepley was any of the above useful ? Can you pass specific configuration options for external modules to your Build System ?

Yes, that worked. I am now going through the C++11 stuff that is broken on my Mac :) My C++ vendetta is growing...

Matt

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/meshadaptation/pragmatic/issues/122#issuecomment-480579738, or mute the thread https://github.com/notifications/unsubscribe-auth/AAjoiXq-eyXwlnsSOthSmiOEOIf_lO52ks5vec8xgaJpZM4cbT7d .

-- 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/

[taupalosaurus]

In conclusion, was there anything wrong on our side regarding this issue ? Except the use of Cmake and C++ ?