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 ***@***.***> 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 <#122 (comment)>, 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++ ?