On Mon, 28 Jul 2014 18:22:47 -0600 "Chang, Christopher" <[email protected]> wrote:
> Sorry Jan, re-posting log file to list. > > I was not aware of that log file (CMake ignorance on my part), > thanks. Looking at it, I can see what the issue was, but I didn't > know it was related to PETSc. > > Now when the build stops, I have two files in CMakeFiles, > petsc_test_lib.cpp and petsc_config_test_version.cpp . I can build a > test binary from petsc_test_lib.cpp by hand-it requires adding > -mt_mpi to the compiler command. I have that flag defined in > CMakeCache.txt under CMAKE_C_FLAGS, but it seems like it might not be > propagating down to the build for testing PETSc (if it were, I > shouldn't be getting the error in CMakeError.log). Is there a log > containing the failed compilation commands? I don't see anything in > CMakeFiles/ . Linker command is also in CMakeError.log, just before error message. You're right, -mt_mpi is not there. Shouldn't it be added to CMAKE_EXE_LINKER_FLAGS or MPI_C_LINK_FLAGS rather than to CMAKE_C_FLAGS? Jan > > Thanks, > > Chris > > -----Original Message----- > From: Jan Blechta [mailto:[email protected]] > Sent: Monday, July 28, 2014 3:36 PM > To: Chang, Christopher > Cc: [email protected] > Subject: Re: [FEniCS-support] Installation--PETSc test failing > > [Please, post CMakeError.log again to > [email protected] and keep it in CC] > > There is problem in linking > --------------------------------------------------------------- > ld: MPIR_Thread: TLS definition > in /nopt/nrel/apps/impi/4.1.1.036/intel64/lib/libmpi_dbg_mt.so > section .tbss mismatches non-TLS definition > in /nopt/nrel/apps/impi/4.1.1.036/intel64/lib/libmpi.so > section .bss /nopt/nrel/apps/impi/4.1.1.036/intel64/lib/libmpi_dbg_mt.so: > could not read symbols: Bad value > --------------------------------------------------------------- which > guides us to some discussions, e.g. > https://software.intel.com/en-us/forums/topic/392483. You could try > disabling threads and/or OpenMP in PETSc. > > Jan > > > On Mon, 28 Jul 2014 14:57:15 -0600 > "Chang, Christopher" <[email protected]> wrote: > > > Jan, > > > > Thanks! Attached. > > > > -----Original Message----- > > From: Jan Blechta [mailto:[email protected]] > > Sent: Monday, July 28, 2014 2:20 PM > > To: Chang, Christopher > > Cc: [email protected] > > Subject: Re: [FEniCS-support] Installation--PETSc test failing > > > > Ok. Attach, please, CMakeError.log. > > > > Jan > > > > > > On Mon, 28 Jul 2014 13:47:11 -0600 > > "Chang, Christopher" <[email protected]> wrote: > > > > > Hi Jan, > > > > > > Thanks. I realized I left out some other useful information in > > > addition to the version. I should have said, I'm trying to build > > > Dolfin. UFL, FFC, Fiat, and instant went in OK. I am building > > > Dolfin from the Git repo, > > > https://bitbucket.org/fenics-project/dolfin.git . The last > > > timestamp in the log was on July 22, 2014. Unfortunately, > > > petsc4py wouldn't build against PETSc < 3.5. I guess I could keep > > > multiple PETSc installation handy, but I was hoping to avoid it. > > > > > > Thanks, > > > > > > Chris > > > > > > -----Original Message----- > > > From: Jan Blechta [mailto:[email protected]] > > > Sent: Monday, July 28, 2014 1:39 PM > > > To: Chang, Christopher > > > Cc: [email protected] > > > Subject: Re: [FEniCS-support] Installation--PETSc test failing > > > > > > Which version of FEniCS are trying to build? Only current FEniCS > > > dev works with PETSc 3.5. Pick PETSc 3.4 for building FEniCS 1.4 > > > release. > > > > > > Jan > > > > > > > > > On Mon, 28 Jul 2014 13:31:32 -0600 > > > "Chang, Christopher" <[email protected]> wrote: > > > > > > > Hi, > > > > > > > > I am trying to install FEniCS from source, building with the > > > > Intel tools. I have succeeded in building and installing PETSc > > > > 3.5.1 and petsc4py, and have pointed CMake to the relevant > > > > locations. However, during the "cmake ." stage of the build, I > > > > get > > > > > > > > -- Checking for package 'PETSc' > > > > -- PETSC_DIR is /home/cchang/petsc-3.5.1 > > > > -- PETSC_ARCH is impi-intel > > > > -- Found petscconf.h > > > > -- Performing test PETSC_TEST_RUNS - Failed > > > > -- Performing test PETSC_TEST_3RD_PARTY_LIBS_RUNS - Failed > > > > -- PETSc configured without Cusp support > > > > -- PETSc could not be found. Be sure to set PETSC_DIR and > > > > PETSC_ARCH. (missing: PETSC_TEST_RUNS) > > > > > > > > I can't troubleshoot further, because I don't know where the > > > > output of whatever tests are being done is. Can anyone tell > > > > what is going wrong with the build from output above? > > > > > > > > Thanks, > > > > > > > > Chris > > > > > > > > -- > > > > Christopher H. Chang, Ph.D. > > > > Computational Scientist IV > > > > National Renewable Energy Laboratory > > > > 15013 Denver West Parkway, MS ESIF301 Golden, CO 80401 Phone > > > > (303) 275-3751 Fax (303) 275-4007 > > > > > > > > > > _______________________________________________ > > > fenics-support mailing list > > > [email protected] > > > http://fenicsproject.org/mailman/listinfo/fenics-support > > > _______________________________________________ fenics-support mailing list [email protected] http://fenicsproject.org/mailman/listinfo/fenics-support
