petsc version 3.1.dfsg-10ubuntu1 failed to build on armel

[Steve Langasek]

petsc version 3.1.dfsg-10ubuntu1 failed to build on armel
Link to failed build: https://launchpad.net/ubuntu/+source/petsc/3.1.dfsg-10ubuntu1/+buildjob/2139778

Direct link to the build log: https://launchpad.net/ubuntu/+source/petsc/3.1.dfsg-10ubuntu1/+buildjob/2139778/+files/buildlog_ubuntu-natty-armel.petsc_3.1.dfsg-10ubuntu1_FAILEDTOBUILD.txt.gz

This log snippet might be of interest, since it triggered the matcher 'Purging chroot-autobuild'.
Excerpt 3041 lines into the build log:

TESTING: check from config.libraries(/build/buildd/petsc-3.1.dfsg/config/BuildSystem/config/libraries.py:133)
TESTING: checkInclude from config.headers(/build/buildd/petsc-3.1.dfsg/config/BuildSystem/config/headers.py:92)
TESTING: checkSharedLibrary from PETSc.packages.hdf5(/build/buildd/petsc-3.1.dfsg/config/BuildSystem/config/package.py:423)
TESTING: alternateConfigureLibrary from PETSc.packages.Dscpack(/build/buildd/petsc-3.1.dfsg/config/BuildSystem/config/package.py:427)
TESTING: configureLibrary from PETSc.packages.blacs(/build/buildd/petsc-3.1.dfsg/config/BuildSystem/config/package.py:370)
TESTING: check from config.libraries(/build/buildd/petsc-3.1.dfsg/config/BuildSystem/config/libraries.py:133)
*******************************************************************************
         UNABLE to CONFIGURE with GIVEN OPTIONS (see configure.log for details):
-------------------------------------------------------------------------------
--with-blacs-lib=['/usr/lib/libblacsCinit-lam.so', '/usr/lib/libblacs-lam.so'] and
--with-blacs-include=/usr/include did not work
*******************************************************************************

make: *** [build-arch] Error 1
dpkg-buildpackage: error: debian/rules build gave error exit status 2
******************************************************************************
Build finished at 20110323-1218
FAILED [dpkg-buildpackage died]
Purging chroot-autobuild/build/buildd/petsc-3.1.dfsg

[Dr. David Alan Gilbert (davidgil-uk)]

Recreated, form the configure.log:

==================================================================================
      Checking for a functional hypre
      Checking for library in User specified root directory HYPRE: ['/usr/lib/libHYPRE.a']
================================================================================
TEST check from config.libraries(/home/daveg/apt/petsc-3.1.dfsg/config/BuildSystem/config/libraries.py:133)
TESTING: check from config.libraries(config/BuildSystem/config/libraries.py:133)
  Checks that the library "libName" contains "funcs", and if it does defines HAVE_LIB"libName"
       - libDir may be a list of directories
       - libName may be a list of library names
          Checking for functions ['HYPRE_IJMatrixCreate'] in library ['/usr/lib/libHYPRE.a'] ['-llapack', '-lblas', '-L/usr/lib/openmpi/lib', '-L/usr/lib/arm-linux-gnueabi/gcc/arm-linux-gnueabi/4.5.2', '-L/usr/lib/arm-linux-gnueabi', '-ldl', '-lmpi', '-lopen-rte', '-lopen-pal', '-lnsl', '-lutil', '-lgcc_s', '-lpthread', '-lmpi_f90', '-lmpi_f77', '-lgfortran', '-lm', '-lm', '-lm', '-lm']
            Pushing language Cxx
sh: /usr/lib/lam/bin/mpicxx -c -o conftest.o -Wall -Wwrite-strings -Wno-strict-aliasing -g conftest.cc
Executing: /usr/lib/lam/bin/mpicxx -c -o conftest.o -Wall -Wwrite-strings -Wno-strict-aliasing -g conftest.cc
sh:
                    Pushing language CXX
                    Popping language CXX
sh: /usr/lib/lam/bin/mpicxx -o conftest -Wall -Wwrite-strings -Wno-strict-aliasing -g conftest.o -Wl,-rpath,/usr/lib -L/usr/lib -lHYPRE -llapack -lblas -L/usr/lib/openmpi/lib -L/usr/lib/arm-linux-gnueabi/gcc/arm-linux-gnueabi/4.5.2 -L/usr/lib/arm-linux-gnueabi -ldl -lmpi -lopen-rte -lopen-pal -lnsl -lutil -lgcc_s -lpthread -lmpi_f90 -lmpi_f77 -lgfortran -lm -lm -lm -lm -L/usr/lib/openmpi/lib -L/usr/lib/arm-linux-gnueabi/gcc/arm-linux-gnueabi/4.5.2 -L/usr/lib/arm-linux-gnueabi -L/usr/lib/arm-linux-gnueabi -ldl -lmpi -lopen-rte -lopen-pal -lnsl -lutil -lgcc_s -lpthread -ldl
Executing: /usr/lib/lam/bin/mpicxx -o conftest -Wall -Wwrite-strings -Wno-strict-aliasing -g conftest.o -Wl,-rpath,/usr/lib -L/usr/lib -lHYPRE -llapack -lblas -L/usr/lib/openmpi/lib -L/usr/lib/arm-linux-gnueabi/gcc/arm-linux-gnueabi/4.5.2 -L/usr/lib/arm-linux-gnueabi -ldl -lmpi -lopen-rte -lopen-pal -lnsl -lutil -lgcc_s -lpthread -lmpi_f90 -lmpi_f77 -lgfortran -lm -lm -lm -lm -L/usr/lib/openmpi/lib -L/usr/lib/arm-linux-gnueabi/gcc/arm-linux-gnueabi/4.5.2 -L/usr/lib/arm-linux-gnueabi -L/usr/lib/arm-linux-gnueabi -ldl -lmpi -lopen-rte -lopen-pal -lnsl -lutil -lgcc_s -lpthread -ldl
sh:
Possible ERROR while running linker: /usr/bin/ld: conftest.o: undefined reference to symbol 'HYPRE_IJMatrixCreate'
/usr/bin/ld: note: 'HYPRE_IJMatrixCreate' is defined in DSO /usr/lib/libHYPRE_IJ_mv-2.4.0.so so try adding it to the linker command line
/usr/lib/libHYPRE_IJ_mv-2.4.0.so: could not read symbols: Invalid operation
collect2: ld returned 1 exit status
 output: ret = 256
error message = {/usr/bin/ld: conftest.o: undefined reference to symbol 'HYPRE_IJMatrixCreate'
/usr/bin/ld: note: 'HYPRE_IJMatrixCreate' is defined in DSO /usr/lib/libHYPRE_IJ_mv-2.4.0.so so try addin...

[Dr. David Alan Gilbert (davidgil-uk)]

Hmm but today I have a different failure:

sh: /usr/lib/lam/bin/mpicxx -o conftest -Wall -Wwrite-strings -Wno-strict-aliasing -g conftest.o -L/usr/lib -L/usr/lib -lblacsCinit-lam -L/usr/lib -L/usr/lib -lblacs-lam -llapack -lblas -L/usr/lib/openmpi/lib -L/usr/lib/arm-linux-gnueabi/gcc/arm-linux-gnueabi/4.5.2 -L/usr/lib/arm-linux-gnueabi -ldl -lmpi -lopen-rte -lopen-pal -lnsl -lutil -lgcc_s -lpthread -lmpi_f90 -lmpi_f77 -lgfortran -lm -lm -lm -lm -L/usr/lib/openmpi/lib -L/usr/lib/arm-linux-gnueabi/gcc/arm-linux-gnueabi/4.5.2 -L/usr/lib/arm-linux-gnueabi -L/usr/lib/arm-linux-gnueabi -ldl -lmpi -lopen-rte -lopen-pal -lnsl -lutil -lgcc_s -lpthread -ldl
Executing: /usr/lib/lam/bin/mpicxx -o conftest -Wall -Wwrite-strings -Wno-strict-aliasing -g conftest.o -L/usr/lib -L/usr/lib -lblacsCinit-lam -L/usr/lib -L/usr/lib -lblacs-lam -llapack -lblas -L/usr/lib/openmpi/lib -L/usr/lib/arm-linux-gnueabi/gcc/arm-linux-gnueabi/4.5.2 -L/usr/lib/arm-linux-gnueabi -ldl -lmpi -lopen-rte -lopen-pal -lnsl -lutil -lgcc_s -lpthread -lmpi_f90 -lmpi_f77 -lgfortran -lm -lm -lm -lm -L/usr/lib/openmpi/lib -L/usr/lib/arm-linux-gnueabi/gcc/arm-linux-gnueabi/4.5.2 -L/usr/lib/arm-linux-gnueabi -L/usr/lib/arm-linux-gnueabi -ldl -lmpi -lopen-rte -lopen-pal -lnsl -lutil -lgcc_s -lpthread -ldl
sh:
Possible ERROR while running linker: /usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_unsigned_short'
/usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_double'
/usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_packed'
/usr/lib/libblacsCinit-lam.so: undefined reference to `lam_mpi_comm_world'
/usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_sum'
/usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_float'
/usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_int'
collect2: ld returned 1 exit status
 output: ret = 256
error message = {/usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_unsigned_short'
/usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_double'
/usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_packed'
/usr/lib/libblacsCinit-lam.so: undefined reference to `lam_mpi_comm_world'
/usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_sum'
/usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_float'
/usr/lib/libblacs-lam.so: undefined reference to `lam_mpi_int'
collect2: ld returned 1 exit status

[Dr. David Alan Gilbert (davidgil-uk)]

I suspect this isn't ARM specific exactly, but it's specific to those platforms (i.e. ARM) that use lam as the default mpi implementation rather than OpenMPI.

If I add a -llam to the configure test case that fails then the link succeeds.

Dave

[Dr. David Alan Gilbert (davidgil-uk)]

MPI makes my heads hurt;

OK - some other observations:

  * petsc has a dependency on libhdf5-mpi-dev
  * libhdf5-mpi-dev is a meta that is supposed to get the right libhdf5 for the platform mpi defaults but actually ends up depending on
       libhdf5-openmpi-dev, mpi-default-dev

      which is somewhere between inconsistent and wrong.

  2) hdf5 build dependencies are a bit odd:
Build-Depends: libmpich1.0-dev (>= 1.2.7-1), zlib1g-dev, lam4-dev (>= 7.1.1-3.2), quilt, libopenmpi-dev [!hp
pa !mips !mipsel !s390 !sh4 !m68k], libjpeg62-dev | libjpeg-dev, debhelper (>> 7), sed (>= 4.1.5), gfortran,
 libibverbs-dev [!hppa !mips !mipsel !s390 !kfreebsd-i386 !kfreebsd-amd64 !hurd-i386 !sh4], sharutils, mpi-d
efault-dev

   I think that needs to lose the libopenmpi-dev dependency completely (with all the predication) and just keep the mpi-default-dev one it already has?

Once you've got the build-deps for petsc installed - and because it has pulled in libopenmpi-dev - if you then try and rebuild libblacs it builds it against openmpi rather than lam.

Dave

[Matthias Klose (doko)]

now built in precise using the MPI libs

[Rolf Leggewie (r0lf)]

oneiric has seen the end of its life and is no longer receiving any updates. Marking the oneiric task for this ticket as "Won't Fix".