Nicholas Breen <nbr...@debian.org> writes: > It's probably a bug in OpenMPI instead (since the exact same tests pass on x32 > when using MPICH as the MPI implementation). I don't have an x32 system > available for testing, but it would be very interesting to know if the tests > pass with the 3.0.x version of OpenMPI out of experimental.
The only 3.0.x version available for x32 is 3.0.0-1, because newer
versions build-depend on pmix, which in turn build-depends on pandoc,
which is unavailable there. (I think the Haskell stack generally is.)
With that version installed in my laptop's x32 chroot and everything
else left at the versions available in unstable, I get *different*
errors per the attached log, excerpted below. (Using unstable across
the board yields the same errors as on the autobuilder.)
9/24 Test #9: GpuUtilsUnitTests ................***Timeout 30.01 sec
[==========] Running 30 tests from 6 test cases.
[----------] Global test environment set-up.
[----------] 7 tests from HostAllocatorTest/0, where TypeParam = int
[ RUN ] HostAllocatorTest/0.EmptyMemoryAlwaysWorks
[ OK ] HostAllocatorTest/0.EmptyMemoryAlwaysWorks (0 ms)
[ RUN ] HostAllocatorTest/0.VectorsWithDefaultHostAllocatorAlwaysWorks
[ OK ] HostAllocatorTest/0.VectorsWithDefaultHostAllocatorAlwaysWorks
(0 ms)
[ RUN ] HostAllocatorTest/0.TransfersWithoutPinningWork
[ OK ] HostAllocatorTest/0.TransfersWithoutPinningWork (0 ms)
[ RUN ] HostAllocatorTest/0.FillInputAlsoWorksAfterCallingReserve
[ OK ] HostAllocatorTest/0.FillInputAlsoWorksAfterCallingReserve (0 ms)
[ RUN ] HostAllocatorTest/0.ChangingPinningPolicyRequiresCuda
[WARNING]
/home/amu/src/gromacs/gromacs-git/src/external/gmock-1.7.0/gtest/src/gtest-death-test.cc:825::
Death tests use fork(), which is unsafe particularly in a threaded context.
For this test, Google Test couldn't detect the number of threads.
[...]
13/24 Test #13: MdrunUtilityMpiUnitTests .........***Failed 0.02 sec
--------------------------------------------------------------------------
There are not enough slots available in the system to satisfy the 4 slots
that were requested by the application:
/home/amu/src/gromacs/gromacs-git/build/openmpi/bin/mdrunutility-mpi-test
Either request fewer slots for your application, or make more slots available
for use.
--------------------------------------------------------------------------
[...]
The following tests FAILED:
9 - GpuUtilsUnitTests (Timeout)
13 - MdrunUtilityMpiUnitTests (Failed)
Errors while running CTest
I can retest on a more powerful system if you think that's likely to
make a difference; I just didn't yet have an x32 chroot there at all,
and would need to reboot it to enable the necessary kernel parameter
(syscall.x32=y).
I hope that information helps!
--
Aaron M. Ucko, KB1CJC (amu at alum.mit.edu, ucko at debian.org)
http://www.mit.edu/~amu/ | http://stuff.mit.edu/cgi/finger/?a...@monk.mit.edu
gromacs_2018-1_x64.build.xz
Description: Binary data
