On Wed, Oct 30, 2013 at 2:14 PM, Anders Logg <[email protected]> wrote:
> On Wed, Oct 30, 2013 at 11:25:06AM +0100, Johannes Ring wrote:
>> On Tue, Oct 29, 2013 at 9:49 PM, Anders Logg <[email protected]> wrote:
>> > On Tue, Oct 29, 2013 at 10:12:35AM +0000, Garth N. Wells wrote:
>> >> On 2013-10-29 10:03, Anders Logg wrote:
>> >> >On Tue, Oct 29, 2013 at 10:30:03AM +0100, Johannes Ring wrote:
>> >> >>On Tue, Oct 29, 2013 at 10:06 AM, Martin Sandve Alnęs
>> >> >><[email protected]> wrote:
>> >> >>> Some of the buildbots have been offline for a while.
>> >> >>
>> >> >>I have restarted one (wheezy-amd64) and disabled the others now
>> >> >>(osx-10.6 and sid-amd64). The osx-10.6 buildbot machine has been
>> >> >>upgraded to OS X 10.8, but is currently not ready to be used as a
>> >> >>buildbot slave.
>> >> >
>> >> >Great. I think we should only enable buildbots that are known to work.
>> >> >
>> >> >>> Some of them have strange timeouts.
>> >> >>
>> >> >>sid-amd64 hang when compiling the fem convergence benchmark. I thought
>> >> >>this had been fixed when gcc 4.8.2 was uploaded to Debian unstable,
>> >> >>but it was not. Can someone else reproduce this timeout when compiling
>> >> >>this benchmark with gcc 4.8? We might want to disable building of this
>> >> >>benchmark.
>> >
>> > Where is the sid-amd64 buildbot? Is it offline? I don't see it in the list.
>>
>> I removed it since it was just hanging, but I will add it back when we
>> disable the fem convergence benchmark.
>
> ok.
>
>> >> >It hangs for me with gcc 4.8.1. Does anyone know when a new gcc will
>> >> >enter Ubuntu?
>> >> >
>> >>
>> >> I wouldn't hold your breath - Ubuntu has a poor track record in
>> >> releasing bug fixes for development tools.
>> >>
>> >> >The benchmark builds fine if I disable P5 for R3. I suggest we disable
>> >> >that benchmark for now and reenable it later. We should open an issue
>> >> >so we don't forget it. (I can do this later today.)
>> >> >
>> >> >>> And at least two display this error:
>> >> >>>
>> >> >>> dolfin-master-full-precise-amd64
>> >> >>>
>> >> >>> [  0%] Building CXX object
>> >> >>> test/unit/la/cpp/CMakeFiles/test_LinearOperator.dir/LinearOperator.cpp.o
>> >> >>> Linking CXX executable test_LinearOperator
>> >> >>> /home/buildbot/fenicsbbot/master/dolfin-full/lib/libdolfin.so: 
>> >> >>> undefined
>> >> >>> reference to `METIS_Free'
>> >> >>
>> >> >>That error was first encountered in this build:
>> >> >>
>> >> >>http://fenicsproject.org:8010/builders/dolfin-master-full-precise-amd64/builds/417
>> >> >
>> >> >Looks like Garth?
>> >> >
>> >>
>> >> The correlation between code changes and buildbot errors has been
>> >> very weak of late. The error is a linking problem which is probably
>> >> due to a library configuration problem on the buildbot.
>> >
>> > Can we get it fixed?
>>
>> This was fixed after I removed the libparmetis-dev Debian package from
>> the buildbot. It is strange though that it suddenly (in build 417)
>> started to pick up libparmetis.so from /usr/lib instead of the locally
>> installed library:
>>
>> http://fenicsproject.org:8010/builders/dolfin-master-full-precise-amd64/builds/416/steps/configure%20%28enable%20all%29/logs/CMakeCache.txt
>> http://fenicsproject.org:8010/builders/dolfin-master-full-precise-amd64/builds/417/steps/configure%20%28enable%20all%29/logs/CMakeCache.txt
>>
>> The Debian package is version 3.1.1, so it shouldn't use that one
>> anyway since we require 4.0.2.
>>
>> > I'm seeing the following errors at the moment:
>> >
>> > master:
>> >
>> > * ImportError:libteuchos.so: cannot open shared object file: No such file 
>> > or directory;
>>
>> This happened after I upgraded to Trilinos 11.4.1 but it is fixed now
>> after everything was rebuilt last night. However, there is a new error
>> on this buildbot slave (wheezy-amd64):
>>
>> $ mpirun -np 3 ./demo_bcs
>> [debian-bbot:22320] *** An error occurred in MPI_Barrier
>> [debian-bbot:22320] *** on communicator MPI_COMM_WORLD
>> [debian-bbot:22320] *** MPI_ERR_COMM: invalid communicator
>> [debian-bbot:22320] *** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)
>> [debian-bbot:22319] 2 more processes have sent help message
>> help-mpi-errors.txt / mpi_errors_are_fatal
>> [debian-bbot:22319] Set MCA parameter "orte_base_help_aggregate" to 0
>> to see all help / error messages
>> $ OMPI_MCA_orte_base_help_aggregate=0 mpirun -np 3 ./demo_bcs
>> [debian-bbot:22324] *** An error occurred in MPI_Barrier
>> [debian-bbot:22324] *** on communicator MPI_COMM_WORLD
>> [debian-bbot:22324] *** MPI_ERR_COMM: invalid communicator
>> [debian-bbot:22324] *** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)
>> [debian-bbot:22325] *** An error occurred in MPI_Barrier
>> [debian-bbot:22325] *** on communicator MPI_COMM_WORLD
>> [debian-bbot:22325] *** MPI_ERR_COMM: invalid communicator
>> [debian-bbot:22325] *** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)
>> [debian-bbot:22326] *** An error occurred in MPI_Barrier
>> [debian-bbot:22326] *** on communicator MPI_COMM_WORLD
>> [debian-bbot:22326] *** MPI_ERR_COMM: invalid communicator
>> [debian-bbot:22326] *** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)
>
> I have no idea what this is, but since it appears in two places, both
> Wheezy in next and master, I would assume a problem with MPI in Wheezy?

I am trying to figure out this one.

>> > * ImportError:libteuchos.so: cannot open shared object file: No such file 
>> > or directory;
>> > * Strange segfaults on osx-10.7
>>
>> I am looking into this.
>
> Great.

Simply restarting the buildbot process fixed this problem.

Johannes
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics

Reply via email to