Open fgeorgatos opened 11 years ago
The relevant errors messages from the above:
SCOTCH should be compiled without SCOTCH_PTHREAD
(0): ERROR: SCOTCH_dgraphInit: Scotch compiled with SCOTCH_PTHREAD and program not launched with MPI_THREAD_MULTIPLE
...
-- Performing test SCOTCH_ZLIB_TEST_RUNS
-- Performing test SCOTCH_ZLIB_TEST_RUNS - Failed
I don't know what's causing these because we're not seeing any problems like this on our side, although the mention of zlib
makes me shiver (cfr. https://github.com/hpcugent/easybuild-framework/issues/108).
Will figure this out post v1.3.0, I don't consider this to be a blocker.
zlib makes me shiver
Shiver not: something is seriously fishy in relation to zlib/zlib-devel/DOLFIN/Boost and friends. It may be needed to find out in a hackathon expedition, most likely.
Will figure this out post v1.3.0, I don't consider this to be a blocker.
It never was: I doubt anybody outside of the RHEL* family ever built it successfully until now, so there is no backward compatibility issue at all, AFAIK :-P
hmm, SCOTCH_PTHREAD should be enabled in non intel mpi builds according to: see https://github.com/hpcugent/easybuild-easyblocks/blob/develop/easybuild/easyblocks/s/scotch.py
if not self.toolchain.mpi_family() == toolchain.INTELMPI: #@UndefinedVariable
cflags += " -DSCOTCH_PTHREAD"
So I'm not sure why this is the case. I can faintly remember I needed to add thi
The zlib test runs are just failing because the scotch test runs failed...
@JensTimmerman: Where did you get that -DSCOTCH_PTHREAD
is only required for Intel builds? Maybe it should always be included, but we just don't notice?
Boegel, it's enabled in all non intel builds, I don't know why this is failing now, is the MPI_THREAD_MULTIPLE option somehow removed in dolfin?
Anyway, fotis, maybe you should look into building dolfin 1.0.1, it is shipping 2 of our patches by default now, and iirc it had some SCOTCH fix in its somwhere aswell.
This needs to be revisited for v1.4. Is this still an issue?
@fgeorgatos: Any updates on this?
Hi Ken,
On Thu, May 23, 2013 at 10:40 PM, Kenneth Hoste notifications@github.comwrote:
@fgeorgatos https://github.com/fgeorgatos: Any updates on this?
I suspect a little bit of libboost saga still going on; in one of our systems we see:
sw@d-cluster1-1:~$ time echo easyconfigs/d/DOLFIN/*eb|xargs -n1 eb
...
== sanity checking...
ERROR: EasyBuild encountered an exception (at easybuild/main.py:817 in
build_and_install_software): autoBuild Failed (last 300 chars):
ftware/DOLFIN/1.0.0-goalf-1.1.0-no-OFED-Python-2.7.3/lib/libdolfin.so:
undefined reference to 'boost::system::generic_category()'
collect2: ld returned 1 exit status
make[2]: *** [demo_stokes-iterative] Error 1
make[1]: *** [CMakeFiles/demo_stokes-iterative.dir/all] Error 2
make: *** [all] Error 2
== sanity checking...
ERROR: EasyBuild encountered an exception (at easybuild/main.py:817 in
build_and_install_software): autoBuild Failed (last 300 chars):
ftware/DOLFIN/1.0.0-goolf-1.4.10-Python-2.7.3/lib/libdolfin.so: undefined
reference to 'boost::system::generic_category()'
collect2: error: ld returned 1 exit status
make[2]: *** [demo_stokes-iterative] Error 1
make[1]: *** [CMakeFiles/demo_stokes-iterative.dir/all] Error 2
make: *** [all] Error 2
== sanity checking...
ERROR: EasyBuild encountered an exception (at easybuild/main.py:817 in
build_and_install_software): autoBuild Failed (last 300 chars): t'
/opt/apps/HPCBIOS.20130517/software/DOLFIN/1.0.0-ictce-4.0.6-Python-2.7.3/lib/libdolfin.so:
undefined reference to 'boost::filesystem3::path::parent_path() const'
make[2]: *** [demo_stokes-iterative] Error 1
make[1]: *** [CMakeFiles/demo_stokes-iterative.dir/all] Error 2
make: *** [all] Error 2
ps. We have no DOLFIN users over here, but we may care about this as part of the PRACE prereqs; (see fenics) ie. low priority so far for us but, eventually it would be nice to get it ironed out.
(I tried fixing the comment above w.r.t. formatting, but GitHub isn't doing what it's supposed to be doing, very weird, so pasted below)
Hi Ken,
On Thu, May 23, 2013 at 10:40 PM, Kenneth Hoste notifications@github.comwrote:
@fgeorgatos https://github.com/fgeorgatos: Any updates on this?
I suspect a little bit of libboost saga still going on; in one of our systems we see:
sw@d-cluster1-1:~$ time echo easyconfigs/d/DOLFIN/*eb|xargs -n1 eb
...
== sanity checking...
ERROR: EasyBuild encountered an exception (at easybuild/main.py:817 in
build_and_install_software): autoBuild Failed (last 300 chars):
ftware/DOLFIN/1.0.0-goalf-1.1.0-no-OFED-Python-2.7.3/lib/libdolfin.so:
undefined reference to 'boost::system::generic_category()'
collect2: ld returned 1 exit status
make[2]: *** [demo_stokes-iterative] Error 1
make[1]: *** [CMakeFiles/demo_stokes-iterative.dir/all] Error 2
make: *** [all] Error 2
== sanity checking...
ERROR: EasyBuild encountered an exception (at easybuild/main.py:817 in
build_and_install_software): autoBuild Failed (last 300 chars):
ftware/DOLFIN/1.0.0-goolf-1.4.10-Python-2.7.3/lib/libdolfin.so: undefined
reference to 'boost::system::generic_category()'
collect2: error: ld returned 1 exit status
make[2]: *** [demo_stokes-iterative] Error 1
make[1]: *** [CMakeFiles/demo_stokes-iterative.dir/all] Error 2
make: *** [all] Error 2
== sanity checking...
ERROR: EasyBuild encountered an exception (at easybuild/main.py:817 in
build_and_install_software): autoBuild Failed (last 300 chars): t'
/opt/apps/HPCBIOS.20130517/software/DOLFIN/1.0.0-ictce-4.0.6-Python-2.7.3/lib/libdolfin.so:
undefined reference to 'boost::filesystem3::path::parent_path() const'
make[2]: *** [demo_stokes-iterative] Error 1
make[1]: *** [CMakeFiles/demo_stokes-iterative.dir/all] Error 2
make: *** [all] Error 2
ps. We have no DOLFIN users over here, but we may care about this as part of the PRACE prereqs; (see fenics) ie. low priority so far for us but, eventually it would be nice to get it ironed out.
@fgeorgatos: Can you check whether that system as a system-wide Boost installed? And if so, which version?
hm... yes, I discovered an interesting situation as regards packages provided libboost: v1.42.0-4 for libs AND headers, along with some extra headers from v1.42.0.1 (debian provides distinct packages for that). In theory, building DOLFIN should be independent from that, but I'm well aware of the mess that can arise out of it...
It's possible that EB isn't properly specifying Boost paths during the DOLFIN build, and thus is able to resort to the system-provided Boost libs/includes.
We'll have to look into this (maybe in combination with testing the build of DOLFIN 1.0)...
this has to be retargeted for v1.6 or even later.
Just reporting that DOLFIN is still tricky over here, this is nothing of urgency though:
sw@gaia-43:~$ time eb DOLFIN-1.0.0-goolf-1.4.10-Python-2.7.3.eb --strict=ignore -r
== temporary log file in case of crash /tmp/easybuild-srLuh9.log
== resolving dependencies ...
== processing EasyBuild easyconfig /opt/apps/HPCBIOS.20131117/software/EasyBuild/1.9.0/lib/python2.6/site-packages/easybuild_easyconfigs-1.9.0.0-py2.6.egg/easybuild/easyconfigs/d/DOLFIN/DOLFIN-1.0.0-goolf-1.4.10-Python-2.7.3.eb
== building and installing DOLFIN/1.0.0-goolf-1.4.10-Python-2.7.3...
== fetching files...
== creating build dir, resetting environment...
== unpacking...
== patching...
== preparing...
== configuring...
== building...
== testing...
== installing...
== taking care of extensions...
== packaging...
== postprocessing...
== sanity checking...
ERROR: EasyBuild encountered an exception (at easybuild/main.py:770 in build_and_install_software): autoBuild Failed (last 300 chars): 10-Python-2.7.3/share/dolfin/demo/pde/stokes-iterative/cpp'
make[1]: *** [CMakeFiles/demo_stokes-iterative.dir/all] Error 2
make[1]: Leaving directory `/mnt/nfs/apps/HPCBIOS.20131117/software/DOLFIN/1.0.0-goolf-1.4.10-Python-2.7.3/share/dolfin/demo/pde/stokes-iterative/cpp'
make: *** [all] Error 2
)
real 7m28.507s
user 17m4.584s
sys 1m26.717s
sw@gaia-43:~$
ImportError: /dev/shm/DOLFIN/1.0.0/goolf-1.4.10-Python-2.7.3/dolfin-1.0.0/dolfin/libdolfin.so.1.0: undefined symbol: _ZN5boost11filesystem36detail6statusERK
If somebody could dump his output of the (equivalent of the) following command, it could be interesting:
ldd /opt/apps/HPCBIOS.20131117/software/DOLFIN/1.0.0-goolf-1.4.10-Python-2.7.3/lib/libdolfin.so|grep -i boost
@fgeorgatos:
$ ldd /user/scratchdelcatty/gent/vsc400/vsc40023/easybuild_REGTEST/SL6/sandybridge/software/DOLFIN/1.0.0-goolf-1.4.10-Python-2.7.3/lib/libdolfin.so | grep -i boost
/user/scratchdelcatty/gent/vsc400/vsc40023/easybuild_REGTEST/SL6/sandybridge/software/DOLFIN/1.0.0-goolf-1.4.10-Python-2.7.3/lib/libdolfin.so: /lib64/libz.so.1: version `ZLIB_1.2.3.3' not found (required by /user/scratchdelcatty/gent/vsc400/vsc40023/easybuild_REGTEST/SL6/sandybridge/software/libxml2/2.8.0-goolf-1.4.10/lib/libxml2.so.2)
libboost_filesystem.so.1.49.0 => /user/scratchdelcatty/gent/vsc400/vsc40023/easybuild_REGTEST/SL6/sandybridge/software/Boost/1.49.0-goolf-1.4.10-Python-2.7.3/lib/libboost_filesystem.so.1.49.0 (0x00002ad0500f4000)
libboost_program_options.so.1.49.0 => /user/scratchdelcatty/gent/vsc400/vsc40023/easybuild_REGTEST/SL6/sandybridge/software/Boost/1.49.0-goolf-1.4.10-Python-2.7.3/lib/libboost_program_options.so.1.49.0 (0x00002ad050314000)
libboost_system.so.1.49.0 => /user/scratchdelcatty/gent/vsc400/vsc40023/easybuild_REGTEST/SL6/sandybridge/software/Boost/1.49.0-goolf-1.4.10-Python-2.7.3/lib/libboost_system.so.1.49.0 (0x00002ad050574000)
libboost_thread.so.1.49.0 => /user/scratchdelcatty/gent/vsc400/vsc40023/easybuild_REGTEST/SL6/sandybridge/software/Boost/1.49.0-goolf-1.4.10-Python-2.7.3/lib/libboost_thread.so.1.49.0 (0x00002ad050778000)
libboost_iostreams.so.1.49.0 => /user/scratchdelcatty/gent/vsc400/vsc40023/easybuild_REGTEST/SL6/sandybridge/software/Boost/1.49.0-goolf-1.4.10-Python-2.7.3/lib/libboost_iostreams.so.1.49.0 (0x00002ad050993000)
libboost_math_tr1.so.1.49.0 => /user/scratchdelcatty/gent/vsc400/vsc40023/easybuild_REGTEST/SL6/sandybridge/software/Boost/1.49.0-goolf-1.4.10-Python-2.7.3/lib/libboost_math_tr1.so.1.49.0 (0x00002ad050bbb000)
libboost_mpi.so.1.49.0 => /user/scratchdelcatty/gent/vsc400/vsc40023/easybuild_REGTEST/SL6/sandybridge/software/Boost/1.49.0-goolf-1.4.10-Python-2.7.3/lib/libboost_mpi.so.1.49.0 (0x00002ad050df7000)
libboost_serialization.so.1.49.0 => /user/scratchdelcatty/gent/vsc400/vsc40023/easybuild_REGTEST/SL6/sandybridge/software/Boost/1.49.0-goolf-1.4.10-Python-2.7.3/lib/libboost_serialization.so.1.49.0 (0x00002ad051026000)
Hi,
i just hope this is not an error on my end but, can't be sure (and you hope vice-versa :)
The answer is no, not really, partly because of the elusive MTL4 dependency; as regards the latter, I just obtained a "proper" copy via another channel. fyi. My default builds are more less the ones at: http://hpcbios.readthedocs.org/en/latest/HPCBIOS_2012-92.html
So, the DOLFIN build insists failing for me, here's my current log (at least the last part of it which is the most interesting).
fyi. I have tried to rebuild SCOTCH, just in case;
N.B. notice the errors around MPI_THREAD_MULTIPLE... what do you see instead?