[ewg] ofed 1.5.1 mvapich-intel build hang on fc12

[Jeremy Enos]

  Sorry-
Yes, the first error was against pgi- don't know why I wrote intel.  I 
don't know if the pgi completed on the subsequent run or not.  I'm 
building from a tree in /tmp w/ 22GB of free space, so network and disk 
issues shouldn't impede anything.  I'm not sure why bash would be 
different here than anywhere else- I'm not using any rpms that aren't 
distro provided.
I'll try a fresh attempt at both 1.5.1 and 1.5.2rc5 against intel 073 
that I just got today and get back to this.
thx-

     Jeremy

On 9/2/2010 7:23 AM, Jonathan Perkins wrote:
> Hi Jeremy, the first mail you sent was the invocation of building the
> mvapich2 rpm with the pgi compilers.  I'm not sure if this was just a
> copy and paste mistake or not.
>
> The 2nd mail seems to indicate that a source file was not found.  This
> is very strange.  I'm under the assumption that the gcc and perhaps
> the pgi compilations have already completed.  If that is the case then
> this error should not have been seen with the intel compiler unless
> you're working on an unstable filesystem.  Are you running out of disk
> space on the root filesystem?  Is this filesystem mounted over a
> network?
>
> This 3rd mail is showing an error from a later rpm and not the
> mvapich2 rpm (with intel compilers).  I believe this means that you
> got this to build successfully on another try but now you ran into a
> problem with the mpitests mvapich2 rpm (with intel compilers).  Does
> this sound correct?
>
> The error you're seeing now is coming from rpm and it seems that it
> can't find these libraries.  I'm not exactly sure what may be
> happening in this situation.  Is it possible that the shell for the
> rpmbuild is not taking your environment variables for some reason?
>
> On Thu, Sep 2, 2010 at 6:36 AM, Jeremy Enos<jenos at ncsa.uiuc.edu>  wrote:
>>   Now failing w/ a different error- anyone know what this means?
>>
>> Failed to install mpitests_mvapich2_intel RPM
>> See /tmp/OFED.1368.logs/mpitests_mvapich2_intel.rpminstall.log
>> You have new mail in /var/spool/mail/root
>> [root at ac42 OFED-1.5.1]# which icc
>> /opt/intel/Compiler/11.1/038/bin/intel64/icc
>> You have new mail in /var/spool/mail/root
>> [root at ac42 OFED-1.5.1]# tail -500
>> /tmp/OFED.1368.logs/mpitests_mvapich2_intel.rpminstall.log
>> error: Failed dependencies:
>>          libimf.so()(64bit) is needed by
>> mpitests_mvapich2_intel-3.2-916.x86_64
>>          libintlc.so.5()(64bit) is needed by
>> mpitests_mvapich2_intel-3.2-916.x86_64
>>          libsvml.so()(64bit) is needed by
>> mpitests_mvapich2_intel-3.2-916.x86_64
>> [root at ac42 OFED-1.5.1]#
>>
>> It's the same with either intel 038 or 072.
>>
>> I checked, those files are there:
>> [root at ac42 OFED-1.5.1]# find /opt/intel/ -name libimf.so
>> /opt/intel/Compiler/11.0/083/lib/intel64/libimf.so
>> /opt/intel/Compiler/11.1/072/lib/intel64/libimf.so
>> /opt/intel/Compiler/11.1/038/lib/intel64/libimf.so
>> [root at ac42 OFED-1.5.1]# find /opt/intel/ -name lib^C^C
>> [root at ac42 OFED-1.5.1]# ls
>> /opt/intel/Compiler/11.1/038/lib/intel64/libimf.so
>> /opt/intel/Compiler/11.1/038/lib/intel64/libimf.so
>> [root at ac42 OFED-1.5.1]# ls
>> /opt/intel/Compiler/11.1/038/lib/intel64/libintlc.so.5
>> /opt/intel/Compiler/11.1/038/lib/intel64/libintlc.so.5
>> [root at ac42 OFED-1.5.1]# ls
>> /opt/intel/Compiler/11.1/038/lib/intel64/libsvml.so
>> /opt/intel/Compiler/11.1/038/lib/intel64/libsvml.so
>> [root at ac42 OFED-1.5.1]# echo $LD_LIBRARY_PATH | grep -c
>> /opt/intel/Compiler/11.1/038/lib/intel64
>> 1
>> [root at ac42 OFED-1.5.1]#
>>
>>
>> On 9/1/2010 3:12 AM, Jeremy Enos wrote:
>>>   Looks like configure was just taking a _really_ long time.  It
>>> eventually finished and failed.  Log attached.
>>>
>>> icc: error #10236: File not found:  'darray_support.c'
>>> icc: command line error: no files specified; for help type "icc -help"
>>>
>>> [root at ac42 OFED-1.5.1]# icc --version
>>> icc (ICC) 11.1 20100414
>>> Copyright (C) 1985-2010 Intel Corporation.  All rights reserved.
>>>
>>> thx-
>>>
>>>      Jeremy
>>>
>>> On 9/1/2010 2:30 AM, Jeremy Enos wrote:
>>>>   On fc12, the mvapich intel build hangs on fc12 (up to date today).
>>>>
>>>> (this command never returns)
>>>> Running  rpmbuild --rebuild  --define '_topdir /var/tmp//OFED_topdir'
>>>> --define 'dist %{nil}' --t
>>>> arget x86_64 --define '_name mvapich2_pgi' --define 'impl ofa'
>>>> --define 'rdma --with-rdma=gen2'
>>>> --define 'ib_include --with-ib-include=/usr/include' --define
>>>> 'ib_libpath --with-ib-libpath=/usr
>>>> /lib64' --define 'shared_libs 1' --define 'romio 1' --define
>>>> 'comp_env CC=pgcc CXX=pgCC F77=pgf7
>>>> 7 F90=pgf90' --define 'auto_req 0' --define 'mpi_selector
>>>> /usr/bin/mpi-selector' --define '_prefix /usr/mpi/pgi/mvapich2-1.4.1'
>>>> /tmp/OFED-1.5.1/SRPMS/mvapich2-1.4.1-1.src.rpm
>>>>
>>>>      Jeremy
>>>>
>> _______________________________________________
>> ewg mailing list
>> ewg at lists.openfabrics.org
>> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
>>
>>
>
>