[Bioclusters] mpiBlast OS X

Aaron Darling bioclusters@bioinformatics.org
Fri, 01 Oct 2004 09:41:05 -0500 

Another method I've used is to simply grep available library dirs for 
the missing symbol.  Unfortunately that method produces both the library 
defining the symbol and any other libraries referencing it, so it can be 
confusing which one to link...
I used to do this quite frequently on windows before I discovered MSDN 
:)  I'm a CodeWarrior so MSDN isn't built in to the IDE.

Joe Landman wrote:

> (popping head up for a moment)
>
> If you have the readelf/objdump tools around, and you know which 
> object file or library is complaining about a missing link, you can 
> walk the cat backwards (start at the object file or library with the 
> missing link, look at the symbols, and try to see in /usr/lib or 
> similar, which library has these symbols ... a simple foreach loop 
> with a readelf/objdump and a grep is quite helpful here).
>
> I have used this methodology with a compiler vendors compilers 
> recently to fix some chemistry code support.  It is annoying and time 
> consuming, but not impossible.
>
> Joe
>
> Aaron Darling wrote:
>
>> I haven't tried it (I don't have an OS X machine handy at the 
>> moment), but based on a perusal of NCBI's darwin.ncbi.mk file, it 
>> looks like they add "-framework CoreServices" and "-lc" to the link 
>> command.  You may also need "-framework Carbon", but I'm guessing not 
>> since that looks like it's just for gui apps on OS X.
>> Good luck, hope it works this time around!
>> -Aaron
>>
>> Nick Brockner wrote:
>>
>>> I now have it almost linking, except for this:
>>>
>>> ld: Undefined symbols:
>>> _DisposeHandle
>>> _HLock
>>> _HUnlock
>>> _MemError
>>> _NewHandle
>>> _SetHandleSize
>>> _FSPathMakeRef
>>> mpiCC: No such file or directory
>>>
>>> Using this command to link:
>>>
>>> mpiCC  -O2 -DINSTALL_PREFIX="\"/usr/local\"" -g -O2   -o mpiblast  
>>> mpiblast-blastjob.o mpiblast-db_spec.o mpiblast-blast_hooks.o 
>>> mpiblast-embed_rank.o mpiblast-file_util.o mpiblast-fragment_list.o 
>>> mpiblast-mpiblast_util.o mpiblast-mpiblast.o 
>>> mpiblast-mpiblast_config.o mpiblast-ncbi_sizeof.o 
>>> mpiblast-distributed_bioseq.o /usr/local/ncbi/build/ncbithr.o  
>>> -lncbitool -lncbiobj -lncbi -L/usr/local/ncbi/lib/  -lm -lpthread 
>>> -lstdc++
>>>
>>> I searched all over and cannot find what to link with for these.  I 
>>> appreciate all the help you have given so far with this.
>>>
>>> Thanks,
>>>
>>> -Nick
>>> On Sep 30, 2004, at 5:49 PM, Aaron Darling wrote:
>>>
>>>> Yes, looks like something wrong with the build environment.  Looks 
>>>> like it's not linking to the c++ library.  Usually the GNU standard 
>>>> c++ library can be linked with -lstdc++.  You could either add that 
>>>> to the CXXFLAGS environment variable and reconfigure or just copy 
>>>> the link command, add -lstdc++, and run it from within the src 
>>>> dir.  In the past we've only tested releases on Linux and Windows, 
>>>> sounds like that needs to change :)
>>>>
>>>> -Aaron
>>>>
>>>> Nick Brockner wrote:
>>>>
>>>>> Ok, now that getopt() is out of the way it compiles, but then when 
>>>>> it goes to link I get this garbage:
>>>>>
>>>>> ld: Undefined symbols:
>>>>> std::basic_ostream<char, std::char_traits<char> 
>>>>> >::operator<<(std::basic_ostream<char, std::char_traits<char> >& 
>>>>> (*)(std::basic_ostream<char, std::char_traits<char> >&))
>>>>> std::basic_ostream<char, std::char_traits<char> >::operator<<(double)
>>>>> std::basic_ostream<char, std::char_traits<char> >::operator<<(long)
>>>>> std::basic_ostream<char, std::char_traits<char> 
>>>>> >::operator<<(unsigned long)
>>>>> std::basic_string<char, std::char_traits<char>, 
>>>>> std::allocator<char> >::_Rep::_M_destroy(std::allocator<char> const&)
>>>>> .
>>>>> .
>>>>> .
>>>>>
>>>>> I think this has nothing to do with the mpiBlast code, and has 
>>>>> something to do with Apple's build environment.  Difficult build, 
>>>>> this is.
>>>>>
>>>>> Thanks,
>>>>>
>>>>> -Nick Brockner
>>>>> On Sep 30, 2004, at 3:35 PM, Aaron Darling wrote:
>>>>>
>>>>>> This happens on OS X, Solaris, and AIX because our getopt() 
>>>>>> detection was broken in 1.2.1.  A quick fix is to comment out 
>>>>>> line 147 of mpiBLAST-1.2.1/src/getopt.h
>>>>>> e.g. change:
>>>>>> extern int getopt ();
>>>>>> to:
>>>>>> /* extern int getopt (); */
>>>>>>
>>>>>> -Aaron
>>>>>>
>>>>>> Nick Brockner wrote:
>>>>>>
>>>>>>> Hi Aaron,
>>>>>>>
>>>>>>> Thanks for the pointer.  Using the Oct. 2003 ncbi toolkit worked 
>>>>>>> for the previous error, but now I have a more sinister looking 
>>>>>>> error:
>>>>>>>
>>>>>>> In file included from /usr/include/pthread.h:61,
>>>>>>>                  from 
>>>>>>> /usr/include/gcc/darwin/3.3/c++/ppc-darwin/bits/gthr-default.h:37,
>>>>>>>                  from 
>>>>>>> /usr/include/gcc/darwin/3.3/c++/ppc-darwin/bits/gthr.h:98,
>>>>>>>                  from 
>>>>>>> /usr/include/gcc/darwin/3.3/c++/ppc-darwin/bits/c++io.h:37,
>>>>>>>                  from 
>>>>>>> /usr/include/gcc/darwin/3.3/c++/bits/fpos.h:44,
>>>>>>>                  from /usr/include/gcc/darwin/3.3/c++/iosfwd:49,
>>>>>>>                  from 
>>>>>>> /usr/include/gcc/darwin/3.3/c++/bits/stl_algobase.h:70,
>>>>>>>                  from /usr/include/gcc/darwin/3.3/c++/algorithm:66,
>>>>>>>                  from mpiblast.cpp:86:
>>>>>>> /usr/include/unistd.h:183: error: declaration of C function `int 
>>>>>>> getopt(int,
>>>>>>>    char* const*, const char*)' conflicts with
>>>>>>> getopt.h:147: error: previous declaration `int getopt()' here
>>>>>>> In file included from file_util.hpp:58,
>>>>>>>                  from mpiblast.hpp:71,
>>>>>>>                  from mpiblast.cpp:88:
>>>>>>>
>>>>>>>
>>>>>>> Has anyone seen this before? Looks OS X specific.
>>>>>>>
>>>>>>> Thanks,
>>>>>>>
>>>>>>> -Nick Brockner
>>>>>>>
>>>>>>> On Sep 30, 2004, at 2:51 PM, Aaron Darling wrote:
>>>>>>>
>>>>>>>> Assuming you are using mpiBLAST 1.2.1, you need to use the 
>>>>>>>> October 2003 release of the NCBI Toolbox from here:
>>>>>>>> ftp://ftp.ncbi.nih.gov/toolbox/ncbi_tools/old/
>>>>>>>>
>>>>>>>> Each version of mpiBLAST is designed to compile against a 
>>>>>>>> specific version of the NCBI Toolbox, as mentioned on our 
>>>>>>>> download page here:
>>>>>>>>
>>>>>>>> http://mpiblast.lanl.gov/download.html
>>>>>>>>
>>>>>>>> Later toolbox releases changed the API to BlastMakeMultQueries 
>>>>>>>> and BXMLPrintOutput, which is why you are seeing those build 
>>>>>>>> errors.
>>>>>>>>
>>>>>>>> Regards,
>>>>>>>> -Aaron
>>>>>>>>
>>>>>>>> Nick Brockner wrote:
>>>>>>>>
>>>>>>>>> Hi All,
>>>>>>>>>
>>>>>>>>> I think I am missing something here, but attempting to compile 
>>>>>>>>> mpiBlast gets me the following:
>>>>>>>>>
>>>>>>>>> blast_hooks.c: In function `getFakeBioseq':
>>>>>>>>> blast_hooks.c:785: error: too few arguments to function 
>>>>>>>>> `BlastMakeMultQueries'
>>>>>>>>> blast_hooks.c: In function `outputResults':
>>>>>>>>> blast_hooks.c:1334: error: too few arguments to function 
>>>>>>>>> `BXMLPrintOutput'
>>>>>>>>> blast_hooks.c:1345: error: too few arguments to function 
>>>>>>>>> `BXMLPrintOutput'
>>>>>>>>> blast_hooks.c:1493: error: too few arguments to function 
>>>>>>>>> `BXMLPrintOutput'
>>>>>>>>> blast_hooks.c:1498: error: too few arguments to function 
>>>>>>>>> `BXMLPrintOutput'
>>>>>>>>> mpicc: No such file or directory
>>>>>>>>> make[1]: *** [mpiblast-blast_hooks.o] Error 1
>>>>>>>>> make: *** [all-recursive] Error 1
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> My setup:
>>>>>>>>> OS X Server 10.3.5
>>>>>>>>> gcc version 3.3 20030304 (Apple Computer, Inc. build 1495)
>>>>>>>>> NCBI toolkit installed
>>>>>>>>> LAM-MPI version 7.0.6
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Thanks,
>>>>>>>>>
>>>>>>>>> -Nick Brockner
>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> Bioclusters maillist  -  Bioclusters@bioinformatics.org
>>>>>>>>> https://bioinformatics.org/mailman/listinfo/bioclusters
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> Bioclusters maillist  -  Bioclusters@bioinformatics.org
>>>>>>>> https://bioinformatics.org/mailman/listinfo/bioclusters
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Bioclusters maillist  -  Bioclusters@bioinformatics.org
>>>>>>> https://bioinformatics.org/mailman/listinfo/bioclusters
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> Bioclusters maillist  -  Bioclusters@bioinformatics.org
>>>>>> https://bioinformatics.org/mailman/listinfo/bioclusters
>>>>>>
>>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Bioclusters maillist  -  Bioclusters@bioinformatics.org
>>>>> https://bioinformatics.org/mailman/listinfo/bioclusters
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Bioclusters maillist  -  Bioclusters@bioinformatics.org
>>>> https://bioinformatics.org/mailman/listinfo/bioclusters
>>>>
>>>
>>> _______________________________________________
>>> Bioclusters maillist  -  Bioclusters@bioinformatics.org
>>> https://bioinformatics.org/mailman/listinfo/bioclusters
>>
>>
>>
>> _______________________________________________
>> Bioclusters maillist  -  Bioclusters@bioinformatics.org
>> https://bioinformatics.org/mailman/listinfo/bioclusters
>
>
>