> On 6 Oct 2013, at 00:28, Dominique Orban <[email protected]> wrote:
> 
> Installing Swig 2.0.11 seems to have resolved the previous issue.
> 
> I'm now using the most recent TAO 2.2 and that doesn't get detected anymore. 
> CMakeFiles/CMakeErrors.log doesn't say a thing about TAO and this is all I 
> see during the configuration:
> 
> -- Checking for package 'TAO'
> -- TAO_DIR is /usr/local/Cellar/tao/2.2
> /tmp/dolfin/build/Makefile.tao:7: warning: overriding commands for target 
> `show'
> /usr/local/Cellar/tao/2.2/conf/tao_rules:25: warning: ignoring old commands 
> for target `show'
> /tmp/dolfin/build/Makefile.tao:7: warning: overriding commands for target 
> `show'
> /usr/local/Cellar/tao/2.2/conf/tao_rules:25: warning: ignoring old commands 
> for target `show'
> -- Performing test TAO_TEST_RUNS - Failed
> -- Performing test TAO_TEST_3RD_PARTY_LIBS_RUNS - Failed
> -- TAO could not be found. Be sure to set TAO_DIR, PETSC_DIR, and PETSC_ARCH. 
> (missing:  TAO_TEST_RUNS)
> 
> Any idea?
> 
> Also, any chance to merge my patch for FindPaSTiX.cmake below? It's still 
> necessary. Shall I submit a pull request?

Yes, go ahead and submit a pull request.

Garth


> 
> Thanks.
> 
> Dominique
> 
> 
>> On 2013-09-20, at 11:08 AM, Johan Hake <[email protected]> wrote:
>> 
>> That is a strange SWIG related linking error... What version of SWIG do you 
>> use? I would recommend version 2.0.9 or later.
>> 
>> Johan
>> 
>> 
>> On Fri, Sep 20, 2013 at 4:12 PM, Dominique Orban <[email protected]> 
>> wrote:
>> 
>> On 2013-09-19, at 4:21 PM, Garth N. Wells <[email protected]> wrote:
>> 
>>>> On 19 September 2013 21:13, Dominique Orban <[email protected]> 
>>>> wrote:
>>>> 
>>>>> On 2013-09-18, at 4:03 PM, Garth N. Wells <[email protected]> wrote:
>>>>> 
>>>>>> On 18 September 2013 20:33, Johan Hake <[email protected]> wrote:
>>>>>> There are no documentation of what software versions you need. This 
>>>>>> should
>>>>>> be taken care of by CMake. If you for example need a higher version of 
>>>>>> some
>>>>>> software CMake should tell you that. You should also be able to look at:
>>>>>> 
>>>>>> build/CMakeFiles/CMakeError.log
>>>>>> 
>>>>>> to check why the SLEPC/PASTIX tests did not pass.
>>>>> 
>>>>> Unfortunately the SLEPc CMake tests are not very verbose - they do not
>>>>> leave a trail in CMakeError.log.
>>>>> 
>>>>> PaStiX is notoriously difficult (I have also seen your message on the
>>>>> PaStiX list ;)), but should leave some info behind CMakeError.log.
>>>> 
>>>> Actually, I missed this bit in the screen output:
>>>> 
>>>> -- Performing Test PASTIX_TEST_RUNS
>>>> CMake Error: The following variables are used in this project, but they 
>>>> are set to NOTFOUND.
>>>> Please set them or make sure they are set and tested correctly in the 
>>>> CMake files:
>>>> RT_LIBRARY
>>>>   linked by target "cmTryCompileExec4032762153" in directory 
>>>> /tmp/dolfin/build/CMakeFiles/CMakeTmp
>>>> 
>>>> 
>>>> It looks like CMake is looking for librt. A quick search reveals that it 
>>>> may not be relevant on OSX:
>>>> 
>>>> https://github.com/wjwwood/serial/issues/24
>>>> http://dev.deluge-torrent.org/ticket/37
>>>> http://goo.gl/X2LVXu (see the answer with 8 points).
>>>> 
>>>> That may explain why PaSTiX isn't detected... ?!
>>> 
>>> Very likely. You could try putting a test in FindPaStiX.cmake to only
>>> look for librt on non-OSX systems.
>> 
>> 
>> I don't speak CMake but this seems to do it:
>> 
>> diff --git a/cmake/modules/FindPaStiX.cmake b/cmake/modules/FindPaStiX.cmake
>> index 1b53fa7..be2f60a 100644
>> --- a/cmake/modules/FindPaStiX.cmake
>> +++ b/cmake/modules/FindPaStiX.cmake
>> @@ -24,7 +24,7 @@ find_library(PASTIX_LIBRARY pastix
>> 
>> set(PASTIX_LIBRARIES ${PASTIX_LIBRARY})
>> 
>> -if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU")
>> +if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU" AND NOT APPLE)
>>   # Check for rt library
>>   find_library(RT_LIBRARY rt
>>     DOC "The RT library"
>> 
>> 
>> Now everything is correctly detected and compilation almost completes. It 
>> gets stuck here:
>> 
>> --
>> [ 95%] Swig source
>> Scanning dependencies of target _la
>> [ 97%] Building CXX object 
>> dolfin/swig/modules/la/CMakeFiles/_la.dir/modulePYTHON_wrap.cxx.o
>> Linking CXX shared module _la.so
>> Undefined symbols for architecture x86_64:
>>  "dolfin::LinearOperator::LinearOperator(void const**)", referenced from:
>>      SwigDirector_LinearOperator::SwigDirector_LinearOperator(_object*) in 
>> modulePYTHON_wrap.cxx.o
>>      SwigDirector_LinearOperator::SwigDirector_LinearOperator(_object*) in 
>> modulePYTHON_wrap.cxx.o
>>  "dolfin::LinearOperator::LinearOperator(void const**, dolfin::GenericVector 
>> const&, dolfin::GenericVector const&)", referenced from:
>>      SwigDirector_LinearOperator::SwigDirector_LinearOperator(_object*, 
>> dolfin::GenericVector const&, dolfin::GenericVector const&) in 
>> modulePYTHON_wrap.cxx.o
>>      SwigDirector_LinearOperator::SwigDirector_LinearOperator(_object*, 
>> dolfin::GenericVector const&, dolfin::GenericVector const&) in 
>> modulePYTHON_wrap.cxx.o
>>  "dolfin::PETScLinearOperator::PETScLinearOperator(void const**)", 
>> referenced from:
>>      
>> SwigDirector_PETScLinearOperator::SwigDirector_PETScLinearOperator(_object*) 
>> in modulePYTHON_wrap.cxx.o
>>      
>> SwigDirector_PETScLinearOperator::SwigDirector_PETScLinearOperator(_object*) 
>> in modulePYTHON_wrap.cxx.o
>>  "dolfin::uBLASLinearOperator::uBLASLinearOperator(void const**)", 
>> referenced from:
>>      
>> SwigDirector_uBLASLinearOperator::SwigDirector_uBLASLinearOperator(_object*) 
>> in modulePYTHON_wrap.cxx.o
>>      
>> SwigDirector_uBLASLinearOperator::SwigDirector_uBLASLinearOperator(_object*) 
>> in modulePYTHON_wrap.cxx.o
>> ld: symbol(s) not found for architecture x86_64
>> collect2: error: ld returned 1 exit status
>> make[2]: *** [dolfin/swig/modules/la/_la.so] Error 1
>> make[1]: *** [dolfin/swig/modules/la/CMakeFiles/_la.dir/all] Error 2
>> make: *** [all] Error 2
>> ---
>> 
>> Looks like some libraries may be missing. I'm using the latest of everything 
>> from the git repos.
>> 
>> Thanks!
>> 
>> Dominique
> 
> 
> 
> 
> 
_______________________________________________
fenics-support mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics-support

Reply via email to