[ITK Community] Wrapping ITK with homebrewed python on OS 10.8.5

[Michka Popoff]

Thanks for the answer.

For the moment I am able to work on my project, so I am not in a hurry. But I will need ITK with python wrappings in the future on OS X (and I think it would be a nice feature for other people too.) I am already using VTK intensively with python, so the next step would be to have ITK working out of the box on mac.

I would really like to help on this project and can free some time for it. I am available for testing, but would also like to participate in the coding effort. But I am afraid that my competences in c++ to python wrapping are very limited, I am a python dev and did use some c++ a long time ago, so I don't know if I would be able to do it (seems quite tough)
I do not realize exactly how much work there is to do, and how long it would take.

If you think that I could give I it a try, I would perhaps need some guidance to start with. I'll read the gccxml code to see if I can find out how it works first :)

Michka

On 29 oct. 2013, at 19:01, Matt McCormick <matt.mccormick at kitware.com> wrote:

> Hi Michka,
> 
> First, thank you very much for maintaining the Homebrew build of ITK!
> This is very important.
> 
> Thank you for the detailed report.  This is a known issue with GCCXML
> and challenges in its maintenance.  As recently as yesterday, we
> brainstormed with Brad King, the author of GCCXML, on how to address
> these challenges.  As you have discovered, explicit support must be
> added to GCCXML to handle the system headers that are specific to each
> compiler.  This is very burdensome, and so support is not present for
> the new Clang or newer versions of Visual Studio.  A replacement for
> GCCXML based on libclang should be more maintainable, but it is an
> open question on how the bolus of effort required will take place.
> 
> In sum, there is not an easy fix here for the near future.
> 
> Thanks,
> Matt
> 
> 
>> I am currently trying to wrap ITK with python. Python 2.7.5 was installed with homebrew on OS 10.8.5.  tested with ITK 4.4.2.
>> The problem is exactly the same on OS 10.7.5.
>> 
>> 
>> Some relevant options I defined in cmake for ITK :
>> 
>> BUILD_EXAMPLES = OFF
>> BUILD_DOCUMENTATION = OFF
>> BUILD_TESTING = OFF
>> BUILD_SHARED_LIBS = ON
>> CMAKE_BUILD_TYPE = Release
>> Module_ITKVtkGlue = ON
>> ITK_WRAP_PYTHON = ON
>> PYTHON_EXECUTABLE = /usr/local/bin/python
>> PYTHON_INCLUDE_DIR = /usr/local/Cellar/python/2.7.5/Frameworks/Python.framework/Versions/2.7/include/python2.7
>> PYTHON_LIBRARY = /usr/local/Cellar/python/2.7.5/Frameworks/Python.framework/Versions/2.7/lib/libpython2.7.dylib
>> ITK_USE_SYSTEM_GCCXML = OFF
>> ITK_WRAP_GCCXML = ON
>> CMAKE_CXX_COMPILER = /usr/bin/c++
>> CMAKE_C_COMPILER = /usr/bin/cc
>> 
>> Gccxml compiles correctly, but then when trying to generate the first xml file (Generating vcl_complex.xml), it fails :
>> 
>> In file included from /usr/include/c++/4.2.1/cstddef:51,
>>                 from /usr/include/c++/4.2.1/cstring:51,
>>                 from /usr/include/c++/4.2.1/bits/char_traits.h:45,
>>                 from /usr/include/c++/4.2.1/string:47,
>>                 from /tmp/insighttoolkit-nNsV/InsightToolkit-4.4.2/Modules/Core/Common/include/itkMacro.h:46,
>>                 from /tmp/insighttoolkit-nNsV/InsightToolkit-4.4.2/Modules/Core/Common/include/itkLightObject.h:21,
>>                 from /tmp/insighttoolkit-nNsV/InsightToolkit-4.4.2/Modules/Core/Common/include/itkObject.h:31,
>>                 from /tmp/insighttoolkit-nNsV/InsightToolkit-4.4.2/Modules/Core/Common/include/itkCommand.h:21,
>>                 from /tmp/insighttoolkit-nNsV/InsightToolkit-4.4.2/itk-build/Wrapping/Modules/ITKCommon/vcl_complex.cxx:1:
>> /usr/bin/../lib/clang/5.0/include/stddef.h:29:42: error: missing binary operator before token "("
>> /usr/bin/../lib/clang/5.0/include/stddef.h:37:39: error: missing binary operator before token "("
>> /usr/bin/../lib/clang/5.0/include/stddef.h:48:42: error: missing binary operator before token "("
>> 
>> I played a little bit around and was wondering about why there are two cstddef files, one in /usr/include/c++/, the other in /usr/bin/../lib/clang/ …
>> 
>> I then renamed the /usr/include to /usr/_include (this is a very dirty hack), and the compilation went to the end (and ITK works in python !).
>> 
>> I asked the same question on the gccxml mailing list, and got no answer for the moment.
>> 
>> My questions are the following :
>> 
>> 1) Why are there two includes folders, and to what compilers do they belong ? I guess one is from clang, the other from llvm-g++ ?
>> 2) Why does the compilation work when I rename the include folder so that then the compiler can't find it ?
>> 3) Is there an options somewhere in cmake to tell gccxml which include folder to use ? Setting the CMAKE_CXX_COMPILER to llvm-g++ is not a solution, you get the same errors.
>> 4) What is the status of the support of Clang for wrapping ITK with python ? (OS 10.9 is out, and Clang only)
>> 
>> I would be pleased to do some further tests or to invest some time in fixing this, but my c++ knowledge is rather bad. But once it is working I will upgrade the install formulas in homebrew so that people can install ITK on OS X with an one-liner :)
>> 
>> Michka
>> 
>> _______________________________________________
>> Community mailing list
>> Community at itk.org
>> http://public.kitware.com/cgi-bin/mailman/listinfo/community