I am having troubles building a conda-forge package for ANTs using the libitk-devel package. I filed an issue upstream with details about the errors in the build log. They reckon it may come from differences in ITK configurations between this conda-forge package and what ANTs expect.

Could you help me identify what's missing and where the issue should be fixed?

Thanks.

I was trying to build RTK/SimpleRTK using the ITK package from conda-forge. Within my conda environment, the following file: lib/cmake/ITK-4.11/Modules/ITKHDF5.cmake has hardcoded paths for libraries such as libpthread.so which do not match their location on my Ubuntu installation.

Specifically the variable ITKHDF5_LIBRARIES is set as:

set(ITKHDF5_LIBRARIES "/home/lee8rx/miniconda3/envs/pyir/lib/libhdf5_cpp.so;/home/lee8rx/miniconda3/envs/pyir/lib/libhdf5.so;/usr/lib64/librt.so;/usr/lib64/libpthread.so;/home/lee8rx/miniconda3/envs/pyir/lib/libz.so;/usr/lib64/libdl.so;/usr/lib64/libm.so")

When building RTK, I get errors related to these libraries not being found because they are in /usr/lib/x86_64-linux-gnu/ rather than /usr/lib64/ on Ubuntu 14.04.

A workaround seems to be to replace all explicit links such as /usr/lib64/libpthread.so with just the name pthread, etc. I suppose this could be done manually within build.sh using sed or a similar tool, or is there perhaps a better solution?

The same thing happens with VTK as reported here: conda-forge/vtk-feedstock#8
and in Continuum's packages as well: ContinuumIO/anaconda-issues#1338

I see

Hey there,

is there any reason for restricting the hdf5 version to 1.8.18.*?
I tried building it against 1.10 and it seems to work.

Cheers and thanks for maintaining the recipe

@conda-forge-admin, please update circle

https://conda-forge.org/docs/maintainer/infrastructure.html#conda-forge-admin-please-update-circle

@blowekamp

We found this on OSX with current conda. The cmake file refers to
/Users/travis/miniconda3/conda-bld/libitk_1546796752855/_build_env/lib/libz.dylib
and therefore all other ITK libraries depend on this file.

Here's an example of a Travis log for our build [https://api.travis-ci.org/v3/job/480922636/log.txt], which seems to say it's using conda-4.5.12 and 4.13.0-hab14342_1001. Not that the log file is somewhat misleading as there /Users/travis/miniconda3/ does exist (while when building on local machine it doesn't of course).

More details are on SyneRBI/SIRF#295

The README need to be update to properly link to the libitk packages.

The old "itk" packages need to be marked as broken.

Since this package just provides the C++ headers / libraries, it could be renamed to libitk, which would be consistent libxml2, libdap4, libiconv, libtiff, etc., and would facilitate installation of the Python bindings package with conda install itk.

I have a package (convert3d) which requires one of the ITK remote modules (MorphologicalContourInterpolation).

Could we provide it through this recipe or should it be maintained in its own?

What do you reckon?

Comment:

Our STIR feedstock uses libitk in our recipe as follows:

requirements:
  build:
    - {{ compiler('c') }}
    - {{ compiler('cxx') }}
    - cmake
    - ninja
  host:
    - libitk-devel
    - libitk <5.2  # [win]
    - libitk  # [not win]
  run:
    - {{ pin_compatible('libitk', min_pin='x.x', max_pin='x.x') }}

(cutting out non-ITK stuff). The pin_compatible line was put in by @carterbox in an attempt to fix the problem that the tests or actual downloaded package uses a different version of ITK than what was used to build, resulting in shared-libraries not found, but it is still happening, see
conda-forge/stir-feedstock#34

Any suggestions how we should fix this properly?

Issue:
On Windows the itkdevel package contains build paths.

grep -r libitk_[0-9]\* Library/lib/cmake/
Library/lib/cmake//ITK-5.0/Modules/ITKPNG.cmake:set(ITKPNG_LIBRARIES "D:/bld/libitk_1574524116751/_h_env/Library/lib/libpng.lib;D:/bld/libitk_1574524116751/_h_env/Library/lib/z.lib")
Library/lib/cmake//ITK-5.0/Modules/ITKPNG.cmake:set(ITKPNG_INCLUDE_DIRS "${ITK_INSTALL_PREFIX}/include/ITK-5.0;D:/bld/libitk_1574524116751/_h_env/Library/include;D:/bld/libitk_1574524116751/_h_env/Library/include;D:/bld/libitk_1574524116751/_h_env/Library/include;D:/bld/libitk_1574524116751/_h_env/Library/include")
Library/lib/cmake//ITK-5.0/Modules/ITKTIFF.cmake:set(ITKTIFF_LIBRARIES "D:/bld/libitk_1574524116751/_h_env/Library/lib/tiff.lib")
Library/lib/cmake//ITK-5.0/Modules/ITKTIFF.cmake:set(ITKTIFF_INCLUDE_DIRS "${ITK_INSTALL_PREFIX}/include/ITK-5.0;D:/bld/libitk_1574524116751/_h_env/Library/include")
Library/lib/cmake//ITK-5.0/Modules/ITKJPEG.cmake:set(ITKJPEG_LIBRARIES "D:/bld/libitk_1574524116751/_h_env/Library/lib/jpeg.lib")
Library/lib/cmake//ITK-5.0/Modules/ITKJPEG.cmake:set(ITKJPEG_INCLUDE_DIRS "${ITK_INSTALL_PREFIX}/include/ITK-5.0;D:/bld/libitk_1574524116751/_h_env/Library/include")

Please see: conda-forge/gdcm-feedstock#3

The current gdcm package does not contain header files it only contains libraries and python wrapping. Additionally it also has a runtime python dependency. As we don't want to introduce a python dependency for this ITK package, it may be best to create a separate libgdcm package.

Issue:
Using this package on ubuntu 18.04 leads to the following error

In file included from /home/sirfuser/devel/STIR/src/IO/ITKImageInputFileFormat.cxx:41:0:
/home/sirfuser/miniconda3/include/ITK-5.0/itkGDCMSeriesFileNames.h:25:10: fatal error: gdcmSerieHelper.h: No such file or directory
 #include "gdcmSerieHelper.h"
          ^~~~~~~~~~~~~~~~~~~
compilation terminated.

My understanding is that this is because I have no system GDCM, and this package does not depend on gdcm, see #1.

sudo apt install libgdcm2-dev

did not help even (just like here) though this installs /usr/include/gdcm-2.8/gdcmSerieHelper.h, which might be an ITK issue?

$ conda info

     active environment : base
    active env location : /home/sirfuser/miniconda3
            shell level : 1
       user config file : /home/sirfuser/.condarc
 populated config files : /home/sirfuser/.condarc
          conda version : 4.8.0
    conda-build version : 3.18.11
         python version : 3.7.4.final.0
       virtual packages : __glibc=2.27
       base environment : /home/sirfuser/miniconda3  (writable)
           channel URLs : https://conda.anaconda.org/conda-forge/linux-64
                          https://conda.anaconda.org/conda-forge/noarch
                          https://repo.anaconda.com/pkgs/main/linux-64
                          https://repo.anaconda.com/pkgs/main/noarch
                          https://repo.anaconda.com/pkgs/r/linux-64
                          https://repo.anaconda.com/pkgs/r/noarch
          package cache : /home/sirfuser/miniconda3/pkgs
                          /home/sirfuser/.conda/pkgs
       envs directories : /home/sirfuser/miniconda3/envs
                          /home/sirfuser/.conda/envs
               platform : linux-64
             user-agent : conda/4.8.0 requests/2.22.0 CPython/3.7.4 Linux/4.15.0-29-generic ubuntu/18.04.2 glibc/2.27
                UID:GID : 1001:1001
             netrc file : None
           offline mode : False