6

I am using Docker with image of Ubuntu 16.04. I am using this docker to run:

  • Import C++ classes in python with the help of boost.python
  • Import matlab's python package in python with the help Matlab Runtime (MCR).

Before I have installed MCR in the docker, I have installed boost using:

$ sudo apt-get install -y libboost-all-dev

and I was able to wrap C++ classes and call them in python. Then I have installed MCR and one of the requirement is to set LD_LIBRARY_PATH to Matlab's library.

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/MATLAB/MATLAB_Runtime/v94/runtime/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/bin/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/sys/os/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/extern/bin/glnxa64

It worked, I was able to run matlab's python package in python but testing the C++ classes wrappers failed due to change of LD_LIBRARY_PATH. For example, I get the following error when I try to import C++ class in python:

 /usr/lib/x86_64-linux-gnu/libpython3.4m.so.1.0: undefined symbol: XML_SetHashSalt

If I unset LD_LIBRARY_PATH C++ import works but MCR fails. So the question what LD_LIBRARY_PATH should be in order to run both C++ classes and Matlab's python package successfully?

I have tried to add path to boost libraries to LD_LIBRARY_PATH but didn't work, so the env value would be:

/usr/lib/x86_64-linux-gnu:/usr/local/MATLAB/MATLAB_Runtime/v94/runtime/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/bin/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/sys/os/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/extern/bin/glnxa64

EDIT

the dependencies of C++ library file before setting the LD_LIBRARY_PATH:

linux-vdso.so.1 (0x00007ffcee0dc000)
    libpython3.4m.so.1.0 => /usr/lib/x86_64-linux-gnu/libpython3.4m.so.1.0 (0x00007f9d69e59000)
    libboost_python-py34.so.1.55.0 => /usr/lib/x86_64-linux-gnu/libboost_python-py34.so.1.55.0 (0x00007f9d69c09000)
    libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007f9d698fe000)
    libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f9d695fd000)
    libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007f9d693e7000)
    libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f9d6903c000)
    librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007f9d68e34000)
    libexpat.so.1 => /lib/x86_64-linux-gnu/libexpat.so.1 (0x00007f9d68c0b000)
    libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007f9d689f0000)
    libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f9d687d3000)
    libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f9d685cf000)
    libutil.so.1 => /lib/x86_64-linux-gnu/libutil.so.1 (0x00007f9d683cc000)
    /lib64/ld-linux-x86-64.so.2 (0x00007f9d6a6c3000)

After setting the LD_LIBRARY_PATH:

linux-vdso.so.1 (0x00007ffc42e9b000)
    libpython3.4m.so.1.0 => /usr/lib/x86_64-linux-gnu/libpython3.4m.so.1.0 (0x00007fad9635b000)
    libboost_python-py34.so.1.55.0 => /usr/lib/x86_64-linux-gnu/libboost_python-py34.so.1.55.0 (0x00007fad9610b000)
    libstdc++.so.6 => /usr/local/MATLAB/MATLAB_Runtime/v94/sys/os/glnxa64/libstdc++.so.6 (0x00007fad95d8a000)
    libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fad95a89000)
    libgcc_s.so.1 => /usr/local/MATLAB/MATLAB_Runtime/v94/sys/os/glnxa64/libgcc_s.so.1 (0x00007fad95873000)
    libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fad954c8000)
    librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007fad952c0000)
    libexpat.so.1 => /usr/local/MATLAB/MATLAB_Runtime/v94/bin/glnxa64/libexpat.so.1 (0x00007fad95095000)
    libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007fad94e7a000)
    libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fad94c5d000)
    libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fad94a59000)
    libutil.so.1 => /lib/x86_64-linux-gnu/libutil.so.1 (0x00007fad94856000)
    /lib64/ld-linux-x86-64.so.2 (0x00007fad96bc5000)

3 Answers 3

7
+50

This Stack Exchange Q&A has various ways of setting LD_LIBRARY_PATH the top voted answer suggests this is the best way:

sudo -H gedit /etc/ld.so.conf.d/randomLibs.conf

inside the file you are supposed to write the complete path to the directory that contains all the libraries that you wish to add to the system, for example

/home/linux/myLocalLibs

remember to add only the path to the dir, not the full path for the file, all the libs inside that path will be automatically indexed.

Save and run sudo ldconfig to update the system with this libs.

0
2

Do you have to export the LD_LIBRARY_PATH? can't you just preface the matlab call with it, e.g.

LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/MATLAB/MATLAB_Runtime/v94/runtime/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/bin/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/sys/os/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/extern/bin/glnxa64 /usr/local/MATLAB/MATLAB_Runtime/...

(I do not exactly know what the executable is called like)

You could put that in a little skript

Secondly, it could be possible to change the rpath of the executable or some library, so it looks at the correct place directly. I use patchelf for this

patchelf --set-rpath 

Since there are multiple paths involved it could get a bit tricky, though

2
  • Thanks for your answer, the preface didn't work with me.. I am trying to do the patchelf but it seems complicated. And in the future I might have many C++ classes.. wouldn't this mean that I will need to do it for all of them?
    – Coderji
    Aug 17, 2018 at 13:53
  • you would have to do this for every executable / shared library. But If you compile them yourself you must be able to set this at compile time ( or link time to be precise)
    – mbeyss
    Aug 17, 2018 at 14:03
1

The only way it worked for me is by defining LD_PRELOAD and point it to system library. So my current environment would be:

LD_LIBRARY_PATH=/usr/local/MATLAB/MATLAB_Runtime/v94/runtime/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/bin/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/sys/os/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/extern/bin/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v94/sys/opengl/lib/glnxa64
LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libexpat.so

So in docker, I have created a .matlab file that contains the mentioned env variables. Then, in dockerfile I added the following:

services:
  django: &django
    build:
      context: .
      dockerfile: ./compose/local/django/Dockerfile
    depends_on:
      - postgres
    volumes:
      - .:/app
      - shared_upload_tmp:/app_temp
      - media:/app/computation_server/media
    env_file:
      - ./.envs/.local/.django
      - ./.envs/.local/.postgres
      - ./.envs/.production/.matlab
    ports:
      - "8800:8000"
    command: /start

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .