cv_bridge_boost error on ROS example policy

[Titouan Le Hir]

System information

• OS Platform and Distribution : Ubuntu 16.04
• Python version: 3.5.2
• Installed using pip or ROS: ROS Kinetic
• GPU model (if applicable): no GPU

Describe the result you are trying to replicate
I'm trying to use the example ROS policy

Provide the exact sequence of commands / steps that you executed to replicate this result
From Ubuntu Xenial fresh install, installed ROS Kinetic
Then made a Python 3.5.2 virtualenv, cloned and installed GQCNN and autolab_core packages
Installed alll dependancies and built with catkin_make

In order to be able to import cv2 I had to add the following in /my/virtualenv/path/lib/python3.5/site-packages/perception/image.py :

import sys
sys.path.remove('opt/ros/kinetic/lib/python2.7/dist-pkg')
import cv2
sys.path.append('opt/ros/kinetic/lib/python2.7/dist-pkg')

instead of import cv2 only
Maybe it is related

Describe the unexpected behavior
The ROS graspins policy works but the python example doesn't, I get the following error

ImportError: dynamic module does not define module export function (PyInit_cv_bridge_boost)
[INFO] [1590483364.054702]: signal_shutdown [atexit]
[INFO] [1590483364.064572]: topic[/rosout] removing connection to /rosout
[INFO] [1590483364.065849]: atexit

Other info / logs
Here's the traceback

Traceback (most recent call last):
  File "examples/policy_ros.py", line 125, in <module>
    grasp_resp = plan_grasp_segmask(color_im.rosmsg, depth_im.rosmsg,
  File "/home/actemium/gqcnn_ve/gqcnn_ve/lib/python3.5/site-packages/perception/image.py", line 250, in rosmsg
    return cv_bridge.cv2_to_imgmsg(self._data, encoding=self._encoding)
  File "/opt/ros/kinetic/lib/python2.7/dist-packages/cv_bridge/core.py", line 259, in cv2_to_imgmsg
    if self.cvtype_to_name[self.encoding_to_cvtype2(encoding)] != cv_type:
  File "/opt/ros/kinetic/lib/python2.7/dist-packages/cv_bridge/core.py", line 91, in encoding_to_cvtype2
    from cv_bridge.boost.cv_bridge_boost import getCvType

I guess the problem comes from Python 3.5.
I tried using GQCNN with Python 2.7 but I always get a cannot import name ABC error.
I tried building the ROS package with
catkin_make -DPYTHON_EXECUTABLE=/usr/bin/python3 -DPYTHON_INCLUDE_DIR=/usr/include/python3.5m -DPYTHON_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.5m.so
(from https://stackoverflow.com/questions/49221565/unable-to-use-cv-bridge-with-ros-kinetic-and-python3)
but this doesn't solve the issue
I also tried adding the "vision_opencv" ROS package in my workspace but I'd doesn't work either.

Thank you for your answers and your work on this project.

[Titouan Le Hir]

[Update]
I've been able to solve the issue following this tutorial : https://github.com/pulver22/gym-gazebo/wiki/ERROR:-Cv_bridge-for-Python3
So now I source the GQCNN package and I extend it with the cv-bridge python3 package.
But it leads to an other issue which is the following :

libgcc_s.so.1 must be installed for pthread_cancel to work

I have no idea how to solve this, I've made a lot of research on the internet and I have not been able to fix it.

Maybe I've done something wrong in the previous steps ?