sensor_msgs/Image generates float instead of int with python3

edit

I am trying to compile ros from source to run some python code that requires python3 and opencv.

I got it mostly working, I think, but I have this strange error message when I try to publish an image using cv_bridge:

 File "/root/ros_catkin_ws/install_isolated/lib/python3/dist-packages/sensor_msgs/msg/_Image.py", line 134, in serialize
     buff.write(_get_struct_BI().pack(_x.is_bigendian, _x.step))
 struct.error: required argument is not an integer
(...)
rospy.exceptions.ROSSerializationException: field step must be unsigned integer type

I've edited this file and it turns out _x.step is a float, even though on the header of _Image.py it says it should be a UINT32. Manually changing this file and typecasting it into int, works, but I would like to fix what I did wrong.

The code that breaks this is very simple, just try to publish any image:

#!/usr/bin/env python3

import rospy
from sensor_msgs.msg import Image
from cv_bridge import CvBridge, CvBridgeError
import numpy as np

rospy.init_node('blank_image')
image_pub = rospy.Publisher("image_raw",Image, queue_size=1)

bridge = CvBridge()
blank_image = np.zeros((200,600,3), np.uint8)

r = rospy.Rate(10)
try:
    while not rospy.is_shutdown():
        try:
            image_pub.publish(bridge.cv2_to_imgmsg(blank_image, "bgr8"))
        except CvBridgeError as e:
            print(e)
        r.sleep()
    rospy.spin()
except KeyboardInterrupt:
    print("Shutting down")

Manually changing _Image.py is not a very good option since it is autogenerated (although I am not going to compile ros all the time...) I feel like I messed something up and that genpy is not generating the package appropriately.

My ros compilation script is:

#!/usr/bin/env bash
#from http://wiki.ros.org/kinetic/Installation/Source with minor adaptations to make it compile
OLDDIR=$PWD
set -e
rosdep init
rosdep update

echo "deb http://ppa.launchpad.net/ondrej/apache2/ubuntu xenial main " >>  /etc/apt/sources.list
echo "deb-src http://ppa.launchpad.net/ondrej/apache2/ubuntu xenial main " >>  /etc/apt/sources.list
apt-key adv --keyserver keyserver.ubuntu.com --recv-keys 14AA40EC0831756756D7F66C4F4EA0AAE5267A6C

apt-get update

mkdir ~/ros_catkin_ws
cd ~/ros_catkin_ws
# fix.py is a workaround for the issue https://github.com/vcstools/vcstools/issues/147
cp /root/face_recognition/fix.py ./

rosinstall_generator ros_comm sensor_msgs image_transport common_msgs --rosdistro kinetic --deps --wet-only --tar > kinetic-ros_comm-wet.rosinstall

### this needs to be python2.7
python fix.py

wstool init -j`nproc` src kinetic-ros_comm-wet-fixed.rosinstall

rosdep install --from-paths src --ignore-src --rosdistro kinetic -y

./src/catkin/bin/catkin_make_isolated --install -DCMAKE_BUILD_TYPE=Release --cmake-args -DPYTHON_VERSION=3.5

cd  /usr/lib/x86_64-linux-gnu/
ln -s /usr/lib/x86_64-linux-gnu/libboost_python-py35.so libboost_python3.so

cd $OLDDIR

I compile my catkin_ws with my own python3 code, vision_opencv-1.11.16 (I am trying to avoid installing opencv3) and using the line to catkin_make it as:

catkin_make --cmake-args -DPYTHON_VERSION=3.5 -DPYTHON_EXECUTABLE=/usr/bin/python3 -DPYTHON_INCLUDE_DIR=/usr/include/python3.5m -DPYTHON_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.5m.so

List of installed packages (apt list --installed ) is https://pastebin.com/E3W6EH7K

List of pip installed packages (pip freeze) is https://pastebin.com/hYv6skJa

List of pip installed packages (pip3 freeze) is https://pastebin.com/3cZVVgSZ

fix ... (more)