Locobot Python 3 install on Ubuntu 16.04 Issue
swiz23 opened this issue · 14 comments
When running ./locobot_install_all.sh -t full -p 3 -l interbotix, the install process exits early with the message below...
I believe this issue happens when running this line of code (sudo pip install catkin_pkg pyyaml empy rospkg).
That is because in this line of code above (sudo pip install --upgrade pip), you are upgrading pip, which installs the latest version (21.0). Unfortunately, this version drops support for Python2 packages (see here), which is my guess for why these error messages are showing up.
Not sure how to fix this... either don't upgrade pip, or if you do upgrade pip...
- install python3-pip and then use pip3 instead of pip
OR - upgrade pip to version 20.3.4 which still supports Python2.
Could you tell me how exactly to replicate the bug? I just ran an install on a clean machine and it worked fine.
I have a client who encountered this issue. I then tested it myself on a clean machine and replicated the issue myself. I used a clean install of Linux 16.04.7 on a NUC8i3BEH. After logging in for the first time, i just ran sudo apt update and sudo apt upgrade. Then I typed ./locobot_install_all.sh -t full -p 3 -l interbotix in a terminal (after curling the install script), and that was it.
Is it possible that you grabbed the install script from the development branch or did not update the computer? I noticed that the latest version of 'pip' dropped support for Python 2 on January 23rd.
Thanks for the detailed instruction - it very much likely to be the latter. Let me try it and get back to you.
Closing the issue - PR #145 should address this. Thank you swiz23!
Awesome. I saw that some CI checks didn't pass though. How will that affect the installation procedure?
Awesome. I saw that some CI checks didn't pass though. How will that affect the installation procedure?
It doesn't - if you check the commit, the only line changed is to specify a pip version, as you suggested. The CI didn't pass because I manually killed the job and used CI cluster as testing machine (the testing procedure installed from /project folder, instead of installation script, which is why we overlooked this issue).
Ok great!
How do you go about downgrading pip? Now that my system has pip v21, I can't even use the 'pip' command to get a downgraded version. Even just running 'pip' in the terminal results in the error message above.
Try sudo apt-get remove python-pip? And then sudo apt-get install python-pip==20.3.4
Nope, that doesn't work. I'm able to remove python-pip (which seems to only remove v8.1.1 - the default), but running the second commands results in an error. Also, the apt-cache policy command shows v8.1.1 as the only available package.
I think where python-pip stuff is installed and where 'pip' stuff is installed are in two separate locations. If there was just a way to wipe out the pip version 21 packages from the system, I image the upgrade pip command as shown in the install script will work - but not sure how to do this.
Try sudo pip install --force-reinstall pip==20.3.4?
Yup - tried that too, but nothing doing
The ugly way might work - delete the pip package installed at /usr/local/lib/python2.7/dist-packages/pip, then run sudo apt-get install --reinstall python2.7 (which in theory should get the default pip tied to py2.7 back)
So I deleted the package at /usr/local/lib/python2.7/dist-packages/pip. I also deleted /usr/local/bin/pip (at least I think that was the location). Afterwards, when I ran pip -V in the terminal, it pointed to my original python-pip installation (v8.1.1) located at /usr/lib/python2.7/dist-packages (I reinstalled it before doing this whole thing).
Then when I ran the installation script again, it updated to pip 20.3 correctly.
So it doesn't look like there was any need to run sudo apt-get install --reinstall python2.7 at all. Also, it looks like the 'apt' version of pip is installed in /usr/lib/... vs the pip version of pip is installed in /usr/local/lib/...
In any event, thanks for the help!