This lesson is being piloted (Beta version)
If you teach this lesson, please tell the authors and provide feedback by opening an issue in the source repository

Intermediate Research Software Development: Common Issues, Fixes & Tips

Here is a list of issues previous participants of the course encountered and some tips to help you with troubleshooting.

Command Line/Git Bash Issues

Python Hangs in Git Bash

Hanging issues with trying to run Python 3 in Git Bash on Windows (i.e. typing python in the shell, which causes it to just hang with no error message or output). The solution appears to be to use winpty - a Windows software package providing an interface similar to a Unix pty-master for communicating with Windows command line tools. Inside the shell type:

$ alias python="winpty python.exe"

This alias will be valid for the duration of the shell session. For a more permanent solution, from the shell do:

$ echo "alias python='winpty python.exe'" >> ~/.bashrc
$ source ~/.bashrc

(and from there on remember to invoke Python as python or whatever command you aliased it to). Read more details on the issue at Stack Overflow or Superuser.

Customising Command Line Prompt

Minor annoyance with the ultra long prompt command line sometimes gives you - if you do not want a reminder of the current working directory, you can set it to just $ by typing the following in your command line: export PS1="$ ". More details on command line prompt customisation can be found in this guide.

Git/GitHub Issues

Connection Issues When Accessing GitHub Using Git Over VPN or Protected Networks - Proxy Needed

When accessing external services and websites (such as GitHub using git or to install Python packages with pip), you may experience connection errors (e.g. similar to fatal: unable to access '....': Failed connect to github.com) or a connection that hangs. This may indicate that they need to configure a proxy server user by your organisation to tunnel SSH traffic through a HTTP proxy.

To get git to work through a proxy server in Windows, you’ll need connect.exe program that comes with GitBash (which you should have installed as part of setup, so no additional installation is needed). If installed in the default location, this file should be found at C:\Program Files\Git\mingw64\bin\connect.exe. Next, you’ll need to modify your ssh config file (typically in ~/.ssh/config) and add the following:

Host github.com
    ProxyCommand "C:/Program Files/Git/mingw64/bin/connect.exe" -H <proxy-server>:<proxy-port> %h %p
    TCPKeepAlive yes
    IdentitiesOnly yes
    User git
    Port 22
    Hostname github.com

Mac and Linux users can use the Corkscrew tool for tunneling SSH through HTTP proxies, which would have to be installed separately. Next, you’ll need to modify your SSH config file (typically in ~/.ssh/config) and add the following:

Host github.com
    ProxyCommand corkscrew <proxy-server> <proxy-port> %h %p
    TCPKeepAlive yes
    IdentitiesOnly yes
    User git
    Port 22
    Hostname github.com

Creating a GitHub Key Without ‘Workflow’ Authorisation Scope

If a learner creates a GitHub authentication token but forgets to check ‘workflow’ scope (to allow the token to be used to update GitHub Action workflows) they will get the following error when trying to push a new workflow (when adding the pytest action in Section 2) to GitHub:

! [remote rejected] test-suite -> test-suite (refusing to allow an OAuth App to create or update workflow `.github/workflows/main.yml` without `workflow` scope`

The solution is to generate a new token with the correct scope/usage permissions and clear the local credential cache (if that’s where the token has been saved). In same cases, simply clearing credential cache was not enough and updating to Git 2.29 was needed.

Please tell me who you are Git Error

If you experience the following error the first time you do a Git commit, you may not have configured your identity with Git on your machine:

fatal: unable to auto-detect email address
*** Please tell me who you are

This can be configured from the command line as follows:

$ git config --global user.name "Your Name"
$ git config --global user.email "name@example.com"

The option --global tells Git to use these settings “globally” (i.e. for every project that uses Git for version control on your machine). If you use different identifies for different projects, then you should not use the --global option. Make sure to use the same email address you used to open an account on GitHub that you are using for this course.

At this point it may also be a good time to configure your favourite text editor with Git, if you have not already done so. For example, to use the editor nano with Git:

$ git config --global core.editor "nano -w"

SSH key authentication issues with Git Bash

Git Bash uses its own SSH library by default, which may result in errors such as the one below even after adding your SSH key correctly:

$ git clone git@github.com:https://github.com/ukaea-rse-training/python-intermediate-inflammation
Cloning into 'python-intermediate-inflammation'...
git@github.com: Permission denied (publickey).
fatal: Could not read from remote repository.

Please make sure you have the correct access rights
and the repository exists.

The solution is to change the SSH library used by Git:

$ git config --global core.sshCommand C:/windows/System32/OpenSSH/ssh.exe

Python, pip, venv & Installing Packages Issues

Issues With Numpy (and Potentially Other Packages) on New M1 Macs

When using numpy package installed via pip on a command line on a new Apple M1 Mac, you get a failed installation with the error:

...
mach-o file, but is an incompatible architecture (have 'x86_64', need 'arm64e').
...

Numpy is a package heavily optimised for performance, and many parts of it are written in C and compiled for specific architectures, such as Intel (x86_64, x86_32, etc.) or Apple’s M1 (arm64e). In this instance, pip is obtaining a version of numpy with the incorrect compiled binaries, instead of the ones needed for Apple’s M1 Mac. One way that was found to work was to install numpy via PyCharm into your environment instead, which seems able to determine the correct packages to download and install.

Python 3 Installed but not Found When Using python3 Command

Python 3 installed on some Windows machines may not be accessible using the python3 command from the command line, but works fine when invoked via the command python.

Connection Issues When Installing Packages With pip Over VPN or Protected Networks - Proxy Needed

If you encounter issues when trying to install packages with pip over your organisational network - it may be because your may need to use a proxy provided by your organisation. In order to get pip to use the proxy, you need to add an additional parameter when installing packages with pip:

$ python3 -m pip install --proxy <proxy-url> <name of package>`

To keep these settings permanently, you may want to add the following to your .zshrc/.bashrc file to avoid having to specify the proxy for each session, and restart your command line terminal:

# call set_proxies to set proxies and unset_proxies to remove them
set_proxies() {
export {http,https,ftp}_proxy='<proxy-url>'
export {HTTP,HTTPS,FTP}_PROXY='<proxy-url>'
export NO_PROXY=localhost,127.0.0.1,10.96.0.0/12,192.168.99.0/24,192.168.39.0/24,192.168.64.2,.<proxy-url>, <proxy-url>
}

unset_proxies() {
export {http,https,ftp}_proxy=
export {HTTP,HTTPS,FTP}_PROXY=
export NO_PROXY=
}

PyCharm Issues

Using GitBash from PyCharm

To embed Git Bash in PyCharm as external tool and work with it in PyCharm window, from Settings select “Tools->Terminal->Shell path” and enter “C:\Program Files\Git\bin\sh.exe” --login. See more details on Stack Overflow.

Virtual Environments Issue "no such option: –build-dir"

Using PyCharm to add a package to a virtual environment created from the command line using venv can fail with error "no such option: –build-dir", which appears to be caused by the latest version of pip (20.3) where the flag -build-dir was removed but is required by PyCharm to install packages. A workaround is to:

See the issue for more details. This issue seems to only occur with older versions of PyCharm - recent versions should be fine.

Invalid YAML Issue

If YAML is copy+pasted from the course material, it might not get pasted correctly in PyCharm and some extra indentation may occur. Annoyingly, PyCharm will not flag this up as invalid YAML and learners may get all sort of different issues and errors with these files - e.g. ‘actions must start with run or uses’ with GitHub Actions workflows.

An example of incorrect extra indentation:

steps:
    - name: foo
      uses: bar

Instead of

steps:
  - name: foo
    uses: bar