无法卸载numpy：Fedora系统pip卸载权限问题求助

Hey there! I’ve run into similar permission headaches before when mixing sudo with Python package installs, so let’s break this down and get it sorted out.

Why This Happens #

Since you installed pandas with sudo, the system-level Python packages (including numpy, which is a core pandas dependency) are owned by the root user. When you try to uninstall with regular pip3, your user account doesn’t have the permissions to delete root-owned files. The "invalid cross-device link" error might pop up if some numpy files are stored across different mounted filesystems on your Fedora system.

Step-by-Step Solutions #

1. Uninstall with Sudo (Most Likely Fix)

Run the uninstall command with sudo to gain the necessary root permissions:

sudo pip3 uninstall numpy -vvv --log pip.txt

This should bypass the Errno 13 permission error because you’re acting as the root user who owns the package files.

2. Manual Removal If Cross-Device Links Persist

If the cross-device link error still sticks around, you can manually delete numpy’s installation directory:

• First, find exactly where numpy is installed:

  pip3 show numpy
  

  Look for the Location line (it’ll look something like /usr/local/lib/python3.6/site-packages).
• Delete the numpy directory and any related leftover files with sudo:

  sudo rm -rf /usr/local/lib/python3.6/site-packages/numpy*
  

  The * ensures you catch all numpy-related subdirectories and stray files.

3. Verify Uninstallation

Check if numpy is gone for good by running:

pip3 list | grep numpy

If there’s no output, numpy has been successfully uninstalled.

Pro Tip for Future Package Management #

As a Python beginner, I’d strongly recommend using virtual environments (like venv or virtualenv) instead of installing packages system-wide with sudo. Virtual environments let you isolate dependencies per project, avoid permission issues entirely, and keep your system Python clean. Here’s a quick start for venv:

# Create a virtual environment for your project
python3 -m venv my_project_env
# Activate it
source my_project_env/bin/activate
# Now install packages without sudo
pip install pandas numpy

When you’re done working, just run deactivate to exit the environment.

内容的提问来源于stack exchange，提问作者audiFanatic