I/O operation failed during extraction

[wpcarro]

Question

uv add numcodecs is failing, but I'm not sure how to troubleshoot the root cause

All I see is Streaming failed for numcodecs==0.16.4; downloading wheel to disk (I/O operation failed during extraction)

Any troubleshooting tips? I'd like to be able to self-service better

Platform

Linux 6.12.22+bpo-amd64 x86_64 GNU/Linux

Version

uv 0.8.14

[nooscraft]

Yeah, that error message is pretty unhelpful, it's hiding the actual I/O error details. The underlying std::io::Error should tell us what's really going on, but it's not being surfaced properly. That's definitely something we should fix.

In the meantime, here's are a few troubleshooting steps:

Disk space - Check with:

df -h ~/.cache/uv

If it's full, either free up space or point uv to a different cache location:

UV_CACHE_DIR=/somewhere/with/space uv add numcodecs

Permissions - Quick test:

touch ~/.cache/uv/test && rm ~/.cache/uv/test

If that fails, the cache dir probably has wrong ownership. Fix it or use a different location.

Corrupted cache - The wheel might be borked in cache. Try:

uv cache clean numcodecs
uv add numcodecs

Filesystem issues - Check if the mount is weird or read-only:

mount | grep cache
touch ~/.cache/uv/test_write && rm ~/.cache/uv/test_write

To get the actual error details, run with backtrace:

RUST_BACKTRACE=1 uv add numcodecs 2>&1 | tee uv-error.log

That should show the real std::io::Error with the OS error code, which will tell us exactly what's failing.

You can also manually test the wheel to see if it's corrupted:

# Grab the wheel URL from PyPI
curl -O <wheel-url>
unzip -t <wheel-file>.whl

Start with the cache clean and disk space steps, and see how it goes. If it's still failing, paste the backtrace output

[konstin]

The message you share should be a warning, can you share the complete output you get?