gh-74623: Avoid untar errors when write-protected files are tarred twice

[arhadthedev]

This is a manual rebase of gh-1808.

The original author commited using a email slightly different from the one the new CLA is signed with. So I squashed everything to add the expected e-mail as Co-authored-by.

• Issue: tarfile would fail to extract tarballs with files under R/O directories (twice) #74623

[westurner]

I'm encountering what I believe is this issue with the build context tarfile passed to repo2podman from repo2docker, which has multiple copies of the same file in the archive (acccording to less archive.tar | grep '.git/objects' | sort -u | wc -l)

I can confirm that copying the tarfile.py from this PR into sys.path before python.zip results in successful tarfile.extractall()

Is there a good reason to not merge this PR?

https://devguide.python.org/core-developers/experts/index.html :

• @gustaebel
• @ethanfurman

[westurner]

I was getting PermissionError: [Errno 13] Permission denied:, which doesn't explain that it's due to ~ Counter()[filename] > 1

[westurner]

• tests: Lib/test/test_tarfile.py
• docs: https://github.com/python/cpython/blob/main/Doc/library/tarfile.rst
• release blurb?

[westurner]

How does tarfile.extractall() differ from tar -xzvf; before and after this PR?