📝 Description

Summary

Replace os.path.commonpath() with Path.relative_to() in SafeTarExtractor._validate_member() to fix a path containment bypass. The commonpath approach (like str.startswith()) can be defeated by prefix collisions — e.g., extracting to /tmp/a with a member ../ab/evil.txt resolves to /tmp/ab/evil.txt, and commonpath returns /tmp/a which matches, allowing the traversal.

Changes

security/tar_safe.py

• _validate_member(): Replaced the os.path.commonpath() containment check with Path.relative_to(). The new approach raises ValueError if the resolved target path is not a true subpath of the extract path, correctly catching prefix-based bypasses (e.g., /tmp/a vs /tmp/ab). The old commonpath-based fallback for unresolvable paths was also removed since relative_to handles all cases cleanly.

tests/test_security.py

• test_prefix_bypass_path_traversal: New test that verifies a tar member targeting a sibling directory with a shared prefix (e.g., /tmp/a vs /tmp/ab) is correctly blocked — a case the old commonpath-based check would miss.
  </｜DSML｜parameter>
  </｜DSML｜invoke>
  </｜DSML｜tool_calls>

🔗 Linked Issue

Closes #1543

📋 Contribution Checklist

• I have verified that my files are placed in the correct directory.
• I have tested my changes thoroughly on my local machine.
• GSSoC 2026: I have been formally assigned to this issue and noted it above.