[BHARTIYAYASH]

Problem

With core.quotePath=false, Git can mix raw UTF-8 with C-style escapes in a quoted status path. Repo.untracked_files converts the entire quoted path to ASCII before decoding the escapes, so a valid filename such as café\x7f.txt raises UnicodeEncodeError. This also affects Repo.is_dirty(untracked_files=True).

The reproduction uses an actual temporary repository and Git's status output; DEL is permitted in Windows filenames, so it also runs on Windows.

Change

Reuse _unquote_path, which Repo.ignored already uses, and retain the existing string return type. The regression checks both core.quotePath=true and false, along with ordinary ASCII and Unicode paths.

Verification

• The new regression failed before the implementation change.
• Related filename tests: 4 passed, 1 skipped, 2 subtests passed.
• Repository/diff modules: 151 passed, 3 skipped, 2 xfailed, 16 subtests passed; three other tests initially failed because this machine defaults to main. All three passed when rerun with a process-local init.defaultBranch=master setting.
• Ruff lint and formatting, mypy on the changed source module, and git diff --check passed.
• Additional core-suite run: 1,290 passed, 58 skipped, 8 xfailed, and 88 subtests passed. It stopped at three test_move_rejects_intermediate_symlink setup failures because Windows denied os.symlink with WinError 1314. All three failures were reproduced on unmodified upstream ccc48117; the complete core suite was not run to completion.

Prepared and submitted by OpenAI Codex, an AI agent operating through BHARTIYAYASH's account. Human review is pending.

[Copilot]

🟢 Approval recommended

The narrow decoder change and targeted regression coverage leave no identified blocking issues.

No open issues 🎉

What changed in this PR

This PR fixes Unicode filename handling in Repo.untracked_files when Git mixes raw UTF-8 with escaped characters, also addressing the affected Repo.is_dirty(untracked_files=True) call.

Changes:

• Reuse _unquote_path instead of the ASCII-only decoding sequence while preserving string results.
• Add regression coverage for both core.quotePath settings, mixed Unicode and escaped characters, and ordinary filenames.

Review by GitHub Copilot; tests were not independently run.

File	Description
test/​test_repo.py	Tests untracked filenames and dirty-state detection under both quoting settings.
git/​repo/​base.py	Uses the existing path decoder to handle quoted Unicode filenames.

_{Review effort: Balanced}