Test flakiness: find_repo_root walks above test tmp_path and finds ancestor .git

[azizur100389]

Summary

Two tests in tests/test_incremental.py fail on any developer machine where an ancestor of pytest's tmp_path (e.g. the user's home directory) contains a .git repository:

• TestFindRepoRoot::test_returns_none_without_git
• TestFindProjectRoot::test_falls_back_to_start

Reproduction

On Windows 11, pytest creates tmp_path under C:/Users/<username>/AppData/Local/Temp/pytest-of-<username>/.... If the user has run git init anywhere under C:/Users/<username>/ (for example, a dotfiles repo at ~/.git — very common on developer machines), then find_repo_root() walking up from tmp_path/no_git finds C:/Users/<username>/.git and returns the home directory instead of None.

Verified on clean origin/main:

tests\test_incremental.py:40: AssertionError
  find_repo_root(WindowsPath('C:/Users/erazi/AppData/Local/Temp/pytest-of-erazi/.../no_git'))
  returned WindowsPath('C:/Users/erazi') instead of None

The exact same failure mode affects Linux users who have a git-initialized home directory (e.g. chezmoi, yadm, or a bare git init ~).

Root cause

find_repo_root() in code_review_graph/incremental.py:74 has no way to bound its ancestor walk:

def find_repo_root(start: Path | None = None) -> Optional[Path]:
    current = start or Path.cwd()
    while current != current.parent:
        if (current / ".git").exists():
            return current
        current = current.parent
    ...

It walks all the way to the filesystem root. This is the correct behavior in production (a nested script inside a git repo should find that repo), but it makes the function un-testable in environments where the tmp path is inside a git-initialized ancestor.

Fix proposal

Add an optional stop_at: Path | None = None parameter to both find_repo_root() and find_project_root(). When provided, the walk halts after examining stop_at and returns None (for find_repo_root) or the start path (for find_project_root) rather than crossing above the boundary.

def find_repo_root(
    start: Path | None = None,
    stop_at: Path | None = None,
) -> Optional[Path]:
    current = start or Path.cwd()
    while current != current.parent:
        if (current / ".git").exists():
            return current
        if stop_at is not None and current == stop_at:
            return None
        current = current.parent
    ...

Semantics:

• stop_at=None (default): existing behavior, walk all the way to the filesystem root
• stop_at=<some path>: the walk examines stop_at for .git and then stops; it never crosses above stop_at

This is:

• Fully backward-compatible — all 7 production callers pass no stop_at
• Testable — the 2 flaky tests can pass stop_at=tmp_path to bound the walk
• Useful beyond tests — any caller that wants to bound the ancestor walk (embedded environments, multi-repo orchestrators, CI containers with bind-mounted volumes) can use the same mechanism

Not a product bug

Production behavior is correct. This is a test-only improvement to make find_repo_root env-independent, plus a small product API addition that's useful in its own right.

A PR implementing this with 2 test fixes + 1 new positive regression test for the stop_at parameter will follow.