dulwich/dulwich/porcelain.py

# e porcelain.py -- Porcelain-like layer on top of Dulwich
# Copyright (C) 2013 Jelmer Vernooij <jelmer@jelmer.uk>
#
# SPDX-License-Identifier: Apache-2.0 OR GPL-2.0-or-later
# Dulwich is dual-licensed under the Apache License, Version 2.0 and the GNU
# General Public License as published by the Free Software Foundation; version 2.0
# or (at your option) any later version. You can redistribute it and/or
# modify it under the terms of either of these two licenses.
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# You should have received a copy of the licenses; if not, see
# <http://www.gnu.org/licenses/> for a copy of the GNU General Public License
# and <http://www.apache.org/licenses/LICENSE-2.0> for a copy of the Apache
# License, Version 2.0.
#

"""Simple wrapper that provides porcelain-like functions on top of Dulwich.

Currently implemented:
 * archive
 * add
 * bisect{_start,_bad,_good,_skip,_reset,_log,_replay}
 * branch{_create,_delete,_list}
 * check_ignore
 * checkout
 * checkout_branch
 * clone
 * cone mode{_init, _set, _add}
 * commit
 * commit_tree
 * daemon
 * describe
 * diff_tree
 * fetch
 * filter_branch
 * for_each_ref
 * grep
 * init
 * interpret_trailers
 * ls_files
 * ls_remote
 * ls_tree
 * mailsplit
 * merge
 * merge_tree
 * mv/move
 * prune
 * pull
 * push
 * rm
 * remote{_add}
 * receive_pack
 * replace{_create,_delete,_list}
 * rerere{_status,_diff,_forget,_clear,_gc}
 * reset
 * revert
 * sparse_checkout
 * submodule_add
 * submodule_init
 * submodule_list
 * rev_list
 * tag{_create,_delete,_list}
 * upload_pack
 * update_server_info
 * var
 * write_commit_graph
 * status
 * shortlog
 * symbolic_ref
 * worktree{_add,_list,_remove,_prune,_lock,_unlock,_move}

These functions are meant to behave similarly to the git subcommands.
Differences in behaviour are considered bugs.

Note: one of the consequences of this is that paths tend to be
interpreted relative to the current working directory rather than relative
to the repository root.

Functions should generally accept both unicode strings and bytestrings
"""

import datetime
import fnmatch
import logging
import os
import posixpath
import re
import stat
import sys
import time
from collections import namedtuple
from collections.abc import Callable, Iterable, Iterator, Sequence
from collections.abc import Set as AbstractSet
from contextlib import AbstractContextManager, closing, contextmanager
from dataclasses import dataclass
from io import BytesIO, RawIOBase
from pathlib import Path
from typing import (
    IO,
    TYPE_CHECKING,
    Any,
    BinaryIO,
    TextIO,
    TypedDict,
    TypeVar,
    cast,
    overload,
)

if sys.version_info >= (3, 12):
    from typing import override
else:
    from typing_extensions import override

from ._typing import Buffer

if TYPE_CHECKING:
    import urllib3

    from .filter_branch import CommitData
    from .gc import GCStats
    from .maintenance import MaintenanceResult

from . import replace_me
from .archive import tar_stream
from .bisect import BisectState
from .client import (
    FetchPackResult,
    LsRemoteResult,
    SendPackResult,
    get_transport_and_path,
)
from .config import Config, ConfigFile, StackedConfig, read_submodules
from .diff_tree import (
    CHANGE_ADD,
    CHANGE_COPY,
    CHANGE_DELETE,
    CHANGE_MODIFY,
    CHANGE_RENAME,
    RENAME_CHANGE_TYPES,
    TreeChange,
    tree_changes,
)
from .errors import SendPackError
from .graph import can_fast_forward
from .ignore import IgnoreFilterManager
from .index import (
    ConflictedIndexEntry,
    Index,
    IndexEntry,
    _fs_to_tree_path,
    blob_from_path_and_stat,
    build_file_from_blob,
    build_index_from_tree,
    get_unstaged_changes,
    index_entry_from_stat,
    symlink,
    update_working_tree,
    validate_path_element_default,
    validate_path_element_hfs,
    validate_path_element_ntfs,
)
from .object_store import BaseObjectStore, tree_lookup_path
from .objects import (
    Blob,
    Commit,
    ObjectID,
    Tag,
    Tree,
    TreeEntry,
    format_timezone,
    parse_timezone,
    pretty_format_tree_entry,
)
from .objectspec import (
    parse_commit,
    parse_object,
    parse_ref,
    parse_reftuples,
    parse_tree,
)
from .pack import UnpackedObject, write_pack_from_container, write_pack_index
from .patch import (
    MailinfoResult,
    get_summary,
    write_commit_patch,
    write_object_diff,
    write_tree_diff,
)
from .protocol import ZERO_SHA, Protocol
from .refs import (
    HEADREF,
    LOCAL_BRANCH_PREFIX,
    LOCAL_NOTES_PREFIX,
    LOCAL_REMOTE_PREFIX,
    LOCAL_REPLACE_PREFIX,
    LOCAL_TAG_PREFIX,
    DictRefsContainer,
    Ref,
    SymrefLoop,
    _import_remote_refs,
    filter_ref_prefix,
    local_branch_name,
    local_replace_name,
    local_tag_name,
    parse_remote_ref,
    shorten_ref_name,
)
from .repo import BaseRepo, Repo, get_user_identity
from .server import (
    FileSystemBackend,
    ReceivePackHandler,
    TCPGitServer,
    UploadPackHandler,
)
from .server import update_server_info as server_update_server_info
from .sparse_patterns import (
    SparseCheckoutConflictError,
    apply_included_paths,
    determine_included_paths,
)
from .trailers import add_trailer_to_message, format_trailers, parse_trailers

# Module level tuple definition for status output
GitStatus = namedtuple("GitStatus", "staged unstaged untracked")

# TypeVar for preserving BaseRepo subclass types
T = TypeVar("T", bound="BaseRepo")

# Type alias for common repository parameter pattern
RepoPath = str | os.PathLike[str] | Repo


class TransportKwargs(TypedDict, total=False):
    """Keyword arguments accepted by get_transport_and_path."""

    operation: str | None
    thin_packs: bool
    report_activity: Callable[[int, str], None] | None
    quiet: bool
    include_tags: bool
    username: str | None
    password: str | None
    key_filename: str | None
    ssh_command: str | None
    pool_manager: "urllib3.PoolManager | None"


@dataclass
class CountObjectsResult:
    """Result of counting objects in a repository.

    Attributes:
      count: Number of loose objects
      size: Total size of loose objects in bytes
      in_pack: Number of objects in pack files
      packs: Number of pack files
      size_pack: Total size of pack files in bytes
    """

    count: int
    size: int
    in_pack: int | None = None
    packs: int | None = None
    size_pack: int | None = None


class NoneStream(RawIOBase):
    """Fallback if stdout or stderr are unavailable, does nothing."""

    def read(self, size: int = -1) -> None:
        """Read from stream (returns None as this is a null stream)."""
        return None

    def readall(self) -> bytes:
        """Read all bytes (returns empty bytes).

        Returns:
          Empty bytes object
        """
        return b""

    if sys.version_info >= (3, 12):

        @override
        def readinto(self, b: Buffer) -> int | None:
            return 0

        @override
        def write(self, b: Buffer) -> int | None:
            return len(cast(bytes, b)) if b else 0

    else:

        @override
        def readinto(self, b: bytearray | memoryview) -> int | None:  # type: ignore[override]
            return 0

        @override
        def write(self, b: bytes | bytearray | memoryview) -> int | None:  # type: ignore[override]
            return len(b) if b else 0


default_bytes_out_stream: BinaryIO = cast(
    BinaryIO, getattr(sys.stdout, "buffer", None) or NoneStream()
)
default_bytes_err_stream: BinaryIO = cast(
    BinaryIO, getattr(sys.stderr, "buffer", None) or NoneStream()
)


DEFAULT_ENCODING = "utf-8"


class Error(Exception):
    """Porcelain-based error."""

    def __init__(self, msg: str) -> None:
        """Initialize Error with message."""
        super().__init__(msg)


class RemoteExists(Error):
    """Raised when the remote already exists."""


class TimezoneFormatError(Error):
    """Raised when the timezone cannot be determined from a given string."""


class CheckoutError(Error):
    """Indicates that a checkout cannot be performed."""


def parse_timezone_format(tz_str: str) -> int:
    """Parse given string and attempt to return a timezone offset.

    Different formats are considered in the following order:

     - Git internal format: <unix timestamp> <timezone offset>
     - RFC 2822: e.g. Mon, 20 Nov 1995 19:12:08 -0500
     - ISO 8601: e.g. 1995-11-20T19:12:08-0500

    Args:
      tz_str: datetime string
    Returns: Timezone offset as integer
    Raises:
      TimezoneFormatError: if timezone information cannot be extracted
    """
    import re

    # Git internal format
    internal_format_pattern = re.compile("^[0-9]+ [+-][0-9]{,4}$")
    if re.match(internal_format_pattern, tz_str):
        try:
            tz_internal = parse_timezone(tz_str.split(" ")[1].encode(DEFAULT_ENCODING))
            return tz_internal[0]
        except ValueError:
            pass

    # RFC 2822
    import email.utils

    rfc_2822 = email.utils.parsedate_tz(tz_str)
    if rfc_2822 and rfc_2822[9] is not None:
        return rfc_2822[9]

    # ISO 8601

    # Supported offsets:
    # sHHMM, sHH:MM, sHH
    iso_8601_pattern = re.compile(
        "[0-9] ?([+-])([0-9]{2})(?::(?=[0-9]{2}))?([0-9]{2})?$"
    )
    match = re.search(iso_8601_pattern, tz_str)
    total_secs = 0
    if match:
        sign, hours, minutes = match.groups()
        total_secs += int(hours) * 3600
        if minutes:
            total_secs += int(minutes) * 60
        total_secs = -total_secs if sign == "-" else total_secs
        return total_secs

    # YYYY.MM.DD, MM/DD/YYYY, DD.MM.YYYY contain no timezone information

    raise TimezoneFormatError(tz_str)


def get_user_timezones() -> tuple[int, int]:
    """Retrieve local timezone as described in git documentation.

    https://raw.githubusercontent.com/git/git/v2.3.0/Documentation/date-formats.txt
    Returns: A tuple containing author timezone, committer timezone.
    """
    local_timezone = time.localtime().tm_gmtoff

    if os.environ.get("GIT_AUTHOR_DATE"):
        author_timezone = parse_timezone_format(os.environ["GIT_AUTHOR_DATE"])
    else:
        author_timezone = local_timezone
    if os.environ.get("GIT_COMMITTER_DATE"):
        commit_timezone = parse_timezone_format(os.environ["GIT_COMMITTER_DATE"])
    else:
        commit_timezone = local_timezone

    return author_timezone, commit_timezone


@overload
def open_repo(path_or_repo: T) -> AbstractContextManager[T]: ...


@overload
def open_repo(
    path_or_repo: str | os.PathLike[str],
) -> AbstractContextManager[Repo]: ...


def open_repo(
    path_or_repo: str | os.PathLike[str] | T,
) -> AbstractContextManager[T | Repo]:
    """Open an argument that can be a repository or a path for a repository."""
    if isinstance(path_or_repo, BaseRepo):
        return _noop_context_manager(path_or_repo)
    return Repo(path_or_repo)


@contextmanager
def _noop_context_manager(obj: T) -> Iterator[T]:
    """Context manager that has the same api as closing but does nothing."""
    yield obj


def _get_reflog_message(
    default_message: bytes, explicit_message: bytes | None = None
) -> bytes:
    """Get reflog message, checking GIT_REFLOG_ACTION environment variable.

    Args:
      default_message: Default message to use if no explicit message or env var
      explicit_message: Explicit message passed as argument (takes precedence)

    Returns:
      The reflog message with priority:
        1. explicit_message if provided
        2. GIT_REFLOG_ACTION environment variable if set
        3. default_message otherwise
    """
    if explicit_message is not None:
        return explicit_message

    env_action = os.environ.get("GIT_REFLOG_ACTION")
    if env_action is not None:
        return env_action.encode("utf-8")

    return default_message


@overload
def open_repo_closing(path_or_repo: T) -> AbstractContextManager[T]: ...


@overload
def open_repo_closing(
    path_or_repo: str | bytes | os.PathLike[str],
) -> AbstractContextManager[Repo]: ...


def open_repo_closing(
    path_or_repo: str | bytes | os.PathLike[str] | T,
) -> AbstractContextManager[T | Repo]:
    """Open an argument that can be a repository or a path for a repository.

    returns a context manager that will close the repo on exit if the argument
    is a path, else does nothing if the argument is a repo.
    """
    if isinstance(path_or_repo, BaseRepo):
        return _noop_context_manager(path_or_repo)
    return closing(Repo(path_or_repo))


def path_to_tree_path(
    repopath: str | bytes | os.PathLike[str],
    path: str | bytes | os.PathLike[str],
    tree_encoding: str = DEFAULT_ENCODING,
) -> bytes:
    """Convert a path to a path usable in an index, e.g. bytes and relative to the repository root.

    Args:
      repopath: Repository path, absolute or relative to the cwd
      path: A path, absolute or relative to the cwd
      tree_encoding: Encoding to use for tree paths
    Returns: A path formatted for use in e.g. an index
    """
    # Resolve might returns a relative path on Windows
    # https://bugs.python.org/issue38671
    if sys.platform == "win32":
        path = os.path.abspath(path)

    # Convert bytes paths to str for Path
    if isinstance(path, bytes):
        path = os.fsdecode(path)
    path = Path(path)
    resolved_path = path.resolve()

    # Resolve and abspath seems to behave differently regarding symlinks,
    # as we are doing abspath on the file path, we need to do the same on
    # the repo path or they might not match
    if sys.platform == "win32":
        repopath = os.path.abspath(repopath)

    # Convert bytes paths to str for Path
    if isinstance(repopath, bytes):
        repopath = os.fsdecode(repopath)
    repopath = Path(repopath).resolve()

    try:
        relpath = resolved_path.relative_to(repopath)
    except ValueError:
        # If path is a symlink that points to a file outside the repo, we
        # want the relpath for the link itself, not the resolved target
        if path.is_symlink():
            parent = path.parent.resolve()
            relpath = (parent / path.name).relative_to(repopath)
        else:
            raise
    if sys.platform == "win32":
        return str(relpath).replace(os.path.sep, "/").encode(tree_encoding)
    else:
        return bytes(relpath)


class DivergedBranches(Error):
    """Branches have diverged and fast-forward is not possible."""

    def __init__(self, current_sha: bytes, new_sha: bytes) -> None:
        """Initialize DivergedBranches error with current and new SHA values."""
        self.current_sha = current_sha
        self.new_sha = new_sha


def check_diverged(repo: BaseRepo, current_sha: ObjectID, new_sha: ObjectID) -> None:
    """Check if updating to a sha can be done with fast forwarding.

    Args:
      repo: Repository object
      current_sha: Current head sha
      new_sha: New head sha
    """
    try:
        can = can_fast_forward(repo, current_sha, new_sha)
    except KeyError:
        can = False
    if not can:
        raise DivergedBranches(current_sha, new_sha)


def archive(
    repo: str | BaseRepo,
    committish: str | bytes | Commit | Tag | None = None,
    outstream: BinaryIO | RawIOBase = default_bytes_out_stream,
    errstream: BinaryIO | RawIOBase = default_bytes_err_stream,
) -> None:
    """Create an archive.

    Args:
      repo: Path of repository for which to generate an archive.
      committish: Commit SHA1 or ref to use
      outstream: Output stream (defaults to stdout)
      errstream: Error stream (defaults to stderr)
    """
    if committish is None:
        committish = "HEAD"
    with open_repo_closing(repo) as repo_obj:
        c = parse_commit(repo_obj, committish)
        tree = repo_obj.object_store[c.tree]
        assert isinstance(tree, Tree)
        for chunk in tar_stream(repo_obj.object_store, tree, c.commit_time):
            outstream.write(chunk)


def update_server_info(repo: RepoPath = ".") -> None:
    """Update server info files for a repository.

    Args:
      repo: path to the repository
    """
    with open_repo_closing(repo) as r:
        server_update_server_info(r)


def write_commit_graph(repo: RepoPath = ".", reachable: bool = True) -> None:
    """Write a commit graph file for a repository.

    Args:
      repo: path to the repository or a Repo object
      reachable: if True, include all commits reachable from refs.
                 if False, only include direct ref targets.
    """
    with open_repo_closing(repo) as r:
        # Get all refs
        refs = list(r.refs.as_dict().values())
        if refs:
            r.object_store.write_commit_graph(refs, reachable=reachable)


def symbolic_ref(repo: RepoPath, ref_name: str | bytes, force: bool = False) -> None:
    """Set git symbolic ref into HEAD.

    Args:
      repo: path to the repository
      ref_name: short name of the new ref
      force: force settings without checking if it exists in refs/heads
    """
    with open_repo_closing(repo) as repo_obj:
        ref_path = _make_branch_ref(ref_name)
        if not force and ref_path not in repo_obj.refs.keys():
            ref_name_str = (
                ref_name.decode("utf-8", "replace")
                if isinstance(ref_name, bytes)
                else ref_name
            )
            raise Error(f"fatal: ref `{ref_name_str}` is not a ref")
        repo_obj.refs.set_symbolic_ref(HEADREF, ref_path)


def pack_refs(repo: RepoPath, all: bool = False) -> None:
    """Pack loose references into packed-refs file."""
    with open_repo_closing(repo) as repo_obj:
        repo_obj.refs.pack_refs(all=all)


def _get_variables(repo: RepoPath = ".") -> dict[str, str]:
    """Internal function to get all Git logical variables.

    Args:
      repo: Path to the repository

    Returns:
      A dictionary of all logical variables with values
    """
    with open_repo_closing(repo) as repo_obj:
        config = repo_obj.get_config_stack()

        # Define callbacks for each logical variable
        def get_author_ident() -> str | None:
            """Get GIT_AUTHOR_IDENT."""
            try:
                author_identity = get_user_identity(config, kind="AUTHOR")
                author_tz, _ = get_user_timezones()
                timestamp = int(time.time())
                return f"{author_identity.decode('utf-8', 'replace')} {timestamp} {author_tz:+05d}"
            except Exception:
                return None

        def get_committer_ident() -> str | None:
            """Get GIT_COMMITTER_IDENT."""
            try:
                committer_identity = get_user_identity(config, kind="COMMITTER")
                _, committer_tz = get_user_timezones()
                timestamp = int(time.time())
                return f"{committer_identity.decode('utf-8', 'replace')} {timestamp} {committer_tz:+05d}"
            except Exception:
                return None

        def get_editor() -> str | None:
            """Get GIT_EDITOR."""
            editor = os.environ.get("GIT_EDITOR")
            if editor is None:
                try:
                    editor_bytes = config.get(("core",), "editor")
                    editor = editor_bytes.decode("utf-8", "replace")
                except KeyError:
                    editor = os.environ.get("VISUAL") or os.environ.get("EDITOR")
            return editor

        def get_sequence_editor() -> str | None:
            """Get GIT_SEQUENCE_EDITOR."""
            sequence_editor = os.environ.get("GIT_SEQUENCE_EDITOR")
            if sequence_editor is None:
                try:
                    seq_editor_bytes = config.get(("sequence",), "editor")
                    sequence_editor = seq_editor_bytes.decode("utf-8", "replace")
                except KeyError:
                    # Falls back to GIT_EDITOR if not set
                    sequence_editor = get_editor()
            return sequence_editor

        def get_pager() -> str | None:
            """Get GIT_PAGER."""
            pager = os.environ.get("GIT_PAGER")
            if pager is None:
                try:
                    pager_bytes = config.get(("core",), "pager")
                    pager = pager_bytes.decode("utf-8", "replace")
                except KeyError:
                    pager = os.environ.get("PAGER")
            return pager

        def get_default_branch() -> str:
            """Get GIT_DEFAULT_BRANCH."""
            try:
                default_branch_bytes = config.get(("init",), "defaultBranch")
                return default_branch_bytes.decode("utf-8", "replace")
            except KeyError:
                # Git's default is "master"
                return "master"

        # Dictionary mapping variable names to their getter callbacks
        variable_callbacks: dict[str, Callable[[], str | None]] = {
            "GIT_AUTHOR_IDENT": get_author_ident,
            "GIT_COMMITTER_IDENT": get_committer_ident,
            "GIT_EDITOR": get_editor,
            "GIT_SEQUENCE_EDITOR": get_sequence_editor,
            "GIT_PAGER": get_pager,
            "GIT_DEFAULT_BRANCH": get_default_branch,
        }

        # Build the variables dictionary by calling callbacks
        variables: dict[str, str] = {}
        for var_name, callback in variable_callbacks.items():
            value = callback()
            if value is not None:
                variables[var_name] = value

        return variables


def var_list(repo: RepoPath = ".") -> dict[str, str]:
    """List all Git logical variables.

    Args:
      repo: Path to the repository

    Returns:
      A dictionary of all logical variables with their values
    """
    return _get_variables(repo)


def var(repo: RepoPath = ".", variable: str = "GIT_AUTHOR_IDENT") -> str:
    """Get the value of a specific Git logical variable.

    Args:
      repo: Path to the repository
      variable: The variable to query (e.g., 'GIT_AUTHOR_IDENT')

    Returns:
      The value of the requested variable as a string

    Raises:
      KeyError: If the requested variable has no value
    """
    variables = _get_variables(repo)
    if variable in variables:
        return variables[variable]
    else:
        raise KeyError(f"Variable {variable} has no value")


def commit(
    repo: RepoPath = ".",
    message: str | bytes | Callable[[Any, Commit], bytes] | None = None,
    author: bytes | None = None,
    author_timezone: int | None = None,
    committer: bytes | None = None,
    commit_timezone: int | None = None,
    encoding: bytes | None = None,
    no_verify: bool = False,
    signoff: bool | None = None,
    all: bool = False,
    amend: bool = False,
    sign: bool | None = None,
) -> bytes:
    """Create a new commit.

    Args:
      repo: Path to repository
      message: Optional commit message (string/bytes or callable that takes
        (repo, commit) and returns bytes)
      author: Optional author name and email
      author_timezone: Author timestamp timezone
      committer: Optional committer name and email
      commit_timezone: Commit timestamp timezone
      encoding: Encoding to use for commit message
      no_verify: Skip pre-commit and commit-msg hooks
      signoff: Add Signed-off-by line to commit message. If None, uses format.signoff config.
      all: Automatically stage all tracked files that have been modified
      amend: Replace the tip of the current branch by creating a new commit
      sign: GPG sign the commit. If None, uses commit.gpgsign config.
        If True, signs with default GPG key. If False, does not sign.
    Returns: SHA1 of the new commit
    """
    encoding_str = encoding.decode("ascii") if encoding else DEFAULT_ENCODING
    if isinstance(message, str):
        message = message.encode(encoding_str)
    if isinstance(author, str):
        author = author.encode(encoding_str)
    if isinstance(committer, str):
        committer = committer.encode(encoding_str)
    local_timezone = get_user_timezones()
    if author_timezone is None:
        author_timezone = local_timezone[0]
    if commit_timezone is None:
        commit_timezone = local_timezone[1]

    with open_repo_closing(repo) as r:
        # Handle amend logic
        merge_heads = None
        if amend:
            try:
                head_commit = r[r.head()]
                assert isinstance(head_commit, Commit)
            except KeyError:
                raise ValueError("Cannot amend: no existing commit found")

            # If message not provided, use the message from the current HEAD
            if message is None:
                message = head_commit.message
            # If author not provided, use the author from the current HEAD
            if author is None:
                author = head_commit.author
                if author_timezone is None:
                    author_timezone = head_commit.author_timezone
            # Use the parent(s) of the current HEAD as our parent(s)
            merge_heads = list(head_commit.parents)

        # If -a flag is used, stage all modified tracked files
        if all:
            index = r.open_index()
            normalizer = r.get_blob_normalizer()

            # Create a wrapper that handles the bytes -> Blob conversion
            if normalizer is not None:

                def filter_callback(data: bytes, path: bytes) -> bytes:
                    blob = Blob()
                    blob.data = data
                    normalized_blob = normalizer.checkin_normalize(blob, path)
                    data_bytes: bytes = normalized_blob.data
                    return data_bytes
            else:
                filter_callback = None

            unstaged_changes = list(
                get_unstaged_changes(index, r.path, filter_callback)
            )

            if unstaged_changes:
                # Convert bytes paths to strings for add function
                modified_files: list[str | bytes | os.PathLike[str]] = []
                for path in unstaged_changes:
                    if isinstance(path, bytes):
                        modified_files.append(path.decode())
                    else:
                        modified_files.append(path)

                add(r, paths=modified_files)

        # For amend, create dangling commit to avoid adding current HEAD as parent
        if amend:
            commit_sha = r.get_worktree().commit(
                message=message,
                author=author,
                author_timezone=author_timezone,
                committer=committer,
                commit_timezone=commit_timezone,
                encoding=encoding,
                no_verify=no_verify,
                sign=sign,
                signoff=signoff,
                merge_heads=merge_heads,
                ref=None,
            )
            # Update HEAD to point to the new commit with reflog message
            try:
                old_head = r.refs[HEADREF]
            except KeyError:
                old_head = None

            # Get the actual commit message from the created commit
            commit_obj = r[commit_sha]
            assert isinstance(commit_obj, Commit)
            commit_message = commit_obj.message
            default_message = b"commit (amend): " + commit_message
            # Truncate message if too long for reflog
            if len(default_message) > 100:
                default_message = default_message[:97] + b"..."
            reflog_message = _get_reflog_message(default_message)

            r.refs.set_if_equals(HEADREF, old_head, commit_sha, message=reflog_message)
            return commit_sha
        else:
            return r.get_worktree().commit(
                message=message,
                author=author,
                author_timezone=author_timezone,
                committer=committer,
                commit_timezone=commit_timezone,
                encoding=encoding,
                no_verify=no_verify,
                sign=sign,
                signoff=signoff,
                merge_heads=merge_heads,
            )


def commit_tree(
    repo: RepoPath,
    tree: ObjectID,
    message: str | bytes | None = None,
    author: bytes | None = None,
    committer: bytes | None = None,
) -> ObjectID:
    """Create a new commit object.

    Args:
      repo: Path to repository
      tree: An existing tree object
      message: Commit message
      author: Optional author name and email
      committer: Optional committer name and email
    """
    with open_repo_closing(repo) as r:
        if isinstance(message, str):
            message = message.encode(DEFAULT_ENCODING)
        return r.get_worktree().commit(
            message=message, tree=tree, committer=committer, author=author
        )


def interpret_trailers(
    message: str | bytes,
    *,
    trailers: list[tuple[str, str]] | None = None,
    trim_empty: bool = False,
    only_trailers: bool = False,
    only_input: bool = False,
    unfold: bool = False,
    parse: bool = False,
    where: str = "end",
    if_exists: str = "addIfDifferentNeighbor",
    if_missing: str = "add",
    separators: str = ":",
) -> bytes:
    r"""Parse and manipulate trailers in a commit message.

    This function implements the functionality of ``git interpret-trailers``,
    allowing parsing and manipulation of structured metadata (trailers) in
    commit messages.

    Trailers are key-value pairs at the end of commit messages, formatted like:
        Signed-off-by: Alice <alice@example.com>
        Reviewed-by: Bob <bob@example.com>

    Args:
        message: The commit message (string or bytes)
        trailers: List of (key, value) tuples to add as new trailers
        trim_empty: Remove trailers with empty values
        only_trailers: Output only the trailers, not the message body
        only_input: Don't add new trailers, only parse existing ones
        unfold: Join multiline trailer values into a single line
        parse: Shorthand for --only-trailers --only-input --unfold
        where: Where to add new trailers ('end', 'start', 'after', 'before')
        if_exists: How to handle duplicate keys
            - 'add': Always add
            - 'replace': Replace all existing
            - 'addIfDifferent': Add only if value differs from all existing
            - 'addIfDifferentNeighbor': Add only if value differs from neighbors
            - 'doNothing': Don't add if key exists
        if_missing: What to do if key doesn't exist ('add' or 'doNothing')
        separators: Valid separator characters (default ':')

    Returns:
        The processed message as bytes

    Examples:
        >>> msg = b"Subject\\n\\nBody text\\n"
        >>> interpret_trailers(msg, trailers=[("Signed-off-by", "Alice <alice@example.com>")])
        b'Subject\\n\\nBody text\\n\\nSigned-off-by: Alice <alice@example.com>\\n'

        >>> msg = b"Subject\\n\\nSigned-off-by: Alice\\n"
        >>> interpret_trailers(msg, only_trailers=True)
        b'Signed-off-by: Alice\\n'
    """
    # Handle --parse shorthand
    if parse:
        only_trailers = True
        only_input = True
        unfold = True

    # Convert message to bytes
    if isinstance(message, str):
        message_bytes = message.encode("utf-8")
    else:
        message_bytes = message

    # Parse existing trailers
    _message_body, parsed_trailers = parse_trailers(message_bytes, separators)

    # Apply unfold if requested
    if unfold:
        for trailer in parsed_trailers:
            # Replace newlines and multiple spaces with single space
            trailer.value = " ".join(trailer.value.split())

    # Apply trim_empty if requested
    if trim_empty:
        parsed_trailers = [t for t in parsed_trailers if t.value.strip()]

    # Add new trailers if requested and not only_input
    if not only_input and trailers:
        for key, value in trailers:
            message_bytes = add_trailer_to_message(
                message_bytes,
                key,
                value,
                separators[0],  # Use first separator as default
                where=where,
                if_exists=if_exists,
                if_missing=if_missing,
            )
        # Re-parse to get updated trailers for output
        if only_trailers:
            _message_body, parsed_trailers = parse_trailers(message_bytes, separators)

    # Return based on only_trailers flag
    if only_trailers:
        return format_trailers(parsed_trailers)
    else:
        return message_bytes


def stripspace(
    text: str | bytes,
    *,
    strip_comments: bool = False,
    comment_char: str = "#",
    comment_lines: bool = False,
) -> bytes:
    r"""Strip unnecessary whitespace from text.

    This function implements the functionality of ``git stripspace``, commonly
    used to clean up commit messages and other text content.

    Args:
        text: The text to process (string or bytes)
        strip_comments: If True, remove lines that begin with comment_char
        comment_char: The comment character to use (default: "#")
        comment_lines: If True, prepend comment_char to each line

    Returns:
        The processed text as bytes

    The function performs the following operations:
        1. If comment_lines is True, prepend comment_char + space to each line
        2. Strip trailing whitespace from each line
        3. If strip_comments is True, remove lines starting with comment_char
        4. Collapse multiple consecutive blank lines into a single blank line
        5. Remove leading blank lines
        6. Remove trailing blank lines
        7. Ensure the text ends with a newline (unless empty)

    Examples:
        >>> stripspace(b"  hello  \\n\\n\\nworld  \\n\\n")
        b'hello\\n\\nworld\\n'

        >>> stripspace(b"# comment\\ntext\\n", strip_comments=True)
        b'text\\n'

        >>> stripspace(b"line\\n", comment_lines=True)
        b'# line\\n'
    """
    from .stripspace import stripspace as _stripspace

    # Convert text to bytes
    if isinstance(text, str):
        text_bytes = text.encode("utf-8")
    else:
        text_bytes = text

    # Convert comment_char to bytes
    comment_char_bytes = (
        comment_char.encode("utf-8") if isinstance(comment_char, str) else comment_char
    )

    return _stripspace(
        text_bytes,
        strip_comments=strip_comments,
        comment_char=comment_char_bytes,
        comment_lines=comment_lines,
    )


def init(
    path: str | os.PathLike[str] = ".",
    *,
    bare: bool = False,
    symlinks: bool | None = None,
) -> Repo:
    """Create a new git repository.

    Args:
      path: Path to repository.
      bare: Whether to create a bare repository.
      symlinks: Whether to create actual symlinks (defaults to autodetect)
    Returns: A Repo instance
    """
    if not os.path.exists(path):
        os.mkdir(path)

    if bare:
        return Repo.init_bare(path)
    else:
        return Repo.init(path, symlinks=symlinks)


def _filter_transport_kwargs(**kwargs: object) -> TransportKwargs:
    """Filter kwargs to only include parameters accepted by get_transport_and_path.

    Args:
      **kwargs: Arbitrary keyword arguments

    Returns:
      Dictionary containing only the kwargs that get_transport_and_path accepts
    """
    valid_params = {
        "operation",
        "thin_packs",
        "report_activity",
        "quiet",
        "include_tags",
        "username",
        "password",
        "key_filename",
        "ssh_command",
        "pool_manager",
    }
    return cast(TransportKwargs, {k: v for k, v in kwargs.items() if k in valid_params})


def clone(
    source: str | bytes | Repo,
    target: str | os.PathLike[str] | None = None,
    bare: bool = False,
    checkout: bool | None = None,
    errstream: BinaryIO | RawIOBase = default_bytes_err_stream,
    outstream: BinaryIO | None = None,
    origin: str | None = "origin",
    depth: int | None = None,
    branch: str | bytes | None = None,
    config: Config | None = None,
    filter_spec: str | None = None,
    protocol_version: int | None = None,
    recurse_submodules: bool = False,
    **kwargs: str | bytes | Sequence[str | bytes],
) -> Repo:
    """Clone a local or remote git repository.

    Args:
      source: Path or URL for source repository
      target: Path to target repository (optional)
      bare: Whether or not to create a bare repository
      checkout: Whether or not to check-out HEAD after cloning
      errstream: Optional stream to write progress to
      outstream: Optional stream to write progress to (deprecated)
      origin: Name of remote from the repository used to clone
      depth: Depth to fetch at
      branch: Optional branch or tag to be used as HEAD in the new repository
        instead of the cloned repository's HEAD.
      config: Configuration to use
      filter_spec: A git-rev-list-style object filter spec, as an ASCII string.
        Only used if the server supports the Git protocol-v2 'filter'
        feature, and ignored otherwise.
      protocol_version: desired Git protocol version. By default the highest
        mutually supported protocol version will be used.
      recurse_submodules: Whether to initialize and clone submodules
      **kwargs: Additional keyword arguments including refspecs to fetch.
        Can be a bytestring, a string, or a list of bytestring/string.

    Returns: The new repository
    """
    if outstream is not None:
        import warnings

        warnings.warn(
            "outstream= has been deprecated in favour of errstream=.",
            DeprecationWarning,
            stacklevel=3,
        )
        # TODO(jelmer): Capture logging output and stream to errstream

    if config is None:
        config = StackedConfig.default()

    if checkout is None:
        checkout = not bare
    if checkout and bare:
        raise Error("checkout and bare are incompatible")

    if target is None:
        if isinstance(source, Repo):
            raise ValueError("target must be specified when cloning from a Repo object")
        elif isinstance(source, bytes):
            target = source.split(b"/")[-1].decode()
        else:
            target = source.split("/")[-1]

    if isinstance(branch, str):
        branch = branch.encode(DEFAULT_ENCODING)

    mkdir = not os.path.exists(target)

    if isinstance(source, Repo):
        # For direct repo cloning, use LocalGitClient
        from .client import GitClient, LocalGitClient

        client: GitClient = LocalGitClient(config=config)
        path = source.path
    else:
        source_str = source.decode() if isinstance(source, bytes) else source
        transport_kwargs = _filter_transport_kwargs(**kwargs)
        (client, path) = get_transport_and_path(
            source_str, config=config, **transport_kwargs
        )

    filter_spec_bytes: bytes | None = None
    if filter_spec:
        filter_spec_bytes = filter_spec.encode("ascii")

    repo = client.clone(
        path,
        str(target),  # Convert PathLike to str
        mkdir=mkdir,
        bare=bare,
        origin=origin,
        checkout=checkout,
        branch=branch.decode() if branch else None,  # Convert bytes to str
        progress=lambda data: (errstream.write(data), None)[1],
        depth=depth,
        filter_spec=filter_spec_bytes,
        protocol_version=protocol_version,
    )

    # Initialize and update submodules if requested
    if recurse_submodules and not bare:
        try:
            submodule_init(repo)
            submodule_update(repo, init=True, recursive=True)
        except FileNotFoundError as e:
            # .gitmodules file doesn't exist - no submodules to process
            logging.debug("No .gitmodules file found: %s", e)
        except KeyError as e:
            # Submodule configuration missing
            logging.warning("Submodule configuration error: %s", e)
            if errstream:
                errstream.write(
                    f"Warning: Submodule configuration error: {e}\n".encode()
                )

    return repo


def add(
    repo: str | os.PathLike[str] | Repo = ".",
    paths: Sequence[str | bytes | os.PathLike[str]]
    | str
    | bytes
    | os.PathLike[str]
    | None = None,
) -> tuple[list[str], set[str]]:
    """Add files to the staging area.

    Args:
      repo: Repository for the files
      paths: Paths to add. If None, stages all untracked and modified files from the
        current working directory (mimicking 'git add .' behavior).
    Returns: Tuple with set of added files and ignored files

    If the repository contains ignored directories, the returned set will
    contain the path to an ignored directory (with trailing slash). Individual
    files within ignored directories will not be returned.

    Note: When paths=None, this function adds all untracked and modified files
    from the entire repository, mimicking 'git add -A' behavior.
    """
    ignored = set()
    with open_repo_closing(repo) as r:
        repo_path = Path(r.path).resolve()
        ignore_manager = IgnoreFilterManager.from_repo(r)

        # Get unstaged changes once for the entire operation
        index = r.open_index()
        normalizer = r.get_blob_normalizer()
        if normalizer is not None:

            def filter_callback(data: bytes, path: bytes) -> bytes:
                blob = Blob()
                blob.data = data
                normalized_blob = normalizer.checkin_normalize(blob, path)
                data_bytes: bytes = normalized_blob.data
                return data_bytes
        else:
            filter_callback = None

        # Check if core.preloadIndex is enabled
        config = r.get_config_stack()
        preload_index = config.get_boolean(b"core", b"preloadIndex", False)

        all_unstaged_paths = list(
            get_unstaged_changes(index, r.path, filter_callback, preload_index)
        )

        if not paths:
            # When no paths specified, add all untracked and modified files from repo root
            paths = [str(repo_path)]
        relpaths = []
        if isinstance(paths, (str, bytes, os.PathLike)):
            paths = [paths]
        for p in paths:
            # Handle bytes paths by decoding them
            if isinstance(p, bytes):
                p = p.decode("utf-8")
            path = Path(p)
            if not path.is_absolute():
                # Make relative paths relative to the repo directory
                path = repo_path / path

            # Don't resolve symlinks completely - only resolve the parent directory
            # to avoid issues when symlinks point outside the repository
            if path.is_symlink():
                # For symlinks, resolve only the parent directory
                parent_resolved = path.parent.resolve()
                resolved_path = parent_resolved / path.name
            else:
                # For regular files/dirs, resolve normally
                resolved_path = path.resolve()

            try:
                relpath = str(resolved_path.relative_to(repo_path)).replace(os.sep, "/")
            except ValueError as e:
                # Path is not within the repository
                p_str = p.decode() if isinstance(p, bytes) else str(p)
                raise ValueError(
                    f"Path {p_str} is not within repository {repo_path}"
                ) from e

            # Handle directories by scanning their contents
            if resolved_path.is_dir():
                # Check if the directory itself is ignored
                dir_relpath = posixpath.join(relpath, "") if relpath != "." else ""
                if dir_relpath and ignore_manager.is_ignored(dir_relpath):
                    ignored.add(dir_relpath)
                    continue

                # When adding a directory, add all untracked files within it
                current_untracked = list(
                    get_untracked_paths(
                        str(resolved_path),
                        str(repo_path),
                        index,
                    )
                )
                for untracked_path in current_untracked:
                    # If we're scanning a subdirectory, adjust the path
                    if relpath != ".":
                        untracked_path = posixpath.join(relpath, untracked_path)

                    if not ignore_manager.is_ignored(untracked_path):
                        relpaths.append(untracked_path)
                    else:
                        ignored.add(untracked_path)

                # Also add unstaged (modified) files within this directory
                for unstaged_path in all_unstaged_paths:
                    if isinstance(unstaged_path, bytes):
                        unstaged_path_str = unstaged_path.decode("utf-8")
                    else:
                        unstaged_path_str = unstaged_path

                    # Check if this unstaged file is within the directory we're processing
                    unstaged_full_path = repo_path / unstaged_path_str
                    try:
                        unstaged_full_path.relative_to(resolved_path)
                        # File is within this directory, add it
                        if not ignore_manager.is_ignored(unstaged_path_str):
                            relpaths.append(unstaged_path_str)
                        else:
                            ignored.add(unstaged_path_str)
                    except ValueError:
                        # File is not within this directory, skip it
                        continue
                continue

            # FIXME: Support patterns
            if ignore_manager.is_ignored(relpath):
                ignored.add(relpath)
                continue
            relpaths.append(relpath)
        r.get_worktree().stage(relpaths)
    return (relpaths, ignored)


def _is_subdir(
    subdir: str | os.PathLike[str], parentdir: str | os.PathLike[str]
) -> bool:
    """Check whether subdir is parentdir or a subdir of parentdir.

    If parentdir or subdir is a relative path, it will be disamgibuated
    relative to the pwd.
    """
    parentdir_abs = os.path.realpath(parentdir) + os.path.sep
    subdir_abs = os.path.realpath(subdir) + os.path.sep
    return subdir_abs.startswith(parentdir_abs)


# TODO: option to remove ignored files also, in line with `git clean -fdx`
def clean(
    repo: str | os.PathLike[str] | Repo = ".",
    target_dir: str | os.PathLike[str] | None = None,
) -> None:
    """Remove any untracked files from the target directory recursively.

    Equivalent to running ``git clean -fd`` in target_dir.

    Args:
      repo: Repository where the files may be tracked
      target_dir: Directory to clean - current directory if None
    """
    if target_dir is None:
        target_dir = os.getcwd()

    with open_repo_closing(repo) as r:
        if not _is_subdir(target_dir, r.path):
            raise Error("target_dir must be in the repo's working dir")

        config = r.get_config_stack()
        config.get_boolean((b"clean",), b"requireForce", True)

        # TODO(jelmer): if require_force is set, then make sure that -f, -i or
        # -n is specified.

        index = r.open_index()
        ignore_manager = IgnoreFilterManager.from_repo(r)

        paths_in_wd = _walk_working_dir_paths(target_dir, r.path)
        # Reverse file visit order, so that files and subdirectories are
        # removed before containing directory
        for ap, is_dir in reversed(list(paths_in_wd)):
            # target_dir and r.path are both str, so ap must be str
            assert isinstance(ap, str)
            if is_dir:
                # All subdirectories and files have been removed if untracked,
                # so dir contains no tracked files iff it is empty.
                is_empty = len(os.listdir(ap)) == 0
                if is_empty:
                    os.rmdir(ap)
            else:
                ip = path_to_tree_path(r.path, ap)
                is_tracked = ip in index

                rp = os.path.relpath(ap, r.path)
                is_ignored = ignore_manager.is_ignored(rp)

                if not is_tracked and not is_ignored:
                    os.remove(ap)


def remove(
    repo: str | os.PathLike[str] | Repo = ".",
    paths: Sequence[str | bytes | os.PathLike[str]] = [],
    cached: bool = False,
) -> None:
    """Remove files from the staging area.

    Args:
      repo: Repository for the files
      paths: Paths to remove. Can be absolute or relative to the repository root.
      cached: Only remove from index, not from working directory
    """
    with open_repo_closing(repo) as r:
        index = r.open_index()
        blob_normalizer = r.get_blob_normalizer()

        for p in paths:
            # If path is absolute, use it as-is. Otherwise, treat it as relative to repo
            p_str = os.fsdecode(p) if isinstance(p, bytes) else str(p)
            if os.path.isabs(p_str):
                full_path = p_str
            else:
                # Treat relative paths as relative to the repository root
                full_path = os.path.join(r.path, p_str)
            tree_path = path_to_tree_path(r.path, full_path)
            # Convert to bytes for file operations
            full_path_bytes = os.fsencode(full_path)
            try:
                entry = index[tree_path]
                if isinstance(entry, ConflictedIndexEntry):
                    raise Error(f"{p_str} has conflicts in the index")
                index_sha = entry.sha
            except KeyError as exc:
                raise Error(f"{p_str} did not match any files") from exc

            if not cached:
                try:
                    st = os.lstat(full_path_bytes)
                except OSError:
                    pass
                else:
                    try:
                        blob = blob_from_path_and_stat(full_path_bytes, st)
                        # Apply checkin normalization to compare apples to apples
                        if blob_normalizer is not None:
                            blob = blob_normalizer.checkin_normalize(blob, tree_path)
                    except OSError:
                        pass
                    else:
                        try:
                            head_commit = r[r.head()]
                            assert isinstance(head_commit, Commit)
                            committed_sha = tree_lookup_path(
                                r.__getitem__, head_commit.tree, tree_path
                            )[1]
                        except KeyError:
                            committed_sha = None

                        if blob.id != index_sha and index_sha != committed_sha:
                            raise Error(
                                "file has staged content differing "
                                f"from both the file and head: {p_str}"
                            )

                        if index_sha != committed_sha:
                            raise Error(f"file has staged changes: {p_str}")
                        os.remove(full_path_bytes)
            del index[tree_path]
        index.write()


rm = remove


def mv(
    repo: str | os.PathLike[str] | Repo,
    source: str | bytes | os.PathLike[str],
    destination: str | bytes | os.PathLike[str],
    force: bool = False,
) -> None:
    """Move or rename a file, directory, or symlink.

    Args:
      repo: Path to the repository
      source: Path to move from
      destination: Path to move to
      force: Force move even if destination exists

    Raises:
      Error: If source doesn't exist, is not tracked, or destination already exists (without force)
    """
    with open_repo_closing(repo) as r:
        index = r.open_index()

        # Handle paths - convert to string if necessary
        if isinstance(source, bytes):
            source = source.decode(sys.getfilesystemencoding())
        elif hasattr(source, "__fspath__"):
            source = os.fspath(source)
        else:
            source = str(source)

        if isinstance(destination, bytes):
            destination = destination.decode(sys.getfilesystemencoding())
        elif hasattr(destination, "__fspath__"):
            destination = os.fspath(destination)
        else:
            destination = str(destination)

        # Get full paths
        if os.path.isabs(source):
            source_full_path = source
        else:
            # Treat relative paths as relative to the repository root
            source_full_path = os.path.join(r.path, source)

        if os.path.isabs(destination):
            destination_full_path = destination
        else:
            # Treat relative paths as relative to the repository root
            destination_full_path = os.path.join(r.path, destination)

        # Check if destination is a directory
        if os.path.isdir(destination_full_path):
            # Move source into destination directory
            basename = os.path.basename(source_full_path)
            destination_full_path = os.path.join(destination_full_path, basename)

        # Convert to tree paths for index
        source_tree_path = path_to_tree_path(r.path, source_full_path)
        destination_tree_path = path_to_tree_path(r.path, destination_full_path)

        # Check if source exists in index
        if source_tree_path not in index:
            raise Error(f"source '{source}' is not under version control")

        # Check if source exists in filesystem
        if not os.path.exists(source_full_path):
            raise Error(f"source '{source}' does not exist")

        # Check if destination already exists
        if os.path.exists(destination_full_path) and not force:
            raise Error(f"destination '{destination}' already exists (use -f to force)")

        # Check if destination is already in index
        if destination_tree_path in index and not force:
            raise Error(
                f"destination '{destination}' already exists in index (use -f to force)"
            )

        # Get the index entry for the source
        source_entry = index[source_tree_path]

        # Convert to bytes for file operations
        source_full_path_bytes = os.fsencode(source_full_path)
        destination_full_path_bytes = os.fsencode(destination_full_path)

        # Create parent directory for destination if needed
        dest_dir = os.path.dirname(destination_full_path_bytes)
        if dest_dir and not os.path.exists(dest_dir):
            os.makedirs(dest_dir)

        # Move the file in the filesystem
        if os.path.exists(destination_full_path_bytes) and force:
            os.remove(destination_full_path_bytes)
        os.rename(source_full_path_bytes, destination_full_path_bytes)

        # Update the index
        del index[source_tree_path]
        index[destination_tree_path] = source_entry
        index.write()


move = mv


def commit_decode(
    commit: Commit, contents: bytes, default_encoding: str = DEFAULT_ENCODING
) -> str:
    """Decode commit contents using the commit's encoding or default."""
    if commit.encoding:
        encoding = commit.encoding.decode("ascii")
    else:
        encoding = default_encoding
    return contents.decode(encoding, "replace")


def commit_encode(
    commit: Commit, contents: str, default_encoding: str = DEFAULT_ENCODING
) -> bytes:
    """Encode commit contents using the commit's encoding or default."""
    if commit.encoding:
        encoding = commit.encoding.decode("ascii")
    else:
        encoding = default_encoding
    return contents.encode(encoding)


def print_commit(
    commit: Commit,
    decode: Callable[[bytes], str],
    outstream: TextIO = sys.stdout,
) -> None:
    """Write a human-readable commit log entry.

    Args:
      commit: A `Commit` object
      decode: Function to decode commit data
      outstream: A stream file to write to
    """
    outstream.write("-" * 50 + "\n")
    outstream.write("commit: " + commit.id.decode("ascii") + "\n")
    if len(commit.parents) > 1:
        outstream.write(
            "merge: "
            + "...".join([c.decode("ascii") for c in commit.parents[1:]])
            + "\n"
        )
    outstream.write("Author: " + decode(commit.author) + "\n")
    if commit.author != commit.committer:
        outstream.write("Committer: " + decode(commit.committer) + "\n")

    time_tuple = time.gmtime(commit.author_time + commit.author_timezone)
    time_str = time.strftime("%a %b %d %Y %H:%M:%S", time_tuple)
    timezone_str = format_timezone(commit.author_timezone).decode("ascii")
    outstream.write("Date:   " + time_str + " " + timezone_str + "\n")
    if commit.message:
        outstream.write("\n")
        outstream.write(decode(commit.message) + "\n")
        outstream.write("\n")


def print_tag(
    tag: Tag, decode: Callable[[bytes], str], outstream: TextIO = sys.stdout
) -> None:
    """Write a human-readable tag.

    Args:
      tag: A `Tag` object
      decode: Function for decoding bytes to unicode string
      outstream: A stream to write to
    """
    outstream.write("Tagger: " + decode(tag.tagger) + "\n")
    time_tuple = time.gmtime(tag.tag_time + tag.tag_timezone)
    time_str = time.strftime("%a %b %d %Y %H:%M:%S", time_tuple)
    timezone_str = format_timezone(tag.tag_timezone).decode("ascii")
    outstream.write("Date:   " + time_str + " " + timezone_str + "\n")
    outstream.write("\n")
    outstream.write(decode(tag.message))
    outstream.write("\n")


def show_blob(
    repo: RepoPath,
    blob: Blob,
    decode: Callable[[bytes], str],
    outstream: TextIO = sys.stdout,
) -> None:
    """Write a blob to a stream.

    Args:
      repo: A `Repo` object
      blob: A `Blob` object
      decode: Function for decoding bytes to unicode string
      outstream: A stream file to write to
    """
    outstream.write(decode(blob.data))


def show_commit(
    repo: RepoPath,
    commit: Commit,
    decode: Callable[[bytes], str],
    outstream: TextIO = sys.stdout,
) -> None:
    """Show a commit to a stream.

    Args:
      repo: A `Repo` object
      commit: A `Commit` object
      decode: Function for decoding bytes to unicode string
      outstream: Stream to write to
    """
    from .diff import ColorizedDiffStream

    # Create a wrapper for ColorizedDiffStream to handle string/bytes conversion
    class _StreamWrapper:
        def __init__(self, stream: "ColorizedDiffStream") -> None:
            self.stream = stream

        def write(self, data: str | bytes) -> None:
            if isinstance(data, str):
                # Convert string to bytes for ColorizedDiffStream
                self.stream.write(data.encode("utf-8"))
            else:
                self.stream.write(data)

    with open_repo_closing(repo) as r:
        # Use wrapper for ColorizedDiffStream, direct stream for others
        if isinstance(outstream, ColorizedDiffStream):
            wrapped_stream = _StreamWrapper(outstream)
            print_commit(commit, decode=decode, outstream=wrapped_stream)
            # Write diff directly to the ColorizedDiffStream as bytes
            write_tree_diff(
                outstream,
                r.object_store,
                commit.parents[0] if commit.parents else None,
                commit.tree,
            )
        else:
            print_commit(commit, decode=decode, outstream=outstream)
            if commit.parents:
                parent_commit = r[commit.parents[0]]
                assert isinstance(parent_commit, Commit)
                base_tree = parent_commit.tree
            else:
                base_tree = None
            # Traditional path: buffer diff and write as decoded text
            diffstream = BytesIO()
            write_tree_diff(diffstream, r.object_store, base_tree, commit.tree)
            diffstream.seek(0)
            outstream.write(commit_decode(commit, diffstream.getvalue()))


def show_tree(
    repo: RepoPath,
    tree: Tree,
    decode: Callable[[bytes], str],
    outstream: TextIO = sys.stdout,
) -> None:
    """Print a tree to a stream.

    Args:
      repo: A `Repo` object
      tree: A `Tree` object
      decode: Function for decoding bytes to unicode string
      outstream: Stream to write to
    """
    for n in tree:
        outstream.write(decode(n) + "\n")


def show_tag(
    repo: RepoPath,
    tag: Tag,
    decode: Callable[[bytes], str],
    outstream: TextIO = sys.stdout,
) -> None:
    """Print a tag to a stream.

    Args:
      repo: A `Repo` object
      tag: A `Tag` object
      decode: Function for decoding bytes to unicode string
      outstream: Stream to write to
    """
    with open_repo_closing(repo) as r:
        print_tag(tag, decode, outstream)
        obj = r[tag.object[1]]
        assert isinstance(obj, (Tree, Blob, Commit, Tag))
        show_object(repo, obj, decode, outstream)


def show_object(
    repo: RepoPath,
    obj: Tree | Blob | Commit | Tag,
    decode: Callable[[bytes], str],
    outstream: TextIO,
) -> None:
    """Show details of a git object."""
    handlers: dict[bytes, Callable[[RepoPath, Any, Any, TextIO], None]] = {
        b"tree": show_tree,
        b"blob": show_blob,
        b"commit": show_commit,
        b"tag": show_tag,
    }
    handler = handlers.get(obj.type_name)
    if handler is None:
        raise ValueError(f"Unknown object type: {obj.type_name.decode()}")
    handler(repo, obj, decode, outstream)


def print_name_status(changes: Iterator[TreeChange]) -> Iterator[str]:
    """Print a simple status summary, listing changed files."""
    for change in changes:
        if not change:
            continue
        if isinstance(change, list):
            change = change[0]
        if change.type == CHANGE_ADD:
            assert change.new is not None
            path1 = change.new.path
            assert path1 is not None
            path2 = b""
            kind = "A"
        elif change.type == CHANGE_DELETE:
            assert change.old is not None
            path1 = change.old.path
            assert path1 is not None
            path2 = b""
            kind = "D"
        elif change.type == CHANGE_MODIFY:
            assert change.new is not None
            path1 = change.new.path
            assert path1 is not None
            path2 = b""
            kind = "M"
        elif change.type in RENAME_CHANGE_TYPES:
            assert change.old is not None and change.new is not None
            path1 = change.old.path
            assert path1 is not None
            path2_opt = change.new.path
            assert path2_opt is not None
            path2 = path2_opt
            if change.type == CHANGE_RENAME:
                kind = "R"
            elif change.type == CHANGE_COPY:
                kind = "C"
        path1_str = (
            path1.decode("utf-8", errors="replace")
            if isinstance(path1, bytes)
            else path1
        )
        path2_str = (
            path2.decode("utf-8", errors="replace")
            if isinstance(path2, bytes)
            else path2
        )
        yield f"{kind:<8}{path1_str:<20}{path2_str:<20}"


def log(
    repo: RepoPath = ".",
    paths: Sequence[str | bytes] | None = None,
    outstream: TextIO = sys.stdout,
    max_entries: int | None = None,
    reverse: bool = False,
    name_status: bool = False,
) -> None:
    """Write commit logs.

    Args:
      repo: Path to repository
      paths: Optional set of specific paths to print entries for
      outstream: Stream to write log output to
      reverse: Reverse order in which entries are printed
      name_status: Print name status
      max_entries: Optional maximum number of entries to display
    """
    with open_repo_closing(repo) as r:
        try:
            include = [r.head()]
        except KeyError:
            include = []
        # Convert paths to bytes if needed
        paths_bytes = None
        if paths:
            paths_bytes = [p.encode() if isinstance(p, str) else p for p in paths]

        walker = r.get_walker(
            include=include, max_entries=max_entries, paths=paths_bytes, reverse=reverse
        )
        for entry in walker:

            def decode_wrapper(x: bytes) -> str:
                return commit_decode(entry.commit, x)

            print_commit(entry.commit, decode_wrapper, outstream)
            if name_status:
                outstream.writelines(
                    [
                        line + "\n"
                        for line in print_name_status(
                            cast(Iterator[TreeChange], entry.changes())
                        )
                    ]
                )


# TODO(jelmer): better default for encoding?
def show(
    repo: RepoPath = ".",
    objects: Sequence[str | bytes] | None = None,
    outstream: TextIO = sys.stdout,
    default_encoding: str = DEFAULT_ENCODING,
) -> None:
    """Print the changes in a commit.

    Args:
      repo: Path to repository
      objects: Objects to show (defaults to [HEAD])
      outstream: Stream to write to
      default_encoding: Default encoding to use if none is set in the
        commit
    """
    if objects is None:
        objects = ["HEAD"]
    if isinstance(objects, (str, bytes)):
        objects = [objects]
    with open_repo_closing(repo) as r:
        for objectish in objects:
            o = parse_object(r, objectish)
            if isinstance(o, Commit):

                def decode(x: bytes) -> str:
                    return commit_decode(o, x, default_encoding)

            else:

                def decode(x: bytes) -> str:
                    return x.decode(default_encoding)

            assert isinstance(o, (Tree, Blob, Commit, Tag))
            show_object(r, o, decode, outstream)


def diff_tree(
    repo: RepoPath,
    old_tree: str | bytes | Tree,
    new_tree: str | bytes | Tree,
    outstream: BinaryIO = default_bytes_out_stream,
) -> None:
    """Compares the content and mode of blobs found via two tree objects.

    Args:
      repo: Path to repository
      old_tree: Id of old tree
      new_tree: Id of new tree
      outstream: Stream to write to
    """
    with open_repo_closing(repo) as r:
        if isinstance(old_tree, Tree):
            old_tree_id: ObjectID | None = old_tree.id
        elif isinstance(old_tree, str):
            old_tree_id = ObjectID(old_tree.encode())
        else:
            old_tree_id = ObjectID(old_tree)

        if isinstance(new_tree, Tree):
            new_tree_id: ObjectID | None = new_tree.id
        elif isinstance(new_tree, str):
            new_tree_id = ObjectID(new_tree.encode())
        else:
            new_tree_id = ObjectID(new_tree)

        write_tree_diff(outstream, r.object_store, old_tree_id, new_tree_id)


def diff(
    repo: RepoPath = ".",
    commit: str | bytes | Commit | None = None,
    commit2: str | bytes | Commit | None = None,
    staged: bool = False,
    paths: Sequence[str | bytes] | None = None,
    outstream: BinaryIO = default_bytes_out_stream,
    diff_algorithm: str | None = None,
) -> None:
    """Show diff.

    Args:
      repo: Path to repository
      commit: First commit to compare. If staged is True, compare
              index to this commit. If staged is False, compare working tree
              to this commit. If None, defaults to HEAD for staged and index
              for unstaged.
      commit2: Second commit to compare against first commit. If provided,
               show diff between commit and commit2 (ignoring staged flag).
      staged: If True, show staged changes (index vs commit).
              If False, show unstaged changes (working tree vs commit/index).
              Ignored if commit2 is provided.
      paths: Optional list of paths to limit diff
      outstream: Stream to write to
      diff_algorithm: Algorithm to use for diffing ("myers" or "patience"),
                      defaults to the underlying function's default if None
    """
    from . import diff as diff_module

    with open_repo_closing(repo) as r:
        # Normalize paths to bytes
        byte_paths: list[bytes] | None = None
        if paths is not None and paths:  # Check if paths is not empty
            byte_paths = []
            for p in paths:
                if isinstance(p, str):
                    byte_paths.append(p.encode("utf-8"))
                else:
                    byte_paths.append(p)
        elif paths == []:  # Convert empty list to None
            byte_paths = None
        else:
            byte_paths = None

        # Resolve commit refs to SHAs if provided
        if commit is not None:
            if isinstance(commit, Commit):
                # Already a Commit object
                commit_sha = commit.id
                commit_obj = commit
            else:
                # parse_commit handles both refs and SHAs, and always returns a Commit object
                commit_obj = parse_commit(r, commit)
                commit_sha = commit_obj.id
        else:
            commit_sha = None
            commit_obj = None

        if commit2 is not None:
            # Compare two commits
            if isinstance(commit2, Commit):
                commit2_obj = commit2
            else:
                commit2_obj = parse_commit(r, commit2)

            # Get trees from commits
            old_tree = commit_obj.tree if commit_obj else None
            new_tree = commit2_obj.tree

            # Use tree_changes to get the changes and apply path filtering
            changes = r.object_store.tree_changes(old_tree, new_tree)
            for (oldpath, newpath), (oldmode, newmode), (oldsha, newsha) in changes:
                # Skip if paths are specified and this change doesn't match
                if byte_paths:
                    path_to_check = newpath or oldpath
                    assert path_to_check is not None
                    if not any(
                        path_to_check == p or path_to_check.startswith(p + b"/")
                        for p in byte_paths
                    ):
                        continue

                write_object_diff(
                    outstream,
                    r.object_store,
                    (oldpath, oldmode, oldsha),
                    (newpath, newmode, newsha),
                    diff_algorithm=diff_algorithm,
                )
        elif staged:
            # Show staged changes (index vs commit)
            diff_module.diff_index_to_tree(
                r, outstream, commit_sha, byte_paths, diff_algorithm=diff_algorithm
            )
        elif commit is not None:
            # Compare working tree to a specific commit
            assert (
                commit_sha is not None
            )  # mypy: commit_sha is set when commit is not None
            diff_module.diff_working_tree_to_tree(
                r, outstream, commit_sha, byte_paths, diff_algorithm=diff_algorithm
            )
        else:
            # Compare working tree to index
            diff_module.diff_working_tree_to_index(
                r, outstream, byte_paths, diff_algorithm=diff_algorithm
            )


def rev_list(
    repo: RepoPath,
    commits: Sequence[str | bytes],
    outstream: BinaryIO = default_bytes_out_stream,
) -> None:
    """Lists commit objects in reverse chronological order.

    Args:
      repo: Path to repository
      commits: Commits over which to iterate
      outstream: Stream to write to
    """
    with open_repo_closing(repo) as r:
        for entry in r.get_walker(
            include=[r[c if isinstance(c, bytes) else c.encode()].id for c in commits]
        ):
            outstream.write(entry.commit.id + b"\n")


def _canonical_part(url: str) -> str:
    name = url.rsplit("/", 1)[-1]
    if name.endswith(".git"):
        name = name[:-4]
    return name


def submodule_add(
    repo: str | os.PathLike[str] | Repo,
    url: str,
    path: str | os.PathLike[str] | None = None,
    name: str | None = None,
) -> None:
    """Add a new submodule.

    Args:
      repo: Path to repository
      url: URL of repository to add as submodule
      path: Path where submodule should live
      name: Name for the submodule
    """
    with open_repo_closing(repo) as r:
        if path is None:
            path = os.path.relpath(_canonical_part(url), r.path)
        if name is None:
            name = os.fsdecode(path) if path is not None else None

        if name is None:
            raise Error("Submodule name must be specified or derivable from path")

        # TODO(jelmer): Move this logic to dulwich.submodule
        gitmodules_path = os.path.join(r.path, ".gitmodules")
        try:
            config = ConfigFile.from_path(gitmodules_path)
        except FileNotFoundError:
            config = ConfigFile()
            config.path = gitmodules_path
        config.set(("submodule", name), "url", url)
        config.set(("submodule", name), "path", os.fsdecode(path))
        config.write_to_path()


def submodule_init(repo: str | os.PathLike[str] | Repo) -> None:
    """Initialize submodules.

    Args:
      repo: Path to repository
    """
    with open_repo_closing(repo) as r:
        config = r.get_config()
        gitmodules_path = os.path.join(r.path, ".gitmodules")
        for path, url, name in read_submodules(gitmodules_path):
            config.set((b"submodule", name), b"active", True)
            config.set((b"submodule", name), b"url", url)
        config.write_to_path()


def submodule_list(repo: RepoPath) -> Iterator[tuple[str, str]]:
    """List submodules.

    Args:
      repo: Path to repository
    """
    from .submodule import iter_cached_submodules

    with open_repo_closing(repo) as r:
        head_commit = r[r.head()]
        assert isinstance(head_commit, Commit)
        for path, sha in iter_cached_submodules(r.object_store, head_commit.tree):
            yield path.decode(DEFAULT_ENCODING), sha.decode(DEFAULT_ENCODING)


def submodule_update(
    repo: str | os.PathLike[str] | Repo,
    paths: Sequence[str | bytes | os.PathLike[str]] | None = None,
    init: bool = False,
    force: bool = False,
    recursive: bool = False,
    errstream: BinaryIO | None = None,
) -> None:
    """Update submodules.

    Args:
      repo: Path to repository
      paths: Optional list of specific submodule paths to update. If None, updates all.
      init: If True, initialize submodules first
      force: Force update even if local changes exist
      recursive: If True, recursively update nested submodules
      errstream: Error stream for error messages
    """
    from .submodule import iter_cached_submodules

    with open_repo_closing(repo) as r:
        if init:
            submodule_init(r)

        config = r.get_config()
        gitmodules_path = os.path.join(r.path, ".gitmodules")

        # Get list of submodules to update
        submodules_to_update = []
        head_commit = r[r.head()]
        assert isinstance(head_commit, Commit)
        for path, sha in iter_cached_submodules(r.object_store, head_commit.tree):
            path_str = (
                path.decode(DEFAULT_ENCODING) if isinstance(path, bytes) else path
            )
            if paths is None or path_str in paths:
                submodules_to_update.append((path, sha))

        # Read submodule configuration
        for path, target_sha in submodules_to_update:
            path_str = (
                path.decode(DEFAULT_ENCODING) if isinstance(path, bytes) else path
            )

            # Find the submodule name from .gitmodules
            submodule_name: bytes | None = None
            for sm_path, sm_url, sm_name in read_submodules(gitmodules_path):
                if sm_path == path:
                    submodule_name = sm_name
                    break

            if not submodule_name:
                continue

            # Get the URL from config
            section = (
                b"submodule",
                submodule_name
                if isinstance(submodule_name, bytes)
                else submodule_name.encode(),
            )
            try:
                url_value = config.get(section, b"url")
                if isinstance(url_value, bytes):
                    url = url_value.decode(DEFAULT_ENCODING)
                else:
                    url = url_value
            except KeyError:
                # URL not in config, skip this submodule
                continue

            # Get or create the submodule repository paths
            submodule_path = os.path.join(r.path, path_str)
            submodule_git_dir = os.path.join(r.controldir(), "modules", path_str)

            # Clone or fetch the submodule
            if not os.path.exists(submodule_git_dir):
                # Clone the submodule as bare repository
                os.makedirs(os.path.dirname(submodule_git_dir), exist_ok=True)

                # Clone to the git directory
                sub_repo = clone(url, submodule_git_dir, bare=True, checkout=False)
                sub_repo.close()

                # Create the submodule directory if it doesn't exist
                if not os.path.exists(submodule_path):
                    os.makedirs(submodule_path)

                # Create .git file in the submodule directory
                relative_git_dir = os.path.relpath(submodule_git_dir, submodule_path)
                git_file_path = os.path.join(submodule_path, ".git")
                with open(git_file_path, "w") as f:
                    f.write(f"gitdir: {relative_git_dir}\n")

                # Set up working directory configuration
                with open_repo_closing(submodule_git_dir) as sub_repo:
                    sub_config = sub_repo.get_config()
                    sub_config.set(
                        (b"core",),
                        b"worktree",
                        os.path.abspath(submodule_path).encode(),
                    )
                    sub_config.write_to_path()

                    # Checkout the target commit
                    sub_repo.refs[HEADREF] = target_sha

                    # Build the index and checkout files
                    tree = sub_repo[target_sha]
                    if hasattr(tree, "tree"):  # If it's a commit, get the tree
                        tree_id = tree.tree
                    else:
                        tree_id = target_sha

                    build_index_from_tree(
                        submodule_path,
                        sub_repo.index_path(),
                        sub_repo.object_store,
                        tree_id,
                    )
            else:
                # Fetch and checkout in existing submodule
                with open_repo_closing(submodule_git_dir) as sub_repo:
                    # Fetch from remote
                    client, path_segments = get_transport_and_path(url)
                    client.fetch(path_segments.encode(), sub_repo)

                    # Update to the target commit
                    sub_repo.refs[HEADREF] = target_sha

                    # Reset the working directory
                    reset(sub_repo, "hard", target_sha)

            # Recursively update nested submodules if requested
            if recursive:
                submodule_gitmodules = os.path.join(submodule_path, ".gitmodules")
                if os.path.exists(submodule_gitmodules):
                    submodule_update(
                        submodule_path,
                        paths=None,
                        init=True,  # Always initialize nested submodules
                        force=force,
                        recursive=True,
                        errstream=errstream,
                    )


def tag_create(
    repo: RepoPath,
    tag: str | bytes,
    author: str | bytes | None = None,
    message: str | bytes | None = None,
    annotated: bool = False,
    objectish: str | bytes = "HEAD",
    tag_time: int | None = None,
    tag_timezone: int | None = None,
    sign: bool | None = None,
    encoding: str = DEFAULT_ENCODING,
) -> None:
    """Creates a tag in git via dulwich calls.

    Args:
      repo: Path to repository
      tag: tag string
      author: tag author (optional, if annotated is set)
      message: tag message (optional)
      annotated: whether to create an annotated tag
      objectish: object the tag should point at, defaults to HEAD
      tag_time: Optional time for annotated tag
      tag_timezone: Optional timezone for annotated tag
      sign: GPG Sign the tag (bool, defaults to False,
        pass True to use default GPG key,
        pass a str containing Key ID to use a specific GPG key)
      encoding: Encoding to use for tag messages
    """
    with open_repo_closing(repo) as r:
        object = parse_object(r, objectish)

        if isinstance(tag, str):
            tag = tag.encode(encoding)

        if annotated:
            # Create the tag object
            tag_obj = Tag()
            if author is None:
                author = get_user_identity(r.get_config_stack())
            elif isinstance(author, str):
                author = author.encode(encoding)
            else:
                assert isinstance(author, bytes)
            tag_obj.tagger = author
            if isinstance(message, str):
                message = message.encode(encoding)
            elif isinstance(message, bytes):
                pass
            else:
                message = b""
            tag_obj.message = message + "\n".encode(encoding)
            tag_obj.name = tag
            tag_obj.object = (type(object), object.id)
            if tag_time is None:
                tag_time = int(time.time())
            tag_obj.tag_time = tag_time
            if tag_timezone is None:
                tag_timezone = get_user_timezones()[1]
            elif isinstance(tag_timezone, str):
                tag_timezone = parse_timezone(tag_timezone.encode())
            tag_obj.tag_timezone = tag_timezone

            # Check if we should sign the tag
            config = r.get_config_stack()

            if sign is None:
                # Check tag.gpgSign configuration when sign is not explicitly set
                try:
                    should_sign = config.get_boolean(
                        (b"tag",), b"gpgsign", default=False
                    )
                except KeyError:
                    should_sign = False  # Default to not signing if no config
            else:
                should_sign = sign

            # Get the signing key from config if signing is enabled
            keyid = None
            if should_sign:
                try:
                    keyid_bytes = config.get((b"user",), b"signingkey")
                    keyid = keyid_bytes.decode() if keyid_bytes else None
                except KeyError:
                    keyid = None
                tag_obj.sign(keyid)

            r.object_store.add_object(tag_obj)
            tag_id = tag_obj.id
        else:
            tag_id = object.id

        r.refs[_make_tag_ref(tag)] = tag_id


def verify_commit(
    repo: RepoPath,
    committish: str | bytes = "HEAD",
    keyids: list[str] | None = None,
) -> None:
    """Verify GPG signature on a commit.

    Args:
      repo: Path to repository
      committish: Commit to verify (defaults to HEAD)
      keyids: Optional list of trusted key IDs. If provided, the commit
        must be signed by one of these keys. If not provided, just verifies
        that the commit has a valid signature.

    Raises:
      gpg.errors.BadSignatures: if GPG signature verification fails
      gpg.errors.MissingSignatures: if commit was not signed by a key
        specified in keyids
    """
    with open_repo_closing(repo) as r:
        commit = parse_commit(r, committish)
        commit.verify(keyids)


def verify_tag(
    repo: RepoPath,
    tagname: str | bytes,
    keyids: list[str] | None = None,
) -> None:
    """Verify GPG signature on a tag.

    Args:
      repo: Path to repository
      tagname: Name of tag to verify
      keyids: Optional list of trusted key IDs. If provided, the tag
        must be signed by one of these keys. If not provided, just verifies
        that the tag has a valid signature.

    Raises:
      gpg.errors.BadSignatures: if GPG signature verification fails
      gpg.errors.MissingSignatures: if tag was not signed by a key
        specified in keyids
    """
    with open_repo_closing(repo) as r:
        if isinstance(tagname, str):
            tagname = tagname.encode()
        tag_ref = _make_tag_ref(tagname)
        tag_id = r.refs[tag_ref]
        tag_obj = r[tag_id]
        if not isinstance(tag_obj, Tag):
            raise Error(f"{tagname!r} does not point to a tag object")
        tag_obj.verify(keyids)


def tag_list(repo: RepoPath, outstream: TextIO = sys.stdout) -> list[Ref]:
    """List all tags.

    Args:
      repo: Path to repository
      outstream: Stream to write tags to
    """
    with open_repo_closing(repo) as r:
        tags: list[Ref] = sorted(r.refs.as_dict(Ref(b"refs/tags")))
        return tags


def tag_delete(repo: RepoPath, name: str | bytes) -> None:
    """Remove a tag.

    Args:
      repo: Path to repository
      name: Name of tag to remove
    """
    with open_repo_closing(repo) as r:
        if isinstance(name, bytes):
            names = [name]
        elif isinstance(name, list):
            names = name
        else:
            raise Error(f"Unexpected tag name type {name!r}")
        for name in names:
            del r.refs[_make_tag_ref(name)]


def _make_notes_ref(name: bytes) -> bytes:
    """Make a notes ref name."""
    if name.startswith(b"refs/notes/"):
        return name
    return LOCAL_NOTES_PREFIX + name


def notes_add(
    repo: RepoPath,
    object_sha: bytes,
    note: bytes,
    ref: bytes = b"commits",
    author: bytes | None = None,
    committer: bytes | None = None,
    message: bytes | None = None,
) -> bytes:
    """Add or update a note for an object.

    Args:
      repo: Path to repository
      object_sha: SHA of the object to annotate
      note: Note content
      ref: Notes ref to use (defaults to "commits" for refs/notes/commits)
      author: Author identity (defaults to committer)
      committer: Committer identity (defaults to config)
      message: Commit message for the notes update

    Returns:
      SHA of the new notes commit
    """
    with open_repo_closing(repo) as r:
        # Parse the object to get its SHA
        obj = parse_object(r, object_sha)
        object_sha = obj.id

        if isinstance(note, str):
            note = note.encode(DEFAULT_ENCODING)
        if isinstance(ref, str):
            ref = ref.encode(DEFAULT_ENCODING)

        notes_ref = _make_notes_ref(ref)
        config = r.get_config_stack()

        return r.notes.set_note(
            object_sha,
            note,
            notes_ref,
            author=author,
            committer=committer,
            message=message,
            config=config,
        )


def notes_remove(
    repo: RepoPath,
    object_sha: bytes,
    ref: bytes = b"commits",
    author: bytes | None = None,
    committer: bytes | None = None,
    message: bytes | None = None,
) -> bytes | None:
    """Remove a note for an object.

    Args:
      repo: Path to repository
      object_sha: SHA of the object to remove notes from
      ref: Notes ref to use (defaults to "commits" for refs/notes/commits)
      author: Author identity (defaults to committer)
      committer: Committer identity (defaults to config)
      message: Commit message for the notes removal

    Returns:
      SHA of the new notes commit, or None if no note existed
    """
    with open_repo_closing(repo) as r:
        # Parse the object to get its SHA
        obj = parse_object(r, object_sha)
        object_sha = obj.id

        if isinstance(ref, str):
            ref = ref.encode(DEFAULT_ENCODING)

        notes_ref = _make_notes_ref(ref)
        config = r.get_config_stack()

        return r.notes.remove_note(
            object_sha,
            notes_ref,
            author=author,
            committer=committer,
            message=message,
            config=config,
        )


def notes_show(
    repo: str | os.PathLike[str] | Repo, object_sha: bytes, ref: bytes = b"commits"
) -> bytes | None:
    """Show the note for an object.

    Args:
      repo: Path to repository
      object_sha: SHA of the object
      ref: Notes ref to use (defaults to "commits" for refs/notes/commits)

    Returns:
      Note content as bytes, or None if no note exists
    """
    with open_repo_closing(repo) as r:
        # Parse the object to get its SHA
        obj = parse_object(r, object_sha)
        object_sha = obj.id

        if isinstance(ref, str):
            ref = ref.encode(DEFAULT_ENCODING)

        notes_ref = _make_notes_ref(ref)
        config = r.get_config_stack()

        return r.notes.get_note(object_sha, notes_ref, config=config)


def notes_list(repo: RepoPath, ref: bytes = b"commits") -> list[tuple[ObjectID, bytes]]:
    """List all notes in a notes ref.

    Args:
      repo: Path to repository
      ref: Notes ref to use (defaults to "commits" for refs/notes/commits)

    Returns:
      List of tuples of (object_sha, note_content)
    """
    with open_repo_closing(repo) as r:
        if isinstance(ref, str):
            ref = ref.encode(DEFAULT_ENCODING)

        notes_ref = _make_notes_ref(ref)
        config = r.get_config_stack()

        return r.notes.list_notes(notes_ref, config=config)


def replace_list(repo: RepoPath) -> list[tuple[ObjectID, ObjectID]]:
    """List all replacement refs.

    Args:
      repo: Path to repository

    Returns:
      List of tuples of (object_sha, replacement_sha) where object_sha is the
      object being replaced and replacement_sha is what it's replaced with
    """
    with open_repo_closing(repo) as r:
        replacements: list[tuple[ObjectID, ObjectID]] = []
        for ref in r.refs.keys():
            if ref.startswith(LOCAL_REPLACE_PREFIX):
                object_sha = ObjectID(ref[len(LOCAL_REPLACE_PREFIX) :])
                replacement_sha = r.refs[ref]
                replacements.append((object_sha, replacement_sha))
        return replacements


def replace_delete(repo: RepoPath, object_sha: ObjectID | str) -> None:
    """Delete a replacement ref.

    Args:
      repo: Path to repository
      object_sha: SHA of the object whose replacement should be removed
    """
    with open_repo_closing(repo) as r:
        # Convert to ObjectID if string
        if isinstance(object_sha, str):
            object_sha_id = ObjectID(object_sha.encode("ascii"))
        else:
            object_sha_id = object_sha

        replace_ref = _make_replace_ref(object_sha_id)
        if replace_ref not in r.refs:
            raise KeyError(
                f"No replacement ref found for {object_sha_id.decode('ascii')}"
            )
        del r.refs[replace_ref]


def replace_create(
    repo: RepoPath,
    object_sha: str | ObjectID,
    replacement_sha: str | ObjectID,
) -> None:
    """Create a replacement ref to replace one object with another.

    Args:
      repo: Path to repository
      object_sha: SHA of the object to replace
      replacement_sha: SHA of the replacement object
    """
    with open_repo_closing(repo) as r:
        # Convert to ObjectID if string
        if isinstance(object_sha, str):
            object_sha_id = ObjectID(object_sha.encode("ascii"))
        else:
            object_sha_id = object_sha

        if isinstance(replacement_sha, str):
            replacement_sha_id = ObjectID(replacement_sha.encode("ascii"))
        else:
            replacement_sha_id = replacement_sha

        # Create the replacement ref
        replace_ref = _make_replace_ref(object_sha_id)
        r.refs[replace_ref] = replacement_sha_id


def reset(
    repo: str | os.PathLike[str] | Repo,
    mode: str,
    treeish: str | bytes | Commit | Tree | Tag = "HEAD",
) -> None:
    """Reset current HEAD to the specified state.

    Args:
      repo: Path to repository
      mode: Mode ("hard", "soft", "mixed")
      treeish: Treeish to reset to
    """
    with open_repo_closing(repo) as r:
        # Parse the target tree
        tree = parse_tree(r, treeish)
        # Only parse as commit if treeish is not a Tree object
        if isinstance(treeish, Tree):
            # For Tree objects, we can't determine the commit, skip updating HEAD
            target_commit = None
        else:
            target_commit = parse_commit(r, treeish)

        # Update HEAD to point to the target commit
        if target_commit is not None:
            # Get the current HEAD value for set_if_equals
            try:
                old_head = r.refs[HEADREF]
            except KeyError:
                old_head = None

            # Create reflog message
            treeish_str = (
                treeish.decode("utf-8")
                if isinstance(treeish, bytes)
                else str(treeish)
                if not isinstance(treeish, (Commit, Tree, Tag))
                else target_commit.id.hex()
            )
            default_message = f"reset: moving to {treeish_str}".encode()
            reflog_message = _get_reflog_message(default_message)

            # Update HEAD with reflog message
            r.refs.set_if_equals(
                HEADREF, old_head, target_commit.id, message=reflog_message
            )

        if mode == "soft":
            # Soft reset: only update HEAD, leave index and working tree unchanged
            return

        elif mode == "mixed":
            # Mixed reset: update HEAD and index, but leave working tree unchanged
            from .object_store import iter_tree_contents

            # Open the index
            index = r.open_index()

            # Clear the current index
            index.clear()

            # Populate index from the target tree
            for entry in iter_tree_contents(r.object_store, tree.id):
                # Create an IndexEntry from the tree entry
                # Use zeros for filesystem-specific fields since we're not touching the working tree
                assert (
                    entry.mode is not None
                    and entry.sha is not None
                    and entry.path is not None
                )
                index_entry = IndexEntry(
                    ctime=(0, 0),
                    mtime=(0, 0),
                    dev=0,
                    ino=0,
                    mode=entry.mode,
                    uid=0,
                    gid=0,
                    size=0,  # Size will be 0 since we're not reading from disk
                    sha=entry.sha,
                    flags=0,
                )
                index[entry.path] = index_entry

            # Write the updated index
            index.write()

        elif mode == "hard":
            # Hard reset: update HEAD, index, and working tree
            # For reset --hard, use current index tree as old tree to get proper deletions
            index = r.open_index()
            if len(index) > 0:
                index_tree_id = index.commit(r.object_store)
            else:
                # Empty index
                index_tree_id = None

            # Get configuration for working tree updates
            honor_filemode, validate_path_element, symlink_fn = (
                _get_worktree_update_config(r)
            )

            blob_normalizer = r.get_blob_normalizer()
            changes = tree_changes(
                r.object_store, index_tree_id, tree.id, want_unchanged=True
            )
            update_working_tree(
                r,
                index_tree_id,
                tree.id,
                change_iterator=changes,
                honor_filemode=honor_filemode,
                validate_path_element=validate_path_element,
                symlink_fn=symlink_fn,
                force_remove_untracked=True,
                blob_normalizer=blob_normalizer,
                allow_overwrite_modified=True,  # Allow overwriting modified files
            )
        else:
            raise Error(f"Invalid reset mode: {mode}")


def get_remote_repo(
    repo: Repo, remote_location: str | bytes | None = None
) -> tuple[str | None, str]:
    """Get the remote repository information.

    Args:
      repo: Local repository object
      remote_location: Optional remote name or URL; defaults to branch remote

    Returns:
      Tuple of (remote_name, remote_url) where remote_name may be None
      if remote_location is a URL rather than a configured remote
    """
    config = repo.get_config()
    if remote_location is None:
        remote_location = get_branch_remote(repo)
    if isinstance(remote_location, str):
        encoded_location = remote_location.encode()
    else:
        encoded_location = remote_location

    section = (b"remote", encoded_location)

    remote_name: str | None = None

    if config.has_section(section):
        remote_name = encoded_location.decode()
        encoded_location = config.get(section, "url")
    else:
        remote_name = None

    return (remote_name, encoded_location.decode())


def push(
    repo: RepoPath,
    remote_location: str | bytes | None = None,
    refspecs: str | bytes | Sequence[str | bytes] | None = None,
    outstream: BinaryIO = default_bytes_out_stream,
    errstream: BinaryIO | RawIOBase = default_bytes_err_stream,
    force: bool = False,
    **kwargs: object,
) -> SendPackResult:
    """Remote push with dulwich via dulwich.client.

    Args:
      repo: Path to repository
      remote_location: Location of the remote
      refspecs: Refs to push to remote
      outstream: A stream file to write output
      errstream: A stream file to write errors
      force: Force overwriting refs
      **kwargs: Additional keyword arguments for the client
    """
    # Open the repo
    with open_repo_closing(repo) as r:
        (remote_name, remote_location) = get_remote_repo(r, remote_location)
        # Check if mirror mode is enabled
        mirror_mode = False
        if remote_name:
            try:
                mirror_mode_val = r.get_config_stack().get_boolean(
                    (b"remote", remote_name.encode()), b"mirror"
                )
                if mirror_mode_val is not None:
                    mirror_mode = mirror_mode_val
            except KeyError:
                pass

        if mirror_mode:
            # Mirror mode: push all refs and delete non-existent ones
            refspecs = []
            for ref in r.refs.keys():
                # Push all refs to the same name on remote
                refspecs.append(ref + b":" + ref)
        elif refspecs is None:
            refspecs = [active_branch(r)]

        # Normalize refspecs to bytes
        if isinstance(refspecs, str):
            refspecs_bytes: bytes | list[bytes] = refspecs.encode()
        elif isinstance(refspecs, bytes):
            refspecs_bytes = refspecs
        else:
            refspecs_bytes = []
            for spec in refspecs:
                if isinstance(spec, str):
                    refspecs_bytes.append(spec.encode())
                else:
                    refspecs_bytes.append(spec)

        # Get the client and path
        transport_kwargs = _filter_transport_kwargs(**kwargs)
        client, path = get_transport_and_path(
            remote_location,
            config=r.get_config_stack(),
            **transport_kwargs,
        )

        selected_refs = []
        remote_changed_refs: dict[Ref, ObjectID | None] = {}

        def update_refs(refs: dict[Ref, ObjectID]) -> dict[Ref, ObjectID]:
            remote_refs = DictRefsContainer(refs)  # type: ignore[arg-type]
            selected_refs.extend(
                parse_reftuples(r.refs, remote_refs, refspecs_bytes, force=force)
            )
            new_refs: dict[Ref, ObjectID] = {}

            # In mirror mode, delete remote refs that don't exist locally
            if mirror_mode:
                local_refs = set(r.refs.keys())
                for remote_ref in refs.keys():
                    if remote_ref not in local_refs:
                        new_refs[remote_ref] = ZERO_SHA
                        remote_changed_refs[remote_ref] = None
            # TODO: Handle selected_refs == {None: None}
            for lh, rh, force_ref in selected_refs:
                if lh is None:
                    assert rh is not None
                    new_refs[rh] = ZERO_SHA
                    remote_changed_refs[rh] = None
                else:
                    try:
                        localsha = r.refs[lh]
                    except KeyError as exc:
                        raise Error(
                            f"No valid ref {lh.decode() if isinstance(lh, bytes) else lh} in local repository"
                        ) from exc
                    assert rh is not None
                    if not force_ref and rh in refs:
                        check_diverged(r, refs[rh], localsha)
                    new_refs[rh] = localsha
                    remote_changed_refs[rh] = localsha
            return new_refs

        err_encoding = getattr(errstream, "encoding", None) or DEFAULT_ENCODING
        remote_location = client.get_url(path)
        try:

            def generate_pack_data_wrapper(
                have: AbstractSet[ObjectID],
                want: AbstractSet[ObjectID],
                *,
                ofs_delta: bool = False,
                progress: Callable[..., None] | None = None,
            ) -> tuple[int, Iterator[UnpackedObject]]:
                # Wrap to match the expected signature
                # Convert AbstractSet to set since generate_pack_data expects set
                return r.generate_pack_data(
                    set(have), set(want), progress=progress, ofs_delta=ofs_delta
                )

            result = client.send_pack(
                path.encode(),
                update_refs,
                generate_pack_data=generate_pack_data_wrapper,
                progress=lambda data: (errstream.write(data), None)[1],
            )
        except SendPackError as exc:
            raise Error(
                "Push to " + remote_location + " failed -> " + exc.args[0].decode(),
            ) from exc
        else:
            errstream.write(
                b"Push to " + remote_location.encode(err_encoding) + b" successful.\n"
            )

        for ref, error in (result.ref_status or {}).items():  # type: ignore[assignment]
            if error is not None:
                errstream.write(
                    f"Push of ref {ref.decode('utf-8', 'replace')} failed: {error}\n".encode(
                        err_encoding
                    )
                )
            else:
                errstream.write(
                    f"Ref {ref.decode('utf-8', 'replace')} updated\n".encode()
                )

        if remote_name is not None:
            _import_remote_refs(r.refs, remote_name, remote_changed_refs)

        return result

    # Trigger auto GC if needed
    from .gc import maybe_auto_gc

    with open_repo_closing(repo) as r:
        maybe_auto_gc(r)


def pull(
    repo: RepoPath,
    remote_location: str | bytes | None = None,
    refspecs: str | bytes | Sequence[str | bytes] | None = None,
    outstream: BinaryIO = default_bytes_out_stream,
    errstream: BinaryIO | RawIOBase = default_bytes_err_stream,
    fast_forward: bool = True,
    ff_only: bool = False,
    force: bool = False,
    filter_spec: str | None = None,
    protocol_version: int | None = None,
    **kwargs: object,
) -> None:
    """Pull from remote via dulwich.client.

    Args:
      repo: Path to repository
      remote_location: Location of the remote
      refspecs: refspecs to fetch. Can be a bytestring, a string, or a list of
        bytestring/string.
      outstream: A stream file to write to output
      errstream: A stream file to write to errors
      fast_forward: If True, raise an exception when fast-forward is not possible
      ff_only: If True, only allow fast-forward merges. Raises DivergedBranches
        when branches have diverged rather than performing a merge.
      force: If True, allow overwriting local changes in the working tree.
        If False, pull will abort if it would overwrite uncommitted changes.
      filter_spec: A git-rev-list-style object filter spec, as an ASCII string.
        Only used if the server supports the Git protocol-v2 'filter'
        feature, and ignored otherwise.
      protocol_version: desired Git protocol version. By default the highest
        mutually supported protocol version will be used
      **kwargs: Additional keyword arguments for the client
    """
    # Open the repo
    with open_repo_closing(repo) as r:
        (remote_name, remote_location) = get_remote_repo(r, remote_location)

        selected_refs = []

        if refspecs is None:
            refspecs_normalized: bytes | list[bytes] = [b"HEAD"]
        elif isinstance(refspecs, str):
            refspecs_normalized = refspecs.encode()
        elif isinstance(refspecs, bytes):
            refspecs_normalized = refspecs
        else:
            refspecs_normalized = []
            for spec in refspecs:
                if isinstance(spec, str):
                    refspecs_normalized.append(spec.encode())
                else:
                    refspecs_normalized.append(spec)

        def determine_wants(
            remote_refs: dict[Ref, ObjectID], depth: int | None = None
        ) -> list[ObjectID]:
            remote_refs_container = DictRefsContainer(remote_refs)  # type: ignore[arg-type]
            selected_refs.extend(
                parse_reftuples(
                    remote_refs_container, r.refs, refspecs_normalized, force=force
                )
            )
            return [
                remote_refs[lh]
                for (lh, rh, force_ref) in selected_refs
                if lh is not None
                and lh in remote_refs
                and remote_refs[lh] not in r.object_store
            ]

        transport_kwargs = _filter_transport_kwargs(**kwargs)
        client, path = get_transport_and_path(
            remote_location,
            config=r.get_config_stack(),
            **transport_kwargs,
        )
        if filter_spec:
            filter_spec_bytes: bytes | None = filter_spec.encode("ascii")
        else:
            filter_spec_bytes = None

        def progress(data: bytes) -> None:
            errstream.write(data)

        fetch_result = client.fetch(
            path.encode(),
            r,
            progress=progress,
            determine_wants=determine_wants,  # type: ignore[arg-type]  # Function matches protocol but mypy can't verify
            filter_spec=filter_spec_bytes,
            protocol_version=protocol_version,
        )

        # Store the old HEAD tree before making changes
        try:
            old_head = r.refs[HEADREF]
            old_commit = r[old_head]
            assert isinstance(old_commit, Commit)
            old_tree_id = old_commit.tree
        except KeyError:
            old_tree_id = None

        merged = False
        for lh, rh, force_ref in selected_refs:
            if not force_ref and rh is not None and rh in r.refs:
                try:
                    assert lh is not None
                    followed_ref = r.refs.follow(rh)[1]
                    assert followed_ref is not None
                    lh_ref = fetch_result.refs[lh]
                    assert lh_ref is not None
                    check_diverged(r, followed_ref, lh_ref)
                except DivergedBranches as exc:
                    if ff_only or fast_forward:
                        raise
                    else:
                        # Perform merge
                        assert lh is not None
                        merge_ref = fetch_result.refs[lh]
                        assert merge_ref is not None
                        _merge_result, conflicts = _do_merge(r, merge_ref)
                        if conflicts:
                            raise Error(
                                f"Merge conflicts occurred: {conflicts}"
                            ) from exc
                        merged = True
                        # Skip updating ref since merge already updated HEAD
                        continue
            if rh is not None and lh is not None:
                lh_value = fetch_result.refs[lh]
                if lh_value is not None:
                    r.refs[Ref(rh)] = lh_value

        # Only update HEAD if we didn't perform a merge
        if selected_refs and not merged:
            lh, rh, _ = selected_refs[0]
            if lh is not None:
                ref_value = fetch_result.refs[lh]
                if ref_value is not None:
                    r[b"HEAD"] = ref_value

        # Update working tree to match the new HEAD
        # Skip if merge was performed as merge already updates the working tree
        if not merged and old_tree_id is not None:
            head_commit = r[b"HEAD"]
            assert isinstance(head_commit, Commit)
            new_tree_id = head_commit.tree
            blob_normalizer = r.get_blob_normalizer()
            changes = tree_changes(r.object_store, old_tree_id, new_tree_id)
            update_working_tree(
                r,
                old_tree_id,
                new_tree_id,
                change_iterator=changes,
                blob_normalizer=blob_normalizer,
                allow_overwrite_modified=force,
            )
        if remote_name is not None:
            _import_remote_refs(r.refs, remote_name, fetch_result.refs)

    # Trigger auto GC if needed
    from .gc import maybe_auto_gc

    with open_repo_closing(repo) as r:
        maybe_auto_gc(r)


def status(
    repo: str | os.PathLike[str] | Repo = ".",
    ignored: bool = False,
    untracked_files: str = "normal",
) -> GitStatus:
    """Returns staged, unstaged, and untracked changes relative to the HEAD.

    Args:
      repo: Path to repository or repository object
      ignored: Whether to include ignored files in untracked
      untracked_files: How to handle untracked files, defaults to "all":
          "no": do not return untracked files
          "normal": return untracked directories, not their contents
          "all": include all files in untracked directories
        Using untracked_files="no" can be faster than "all" when the worktree
          contains many untracked files/directories.
        Using untracked_files="normal" provides a good balance, only showing
          directories that are entirely untracked without listing all their contents.

    Returns: GitStatus tuple,
        staged -  dict with lists of staged paths (diff index/HEAD)
        unstaged -  list of unstaged paths (diff index/working-tree)
        untracked - list of untracked, un-ignored & non-.git paths
    """
    with open_repo_closing(repo) as r:
        # Open the index once and reuse it for both staged and unstaged checks
        index = r.open_index()
        # 1. Get status of staged
        tracked_changes = get_tree_changes(r, index)
        # 2. Get status of unstaged
        normalizer = r.get_blob_normalizer()

        # Create a wrapper that handles the bytes -> Blob conversion
        if normalizer is not None:

            def filter_callback(data: bytes, path: bytes) -> bytes:
                blob = Blob()
                blob.data = data
                normalized_blob = normalizer.checkin_normalize(blob, path)
                result_data: bytes = normalized_blob.data
                return result_data
        else:
            filter_callback = None

        # Check if core.preloadIndex is enabled
        config = r.get_config_stack()
        preload_index = config.get_boolean(b"core", b"preloadIndex", False)

        unstaged_changes = list(
            get_unstaged_changes(index, r.path, filter_callback, preload_index)
        )

        untracked_paths = get_untracked_paths(
            r.path,
            r.path,
            index,
            exclude_ignored=not ignored,
            untracked_files=untracked_files,
        )
        if sys.platform == "win32":
            untracked_changes = [
                path.replace(os.path.sep, "/") for path in untracked_paths
            ]
        else:
            untracked_changes = list(untracked_paths)

        return GitStatus(tracked_changes, unstaged_changes, untracked_changes)


def shortlog(
    repo: str | os.PathLike[str] | Repo,
    summary_only: bool = False,
    sort_by_commits: bool = False,
) -> list[dict[str, str]]:
    """Summarize commits by author, like git shortlog.

    Args:
        repo: Path to repository or Repo object.
        summary_only: If True, only show counts per author.
        sort_by_commits: If True, sort authors by number of commits.

    Returns:
        A list where each item is a dict containing:

            - "author": the author's name as a string
            - "messages": all commit messages concatenated into a single string
    """
    with open_repo_closing(repo) as r:
        walker = r.get_walker()
        authors: dict[str, list[str]] = {}

        for entry in walker:
            commit = entry.commit

            author = commit.author.decode(commit.encoding or "utf-8")
            message = commit.message.decode(commit.encoding or "utf-8").strip()

            authors.setdefault(author, []).append(message)

        # Convert messages to single string per author
        items: list[dict[str, str]] = [
            {"author": author, "messages": "\n".join(msgs)}
            for author, msgs in authors.items()
        ]

        if sort_by_commits:
            # Sort by number of commits (lines in messages)
            items.sort(key=lambda x: len(x["messages"].splitlines()), reverse=True)

        return items


def _walk_working_dir_paths(
    frompath: str | bytes | os.PathLike[str],
    basepath: str | bytes | os.PathLike[str],
    prune_dirnames: Callable[[str, list[str]], list[str]] | None = None,
) -> Iterator[tuple[str | bytes, bool]]:
    """Get path, is_dir for files in working dir from frompath.

    Args:
      frompath: Path to begin walk
      basepath: Path to compare to
      prune_dirnames: Optional callback to prune dirnames during os.walk
        dirnames will be set to result of prune_dirnames(dirpath, dirnames)
    """
    # Convert paths to strings for os.walk compatibility

    for dirpath, dirnames, filenames in os.walk(frompath):  # type: ignore[type-var]
        # Skip .git and below.
        if ".git" in dirnames:
            dirnames.remove(".git")
            if dirpath != basepath:
                continue

        if ".git" in filenames:
            filenames.remove(".git")
            if dirpath != basepath:
                continue

        if dirpath != frompath:
            yield dirpath, True  # type: ignore[misc]

        for filename in filenames:
            filepath = os.path.join(dirpath, filename)  # type: ignore[call-overload]
            yield filepath, False

        if prune_dirnames:
            dirnames[:] = prune_dirnames(dirpath, dirnames)  # type: ignore[arg-type]


def get_untracked_paths(
    frompath: str | bytes | os.PathLike[str],
    basepath: str | bytes | os.PathLike[str],
    index: Index,
    exclude_ignored: bool = False,
    untracked_files: str = "all",
) -> Iterator[str]:
    """Get untracked paths.

    Args:
      frompath: Path to walk
      basepath: Path to compare to
      index: Index to check against
      exclude_ignored: Whether to exclude ignored paths
      untracked_files: How to handle untracked files:
        - "no": return an empty list
        - "all": return all files in untracked directories
        - "normal": return untracked directories without listing their contents

    Note: ignored directories will never be walked for performance reasons.
      If exclude_ignored is False, only the path to an ignored directory will
      be yielded, no files inside the directory will be returned
    """
    if untracked_files not in ("no", "all", "normal"):
        raise ValueError("untracked_files must be one of (no, all, normal)")

    if untracked_files == "no":
        return

    # Normalize paths to str
    frompath_str = os.fsdecode(os.fspath(frompath))
    basepath_str = os.fsdecode(os.fspath(basepath))

    with open_repo_closing(basepath_str) as r:
        ignore_manager = IgnoreFilterManager.from_repo(r)

    ignored_dirs = []
    # List to store untracked directories found during traversal
    untracked_dir_list = []

    def directory_has_non_ignored_files(dir_path: str, base_rel_path: str) -> bool:
        """Recursively check if directory contains any non-ignored files."""
        try:
            for entry in os.listdir(dir_path):
                entry_path = os.path.join(dir_path, entry)
                rel_entry = os.path.join(base_rel_path, entry)

                if os.path.isfile(entry_path):
                    if ignore_manager.is_ignored(rel_entry) is not True:
                        return True
                elif os.path.isdir(entry_path):
                    if directory_has_non_ignored_files(entry_path, rel_entry):
                        return True
            return False
        except OSError:
            # If we can't read the directory, assume it has non-ignored files
            return True

    def prune_dirnames(dirpath: str, dirnames: list[str]) -> list[str]:
        for i in range(len(dirnames) - 1, -1, -1):
            path = os.path.join(dirpath, dirnames[i])
            ip = os.path.join(os.path.relpath(path, basepath_str), "")

            # Check if directory is ignored
            if ignore_manager.is_ignored(ip) is True:
                if not exclude_ignored:
                    ignored_dirs.append(
                        os.path.join(os.path.relpath(path, frompath_str), "")
                    )
                del dirnames[i]
                continue

            # For "normal" mode, check if the directory is entirely untracked
            if untracked_files == "normal":
                # Convert directory path to tree path for index lookup
                dir_tree_path = path_to_tree_path(basepath_str, path)

                # Check if any file in this directory is tracked
                dir_prefix = dir_tree_path + b"/" if dir_tree_path else b""
                has_tracked_files = any(name.startswith(dir_prefix) for name in index)

                if not has_tracked_files:
                    # This directory is entirely untracked
                    rel_path_base = os.path.relpath(path, basepath_str)
                    rel_path_from = os.path.join(
                        os.path.relpath(path, frompath_str), ""
                    )

                    # If excluding ignored, check if directory contains any non-ignored files
                    if exclude_ignored:
                        if not directory_has_non_ignored_files(path, rel_path_base):
                            # Directory only contains ignored files, skip it
                            del dirnames[i]
                            continue

                    # Check if it should be excluded due to ignore rules
                    is_ignored = ignore_manager.is_ignored(rel_path_base)
                    if not exclude_ignored or not is_ignored:
                        untracked_dir_list.append(rel_path_from)
                    del dirnames[i]

        return dirnames

    # For "all" mode, use the original behavior
    if untracked_files == "all":
        for ap, is_dir in _walk_working_dir_paths(
            frompath_str, basepath_str, prune_dirnames=prune_dirnames
        ):
            # frompath_str and basepath_str are both str, so ap must be str
            assert isinstance(ap, str)
            if not is_dir:
                ip = path_to_tree_path(basepath_str, ap)
                if ip not in index:
                    if not exclude_ignored or not ignore_manager.is_ignored(
                        os.path.relpath(ap, basepath_str)
                    ):
                        yield os.path.relpath(ap, frompath_str)
    else:  # "normal" mode
        # Walk directories, handling both files and directories
        for ap, is_dir in _walk_working_dir_paths(
            frompath_str, basepath_str, prune_dirnames=prune_dirnames
        ):
            # frompath_str and basepath_str are both str, so ap must be str
            assert isinstance(ap, str)
            # This part won't be reached for pruned directories
            if is_dir:
                # Check if this directory is entirely untracked
                dir_tree_path = path_to_tree_path(basepath_str, ap)
                dir_prefix = dir_tree_path + b"/" if dir_tree_path else b""
                has_tracked_files = any(name.startswith(dir_prefix) for name in index)
                if not has_tracked_files:
                    if not exclude_ignored or not ignore_manager.is_ignored(
                        os.path.relpath(ap, basepath_str)
                    ):
                        yield os.path.join(os.path.relpath(ap, frompath_str), "")
            else:
                # Check individual files in directories that contain tracked files
                ip = path_to_tree_path(basepath_str, ap)
                if ip not in index:
                    if not exclude_ignored or not ignore_manager.is_ignored(
                        os.path.relpath(ap, basepath_str)
                    ):
                        yield os.path.relpath(ap, frompath_str)

        # Yield any untracked directories found during pruning
        yield from untracked_dir_list

    yield from ignored_dirs


def grep(
    repo: RepoPath,
    pattern: str | bytes,
    *,
    outstream: TextIO = sys.stdout,
    rev: str | bytes | None = None,
    pathspecs: Sequence[str | bytes] | None = None,
    ignore_case: bool = False,
    line_number: bool = False,
    max_depth: int | None = None,
    respect_ignores: bool = True,
) -> None:
    """Search for a pattern in tracked files.

    Args:
      repo: Path to repository or Repo object
      pattern: Regular expression pattern to search for
      outstream: Stream to write results to
      rev: Revision to search in (defaults to HEAD)
      pathspecs: Optional list of path patterns to limit search
      ignore_case: Whether to perform case-insensitive matching
      line_number: Whether to output line numbers
      max_depth: Maximum directory depth to search
      respect_ignores: Whether to respect .gitignore patterns
    """
    from .object_store import iter_tree_contents

    # Compile the pattern
    flags = re.IGNORECASE if ignore_case else 0
    try:
        if isinstance(pattern, bytes):
            compiled_pattern = re.compile(pattern, flags)
        else:
            compiled_pattern = re.compile(pattern.encode("utf-8"), flags)
    except re.error as e:
        raise ValueError(f"Invalid regular expression: {e}") from e

    with open_repo_closing(repo) as r:
        # Get the tree to search
        if rev is None:
            try:
                commit = r[b"HEAD"]
                assert isinstance(commit, Commit)
            except KeyError as e:
                raise ValueError("No HEAD commit found") from e
        else:
            rev_bytes = rev if isinstance(rev, bytes) else rev.encode("utf-8")
            commit_obj = parse_commit(r, rev_bytes)
            if commit_obj is None:
                raise ValueError(f"Invalid revision: {rev}")
            commit = commit_obj

        tree = r[commit.tree]
        assert isinstance(tree, Tree)

        # Set up ignore filter if requested
        ignore_manager = None
        if respect_ignores:
            ignore_manager = IgnoreFilterManager.from_repo(r)

        # Convert pathspecs to bytes
        pathspecs_bytes: list[bytes] | None = None
        if pathspecs:
            pathspecs_bytes = [
                p if isinstance(p, bytes) else p.encode("utf-8") for p in pathspecs
            ]

        # Iterate through all files in the tree
        for entry in iter_tree_contents(r.object_store, tree.id):
            path, mode, sha = entry.path, entry.mode, entry.sha
            assert path is not None
            assert mode is not None
            assert sha is not None

            # Skip directories
            if stat.S_ISDIR(mode):
                continue

            # Check max depth
            if max_depth is not None:
                depth = path.count(b"/")
                if depth > max_depth:
                    continue

            # Check pathspecs
            if pathspecs_bytes:
                matches_pathspec = False
                for pathspec in pathspecs_bytes:
                    # Simple prefix matching (could be enhanced with full pathspec support)
                    if path.startswith(pathspec) or fnmatch.fnmatch(
                        path.decode("utf-8", errors="replace"),
                        pathspec.decode("utf-8", errors="replace"),
                    ):
                        matches_pathspec = True
                        break
                if not matches_pathspec:
                    continue

            # Check ignore patterns
            if ignore_manager:
                path_str = path.decode("utf-8", errors="replace")
                if ignore_manager.is_ignored(path_str) is True:
                    continue

            # Get the blob content
            blob = r[sha]
            assert isinstance(blob, Blob)

            # Search for pattern in the blob
            content = blob.data
            lines = content.split(b"\n")

            for line_num, line in enumerate(lines, 1):
                if compiled_pattern.search(line):
                    path_str = path.decode("utf-8", errors="replace")
                    line_str = line.decode("utf-8", errors="replace")
                    if line_number:
                        outstream.write(f"{path_str}:{line_num}:{line_str}\n")
                    else:
                        outstream.write(f"{path_str}:{line_str}\n")


def get_tree_changes(
    repo: RepoPath, index: Index | None = None
) -> dict[str, list[str | bytes]]:
    """Return add/delete/modify changes to tree by comparing index to HEAD.

    Args:
      repo: repo path or object
      index: optional Index object to reuse (avoids re-opening the index)
    Returns: dict with lists for each type of change
    """
    with open_repo_closing(repo) as r:
        if index is None:
            index = r.open_index()

        # Compares the Index to the HEAD & determines changes
        # Iterate through the changes and report add/delete/modify
        # TODO: call out to dulwich.diff_tree somehow.
        tracked_changes: dict[str, list[str | bytes]] = {
            "add": [],
            "delete": [],
            "modify": [],
        }
        try:
            head_commit = r[b"HEAD"]
            assert isinstance(head_commit, Commit)
            tree_id = head_commit.tree
        except KeyError:
            tree_id = None

        for change in index.changes_from_tree(r.object_store, tree_id):
            if not change[0][0]:
                assert change[0][1] is not None
                tracked_changes["add"].append(change[0][1])
            elif not change[0][1]:
                assert change[0][0] is not None
                tracked_changes["delete"].append(change[0][0])
            elif change[0][0] == change[0][1]:
                assert change[0][0] is not None
                tracked_changes["modify"].append(change[0][0])
            else:
                raise NotImplementedError("git mv ops not yet supported")
        return tracked_changes


def daemon(
    path: str | os.PathLike[str] = ".",
    address: str | None = None,
    port: int | None = None,
) -> None:
    """Run a daemon serving Git requests over TCP/IP.

    Args:
      path: Path to the directory to serve.
      address: Optional address to listen on (defaults to ::)
      port: Optional port to listen on (defaults to TCP_GIT_PORT)
    """
    # TODO(jelmer): Support git-daemon-export-ok and --export-all.
    backend = FileSystemBackend(os.fspath(path))
    server = TCPGitServer(backend, address or "localhost", port or 9418)
    server.serve_forever()


def web_daemon(
    path: str | os.PathLike[str] = ".",
    address: str | None = None,
    port: int | None = None,
) -> None:
    """Run a daemon serving Git requests over HTTP.

    Args:
      path: Path to the directory to serve
      address: Optional address to listen on (defaults to ::)
      port: Optional port to listen on (defaults to 80)
    """
    from .web import (
        WSGIRequestHandlerLogger,
        WSGIServerLogger,
        make_server,
        make_wsgi_chain,
    )

    backend = FileSystemBackend(os.fspath(path))
    app = make_wsgi_chain(backend)
    server = make_server(
        address or "::",
        port or 80,
        app,
        handler_class=WSGIRequestHandlerLogger,
        server_class=WSGIServerLogger,
    )
    server.serve_forever()


def upload_pack(
    path: str | os.PathLike[str] = ".",
    inf: BinaryIO | None = None,
    outf: BinaryIO | None = None,
) -> int:
    """Upload a pack file after negotiating its contents using smart protocol.

    Args:
      path: Path to the repository
      inf: Input stream to communicate with client
      outf: Output stream to communicate with client
    """
    if outf is None:
        outf = sys.stdout.buffer
    if inf is None:
        inf = sys.stdin.buffer

    assert outf is not None
    assert inf is not None

    path = os.path.expanduser(path)
    backend = FileSystemBackend(path)

    def send_fn(data: bytes) -> None:
        outf.write(data)
        outf.flush()

    proto = Protocol(inf.read, send_fn)
    handler = UploadPackHandler(backend, [path], proto)
    # FIXME: Catch exceptions and write a single-line summary to outf.
    handler.handle()
    return 0


def receive_pack(
    path: str | os.PathLike[str] = ".",
    inf: BinaryIO | None = None,
    outf: BinaryIO | None = None,
) -> int:
    """Receive a pack file after negotiating its contents using smart protocol.

    Args:
      path: Path to the repository
      inf: Input stream to communicate with client
      outf: Output stream to communicate with client
    """
    if outf is None:
        outf = sys.stdout.buffer
    if inf is None:
        inf = sys.stdin.buffer

    assert outf is not None
    assert inf is not None

    path = os.path.expanduser(path)
    backend = FileSystemBackend(path)

    def send_fn(data: bytes) -> None:
        outf.write(data)
        outf.flush()

    proto = Protocol(inf.read, send_fn)
    handler = ReceivePackHandler(backend, [path], proto)
    # FIXME: Catch exceptions and write a single-line summary to outf.
    handler.handle()
    return 0


def _make_branch_ref(name: str | bytes) -> Ref:
    if isinstance(name, str):
        name = name.encode(DEFAULT_ENCODING)
    return local_branch_name(name)


def _make_tag_ref(name: str | bytes) -> Ref:
    if isinstance(name, str):
        name = name.encode(DEFAULT_ENCODING)
    return local_tag_name(name)


def _make_replace_ref(name: str | bytes | ObjectID) -> Ref:
    if isinstance(name, str):
        name = name.encode(DEFAULT_ENCODING)
    return local_replace_name(name)


def branch_delete(repo: RepoPath, name: str | bytes | Sequence[str | bytes]) -> None:
    """Delete a branch.

    Args:
      repo: Path to the repository
      name: Name of the branch
    """
    with open_repo_closing(repo) as r:
        if isinstance(name, (list, tuple)):
            names = name
        else:
            names = [name]
        for branch_name in names:
            del r.refs[_make_branch_ref(branch_name)]


def branch_create(
    repo: str | os.PathLike[str] | Repo,
    name: str | bytes,
    objectish: str | bytes | None = None,
    force: bool = False,
) -> None:
    """Create a branch.

    Args:
      repo: Path to the repository
      name: Name of the new branch
      objectish: Target object to point new branch at (defaults to HEAD)
      force: Force creation of branch, even if it already exists
    """
    with open_repo_closing(repo) as r:
        if objectish is None:
            objectish = "HEAD"

        # Try to expand branch shorthand before parsing
        original_objectish = objectish
        objectish_bytes = (
            objectish.encode(DEFAULT_ENCODING)
            if isinstance(objectish, str)
            else objectish
        )

        if Ref(b"refs/remotes/" + objectish_bytes) in r.refs:
            objectish = b"refs/remotes/" + objectish_bytes
        elif local_branch_name(objectish_bytes) in r.refs:
            objectish = local_branch_name(objectish_bytes)

        object = parse_object(r, objectish)
        refname = _make_branch_ref(name)
        default_message = (
            b"branch: Created from " + original_objectish.encode(DEFAULT_ENCODING)
            if isinstance(original_objectish, str)
            else b"branch: Created from " + original_objectish
        )
        ref_message = _get_reflog_message(default_message)
        if force:
            r.refs.set_if_equals(refname, None, object.id, message=ref_message)
        else:
            if not r.refs.add_if_new(refname, object.id, message=ref_message):
                name_str = name.decode() if isinstance(name, bytes) else name
                raise Error(f"Branch with name {name_str} already exists.")

        # Check if we should set up tracking
        config = r.get_config_stack()
        try:
            auto_setup_merge = config.get((b"branch",), b"autoSetupMerge").decode()
        except KeyError:
            auto_setup_merge = "true"  # Default value

        # Determine if the objectish refers to a remote-tracking branch
        objectish_ref = None
        if original_objectish != "HEAD":
            # Try to resolve objectish as a ref
            objectish_bytes = (
                original_objectish.encode(DEFAULT_ENCODING)
                if isinstance(original_objectish, str)
                else original_objectish
            )

            if Ref(objectish_bytes) in r.refs:
                objectish_ref = objectish_bytes
            elif Ref(b"refs/remotes/" + objectish_bytes) in r.refs:
                objectish_ref = b"refs/remotes/" + objectish_bytes
            elif local_branch_name(objectish_bytes) in r.refs:
                objectish_ref = local_branch_name(objectish_bytes)
        else:
            # HEAD might point to a remote-tracking branch
            head_ref = r.refs.follow(HEADREF)[0][1]
            if head_ref.startswith(b"refs/remotes/"):
                objectish_ref = head_ref

        # Set up tracking if appropriate
        if objectish_ref and (
            (auto_setup_merge == "always")
            or (
                auto_setup_merge == "true"
                and objectish_ref.startswith(b"refs/remotes/")
            )
        ):
            # Extract remote name and branch from the ref
            if objectish_ref.startswith(b"refs/remotes/"):
                parts = objectish_ref[len(b"refs/remotes/") :].split(b"/", 1)
                if len(parts) == 2:
                    remote_name = parts[0]
                    remote_branch = local_branch_name(parts[1])

                    # Set up tracking
                    repo_config = r.get_config()
                    branch_name_bytes = (
                        name.encode(DEFAULT_ENCODING) if isinstance(name, str) else name
                    )
                    repo_config.set(
                        (b"branch", branch_name_bytes), b"remote", remote_name
                    )
                    repo_config.set(
                        (b"branch", branch_name_bytes), b"merge", remote_branch
                    )
                    repo_config.write_to_path()


def filter_branches_by_pattern(branches: Iterable[bytes], pattern: str) -> list[bytes]:
    """Filter branches by fnmatch pattern.

    Args:
        branches: Iterable of branch names as bytes
        pattern: Pattern to match against

    Returns:
        List of filtered branch names
    """
    return [
        branch for branch in branches if fnmatch.fnmatchcase(branch.decode(), pattern)
    ]


def branch_list(repo: RepoPath) -> list[Ref]:
    """List all branches.

    Args:
      repo: Path to the repository
    Returns:
      List of branch names (without refs/heads/ prefix)
    """
    with open_repo_closing(repo) as r:
        branches: list[Ref] = list(r.refs.keys(base=Ref(LOCAL_BRANCH_PREFIX)))

        # Check for branch.sort configuration
        config = r.get_config_stack()
        try:
            sort_key = config.get((b"branch",), b"sort").decode()
        except KeyError:
            # Default is refname (alphabetical)
            sort_key = "refname"

        # Parse sort key
        reverse = False
        if sort_key.startswith("-"):
            reverse = True
            sort_key = sort_key[1:]

        # Apply sorting
        if sort_key == "refname":
            # Simple alphabetical sort (default)
            branches.sort(reverse=reverse)
        elif sort_key in ("committerdate", "authordate"):
            # Sort by date
            def get_commit_date(branch_name: bytes) -> int:
                ref = local_branch_name(branch_name)
                sha = r.refs[ref]
                commit = r.object_store[sha]
                assert isinstance(commit, Commit)
                if sort_key == "committerdate":
                    return cast(int, commit.commit_time)
                else:  # authordate
                    return cast(int, commit.author_time)

            # Sort branches by date
            # Note: Python's sort naturally orders smaller values first (ascending)
            # For dates, this means oldest first by default
            # Use a stable sort with branch name as secondary key for consistent ordering
            if reverse:
                # For reverse sort, we want newest dates first but alphabetical names second
                branches.sort(key=lambda b: (-get_commit_date(b), b))
            else:
                branches.sort(key=lambda b: (get_commit_date(b), b))
        else:
            # Unknown sort key, fall back to default
            branches.sort()

        return branches


def branch_remotes_list(repo: RepoPath) -> list[bytes]:
    """List the short names of all remote branches.

    Args:
      repo: Path to the repository
    Returns:
      List of branch names (without refs/remotes/ prefix, and without remote name; e.g. 'main' from 'origin/main')
    """
    with open_repo_closing(repo) as r:
        branches = [bytes(ref) for ref in r.refs.keys(base=Ref(LOCAL_REMOTE_PREFIX))]

        config = r.get_config_stack()
        try:
            sort_key = config.get((b"branch",), b"sort").decode()
        except KeyError:
            # Default is refname (alphabetical)
            sort_key = "refname"

        # Parse sort key
        reverse = False
        if sort_key.startswith("-"):
            reverse = True
            sort_key = sort_key[1:]

        # Apply sorting
        if sort_key == "refname":
            # Simple alphabetical sort (default)
            branches.sort(reverse=reverse)
        elif sort_key in ("committerdate", "authordate"):
            # Sort by date
            def get_commit_date(branch_name: bytes) -> int:
                ref = LOCAL_REMOTE_PREFIX + branch_name
                sha = r.refs[Ref(ref)]
                commit = r.object_store[sha]
                assert isinstance(commit, Commit)
                if sort_key == "committerdate":
                    return cast(int, commit.commit_time)
                else:  # authordate
                    return cast(int, commit.author_time)

            # Sort branches by date
            # Note: Python's sort naturally orders smaller values first (ascending)
            # For dates, this means oldest first by default
            # Use a stable sort with branch name as secondary key for consistent ordering
            if reverse:
                # For reverse sort, we want newest dates first but alphabetical names second
                branches.sort(key=lambda b: (-get_commit_date(b), b))
            else:
                branches.sort(key=lambda b: (get_commit_date(b), b))
        else:
            # Unknown sort key
            raise ValueError(f"Unknown sort key: {sort_key}")

        return branches


def _get_branch_merge_status(repo: RepoPath) -> Iterator[tuple[bytes, bool]]:
    """Get merge status for all branches relative to current HEAD.

    Args:
        repo: Path to the repository

    Yields:
        tuple of (``branch_name``, ``is_merged``) where:

        - ``branch_name``: Branch name without refs/heads/ prefix
        - ``is_merged``: True if branch is merged into HEAD, False otherwise
    """
    with open_repo_closing(repo) as r:
        current_sha = r.refs[HEADREF]

        for branch_ref, branch_sha in r.refs.as_dict(base=Ref(b"refs/heads/")).items():
            # Check if branch is an ancestor of HEAD (fully merged)
            is_merged = can_fast_forward(r, branch_sha, current_sha)
            yield branch_ref, is_merged


def merged_branches(repo: RepoPath) -> Iterator[bytes]:
    """List branches that have been merged into the current branch.

    Args:
      repo: Path to the repository
    Yields:
      Branch names (without refs/heads/ prefix) that are merged
      into the current HEAD
    """
    for branch_name, is_merged in _get_branch_merge_status(repo):
        if is_merged:
            yield branch_name


def no_merged_branches(repo: RepoPath) -> Iterator[bytes]:
    """List branches that have been merged into the current branch.

    Args:
      repo: Path to the repository
    Yields:
      Branch names (without refs/heads/ prefix) that are merged
      into the current HEAD
    """
    for branch_name, is_merged in _get_branch_merge_status(repo):
        if not is_merged:
            yield branch_name


def branches_containing(repo: RepoPath, commit: str) -> Iterator[bytes]:
    """List branches that contain the specified commit.

    Args:
        repo: Path to the repository
        commit: Commit-ish string (SHA, branch name, tag, etc.)

    Yields:
        Branch names (without refs/heads/ prefix) that contain the commit

    Raises:
        ValueError: If the commit reference is malformed
        KeyError: If the commit reference does not exist
    """
    with open_repo_closing(repo) as r:
        commit_obj = parse_commit(r, commit)
        commit_sha = commit_obj.id

        for branch_ref, branch_sha in r.refs.as_dict(
            base=Ref(LOCAL_BRANCH_PREFIX)
        ).items():
            if can_fast_forward(r, commit_sha, branch_sha):
                yield branch_ref


def active_branch(repo: RepoPath) -> bytes:
    """Return the active branch in the repository, if any.

    Args:
      repo: Repository to open
    Returns:
      branch name
    Raises:
      KeyError: if the repository does not have a working tree
      IndexError: if HEAD is floating
    """
    with open_repo_closing(repo) as r:
        active_ref = r.refs.follow(HEADREF)[0][1]
        if not active_ref.startswith(LOCAL_BRANCH_PREFIX):
            raise ValueError(active_ref)
        return active_ref[len(LOCAL_BRANCH_PREFIX) :]


def get_branch_remote(repo: str | os.PathLike[str] | Repo) -> bytes:
    """Return the active branch's remote name, if any.

    Args:
      repo: Repository to open
    Returns:
      remote name
    Raises:
      KeyError: if the repository does not have a working tree
    """
    with open_repo_closing(repo) as r:
        branch_name = active_branch(r.path)
        config = r.get_config()
        try:
            remote_name = config.get((b"branch", branch_name), b"remote")
        except KeyError:
            remote_name = b"origin"
    return remote_name


def get_branch_merge(repo: RepoPath, branch_name: bytes | None = None) -> bytes:
    """Return the branch's merge reference (upstream branch), if any.

    Args:
      repo: Repository to open
      branch_name: Name of the branch (defaults to active branch)

    Returns:
      merge reference name (e.g. b"refs/heads/main")

    Raises:
      KeyError: if the branch does not have a merge configuration
    """
    with open_repo_closing(repo) as r:
        if branch_name is None:
            branch_name = active_branch(r.path)
        config = r.get_config()
        return config.get((b"branch", branch_name), b"merge")


def set_branch_tracking(
    repo: str | os.PathLike[str] | Repo,
    branch_name: bytes,
    remote_name: bytes,
    remote_ref: bytes,
) -> None:
    """Set up branch tracking configuration.

    Args:
      repo: Repository to open
      branch_name: Name of the local branch
      remote_name: Name of the remote (e.g. b"origin")
      remote_ref: Remote reference to track (e.g. b"refs/heads/main")
    """
    with open_repo_closing(repo) as r:
        config = r.get_config()
        config.set((b"branch", branch_name), b"remote", remote_name)
        config.set((b"branch", branch_name), b"merge", remote_ref)
        config.write_to_path()


def fetch(
    repo: RepoPath,
    remote_location: str | bytes | None = None,
    outstream: TextIO = sys.stdout,
    errstream: BinaryIO | RawIOBase = default_bytes_err_stream,
    message: bytes | None = None,
    depth: int | None = None,
    prune: bool = False,
    prune_tags: bool = False,
    force: bool = False,
    operation: str | None = None,
    thin_packs: bool = True,
    report_activity: Callable[[int, str], None] | None = None,
    quiet: bool = False,
    include_tags: bool = False,
    username: str | None = None,
    password: str | None = None,
    key_filename: str | None = None,
    ssh_command: str | None = None,
    shallow_since: str | None = None,
    shallow_exclude: list[str] | None = None,
) -> FetchPackResult:
    """Fetch objects from a remote server.

    Args:
      repo: Path to the repository
      remote_location: String identifying a remote server
      outstream: Output stream (defaults to stdout)
      errstream: Error stream (defaults to stderr)
      message: Reflog message (defaults to b"fetch: from <remote_name>")
      depth: Depth to fetch at
      prune: Prune remote removed refs
      prune_tags: Prune remote removed tags
      force: Force fetching even if it would overwrite local changes
      operation: Git operation for authentication (e.g., "fetch")
      thin_packs: Whether to use thin packs
      report_activity: Optional callback for reporting transport activity
      quiet: Whether to suppress progress output
      include_tags: Whether to include tags
      username: Username for authentication
      password: Password for authentication
      key_filename: SSH key filename
      ssh_command: SSH command to use
      shallow_since: Deepen or shorten the history to include commits after this date
      shallow_exclude: Deepen or shorten the history to exclude commits reachable from these refs
    Returns:
      Dictionary with refs on the remote
    """
    with open_repo_closing(repo) as r:
        (remote_name, remote_location) = get_remote_repo(r, remote_location)
        default_message = b"fetch: from " + remote_location.encode(DEFAULT_ENCODING)
        message = _get_reflog_message(default_message, message)
        client, path = get_transport_and_path(
            remote_location,
            config=r.get_config_stack(),
            operation=operation,
            thin_packs=thin_packs,
            report_activity=report_activity,
            quiet=quiet,
            include_tags=include_tags,
            username=username,
            password=password,
            key_filename=key_filename,
            ssh_command=ssh_command,
        )

        def progress(data: bytes) -> None:
            errstream.write(data)

        fetch_result = client.fetch(
            path.encode(),
            r,
            progress=progress,
            depth=depth,
            shallow_since=shallow_since,
            shallow_exclude=shallow_exclude,
        )
        if remote_name is not None:
            _import_remote_refs(
                r.refs,
                remote_name,
                fetch_result.refs,
                message,
                prune=prune,
                prune_tags=prune_tags,
            )

    # Trigger auto GC if needed
    from .gc import maybe_auto_gc

    with open_repo_closing(repo) as r:
        maybe_auto_gc(r)

    return fetch_result


def for_each_ref(
    repo: Repo | str = ".",
    pattern: str | bytes | None = None,
) -> list[tuple[bytes, bytes, bytes]]:
    """Iterate over all refs that match the (optional) pattern.

    Args:
      repo: Path to the repository
      pattern: Optional glob (7) patterns to filter the refs with
    Returns: List of bytes tuples with: (sha, object_type, ref_name)
    """
    if isinstance(pattern, str):
        pattern = os.fsencode(pattern)

    with open_repo_closing(repo) as r:
        refs = r.get_refs()

    if pattern:
        matching_refs: dict[Ref, ObjectID] = {}
        pattern_parts = pattern.split(b"/")
        for ref, sha in refs.items():
            matches = False

            # git for-each-ref uses glob (7) style patterns, but fnmatch
            # is greedy and also matches slashes, unlike glob.glob.
            # We have to check parts of the pattern individually.
            # See https://github.com/python/cpython/issues/72904
            ref_parts = ref.split(b"/")
            if len(ref_parts) > len(pattern_parts):
                continue

            for pat, ref_part in zip(pattern_parts, ref_parts):
                matches = fnmatch.fnmatchcase(ref_part, pat)
                if not matches:
                    break

            if matches:
                matching_refs[ref] = sha

        refs = matching_refs

    ret: list[tuple[bytes, bytes, bytes]] = [
        (sha, r.get_object(sha).type_name, ref)
        for ref, sha in sorted(
            refs.items(),
            key=lambda ref_sha: ref_sha[0],
        )
        if ref != b"HEAD"
    ]

    return ret


def show_ref(
    repo: Repo | str = ".",
    patterns: list[str | bytes] | None = None,
    head: bool = False,
    branches: bool = False,
    tags: bool = False,
    dereference: bool = False,
    verify: bool = False,
) -> list[tuple[bytes, bytes]]:
    """List references in a local repository.

    Args:
      repo: Path to the repository
      patterns: Optional list of patterns to filter refs (matched from the end)
      head: Show the HEAD reference
      branches: Limit to local branches (refs/heads/)
      tags: Limit to local tags (refs/tags/)
      dereference: Dereference tags into object IDs
      verify: Enable stricter reference checking (exact path match)
    Returns: List of tuples with (sha, ref_name) or (sha, ref_name^{}) for dereferenced tags
    """
    # Convert string patterns to bytes
    byte_patterns: list[bytes] | None = None
    if patterns:
        byte_patterns = [os.fsencode(p) if isinstance(p, str) else p for p in patterns]

    with open_repo_closing(repo) as r:
        refs = r.get_refs()

        # Filter by branches/tags if specified
        if branches or tags:
            prefixes = []
            if branches:
                prefixes.append(LOCAL_BRANCH_PREFIX)
            if tags:
                prefixes.append(LOCAL_TAG_PREFIX)
            filtered_refs = filter_ref_prefix(refs, prefixes)
        else:
            # By default, show tags, heads, and remote refs (but not HEAD)
            filtered_refs = filter_ref_prefix(refs, [b"refs/"])

        # Add HEAD if requested
        if head and HEADREF in refs:
            filtered_refs[HEADREF] = refs[HEADREF]

        # Filter by patterns if specified
        if byte_patterns:
            matching_refs: dict[Ref, ObjectID] = {}
            for ref, sha in filtered_refs.items():
                for pattern in byte_patterns:
                    if verify:
                        # Verify mode requires exact match
                        if ref == pattern:
                            matching_refs[ref] = sha
                            break
                    else:
                        # Pattern matching from the end of the full name
                        # Only complete parts are matched
                        # E.g., "master" matches "refs/heads/master" but not "refs/heads/mymaster"
                        pattern_parts = pattern.split(b"/")
                        ref_parts = ref.split(b"/")

                        # Try to match from the end
                        if len(pattern_parts) <= len(ref_parts):
                            # Check if the end of ref matches the pattern
                            matches = True
                            for i in range(len(pattern_parts)):
                                if (
                                    ref_parts[-(len(pattern_parts) - i)]
                                    != pattern_parts[i]
                                ):
                                    matches = False
                                    break
                            if matches:
                                matching_refs[ref] = sha
                                break
            filtered_refs = matching_refs

        # Sort by ref name
        sorted_refs = sorted(filtered_refs.items(), key=lambda x: x[0])

        # Build result list
        result: list[tuple[bytes, bytes]] = []
        for ref, sha in sorted_refs:
            result.append((sha, ref))

            # Dereference tags if requested
            if dereference and ref.startswith(LOCAL_TAG_PREFIX):
                try:
                    obj = r.get_object(sha)
                    # Peel tag objects to get the underlying commit/object
                    while obj.type_name == b"tag":
                        assert isinstance(obj, Tag)
                        _obj_class, sha = obj.object
                        obj = r.get_object(sha)
                    result.append((sha, ref + b"^{}"))
                except KeyError:
                    # Object not found, skip dereferencing
                    pass

    return result


def show_branch(
    repo: Repo | str = ".",
    branches: list[str | bytes] | None = None,
    all_branches: bool = False,
    remotes: bool = False,
    current: bool = False,
    topo_order: bool = False,
    more: int | None = None,
    list_branches: bool = False,
    independent_branches: bool = False,
    merge_base: bool = False,
) -> list[str]:
    """Display branches and their commits.

    Args:
      repo: Path to the repository
      branches: List of specific branches to show (default: all local branches)
      all_branches: Show both local and remote branches
      remotes: Show only remote branches
      current: Include current branch if not specified
      topo_order: Show in topological order instead of chronological
      more: Show N more commits beyond common ancestor (negative to show only headers)
      list_branches: Synonym for more=-1 (show only branch headers)
      independent_branches: Show only branches not reachable from others
      merge_base: Show merge bases instead of commit list

    Returns:
      List of output lines
    """
    from .graph import find_octopus_base, independent

    output_lines: list[str] = []

    with open_repo_closing(repo) as r:
        refs = r.get_refs()

        # Determine which branches to show
        branch_refs: dict[Ref, ObjectID] = {}

        if branches:
            # Specific branches requested
            for branch in branches:
                branch_bytes = (
                    os.fsencode(branch) if isinstance(branch, str) else branch
                )
                # Try as full ref name first
                branch_ref_check = Ref(branch_bytes)
                if branch_ref_check in refs:
                    branch_refs[branch_ref_check] = refs[branch_ref_check]
                else:
                    # Try as branch name
                    branch_ref = local_branch_name(branch_bytes)
                    if branch_ref in refs:
                        branch_refs[branch_ref] = refs[branch_ref]
                    # Try as remote branch
                    else:
                        remote_ref = Ref(LOCAL_REMOTE_PREFIX + branch_bytes)
                        if remote_ref in refs:
                            branch_refs[remote_ref] = refs[remote_ref]
        else:
            # Default behavior: show local branches
            if all_branches:
                # Show both local and remote branches
                branch_refs = filter_ref_prefix(
                    refs, [LOCAL_BRANCH_PREFIX, LOCAL_REMOTE_PREFIX]
                )
            elif remotes:
                # Show only remote branches
                branch_refs = filter_ref_prefix(refs, [LOCAL_REMOTE_PREFIX])
            else:
                # Show only local branches
                branch_refs = filter_ref_prefix(refs, [LOCAL_BRANCH_PREFIX])

        # Add current branch if requested and not already included
        if current:
            try:
                head_refs, _ = r.refs.follow(HEADREF)
                if head_refs:
                    head_ref = head_refs[0]
                    if head_ref not in branch_refs and head_ref in refs:
                        branch_refs[head_ref] = refs[head_ref]
            except (KeyError, TypeError):
                # HEAD doesn't point to a branch or doesn't exist
                pass

        if not branch_refs:
            return output_lines

        # Sort branches for consistent output
        sorted_branches = sorted(branch_refs.items(), key=lambda x: x[0])
        branch_sha_list: list[ObjectID] = [sha for _, sha in sorted_branches]

        # Handle --independent flag
        if independent_branches:
            independent_shas = independent(r, branch_sha_list)
            for ref_name, sha in sorted_branches:
                if sha in independent_shas:
                    ref_str = os.fsdecode(shorten_ref_name(ref_name))
                    output_lines.append(ref_str)
            return output_lines

        # Handle --merge-base flag
        if merge_base:
            if len(branch_sha_list) < 2:
                # Need at least 2 branches for merge base
                return output_lines

            merge_bases = find_octopus_base(r, branch_sha_list)
            for sha in merge_bases:
                output_lines.append(sha.decode("ascii"))
            return output_lines

        # Get current branch for marking
        current_branch: bytes | None = None
        try:
            head_refs, _ = r.refs.follow(HEADREF)
            if head_refs:
                current_branch = head_refs[0]
        except (KeyError, TypeError):
            pass

        # Collect commit information for each branch
        branch_commits: list[tuple[bytes, str]] = []  # (sha, message)
        for ref_name, sha in sorted_branches:
            try:
                commit = r[sha]
                if hasattr(commit, "message"):
                    message = commit.message.decode("utf-8", errors="replace").split(
                        "\n"
                    )[0]
                else:
                    message = ""
                branch_commits.append((sha, message))
            except KeyError:
                branch_commits.append((sha, ""))

        # Handle --list flag (show only branch headers)
        if list_branches or (more is not None and more < 0):
            # Just show the branch headers
            for i, (ref_name, sha) in enumerate(sorted_branches):
                is_current = ref_name == current_branch
                marker = "*" if is_current else "!"
                # Create spacing for alignment
                prefix = " " * i + marker + " " * (len(sorted_branches) - i - 1)
                ref_str = os.fsdecode(shorten_ref_name(ref_name))
                _, message = branch_commits[i]
                output_lines.append(f"{prefix}[{ref_str}] {message}")
            return output_lines

        # Build commit history for visualization
        # Collect all commits reachable from any branch
        all_commits: dict[
            bytes, tuple[int, list[bytes], str]
        ] = {}  # sha -> (timestamp, parents, message)

        def collect_commits(sha: bytes, branch_idx: int, visited: set[bytes]) -> None:
            """Recursively collect commits."""
            if sha in visited:
                return
            visited.add(sha)

            try:
                commit = r[sha]
                if not hasattr(commit, "commit_time"):
                    return

                timestamp = commit.commit_time
                parents = commit.parents if hasattr(commit, "parents") else []
                message = (
                    commit.message.decode("utf-8", errors="replace").split("\n")[0]
                    if hasattr(commit, "message")
                    else ""
                )

                if sha not in all_commits:
                    all_commits[sha] = (timestamp, parents, message)

                # Recurse to parents
                for parent in parents:
                    collect_commits(parent, branch_idx, visited)
            except KeyError:
                # Commit not found, stop traversal
                pass

        # Collect commits from all branches
        for i, (_, sha) in enumerate(sorted_branches):
            collect_commits(sha, i, set())

        # Find common ancestor
        common_ancestor_sha = None
        if len(branch_sha_list) >= 2:
            try:
                merge_bases = find_octopus_base(r, branch_sha_list)
                if merge_bases:
                    common_ancestor_sha = merge_bases[0]
            except (KeyError, IndexError):
                pass

        # Sort commits (chronological by default, or topological if requested)
        if topo_order:
            # Topological sort is more complex, for now use chronological
            # TODO: Implement proper topological ordering
            sorted_commits = sorted(all_commits.items(), key=lambda x: -x[1][0])
        else:
            # Reverse chronological order (newest first)
            sorted_commits = sorted(all_commits.items(), key=lambda x: -x[1][0])

        # Determine how many commits to show
        if more is not None:
            # Find index of common ancestor
            if common_ancestor_sha and common_ancestor_sha in all_commits:
                ancestor_idx = next(
                    (
                        i
                        for i, (sha, _) in enumerate(sorted_commits)
                        if sha == common_ancestor_sha
                    ),
                    None,
                )
                if ancestor_idx is not None:
                    # Show commits up to ancestor + more
                    sorted_commits = sorted_commits[: ancestor_idx + 1 + more]

        # Determine which branches contain which commits
        branch_contains: list[set[bytes]] = []
        for ref_name, sha in sorted_branches:
            reachable = set()

            def mark_reachable(commit_sha: bytes) -> None:
                if commit_sha in reachable:
                    return
                reachable.add(commit_sha)
                if commit_sha in all_commits:
                    _, parents, _ = all_commits[commit_sha]
                    for parent in parents:
                        mark_reachable(parent)

            mark_reachable(sha)
            branch_contains.append(reachable)

        # Output branch headers
        for i, (ref_name, sha) in enumerate(sorted_branches):
            is_current = ref_name == current_branch
            marker = "*" if is_current else "!"
            # Create spacing for alignment
            prefix = " " * i + marker + " " * (len(sorted_branches) - i - 1)
            ref_str = os.fsdecode(shorten_ref_name(ref_name))
            _, message = branch_commits[i]
            output_lines.append(f"{prefix}[{ref_str}] {message}")

        # Output separator
        output_lines.append("-" * (len(sorted_branches) + 2))

        # Output commits
        for commit_sha, (_, _, message) in sorted_commits:
            # Build marker string
            markers = []
            for i, (ref_name, branch_sha) in enumerate(sorted_branches):
                if commit_sha == branch_sha:
                    # This is the tip of the branch
                    markers.append("*")
                elif commit_sha in branch_contains[i]:
                    # This commit is in the branch
                    markers.append("+")
                else:
                    # This commit is not in the branch
                    markers.append(" ")

            marker_str = "".join(markers)
            output_lines.append(f"{marker_str} [{message}]")

            # Limit output to 26 branches (git show-branch limitation)
            if len(sorted_branches) > 26:
                break

    return output_lines


def ls_remote(
    remote: str | bytes,
    config: Config | None = None,
    operation: str | None = None,
    thin_packs: bool = True,
    report_activity: Callable[[int, str], None] | None = None,
    quiet: bool = False,
    include_tags: bool = False,
    username: str | None = None,
    password: str | None = None,
    key_filename: str | None = None,
    ssh_command: str | None = None,
) -> LsRemoteResult:
    """List the refs in a remote.

    Args:
      remote: Remote repository location
      config: Configuration to use
      operation: Operation type
      thin_packs: Whether to use thin packs
      report_activity: Function to report activity
      quiet: Whether to suppress output
      include_tags: Whether to include tags
      username: Username for authentication
      password: Password for authentication
      key_filename: SSH key filename
      ssh_command: SSH command to use
    Returns:
      LsRemoteResult object with refs and symrefs
    """
    if config is None:
        config = StackedConfig.default()
    remote_str = remote.decode() if isinstance(remote, bytes) else remote
    client, host_path = get_transport_and_path(
        remote_str,
        config=config,
        operation=operation,
        thin_packs=thin_packs,
        report_activity=report_activity,
        quiet=quiet,
        include_tags=include_tags,
        username=username,
        password=password,
        key_filename=key_filename,
        ssh_command=ssh_command,
    )
    return client.get_refs(
        host_path.encode() if isinstance(host_path, str) else host_path
    )


def repack(repo: RepoPath, write_bitmaps: bool = False) -> None:
    """Repack loose files in a repository.

    Currently this only packs loose objects.

    Args:
      repo: Path to the repository
      write_bitmaps: Whether to write bitmap indexes for packs
    """
    with open_repo_closing(repo) as r:
        r.object_store.pack_loose_objects()
        if write_bitmaps:
            # Update pack cache to pick up newly created packs
            r.object_store._update_pack_cache()
            r.object_store.generate_pack_bitmaps(r.refs.as_dict())


def pack_objects(
    repo: RepoPath,
    object_ids: Sequence[ObjectID],
    packf: BinaryIO,
    idxf: BinaryIO | None,
    delta_window_size: int | None = None,
    deltify: bool | None = None,
    reuse_deltas: bool = True,
    pack_index_version: int | None = None,
) -> None:
    """Pack objects into a file.

    Args:
      repo: Path to the repository
      object_ids: List of object ids to write
      packf: File-like object to write to
      idxf: File-like object to write to (can be None)
      delta_window_size: Sliding window size for searching for deltas;
                         Set to None for default window size.
      deltify: Whether to deltify objects
      reuse_deltas: Allow reuse of existing deltas while deltifying
      pack_index_version: Pack index version to use (1, 2, or 3). If None, uses default version.
    """
    with open_repo_closing(repo) as r:
        entries, data_sum = write_pack_from_container(
            packf.write,
            r.object_store,
            [(oid, None) for oid in object_ids],
            deltify=deltify,
            delta_window_size=delta_window_size,
            reuse_deltas=reuse_deltas,
        )
    if idxf is not None:
        index_entries = sorted([(k, v[0], v[1]) for (k, v) in entries.items()])
        write_pack_index(idxf, index_entries, data_sum, version=pack_index_version)


def ls_tree(
    repo: RepoPath,
    treeish: str | bytes | Commit | Tree | Tag = b"HEAD",
    outstream: TextIO | BinaryIO = sys.stdout,
    recursive: bool = False,
    name_only: bool = False,
) -> None:
    """List contents of a tree.

    Args:
      repo: Path to the repository
      treeish: Tree id to list
      outstream: Output stream (defaults to stdout)
      recursive: Whether to recursively list files
      name_only: Only print item name
    """

    def list_tree(store: BaseObjectStore, treeid: ObjectID, base: bytes) -> None:
        tree = store[treeid]
        assert isinstance(tree, Tree)
        for name, mode, sha in tree.iteritems():
            assert name is not None
            assert mode is not None
            assert sha is not None
            if base:
                name = posixpath.join(base, name)
            if name_only:
                if isinstance(outstream, BinaryIO):
                    outstream.write(name + b"\n")
                else:
                    outstream.write(name.decode("utf-8", "replace") + "\n")
            else:
                formatted = pretty_format_tree_entry(name, mode, sha)
                if isinstance(outstream, BinaryIO):
                    outstream.write(formatted.encode("utf-8"))
                else:
                    outstream.write(formatted)
            if stat.S_ISDIR(mode) and recursive:
                list_tree(store, sha, name)

    with open_repo_closing(repo) as r:
        tree = parse_tree(r, treeish)
        list_tree(r.object_store, tree.id, b"")


def remote_add(
    repo: RepoPath,
    name: bytes | str,
    url: bytes | str,
) -> None:
    """Add a remote.

    Args:
      repo: Path to the repository
      name: Remote name
      url: Remote URL
    """
    if not isinstance(name, bytes):
        name = name.encode(DEFAULT_ENCODING)
    if not isinstance(url, bytes):
        url = url.encode(DEFAULT_ENCODING)
    with open_repo_closing(repo) as r:
        c = r.get_config()
        section = (b"remote", name)
        if c.has_section(section):
            raise RemoteExists(f"Remote {name.decode()} already exists")
        c.set(section, b"url", url)
        c.write_to_path()


def remote_remove(repo: Repo, name: bytes | str) -> None:
    """Remove a remote.

    Args:
      repo: Path to the repository
      name: Remote name
    """
    if not isinstance(name, bytes):
        name = name.encode(DEFAULT_ENCODING)
    with open_repo_closing(repo) as r:
        c = r.get_config()
        section = (b"remote", name)
        del c[section]
        c.write_to_path()


def _quote_path(path: str) -> str:
    """Quote a path using C-style quoting similar to git's core.quotePath.

    Args:
        path: Path to quote

    Returns:
        Quoted path string
    """
    # Check if path needs quoting (non-ASCII or special characters)
    needs_quoting = False
    for char in path:
        if ord(char) > 127 or char in '"\\':
            needs_quoting = True
            break

    if not needs_quoting:
        return path

    # Apply C-style quoting
    quoted = '"'
    for char in path:
        if ord(char) > 127:
            # Non-ASCII character, encode as octal escape
            utf8_bytes = char.encode("utf-8")
            for byte in utf8_bytes:
                quoted += f"\\{byte:03o}"
        elif char == '"':
            quoted += '\\"'
        elif char == "\\":
            quoted += "\\\\"
        else:
            quoted += char
    quoted += '"'
    return quoted


def check_ignore(
    repo: RepoPath,
    paths: Sequence[str | bytes | os.PathLike[str]],
    no_index: bool = False,
    quote_path: bool = True,
) -> Iterator[str]:
    r"""Debug gitignore files.

    Args:
      repo: Path to the repository
      paths: List of paths to check for
      no_index: Don't check index
      quote_path: If True, quote non-ASCII characters in returned paths using
                  C-style octal escapes (e.g. "тест.txt" becomes "\\321\\202\\320\\265\\321\\201\\321\\202.txt").
                  If False, return raw unicode paths.
    Returns: List of ignored files
    """
    with open_repo_closing(repo) as r:
        index = r.open_index()
        ignore_manager = IgnoreFilterManager.from_repo(r)
        for original_path in paths:
            # Convert path to string for consistent handling
            original_path_fspath = os.fspath(original_path)
            # Normalize to str
            original_path_str = os.fsdecode(original_path_fspath)

            if not no_index and path_to_tree_path(r.path, original_path_str) in index:
                continue

            # Preserve whether the original path had a trailing slash
            had_trailing_slash = original_path_str.endswith(("/", os.path.sep))

            if os.path.isabs(original_path_str):
                path = os.path.relpath(original_path_str, r.path)
                # Normalize Windows paths to use forward slashes
                if os.path.sep != "/":
                    path = path.replace(os.path.sep, "/")
            else:
                path = original_path_str

            # Restore trailing slash if it was in the original
            if had_trailing_slash and not path.endswith("/"):
                path = path + "/"

            # For directories, check with trailing slash to get correct ignore behavior
            test_path = path
            path_without_slash = path.rstrip("/")
            is_directory = os.path.isdir(os.path.join(r.path, path_without_slash))

            # If this is a directory path, ensure we test it correctly
            if is_directory and not path.endswith("/"):
                test_path = path + "/"

            if ignore_manager.is_ignored(test_path):
                # Return relative path (like git does) when absolute path was provided
                if os.path.isabs(original_path):
                    output_path = path
                else:
                    output_path = original_path  # type: ignore[assignment]
                yield _quote_path(output_path) if quote_path else output_path


def _get_current_head_tree(repo: Repo) -> ObjectID | None:
    """Get the current HEAD tree ID.

    Args:
      repo: Repository object

    Returns:
      Tree ID of current HEAD, or None if no HEAD exists (empty repo)
    """
    try:
        current_head = repo.refs[HEADREF]
        current_commit = repo[current_head]
        assert isinstance(current_commit, Commit), "Expected a Commit object"
        tree_id: ObjectID = current_commit.tree
        return tree_id
    except KeyError:
        # No HEAD yet (empty repo)
        return None


def _check_uncommitted_changes(
    repo: Repo, target_tree_id: ObjectID, force: bool = False
) -> None:
    """Check for uncommitted changes that would conflict with a checkout/switch.

    Args:
      repo: Repository object
      target_tree_id: Tree ID to check conflicts against
      force: If True, skip the check

    Raises:
      CheckoutError: If there are conflicting local changes
    """
    if force:
        return

    # Get current HEAD tree for comparison
    current_tree_id = _get_current_head_tree(repo)
    if current_tree_id is None:
        # No HEAD yet (empty repo)
        return

    status_report = status(repo)
    changes = []
    # staged is a dict with 'add', 'delete', 'modify' keys
    if isinstance(status_report.staged, dict):
        changes.extend(status_report.staged.get("add", []))
        changes.extend(status_report.staged.get("delete", []))
        changes.extend(status_report.staged.get("modify", []))
    # unstaged is a list
    changes.extend(status_report.unstaged)

    if changes:
        # Check if any changes would conflict with checkout
        target_tree_obj = repo[target_tree_id]
        assert isinstance(target_tree_obj, Tree), "Expected a Tree object"
        target_tree = target_tree_obj
        for change in changes:
            if isinstance(change, str):
                change = change.encode(DEFAULT_ENCODING)

            try:
                target_tree.lookup_path(repo.object_store.__getitem__, change)
            except KeyError:
                # File doesn't exist in target tree - change can be preserved
                pass
            else:
                # File exists in target tree - would overwrite local changes
                raise CheckoutError(
                    f"Your local changes to '{change.decode()}' would be "
                    "overwritten. Please commit or stash before switching."
                )


def _get_worktree_update_config(
    repo: Repo,
) -> tuple[
    bool,
    Callable[[bytes], bool],
    Callable[[str | bytes | os.PathLike[str], str | bytes | os.PathLike[str]], None],
]:
    """Get configuration for working tree updates.

    Args:
      repo: Repository object

    Returns:
      Tuple of (honor_filemode, validate_path_element, symlink_fn)
    """
    config = repo.get_config()
    honor_filemode = config.get_boolean(b"core", b"filemode", os.name != "nt")

    if config.get_boolean(b"core", b"core.protectNTFS", os.name == "nt"):
        validate_path_element = validate_path_element_ntfs
    elif config.get_boolean(b"core", b"core.protectHFS", sys.platform == "darwin"):
        validate_path_element = validate_path_element_hfs
    else:
        validate_path_element = validate_path_element_default

    if config.get_boolean(b"core", b"symlinks", True):

        def symlink_wrapper(
            source: str | bytes | os.PathLike[str],
            target: str | bytes | os.PathLike[str],
        ) -> None:
            symlink(source, target)  # type: ignore[arg-type,unused-ignore]

        symlink_fn = symlink_wrapper
    else:

        def symlink_fallback(
            source: str | bytes | os.PathLike[str],
            target: str | bytes | os.PathLike[str],
        ) -> None:
            mode = "w" + ("b" if isinstance(source, bytes) else "")
            with open(target, mode) as f:
                f.write(source)

        symlink_fn = symlink_fallback

    return honor_filemode, validate_path_element, symlink_fn


def _perform_tree_switch(
    repo: Repo,
    current_tree_id: ObjectID | None,
    target_tree_id: ObjectID,
    force: bool = False,
) -> None:
    """Perform the actual working tree switch.

    Args:
      repo: Repository object
      current_tree_id: Current tree ID (or None for empty repo)
      target_tree_id: Target tree ID to switch to
      force: If True, force removal of untracked files and allow overwriting modified files
    """
    honor_filemode, validate_path_element, symlink_fn = _get_worktree_update_config(
        repo
    )

    # Get blob normalizer for line ending conversion
    blob_normalizer = repo.get_blob_normalizer()

    # Update working tree
    tree_change_iterator: Iterator[TreeChange] = tree_changes(
        repo.object_store, current_tree_id, target_tree_id
    )
    update_working_tree(
        repo,
        current_tree_id,
        target_tree_id,
        change_iterator=tree_change_iterator,
        honor_filemode=honor_filemode,
        validate_path_element=validate_path_element,
        symlink_fn=symlink_fn,
        force_remove_untracked=force,
        blob_normalizer=blob_normalizer,
        allow_overwrite_modified=force,
    )


def update_head(
    repo: RepoPath,
    target: str | bytes,
    detached: bool = False,
    new_branch: str | bytes | None = None,
) -> None:
    """Update HEAD to point at a new branch/commit.

    Note that this does not actually update the working tree.

    Args:
      repo: Path to the repository
      detached: Create a detached head
      target: Branch or committish to switch to
      new_branch: New branch to create
    """
    with open_repo_closing(repo) as r:
        if new_branch is not None:
            to_set = _make_branch_ref(new_branch)
        else:
            to_set = HEADREF
        if detached:
            # TODO(jelmer): Provide some way so that the actual ref gets
            # updated rather than what it points to, so the delete isn't
            # necessary.
            del r.refs[to_set]
            r.refs[to_set] = parse_commit(r, target).id
        else:
            r.refs.set_symbolic_ref(to_set, parse_ref(r, target))
        if new_branch is not None:
            r.refs.set_symbolic_ref(HEADREF, to_set)


def checkout(
    repo: str | os.PathLike[str] | Repo,
    target: str | bytes | Commit | Tag | None = None,
    force: bool = False,
    new_branch: bytes | str | None = None,
    paths: list[bytes | str] | None = None,
) -> None:
    """Switch to a branch or commit, updating both HEAD and the working tree.

    This is similar to 'git checkout', allowing you to switch to a branch,
    tag, or specific commit. Unlike update_head, this function also updates
    the working tree to match the target.

    Args:
      repo: Path to repository or repository object
      target: Branch name, tag, or commit SHA to checkout. If None and paths is specified,
              restores files from HEAD
      force: Force checkout even if there are local changes
      new_branch: Create a new branch at target (like git checkout -b)
      paths: List of specific paths to checkout. If specified, only these paths are updated
             and HEAD is not changed

    Raises:
      CheckoutError: If checkout cannot be performed due to conflicts
      KeyError: If the target reference cannot be found
    """
    with open_repo_closing(repo) as r:
        # Store the original target for later reference checks
        original_target = target

        worktree = r.get_worktree()

        # Handle path-specific checkout (like git checkout -- <paths>)
        if paths is not None:
            # Convert paths to bytes
            byte_paths = []
            for path in paths:
                if isinstance(path, str):
                    byte_paths.append(path.encode(DEFAULT_ENCODING))
                else:
                    byte_paths.append(path)

            # If no target specified, use HEAD
            if target is None:
                try:
                    target = r.refs[HEADREF]
                except KeyError:
                    raise CheckoutError("No HEAD reference found")
            else:
                if isinstance(target, str):
                    target = target.encode(DEFAULT_ENCODING)

            # Get the target commit and tree
            target_tree = parse_tree(r, target)

            # Get blob normalizer for line ending conversion
            blob_normalizer = r.get_blob_normalizer()

            # Restore specified paths from target tree
            for path in byte_paths:
                try:
                    # Look up the path in the target tree
                    mode, sha = target_tree.lookup_path(
                        r.object_store.__getitem__, path
                    )
                    obj = r[sha]
                    assert isinstance(obj, Blob), "Expected a Blob object"
                except KeyError:
                    # Path doesn't exist in target tree
                    pass
                else:
                    # Create directories if needed
                    # Handle path as string
                    if isinstance(path, bytes):
                        path_str = path.decode(DEFAULT_ENCODING)
                    else:
                        path_str = path
                    file_path = os.path.join(r.path, path_str)
                    os.makedirs(os.path.dirname(file_path), exist_ok=True)

                    # Write the file content
                    if stat.S_ISREG(mode):
                        # Apply checkout filters (smudge)
                        if blob_normalizer:
                            obj = blob_normalizer.checkout_normalize(obj, path)

                        flags = os.O_WRONLY | os.O_CREAT | os.O_TRUNC
                        if sys.platform == "win32":
                            flags |= os.O_BINARY

                        with os.fdopen(os.open(file_path, flags, mode), "wb") as f:
                            f.write(obj.data)

                    # Update the index
                    worktree.stage(path)

            return

        # Normal checkout (switching branches/commits)
        if target is None:
            raise ValueError("Target must be specified for branch/commit checkout")

        if isinstance(target, str):
            target_bytes = target.encode(DEFAULT_ENCODING)
        elif isinstance(target, bytes):
            target_bytes = target
        else:
            # For Commit/Tag objects, we'll use their SHA
            target_bytes = target.id

        if isinstance(new_branch, str):
            new_branch = new_branch.encode(DEFAULT_ENCODING)

        # Parse the target to get the commit
        assert (
            original_target is not None
        )  # Guaranteed by earlier check for normal checkout
        target_commit = parse_commit(r, original_target)
        target_tree_id = target_commit.tree

        # Get current HEAD tree for comparison
        current_tree_id = _get_current_head_tree(r)

        # Check for uncommitted changes if not forcing
        if current_tree_id is not None:
            _check_uncommitted_changes(r, target_tree_id, force)

        # Update working tree
        _perform_tree_switch(r, current_tree_id, target_tree_id, force)

        # Update HEAD
        if new_branch:
            # Create new branch and switch to it
            branch_create(r, new_branch, objectish=target_commit.id.decode("ascii"))
            update_head(r, new_branch)

            # Set up tracking if creating from a remote branch
            if isinstance(original_target, bytes) and target_bytes.startswith(
                LOCAL_REMOTE_PREFIX
            ):
                try:
                    remote_name, branch_name = parse_remote_ref(target_bytes)
                    # Set tracking to refs/heads/<branch> on the remote
                    set_branch_tracking(
                        r, new_branch, remote_name, local_branch_name(branch_name)
                    )
                except ValueError:
                    # Invalid remote ref format, skip tracking setup
                    pass
        else:
            # Check if target is a branch name (with or without refs/heads/ prefix)
            branch_ref = None
            if (
                isinstance(original_target, (str, bytes))
                and target_bytes in r.refs.keys()
            ):
                if target_bytes.startswith(LOCAL_BRANCH_PREFIX):
                    branch_ref = target_bytes
            else:
                # Try adding refs/heads/ prefix
                potential_branch = (
                    _make_branch_ref(target_bytes)
                    if isinstance(original_target, (str, bytes))
                    else None
                )
                if potential_branch in r.refs.keys():
                    branch_ref = potential_branch

            if branch_ref:
                # It's a branch - update HEAD symbolically
                update_head(r, branch_ref)
            else:
                # It's a tag, other ref, or commit SHA - detached HEAD
                update_head(r, target_commit.id.decode("ascii"), detached=True)


def restore(
    repo: str | os.PathLike[str] | Repo,
    paths: list[bytes | str],
    source: str | bytes | Commit | Tag | None = None,
    staged: bool = False,
    worktree: bool = True,
) -> None:
    """Restore working tree files.

    This is similar to 'git restore', allowing you to restore specific files
    from a commit or the index without changing HEAD.

    Args:
      repo: Path to repository or repository object
      paths: List of specific paths to restore
      source: Branch name, tag, or commit SHA to restore from. If None, restores
              staged files from HEAD, or worktree files from index
      staged: Restore files in the index (--staged)
      worktree: Restore files in the working tree (default: True)

    Raises:
      CheckoutError: If restore cannot be performed
      ValueError: If neither staged nor worktree is specified
      KeyError: If the source reference cannot be found
    """
    if not staged and not worktree:
        raise ValueError("At least one of staged or worktree must be True")

    with open_repo_closing(repo) as r:
        from .index import _fs_to_tree_path, build_file_from_blob

        # Determine the source tree
        if source is None:
            if staged:
                # Restoring staged files from HEAD
                try:
                    source = r.refs[HEADREF]
                except KeyError:
                    raise CheckoutError("No HEAD reference found")
            elif worktree:
                # Restoring worktree files from index
                from .index import ConflictedIndexEntry, IndexEntry

                index = r.open_index()
                for path in paths:
                    if isinstance(path, str):
                        tree_path = _fs_to_tree_path(path)
                    else:
                        tree_path = path

                    try:
                        index_entry = index[tree_path]
                        if isinstance(index_entry, ConflictedIndexEntry):
                            raise CheckoutError(
                                f"Path '{path if isinstance(path, str) else path.decode(DEFAULT_ENCODING)}' has conflicts"
                            )
                        blob = r[index_entry.sha]
                        assert isinstance(blob, Blob), "Expected a Blob object"

                        full_path = os.path.join(os.fsencode(r.path), tree_path)
                        mode = index_entry.mode

                        # Use build_file_from_blob to write the file
                        build_file_from_blob(blob, mode, full_path)
                    except KeyError:
                        # Path doesn't exist in index
                        raise CheckoutError(
                            f"Path '{path if isinstance(path, str) else path.decode(DEFAULT_ENCODING)}' not in index"
                        )
                return

        # source is not None at this point
        assert source is not None
        # Get the source tree
        source_tree = parse_tree(r, treeish=source)

        # Restore specified paths from source tree
        for path in paths:
            if isinstance(path, str):
                tree_path = _fs_to_tree_path(path)
            else:
                tree_path = path

            try:
                # Look up the path in the source tree
                mode, sha = source_tree.lookup_path(
                    r.object_store.__getitem__, tree_path
                )
                blob = r[sha]
                assert isinstance(blob, Blob), "Expected a Blob object"
            except KeyError:
                # Path doesn't exist in source tree
                raise CheckoutError(
                    f"Path '{path if isinstance(path, str) else path.decode(DEFAULT_ENCODING)}' not found in source"
                )

            full_path = os.path.join(os.fsencode(r.path), tree_path)

            if worktree:
                # Use build_file_from_blob to restore to working tree
                build_file_from_blob(blob, mode, full_path)

            if staged:
                # Update the index with the blob from source
                from .index import IndexEntry

                index = r.open_index()

                # When only updating staged (not worktree), we want to reset the index
                # to the source, but invalidate the stat cache so Git knows to check
                # the worktree file. Use zeros for stat fields.
                if not worktree:
                    # Invalidate stat cache by using zeros
                    new_entry = IndexEntry(
                        ctime=(0, 0),
                        mtime=(0, 0),
                        dev=0,
                        ino=0,
                        mode=mode,
                        uid=0,
                        gid=0,
                        size=0,
                        sha=sha,
                    )
                else:
                    # If we also updated worktree, use actual stat
                    from .index import index_entry_from_stat

                    st = os.lstat(full_path)
                    new_entry = index_entry_from_stat(st, sha, mode)

                index[tree_path] = new_entry
                index.write()


def switch(
    repo: str | os.PathLike[str] | Repo,
    target: str | bytes | Commit | Tag,
    create: str | bytes | None = None,
    force: bool = False,
    detach: bool = False,
) -> None:
    """Switch branches.

    This is similar to 'git switch', allowing you to switch to a different
    branch or commit, updating both HEAD and the working tree.

    Args:
      repo: Path to repository or repository object
      target: Branch name, tag, or commit SHA to switch to
      create: Create a new branch at target before switching (like git switch -c)
      force: Force switch even if there are local changes
      detach: Switch to a commit in detached HEAD state (like git switch --detach)

    Raises:
      CheckoutError: If switch cannot be performed due to conflicts
      KeyError: If the target reference cannot be found
      ValueError: If both create and detach are specified
    """
    if create and detach:
        raise ValueError("Cannot use both create and detach options")

    with open_repo_closing(repo) as r:
        # Store the original target for later reference checks
        original_target = target

        if isinstance(target, str):
            target_bytes = target.encode(DEFAULT_ENCODING)
        elif isinstance(target, bytes):
            target_bytes = target
        else:
            # For Commit/Tag objects, we'll use their SHA
            target_bytes = target.id

        if isinstance(create, str):
            create = create.encode(DEFAULT_ENCODING)

        # Parse the target to get the commit
        target_commit = parse_commit(r, original_target)
        target_tree_id = target_commit.tree

        # Get current HEAD tree for comparison
        current_tree_id = _get_current_head_tree(r)

        # Check for uncommitted changes if not forcing
        if current_tree_id is not None:
            _check_uncommitted_changes(r, target_tree_id, force)

        # Update working tree
        _perform_tree_switch(r, current_tree_id, target_tree_id, force)

        # Update HEAD
        if create:
            # Create new branch and switch to it
            branch_create(r, create, objectish=target_commit.id.decode("ascii"))
            update_head(r, create)

            # Set up tracking if creating from a remote branch
            if isinstance(original_target, bytes) and target_bytes.startswith(
                LOCAL_REMOTE_PREFIX
            ):
                try:
                    remote_name, branch_name = parse_remote_ref(target_bytes)
                    # Set tracking to refs/heads/<branch> on the remote
                    set_branch_tracking(
                        r, create, remote_name, local_branch_name(branch_name)
                    )
                except ValueError:
                    # Invalid remote ref format, skip tracking setup
                    pass
        elif detach:
            # Detached HEAD mode
            update_head(r, target_commit.id.decode("ascii"), detached=True)
        else:
            # Check if target is a branch name (with or without refs/heads/ prefix)
            branch_ref = None
            if (
                isinstance(original_target, (str, bytes))
                and target_bytes in r.refs.keys()
            ):
                if target_bytes.startswith(LOCAL_BRANCH_PREFIX):
                    branch_ref = target_bytes
            else:
                # Try adding refs/heads/ prefix
                potential_branch = (
                    _make_branch_ref(target_bytes)
                    if isinstance(original_target, (str, bytes))
                    else None
                )
                if potential_branch in r.refs.keys():
                    branch_ref = potential_branch

            if branch_ref:
                # It's a branch - update HEAD symbolically
                update_head(r, branch_ref)
            else:
                # It's a tag, other ref, or commit SHA
                # In git switch, this would be an error unless --detach is used
                raise CheckoutError(
                    f"'{target_bytes.decode(DEFAULT_ENCODING)}' is not a branch. "
                    "Use detach=True to switch to a commit in detached HEAD state."
                )


def reset_file(
    repo: Repo,
    file_path: str,
    target: str | bytes | Commit | Tree | Tag = b"HEAD",
    symlink_fn: Callable[
        [str | bytes | os.PathLike[str], str | bytes | os.PathLike[str]], None
    ]
    | None = None,
) -> None:
    """Reset the file to specific commit or branch.

    Args:
      repo: dulwich Repo object
      file_path: file to reset, relative to the repository path
      target: branch or commit or b'HEAD' to reset
      symlink_fn: Function to use for creating symlinks
    """
    tree = parse_tree(repo, treeish=target)
    tree_path = _fs_to_tree_path(file_path)

    file_entry = tree.lookup_path(repo.object_store.__getitem__, tree_path)
    full_path = os.path.join(os.fsencode(repo.path), tree_path)
    blob = repo.object_store[file_entry[1]]
    assert isinstance(blob, Blob)
    mode = file_entry[0]
    build_file_from_blob(blob, mode, full_path, symlink_fn=symlink_fn)


@replace_me(since="0.22.9", remove_in="0.24.0")
def checkout_branch(
    repo: str | os.PathLike[str] | Repo,
    target: bytes | str,
    force: bool = False,
) -> None:
    """Switch branches or restore working tree files.

    This is now a wrapper around the general checkout() function.
    Preserved for backward compatibility.

    Args:
      repo: dulwich Repo object
      target: branch name or commit sha to checkout
      force: true or not to force checkout
    """
    # Simply delegate to the new checkout function
    return checkout(repo, target, force=force)


def sparse_checkout(
    repo: str | os.PathLike[str] | Repo,
    patterns: list[str] | None = None,
    force: bool = False,
    cone: bool | None = None,
) -> None:
    """Perform a sparse checkout in the repository (either 'full' or 'cone mode').

    Perform sparse checkout in either 'cone' (directory-based) mode or
    'full pattern' (.gitignore) mode, depending on the ``cone`` parameter.

    If ``cone`` is ``None``, the mode is inferred from the repository's
    ``core.sparseCheckoutCone`` config setting.

    Steps:
      1) If ``patterns`` is provided, write them to ``.git/info/sparse-checkout``.
      2) Determine which paths in the index are included vs. excluded.
         - If ``cone=True``, use "cone-compatible" directory-based logic.
         - If ``cone=False``, use standard .gitignore-style matching.
      3) Update the index's skip-worktree bits and add/remove files in
         the working tree accordingly.
      4) If ``force=False``, refuse to remove files that have local modifications.

    Args:
      repo: Path to the repository or a Repo object.
      patterns: Optional list of sparse-checkout patterns to write.
      force: Whether to force removal of locally modified files (default False).
      cone: Boolean indicating cone mode (True/False). If None, read from config.

    Returns:
      None
    """
    with open_repo_closing(repo) as repo_obj:
        # --- 0) Possibly infer 'cone' from config ---
        if cone is None:
            cone = repo_obj.get_worktree().infer_cone_mode()

        # --- 1) Read or write patterns ---
        if patterns is None:
            lines = repo_obj.get_worktree().get_sparse_checkout_patterns()
            if lines is None:
                raise Error("No sparse checkout patterns found.")
        else:
            lines = patterns
            repo_obj.get_worktree().set_sparse_checkout_patterns(patterns)

        # --- 2) Determine the set of included paths ---
        index = repo_obj.open_index()
        included_paths = determine_included_paths(index, lines, cone)

        # --- 3) Apply those results to the index & working tree ---
        try:
            apply_included_paths(repo_obj, included_paths, force=force)
        except SparseCheckoutConflictError as exc:
            raise CheckoutError(*exc.args) from exc


def cone_mode_init(repo: str | os.PathLike[str] | Repo) -> None:
    """Initialize a repository to use sparse checkout in 'cone' mode.

    Sets ``core.sparseCheckout`` and ``core.sparseCheckoutCone`` in the config.
    Writes an initial ``.git/info/sparse-checkout`` file that includes only
    top-level files (and excludes all subdirectories), e.g. ``["/*", "!/*/"]``.
    Then performs a sparse checkout to update the working tree accordingly.

    If no directories are specified, then only top-level files are included:
    https://git-scm.com/docs/git-sparse-checkout#_internalscone_mode_handling

    Args:
      repo: Path to the repository or a Repo object.

    Returns:
      None
    """
    with open_repo_closing(repo) as repo_obj:
        repo_obj.get_worktree().configure_for_cone_mode()
        patterns = ["/*", "!/*/"]  # root-level files only
        sparse_checkout(repo_obj, patterns, force=True, cone=True)


def cone_mode_set(
    repo: str | os.PathLike[str] | Repo, dirs: Sequence[str], force: bool = False
) -> None:
    """Overwrite the existing 'cone-mode' sparse patterns with a new set of directories.

    Ensures ``core.sparseCheckout`` and ``core.sparseCheckoutCone`` are enabled.
    Writes new patterns so that only the specified directories (and top-level files)
    remain in the working tree, and applies the sparse checkout update.

    Args:
      repo: Path to the repository or a Repo object.
      dirs: List of directory names to include.
      force: Whether to forcibly discard local modifications (default False).

    Returns:
      None
    """
    with open_repo_closing(repo) as repo_obj:
        repo_obj.get_worktree().configure_for_cone_mode()
        repo_obj.get_worktree().set_cone_mode_patterns(dirs=dirs)
        new_patterns = repo_obj.get_worktree().get_sparse_checkout_patterns()
        # Finally, apply the patterns and update the working tree
        sparse_checkout(repo_obj, new_patterns, force=force, cone=True)


def cone_mode_add(
    repo: str | os.PathLike[str] | Repo, dirs: Sequence[str], force: bool = False
) -> None:
    """Add new directories to the existing 'cone-mode' sparse-checkout patterns.

    Reads the current patterns from ``.git/info/sparse-checkout``, adds pattern
    lines to include the specified directories, and then performs a sparse
    checkout to update the working tree accordingly.

    Args:
      repo: Path to the repository or a Repo object.
      dirs: List of directory names to add to the sparse-checkout.
      force: Whether to forcibly discard local modifications (default False).

    Returns:
      None
    """
    with open_repo_closing(repo) as repo_obj:
        repo_obj.get_worktree().configure_for_cone_mode()
        # Do not pass base patterns as dirs
        base_patterns = ["/*", "!/*/"]
        existing_dirs = [
            pat.strip("/")
            for pat in repo_obj.get_worktree().get_sparse_checkout_patterns()
            if pat not in base_patterns
        ]
        added_dirs = existing_dirs + list(dirs or [])
        repo_obj.get_worktree().set_cone_mode_patterns(dirs=added_dirs)
        new_patterns = repo_obj.get_worktree().get_sparse_checkout_patterns()
        sparse_checkout(repo_obj, patterns=new_patterns, force=force, cone=True)


def check_mailmap(repo: RepoPath, contact: str | bytes) -> bytes:
    """Check canonical name and email of contact.

    Args:
      repo: Path to the repository
      contact: Contact name and/or email
    Returns: Canonical contact data
    """
    with open_repo_closing(repo) as r:
        from .mailmap import Mailmap

        try:
            mailmap = Mailmap.from_path(os.path.join(r.path, ".mailmap"))
        except FileNotFoundError:
            mailmap = Mailmap()
        contact_bytes = (
            contact.encode(DEFAULT_ENCODING) if isinstance(contact, str) else contact
        )
        result = mailmap.lookup(contact_bytes)
        if isinstance(result, bytes):
            return result
        else:
            # Convert tuple back to bytes format
            name, email = result
            if name is None:
                name = b""
            if email is None:
                email = b""
            return name + b" <" + email + b">"


def fsck(repo: RepoPath) -> Iterator[tuple[bytes, Exception]]:
    """Check a repository.

    Args:
      repo: A path to the repository
    Returns: Iterator over errors/warnings
    """
    with open_repo_closing(repo) as r:
        # TODO(jelmer): check pack files
        # TODO(jelmer): check graph
        # TODO(jelmer): check refs
        for sha in r.object_store:
            o = r.object_store[sha]
            try:
                o.check()
            except Exception as e:
                yield (sha, e)


def stash_list(
    repo: str | os.PathLike[str] | Repo,
) -> Iterator[tuple[int, tuple[bytes, bytes]]]:
    """List all stashes in a repository."""
    with open_repo_closing(repo) as r:
        from .stash import Stash

        stash = Stash.from_repo(r)
        entries = stash.stashes()
        # Convert Entry objects to (old_sha, new_sha) tuples
        return enumerate([(entry.old_sha, entry.new_sha) for entry in entries])


def stash_push(repo: str | os.PathLike[str] | Repo) -> None:
    """Push a new stash onto the stack."""
    with open_repo_closing(repo) as r:
        from .stash import Stash

        stash = Stash.from_repo(r)
        stash.push()


def stash_pop(repo: str | os.PathLike[str] | Repo) -> None:
    """Pop a stash from the stack."""
    with open_repo_closing(repo) as r:
        from .stash import Stash

        stash = Stash.from_repo(r)
        stash.pop(0)


def stash_drop(repo: str | os.PathLike[str] | Repo, index: int) -> None:
    """Drop a stash from the stack."""
    with open_repo_closing(repo) as r:
        from .stash import Stash

        stash = Stash.from_repo(r)
        stash.drop(index)


def ls_files(repo: RepoPath) -> list[bytes]:
    """List all files in an index."""
    with open_repo_closing(repo) as r:
        return sorted(r.open_index())


def find_unique_abbrev(
    object_store: BaseObjectStore, object_id: str | bytes, min_length: int = 7
) -> str:
    """Find the shortest unique abbreviation for an object ID.

    Args:
      object_store: Object store to search in
      object_id: The full object ID to abbreviate
      min_length: Minimum length of abbreviation (default 7)

    Returns:
      The shortest unique prefix of the object ID (at least min_length chars)
    """
    if isinstance(object_id, bytes):
        hex_id = object_id.decode("ascii")
    else:
        hex_id = object_id

    # Start with minimum length
    for length in range(min_length, len(hex_id) + 1):
        prefix = hex_id[:length]
        matches = 0

        # Check if this prefix is unique
        for obj_id in object_store:
            if obj_id.decode("ascii").startswith(prefix):
                matches += 1
                if matches > 1:
                    # Not unique, need more characters
                    break

        if matches == 1:
            # Found unique prefix
            return prefix

    # If we get here, return the full ID
    return hex_id


def describe(repo: str | os.PathLike[str] | Repo, abbrev: int | None = None) -> str:
    """Describe the repository version.

    Args:
      repo: git repository
      abbrev: number of characters of commit to take, default is 7
    Returns: a string description of the current git revision

    Examples: "gabcdefh", "v0.1" or "v0.1-5-gabcdefh".
    """
    abbrev_slice = slice(0, abbrev if abbrev is not None else 7)
    # Get the repository
    with open_repo_closing(repo) as r:
        # Get a list of all tags
        refs = r.get_refs()
        tags = {}
        for key, value in refs.items():
            key_str = key.decode()
            obj = r.get_object(value)
            if "tags" not in key_str:
                continue

            _, tag = key_str.rsplit("/", 1)

            if isinstance(obj, Tag):
                # Annotated tag case
                commit = r.get_object(obj.object[1])
            else:
                # Lightweight tag case - obj is already the commit
                commit = obj

            if not isinstance(commit, Commit):
                raise AssertionError(
                    f"Expected Commit object, got {type(commit).__name__}"
                )

            tag_info: list[Any] = [
                datetime.datetime(*time.gmtime(commit.commit_time)[:6]),
                commit.id.decode("ascii"),
            ]
            tags[tag] = tag_info

        # Sort tags by datetime (first element of the value list)
        sorted_tags = sorted(
            tags.items(), key=lambda tag_item: tag_item[1][0], reverse=True
        )

        # Get the latest commit
        latest_commit = r[r.head()]

        # If there are no tags, return the latest commit
        if len(sorted_tags) == 0:
            if abbrev is not None:
                return "g{}".format(latest_commit.id.decode("ascii")[abbrev_slice])
            return f"g{find_unique_abbrev(r.object_store, latest_commit.id)}"

        # We're now 0 commits from the top
        commit_count = 0

        # Walk through all commits
        walker = r.get_walker()
        for entry in walker:
            # Check if tag
            commit_id = entry.commit.id.decode("ascii")
            for tag_item in sorted_tags:
                tag_name = tag_item[0]
                tag_commit = tag_item[1][1]
                if commit_id == tag_commit:
                    if commit_count == 0:
                        return tag_name
                    else:
                        if abbrev is not None:
                            abbrev_hash = latest_commit.id.decode("ascii")[abbrev_slice]
                        else:
                            abbrev_hash = find_unique_abbrev(
                                r.object_store, latest_commit.id
                            )
                        return f"{tag_name}-{commit_count}-g{abbrev_hash}"

            commit_count += 1

        # Return plain commit if no parent tag can be found
        if abbrev is not None:
            return "g{}".format(latest_commit.id.decode("ascii")[abbrev_slice])
        return f"g{find_unique_abbrev(r.object_store, latest_commit.id)}"


def get_object_by_path(
    repo: RepoPath,
    path: str | bytes,
    committish: str | bytes | Commit | Tag | None = None,
) -> Blob | Tree | Commit | Tag:
    """Get an object by path.

    Args:
      repo: A path to the repository
      path: Path to look up
      committish: Commit to look up path in
    Returns: A `ShaFile` object
    """
    if committish is None:
        committish = "HEAD"
    # Get the repository
    with open_repo_closing(repo) as r:
        commit = parse_commit(r, committish)
        base_tree = commit.tree
        if not isinstance(path, bytes):
            path = commit_encode(commit, path)
        (_mode, sha) = tree_lookup_path(r.object_store.__getitem__, base_tree, path)
        obj = r[sha]
        assert isinstance(obj, (Blob, Tree, Commit, Tag))
        return obj


def write_tree(repo: RepoPath) -> bytes:
    """Write a tree object from the index.

    Args:
      repo: Repository for which to write tree
    Returns: tree id for the tree that was written
    """
    with open_repo_closing(repo) as r:
        return r.open_index().commit(r.object_store)


def _do_merge(
    r: Repo,
    merge_commit_id: ObjectID,
    no_commit: bool = False,
    no_ff: bool = False,
    message: bytes | None = None,
    author: bytes | None = None,
    committer: bytes | None = None,
) -> tuple[ObjectID | None, list[bytes]]:
    """Internal merge implementation that operates on an open repository.

    Args:
      r: Open repository object
      merge_commit_id: SHA of commit to merge
      no_commit: If True, do not create a merge commit
      no_ff: If True, force creation of a merge commit
      message: Optional merge commit message
      author: Optional author for merge commit
      committer: Optional committer for merge commit

    Returns:
      Tuple of (merge_commit_sha, conflicts) where merge_commit_sha is None
      if no_commit=True or there were conflicts
    """
    from .graph import find_merge_base
    from .merge import recursive_merge

    # Get HEAD commit
    try:
        head_commit_id = r.refs[HEADREF]
    except KeyError:
        raise Error("No HEAD reference found")

    head_commit = r[head_commit_id]
    assert isinstance(head_commit, Commit), "Expected a Commit object"
    merge_commit = r[merge_commit_id]
    assert isinstance(merge_commit, Commit), "Expected a Commit object"

    # Check if fast-forward is possible
    merge_bases = find_merge_base(r, [head_commit_id, merge_commit_id])

    if not merge_bases:
        raise Error("No common ancestor found")

    # Use the first merge base for fast-forward checks
    base_commit_id = merge_bases[0]

    # Check if we're trying to merge the same commit
    if head_commit_id == merge_commit_id:
        # Already up to date
        return (None, [])

    # Check for fast-forward
    if base_commit_id == head_commit_id and not no_ff:
        # Fast-forward merge
        r.refs[HEADREF] = merge_commit_id
        # Update the working directory
        changes = tree_changes(r.object_store, head_commit.tree, merge_commit.tree)
        update_working_tree(
            r, head_commit.tree, merge_commit.tree, change_iterator=changes
        )
        return (merge_commit_id, [])

    if base_commit_id == merge_commit_id:
        # Already up to date
        return (None, [])

    # Perform recursive merge (handles multiple merge bases automatically)
    gitattributes = r.get_gitattributes()
    config = r.get_config()
    merged_tree, conflicts = recursive_merge(
        r.object_store, merge_bases, head_commit, merge_commit, gitattributes, config
    )

    # Add merged tree to object store
    r.object_store.add_object(merged_tree)

    # Update index and working directory
    changes = tree_changes(r.object_store, head_commit.tree, merged_tree.id)
    update_working_tree(r, head_commit.tree, merged_tree.id, change_iterator=changes)

    if conflicts or no_commit:
        # Don't create a commit if there are conflicts or no_commit is True
        return (None, conflicts)

    # Create merge commit
    merge_commit_obj = Commit()
    merge_commit_obj.tree = merged_tree.id
    merge_commit_obj.parents = [head_commit_id, merge_commit_id]

    # Set author/committer
    if author is None:
        author = get_user_identity(r.get_config_stack())
    if committer is None:
        committer = author

    merge_commit_obj.author = author
    merge_commit_obj.committer = committer

    # Set timestamps
    timestamp = int(time.time())
    timezone = 0  # UTC
    merge_commit_obj.author_time = timestamp
    merge_commit_obj.author_timezone = timezone
    merge_commit_obj.commit_time = timestamp
    merge_commit_obj.commit_timezone = timezone

    # Set commit message
    if message is None:
        message = f"Merge commit '{merge_commit_id.decode()[:7]}'\n".encode()
    merge_commit_obj.message = message.encode() if isinstance(message, str) else message

    # Add commit to object store
    r.object_store.add_object(merge_commit_obj)

    # Update HEAD
    r.refs[HEADREF] = merge_commit_obj.id

    return (merge_commit_obj.id, [])


def _do_octopus_merge(
    r: Repo,
    merge_commit_ids: list[ObjectID],
    no_commit: bool = False,
    no_ff: bool = False,
    message: bytes | None = None,
    author: bytes | None = None,
    committer: bytes | None = None,
) -> tuple[ObjectID | None, list[bytes]]:
    """Internal octopus merge implementation that operates on an open repository.

    Args:
      r: Open repository object
      merge_commit_ids: List of commit SHAs to merge
      no_commit: If True, do not create a merge commit
      no_ff: If True, force creation of a merge commit (ignored for octopus)
      message: Optional merge commit message
      author: Optional author for merge commit
      committer: Optional committer for merge commit

    Returns:
      Tuple of (merge_commit_sha, conflicts) where merge_commit_sha is None
      if no_commit=True or there were conflicts
    """
    from .graph import find_octopus_base
    from .merge import octopus_merge

    # Get HEAD commit
    try:
        head_commit_id = r.refs[HEADREF]
    except KeyError:
        raise Error("No HEAD reference found")

    head_commit = r[head_commit_id]
    assert isinstance(head_commit, Commit), "Expected a Commit object"

    # Get all commits to merge
    other_commits = []
    for merge_commit_id in merge_commit_ids:
        merge_commit = r[merge_commit_id]
        assert isinstance(merge_commit, Commit), "Expected a Commit object"

        # Check if we're trying to merge the same commit as HEAD
        if head_commit_id == merge_commit_id:
            # Skip this commit, it's already merged
            continue

        other_commits.append(merge_commit)

    # If no commits to merge after filtering, we're already up to date
    if not other_commits:
        return (None, [])

    # If only one commit to merge, use regular merge
    if len(other_commits) == 1:
        return _do_merge(
            r, other_commits[0].id, no_commit, no_ff, message, author, committer
        )

    # Find the octopus merge base
    all_commit_ids = [head_commit_id] + [c.id for c in other_commits]
    merge_bases = find_octopus_base(r, all_commit_ids)

    if not merge_bases:
        raise Error("No common ancestor found")

    # Check if this is a fast-forward (HEAD is the merge base)
    # For octopus merges, fast-forward doesn't really apply, so we always create a merge commit

    # Perform octopus merge
    gitattributes = r.get_gitattributes()
    config = r.get_config()
    merged_tree, conflicts = octopus_merge(
        r.object_store, merge_bases, head_commit, other_commits, gitattributes, config
    )

    # Add merged tree to object store
    r.object_store.add_object(merged_tree)

    # Update index and working directory
    changes = tree_changes(r.object_store, head_commit.tree, merged_tree.id)
    update_working_tree(r, head_commit.tree, merged_tree.id, change_iterator=changes)

    if conflicts:
        # Don't create a commit if there are conflicts
        # Octopus merge refuses to proceed with conflicts
        return (None, conflicts)

    if no_commit:
        # Don't create a commit if no_commit is True
        return (None, [])

    # Create merge commit with multiple parents
    merge_commit_obj = Commit()
    merge_commit_obj.tree = merged_tree.id
    merge_commit_obj.parents = [head_commit_id] + [c.id for c in other_commits]

    # Set author/committer
    if author is None:
        author = get_user_identity(r.get_config_stack())
    if committer is None:
        committer = author

    merge_commit_obj.author = author
    merge_commit_obj.committer = committer

    # Set timestamps
    timestamp = int(time.time())
    timezone = 0  # UTC
    merge_commit_obj.author_time = timestamp
    merge_commit_obj.author_timezone = timezone
    merge_commit_obj.commit_time = timestamp
    merge_commit_obj.commit_timezone = timezone

    # Set commit message
    if message is None:
        # Generate default message for octopus merge
        branch_names = []
        for commit_id in merge_commit_ids:
            branch_names.append(commit_id.decode()[:7])
        message = f"Merge commits {', '.join(branch_names)}\n".encode()
    merge_commit_obj.message = message.encode() if isinstance(message, str) else message

    # Add commit to object store
    r.object_store.add_object(merge_commit_obj)

    # Update HEAD
    r.refs[HEADREF] = merge_commit_obj.id

    return (merge_commit_obj.id, [])


def merge(
    repo: str | os.PathLike[str] | Repo,
    committish: str | bytes | Commit | Tag | Sequence[str | bytes | Commit | Tag],
    no_commit: bool = False,
    no_ff: bool = False,
    message: bytes | None = None,
    author: bytes | None = None,
    committer: bytes | None = None,
) -> tuple[bytes | None, list[bytes]]:
    """Merge one or more commits into the current branch.

    Args:
      repo: Repository to merge into
      committish: Commit(s) to merge. Can be a single commit or a sequence of commits.
                  When merging more than two heads, the octopus merge strategy is used.
      no_commit: If True, do not create a merge commit
      no_ff: If True, force creation of a merge commit
      message: Optional merge commit message
      author: Optional author for merge commit
      committer: Optional committer for merge commit

    Returns:
      Tuple of (merge_commit_sha, conflicts) where merge_commit_sha is None
      if no_commit=True or there were conflicts

    Raises:
      Error: If there is no HEAD reference or commit cannot be found
    """
    with open_repo_closing(repo) as r:
        # Handle both single commit and multiple commits
        if isinstance(committish, (list, tuple)):
            # Multiple commits - use octopus merge
            merge_commit_ids = []
            for c in committish:
                try:
                    merge_commit_ids.append(parse_commit(r, c).id)
                except KeyError:
                    raise Error(
                        f"Cannot find commit '{c.decode() if isinstance(c, bytes) else c}'"
                    )

            if len(merge_commit_ids) == 1:
                # Only one commit, use regular merge
                result = _do_merge(
                    r, merge_commit_ids[0], no_commit, no_ff, message, author, committer
                )
            else:
                # Multiple commits, use octopus merge
                result = _do_octopus_merge(
                    r, merge_commit_ids, no_commit, no_ff, message, author, committer
                )
        else:
            # Single commit - use regular merge
            # Type narrowing: committish is not a sequence in this branch
            single_committish = cast(str | bytes | Commit | Tag, committish)
            try:
                merge_commit_id = parse_commit(r, single_committish).id
            except KeyError:
                raise Error(
                    f"Cannot find commit '{single_committish.decode() if isinstance(single_committish, bytes) else single_committish}'"
                )

            result = _do_merge(
                r, merge_commit_id, no_commit, no_ff, message, author, committer
            )

        # Trigger auto GC if needed
        from .gc import maybe_auto_gc

        maybe_auto_gc(r)

        return result


def unpack_objects(
    pack_path: str | os.PathLike[str], target: str | os.PathLike[str] = "."
) -> int:
    """Unpack objects from a pack file into the repository.

    Args:
      pack_path: Path to the pack file to unpack
      target: Path to the repository to unpack into

    Returns:
      Number of objects unpacked
    """
    from .pack import Pack

    with open_repo_closing(target) as r:
        pack_basename = os.path.splitext(pack_path)[0]
        with Pack(pack_basename) as pack:
            count = 0
            for unpacked in pack.iter_unpacked():
                obj = unpacked.sha_file()
                r.object_store.add_object(obj)
                count += 1
            return count


def merge_tree(
    repo: RepoPath,
    base_tree: str | bytes | Tree | Commit | Tag | None,
    our_tree: str | bytes | Tree | Commit | Tag,
    their_tree: str | bytes | Tree | Commit | Tag,
) -> tuple[bytes, list[bytes]]:
    """Perform a three-way tree merge without touching the working directory.

    This is similar to git merge-tree, performing a merge at the tree level
    without creating commits or updating any references.

    Args:
      repo: Repository containing the trees
      base_tree: Tree-ish of the common ancestor (or None for no common ancestor)
      our_tree: Tree-ish of our side of the merge
      their_tree: Tree-ish of their side of the merge

    Returns:
      tuple: A tuple of (merged_tree_id, conflicts) where:
        - merged_tree_id is the SHA-1 of the merged tree
        - conflicts is a list of paths (as bytes) that had conflicts

    Raises:
      KeyError: If any of the tree-ish arguments cannot be resolved
    """
    from .merge import Merger

    with open_repo_closing(repo) as r:
        # Resolve tree-ish arguments to actual trees
        base = parse_tree(r, base_tree) if base_tree else None
        ours = parse_tree(r, our_tree)
        theirs = parse_tree(r, their_tree)

        # Perform the merge
        gitattributes = r.get_gitattributes()
        config = r.get_config()
        merger = Merger(r.object_store, gitattributes, config)
        merged_tree, conflicts = merger.merge_trees(base, ours, theirs)

        # Add the merged tree to the object store
        r.object_store.add_object(merged_tree)

        return merged_tree.id, conflicts


def cherry(
    repo: str | os.PathLike[str] | Repo,
    upstream: str | bytes | None = None,
    head: str | bytes | None = None,
    limit: str | bytes | None = None,
    verbose: bool = False,
) -> list[tuple[str, bytes, bytes | None]]:
    """Find commits not merged upstream.

    Args:
        repo: Repository path or object
        upstream: Upstream branch (default: tracking branch or @{upstream})
        head: Head branch (default: HEAD)
        limit: Limit commits to those after this ref
        verbose: Include commit messages in output

    Returns:
        List of tuples (status, commit_sha, message) where status is '+' or '-'
        '+' means commit is not in upstream, '-' means equivalent patch exists upstream
        message is None unless verbose=True
    """
    from .patch import commit_patch_id

    with open_repo_closing(repo) as r:
        # Resolve upstream
        if upstream is None:
            # Try to find tracking branch
            upstream_found = False
            head_refs, _ = r.refs.follow(HEADREF)
            if head_refs:
                head_ref = head_refs[0]
                if head_ref.startswith(b"refs/heads/"):
                    config = r.get_config()
                    branch_name = head_ref[len(b"refs/heads/") :]

                    try:
                        upstream_ref = config.get((b"branch", branch_name), b"merge")
                    except KeyError:
                        upstream_ref = None

                    if upstream_ref:
                        try:
                            remote_name = config.get(
                                (b"branch", branch_name), b"remote"
                            )
                        except KeyError:
                            remote_name = None

                        if remote_name:
                            # Build the tracking branch ref
                            upstream_refname = Ref(
                                b"refs/remotes/"
                                + remote_name
                                + b"/"
                                + upstream_ref.split(b"/")[-1]
                            )
                            if upstream_refname in r.refs:
                                upstream = upstream_refname
                                upstream_found = True

            if not upstream_found:
                # Default to HEAD^ if no tracking branch found
                head_commit = r[HEADREF]
                if isinstance(head_commit, Commit) and head_commit.parents:
                    upstream = head_commit.parents[0]
                else:
                    raise ValueError("Could not determine upstream branch")

        # Resolve head
        if head is None:
            head = b"HEAD"

        # Convert strings to bytes
        if isinstance(upstream, str):
            upstream = upstream.encode("utf-8")
        if isinstance(head, str):
            head = head.encode("utf-8")
        if limit is not None and isinstance(limit, str):
            limit = limit.encode("utf-8")

        # Resolve refs to commit IDs
        assert upstream is not None
        upstream_obj = r[upstream]
        head_obj = r[head]
        upstream_id = upstream_obj.id
        head_id = head_obj.id

        # Get limit commit ID if specified
        limit_id = None
        if limit is not None:
            limit_id = r[limit].id

        # Find all commits reachable from head but not from upstream
        # This is equivalent to: git rev-list ^upstream head

        # Get commits from head that are not in upstream
        walker = r.get_walker([head_id], exclude=[upstream_id])
        head_commits = []
        for entry in walker:
            commit = entry.commit
            # Apply limit if specified
            if limit_id is not None:
                # Stop when we reach the limit commit
                if commit.id == limit_id:
                    break
            head_commits.append(commit.id)

        # Compute patch IDs for upstream commits
        upstream_walker = r.get_walker([upstream_id])
        upstream_patch_ids = {}  # Maps patch_id -> commit_id for debugging
        for entry in upstream_walker:
            commit = entry.commit
            pid = commit_patch_id(r.object_store, commit.id)
            upstream_patch_ids[pid] = commit.id

        # For each head commit, check if equivalent patch exists in upstream
        results: list[tuple[str, bytes, bytes | None]] = []
        for commit_id in reversed(head_commits):  # Show oldest first
            obj = r.object_store[commit_id]
            assert isinstance(obj, Commit)
            commit = obj

            pid = commit_patch_id(r.object_store, commit_id)

            if pid in upstream_patch_ids:
                status = "-"
            else:
                status = "+"

            message = None
            if verbose:
                message = commit.message.split(b"\n")[0]  # First line only

            results.append((status, commit_id, message))

        return results


def cherry_pick(  # noqa: D417
    repo: str | os.PathLike[str] | Repo,
    committish: str | bytes | Commit | Tag | None,
    no_commit: bool = False,
    continue_: bool = False,
    abort: bool = False,
) -> bytes | None:
    r"""Cherry-pick a commit onto the current branch.

    Args:
      repo: Repository to cherry-pick into
      committish: Commit to cherry-pick (can be None only when resuming or aborting)
      no_commit: If True, do not create a commit after applying changes
      ``continue_``: Resume an in-progress cherry-pick after resolving conflicts if True
      abort: Abort an in-progress cherry-pick

    Returns:
      The SHA of the newly created commit, or None if no_commit=True or there were conflicts

    Raises:
      Error: If there is no HEAD reference, commit cannot be found, or operation fails
    """
    from .merge import three_way_merge

    # Validate that committish is provided when needed
    if not (continue_ or abort) and committish is None:
        raise ValueError("committish is required when not using --continue or --abort")

    with open_repo_closing(repo) as r:
        # Handle abort
        if abort:
            # Clean up any cherry-pick state
            try:
                os.remove(os.path.join(r.controldir(), "CHERRY_PICK_HEAD"))
            except FileNotFoundError:
                pass
            try:
                os.remove(os.path.join(r.controldir(), "MERGE_MSG"))
            except FileNotFoundError:
                pass
            # Reset index to HEAD
            head_commit = r[b"HEAD"]
            assert isinstance(head_commit, Commit)
            r.get_worktree().reset_index(head_commit.tree)
            return None

        # Handle continue
        if continue_:
            # Check if there's a cherry-pick in progress
            cherry_pick_head_path = os.path.join(r.controldir(), "CHERRY_PICK_HEAD")
            try:
                with open(cherry_pick_head_path, "rb") as f:
                    cherry_pick_commit_id = f.read().strip()
                cherry_pick_commit = r[cherry_pick_commit_id]
            except FileNotFoundError:
                raise Error("No cherry-pick in progress")

            # Check for unresolved conflicts
            if r.open_index().has_conflicts():
                raise Error("Unresolved conflicts remain")

            # Create the commit
            tree_id = r.open_index().commit(r.object_store)

            # Read saved message if any
            merge_msg_path = os.path.join(r.controldir(), "MERGE_MSG")
            try:
                with open(merge_msg_path, "rb") as f:
                    message = f.read()
            except FileNotFoundError:
                assert isinstance(cherry_pick_commit, Commit)
                message = cherry_pick_commit.message

            assert isinstance(cherry_pick_commit, Commit)
            new_commit = r.get_worktree().commit(
                message=message,
                tree=tree_id,
                author=cherry_pick_commit.author,
                author_timestamp=cherry_pick_commit.author_time,
                author_timezone=cherry_pick_commit.author_timezone,
            )

            # Clean up state files
            try:
                os.remove(cherry_pick_head_path)
            except FileNotFoundError:
                pass
            try:
                os.remove(merge_msg_path)
            except FileNotFoundError:
                pass

            return new_commit

        # Normal cherry-pick operation
        # Get current HEAD
        try:
            head_commit = r[b"HEAD"]
        except KeyError:
            raise Error("No HEAD reference found")

        # Parse the commit to cherry-pick
        # committish cannot be None here due to validation above
        assert committish is not None
        try:
            cherry_pick_commit = parse_commit(r, committish)
        except KeyError:
            raise Error(
                f"Cannot find commit '{committish.decode() if isinstance(committish, bytes) else committish}'"
            )

        # Check if commit has parents
        assert isinstance(cherry_pick_commit, Commit)
        if not cherry_pick_commit.parents:
            raise Error("Cannot cherry-pick root commit")

        # Get parent of cherry-pick commit
        parent_commit = r[cherry_pick_commit.parents[0]]
        assert isinstance(parent_commit, Commit)

        # Perform three-way merge
        assert isinstance(head_commit, Commit)
        merged_tree, conflicts = three_way_merge(
            r.object_store, parent_commit, head_commit, cherry_pick_commit
        )

        # Add merged tree to object store
        r.object_store.add_object(merged_tree)

        # Update working tree and index
        # Reset index to match merged tree
        r.get_worktree().reset_index(merged_tree.id)

        # Update working tree from the new index
        # Allow overwriting because we're applying the merge result
        assert isinstance(head_commit, Commit)
        changes = tree_changes(r.object_store, head_commit.tree, merged_tree.id)
        update_working_tree(
            r,
            head_commit.tree,
            merged_tree.id,
            change_iterator=changes,
            allow_overwrite_modified=True,
        )

        if conflicts:
            # Save state for later continuation
            with open(os.path.join(r.controldir(), "CHERRY_PICK_HEAD"), "wb") as f:
                f.write(cherry_pick_commit.id + b"\n")

            # Save commit message
            with open(os.path.join(r.controldir(), "MERGE_MSG"), "wb") as f:
                f.write(cherry_pick_commit.message)

            raise Error(
                f"Conflicts in: {', '.join(c.decode('utf-8', 'replace') for c in conflicts)}\n"
                f"Fix conflicts and run 'dulwich cherry-pick --continue'"
            )

        if no_commit:
            return None

        # Create the commit
        new_commit = r.get_worktree().commit(
            message=cherry_pick_commit.message,
            tree=merged_tree.id,
            author=cherry_pick_commit.author,
            author_timestamp=cherry_pick_commit.author_time,
            author_timezone=cherry_pick_commit.author_timezone,
        )

        return new_commit


def revert(
    repo: str | os.PathLike[str] | Repo,
    commits: str | bytes | Commit | Tag | Sequence[str | bytes | Commit | Tag],
    no_commit: bool = False,
    message: str | bytes | None = None,
    author: bytes | None = None,
    committer: bytes | None = None,
) -> bytes | None:
    """Revert one or more commits.

    This creates a new commit that undoes the changes introduced by the
    specified commits. Unlike reset, revert creates a new commit that
    preserves history.

    Args:
      repo: Path to repository or repository object
      commits: List of commit-ish (SHA, ref, etc.) to revert, or a single commit-ish
      no_commit: If True, apply changes to index/working tree but don't commit
      message: Optional commit message (default: "Revert <original subject>")
      author: Optional author for revert commit
      committer: Optional committer for revert commit

    Returns:
      SHA1 of the new revert commit, or None if no_commit=True

    Raises:
      Error: If revert fails due to conflicts or other issues
    """
    from .merge import three_way_merge

    # Normalize commits to a list
    if isinstance(commits, (str, bytes, Commit, Tag)):
        commits = [commits]

    with open_repo_closing(repo) as r:
        # Convert string refs to bytes
        commits_to_revert = []
        for commit_ref in commits:
            if isinstance(commit_ref, str):
                commit_ref = commit_ref.encode("utf-8")
            commit = parse_commit(r, commit_ref)
            commits_to_revert.append(commit)

        # Get current HEAD
        try:
            head_commit_id = r.refs[HEADREF]
        except KeyError:
            raise Error("No HEAD reference found")

        head_commit = r[head_commit_id]
        assert isinstance(head_commit, Commit)
        current_tree = head_commit.tree

        # Process commits in order
        for commit_to_revert in commits_to_revert:
            # For revert, we want to apply the inverse of the commit
            # This means using the commit's tree as "base" and its parent as "theirs"

            if not commit_to_revert.parents:
                raise Error(
                    f"Cannot revert commit {commit_to_revert.id.decode() if isinstance(commit_to_revert.id, bytes) else commit_to_revert.id} - it has no parents"
                )

            # For simplicity, we only handle commits with one parent (no merge commits)
            if len(commit_to_revert.parents) > 1:
                raise Error(
                    f"Cannot revert merge commit {commit_to_revert.id.decode() if isinstance(commit_to_revert.id, bytes) else commit_to_revert.id} - not yet implemented"
                )

            parent_commit = r[commit_to_revert.parents[0]]
            assert isinstance(parent_commit, Commit)

            # Perform three-way merge:
            # - base: the commit we're reverting (what we want to remove)
            # - ours: current HEAD (what we have now)
            # - theirs: parent of commit being reverted (what we want to go back to)
            assert isinstance(commit_to_revert, Commit)
            head_for_merge = r[head_commit_id]
            assert isinstance(head_for_merge, Commit)
            merged_tree, conflicts = three_way_merge(
                r.object_store,
                commit_to_revert,  # base
                head_for_merge,  # ours
                parent_commit,  # theirs
            )

            if conflicts:
                # Update working tree with conflicts
                changes = tree_changes(r.object_store, current_tree, merged_tree.id)
                update_working_tree(
                    r, current_tree, merged_tree.id, change_iterator=changes
                )
                conflicted_paths = [c.decode("utf-8", "replace") for c in conflicts]
                raise Error(f"Conflicts while reverting: {', '.join(conflicted_paths)}")

            # Add merged tree to object store
            r.object_store.add_object(merged_tree)

            # Update working tree
            changes = tree_changes(r.object_store, current_tree, merged_tree.id)
            update_working_tree(
                r, current_tree, merged_tree.id, change_iterator=changes
            )
            current_tree = merged_tree.id

            if not no_commit:
                # Create revert commit
                revert_commit = Commit()
                revert_commit.tree = merged_tree.id
                revert_commit.parents = [head_commit_id]

                # Set author/committer
                if author is None:
                    author = get_user_identity(r.get_config_stack())
                if committer is None:
                    committer = author

                revert_commit.author = author
                revert_commit.committer = committer

                # Set timestamps
                timestamp = int(time.time())
                timezone = 0  # UTC
                revert_commit.author_time = timestamp
                revert_commit.author_timezone = timezone
                revert_commit.commit_time = timestamp
                revert_commit.commit_timezone = timezone

                # Set message
                if message is None:
                    # Extract original commit subject
                    original_message = commit_to_revert.message
                    if isinstance(original_message, bytes):
                        original_message = original_message.decode("utf-8", "replace")
                    subject = original_message.split("\n")[0]
                    message = f'Revert "{subject}"\n\nThis reverts commit {commit_to_revert.id.decode("ascii")}.'.encode()
                elif isinstance(message, str):
                    message = message.encode("utf-8")

                revert_commit.message = message

                # Add commit to object store
                r.object_store.add_object(revert_commit)

                # Update HEAD
                r.refs[HEADREF] = revert_commit.id
                head_commit_id = revert_commit.id

        return head_commit_id if not no_commit else None


def gc(
    repo: RepoPath,
    auto: bool = False,
    aggressive: bool = False,
    prune: bool = True,
    grace_period: int | None = 1209600,  # 2 weeks default
    dry_run: bool = False,
    progress: Callable[[str], None] | None = None,
) -> "GCStats":
    """Run garbage collection on a repository.

    Args:
      repo: Path to the repository or a Repo object
      auto: If True, only run gc if needed
      aggressive: If True, use more aggressive settings
      prune: If True, prune unreachable objects
      grace_period: Grace period in seconds for pruning (default 2 weeks)
      dry_run: If True, only report what would be done
      progress: Optional progress callback

    Returns:
      GCStats object with garbage collection statistics
    """
    from .gc import garbage_collect

    with open_repo_closing(repo) as r:
        return garbage_collect(
            r,
            auto=auto,
            aggressive=aggressive,
            prune=prune,
            grace_period=grace_period,
            dry_run=dry_run,
            progress=progress,
        )


def prune(
    repo: RepoPath,
    grace_period: int | None = None,
    dry_run: bool = False,
    progress: Callable[[str], None] | None = None,
) -> None:
    """Prune/clean up a repository's object store.

    This removes temporary files that were left behind by interrupted
    pack operations.

    Args:
      repo: Path to the repository or a Repo object
      grace_period: Grace period in seconds for removing temporary files
                    (default 2 weeks)
      dry_run: If True, only report what would be done
      progress: Optional progress callback
    """
    with open_repo_closing(repo) as r:
        if progress:
            progress("Pruning temporary files")
        if not dry_run:
            r.object_store.prune(grace_period=grace_period)


def maintenance_run(
    repo: RepoPath,
    tasks: list[str] | None = None,
    auto: bool = False,
    progress: Callable[[str], None] | None = None,
) -> "MaintenanceResult":
    """Run maintenance tasks on a repository.

    Args:
      repo: Path to the repository or a Repo object
      tasks: Optional list of specific task names to run
             (e.g., ['gc', 'commit-graph', 'pack-refs'])
      auto: If True, only run tasks if needed
      progress: Optional progress callback

    Returns:
      MaintenanceResult object with task execution results
    """
    from .maintenance import run_maintenance

    with open_repo_closing(repo) as r:
        return run_maintenance(r, tasks=tasks, auto=auto, progress=progress)


def maintenance_register(repo: RepoPath) -> None:
    """Register a repository for background maintenance.

    This adds the repository to the global maintenance.repo config and sets
    up recommended configuration for scheduled maintenance.

    Args:
      repo: Path to the repository or repository object
    """
    from .maintenance import register_repository

    with open_repo_closing(repo) as r:
        register_repository(r)


def maintenance_unregister(repo: RepoPath, force: bool = False) -> None:
    """Unregister a repository from background maintenance.

    This removes the repository from the global maintenance.repo config.

    Args:
      repo: Path to the repository or repository object
      force: If True, don't error if repository is not registered
    """
    from .maintenance import unregister_repository

    with open_repo_closing(repo) as r:
        unregister_repository(r, force=force)


def count_objects(repo: RepoPath = ".", verbose: bool = False) -> CountObjectsResult:
    """Count unpacked objects and their disk usage.

    Args:
      repo: Path to repository or repository object
      verbose: Whether to return verbose information

    Returns:
      CountObjectsResult object with detailed statistics
    """
    with open_repo_closing(repo) as r:
        object_store = r.object_store

        # Count loose objects
        loose_count = 0
        loose_size = 0
        for sha in object_store._iter_loose_objects():
            loose_count += 1
            from .object_store import DiskObjectStore

            assert isinstance(object_store, DiskObjectStore)
            path = object_store._get_shafile_path(sha)
            try:
                stat_info = os.stat(path)
                # Git uses disk usage, not file size. st_blocks is always in
                # 512-byte blocks per POSIX standard
                st_blocks = getattr(stat_info, "st_blocks", None)
                if st_blocks is not None:
                    # Available on Linux and macOS
                    loose_size += st_blocks * 512
                else:
                    # Fallback for Windows
                    loose_size += stat_info.st_size
            except FileNotFoundError:
                # Object may have been removed between iteration and stat
                pass

        if not verbose:
            return CountObjectsResult(count=loose_count, size=loose_size)

        # Count pack information
        pack_count = len(object_store.packs)
        in_pack_count = 0
        pack_size = 0

        for pack in object_store.packs:
            in_pack_count += len(pack)
            # Get pack file size
            pack_path = pack._data_path
            try:
                pack_size += os.path.getsize(pack_path)
            except FileNotFoundError:
                pass
            # Get index file size
            idx_path = pack._idx_path
            try:
                pack_size += os.path.getsize(idx_path)
            except FileNotFoundError:
                pass

        return CountObjectsResult(
            count=loose_count,
            size=loose_size,
            in_pack=in_pack_count,
            packs=pack_count,
            size_pack=pack_size,
        )


def is_interactive_rebase(repo: Repo | str) -> bool:
    """Check if an interactive rebase is in progress.

    Args:
      repo: Repository to check

    Returns:
      True if interactive rebase is in progress, False otherwise
    """
    with open_repo_closing(repo) as r:
        state_manager = r.get_rebase_state_manager()
        if not state_manager.exists():
            return False

        # Check if todo file exists
        todo = state_manager.load_todo()
        return todo is not None


def rebase(
    repo: Repo | str,
    upstream: bytes | str,
    onto: bytes | str | None = None,
    branch: bytes | str | None = None,
    abort: bool = False,
    continue_rebase: bool = False,
    skip: bool = False,
    interactive: bool = False,
    edit_todo: bool = False,
) -> list[bytes]:
    """Rebase commits onto another branch.

    Args:
      repo: Repository to rebase in
      upstream: Upstream branch/commit to rebase onto
      onto: Specific commit to rebase onto (defaults to upstream)
      branch: Branch to rebase (defaults to current branch)
      abort: Abort an in-progress rebase
      continue_rebase: Continue an in-progress rebase
      skip: Skip current commit and continue rebase
      interactive: Start an interactive rebase
      edit_todo: Edit the todo list of an interactive rebase

    Returns:
      List of new commit SHAs created by rebase

    Raises:
      Error: If rebase fails or conflicts occur
    """
    # TODO: Avoid importing from .cli
    from .cli import launch_editor
    from .rebase import (
        RebaseConflict,
        RebaseError,
        Rebaser,
        process_interactive_rebase,
        start_interactive,
    )
    from .rebase import (
        edit_todo as edit_todo_func,
    )

    with open_repo_closing(repo) as r:
        rebaser = Rebaser(r)

        if abort:
            try:
                rebaser.abort()
                return []
            except RebaseError as e:
                raise Error(str(e))

        if edit_todo:
            # Edit the todo list of an interactive rebase
            try:
                edit_todo_func(r, launch_editor)
                print("Todo list updated. Continue with 'rebase --continue'")
                return []
            except RebaseError as e:
                raise Error(str(e))

        if continue_rebase:
            try:
                if interactive:
                    # Continue interactive rebase
                    is_complete, pause_reason = process_interactive_rebase(
                        r, editor_callback=launch_editor
                    )
                    if is_complete:
                        return [c.id for c in rebaser._done]
                    else:
                        if pause_reason == "conflict":
                            raise Error("Conflicts detected. Resolve and continue.")
                        elif pause_reason == "edit":
                            print("Stopped for editing. Make changes and continue.")
                        elif pause_reason == "break":
                            print("Rebase paused at break. Continue when ready.")
                        else:
                            print(f"Rebase paused: {pause_reason}")
                        return []
                else:
                    # Continue regular rebase
                    result = rebaser.continue_()
                    if result is None:
                        # Rebase complete
                        return [c.id for c in rebaser._done]
                    elif isinstance(result, tuple) and result[1]:
                        # Still have conflicts
                        raise Error(
                            f"Conflicts in: {', '.join(f.decode('utf-8', 'replace') for f in result[1])}"
                        )
            except RebaseError as e:
                raise Error(str(e))

        # Convert string refs to bytes
        if isinstance(upstream, str):
            upstream = upstream.encode("utf-8")
        if isinstance(onto, str):
            onto = onto.encode("utf-8") if onto else None
        if isinstance(branch, str):
            branch = branch.encode("utf-8") if branch else None

        try:
            if interactive:
                # Start interactive rebase
                todo = start_interactive(r, upstream, onto, branch, launch_editor)

                # Process the todo list
                is_complete, pause_reason = process_interactive_rebase(
                    r, todo, editor_callback=launch_editor
                )

                if is_complete:
                    return [c.id for c in rebaser._done]
                else:
                    if pause_reason == "conflict":
                        raise Error("Conflicts detected. Resolve and continue.")
                    elif pause_reason == "edit":
                        print("Stopped for editing. Make changes and continue.")
                    elif pause_reason == "break":
                        print("Rebase paused at break. Continue when ready.")
                    else:
                        print(f"Rebase paused: {pause_reason}")
                    return []
            else:
                # Regular rebase
                rebaser.start(upstream, onto, branch)

                # Continue rebase automatically
                result = rebaser.continue_()
                if result is not None:
                    # Conflicts
                    raise RebaseConflict(result[1])

                # Return the SHAs of the rebased commits
                return [c.id for c in rebaser._done]

        except RebaseConflict as e:
            raise Error(str(e))
        except RebaseError as e:
            raise Error(str(e))


def annotate(
    repo: RepoPath,
    path: str | bytes,
    committish: str | bytes | Commit | Tag | None = None,
) -> list[tuple[tuple[Commit, TreeEntry], bytes]]:
    """Annotate the history of a file.

    :param repo: Path to the repository
    :param path: Path to annotate
    :param committish: Commit id to find path in
    :return: List of ((Commit, TreeChange), line) tuples
    """
    if committish is None:
        committish = "HEAD"
    from .annotate import annotate_lines

    with open_repo_closing(repo) as r:
        commit_id = parse_commit(r, committish).id
        # Ensure path is bytes
        if isinstance(path, str):
            path = path.encode()
        return annotate_lines(r.object_store, commit_id, path)


blame = annotate


def filter_branch(
    repo: RepoPath = ".",
    branch: str | bytes = "HEAD",
    *,
    filter_fn: Callable[[Commit], "CommitData | None"] | None = None,
    filter_author: Callable[[bytes], bytes | None] | None = None,
    filter_committer: Callable[[bytes], bytes | None] | None = None,
    filter_message: Callable[[bytes], bytes | None] | None = None,
    tree_filter: Callable[[ObjectID, str], ObjectID | None] | None = None,
    index_filter: Callable[[ObjectID, str], ObjectID | None] | None = None,
    parent_filter: Callable[[Sequence[ObjectID]], list[ObjectID]] | None = None,
    commit_filter: Callable[[Commit, ObjectID], ObjectID | None] | None = None,
    subdirectory_filter: str | bytes | None = None,
    prune_empty: bool = False,
    tag_name_filter: Callable[[bytes], bytes | None] | None = None,
    force: bool = False,
    keep_original: bool = True,
    refs: list[bytes] | None = None,
) -> dict[ObjectID, ObjectID]:
    """Rewrite branch history by creating new commits with filtered properties.

    This is similar to git filter-branch, allowing you to rewrite commit
    history by modifying trees, parents, author, committer, or commit messages.

    Args:
      repo: Path to repository
      branch: Branch to rewrite (defaults to HEAD)
      filter_fn: Optional callable that takes a Commit object and returns
        a dict of updated fields (author, committer, message, etc.)
      filter_author: Optional callable that takes author bytes and returns
        updated author bytes or None to keep unchanged
      filter_committer: Optional callable that takes committer bytes and returns
        updated committer bytes or None to keep unchanged
      filter_message: Optional callable that takes commit message bytes
        and returns updated message bytes
      tree_filter: Optional callable that takes (tree_sha, temp_dir) and returns
        new tree SHA after modifying working directory
      index_filter: Optional callable that takes (tree_sha, temp_index_path) and
        returns new tree SHA after modifying index
      parent_filter: Optional callable that takes parent list and returns
        modified parent list
      commit_filter: Optional callable that takes (Commit, tree_sha) and returns
        new commit SHA or None to skip commit
      subdirectory_filter: Optional subdirectory path to extract as new root
      prune_empty: Whether to prune commits that become empty
      tag_name_filter: Optional callable to rename tags
      force: Force operation even if branch has been filtered before
      keep_original: Keep original refs under refs/original/
      refs: List of refs to rewrite (defaults to [branch])

    Returns:
      Dict mapping old commit SHAs to new commit SHAs

    Raises:
      Error: If branch is already filtered and force is False
    """
    from .filter_branch import CommitFilter, filter_refs

    with open_repo_closing(repo) as r:
        # Parse branch/committish
        if isinstance(branch, str):
            branch = branch.encode()

        # Determine which refs to process
        if refs is None:
            if branch == b"HEAD":
                # Resolve HEAD to actual branch
                try:
                    resolved = r.refs.follow(HEADREF)
                    if resolved and resolved[0]:
                        # resolved is a list of (refname, sha) tuples
                        resolved_ref = resolved[0][-1]
                        if resolved_ref and resolved_ref != b"HEAD":
                            refs = [resolved_ref]
                        else:
                            # HEAD points directly to a commit
                            refs = [b"HEAD"]
                    else:
                        refs = [b"HEAD"]
                except SymrefLoop:
                    refs = [b"HEAD"]
            else:
                # Convert branch name to full ref if needed
                if not branch.startswith(b"refs/"):
                    branch = local_branch_name(branch)
                refs = [branch]

        # Convert subdirectory filter to bytes if needed
        if subdirectory_filter:
            if isinstance(subdirectory_filter, str):
                subdirectory_filter = subdirectory_filter.encode()
        else:
            subdirectory_filter = None

        # Create commit filter
        filter_obj = CommitFilter(
            r.object_store,
            filter_fn=filter_fn,
            filter_author=filter_author,
            filter_committer=filter_committer,
            filter_message=filter_message,
            tree_filter=tree_filter,
            index_filter=index_filter,
            parent_filter=parent_filter,
            commit_filter=commit_filter,
            subdirectory_filter=subdirectory_filter,
            prune_empty=prune_empty,
            tag_name_filter=tag_name_filter,
        )

        # Tag callback for renaming tags
        def rename_tag(old_ref: Ref, new_ref: Ref) -> None:
            # Copy tag to new name
            r.refs[new_ref] = r.refs[old_ref]
            # Delete old tag
            del r.refs[old_ref]

        # Filter refs
        try:
            return filter_refs(
                r.refs,
                r.object_store,
                refs,
                filter_obj,
                keep_original=keep_original,
                force=force,
                tag_callback=rename_tag if tag_name_filter else None,
            )
        except ValueError as e:
            raise Error(str(e)) from e


def format_patch(
    repo: RepoPath = ".",
    committish: ObjectID | tuple[ObjectID, ObjectID] | None = None,
    outstream: TextIO = sys.stdout,
    outdir: str | os.PathLike[str] | None = None,
    n: int = 1,
    stdout: bool = False,
    version: str | None = None,
) -> list[str]:
    """Generate patches suitable for git am.

    Args:
      repo: Path to repository
      committish: Commit-ish or commit range to generate patches for.
        Can be a single commit id, or a tuple of (start, end) commit ids
        for a range. If None, formats the last n commits from HEAD.
      outstream: Stream to write to if stdout=True
      outdir: Directory to write patch files to (default: current directory)
      n: Number of patches to generate if committish is None
      stdout: Write patches to stdout instead of files
      version: Version string to include in patches (default: Dulwich version)

    Returns:
      List of patch filenames that were created (empty if stdout=True)
    """
    if outdir is None:
        outdir = "."

    filenames = []

    with open_repo_closing(repo) as r:
        # Determine which commits to format
        commits_to_format = []

        if committish is None:
            # Get the last n commits from HEAD
            try:
                walker = r.get_walker()
                for entry in walker:
                    commits_to_format.append(entry.commit)
                    if len(commits_to_format) >= n:
                        break
                commits_to_format.reverse()
            except KeyError:
                # No HEAD or empty repository
                pass
        elif isinstance(committish, tuple):
            # Handle commit range (start, end)
            start_commit, end_commit = committish

            # Extract commit IDs from commit objects if needed
            start_id = (
                start_commit.id if isinstance(start_commit, Commit) else start_commit
            )
            end_id = end_commit.id if isinstance(end_commit, Commit) else end_commit

            # Walk from end back to start
            walker = r.get_walker(include=[end_id], exclude=[start_id])
            for entry in walker:
                commits_to_format.append(entry.commit)
            commits_to_format.reverse()
        else:
            # Single commit
            commit = r.object_store[committish]
            assert isinstance(commit, Commit)
            commits_to_format.append(commit)

        # Generate patches
        total = len(commits_to_format)
        for i, commit in enumerate(commits_to_format, 1):
            assert isinstance(commit, Commit)
            # Get the parent
            if commit.parents:
                parent_id = commit.parents[0]
                parent = r.object_store[parent_id]
                assert isinstance(parent, Commit)
            else:
                parent = None

            # Generate the diff
            from io import BytesIO

            diff_content = BytesIO()
            if parent:
                write_tree_diff(
                    diff_content,
                    r.object_store,
                    parent.tree,
                    commit.tree,
                )
            else:
                # Initial commit - diff against empty tree
                write_tree_diff(
                    diff_content,
                    r.object_store,
                    None,
                    commit.tree,
                )

            # Generate patch with commit metadata
            if stdout:
                # Get binary stream from TextIO
                if hasattr(outstream, "buffer"):
                    binary_out: IO[bytes] = outstream.buffer
                else:
                    # Fallback for non-text streams
                    binary_out = outstream  # type: ignore[assignment]
                write_commit_patch(
                    binary_out,
                    commit,
                    diff_content.getvalue(),
                    (i, total),
                    version=version,
                )
            else:
                # Generate filename
                summary = get_summary(commit)
                filename = os.path.join(outdir, f"{i:04d}-{summary}.patch")

                with open(filename, "wb") as f:
                    write_commit_patch(
                        f,
                        commit,
                        diff_content.getvalue(),
                        (i, total),
                        version=version,
                    )
                filenames.append(filename)

    return filenames


def bisect_start(
    repo: str | os.PathLike[str] | Repo = ".",
    bad: str | bytes | Commit | Tag | None = None,
    good: str
    | bytes
    | Commit
    | Tag
    | Sequence[str | bytes | Commit | Tag]
    | None = None,
    paths: Sequence[bytes] | None = None,
    no_checkout: bool = False,
    term_bad: str = "bad",
    term_good: str = "good",
) -> bytes | None:
    """Start a new bisect session.

    Args:
        repo: Path to repository or a Repo object
        bad: The bad commit (defaults to HEAD)
        good: List of good commits or a single good commit
        paths: Optional paths to limit bisect to
        no_checkout: If True, don't checkout commits during bisect
        term_bad: Term to use for bad commits (default: "bad")
        term_good: Term to use for good commits (default: "good")
    """
    with open_repo_closing(repo) as r:
        state = BisectState(r)

        # Convert single good commit to sequence
        if good is not None and isinstance(good, (str, bytes, Commit, Tag)):
            good = [good]

        # Parse commits
        bad_sha = parse_commit(r, bad).id if bad else None
        good_shas = [parse_commit(r, g).id for g in good] if good else None

        state.start(bad_sha, good_shas, paths, no_checkout, term_bad, term_good)

        # Return the next commit to test if we have both good and bad
        if bad_sha and good_shas:
            next_sha = state._find_next_commit()
            if next_sha and not no_checkout:
                # Checkout the next commit
                old_commit = r[r.head()]
                assert isinstance(old_commit, Commit)
                old_tree = old_commit.tree if r.head() else None
                r.refs[HEADREF] = next_sha
                commit = r[next_sha]
                assert isinstance(commit, Commit)
                changes = tree_changes(r.object_store, old_tree, commit.tree)
                update_working_tree(r, old_tree, commit.tree, change_iterator=changes)
            return next_sha
        return None


def bisect_bad(
    repo: str | os.PathLike[str] | Repo = ".",
    rev: str | bytes | Commit | Tag | None = None,
) -> bytes | None:
    """Mark a commit as bad.

    Args:
        repo: Path to repository or a Repo object
        rev: Commit to mark as bad (defaults to HEAD)

    Returns:
        The SHA of the next commit to test, or None if bisect is complete
    """
    with open_repo_closing(repo) as r:
        state = BisectState(r)
        rev_sha = parse_commit(r, rev).id if rev else None
        next_sha = state.mark_bad(rev_sha)

        if next_sha:
            # Checkout the next commit
            old_commit = r[r.head()]
            assert isinstance(old_commit, Commit)
            old_tree = old_commit.tree if r.head() else None
            r.refs[HEADREF] = next_sha
            commit = r[next_sha]
            assert isinstance(commit, Commit)
            changes = tree_changes(r.object_store, old_tree, commit.tree)
            update_working_tree(r, old_tree, commit.tree, change_iterator=changes)

        return next_sha


def bisect_good(
    repo: str | os.PathLike[str] | Repo = ".",
    rev: str | bytes | Commit | Tag | None = None,
) -> bytes | None:
    """Mark a commit as good.

    Args:
        repo: Path to repository or a Repo object
        rev: Commit to mark as good (defaults to HEAD)

    Returns:
        The SHA of the next commit to test, or None if bisect is complete
    """
    with open_repo_closing(repo) as r:
        state = BisectState(r)
        rev_sha = parse_commit(r, rev).id if rev else None
        next_sha = state.mark_good(rev_sha)

        if next_sha:
            # Checkout the next commit
            old_commit = r[r.head()]
            assert isinstance(old_commit, Commit)
            old_tree = old_commit.tree if r.head() else None
            r.refs[HEADREF] = next_sha
            commit = r[next_sha]
            assert isinstance(commit, Commit)
            changes = tree_changes(r.object_store, old_tree, commit.tree)
            update_working_tree(r, old_tree, commit.tree, change_iterator=changes)

        return next_sha


def bisect_skip(
    repo: str | os.PathLike[str] | Repo = ".",
    revs: str
    | bytes
    | Commit
    | Tag
    | Sequence[str | bytes | Commit | Tag]
    | None = None,
) -> bytes | None:
    """Skip one or more commits.

    Args:
        repo: Path to repository or a Repo object
        revs: List of commits to skip (defaults to [HEAD])

    Returns:
        The SHA of the next commit to test, or None if bisect is complete
    """
    with open_repo_closing(repo) as r:
        state = BisectState(r)

        if revs is None:
            rev_shas = None
        else:
            # Convert single rev to sequence
            if isinstance(revs, (str, bytes, Commit, Tag)):
                revs = [revs]
            rev_shas = [parse_commit(r, rev).id for rev in revs]

        next_sha = state.skip(rev_shas)

        if next_sha:
            # Checkout the next commit
            old_commit = r[r.head()]
            assert isinstance(old_commit, Commit)
            old_tree = old_commit.tree if r.head() else None
            r.refs[HEADREF] = next_sha
            commit = r[next_sha]
            assert isinstance(commit, Commit)
            changes = tree_changes(r.object_store, old_tree, commit.tree)
            update_working_tree(r, old_tree, commit.tree, change_iterator=changes)

        return next_sha


def bisect_reset(
    repo: str | os.PathLike[str] | Repo = ".",
    commit: str | bytes | Commit | Tag | None = None,
) -> None:
    """Reset bisect state and return to original branch/commit.

    Args:
        repo: Path to repository or a Repo object
        commit: Optional commit to reset to (defaults to original branch/commit)
    """
    with open_repo_closing(repo) as r:
        state = BisectState(r)
        # Get old tree before reset
        try:
            old_commit = r[r.head()]
            assert isinstance(old_commit, Commit)
            old_tree = old_commit.tree
        except KeyError:
            old_tree = None

        commit_sha = parse_commit(r, commit).id if commit else None
        state.reset(commit_sha)

        # Update working tree to new HEAD
        try:
            new_head = r.head()
            if new_head:
                new_commit = r[new_head]
                assert isinstance(new_commit, Commit)
                changes = tree_changes(r.object_store, old_tree, new_commit.tree)
                update_working_tree(
                    r, old_tree, new_commit.tree, change_iterator=changes
                )
        except KeyError:
            # No HEAD after reset
            pass


def bisect_log(repo: str | os.PathLike[str] | Repo = ".") -> str:
    """Get the bisect log.

    Args:
        repo: Path to repository or a Repo object

    Returns:
        The bisect log as a string
    """
    with open_repo_closing(repo) as r:
        state = BisectState(r)
        return state.get_log()


def bisect_replay(
    repo: str | os.PathLike[str] | Repo,
    log_file: str | os.PathLike[str] | BinaryIO,
) -> None:
    """Replay a bisect log.

    Args:
        repo: Path to repository or a Repo object
        log_file: Path to the log file or file-like object
    """
    with open_repo_closing(repo) as r:
        state = BisectState(r)

        if isinstance(log_file, (str, os.PathLike)):
            with open(log_file) as f:
                log_content = f.read()
        else:
            content = log_file.read()
            log_content = content.decode() if isinstance(content, bytes) else content

        state.replay(log_content)


def reflog(
    repo: RepoPath = ".", ref: str | bytes = b"HEAD", all: bool = False
) -> Iterator[Any | tuple[bytes, Any]]:
    """Show reflog entries for a reference or all references.

    Args:
        repo: Path to repository or a Repo object
        ref: Reference name (defaults to HEAD)
        all: If True, show reflogs for all refs (ignores ref parameter)

    Yields:
        If all=False: ReflogEntry objects
        If all=True: Tuples of (ref_name, ReflogEntry) for all refs with reflogs
    """
    import os

    from .reflog import iter_reflogs

    if isinstance(ref, str):
        ref = ref.encode("utf-8")

    with open_repo_closing(repo) as r:
        if not all:
            yield from r.read_reflog(ref)
        else:
            logs_dir = os.path.join(r.controldir(), "logs")
            # Use iter_reflogs to discover all reflogs
            for ref_bytes in iter_reflogs(logs_dir):
                # Read the reflog entries for this ref
                for entry in r.read_reflog(ref_bytes):
                    yield (ref_bytes, entry)


def reflog_expire(
    repo: RepoPath = ".",
    ref: str | bytes | None = None,
    all: bool = False,
    expire_time: int | None = None,
    expire_unreachable_time: int | None = None,
    dry_run: bool = False,
) -> dict[bytes, int]:
    """Expire reflog entries based on age and reachability.

    Args:
        repo: Path to repository or a Repo object
        ref: Reference name (if not using --all)
        all: If True, expire reflogs for all refs
        expire_time: Expire entries older than this timestamp (seconds since epoch)
        expire_unreachable_time: Expire unreachable entries older than this timestamp
        dry_run: If True, show what would be expired without making changes

    Returns:
        Dictionary mapping ref names to number of expired entries
    """
    import os
    import time

    from .reflog import expire_reflog, iter_reflogs

    if not all and ref is None:
        raise ValueError("Must specify either ref or all=True")

    if isinstance(ref, str):
        ref = ref.encode("utf-8")

    # Default expire times if not specified
    if expire_time is None and expire_unreachable_time is None:
        # Default: expire entries older than 90 days, unreachable older than 30 days
        now = int(time.time())
        expire_time = now - (90 * 24 * 60 * 60)
        expire_unreachable_time = now - (30 * 24 * 60 * 60)

    result = {}

    with open_repo_closing(repo) as r:
        # Determine which refs to process
        refs_to_process: list[bytes] = []
        if all:
            logs_dir = os.path.join(r.controldir(), "logs")
            refs_to_process = list(iter_reflogs(logs_dir))
        else:
            assert ref is not None  # Already checked above
            refs_to_process = [ref]

        # Build set of reachable objects if we have unreachable expiration time
        reachable_objects: set[ObjectID] | None = None
        if expire_unreachable_time is not None:
            from .gc import find_reachable_objects

            reachable_objects = find_reachable_objects(
                r.object_store, r.refs, include_reflogs=False
            )

        # Process each ref
        for ref_name in refs_to_process:
            reflog_path = r._reflog_path(ref_name)
            if not os.path.exists(reflog_path):
                continue

            # Create reachability checker
            def is_reachable(sha: bytes) -> bool:
                if reachable_objects is None:
                    # No unreachable expiration, so assume everything is reachable
                    return True
                return sha in reachable_objects

            # Open the reflog file
            if dry_run:
                # For dry run, just read and count what would be expired
                with open(reflog_path, "rb") as f:
                    from .reflog import read_reflog

                    count = 0
                    for entry in read_reflog(f):
                        is_obj_reachable = is_reachable(entry.new_sha)
                        should_expire = False

                        if is_obj_reachable and expire_time is not None:
                            if entry.timestamp < expire_time:
                                should_expire = True
                        elif (
                            not is_obj_reachable and expire_unreachable_time is not None
                        ):
                            if entry.timestamp < expire_unreachable_time:
                                should_expire = True

                        if should_expire:
                            count += 1

                    result[ref_name] = count
            else:
                # Actually expire entries
                with open(reflog_path, "r+b") as f:  # type: ignore[assignment]
                    count = expire_reflog(
                        f,
                        expire_time=expire_time,
                        expire_unreachable_time=expire_unreachable_time,
                        reachable_checker=is_reachable,
                    )
                    result[ref_name] = count

    return result


def reflog_delete(
    repo: RepoPath = ".",
    ref: str | bytes = b"HEAD",
    index: int = 0,
    rewrite: bool = False,
) -> None:
    """Delete a specific reflog entry.

    Args:
        repo: Path to repository or a Repo object
        ref: Reference name
        index: Reflog entry index (0 = newest, in Git reflog order)
        rewrite: If True, rewrite old_sha of subsequent entries to maintain consistency
    """
    import os

    from .reflog import drop_reflog_entry

    if isinstance(ref, str):
        ref = ref.encode("utf-8")

    with open_repo_closing(repo) as r:
        reflog_path = r._reflog_path(ref)
        if not os.path.exists(reflog_path):
            raise ValueError(f"No reflog for ref {ref.decode()}")

        with open(reflog_path, "r+b") as f:
            drop_reflog_entry(f, index, rewrite=rewrite)


def lfs_track(
    repo: str | os.PathLike[str] | Repo = ".",
    patterns: Sequence[str] | None = None,
) -> list[str]:
    """Track file patterns with Git LFS.

    Args:
      repo: Path to repository
      patterns: List of file patterns to track (e.g., ["*.bin", "*.pdf"])
                If None, returns current tracked patterns

    Returns:
      List of tracked patterns
    """
    from .attrs import GitAttributes

    with open_repo_closing(repo) as r:
        gitattributes_path = os.path.join(r.path, ".gitattributes")

        # Load existing GitAttributes
        if os.path.exists(gitattributes_path):
            gitattributes = GitAttributes.from_file(gitattributes_path)
        else:
            gitattributes = GitAttributes()

        if patterns is None:
            # Return current LFS tracked patterns
            tracked = []
            for pattern_obj, attrs in gitattributes:
                if attrs.get(b"filter") == b"lfs":
                    tracked.append(pattern_obj.pattern.decode())
            return tracked

        # Add new patterns
        for pattern in patterns:
            # Ensure pattern is bytes
            pattern_bytes = pattern.encode() if isinstance(pattern, str) else pattern

            # Set LFS attributes for the pattern
            gitattributes.set_attribute(pattern_bytes, b"filter", b"lfs")
            gitattributes.set_attribute(pattern_bytes, b"diff", b"lfs")
            gitattributes.set_attribute(pattern_bytes, b"merge", b"lfs")
            gitattributes.set_attribute(pattern_bytes, b"text", False)

        # Write updated attributes
        gitattributes.write_to_file(gitattributes_path)

        # Stage the .gitattributes file
        add(r, [".gitattributes"])

        return lfs_track(r)  # Return updated list


def lfs_untrack(
    repo: str | os.PathLike[str] | Repo = ".",
    patterns: Sequence[str] | None = None,
) -> list[str]:
    """Untrack file patterns from Git LFS.

    Args:
      repo: Path to repository
      patterns: List of file patterns to untrack

    Returns:
      List of remaining tracked patterns
    """
    from .attrs import GitAttributes

    if not patterns:
        return lfs_track(repo)

    with open_repo_closing(repo) as r:
        gitattributes_path = os.path.join(r.path, ".gitattributes")

        if not os.path.exists(gitattributes_path):
            return []

        # Load existing GitAttributes
        gitattributes = GitAttributes.from_file(gitattributes_path)

        # Remove specified patterns
        for pattern in patterns:
            pattern_bytes = pattern.encode() if isinstance(pattern, str) else pattern

            # Check if pattern is tracked by LFS
            for pattern_obj, attrs in list(gitattributes):
                if (
                    pattern_obj.pattern == pattern_bytes
                    and attrs.get(b"filter") == b"lfs"
                ):
                    gitattributes.remove_pattern(pattern_bytes)
                    break

        # Write updated attributes
        gitattributes.write_to_file(gitattributes_path)

        # Stage the .gitattributes file
        add(r, [".gitattributes"])

        return lfs_track(r)  # Return updated list


def lfs_init(repo: str | os.PathLike[str] | Repo = ".") -> None:
    """Initialize Git LFS in a repository.

    Args:
      repo: Path to repository

    Returns:
      None
    """
    from .lfs import LFSStore

    with open_repo_closing(repo) as r:
        # Create LFS store
        LFSStore.from_repo(r, create=True)

        # Set up Git config for LFS
        config = r.get_config()
        config.set((b"filter", b"lfs"), b"process", b"git-lfs filter-process")
        config.set((b"filter", b"lfs"), b"required", b"true")
        config.set((b"filter", b"lfs"), b"clean", b"git-lfs clean -- %f")
        config.set((b"filter", b"lfs"), b"smudge", b"git-lfs smudge -- %f")
        config.write_to_path()


def lfs_clean(
    repo: str | os.PathLike[str] | Repo = ".",
    path: str | os.PathLike[str] | None = None,
) -> bytes:
    """Clean a file by converting it to an LFS pointer.

    Args:
      repo: Path to repository
      path: Path to file to clean (relative to repo root)

    Returns:
      LFS pointer content as bytes
    """
    from .lfs import LFSFilterDriver, LFSStore

    with open_repo_closing(repo) as r:
        if path is None:
            raise ValueError("Path must be specified")

        # Get LFS store
        lfs_store = LFSStore.from_repo(r)
        filter_driver = LFSFilterDriver(lfs_store, config=r.get_config())

        # Read file content
        full_path = os.path.join(r.path, path)
        with open(full_path, "rb") as f:
            content = f.read()

        # Clean the content (convert to LFS pointer)
        return filter_driver.clean(content)


def lfs_smudge(
    repo: str | os.PathLike[str] | Repo = ".",
    pointer_content: bytes | None = None,
) -> bytes:
    """Smudge an LFS pointer by retrieving the actual content.

    Args:
      repo: Path to repository
      pointer_content: LFS pointer content as bytes

    Returns:
      Actual file content as bytes
    """
    from .lfs import LFSFilterDriver, LFSStore

    with open_repo_closing(repo) as r:
        if pointer_content is None:
            raise ValueError("Pointer content must be specified")

        # Get LFS store
        lfs_store = LFSStore.from_repo(r)
        filter_driver = LFSFilterDriver(lfs_store, config=r.get_config())

        # Smudge the pointer (retrieve actual content)
        return filter_driver.smudge(pointer_content)


def lfs_ls_files(
    repo: str | os.PathLike[str] | Repo = ".",
    ref: str | bytes | None = None,
) -> list[tuple[bytes, str, int]]:
    """List files tracked by Git LFS.

    Args:
      repo: Path to repository
      ref: Git ref to check (defaults to HEAD)

    Returns:
      List of (path, oid, size) tuples for LFS files
    """
    from .lfs import LFSPointer
    from .object_store import iter_tree_contents

    with open_repo_closing(repo) as r:
        if ref is None:
            ref = b"HEAD"
        elif isinstance(ref, str):
            ref = ref.encode()

        # Get the commit and tree
        try:
            commit = r[ref]
            assert isinstance(commit, Commit)
            tree = r[commit.tree]
            assert isinstance(tree, Tree)
        except KeyError:
            return []

        lfs_files = []

        # Walk the tree
        for path, mode, sha in iter_tree_contents(r.object_store, tree.id):
            assert path is not None
            assert mode is not None
            assert sha is not None
            if not stat.S_ISREG(mode):
                continue

            # Check if it's an LFS pointer
            obj = r.object_store[sha]
            if not isinstance(obj, Blob):
                raise AssertionError(f"Expected Blob object, got {type(obj).__name__}")
            pointer = LFSPointer.from_bytes(obj.data)
            if pointer is not None:
                lfs_files.append((path, pointer.oid, pointer.size))

        return lfs_files


def lfs_migrate(
    repo: str | os.PathLike[str] | Repo = ".",
    include: list[str] | None = None,
    exclude: list[str] | None = None,
    everything: bool = False,
) -> int:
    """Migrate files to Git LFS.

    Args:
      repo: Path to repository
      include: Patterns of files to include
      exclude: Patterns of files to exclude
      everything: Migrate all files above a certain size

    Returns:
      Number of migrated files
    """
    from .lfs import LFSFilterDriver, LFSStore

    with open_repo_closing(repo) as r:
        # Initialize LFS if needed
        lfs_store = LFSStore.from_repo(r, create=True)
        filter_driver = LFSFilterDriver(lfs_store, config=r.get_config())

        # Get current index
        index = r.open_index()

        migrated = 0

        # Determine files to migrate
        files_to_migrate = []

        if everything:
            # Migrate all files above 100MB
            for path, entry in index.items():
                full_path = os.path.join(r.path, path.decode())
                if os.path.exists(full_path):
                    size = os.path.getsize(full_path)
                    if size > 100 * 1024 * 1024:  # 100MB
                        files_to_migrate.append(path.decode())
        else:
            # Use include/exclude patterns
            for path, entry in index.items():
                path_str = path.decode()

                # Check include patterns
                if include:
                    matched = any(
                        fnmatch.fnmatch(path_str, pattern) for pattern in include
                    )
                    if not matched:
                        continue

                # Check exclude patterns
                if exclude:
                    excluded = any(
                        fnmatch.fnmatch(path_str, pattern) for pattern in exclude
                    )
                    if excluded:
                        continue

                files_to_migrate.append(path_str)

        # Migrate files
        for path_str in files_to_migrate:
            full_path = os.path.join(r.path, path_str)
            if not os.path.exists(full_path):
                continue

            # Read file content
            with open(full_path, "rb") as f:
                content = f.read()

            # Convert to LFS pointer
            pointer_content = filter_driver.clean(content)

            # Write pointer back to file
            with open(full_path, "wb") as f:
                f.write(pointer_content)

            # Create blob for pointer content and update index
            blob = Blob()
            blob.data = pointer_content
            r.object_store.add_object(blob)

            st = os.stat(full_path)
            index_entry = index_entry_from_stat(st, blob.id, 0)
            path_bytes = path_str.encode() if isinstance(path_str, str) else path_str
            index[path_bytes] = index_entry

            migrated += 1

        # Write updated index
        index.write()

        # Track patterns if include was specified
        if include:
            lfs_track(r, include)

        return migrated


def lfs_pointer_check(
    repo: str | os.PathLike[str] | Repo = ".",
    paths: Sequence[str] | None = None,
) -> dict[str, Any | None]:
    """Check if files are valid LFS pointers.

    Args:
      repo: Path to repository
      paths: List of file paths to check (if None, check all files)

    Returns:
      Dict mapping paths to LFSPointer objects (or None if not a pointer)
    """
    from .lfs import LFSPointer

    with open_repo_closing(repo) as r:
        results = {}

        if paths is None:
            # Check all files in index
            index = r.open_index()
            paths = [path.decode() for path in index]

        for path in paths:
            full_path = os.path.join(r.path, path)
            if os.path.exists(full_path):
                try:
                    with open(full_path, "rb") as f:
                        content = f.read()
                    pointer = LFSPointer.from_bytes(content)
                    results[path] = pointer
                except OSError:
                    results[path] = None
            else:
                results[path] = None

        return results


def lfs_fetch(
    repo: str | os.PathLike[str] | Repo = ".",
    remote: str = "origin",
    refs: list[str | bytes] | None = None,
) -> int:
    """Fetch LFS objects from remote.

    Args:
      repo: Path to repository
      remote: Remote name (default: origin)
      refs: Specific refs to fetch LFS objects for (default: all refs)

    Returns:
      Number of objects fetched
    """
    from .lfs import LFSClient, LFSPointer, LFSStore

    with open_repo_closing(repo) as r:
        # Get LFS server URL from config
        config = r.get_config()
        lfs_url_bytes = config.get((b"lfs",), b"url")
        if not lfs_url_bytes:
            # Try remote URL
            remote_url = config.get((b"remote", remote.encode()), b"url")
            if remote_url:
                # Append /info/lfs to remote URL
                remote_url_str = remote_url.decode()
                if remote_url_str.endswith(".git"):
                    remote_url_str = remote_url_str[:-4]
                lfs_url = f"{remote_url_str}/info/lfs"
            else:
                raise ValueError(f"No LFS URL configured for remote {remote}")
        else:
            lfs_url = lfs_url_bytes.decode()

        # Get authentication
        auth = None
        # TODO: Support credential helpers and other auth methods

        # Create LFS client and store
        client = LFSClient(lfs_url, auth)
        store = LFSStore.from_repo(r)

        # Find all LFS pointers in the refs
        pointers_to_fetch = []

        if refs is None:
            # Get all refs
            refs = list(r.refs.keys())

        for ref in refs:
            if isinstance(ref, str):
                ref_key = Ref(ref.encode())
            elif isinstance(ref, bytes):
                ref_key = Ref(ref)
            else:
                ref_key = ref
            try:
                commit = r[r.refs[ref_key]]
            except KeyError:
                continue

            # Walk the commit tree
            assert isinstance(commit, Commit)
            for path, mode, sha in r.object_store.iter_tree_contents(commit.tree):
                assert sha is not None
                try:
                    obj = r.object_store[sha]
                except KeyError:
                    pass
                else:
                    if isinstance(obj, Blob):
                        pointer = LFSPointer.from_bytes(obj.data)
                        if pointer and pointer.is_valid_oid():
                            # Check if we already have it
                            try:
                                with store.open_object(pointer.oid):
                                    pass  # Object exists, no need to fetch
                            except KeyError:
                                pointers_to_fetch.append((pointer.oid, pointer.size))

        # Fetch missing objects
        fetched = 0
        for oid, size in pointers_to_fetch:
            content = client.download(oid, size)
            store.write_object([content])
            fetched += 1

        return fetched


def lfs_pull(repo: str | os.PathLike[str] | Repo = ".", remote: str = "origin") -> int:
    """Pull LFS objects for current checkout.

    Args:
      repo: Path to repository
      remote: Remote name (default: origin)

    Returns:
      Number of objects fetched
    """
    from .lfs import LFSPointer, LFSStore

    with open_repo_closing(repo) as r:
        # First do a fetch for HEAD
        fetched = lfs_fetch(repo, remote, [b"HEAD"])

        # Then checkout LFS files in working directory
        store = LFSStore.from_repo(r)
        index = r.open_index()

        for path, entry in index.items():
            full_path = os.path.join(r.path, path.decode())
            if os.path.exists(full_path):
                with open(full_path, "rb") as f:
                    content = f.read()

                pointer = LFSPointer.from_bytes(content)
                if pointer and pointer.is_valid_oid():
                    try:
                        # Replace pointer with actual content
                        with store.open_object(pointer.oid) as lfs_file:
                            lfs_content = lfs_file.read()
                        with open(full_path, "wb") as f:
                            f.write(lfs_content)
                    except KeyError:
                        # Object not available
                        pass

        return fetched


def lfs_push(
    repo: str | os.PathLike[str] | Repo = ".",
    remote: str = "origin",
    refs: list[str | bytes] | None = None,
) -> int:
    """Push LFS objects to remote.

    Args:
      repo: Path to repository
      remote: Remote name (default: origin)
      refs: Specific refs to push LFS objects for (default: current branch)

    Returns:
      Number of objects pushed
    """
    from .lfs import LFSClient, LFSPointer, LFSStore

    with open_repo_closing(repo) as r:
        # Get LFS server URL from config
        config = r.get_config()
        lfs_url_bytes = config.get((b"lfs",), b"url")
        if not lfs_url_bytes:
            # Try remote URL
            remote_url = config.get((b"remote", remote.encode()), b"url")
            if remote_url:
                # Append /info/lfs to remote URL
                remote_url_str = remote_url.decode()
                if remote_url_str.endswith(".git"):
                    remote_url_str = remote_url_str[:-4]
                lfs_url = f"{remote_url_str}/info/lfs"
            else:
                raise ValueError(f"No LFS URL configured for remote {remote}")
        else:
            lfs_url = lfs_url_bytes.decode()

        # Get authentication
        auth = None
        # TODO: Support credential helpers and other auth methods

        # Create LFS client and store
        client = LFSClient(lfs_url, auth)
        store = LFSStore.from_repo(r)

        # Find all LFS objects to push
        if refs is None:
            # Push current branch
            head_ref = r.refs.read_ref(HEADREF)
            refs = [head_ref] if head_ref else []

        objects_to_push = set()

        for ref in refs:
            if isinstance(ref, str):
                ref_bytes = ref.encode()
            else:
                ref_bytes = ref
            try:
                if ref_bytes.startswith(b"refs/"):
                    commit = r[r.refs[Ref(ref_bytes)]]
                else:
                    commit = r[ref_bytes]
            except KeyError:
                continue

            # Walk the commit tree
            assert isinstance(commit, Commit)
            for path, mode, sha in r.object_store.iter_tree_contents(commit.tree):
                assert sha is not None
                try:
                    obj = r.object_store[sha]
                except KeyError:
                    pass
                else:
                    if isinstance(obj, Blob):
                        pointer = LFSPointer.from_bytes(obj.data)
                        if pointer and pointer.is_valid_oid():
                            objects_to_push.add((pointer.oid, pointer.size))

        # Push objects
        pushed = 0
        for oid, size in objects_to_push:
            try:
                with store.open_object(oid) as f:
                    content = f.read()
            except KeyError:
                # Object not in local store
                logging.warn("LFS object %s not found locally", oid)
            else:
                client.upload(oid, size, content)
                pushed += 1

        return pushed


def lfs_status(repo: str | os.PathLike[str] | Repo = ".") -> dict[str, list[str]]:
    """Show status of LFS files.

    Args:
      repo: Path to repository

    Returns:
      Dict with status information
    """
    from .lfs import LFSPointer, LFSStore

    with open_repo_closing(repo) as r:
        store = LFSStore.from_repo(r)
        index = r.open_index()

        status: dict[str, list[str]] = {
            "tracked": [],
            "not_staged": [],
            "not_committed": [],
            "not_pushed": [],
            "missing": [],
        }

        # Check working directory files
        for path, entry in index.items():
            path_str = path.decode()
            full_path = os.path.join(r.path, path_str)

            if os.path.exists(full_path):
                with open(full_path, "rb") as f:
                    content = f.read()

                pointer = LFSPointer.from_bytes(content)
                if pointer and pointer.is_valid_oid():
                    status["tracked"].append(path_str)

                    # Check if object exists locally
                    try:
                        with store.open_object(pointer.oid):
                            pass  # Object exists locally
                    except KeyError:
                        status["missing"].append(path_str)

                    # Check if file has been modified
                    if isinstance(entry, ConflictedIndexEntry):
                        continue  # Skip conflicted entries
                    try:
                        staged_obj = r.object_store[entry.sha]
                    except KeyError:
                        pass
                    else:
                        if not isinstance(staged_obj, Blob):
                            raise AssertionError(
                                f"Expected Blob object, got {type(staged_obj).__name__}"
                            )
                        staged_pointer = LFSPointer.from_bytes(staged_obj.data)
                        if staged_pointer and staged_pointer.oid != pointer.oid:
                            status["not_staged"].append(path_str)

        # TODO: Check for not committed and not pushed files

        return status


def worktree_list(repo: RepoPath = ".") -> list[Any]:
    """List all worktrees for a repository.

    Args:
        repo: Path to repository

    Returns:
        List of WorkTreeInfo objects
    """
    from .worktree import list_worktrees

    with open_repo_closing(repo) as r:
        return list_worktrees(r)


def worktree_add(
    repo: RepoPath = ".",
    path: str | os.PathLike[str] | None = None,
    branch: str | bytes | None = None,
    commit: str | bytes | None = None,
    detach: bool = False,
    force: bool = False,
) -> str:
    """Add a new worktree.

    Args:
        repo: Path to repository
        path: Path for new worktree
        branch: Branch to checkout (creates if doesn't exist)
        commit: Specific commit to checkout
        detach: Create with detached HEAD
        force: Force creation even if branch is already checked out

    Returns:
        Path to the newly created worktree
    """
    from .worktree import add_worktree

    if path is None:
        raise ValueError("Path is required for worktree add")

    with open_repo_closing(repo) as r:
        commit_bytes = commit.encode() if isinstance(commit, str) else commit
        commit_id = ObjectID(commit_bytes) if commit_bytes is not None else None
        wt_repo = add_worktree(
            r, path, branch=branch, commit=commit_id, detach=detach, force=force
        )
        return wt_repo.path


def worktree_remove(
    repo: RepoPath = ".",
    path: str | os.PathLike[str] | None = None,
    force: bool = False,
) -> None:
    """Remove a worktree.

    Args:
        repo: Path to repository
        path: Path to worktree to remove
        force: Force removal even if there are local changes
    """
    from .worktree import remove_worktree

    if path is None:
        raise ValueError("Path is required for worktree remove")

    with open_repo_closing(repo) as r:
        remove_worktree(r, path, force=force)


def worktree_prune(
    repo: RepoPath = ".", dry_run: bool = False, expire: int | None = None
) -> list[str]:
    """Prune worktree administrative files.

    Args:
        repo: Path to repository
        dry_run: Only show what would be removed
        expire: Only prune worktrees older than this many seconds

    Returns:
        List of pruned worktree names
    """
    from .worktree import prune_worktrees

    with open_repo_closing(repo) as r:
        return prune_worktrees(r, expire=expire, dry_run=dry_run)


def worktree_lock(
    repo: RepoPath = ".",
    path: str | os.PathLike[str] | None = None,
    reason: str | None = None,
) -> None:
    """Lock a worktree to prevent it from being pruned.

    Args:
        repo: Path to repository
        path: Path to worktree to lock
        reason: Optional reason for locking
    """
    from .worktree import lock_worktree

    if path is None:
        raise ValueError("Path is required for worktree lock")

    with open_repo_closing(repo) as r:
        lock_worktree(r, path, reason=reason)


def worktree_unlock(
    repo: RepoPath = ".", path: str | os.PathLike[str] | None = None
) -> None:
    """Unlock a worktree.

    Args:
        repo: Path to repository
        path: Path to worktree to unlock
    """
    from .worktree import unlock_worktree

    if path is None:
        raise ValueError("Path is required for worktree unlock")

    with open_repo_closing(repo) as r:
        unlock_worktree(r, path)


def worktree_move(
    repo: RepoPath = ".",
    old_path: str | os.PathLike[str] | None = None,
    new_path: str | os.PathLike[str] | None = None,
) -> None:
    """Move a worktree to a new location.

    Args:
        repo: Path to repository
        old_path: Current path of worktree
        new_path: New path for worktree
    """
    from .worktree import move_worktree

    if old_path is None or new_path is None:
        raise ValueError("Both old_path and new_path are required for worktree move")

    with open_repo_closing(repo) as r:
        move_worktree(r, old_path, new_path)


def worktree_repair(
    repo: RepoPath = ".",
    paths: list[str | os.PathLike[str]] | None = None,
) -> list[str]:
    """Repair worktree administrative files.

    Args:
        repo: Path to repository
        paths: Optional list of worktree paths to repair. If None, repairs
               connections from the main repository to all linked worktrees.

    Returns:
        List of repaired worktree paths
    """
    from .worktree import repair_worktree

    with open_repo_closing(repo) as r:
        return repair_worktree(r, paths=paths)


def merge_base(
    repo: RepoPath = ".",
    committishes: Sequence[str | bytes] | None = None,
    all: bool = False,
    octopus: bool = False,
) -> list[ObjectID]:
    """Find the best common ancestor(s) between commits.

    Args:
        repo: Path to repository
        committishes: List of commit references (branches, tags, commit IDs)
        all: If True, return all merge bases, not just one
        octopus: If True, find merge base of all commits (n-way merge)

    Returns:
        List of commit IDs that are merge bases
    """
    from .graph import find_merge_base, find_octopus_base
    from .objectspec import parse_object

    if committishes is None or len(committishes) < 2:
        raise ValueError("At least two commits are required")

    with open_repo_closing(repo) as r:
        # Resolve committish references to commit IDs
        commit_ids = []
        for committish in committishes:
            obj = parse_object(r, committish)
            if not isinstance(obj, Commit):
                raise ValueError(f"Expected commit, got {obj.type_name.decode()}")
            commit_ids.append(obj.id)

        # Find merge base
        if octopus:
            result = find_octopus_base(r, commit_ids)
        else:
            result = find_merge_base(r, commit_ids)

        # Return first result only if all=False
        if not all and result:
            return [result[0]]
        return result


def is_ancestor(
    repo: RepoPath = ".",
    ancestor: str | bytes | None = None,
    descendant: str | bytes | None = None,
) -> bool:
    """Check if one commit is an ancestor of another.

    Args:
        repo: Path to repository
        ancestor: Commit that might be the ancestor
        descendant: Commit that might be the descendant

    Returns:
        True if ancestor is an ancestor of descendant, False otherwise
    """
    from .graph import find_merge_base
    from .objectspec import parse_object

    if ancestor is None or descendant is None:
        raise ValueError("Both ancestor and descendant are required")

    with open_repo_closing(repo) as r:
        # Resolve committish references to commit IDs
        ancestor_obj = parse_object(r, ancestor)
        if not isinstance(ancestor_obj, Commit):
            raise ValueError(f"Expected commit, got {ancestor_obj.type_name.decode()}")
        descendant_obj = parse_object(r, descendant)
        if not isinstance(descendant_obj, Commit):
            raise ValueError(
                f"Expected commit, got {descendant_obj.type_name.decode()}"
            )

        # If ancestor is the merge base of (ancestor, descendant), then it's an ancestor
        merge_bases = find_merge_base(r, [ancestor_obj.id, descendant_obj.id])
        return merge_bases == [ancestor_obj.id]


def independent_commits(
    repo: RepoPath = ".",
    committishes: Sequence[str | bytes] | None = None,
) -> list[ObjectID]:
    """Filter commits to only those that are not reachable from others.

    Args:
        repo: Path to repository
        committishes: List of commit references to filter

    Returns:
        List of commit IDs that are not ancestors of any other commits in the list
    """
    from .graph import independent
    from .objectspec import parse_object

    if committishes is None or len(committishes) == 0:
        return []

    with open_repo_closing(repo) as r:
        # Resolve committish references to commit IDs
        commit_ids = []
        for committish in committishes:
            obj = parse_object(r, committish)
            if not isinstance(obj, Commit):
                raise ValueError(f"Expected commit, got {obj.type_name.decode()}")
            commit_ids.append(obj.id)

        # Filter to independent commits
        return independent(r, commit_ids)


def mailsplit(
    input_path: str | os.PathLike[str] | IO[bytes] | None = None,
    output_dir: str | os.PathLike[str] = ".",
    start_number: int = 1,
    precision: int = 4,
    keep_cr: bool = False,
    mboxrd: bool = False,
    is_maildir: bool = False,
) -> list[str]:
    r"""Split an mbox file or Maildir into individual message files.

    This is similar to git mailsplit.

    Args:
        input_path: Path to mbox file, Maildir, or file-like object. If None, reads from stdin.
        output_dir: Directory where individual messages will be written
        start_number: Starting number for output files (default: 1)
        precision: Number of digits for output filenames (default: 4)
        keep_cr: If True, preserve \r in lines ending with \r\n (default: False)
        mboxrd: If True, treat input as mboxrd format and reverse escaping (default: False)
        is_maildir: If True, treat input_path as a Maildir (default: False)

    Returns:
        List of output file paths that were created

    Raises:
        ValueError: If output_dir doesn't exist or input is invalid
        OSError: If there are issues reading/writing files
    """
    from .mbox import split_maildir, split_mbox

    if is_maildir:
        if input_path is None:
            raise ValueError("input_path is required for Maildir splitting")
        if not isinstance(input_path, (str, bytes, os.PathLike)):
            raise ValueError("Maildir splitting requires a path, not a file object")
        # Convert PathLike to str for split_maildir
        maildir_path: str | bytes = (
            os.fspath(input_path) if isinstance(input_path, os.PathLike) else input_path
        )
        out_dir: str | bytes = (
            os.fspath(output_dir) if isinstance(output_dir, os.PathLike) else output_dir
        )
        return split_maildir(
            maildir_path,
            out_dir,
            start_number=start_number,
            precision=precision,
            keep_cr=keep_cr,
        )
    else:
        if input_path is None:
            # Read from stdin
            input_file: str | bytes | BinaryIO = sys.stdin.buffer
        else:
            # Convert PathLike to str if needed
            if isinstance(input_path, os.PathLike):
                input_file = os.fspath(input_path)
            else:
                # input_path is either str or IO[bytes] here
                input_file = cast(str | BinaryIO, input_path)

        out_dir = (
            os.fspath(output_dir) if isinstance(output_dir, os.PathLike) else output_dir
        )
        return split_mbox(
            input_file,
            out_dir,
            start_number=start_number,
            precision=precision,
            keep_cr=keep_cr,
            mboxrd=mboxrd,
        )


def mailinfo(
    input_path: str | os.PathLike[str] | IO[bytes] | IO[str] | None = None,
    msg_file: str | os.PathLike[str] | None = None,
    patch_file: str | os.PathLike[str] | None = None,
    keep_subject: bool = False,
    keep_non_patch: bool = False,
    encoding: str | None = None,
    scissors: bool = False,
    message_id: bool = False,
) -> MailinfoResult:
    """Extract patch information from an email message.

    This is similar to git mailinfo.

    Args:
        input_path: Path to email file or file-like object. If None, reads from stdin.
        msg_file: Path to write commit message. If None, message not written to file.
        patch_file: Path to write patch content. If None, patch not written to file.
        keep_subject: If True, keep subject intact without munging (-k)
        keep_non_patch: If True, only strip [PATCH] from brackets (-b)
        encoding: Character encoding to use (default: detect from message)
        scissors: If True, remove everything before scissors line
        message_id: If True, include Message-ID in commit message (-m)

    Returns:
        MailinfoResult with parsed information

    Raises:
        ValueError: If message is malformed or missing required fields
        OSError: If there are issues reading/writing files

    Example:
        >>> result = mailinfo("patch.eml", "msg", "patch")
        >>> print(f"Author: {result.author_name} <{result.author_email}>")
        >>> print(f"Subject: {result.subject}")
    """
    from .mbox import mailinfo as mbox_mailinfo

    if input_path is None:
        # Read from stdin
        input_file: str | bytes | BinaryIO | TextIO = sys.stdin.buffer
    else:
        # Convert PathLike to str if needed
        if isinstance(input_path, os.PathLike):
            input_file = os.fspath(input_path)
        else:
            # input_path is either str or IO[bytes] or IO[str] here
            input_file = cast(str | BinaryIO | TextIO, input_path)

    result = mbox_mailinfo(
        input_file,
        keep_subject=keep_subject,
        keep_non_patch=keep_non_patch,
        encoding=encoding,
        scissors=scissors,
        message_id=message_id,
    )

    # Write message to file if requested
    if msg_file is not None:
        msg_path = (
            os.fspath(msg_file) if isinstance(msg_file, os.PathLike) else msg_file
        )
        with open(msg_path, "w", encoding=encoding or "utf-8") as f:
            f.write(result.message)
            if not result.message.endswith("\n"):
                f.write("\n")

    # Write patch to file if requested
    if patch_file is not None:
        patch_path = (
            os.fspath(patch_file) if isinstance(patch_file, os.PathLike) else patch_file
        )
        with open(patch_path, "w", encoding=encoding or "utf-8") as f:
            f.write(result.patch)

    return result


def rerere(repo: RepoPath = ".") -> tuple[list[tuple[bytes, str]], list[bytes]]:
    """Record current conflict resolutions and apply known resolutions.

    This reads conflicted files from the working tree and records them
    in the rerere cache. If rerere.autoupdate is enabled and a known
    resolution exists, it will be automatically applied.

    Args:
        repo: Path to the repository

    Returns:
        Tuple of:
        - List of tuples (path, conflict_id) for recorded conflicts
        - List of paths where resolutions were automatically applied
    """
    from .rerere import _has_conflict_markers, rerere_auto

    with open_repo_closing(repo) as r:
        # Get conflicts from the index (if available)
        index = r.open_index()
        conflicts = []

        for path, entry in index.items():
            if isinstance(entry, ConflictedIndexEntry):
                conflicts.append(path)

        # Also scan working tree for files with conflict markers
        # This is needed because merge() doesn't always create ConflictedIndexEntry
        if not conflicts:
            working_tree = r.path
            for path in index:
                file_path = os.path.join(working_tree, os.fsdecode(path))
                try:
                    with open(file_path, "rb") as f:
                        content = f.read()
                    if _has_conflict_markers(content):
                        conflicts.append(path)
                except (FileNotFoundError, IsADirectoryError, PermissionError):
                    pass

        # Record conflicts and apply known resolutions
        working_tree = r.path
        return rerere_auto(r, working_tree, conflicts)


def rerere_status(repo: RepoPath = ".") -> list[tuple[str, bool]]:
    """Get the status of all conflicts in the rerere cache.

    Args:
        repo: Path to the repository

    Returns:
        List of tuples (conflict_id, has_resolution)
    """
    from .rerere import RerereCache

    with open_repo_closing(repo) as r:
        cache = RerereCache.from_repo(r)
        return cache.status()


def rerere_diff(
    repo: RepoPath = ".", conflict_id: str | None = None
) -> list[tuple[str, bytes, bytes | None]]:
    """Show differences for recorded rerere conflicts.

    Args:
        repo: Path to the repository
        conflict_id: Optional specific conflict ID to show

    Returns:
        List of tuples (conflict_id, preimage, postimage)
    """
    from .rerere import RerereCache

    with open_repo_closing(repo) as r:
        cache = RerereCache.from_repo(r)

        if conflict_id:
            preimage, postimage = cache.diff(conflict_id)
            if preimage is not None:
                return [(conflict_id, preimage, postimage)]
            return []

        # Show all conflicts
        results = []
        for cid, _has_res in cache.status():
            preimage, postimage = cache.diff(cid)
            if preimage is not None:
                results.append((cid, preimage, postimage))
        return results


def rerere_forget(repo: RepoPath = ".", pathspec: str | bytes | None = None) -> None:
    """Forget recorded rerere resolutions for a pathspec.

    Args:
        repo: Path to the repository
        pathspec: Path to forget (currently not implemented, forgets all)
    """
    from .rerere import RerereCache

    with open_repo_closing(repo) as r:
        cache = RerereCache.from_repo(r)

        if pathspec:
            # TODO: Implement pathspec matching
            # For now, we need to track which conflict IDs correspond to which paths
            raise NotImplementedError("Pathspec matching not yet implemented")

        # Forget all conflicts (this is when called with no pathspec after resolving)
        cache.clear()


def rerere_clear(repo: RepoPath = ".") -> None:
    """Clear all recorded rerere resolutions.

    Args:
        repo: Path to the repository
    """
    from .rerere import RerereCache

    with open_repo_closing(repo) as r:
        cache = RerereCache.from_repo(r)
        cache.clear()


def rerere_gc(repo: RepoPath = ".", max_age_days: int = 60) -> None:
    """Garbage collect old rerere resolutions.

    Args:
        repo: Path to the repository
        max_age_days: Maximum age in days for keeping resolutions
    """
    from .rerere import RerereCache

    with open_repo_closing(repo) as r:
        cache = RerereCache.from_repo(r)
        cache.gc(max_age_days)