# Copyright (C) 2008, 2009 Michael Trier (mtrier@gmail.com) and contributors
#
# This module is part of GitPython and is released under the
# 3-Clause BSD License: https://opensource.org/license/bsd-3-clause/
from __future__ import annotations
__all__ = ["Git"]
import contextlib
import io
import itertools
import logging
import os
import re
import shutil
import signal
import subprocess
import sys
import threading
import time
from subprocess import DEVNULL, PIPE, Popen
from textwrap import dedent
# typing ---------------------------------------------------------------------------
from typing import (
IO,
TYPE_CHECKING,
Any,
AnyStr,
BinaryIO,
Callable,
Dict,
Iterator,
List,
Mapping,
Optional,
Sequence,
TextIO,
Tuple,
Union,
cast,
overload,
)
from git.compat import defenc, force_bytes, safe_decode
from git import _backend
from git.exc import (
CommandError,
GitCommandError,
GitCommandNotFound,
UnsafeOptionError,
UnsafeProtocolError,
UnsupportedOperation,
)
from git.util import (
cygpath,
expand_path,
is_cygwin_git,
patch_env,
remove_password_if_present,
stream_copy,
)
if sys.version_info >= (3, 10):
from typing import TypeAlias
else:
from typing_extensions import TypeAlias
from git.types import TBD, Literal, PathLike
if TYPE_CHECKING:
from git.diff import DiffIndex
from git.repo.base import Repo
# ---------------------------------------------------------------------------------
execute_kwargs = {
"istream",
"with_extended_output",
"with_exceptions",
"as_process",
"output_stream",
"stdout_as_string",
"kill_after_timeout",
"with_stdout",
"universal_newlines",
"shell",
"env",
"max_chunk_size",
"strip_newline_in_stdout",
}
_logger = logging.getLogger(__name__)
# ==============================================================================
## @name Utilities
# ------------------------------------------------------------------------------
# Documentation
## @{
def _kill_process(pid: int) -> bool:
"""Kill a POSIX process and its descendants, returning whether it was killed."""
# Collect descendants before signalling, while their parent PIDs still identify them.
pids = [pid]
for parent_pid in pids:
try:
try:
p = Popen(["pgrep", "-P", str(parent_pid)], stdout=PIPE)
except FileNotFoundError:
# POSIX ps does not support selecting by parent PID.
ps_args = ["ps", "-ef"] if sys.platform == "cygwin" else ["ps", "-A", "-o", "pid=", "-o", "ppid="]
with Popen(ps_args, stdout=PIPE) as p:
if p.stdout is not None:
for line in p.stdout:
fields = line.split()
if sys.platform == "cygwin":
fields = fields[1:3] # ps -ef starts with UID, PID, PPID.
if len(fields) == 2 and all(field.isdigit() for field in fields):
if int(fields[1]) == parent_pid:
pids.append(int(fields[0]))
else:
with p:
if p.stdout is not None:
for line in p.stdout:
if line.strip().isdigit():
pids.append(int(line))
except OSError as ex:
_logger.info("Unable to enumerate child processes: %r", ex)
killed = False
if sys.platform != "win32":
for process_pid in pids:
try:
os.kill(process_pid, signal.SIGKILL)
killed = killed or process_pid == pid
except OSError:
pass
return killed
def handle_process_output(
process: Union["Git.AutoInterrupt", Popen],
stdout_handler: Union[
None,
Callable[[AnyStr], None],
Callable[[List[AnyStr]], None],
Callable[[bytes, "Repo", "DiffIndex"], None],
],
stderr_handler: Union[None, Callable[[AnyStr], None], Callable[[List[AnyStr]], None]],
finalizer: Union[None, Callable[[Union[Popen, "Git.AutoInterrupt"]], None]] = None,
decode_streams: bool = True,
kill_after_timeout: Union[None, float] = None,
) -> None:
R"""Register for notifications to learn that process output is ready to read, and
dispatch lines to the respective line handlers.
This function returns once the finalizer returns.
:param process:
:class:`subprocess.Popen` instance.
:param stdout_handler:
f(stdout_line_string), or ``None``.
:param stderr_handler:
f(stderr_line_string), or ``None``.
:param finalizer:
f(proc) - wait for proc to finish.
:param decode_streams:
Assume stdout/stderr streams are binary and decode them before pushing their
contents to handlers.
This defaults to ``True``. Set it to ``False`` if:
- ``universal_newlines == True``, as then streams are in text mode, or
- decoding must happen later, such as for :class:`~git.diff.Diff`\s.
:param kill_after_timeout:
:class:`float` or ``None``, Default = ``None``
To specify a timeout in seconds for the git command, after which the process
should be killed.
"""
# Use 2 "pump" threads and wait for both to finish.
def pump_stream(
cmdline: List[str],
name: str,
stream: Union[BinaryIO, TextIO],
is_decode: bool,
handler: Union[None, Callable[[Union[bytes, str]], None]],
) -> None:
try:
for line in stream:
if handler:
if is_decode:
assert isinstance(line, bytes)
line_str = line.decode(defenc)
handler(line_str)
else:
handler(line)
except Exception as ex:
_logger.error(f"Pumping {name!r} of cmd({remove_password_if_present(cmdline)}) failed due to: {ex!r}")
if not (isinstance(ex, ValueError) and stream.closed):
# Only reraise if the error was not due to the stream closing.
raise CommandError([f""] + remove_password_if_present(cmdline), ex) from ex
finally:
stream.close()
if hasattr(process, "proc"):
process = cast("Git.AutoInterrupt", process)
cmdline: str | Tuple[str, ...] | List[str] = getattr(process.proc, "args", "")
p_stdout = process.proc.stdout if process.proc else None
p_stderr = process.proc.stderr if process.proc else None
else:
process = cast(Popen, process) # type: ignore[redundant-cast]
cmdline = getattr(process, "args", "")
p_stdout = process.stdout
p_stderr = process.stderr
if not isinstance(cmdline, (tuple, list)):
cmdline = cmdline.split()
pumps: List[Tuple[str, IO, Callable[..., None] | None]] = []
if p_stdout:
pumps.append(("stdout", p_stdout, stdout_handler))
if p_stderr:
pumps.append(("stderr", p_stderr, stderr_handler))
threads: List[threading.Thread] = []
for name, stream, handler in pumps:
t = threading.Thread(target=pump_stream, args=(cmdline, name, stream, decode_streams, handler))
t.daemon = True
t.start()
threads.append(t)
# Wait for output handlers to finish before finalizing.
# If the child needs stdin, the caller must arrange to feed/close it
# before this call or concurrently; this function only drains output.
deadline = None if kill_after_timeout is None else time.monotonic() + kill_after_timeout
for t in threads:
t.join(timeout=None if deadline is None else max(0, deadline - time.monotonic()))
if t.is_alive():
if isinstance(process, Git.AutoInterrupt):
if sys.platform != "win32" and process.proc is not None and process.proc.poll() is None:
_kill_process(process.proc.pid)
process._terminate()
else: # Don't want to deal with the other case.
raise RuntimeError(
"Thread join() timed out in cmd.handle_process_output()."
f" kill_after_timeout={kill_after_timeout} seconds"
)
process._timeout_error = (
f"error: process killed because it timed out. kill_after_timeout={kill_after_timeout} seconds"
)
break
if finalizer:
finalizer(process)
safer_popen: Callable[..., Popen]
if sys.platform == "win32":
def _safer_popen_windows(
command: Union[str, Sequence[Any]],
*,
shell: bool = False,
env: Optional[Mapping[str, str]] = None,
**kwargs: Any,
) -> Popen:
"""Call :class:`subprocess.Popen` on Windows but don't include a CWD in the
search.
This avoids an untrusted search path condition where a file like ``git.exe`` in
a malicious repository would be run when GitPython operates on the repository.
The process using GitPython may have an untrusted repository's working tree as
its current working directory. Some operations may temporarily change to that
directory before running a subprocess. In addition, while by default GitPython
does not run external commands with a shell, it can be made to do so, in which
case the CWD of the subprocess, which GitPython usually sets to a repository
working tree, can itself be searched automatically by the shell. This wrapper
covers all those cases.
:note:
This currently works by setting the
:envvar:`NoDefaultCurrentDirectoryInExePath` environment variable during
subprocess creation. It also takes care of passing Windows-specific process
creation flags, but that is unrelated to path search.
:note:
The current implementation contains a race condition on :attr:`os.environ`.
GitPython isn't thread-safe, but a program using it on one thread should
ideally be able to mutate :attr:`os.environ` on another, without
unpredictable results. See comments in:
https://github.com/gitpython-developers/GitPython/pull/1650
"""
# CREATE_NEW_PROCESS_GROUP is needed for some ways of killing it afterwards.
# https://docs.python.org/3/library/subprocess.html#subprocess.Popen.send_signal
# https://docs.python.org/3/library/subprocess.html#subprocess.CREATE_NEW_PROCESS_GROUP
creationflags = subprocess.CREATE_NO_WINDOW | subprocess.CREATE_NEW_PROCESS_GROUP
# When using a shell, the shell is the direct subprocess, so the variable must
# be set in its environment, to affect its search behavior.
if shell:
# The original may be immutable, or the caller may reuse it. Mutate a copy.
env = {} if env is None else dict(env)
env["NoDefaultCurrentDirectoryInExePath"] = "1" # The "1" can be any value.
# When not using a shell, the current process does the search in a
# CreateProcessW API call, so the variable must be set in our environment. With
# a shell, that's unnecessary if https://github.com/python/cpython/issues/101283
# is patched. In Python versions where it is unpatched, in the rare case the
# ComSpec environment variable is unset, the search for the shell itself is
# unsafe. Setting NoDefaultCurrentDirectoryInExePath in all cases, as done here,
# is simpler and protects against that. (As above, the "1" can be any value.)
with patch_env("NoDefaultCurrentDirectoryInExePath", "1"):
return Popen(
command,
shell=shell,
env=env,
creationflags=creationflags,
**kwargs,
)
safer_popen = _safer_popen_windows
else:
safer_popen = Popen
def dashify(string: str) -> str:
return string.replace("_", "-")
def slots_to_dict(self: "Git", exclude: Sequence[str] = ()) -> Dict[str, Any]:
return {s: getattr(self, s) for s in self.__slots__ if s not in exclude}
def dict_to_slots_and__excluded_are_none(self: object, d: Mapping[str, Any], excluded: Sequence[str] = ()) -> None:
for k, v in d.items():
setattr(self, k, v)
for k in excluded:
setattr(self, k, None)
## -- End Utilities -- @}
class _AutoInterrupt:
"""Process wrapper that terminates the wrapped process on finalization.
This kills/interrupts the stored process instance once this instance goes out of
scope. It is used to prevent processes piling up in case iterators stop reading.
All attributes are wired through to the contained process object.
The wait method is overridden to perform automatic status code checking and possibly
raise.
"""
__slots__ = ("proc", "args", "status", "_timeout_error")
# If this is non-zero it will override any status code during _terminate, used
# to prevent race conditions in testing.
_status_code_if_terminate: int = 0
def __init__(self, proc: Union[None, subprocess.Popen], args: Any) -> None:
self.proc = proc
self.args = args
self.status: Union[int, None] = None
self._timeout_error: Optional[str] = None
def _terminate(self) -> None:
"""Terminate the underlying process."""
if self.proc is None:
return
proc = self.proc
self.proc = None
try:
if proc.stdin:
# A timed-out process may already have exited before input is flushed.
with contextlib.suppress(BrokenPipeError):
proc.stdin.close()
# Did the process finish already so we have a return code?
try:
if proc.poll() is not None:
self.status = self._status_code_if_terminate or proc.poll()
return
except OSError as ex:
_logger.info("Ignored error after process had died: %r", ex)
# It can be that nothing really exists anymore...
if getattr(os, "kill", None) is None:
return
# Try to kill it.
try:
proc.terminate()
except (OSError, AttributeError) as ex:
# On interpreter shutdown (notably on Windows), parts of the stdlib used by
# subprocess can already be torn down (e.g. `subprocess._winapi` becomes None),
# which can cause AttributeError during terminate(). In that case, we prefer
# to silently ignore to avoid noisy "Exception ignored in: __del__" messages.
_logger.info("Ignored error while terminating process: %r", ex)
return
# END exception handling
finally:
if proc.stdout:
proc.stdout.close()
if proc.stderr:
proc.stderr.close()
try:
status = proc.wait()
self.status = self._status_code_if_terminate or status
except (OSError, AttributeError) as ex:
_logger.info("Ignored error while waiting for terminated process: %r", ex)
def __del__(self) -> None:
self._terminate()
def __getattr__(self, attr: str) -> Any:
return getattr(self.proc, attr)
# TODO: Bad choice to mimic `proc.wait()` but with different args.
def wait(self, stderr: Union[None, str, bytes] = b"") -> int:
"""Wait for the process and return its status code.
:param stderr:
Previously read value of stderr, in case stderr is already closed.
:warn:
May deadlock if output or error pipes are used and not handled separately.
:raise git.exc.GitCommandError:
If the return status is not 0 or output handling timed out.
"""
status: Union[int, None]
if self.proc is not None:
status = self.proc.wait()
p_stderr = self.proc.stderr
else: # Assume the underlying proc was killed earlier or never existed.
status = self.status
p_stderr = None
stderr_b = force_bytes(data=stderr, encoding="utf-8") or b""
if self._timeout_error is not None:
stderr_b += (b"\n" if stderr_b else b"") + self._timeout_error.encode("utf-8")
status = status or 1
def read_all_from_possibly_closed_stream(stream: Union[IO[bytes], None]) -> bytes:
if stream:
try:
return stderr_b + force_bytes(stream.read())
except (OSError, ValueError):
return stderr_b or b""
else:
return stderr_b or b""
# END status handling
if status != 0:
errstr = read_all_from_possibly_closed_stream(p_stderr)
_logger.debug("AutoInterrupt wait stderr: %r" % (errstr,))
raise GitCommandError(remove_password_if_present(self.args), status, errstr)
return status
_AutoInterrupt.__name__ = "AutoInterrupt"
_AutoInterrupt.__qualname__ = "Git.AutoInterrupt"
class _CatFileContentStream:
"""Object representing a sized read-only stream returning the contents of
an object.
This behaves like a stream, but counts the data read and simulates an empty stream
once our sized content region is empty.
If not all data are read to the end of the object's lifetime, we read the rest to
ensure the underlying stream continues to work.
"""
__slots__ = ("_stream", "_nbr", "_size")
def __init__(self, size: int, stream: IO[bytes]) -> None:
self._stream = stream
self._size = size
self._nbr = 0 # Number of bytes read.
# Special case: If the object is empty, has null bytes, get the final
# newline right away.
if size == 0:
stream.read(1)
# END handle empty streams
def read(self, size: int = -1) -> bytes:
bytes_left = self._size - self._nbr
if bytes_left == 0:
return b""
if size > -1:
# Ensure we don't try to read past our limit.
size = min(bytes_left, size)
else:
# They try to read all, make sure it's not more than what remains.
size = bytes_left
# END check early depletion
data = self._stream.read(size)
self._nbr += len(data)
# Check for depletion, read our final byte to make the stream usable by
# others.
if self._size - self._nbr == 0:
self._stream.read(1) # final newline
# END finish reading
return data
def readline(self, size: int = -1) -> bytes:
if self._nbr == self._size:
return b""
# Clamp size to lowest allowed value.
bytes_left = self._size - self._nbr
if size > -1:
size = min(bytes_left, size)
else:
size = bytes_left
# END handle size
data = self._stream.readline(size)
self._nbr += len(data)
# Handle final byte.
if self._size - self._nbr == 0:
self._stream.read(1)
# END finish reading
return data
def readlines(self, size: int = -1) -> List[bytes]:
if self._nbr == self._size:
return []
# Leave all additional logic to our readline method, we just check the size.
out = []
nbr = 0
while True:
line = self.readline()
if not line:
break
out.append(line)
if size > -1:
nbr += len(line)
if nbr > size:
break
# END handle size constraint
# END readline loop
return out
# skipcq: PYL-E0301
def __iter__(self) -> "Git.CatFileContentStream":
return self
def __next__(self) -> bytes:
line = self.readline()
if not line:
raise StopIteration
return line
next = __next__
def __del__(self) -> None:
bytes_left = self._size - self._nbr
if bytes_left:
# Read and discard - seeking is impossible within a stream.
# This includes any terminating newline.
self._stream.read(bytes_left + 1)
# END handle incomplete read
_CatFileContentStream.__name__ = "CatFileContentStream"
_CatFileContentStream.__qualname__ = "Git.CatFileContentStream"
class Git:
"""The Git class manages communication with the Git binary.
It provides a convenient interface to calling the Git binary, such as in::
g = Git( git_dir )
g.init() # calls 'git init' program
rval = g.ls_files() # calls 'git ls-files' program
Debugging:
* Set the :envvar:`GIT_PYTHON_TRACE` environment variable to print each invocation
of the command to stdout.
* Set its value to ``full`` to see details about the returned values.
"""
__slots__ = (
"_working_dir",
"cat_file_all",
"cat_file_header",
"_version_info",
"_version_info_token",
"_git_options",
"_persistent_git_options",
"_environment",
"_repository_environment",
"_repo",
)
_excluded_ = (
"cat_file_all",
"cat_file_header",
"_version_info",
"_version_info_token",
"_repo",
)
_version_info: Optional[Tuple[int, ...]]
_version_info_token: object
# Match Git's leading transport selector, including an empty helper name.
# Git also selects the command-executing ext helper for an ext:// URL.
re_unsafe_protocol = re.compile(r"([A-Za-z0-9][A-Za-z0-9+.-]*|)::|ext://")
unsafe_git_ls_remote_options = [
# This option allows arbitrary command execution in git-ls-remote.
"--upload-pack",
"--exec",
]
unsafe_git_pathspec_from_file_options = [
# Reads pathspecs from a caller-controlled file. Some commands include an
# unmatched pathspec in their error output, which can disclose the file.
"--pathspec-from-file",
]
def __getstate__(self) -> Dict[str, Any]:
return slots_to_dict(self, exclude=self._excluded_)
def __setstate__(self, d: Dict[str, Any]) -> None:
dict_to_slots_and__excluded_are_none(self, d, excluded=self._excluded_)
# CONFIGURATION
git_exec_name = "git"
"""Default git command that should work on Linux, Windows, and other systems."""
GIT_PYTHON_TRACE = os.environ.get("GIT_PYTHON_TRACE", False)
"""Enables debugging of GitPython's git commands."""
_git_exec_env_var = "GIT_PYTHON_GIT_EXECUTABLE"
_refresh_env_var = "GIT_PYTHON_REFRESH"
GIT_PYTHON_GIT_EXECUTABLE = None
"""Provide the full path to the git executable. Otherwise it assumes git is in the
executable search path.
:note:
The git executable is actually found during the refresh step in the top level
``__init__``. It can also be changed by explicitly calling :func:`git.refresh`.
"""
_refresh_token = object() # Since None would match an initial _version_info_token.
_version_check_cache: Dict[Tuple[Any, ...], Tuple[int, ...]] = {}
@classmethod
def refresh(cls, path: Union[None, PathLike] = None) -> bool:
"""Update information about the git executable :class:`Git` objects will use.
Called by the :func:`git.refresh` function in the top level ``__init__``.
:param path:
Optional path to the git executable. If not absolute, it is resolved
immediately, relative to the current directory. (See note below.)
:note:
The top-level :func:`git.refresh` should be preferred because it calls this
method and may also update other state accordingly.
:note:
There are three different ways to specify the command that refreshing causes
to be used for git:
1. Pass no `path` argument and do not set the
:envvar:`GIT_PYTHON_GIT_EXECUTABLE` environment variable. The command
name ``git`` is used. It is looked up in a path search by the system, in
each command run (roughly similar to how git is found when running
``git`` commands manually). This is usually the desired behavior.
2. Pass no `path` argument but set the :envvar:`GIT_PYTHON_GIT_EXECUTABLE`
environment variable. The command given as the value of that variable is
used. This may be a simple command or an arbitrary path. It is looked up
in each command run. Setting :envvar:`GIT_PYTHON_GIT_EXECUTABLE` to
``git`` has the same effect as not setting it.
3. Pass a `path` argument. This path, if not absolute, is immediately
resolved, relative to the current directory. This resolution occurs at
the time of the refresh. When git commands are run, they are run using
that previously resolved path. If a `path` argument is passed, the
:envvar:`GIT_PYTHON_GIT_EXECUTABLE` environment variable is not
consulted.
:note:
Refreshing always sets the :attr:`Git.GIT_PYTHON_GIT_EXECUTABLE` class
attribute, which can be read on the :class:`Git` class or any of its
instances to check what command is used to run git. This attribute should
not be confused with the related :envvar:`GIT_PYTHON_GIT_EXECUTABLE`
environment variable. The class attribute is set no matter how refreshing is
performed.
"""
# Discern which path to refresh with.
if path is not None:
new_git = os.path.expanduser(path)
new_git = os.path.abspath(new_git)
else:
new_git = os.environ.get(cls._git_exec_env_var, cls.git_exec_name)
# Keep track of the old and new git executable path.
old_git = cls.GIT_PYTHON_GIT_EXECUTABLE
old_refresh_token = cls._refresh_token
cls.GIT_PYTHON_GIT_EXECUTABLE = new_git
cls._refresh_token = object()
# Test if the new git executable path is valid. A GitCommandNotFound error is
# raised by us. A PermissionError is raised if the git executable cannot be
# executed for whatever reason.
has_git = False
try:
cls().version()
has_git = True
except (GitCommandNotFound, PermissionError):
pass
# Warn or raise exception if test failed.
if not has_git:
err = (
dedent(
"""\
Bad git executable.
The git executable must be specified in one of the following ways:
- be included in your $PATH
- be set via $%s
- explicitly set via git.refresh()
"""
)
% cls._git_exec_env_var
)
# Revert to whatever the old_git was.
cls.GIT_PYTHON_GIT_EXECUTABLE = old_git
cls._refresh_token = old_refresh_token
if old_git is None:
# On the first refresh (when GIT_PYTHON_GIT_EXECUTABLE is None) we only
# are quiet, warn, or error depending on the GIT_PYTHON_REFRESH value.
# Determine what the user wants to happen during the initial refresh. We
# expect GIT_PYTHON_REFRESH to either be unset or be one of the
# following values:
#
# 0|q|quiet|s|silence|silent|n|none
# 1|w|warn|warning|l|log
# 2|r|raise|e|error|exception
mode = os.environ.get(cls._refresh_env_var, "raise").lower()
quiet = ["quiet", "q", "silence", "s", "silent", "none", "n", "0"]
warn = ["warn", "w", "warning", "log", "l", "1"]
error = ["error", "e", "exception", "raise", "r", "2"]
if mode in quiet:
pass
elif mode in warn or mode in error:
err = dedent(
"""\
%s
All git commands will error until this is rectified.
This initial message can be silenced or aggravated in the future by setting the
$%s environment variable. Use one of the following values:
- %s: for no message or exception
- %s: for a warning message (logging level CRITICAL, displayed by default)
- %s: for a raised exception
Example:
export %s=%s
"""
) % (
err,
cls._refresh_env_var,
"|".join(quiet),
"|".join(warn),
"|".join(error),
cls._refresh_env_var,
quiet[0],
)
if mode in warn:
_logger.critical(err)
else:
raise ImportError(err)
else:
err = dedent(
"""\
%s environment variable has been set but it has been set with an invalid value.
Use only the following values:
- %s: for no message or exception
- %s: for a warning message (logging level CRITICAL, displayed by default)
- %s: for a raised exception
"""
) % (
cls._refresh_env_var,
"|".join(quiet),
"|".join(warn),
"|".join(error),
)
raise ImportError(err)
# We get here if this was the initial refresh and the refresh mode was
# not error. Go ahead and set the GIT_PYTHON_GIT_EXECUTABLE such that we
# discern the difference between the first refresh at import time
# and subsequent calls to git.refresh or this refresh method.
cls.GIT_PYTHON_GIT_EXECUTABLE = cls.git_exec_name
else:
# After the first refresh (when GIT_PYTHON_GIT_EXECUTABLE is no longer
# None) we raise an exception.
raise GitCommandNotFound(new_git, err)
return has_git
@classmethod
def is_cygwin(cls) -> bool:
return is_cygwin_git(cls.GIT_PYTHON_GIT_EXECUTABLE)
@overload
@classmethod
def polish_url(cls, url: str, is_cygwin: Literal[False] = ...) -> str: ...
@overload
@classmethod
def polish_url(cls, url: str, is_cygwin: Union[None, bool] = None) -> str: ...
@classmethod
def polish_url(cls, url: str, is_cygwin: Union[None, bool] = None) -> PathLike:
"""Remove any backslashes from URLs to be written in config files.
Windows might create config files containing paths with backslashes, but git
stops liking them as it will escape the backslashes. Hence we undo the escaping
just to be sure.
"""
if is_cygwin is None:
is_cygwin = cls.is_cygwin()
if is_cygwin:
url = cygpath(url)
else:
if url.startswith("~"):
url = os.path.expanduser(url)
url = url.replace("\\\\", "\\").replace("\\", "/")
return url
@classmethod
def check_unsafe_protocols(cls, url: str) -> None:
"""Check for unsafe protocols.
Apart from the usual protocols (http, git, ssh), Git allows "remote helpers"
that have the form ``::``. One of these helpers (``ext::``)
can be used to invoke any arbitrary command. Git also selects that helper
for ``ext://`` URLs and interprets the URL as a command path.
See:
- https://git-scm.com/docs/gitremote-helpers
- https://git-scm.com/docs/git-remote-ext
"""
match = cls.re_unsafe_protocol.match(url)
if match:
protocol = match.group(0)
raise UnsafeProtocolError(
f"The `{protocol}` protocol looks suspicious, use `allow_unsafe_protocols=True` to allow it."
)
def _check_unsafe_protocols_in_args(self, args: Sequence[Any], kwargs: Mapping[str, Any]) -> None:
"""Check positional operands and standalone values in rendered command options.
A short flag's split value can become the repository operand before ``--``.
"""
for arg in self._unpack_args(args) + self.transform_kwargs(**kwargs):
self.check_unsafe_protocols(arg)
@classmethod
def _canonicalize_option_name(cls, option: str) -> str:
"""Return the option name used for unsafe-option checks.
Examples:
``"--upload-pack=/tmp/helper"`` -> ``"upload-pack"``
``"upload_pack"`` -> ``"upload-pack"``
``"--config core.filemode=false"`` -> ``"config"``
"""
option_name = option.lstrip("-").split("=", 1)[0]
option_tokens = option_name.split(None, 1)
if not option_tokens:
return ""
return dashify(option_tokens[0])
@classmethod
def check_unsafe_options(
cls, options: List[str], unsafe_options: List[str], clusterable_short_options: str = "46flnqsv"
) -> None:
"""Raise :class:`~git.exc.UnsafeOptionError` for blocked option spellings.
In addition to exact matches, this rejects abbreviated long options accepted
by Git (for example, ``--upl`` for ``--upload-pack``) and unsafe short options
whose values are joined to the same token, including after clusterable flags
(for example, ``-uVALUE`` and ``-fuVALUE``).
A list containing only bare names is treated as normalized keyword arguments,
so multi-character names such as ``upload_p`` are checked as long-option
abbreviations. If any item starts with ``-``, the list is treated as tokenized
command-line input: bare items can be option values and are not checked as
abbreviations. Thus ``["--origin", "upload"]`` is allowed. Single-dash options
use short-option parsing rather than broad prefix matching, preserving safe
attached values such as ``-oupstream`` and ``-bcurrent``.
Some options passed to ``git `` can execute arbitrary commands and
are therefore blocked by default unless the caller explicitly allows them.
"""
# Options can be of the form `foo`, `--foo`, `--foo bar`, or `--foo=bar`.
# Git accepts any unambiguous prefix of a long option, so an abbreviated
# spelling such as `--upl` for `--upload-pack` must be rejected too. An
# option is unsafe if its canonical name is a prefix of any blocked
# option's canonical name. Only long options and multi-character kwargs
# can be abbreviations; single-character short options remain exact-match
# only.
canonical_unsafe_options = {cls._canonicalize_option_name(option): option for option in unsafe_options}
unsafe_short_options = {
canonical: option
for canonical, option in canonical_unsafe_options.items()
if option.startswith("-") and not option.startswith("--") and len(canonical) == 1
}
# These value-less Git flags can be clustered before another short option
# (for example, ``-fuVALUE``). Stop at any other character because it may
# begin an attached value, as ``o`` does in the safe option ``-oupstream``.
clusterable_short_options_set = frozenset(clusterable_short_options)
options_are_kwargs = all(not option.startswith("-") for option in options)
for option in options:
candidate = cls._canonicalize_option_name(option)
if not candidate:
continue
unsafe_option = canonical_unsafe_options.get(candidate)
if unsafe_option is not None:
raise UnsafeOptionError(f"{unsafe_option} is not allowed, use `allow_unsafe_options=True` to allow it.")
option_token = option.split("=", 1)[0].split(None, 1)[0]
if option_token.startswith("-") and not option_token.startswith("--"):
for option_char in option_token[1:]:
unsafe_option = unsafe_short_options.get(option_char)
if unsafe_option is not None:
raise UnsafeOptionError(
f"{unsafe_option} is not allowed, use `allow_unsafe_options=True` to allow it."
)
if option_char not in clusterable_short_options_set:
break
if not (option.startswith("--") or (options_are_kwargs and len(candidate) > 1)):
continue
for canonical, unsafe_option in canonical_unsafe_options.items():
if canonical.startswith(candidate):
raise UnsafeOptionError(
f"{unsafe_option} is not allowed, use `allow_unsafe_options=True` to allow it."
)
@classmethod
def _option_candidates(cls, args: Sequence[Any] = (), kwargs: Optional[Mapping[str, Any]] = None) -> List[str]:
"""Collect possible option spellings before command-line transformation."""
options = [
option for option in cls._unpack_args([arg for arg in args if arg is not None]) if option.startswith("-")
]
if kwargs:
split_single_char_options = kwargs.get("split_single_char_options", True)
for key, value in kwargs.items():
values = value if isinstance(value, (list, tuple)) else (value,)
if any(value is True or (value is not False and value is not None) for value in values):
key = str(key)
if len(key) != 1:
options.append(f"--{dashify(key)}")
elif split_single_char_options:
options.append(f"-{key}")
options.extend(
str(value)
for value in values
if value is not True and value not in (False, None) and str(value).startswith("-")
)
else:
options.extend(
f"-{key}" if value is True else f"-{key}{value}"
for value in values
if value is True or (value is not False and value is not None)
)
return options
@staticmethod
def _check_operand(value: Any, label: str = "operand") -> str:
"""Validate a single name/revision, before Git can interpret it as an option.
Paths and free-form payloads need their own validation and framing instead.
"""
value = safe_decode(value) if isinstance(value, bytes) else str(value)
if value.startswith("-") or any(char in value for char in "\0\r\n"):
raise UnsafeOptionError(f"Invalid {label}: {value!r}")
return value
def _require_version(self) -> None:
key = None
if self._version_info_token is not self._refresh_token:
key = self._version_check_key()
if key is not None:
version = self._version_check_cache.get(key)
if version is not None:
self._version_info = version
self._version_info_token = key[0]
version = self.version_info
if version < (2, 52):
raise UnsupportedOperation("GitPython requires Git 2.52 or newer for repository operations")
if key is not None and key not in self._version_check_cache:
# ponytail: clear at 128 contexts; use LRU if diverse environments churn.
if len(self._version_check_cache) >= 128:
self._version_check_cache.clear()
self._version_check_cache[key] = version
def _version_check_key(self) -> Optional[Tuple[Any, ...]]:
"""Identify the executable and context of a minimum-version probe."""
executable = self.GIT_PYTHON_GIT_EXECUTABLE
if not executable or self._git_options or self._persistent_git_options:
return None
environment = {**os.environ, "LANGUAGE": "C", "LC_ALL": "C", **self._environment}
try:
cwd = os.path.realpath(self._working_dir or os.getcwd())
if not os.access(cwd, os.X_OK):
return None
resolved: Optional[str]
if os.path.isabs(executable):
resolved = executable
elif sys.platform == "win32":
# CreateProcess searches in the parent context, unlike execvpe.
return None
elif os.path.dirname(executable):
resolved = os.path.join(cwd, executable)
else:
path = environment.get("PATH")
path = os.defpath if path is None else path
resolved = shutil.which(
executable, path=os.pathsep.join(os.path.join(cwd, p) for p in path.split(os.pathsep))
)
if resolved is None:
return None
info = os.stat(resolved)
except OSError:
return None
return (
self._refresh_token,
type(self),
executable,
resolved,
cwd,
tuple(sorted(environment.items())),
(info.st_dev, info.st_ino, info.st_size, info.st_mtime_ns, info.st_ctime_ns),
)
def _call_process_safe(
self,
method: str,
*args: Any,
_allow_hooks: bool = False,
_allow_network: bool = False,
_config: Sequence[str] = (),
**kwargs: Any,
) -> Any:
"""Run library-owned plumbing without introducing executable configuration.
Callers validate their operands and any forwarded options using the existing
command-specific guards. This does not restrict the public raw Git interface.
"""
self._check_operand(method, "command")
if kwargs.get("shell"):
raise UnsafeOptionError("Library operations cannot run through a shell")
if _allow_hooks and method != "hook":
raise UnsafeOptionError("Only explicit hook operations may enable hooks")
insertion = kwargs.get("insert_kwargs_after")
if insertion is not None and not (
method == "remote" and args and args[0] == insertion and insertion in ("add", "set-url", "update")
):
raise UnsafeOptionError("Library command options cannot be reordered past safety flags")
for setting in _config:
key, separator, value = setting.partition("=")
if not separator or (
key.lower() not in ("i18n.commitencoding", "diff.mnemonicprefix", "fetch.output", "core.abbrev")
and not (key in ("core.protectHFS", "core.protectNTFS") and value == "true")
and not (key == "protocol.file.allow" and value in ("always", "never", "user"))
and not (key in ("tar.tgz.command", "tar.tar.gz.command") and value == "git archive gzip")
):
raise UnsafeOptionError(f"Unsupported internal Git configuration: {key!r}")
for arg in self._unpack_args([arg for arg in args if arg is not None]) + self.transform_kwargs(
**{key: value for key, value in kwargs.items() if key not in execute_kwargs}
):
if "\0" in arg:
raise UnsafeOptionError("Git arguments cannot contain NUL bytes")
options = ["--no-pager", "--no-optional-locks"]
# These settings would become visible as user configuration in `config`.
if method != "config":
config = ["core.fsmonitor=false", "gc.auto=0", "maintenance.auto=false"]
if not _allow_hooks:
config.append(f"core.hooksPath={os.devnull}")
for setting in [*config, *_config]:
options.extend(("-c", setting))
elif _config:
raise ValueError("Configuration queries must not include synthetic settings")
native = _backend.dispatch(self, method, args, kwargs, _config)
if native is not NotImplemented:
return native
self._require_version()
env = dict(kwargs.pop("env", {}) or {})
env.update(LC_ALL="C", LANGUAGE="C")
if not _allow_network:
env.update(GIT_NO_LAZY_FETCH="1", GIT_TERMINAL_PROMPT="0")
return self._call_process(
method,
*args,
_safe_git_options=options,
shell=False,
env=env,
**{key: value for key, value in kwargs.items() if key != "shell"},
)
AutoInterrupt: TypeAlias = _AutoInterrupt
CatFileContentStream: TypeAlias = _CatFileContentStream
def __init__(self, working_dir: Union[None, PathLike] = None) -> None:
"""Initialize this instance with:
:param working_dir:
Git directory we should work in. If ``None``, we always work in the current
directory as returned by :func:`os.getcwd`.
This is meant to be the working tree directory if available, or the
``.git`` directory in case of bare repositories.
"""
super().__init__()
self._working_dir = expand_path(working_dir)
self._git_options: Union[List[str], Tuple[str, ...]] = ()
self._persistent_git_options: List[str] = []
# Extra environment variables to pass to git commands
self._environment: Dict[str, Optional[str]] = {}
self._repository_environment: Dict[str, Optional[str]] = {}
self._repo: Any = None # Weak reference; the Repo owns native resources.
# Cached version slots
self._version_info = None
self._version_info_token = None
# Cached command slots
self.cat_file_header: Union[None, TBD] = None
self.cat_file_all: Union[None, TBD] = None
def __getattr__(self, name: str) -> Any:
"""A convenience method as it allows to call the command as if it was an object.
:return:
Callable object that will execute call :meth:`_call_process` with your
arguments.
"""
if name.startswith("_"):
return super().__getattribute__(name)
return lambda *args, **kwargs: self._call_process(name, *args, **kwargs)
def set_persistent_git_options(self, **kwargs: Any) -> None:
"""Specify command line options to the git executable for subsequent
subcommand calls.
:param kwargs:
A dict of keyword arguments.
These arguments are passed as in :meth:`_call_process`, but will be passed
to the git command rather than the subcommand.
"""
self._persistent_git_options = self.transform_kwargs(split_single_char_options=True, **kwargs)
def ls_remote(
self,
*args: Any,
allow_unsafe_options: bool = False,
allow_unsafe_protocols: bool = False,
**kwargs: Any,
) -> Union[str, bytes, Tuple[int, Union[str, bytes], str], "Git.AutoInterrupt"]:
"""List references in a remote repository.
:param allow_unsafe_options:
Allow unsafe options, like ``--upload-pack`` or ``--exec``.
:param allow_unsafe_protocols:
Allow unsafe protocols to be used, like ``ext``. Positional arguments
and split short-option values are checked.
"""
if not allow_unsafe_options:
candidate_options = self._option_candidates(args, kwargs)
Git.check_unsafe_options(options=candidate_options, unsafe_options=self.unsafe_git_ls_remote_options)
if not allow_unsafe_protocols:
self._check_unsafe_protocols_in_args(args, kwargs)
return self._call_process("ls_remote", *args, **kwargs)
@property
def working_dir(self) -> Union[None, PathLike]:
""":return: Git directory we are working on"""
return self._working_dir
@property
def version_info(self) -> Tuple[int, ...]:
"""
:return: Tuple with integers representing the major, minor and additional
version numbers as parsed from :manpage:`git-version(1)`. Up to four fields
are used.
This value is generated on demand and is cached.
"""
# Refreshing is global, but version_info caching is per-instance.
refresh_token = self._refresh_token # Copy token in case of concurrent refresh.
# Use the cached version if obtained after the most recent refresh.
if self._version_info_token is refresh_token:
assert self._version_info is not None, "Bug: corrupted token-check state"
return self._version_info
# Run "git version" and parse it.
process_version = cast(str, self._call_process("version", shell=False))
version_string = process_version.split(" ")[2]
version_fields = version_string.split(".")[:4]
leading_numeric_fields = itertools.takewhile(str.isdigit, version_fields)
self._version_info = tuple(map(int, leading_numeric_fields))
# This value will be considered valid until the next refresh.
self._version_info_token = refresh_token
return self._version_info
@overload
def execute(
self,
command: Union[str, Sequence[Any]],
istream: Union[None, int, BinaryIO] = None,
*,
as_process: Literal[True],
**subprocess_kwargs: Any,
) -> "AutoInterrupt": ...
@overload
def execute(
self,
command: Union[str, Sequence[Any]],
istream: Union[None, int, BinaryIO] = None,
*,
as_process: Literal[False] = False,
with_extended_output: Literal[False] = False,
stdout_as_string: Literal[True] = True,
with_stdout: Literal[True] = True,
**subprocess_kwargs: Any,
) -> str: ...
@overload
def execute(
self,
command: Union[str, Sequence[Any]],
istream: Union[None, int, BinaryIO] = None,
*,
as_process: Literal[False] = False,
with_extended_output: Literal[False] = False,
stdout_as_string: Literal[False],
universal_newlines: Literal[False] = False,
with_stdout: Literal[True] = True,
**subprocess_kwargs: Any,
) -> bytes: ...
@overload
def execute(
self,
command: Union[str, Sequence[Any]],
istream: Union[None, int, BinaryIO] = None,
*,
as_process: Literal[False] = False,
with_extended_output: Literal[True],
stdout_as_string: Literal[True] = True,
with_stdout: Literal[True] = True,
**subprocess_kwargs: Any,
) -> Tuple[int, str, str]: ...
@overload
def execute(
self,
command: Union[str, Sequence[Any]],
istream: Union[None, int, BinaryIO] = None,
*,
as_process: Literal[False] = False,
with_extended_output: Literal[True],
stdout_as_string: Literal[False],
universal_newlines: Literal[False] = False,
with_stdout: Literal[True] = True,
**subprocess_kwargs: Any,
) -> Tuple[int, bytes, str]: ...
@overload
def execute(
self,
command: Union[str, Sequence[Any]],
istream: Union[None, int, BinaryIO] = None,
*,
as_process: Literal[False] = False,
with_extended_output: Literal[True],
**subprocess_kwargs: Any,
) -> Tuple[int, Union[str, bytes, None], str]: ...
@overload
def execute(
self,
command: Union[str, Sequence[Any]],
istream: Union[None, int, BinaryIO] = None,
*,
as_process: Literal[False] = False,
with_extended_output: Literal[False] = False,
**subprocess_kwargs: Any,
) -> Union[str, bytes, None]: ...
@overload
def execute(
self,
command: Union[str, Sequence[Any]],
istream: Union[None, int, BinaryIO] = None,
with_extended_output: bool = False,
with_exceptions: bool = True,
as_process: bool = False,
output_stream: Union[None, BinaryIO] = None,
stdout_as_string: bool = True,
kill_after_timeout: Union[None, float] = None,
with_stdout: bool = True,
universal_newlines: bool = False,
shell: Union[None, bool] = None,
env: Union[None, Mapping[str, str]] = None,
max_chunk_size: int = io.DEFAULT_BUFFER_SIZE,
strip_newline_in_stdout: bool = True,
**subprocess_kwargs: Any,
) -> Union[None, str, bytes, Tuple[int, Union[str, bytes, None], str], AutoInterrupt]: ...
def execute(
self,
command: Union[str, Sequence[Any]],
istream: Union[None, int, BinaryIO] = None,
with_extended_output: bool = False,
with_exceptions: bool = True,
as_process: bool = False,
output_stream: Union[None, BinaryIO] = None,
stdout_as_string: bool = True,
kill_after_timeout: Union[None, float] = None,
with_stdout: bool = True,
universal_newlines: bool = False,
shell: Union[None, bool] = None,
env: Union[None, Mapping[str, str]] = None,
max_chunk_size: int = io.DEFAULT_BUFFER_SIZE,
strip_newline_in_stdout: bool = True,
**subprocess_kwargs: Any,
) -> Union[None, str, bytes, Tuple[int, Union[str, bytes, None], str], AutoInterrupt]:
R"""Handle executing the command, and consume and return the returned
information (stdout).
:param command:
The command to execute. A sequence of program arguments is recommended.
A string is also accepted, but its meaning is strongly platform-dependent.
By default, a shell is not used. On Unix-like systems, a string is the whole
program name (so ``"git log -n 1"`` raises :class:`GitCommandNotFound`). On
Windows, the program parses the arguments itself, so multi-word strings can
work but are not portable.
Avoid ``shell=True``: this runs the command in
a shell, which is generally unsafe. The shell interprets metacharacters
such as ``;``, ``|``, ``&``, ``$(...)``, ``$VAR``, ``%VAR%``, and ``^``
(depending on the platform) as syntax. Any untrusted text in the command
can then execute arbitrary OS commands.
Producing a sequence automatically by :func:`shlex.split` and passing it
as the command is far safer than ``shell=True``. But :func:`shlex.split`
parses POSIX shell syntax on all systems, and the result is still unsafe
for anything but *fixed, fully trusted* strings. Do not use it on strings
built by interpolating values: whitespace or quoting in an untrusted value
can still inject arguments. For input derived in any way from untrusted
data, build the argument sequence yourself, while ensuring each argument
is fully sanitized.
:param istream:
Standard input filehandle passed to :class:`subprocess.Popen`.
:param with_extended_output:
Whether to return a (status, stdout, stderr) tuple.
:param with_exceptions:
Whether to raise an exception when git returns a non-zero status.
:param as_process:
Whether to return the created process instance directly from which
streams can be read on demand. This will render `with_extended_output`
and `with_exceptions` ineffective - the caller will have to deal with
the details. It is important to note that the process will be placed
into an :class:`AutoInterrupt` wrapper that will interrupt the process
once it goes out of scope. If you use the command in iterators, you
should pass the whole process instance instead of a single stream.
:param output_stream:
If set to a file-like object, data produced by the git command will be
copied to the given stream instead of being returned as a string.
This feature only has any effect if `as_process` is ``False``.
:param stdout_as_string:
If ``False``, the command's standard output will be bytes. Otherwise, it
will be decoded into a string using the default encoding (usually UTF-8).
The latter can fail, if the output contains binary data.
:param kill_after_timeout:
Specifies a timeout in seconds for the git command, after which the process
should be killed. This will have no effect if `as_process` is set to
``True``. It is set to ``None`` by default and will let the process run
until the timeout is explicitly specified. Uses of this feature should be
carefully considered, due to the following limitations:
1. This feature is not supported at all on Windows.
2. Enumerating child processes requires ``pgrep -P``, or a ``ps`` command
supporting the POSIX ``-A`` and ``-o`` options if ``pgrep`` is not
installed (``ps -ef`` on Cygwin). Effectiveness may vary on systems
without these commands.
3. Descendants are enumerated before signalling. Processes that detach
or spawn after enumeration may not receive signals.
4. `kill_after_timeout` uses ``SIGKILL``, which can have negative side
effects on a repository. For example, stale locks in case of
:manpage:`git-gc(1)` could render the repository incapable of accepting
changes until the lock is manually removed.
:param with_stdout:
If ``True``, default ``True``, we open stdout on the created process.
:param universal_newlines:
If ``True``, pipes will be opened as text, and lines are split at all known
line endings.
:param shell:
Whether to invoke commands through a shell
(see :class:`Popen(..., shell=True) `).
Use a shell explicitly; the default is ``False``.
Passing ``shell=True`` to this or any other GitPython function should be
avoided, as it is unsafe under most circumstances. This is because it is
typically not feasible to fully consider and account for the effect of shell
expansions, especially when passing ``shell=True`` to other methods that
forward it to :meth:`Git.execute`. Passing ``shell=True`` is also no longer
needed (nor useful) to work around any known operating system specific
issues.
On Unix-like systems, when migrating away from passing string commands with
``shell=True``, :func:`shlex.split` may serve as a transitional step in rare
cases, with extreme care. (Drop ``shell=True`` and pass the resulting
sequence as the command.) See the `command` parameter above on the risks.
:param env:
A dictionary of environment variables to be passed to
:class:`subprocess.Popen`.
:param max_chunk_size:
Maximum number of bytes in one chunk of data passed to the `output_stream`
in one invocation of its ``write()`` method. If the given number is not
positive then the default value is used.
:param strip_newline_in_stdout:
Whether to strip the trailing ``\n`` of the command stdout.
:param subprocess_kwargs:
Keyword arguments to be passed to :class:`subprocess.Popen`. Please note
that some of the valid kwargs are already set by this method; the ones you
specify may not be the same ones.
:return:
* str(output), if `extended_output` is ``False`` (Default)
* tuple(int(status), str(stdout), str(stderr)),
if `extended_output` is ``True``
If `output_stream` is ``True``, the stdout value will be your output stream:
* output_stream, if `extended_output` is ``False``
* tuple(int(status), output_stream, str(stderr)),
if `extended_output` is ``True``
Note that git is executed with ``LC_MESSAGES="C"`` to ensure consistent
output regardless of system language.
:raise git.exc.GitCommandError:
:note:
If you add additional keyword arguments to the signature of this method, you
must update the ``execute_kwargs`` variable housed in this module.
"""
# Remove password for the command if present.
redacted_command = remove_password_if_present(command)
if self.GIT_PYTHON_TRACE and (self.GIT_PYTHON_TRACE != "full" or as_process):
_logger.info(" ".join(redacted_command))
# Allow the user to have the command executed in their working dir.
try:
cwd = self._working_dir or os.getcwd() # type: Optional[PathLike]
if not os.access(str(cwd), os.X_OK):
cwd = None
except FileNotFoundError:
cwd = None
# Start the process.
inline_env = env
environment: Dict[str, Optional[str]] = dict(os.environ)
# Attempt to force all output to plain ASCII English, which is what some parsing
# code may expect.
# According to https://askubuntu.com/a/311796, we are setting LANGUAGE as well
# just to be sure.
environment["LANGUAGE"] = "C"
environment["LC_ALL"] = "C"
redirected = subprocess_kwargs.pop("_repository_redirected", False)
automatic = self._repository_environment if redirected else {}
environment.update(
{key: value for key, value in self._environment.items() if key not in automatic or value != automatic[key]}
)
if inline_env is not None:
environment.update(inline_env)
# Internal or per-call None overrides remove inherited variables.
env = {key: value for key, value in environment.items() if value is not None}
if sys.platform == "win32":
if kill_after_timeout is not None:
raise GitCommandError(
redacted_command,
'"kill_after_timeout" feature is not supported on Windows.',
)
cmd_not_found_exception = OSError
else:
cmd_not_found_exception = FileNotFoundError
# END handle
stdout_sink = PIPE if with_stdout else getattr(subprocess, "DEVNULL", None) or open(os.devnull, "wb")
if shell is None:
shell = False
_logger.debug(
"Popen(%s, cwd=%s, stdin=%s, shell=%s, universal_newlines=%s)",
redacted_command,
cwd,
"" if istream else "None",
shell,
universal_newlines,
)
try:
proc = safer_popen(
command,
env=env,
cwd=cwd,
bufsize=-1,
stdin=(istream or DEVNULL),
stderr=PIPE,
stdout=stdout_sink,
shell=shell,
universal_newlines=universal_newlines,
encoding=defenc if universal_newlines else None,
**subprocess_kwargs,
)
except cmd_not_found_exception as err:
raise GitCommandNotFound(redacted_command, err) from err
else:
_backend.record("Git.execute", "CLI process")
owner = self._repo() if self._repo is not None else None
if owner is not None:
owner._gix_state = None
# Replace with a typeguard for Popen[bytes]?
proc.stdout = cast(BinaryIO, proc.stdout)
proc.stderr = cast(BinaryIO, proc.stderr)
if as_process:
return self.AutoInterrupt(proc, command)
watchdog: Optional[threading.Timer] = None
kill_check: Optional[threading.Event] = None
timeout_error: Optional[Callable[[], Union[str, bytes]]] = None
if sys.platform != "win32" and kill_after_timeout is not None:
# Help mypy figure out this is not None even when used inside communicate().
timeout = kill_after_timeout
def kill_process(pid: int) -> None:
if _kill_process(pid):
assert kill_check is not None
kill_check.set()
def make_timeout_error() -> Union[str, bytes]:
err = f'Timeout: the command "{" ".join(redacted_command)}" did not complete in {timeout:g} secs.'
return err if universal_newlines else err.encode(defenc)
def communicate() -> Tuple[Union[str, bytes, None], Union[str, bytes, None]]:
assert watchdog is not None
assert kill_check is not None
watchdog.start()
out, err = proc.communicate()
watchdog.cancel()
if kill_check.is_set():
err = make_timeout_error()
return out, err
# END helpers
kill_check = threading.Event()
watchdog = threading.Timer(timeout, kill_process, args=(proc.pid,))
timeout_error = make_timeout_error
else:
communicate = proc.communicate
# Wait for the process to return.
status = 0
stdout_value: Union[str, bytes, None] = b""
stderr_value: Union[str, bytes, None] = b""
newline = "\n" if universal_newlines else b"\n"
try:
if output_stream is None:
stdout_value, stderr_value = communicate()
# Strip trailing "\n".
if stdout_value is not None and stdout_value.endswith(newline) and strip_newline_in_stdout: # type: ignore[arg-type]
stdout_value = stdout_value[:-1]
if stderr_value is not None and stderr_value.endswith(newline): # type: ignore[arg-type]
stderr_value = stderr_value[:-1]
status = proc.returncode
else:
max_chunk_size = max_chunk_size if max_chunk_size and max_chunk_size > 0 else io.DEFAULT_BUFFER_SIZE
if watchdog is not None:
watchdog.start()
try:
if proc.stdout is not None:
stream_copy(proc.stdout, output_stream, max_chunk_size)
stdout_value = proc.stdout.read()
if proc.stderr is not None:
stderr_value = proc.stderr.read()
status = proc.wait()
finally:
if watchdog is not None:
watchdog.cancel()
# Strip trailing "\n".
if stderr_value is not None and stderr_value.endswith(newline): # type: ignore[arg-type]
stderr_value = stderr_value[:-1]
if kill_check is not None and kill_check.is_set():
assert timeout_error is not None
stderr_value = timeout_error()
# END stdout handling
finally:
if proc.stdout is not None:
proc.stdout.close()
if proc.stderr is not None:
proc.stderr.close()
if self.GIT_PYTHON_TRACE == "full":
cmdstr = " ".join(redacted_command)
def as_text(stdout_value: Union[bytes, str, None]) -> str:
return not output_stream and safe_decode(stdout_value) or ""
# END as_text
if stderr_value:
_logger.info(
"%s -> %d; stdout: '%s'; stderr: '%s'",
cmdstr,
status,
as_text(stdout_value),
safe_decode(stderr_value),
)
elif stdout_value:
_logger.info("%s -> %d; stdout: '%s'", cmdstr, status, as_text(stdout_value))
else:
_logger.info("%s -> %d", cmdstr, status)
# END handle debug printing
if with_exceptions and status != 0:
raise GitCommandError(redacted_command, status, stderr_value, stdout_value)
if isinstance(stdout_value, bytes) and stdout_as_string: # Could also be output_stream.
stdout_value = safe_decode(stdout_value)
# stderr is always captured through PIPE.
assert stderr_value is not None
# Allow access to the command's status code.
if with_extended_output:
return (status, stdout_value, safe_decode(stderr_value))
else:
return stdout_value
def environment(self) -> Dict[str, Optional[str]]:
return self._environment
def update_environment(self, **kwargs: Any) -> Dict[str, Union[str, None]]:
"""Set environment variables for future git invocations. Return all changed
values in a format that can be passed back into this function to revert the
changes.
Examples::
old_env = self.update_environment(PWD='/tmp')
self.update_environment(**old_env)
:param kwargs:
Environment variables to use for git processes.
:return:
Dict that maps environment variables to their old values
"""
old_env = {}
for key, value in kwargs.items():
self._repository_environment.pop(key, None)
# Set value if it is None.
if value is not None:
old_env[key] = self._environment.get(key)
self._environment[key] = value
# Remove key from environment if its value is None.
elif key in self._environment:
old_env[key] = self._environment[key]
del self._environment[key]
return old_env
@contextlib.contextmanager
def custom_environment(self, **kwargs: Any) -> Iterator[None]:
"""A context manager around the above :meth:`update_environment` method to
restore the environment back to its previous state after operation.
Examples::
with self.custom_environment(GIT_SSH='/bin/ssh_wrapper'):
repo.remotes.origin.fetch()
:param kwargs:
See :meth:`update_environment`.
"""
automatic = {key: self._repository_environment[key] for key in kwargs if key in self._repository_environment}
old_env = self.update_environment(**kwargs)
try:
yield
finally:
self.update_environment(**old_env)
self._repository_environment.update(automatic)
def transform_kwarg(self, name: str, value: Any, split_single_char_options: bool) -> List[str]:
if len(name) == 1:
if value is True:
return ["-%s" % name]
elif value not in (False, None):
if split_single_char_options:
return ["-%s" % name, "%s" % value]
else:
return ["-%s%s" % (name, value)]
else:
if value is True:
return ["--%s" % dashify(name)]
elif value is not False and value is not None:
return ["--%s=%s" % (dashify(name), value)]
return []
def transform_kwargs(self, split_single_char_options: bool = True, **kwargs: Any) -> List[str]:
"""Transform Python-style kwargs into git command line options."""
args = []
for k, v in kwargs.items():
if isinstance(v, (list, tuple)):
for value in v:
args += self.transform_kwarg(k, value, split_single_char_options)
else:
args += self.transform_kwarg(k, v, split_single_char_options)
return args
@classmethod
def _unpack_args(cls, arg_list: Sequence[Any]) -> List[str]:
outlist = []
if isinstance(arg_list, (list, tuple)):
for arg in arg_list:
outlist.extend(cls._unpack_args(arg))
else:
outlist.append(os.fsdecode(arg_list) if isinstance(arg_list, os.PathLike) else str(arg_list))
return outlist
def __call__(self, **kwargs: Any) -> "Git":
"""Specify command line options to the git executable for a subcommand call.
:param kwargs:
A dict of keyword arguments.
These arguments are passed as in :meth:`_call_process`, but will be passed
to the git command rather than the subcommand.
Examples::
git(work_tree='/tmp').difftool()
"""
self._git_options = self.transform_kwargs(split_single_char_options=True, **kwargs)
return self
@overload
def _call_process(
self, method: str, *args: None, **kwargs: None
) -> str: ... # If no args were given, execute the call with all defaults.
@overload
def _call_process(
self,
method: str,
istream: int,
as_process: Literal[True],
*args: Any,
**kwargs: Any,
) -> "Git.AutoInterrupt": ...
@overload
def _call_process(
self, method: str, *args: Any, **kwargs: Any
) -> Union[str, bytes, Tuple[int, Union[str, bytes], str], "Git.AutoInterrupt"]: ...
def _call_process(
self, method: str, *args: Any, **kwargs: Any
) -> Union[str, bytes, Tuple[int, Union[str, bytes], str], "Git.AutoInterrupt"]:
"""Run the given git command with the specified arguments and return the result
as a string.
:param method:
The command. Contained ``_`` characters will be converted to hyphens, such
as in ``ls_files`` to call ``ls-files``.
:param args:
The list of arguments. If ``None`` is included, it will be pruned.
This allows your commands to call git more conveniently, as ``None`` is
realized as non-existent.
Positional arguments may intentionally contain command options. Higher-level
APIs must separate their operands with ``--`` where the Git command supports
it, or reject option-shaped operands where Git reparses them internally (for
example, ``pull`` and ``remote update``). Shell quoting cannot prevent Git
from interpreting a leading-dash argument as an option.
:param kwargs:
Contains key-values for the following:
- The :meth:`execute()` kwds, as listed in ``execute_kwargs``.
- "Command options" to be converted by :meth:`transform_kwargs`.
- The ``insert_kwargs_after`` key which its value must match one of
``*args``.
It also contains any command options, to be appended after the matched arg.
Examples::
git.rev_list('master', max_count=10, header=True)
turns into::
git rev-list --max-count=10 --header master
:return:
Same as :meth:`execute`. If no args are given, used :meth:`execute`'s
default (especially ``as_process = False``, ``stdout_as_string = True``) and
return :class:`str`.
"""
# Handle optional arguments prior to calling transform_kwargs.
# Otherwise these'll end up in args, which is bad.
safe_git_options = kwargs.pop("_safe_git_options", ())
exec_kwargs = {k: v for k, v in kwargs.items() if k in execute_kwargs}
opts_kwargs = {k: v for k, v in kwargs.items() if k not in execute_kwargs}
insert_after_this_arg = opts_kwargs.pop("insert_kwargs_after", None)
# Prepare the argument list.
opt_args = self.transform_kwargs(**opts_kwargs)
ext_args = self._unpack_args([a for a in args if a is not None])
if insert_after_this_arg is None:
args_list = opt_args + ext_args
else:
try:
index = ext_args.index(insert_after_this_arg)
except ValueError as err:
raise ValueError(
"Couldn't find argument '%s' in args %s to insert cmd options after"
% (insert_after_this_arg, str(ext_args))
) from err
# END handle error
args_list = ext_args[: index + 1] + opt_args + ext_args[index + 1 :]
# END handle opts_kwargs
call = [self.GIT_PYTHON_GIT_EXECUTABLE]
# Add persistent git options.
call.extend(self._persistent_git_options)
# Add the git options, then reset to empty to avoid side effects.
call.extend(self._git_options)
self._git_options = ()
# Only global options redirect discovery. A subcommand operand named
# -C must not drop the repository locations supplied by Repo.
if any(option == "-C" for option in call[1:]):
exec_kwargs["_repository_redirected"] = True
call.extend(safe_git_options)
call.append(dashify(method))
call.extend(args_list)
return self.execute(call, **exec_kwargs)
def _parse_object_header(self, header_line: Union[str, bytes]) -> Tuple[str, str, int]:
"""
:param header_line:
A line of the form::
type_string size_as_int
:return:
(hex_sha, type_string, size_as_int)
:raise ValueError:
If the header contains indication for an error due to incorrect input sha.
"""
if isinstance(header_line, bytes):
header_line = header_line.decode("ascii", "replace")
tokens = header_line.split()
if len(tokens) != 3:
if not tokens:
err_msg = (
f"SHA is empty, possible dubious ownership in the repository "
f"""at {self._working_dir}.\n If this is unintended run:\n\n """
f""" "git config --global --add safe.directory {self._working_dir}" """
)
raise ValueError(err_msg)
else:
raise ValueError("SHA %s could not be resolved, git returned: %r" % (tokens[0], header_line.strip()))
# END handle actual return value
# END error handling
if (
not re.fullmatch(r"[0-9a-fA-F]+", tokens[0])
or len(tokens[0]) % 2
or tokens[1] not in ("blob", "tree", "commit", "tag")
or not tokens[2].isdigit()
):
raise ValueError("Failed to parse header: %r" % header_line)
return (tokens[0], tokens[1], int(tokens[2]))
def _prepare_ref(self, ref: object) -> bytes:
# `cat-file -Z` separates both requests and responses with NUL, so paths
# containing newlines cannot inject requests or desynchronize the process.
if isinstance(ref, bytes):
refstr: str = ref.decode(defenc, "surrogateescape")
elif not isinstance(ref, str):
refstr = str(ref) # Could be ref-object.
else:
refstr = ref
if "\0" in refstr or refstr.startswith("-"):
raise UnsafeOptionError("Object queries cannot contain NUL or start with '-'")
return refstr.encode(defenc, "surrogateescape") + b"\0"
def _get_persistent_cmd(self, attr_name: str, cmd_name: str, *args: Any, **kwargs: Any) -> "Git.AutoInterrupt":
cur_val = getattr(self, attr_name)
if cur_val is not None:
return cur_val
options: Dict[str, Any] = {"istream": PIPE, "as_process": True}
options.update(kwargs)
cmd = self._call_process_safe(cmd_name, *args, **options)
setattr(self, attr_name, cmd)
cmd = cast("Git.AutoInterrupt", cmd)
return cmd
def __get_object_header(self, cmd: "Git.AutoInterrupt", request: bytes) -> Tuple[str, str, int]:
if cmd.stdin and cmd.stdout:
cmd.stdin.write(request)
cmd.stdin.flush()
header = bytearray()
while True:
char = cmd.stdout.read(1)
if not char:
cmd.wait()
raise ValueError("Git closed the object stream before its response")
if char == b"\0":
break
header.extend(char)
return self._parse_object_header(bytes(header))
else:
raise ValueError("cmd stdin was empty")
def get_object_header(self, ref: Union[str, bytes]) -> Tuple[str, str, int]:
"""Use this method to quickly examine the type and size of the object behind the
given ref.
:note:
The method will only suffer from the costs of command invocation once and
reuses the command in subsequent calls.
:return:
(hexsha, type_string, size_as_int)
"""
request = self._prepare_ref(ref)
native = _backend.object_data(self, request[:-1])
if native is not NotImplemented:
return native
cmd = self._get_persistent_cmd("cat_file_header", "cat_file", batch_check=True, Z=True)
return self.__get_object_header(cmd, request)
def get_object_data(self, ref: Union[str, bytes]) -> Tuple[str, str, int, bytes]:
"""Similar to :meth:`get_object_header`, but returns object data as well.
:return:
(hexsha, type_string, size_as_int, data_string)
:note:
Not threadsafe.
"""
hexsha, typename, size, stream = self.stream_object_data(ref)
data = stream.read(size)
del stream
return (hexsha, typename, size, data)
def stream_object_data(self, ref: Union[str, bytes]) -> Tuple[str, str, int, "Git.CatFileContentStream"]:
"""Similar to :meth:`get_object_data`, but returns the data as a stream.
:return:
(hexsha, type_string, size_as_int, stream)
:note:
This method is not threadsafe. You need one independent :class:`Git`
instance per thread to be safe!
"""
request = self._prepare_ref(ref)
native = _backend.object_data(self, request[:-1], stream=True)
if native is not NotImplemented:
return native
cmd = self._get_persistent_cmd("cat_file_all", "cat_file", batch=True, Z=True)
hexsha, typename, size = self.__get_object_header(cmd, request)
cmd_stdout = cmd.stdout if cmd.stdout is not None else io.BytesIO()
return (hexsha, typename, size, self.CatFileContentStream(size, cmd_stdout))
def clear_cache(self) -> "Git":
"""Clear all kinds of internal caches to release resources.
Currently persistent commands will be interrupted.
:return:
self
"""
for cmd in (self.cat_file_all, self.cat_file_header):
if cmd:
cmd.__del__()
self.cat_file_all = None
self.cat_file_header = None
return self