Coverage for /pythoncovmergedfiles/medio/medio/usr/local/lib/python3.11/site-packages/git/cmd.py: 61%
Shortcuts on this page
r m x toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
1# Copyright (C) 2008, 2009 Michael Trier (mtrier@gmail.com) and contributors
2#
3# This module is part of GitPython and is released under the
4# 3-Clause BSD License: https://opensource.org/license/bsd-3-clause/
6from __future__ import annotations
8__all__ = ["GitMeta", "Git"]
10import contextlib
11import io
12import itertools
13import logging
14import os
15import re
16import signal
17import subprocess
18from subprocess import DEVNULL, PIPE, Popen
19import sys
20from textwrap import dedent
21import threading
22import warnings
24from git.compat import defenc, force_bytes, safe_decode
25from git.exc import (
26 CommandError,
27 GitCommandError,
28 GitCommandNotFound,
29 UnsafeOptionError,
30 UnsafeProtocolError,
31)
32from git.util import (
33 cygpath,
34 expand_path,
35 is_cygwin_git,
36 patch_env,
37 remove_password_if_present,
38 stream_copy,
39)
41# typing ---------------------------------------------------------------------------
43from typing import (
44 Any,
45 AnyStr,
46 BinaryIO,
47 Callable,
48 Dict,
49 IO,
50 Iterator,
51 List,
52 Mapping,
53 Optional,
54 Sequence,
55 TYPE_CHECKING,
56 TextIO,
57 Tuple,
58 Union,
59 cast,
60 overload,
61)
63if sys.version_info >= (3, 10):
64 from typing import TypeAlias
65else:
66 from typing_extensions import TypeAlias
68from git.types import Literal, PathLike, TBD
70if TYPE_CHECKING:
71 from git.diff import DiffIndex
72 from git.repo.base import Repo
74# ---------------------------------------------------------------------------------
76execute_kwargs = {
77 "istream",
78 "with_extended_output",
79 "with_exceptions",
80 "as_process",
81 "output_stream",
82 "stdout_as_string",
83 "kill_after_timeout",
84 "with_stdout",
85 "universal_newlines",
86 "shell",
87 "env",
88 "max_chunk_size",
89 "strip_newline_in_stdout",
90}
92_logger = logging.getLogger(__name__)
95# ==============================================================================
96## @name Utilities
97# ------------------------------------------------------------------------------
98# Documentation
99## @{
102def handle_process_output(
103 process: "Git.AutoInterrupt" | Popen,
104 stdout_handler: Union[
105 None,
106 Callable[[AnyStr], None],
107 Callable[[List[AnyStr]], None],
108 Callable[[bytes, "Repo", "DiffIndex"], None],
109 ],
110 stderr_handler: Union[None, Callable[[AnyStr], None], Callable[[List[AnyStr]], None]],
111 finalizer: Union[None, Callable[[Union[Popen, "Git.AutoInterrupt"]], None]] = None,
112 decode_streams: bool = True,
113 kill_after_timeout: Union[None, float] = None,
114) -> None:
115 R"""Register for notifications to learn that process output is ready to read, and
116 dispatch lines to the respective line handlers.
118 This function returns once the finalizer returns.
120 :param process:
121 :class:`subprocess.Popen` instance.
123 :param stdout_handler:
124 f(stdout_line_string), or ``None``.
126 :param stderr_handler:
127 f(stderr_line_string), or ``None``.
129 :param finalizer:
130 f(proc) - wait for proc to finish.
132 :param decode_streams:
133 Assume stdout/stderr streams are binary and decode them before pushing their
134 contents to handlers.
136 This defaults to ``True``. Set it to ``False`` if:
138 - ``universal_newlines == True``, as then streams are in text mode, or
139 - decoding must happen later, such as for :class:`~git.diff.Diff`\s.
141 :param kill_after_timeout:
142 :class:`float` or ``None``, Default = ``None``
144 To specify a timeout in seconds for the git command, after which the process
145 should be killed.
146 """
148 # Use 2 "pump" threads and wait for both to finish.
149 def pump_stream(
150 cmdline: List[str],
151 name: str,
152 stream: Union[BinaryIO, TextIO],
153 is_decode: bool,
154 handler: Union[None, Callable[[Union[bytes, str]], None]],
155 ) -> None:
156 try:
157 for line in stream:
158 if handler:
159 if is_decode:
160 assert isinstance(line, bytes)
161 line_str = line.decode(defenc)
162 handler(line_str)
163 else:
164 handler(line)
166 except Exception as ex:
167 _logger.error(f"Pumping {name!r} of cmd({remove_password_if_present(cmdline)}) failed due to: {ex!r}")
168 if "I/O operation on closed file" not in str(ex):
169 # Only reraise if the error was not due to the stream closing.
170 raise CommandError([f"<{name}-pump>"] + remove_password_if_present(cmdline), ex) from ex
171 finally:
172 stream.close()
174 if hasattr(process, "proc"):
175 process = cast("Git.AutoInterrupt", process)
176 cmdline: str | Tuple[str, ...] | List[str] = getattr(process.proc, "args", "")
177 p_stdout = process.proc.stdout if process.proc else None
178 p_stderr = process.proc.stderr if process.proc else None
179 else:
180 process = cast(Popen, process) # type: ignore[redundant-cast]
181 cmdline = getattr(process, "args", "")
182 p_stdout = process.stdout
183 p_stderr = process.stderr
185 if not isinstance(cmdline, (tuple, list)):
186 cmdline = cmdline.split()
188 pumps: List[Tuple[str, IO, Callable[..., None] | None]] = []
189 if p_stdout:
190 pumps.append(("stdout", p_stdout, stdout_handler))
191 if p_stderr:
192 pumps.append(("stderr", p_stderr, stderr_handler))
194 threads: List[threading.Thread] = []
196 for name, stream, handler in pumps:
197 t = threading.Thread(target=pump_stream, args=(cmdline, name, stream, decode_streams, handler))
198 t.daemon = True
199 t.start()
200 threads.append(t)
202 # FIXME: Why join? Will block if stdin needs feeding...
203 for t in threads:
204 t.join(timeout=kill_after_timeout)
205 if t.is_alive():
206 if isinstance(process, Git.AutoInterrupt):
207 process._terminate()
208 else: # Don't want to deal with the other case.
209 raise RuntimeError(
210 "Thread join() timed out in cmd.handle_process_output()."
211 f" kill_after_timeout={kill_after_timeout} seconds"
212 )
213 if stderr_handler:
214 error_str: Union[str, bytes] = (
215 f"error: process killed because it timed out. kill_after_timeout={kill_after_timeout} seconds"
216 )
217 if not decode_streams and isinstance(p_stderr, BinaryIO):
218 # Assume stderr_handler needs binary input.
219 error_str = cast(str, error_str)
220 error_str = error_str.encode()
221 # We ignore typing on the next line because mypy does not like the way
222 # we inferred that stderr takes str or bytes.
223 stderr_handler(error_str) # type: ignore[arg-type]
225 if finalizer:
226 finalizer(process)
229safer_popen: Callable[..., Popen]
231if sys.platform == "win32":
233 def _safer_popen_windows(
234 command: Union[str, Sequence[Any]],
235 *,
236 shell: bool = False,
237 env: Optional[Mapping[str, str]] = None,
238 **kwargs: Any,
239 ) -> Popen:
240 """Call :class:`subprocess.Popen` on Windows but don't include a CWD in the
241 search.
243 This avoids an untrusted search path condition where a file like ``git.exe`` in
244 a malicious repository would be run when GitPython operates on the repository.
245 The process using GitPython may have an untrusted repository's working tree as
246 its current working directory. Some operations may temporarily change to that
247 directory before running a subprocess. In addition, while by default GitPython
248 does not run external commands with a shell, it can be made to do so, in which
249 case the CWD of the subprocess, which GitPython usually sets to a repository
250 working tree, can itself be searched automatically by the shell. This wrapper
251 covers all those cases.
253 :note:
254 This currently works by setting the
255 :envvar:`NoDefaultCurrentDirectoryInExePath` environment variable during
256 subprocess creation. It also takes care of passing Windows-specific process
257 creation flags, but that is unrelated to path search.
259 :note:
260 The current implementation contains a race condition on :attr:`os.environ`.
261 GitPython isn't thread-safe, but a program using it on one thread should
262 ideally be able to mutate :attr:`os.environ` on another, without
263 unpredictable results. See comments in:
264 https://github.com/gitpython-developers/GitPython/pull/1650
265 """
266 # CREATE_NEW_PROCESS_GROUP is needed for some ways of killing it afterwards.
267 # https://docs.python.org/3/library/subprocess.html#subprocess.Popen.send_signal
268 # https://docs.python.org/3/library/subprocess.html#subprocess.CREATE_NEW_PROCESS_GROUP
269 creationflags = subprocess.CREATE_NO_WINDOW | subprocess.CREATE_NEW_PROCESS_GROUP
271 # When using a shell, the shell is the direct subprocess, so the variable must
272 # be set in its environment, to affect its search behavior.
273 if shell:
274 # The original may be immutable, or the caller may reuse it. Mutate a copy.
275 env = {} if env is None else dict(env)
276 env["NoDefaultCurrentDirectoryInExePath"] = "1" # The "1" can be any value.
278 # When not using a shell, the current process does the search in a
279 # CreateProcessW API call, so the variable must be set in our environment. With
280 # a shell, that's unnecessary if https://github.com/python/cpython/issues/101283
281 # is patched. In Python versions where it is unpatched, in the rare case the
282 # ComSpec environment variable is unset, the search for the shell itself is
283 # unsafe. Setting NoDefaultCurrentDirectoryInExePath in all cases, as done here,
284 # is simpler and protects against that. (As above, the "1" can be any value.)
285 with patch_env("NoDefaultCurrentDirectoryInExePath", "1"):
286 return Popen(
287 command,
288 shell=shell,
289 env=env,
290 creationflags=creationflags,
291 **kwargs,
292 )
294 safer_popen = _safer_popen_windows
295else:
296 safer_popen = Popen
299def dashify(string: str) -> str:
300 return string.replace("_", "-")
303def slots_to_dict(self: "Git", exclude: Sequence[str] = ()) -> Dict[str, Any]:
304 return {s: getattr(self, s) for s in self.__slots__ if s not in exclude}
307def dict_to_slots_and__excluded_are_none(self: object, d: Mapping[str, Any], excluded: Sequence[str] = ()) -> None:
308 for k, v in d.items():
309 setattr(self, k, v)
310 for k in excluded:
311 setattr(self, k, None)
314## -- End Utilities -- @}
317class _AutoInterrupt:
318 """Process wrapper that terminates the wrapped process on finalization.
320 This kills/interrupts the stored process instance once this instance goes out of
321 scope. It is used to prevent processes piling up in case iterators stop reading.
323 All attributes are wired through to the contained process object.
325 The wait method is overridden to perform automatic status code checking and possibly
326 raise.
327 """
329 __slots__ = ("proc", "args", "status")
331 # If this is non-zero it will override any status code during _terminate, used
332 # to prevent race conditions in testing.
333 _status_code_if_terminate: int = 0
335 def __init__(self, proc: Union[None, subprocess.Popen], args: Any) -> None:
336 self.proc = proc
337 self.args = args
338 self.status: Union[int, None] = None
340 def _terminate(self) -> None:
341 """Terminate the underlying process."""
342 if self.proc is None:
343 return
345 proc = self.proc
346 self.proc = None
347 if proc.stdin:
348 proc.stdin.close()
349 if proc.stdout:
350 proc.stdout.close()
351 if proc.stderr:
352 proc.stderr.close()
353 # Did the process finish already so we have a return code?
354 try:
355 if proc.poll() is not None:
356 self.status = self._status_code_if_terminate or proc.poll()
357 return
358 except OSError as ex:
359 _logger.info("Ignored error after process had died: %r", ex)
361 # It can be that nothing really exists anymore...
362 if os is None or getattr(os, "kill", None) is None:
363 return
365 # Try to kill it.
366 try:
367 proc.terminate()
368 status = proc.wait() # Ensure the process goes away.
370 self.status = self._status_code_if_terminate or status
371 except (OSError, AttributeError) as ex:
372 # On interpreter shutdown (notably on Windows), parts of the stdlib used by
373 # subprocess can already be torn down (e.g. `subprocess._winapi` becomes None),
374 # which can cause AttributeError during terminate(). In that case, we prefer
375 # to silently ignore to avoid noisy "Exception ignored in: __del__" messages.
376 _logger.info("Ignored error while terminating process: %r", ex)
377 # END exception handling
379 def __del__(self) -> None:
380 self._terminate()
382 def __getattr__(self, attr: str) -> Any:
383 return getattr(self.proc, attr)
385 # TODO: Bad choice to mimic `proc.wait()` but with different args.
386 def wait(self, stderr: Union[None, str, bytes] = b"") -> int:
387 """Wait for the process and return its status code.
389 :param stderr:
390 Previously read value of stderr, in case stderr is already closed.
392 :warn:
393 May deadlock if output or error pipes are used and not handled separately.
395 :raise git.exc.GitCommandError:
396 If the return status is not 0.
397 """
398 if stderr is None:
399 stderr_b = b""
400 stderr_b = force_bytes(data=stderr, encoding="utf-8")
401 status: Union[int, None]
402 if self.proc is not None:
403 status = self.proc.wait()
404 p_stderr = self.proc.stderr
405 else: # Assume the underlying proc was killed earlier or never existed.
406 status = self.status
407 p_stderr = None
409 def read_all_from_possibly_closed_stream(stream: Union[IO[bytes], None]) -> bytes:
410 if stream:
411 try:
412 return stderr_b + force_bytes(stream.read())
413 except (OSError, ValueError):
414 return stderr_b or b""
415 else:
416 return stderr_b or b""
418 # END status handling
420 if status != 0:
421 errstr = read_all_from_possibly_closed_stream(p_stderr)
422 _logger.debug("AutoInterrupt wait stderr: %r" % (errstr,))
423 raise GitCommandError(remove_password_if_present(self.args), status, errstr)
424 return status
427_AutoInterrupt.__name__ = "AutoInterrupt"
428_AutoInterrupt.__qualname__ = "Git.AutoInterrupt"
431class _CatFileContentStream:
432 """Object representing a sized read-only stream returning the contents of
433 an object.
435 This behaves like a stream, but counts the data read and simulates an empty stream
436 once our sized content region is empty.
438 If not all data are read to the end of the object's lifetime, we read the rest to
439 ensure the underlying stream continues to work.
440 """
442 __slots__ = ("_stream", "_nbr", "_size")
444 def __init__(self, size: int, stream: IO[bytes]) -> None:
445 self._stream = stream
446 self._size = size
447 self._nbr = 0 # Number of bytes read.
449 # Special case: If the object is empty, has null bytes, get the final
450 # newline right away.
451 if size == 0:
452 stream.read(1)
453 # END handle empty streams
455 def read(self, size: int = -1) -> bytes:
456 bytes_left = self._size - self._nbr
457 if bytes_left == 0:
458 return b""
459 if size > -1:
460 # Ensure we don't try to read past our limit.
461 size = min(bytes_left, size)
462 else:
463 # They try to read all, make sure it's not more than what remains.
464 size = bytes_left
465 # END check early depletion
466 data = self._stream.read(size)
467 self._nbr += len(data)
469 # Check for depletion, read our final byte to make the stream usable by
470 # others.
471 if self._size - self._nbr == 0:
472 self._stream.read(1) # final newline
473 # END finish reading
474 return data
476 def readline(self, size: int = -1) -> bytes:
477 if self._nbr == self._size:
478 return b""
480 # Clamp size to lowest allowed value.
481 bytes_left = self._size - self._nbr
482 if size > -1:
483 size = min(bytes_left, size)
484 else:
485 size = bytes_left
486 # END handle size
488 data = self._stream.readline(size)
489 self._nbr += len(data)
491 # Handle final byte.
492 if self._size - self._nbr == 0:
493 self._stream.read(1)
494 # END finish reading
496 return data
498 def readlines(self, size: int = -1) -> List[bytes]:
499 if self._nbr == self._size:
500 return []
502 # Leave all additional logic to our readline method, we just check the size.
503 out = []
504 nbr = 0
505 while True:
506 line = self.readline()
507 if not line:
508 break
509 out.append(line)
510 if size > -1:
511 nbr += len(line)
512 if nbr > size:
513 break
514 # END handle size constraint
515 # END readline loop
516 return out
518 # skipcq: PYL-E0301
519 def __iter__(self) -> "Git.CatFileContentStream":
520 return self
522 def __next__(self) -> bytes:
523 line = self.readline()
524 if not line:
525 raise StopIteration
527 return line
529 next = __next__
531 def __del__(self) -> None:
532 bytes_left = self._size - self._nbr
533 if bytes_left:
534 # Read and discard - seeking is impossible within a stream.
535 # This includes any terminating newline.
536 self._stream.read(bytes_left + 1)
537 # END handle incomplete read
540_CatFileContentStream.__name__ = "CatFileContentStream"
541_CatFileContentStream.__qualname__ = "Git.CatFileContentStream"
544_USE_SHELL_DEFAULT_MESSAGE = (
545 "Git.USE_SHELL is deprecated, because only its default value of False is safe. "
546 "It will be removed in a future release."
547)
549_USE_SHELL_DANGER_MESSAGE = (
550 "Setting Git.USE_SHELL to True is unsafe and insecure, as the effect of special "
551 "shell syntax cannot usually be accounted for. This can result in a command "
552 "injection vulnerability and arbitrary code execution. Git.USE_SHELL is deprecated "
553 "and will be removed in a future release."
554)
557def _warn_use_shell(*, extra_danger: bool) -> None:
558 warnings.warn(
559 _USE_SHELL_DANGER_MESSAGE if extra_danger else _USE_SHELL_DEFAULT_MESSAGE,
560 DeprecationWarning,
561 stacklevel=3,
562 )
565class _GitMeta(type):
566 """Metaclass for :class:`Git`.
568 This helps issue :class:`DeprecationWarning` if :attr:`Git.USE_SHELL` is used.
569 """
571 def __getattribute(cls, name: str) -> Any:
572 if name == "USE_SHELL":
573 _warn_use_shell(extra_danger=False)
574 return super().__getattribute__(name)
576 def __setattr(cls, name: str, value: Any) -> Any:
577 if name == "USE_SHELL":
578 _warn_use_shell(extra_danger=value)
579 super().__setattr__(name, value)
581 if not TYPE_CHECKING:
582 # To preserve static checking for undefined/misspelled attributes while letting
583 # the methods' bodies be type-checked, these are defined as non-special methods,
584 # then bound to special names out of view of static type checkers. (The original
585 # names invoke name mangling (leading "__") to avoid confusion in other scopes.)
586 __getattribute__ = __getattribute
587 __setattr__ = __setattr
590GitMeta = _GitMeta
591"""Alias of :class:`Git`'s metaclass, whether it is :class:`type` or a custom metaclass.
593Whether the :class:`Git` class has the default :class:`type` as its metaclass or uses a
594custom metaclass is not documented and may change at any time. This statically checkable
595metaclass alias is equivalent at runtime to ``type(Git)``. This should almost never be
596used. Code that benefits from it is likely to be remain brittle even if it is used.
598In view of the :class:`Git` class's intended use and :class:`Git` objects' dynamic
599callable attributes representing git subcommands, it rarely makes sense to inherit from
600:class:`Git` at all. Using :class:`Git` in multiple inheritance can be especially tricky
601to do correctly. Attempting uses of :class:`Git` where its metaclass is relevant, such
602as when a sibling class has an unrelated metaclass and a shared lower bound metaclass
603might have to be introduced to solve a metaclass conflict, is not recommended.
605:note:
606 The correct static type of the :class:`Git` class itself, and any subclasses, is
607 ``Type[Git]``. (This can be written as ``type[Git]`` in Python 3.9 later.)
609 :class:`GitMeta` should never be used in any annotation where ``Type[Git]`` is
610 intended or otherwise possible to use. This alias is truly only for very rare and
611 inherently precarious situations where it is necessary to deal with the metaclass
612 explicitly.
613"""
616class Git(metaclass=_GitMeta):
617 """The Git class manages communication with the Git binary.
619 It provides a convenient interface to calling the Git binary, such as in::
621 g = Git( git_dir )
622 g.init() # calls 'git init' program
623 rval = g.ls_files() # calls 'git ls-files' program
625 Debugging:
627 * Set the :envvar:`GIT_PYTHON_TRACE` environment variable to print each invocation
628 of the command to stdout.
629 * Set its value to ``full`` to see details about the returned values.
630 """
632 __slots__ = (
633 "_working_dir",
634 "cat_file_all",
635 "cat_file_header",
636 "_version_info",
637 "_version_info_token",
638 "_git_options",
639 "_persistent_git_options",
640 "_environment",
641 )
643 _excluded_ = (
644 "cat_file_all",
645 "cat_file_header",
646 "_version_info",
647 "_version_info_token",
648 )
650 re_unsafe_protocol = re.compile(r"(.+)::.+")
652 def __getstate__(self) -> Dict[str, Any]:
653 return slots_to_dict(self, exclude=self._excluded_)
655 def __setstate__(self, d: Dict[str, Any]) -> None:
656 dict_to_slots_and__excluded_are_none(self, d, excluded=self._excluded_)
658 # CONFIGURATION
660 git_exec_name = "git"
661 """Default git command that should work on Linux, Windows, and other systems."""
663 GIT_PYTHON_TRACE = os.environ.get("GIT_PYTHON_TRACE", False)
664 """Enables debugging of GitPython's git commands."""
666 USE_SHELL: bool = False
667 """Deprecated. If set to ``True``, a shell will be used when executing git commands.
669 Code that uses ``USE_SHELL = True`` or that passes ``shell=True`` to any GitPython
670 functions should be updated to use the default value of ``False`` instead. ``True``
671 is unsafe unless the effect of syntax treated specially by the shell is fully
672 considered and accounted for, which is not possible under most circumstances. As
673 detailed below, it is also no longer needed, even where it had been in the past.
675 It is in many if not most cases a command injection vulnerability for an application
676 to set :attr:`USE_SHELL` to ``True``. Any attacker who can cause a specially crafted
677 fragment of text to make its way into any part of any argument to any git command
678 (including paths, branch names, etc.) can cause the shell to read and write
679 arbitrary files and execute arbitrary commands. Innocent input may also accidentally
680 contain special shell syntax, leading to inadvertent malfunctions.
682 In addition, how a value of ``True`` interacts with some aspects of GitPython's
683 operation is not precisely specified and may change without warning, even before
684 GitPython 4.0.0 when :attr:`USE_SHELL` may be removed. This includes:
686 * Whether or how GitPython automatically customizes the shell environment.
688 * Whether, outside of Windows (where :class:`subprocess.Popen` supports lists of
689 separate arguments even when ``shell=True``), this can be used with any GitPython
690 functionality other than direct calls to the :meth:`execute` method.
692 * Whether any GitPython feature that runs git commands ever attempts to partially
693 sanitize data a shell may treat specially. Currently this is not done.
695 Prior to GitPython 2.0.8, this had a narrow purpose in suppressing console windows
696 in graphical Windows applications. In 2.0.8 and higher, it provides no benefit, as
697 GitPython solves that problem more robustly and safely by using the
698 ``CREATE_NO_WINDOW`` process creation flag on Windows.
700 Because Windows path search differs subtly based on whether a shell is used, in rare
701 cases changing this from ``True`` to ``False`` may keep an unusual git "executable",
702 such as a batch file, from being found. To fix this, set the command name or full
703 path in the :envvar:`GIT_PYTHON_GIT_EXECUTABLE` environment variable or pass the
704 full path to :func:`git.refresh` (or invoke the script using a ``.exe`` shim).
706 Further reading:
708 * :meth:`Git.execute` (on the ``shell`` parameter).
709 * https://github.com/gitpython-developers/GitPython/commit/0d9390866f9ce42870d3116094cd49e0019a970a
710 * https://learn.microsoft.com/en-us/windows/win32/procthread/process-creation-flags
711 * https://github.com/python/cpython/issues/91558#issuecomment-1100942950
712 * https://learn.microsoft.com/en-us/windows/win32/api/processthreadsapi/nf-processthreadsapi-createprocessw
713 """
715 _git_exec_env_var = "GIT_PYTHON_GIT_EXECUTABLE"
716 _refresh_env_var = "GIT_PYTHON_REFRESH"
718 GIT_PYTHON_GIT_EXECUTABLE = None
719 """Provide the full path to the git executable. Otherwise it assumes git is in the
720 executable search path.
722 :note:
723 The git executable is actually found during the refresh step in the top level
724 ``__init__``. It can also be changed by explicitly calling :func:`git.refresh`.
725 """
727 _refresh_token = object() # Since None would match an initial _version_info_token.
729 @classmethod
730 def refresh(cls, path: Union[None, PathLike] = None) -> bool:
731 """Update information about the git executable :class:`Git` objects will use.
733 Called by the :func:`git.refresh` function in the top level ``__init__``.
735 :param path:
736 Optional path to the git executable. If not absolute, it is resolved
737 immediately, relative to the current directory. (See note below.)
739 :note:
740 The top-level :func:`git.refresh` should be preferred because it calls this
741 method and may also update other state accordingly.
743 :note:
744 There are three different ways to specify the command that refreshing causes
745 to be used for git:
747 1. Pass no `path` argument and do not set the
748 :envvar:`GIT_PYTHON_GIT_EXECUTABLE` environment variable. The command
749 name ``git`` is used. It is looked up in a path search by the system, in
750 each command run (roughly similar to how git is found when running
751 ``git`` commands manually). This is usually the desired behavior.
753 2. Pass no `path` argument but set the :envvar:`GIT_PYTHON_GIT_EXECUTABLE`
754 environment variable. The command given as the value of that variable is
755 used. This may be a simple command or an arbitrary path. It is looked up
756 in each command run. Setting :envvar:`GIT_PYTHON_GIT_EXECUTABLE` to
757 ``git`` has the same effect as not setting it.
759 3. Pass a `path` argument. This path, if not absolute, is immediately
760 resolved, relative to the current directory. This resolution occurs at
761 the time of the refresh. When git commands are run, they are run using
762 that previously resolved path. If a `path` argument is passed, the
763 :envvar:`GIT_PYTHON_GIT_EXECUTABLE` environment variable is not
764 consulted.
766 :note:
767 Refreshing always sets the :attr:`Git.GIT_PYTHON_GIT_EXECUTABLE` class
768 attribute, which can be read on the :class:`Git` class or any of its
769 instances to check what command is used to run git. This attribute should
770 not be confused with the related :envvar:`GIT_PYTHON_GIT_EXECUTABLE`
771 environment variable. The class attribute is set no matter how refreshing is
772 performed.
773 """
774 # Discern which path to refresh with.
775 if path is not None:
776 new_git = os.path.expanduser(path)
777 new_git = os.path.abspath(new_git)
778 else:
779 new_git = os.environ.get(cls._git_exec_env_var, cls.git_exec_name)
781 # Keep track of the old and new git executable path.
782 old_git = cls.GIT_PYTHON_GIT_EXECUTABLE
783 old_refresh_token = cls._refresh_token
784 cls.GIT_PYTHON_GIT_EXECUTABLE = new_git
785 cls._refresh_token = object()
787 # Test if the new git executable path is valid. A GitCommandNotFound error is
788 # raised by us. A PermissionError is raised if the git executable cannot be
789 # executed for whatever reason.
790 has_git = False
791 try:
792 cls().version()
793 has_git = True
794 except (GitCommandNotFound, PermissionError):
795 pass
797 # Warn or raise exception if test failed.
798 if not has_git:
799 err = (
800 dedent(
801 """\
802 Bad git executable.
803 The git executable must be specified in one of the following ways:
804 - be included in your $PATH
805 - be set via $%s
806 - explicitly set via git.refresh(<full-path-to-git-executable>)
807 """
808 )
809 % cls._git_exec_env_var
810 )
812 # Revert to whatever the old_git was.
813 cls.GIT_PYTHON_GIT_EXECUTABLE = old_git
814 cls._refresh_token = old_refresh_token
816 if old_git is None:
817 # On the first refresh (when GIT_PYTHON_GIT_EXECUTABLE is None) we only
818 # are quiet, warn, or error depending on the GIT_PYTHON_REFRESH value.
820 # Determine what the user wants to happen during the initial refresh. We
821 # expect GIT_PYTHON_REFRESH to either be unset or be one of the
822 # following values:
823 #
824 # 0|q|quiet|s|silence|silent|n|none
825 # 1|w|warn|warning|l|log
826 # 2|r|raise|e|error|exception
828 mode = os.environ.get(cls._refresh_env_var, "raise").lower()
830 quiet = ["quiet", "q", "silence", "s", "silent", "none", "n", "0"]
831 warn = ["warn", "w", "warning", "log", "l", "1"]
832 error = ["error", "e", "exception", "raise", "r", "2"]
834 if mode in quiet:
835 pass
836 elif mode in warn or mode in error:
837 err = dedent(
838 """\
839 %s
840 All git commands will error until this is rectified.
842 This initial message can be silenced or aggravated in the future by setting the
843 $%s environment variable. Use one of the following values:
844 - %s: for no message or exception
845 - %s: for a warning message (logging level CRITICAL, displayed by default)
846 - %s: for a raised exception
848 Example:
849 export %s=%s
850 """
851 ) % (
852 err,
853 cls._refresh_env_var,
854 "|".join(quiet),
855 "|".join(warn),
856 "|".join(error),
857 cls._refresh_env_var,
858 quiet[0],
859 )
861 if mode in warn:
862 _logger.critical(err)
863 else:
864 raise ImportError(err)
865 else:
866 err = dedent(
867 """\
868 %s environment variable has been set but it has been set with an invalid value.
870 Use only the following values:
871 - %s: for no message or exception
872 - %s: for a warning message (logging level CRITICAL, displayed by default)
873 - %s: for a raised exception
874 """
875 ) % (
876 cls._refresh_env_var,
877 "|".join(quiet),
878 "|".join(warn),
879 "|".join(error),
880 )
881 raise ImportError(err)
883 # We get here if this was the initial refresh and the refresh mode was
884 # not error. Go ahead and set the GIT_PYTHON_GIT_EXECUTABLE such that we
885 # discern the difference between the first refresh at import time
886 # and subsequent calls to git.refresh or this refresh method.
887 cls.GIT_PYTHON_GIT_EXECUTABLE = cls.git_exec_name
888 else:
889 # After the first refresh (when GIT_PYTHON_GIT_EXECUTABLE is no longer
890 # None) we raise an exception.
891 raise GitCommandNotFound(new_git, err)
893 return has_git
895 @classmethod
896 def is_cygwin(cls) -> bool:
897 return is_cygwin_git(cls.GIT_PYTHON_GIT_EXECUTABLE)
899 @overload
900 @classmethod
901 def polish_url(cls, url: str, is_cygwin: Literal[False] = ...) -> str: ...
903 @overload
904 @classmethod
905 def polish_url(cls, url: str, is_cygwin: Union[None, bool] = None) -> str: ...
907 @classmethod
908 def polish_url(cls, url: str, is_cygwin: Union[None, bool] = None) -> PathLike:
909 """Remove any backslashes from URLs to be written in config files.
911 Windows might create config files containing paths with backslashes, but git
912 stops liking them as it will escape the backslashes. Hence we undo the escaping
913 just to be sure.
914 """
915 if is_cygwin is None:
916 is_cygwin = cls.is_cygwin()
918 if is_cygwin:
919 url = cygpath(url)
920 else:
921 url = os.path.expandvars(url)
922 if url.startswith("~"):
923 url = os.path.expanduser(url)
924 url = url.replace("\\\\", "\\").replace("\\", "/")
925 return url
927 @classmethod
928 def check_unsafe_protocols(cls, url: str) -> None:
929 """Check for unsafe protocols.
931 Apart from the usual protocols (http, git, ssh), Git allows "remote helpers"
932 that have the form ``<transport>::<address>``. One of these helpers (``ext::``)
933 can be used to invoke any arbitrary command.
935 See:
937 - https://git-scm.com/docs/gitremote-helpers
938 - https://git-scm.com/docs/git-remote-ext
939 """
940 match = cls.re_unsafe_protocol.match(url)
941 if match:
942 protocol = match.group(1)
943 raise UnsafeProtocolError(
944 f"The `{protocol}::` protocol looks suspicious, use `allow_unsafe_protocols=True` to allow it."
945 )
947 @classmethod
948 def _canonicalize_option_name(cls, option: str) -> str:
949 """Return the option name used for unsafe-option checks.
951 Examples:
952 ``"--upload-pack=/tmp/helper"`` -> ``"upload-pack"``
953 ``"upload_pack"`` -> ``"upload-pack"``
954 ``"--config core.filemode=false"`` -> ``"config"``
955 """
956 option_name = option.lstrip("-").split("=", 1)[0]
957 option_tokens = option_name.split(None, 1)
958 if not option_tokens:
959 return ""
960 return dashify(option_tokens[0])
962 @classmethod
963 def check_unsafe_options(cls, options: List[str], unsafe_options: List[str]) -> None:
964 """Check for unsafe options.
966 Some options that are passed to ``git <command>`` can be used to execute
967 arbitrary commands. These are blocked by default.
968 """
969 # Options can be of the form `foo`, `--foo`, `--foo bar`, or `--foo=bar`.
970 canonical_unsafe_options = {cls._canonicalize_option_name(option): option for option in unsafe_options}
971 for option in options:
972 unsafe_option = canonical_unsafe_options.get(cls._canonicalize_option_name(option))
973 if unsafe_option is not None:
974 raise UnsafeOptionError(f"{unsafe_option} is not allowed, use `allow_unsafe_options=True` to allow it.")
976 AutoInterrupt: TypeAlias = _AutoInterrupt
978 CatFileContentStream: TypeAlias = _CatFileContentStream
980 def __init__(self, working_dir: Union[None, PathLike] = None) -> None:
981 """Initialize this instance with:
983 :param working_dir:
984 Git directory we should work in. If ``None``, we always work in the current
985 directory as returned by :func:`os.getcwd`.
986 This is meant to be the working tree directory if available, or the
987 ``.git`` directory in case of bare repositories.
988 """
989 super().__init__()
990 self._working_dir = expand_path(working_dir)
991 self._git_options: Union[List[str], Tuple[str, ...]] = ()
992 self._persistent_git_options: List[str] = []
994 # Extra environment variables to pass to git commands
995 self._environment: Dict[str, str] = {}
997 # Cached version slots
998 self._version_info: Union[Tuple[int, ...], None] = None
999 self._version_info_token: object = None
1001 # Cached command slots
1002 self.cat_file_header: Union[None, TBD] = None
1003 self.cat_file_all: Union[None, TBD] = None
1005 def __getattribute__(self, name: str) -> Any:
1006 if name == "USE_SHELL":
1007 _warn_use_shell(extra_danger=False)
1008 return super().__getattribute__(name)
1010 def __getattr__(self, name: str) -> Any:
1011 """A convenience method as it allows to call the command as if it was an object.
1013 :return:
1014 Callable object that will execute call :meth:`_call_process` with your
1015 arguments.
1016 """
1017 if name.startswith("_"):
1018 return super().__getattribute__(name)
1019 return lambda *args, **kwargs: self._call_process(name, *args, **kwargs)
1021 def set_persistent_git_options(self, **kwargs: Any) -> None:
1022 """Specify command line options to the git executable for subsequent
1023 subcommand calls.
1025 :param kwargs:
1026 A dict of keyword arguments.
1027 These arguments are passed as in :meth:`_call_process`, but will be passed
1028 to the git command rather than the subcommand.
1029 """
1031 self._persistent_git_options = self.transform_kwargs(split_single_char_options=True, **kwargs)
1033 @property
1034 def working_dir(self) -> Union[None, PathLike]:
1035 """:return: Git directory we are working on"""
1036 return self._working_dir
1038 @property
1039 def version_info(self) -> Tuple[int, ...]:
1040 """
1041 :return: Tuple with integers representing the major, minor and additional
1042 version numbers as parsed from :manpage:`git-version(1)`. Up to four fields
1043 are used.
1045 This value is generated on demand and is cached.
1046 """
1047 # Refreshing is global, but version_info caching is per-instance.
1048 refresh_token = self._refresh_token # Copy token in case of concurrent refresh.
1050 # Use the cached version if obtained after the most recent refresh.
1051 if self._version_info_token is refresh_token:
1052 assert self._version_info is not None, "Bug: corrupted token-check state"
1053 return self._version_info
1055 # Run "git version" and parse it.
1056 process_version = self._call_process("version")
1057 version_string = process_version.split(" ")[2]
1058 version_fields = version_string.split(".")[:4]
1059 leading_numeric_fields = itertools.takewhile(str.isdigit, version_fields)
1060 self._version_info = tuple(map(int, leading_numeric_fields))
1062 # This value will be considered valid until the next refresh.
1063 self._version_info_token = refresh_token
1064 return self._version_info
1066 @overload
1067 def execute(
1068 self,
1069 command: Union[str, Sequence[Any]],
1070 *,
1071 as_process: Literal[True],
1072 ) -> "AutoInterrupt": ...
1074 @overload
1075 def execute(
1076 self,
1077 command: Union[str, Sequence[Any]],
1078 *,
1079 as_process: Literal[False] = False,
1080 stdout_as_string: Literal[True],
1081 ) -> Union[str, Tuple[int, str, str]]: ...
1083 @overload
1084 def execute(
1085 self,
1086 command: Union[str, Sequence[Any]],
1087 *,
1088 as_process: Literal[False] = False,
1089 stdout_as_string: Literal[False] = False,
1090 ) -> Union[bytes, Tuple[int, bytes, str]]: ...
1092 @overload
1093 def execute(
1094 self,
1095 command: Union[str, Sequence[Any]],
1096 *,
1097 with_extended_output: Literal[False],
1098 as_process: Literal[False],
1099 stdout_as_string: Literal[True],
1100 ) -> str: ...
1102 @overload
1103 def execute(
1104 self,
1105 command: Union[str, Sequence[Any]],
1106 *,
1107 with_extended_output: Literal[False],
1108 as_process: Literal[False],
1109 stdout_as_string: Literal[False],
1110 ) -> bytes: ...
1112 def execute(
1113 self,
1114 command: Union[str, Sequence[Any]],
1115 istream: Union[None, BinaryIO] = None,
1116 with_extended_output: bool = False,
1117 with_exceptions: bool = True,
1118 as_process: bool = False,
1119 output_stream: Union[None, BinaryIO] = None,
1120 stdout_as_string: bool = True,
1121 kill_after_timeout: Union[None, float] = None,
1122 with_stdout: bool = True,
1123 universal_newlines: bool = False,
1124 shell: Union[None, bool] = None,
1125 env: Union[None, Mapping[str, str]] = None,
1126 max_chunk_size: int = io.DEFAULT_BUFFER_SIZE,
1127 strip_newline_in_stdout: bool = True,
1128 **subprocess_kwargs: Any,
1129 ) -> Union[str, bytes, Tuple[int, Union[str, bytes], str], AutoInterrupt]:
1130 R"""Handle executing the command, and consume and return the returned
1131 information (stdout).
1133 :param command:
1134 The command to execute. A sequence of program arguments is recommended.
1135 A string is also accepted, but its meaning is strongly platform-dependent.
1137 By default, a shell is not used. On Unix-like systems, a string is the whole
1138 program name (so ``"git log -n 1"`` raises :class:`GitCommandNotFound`). On
1139 Windows, the program parses the arguments itself, so multi-word strings can
1140 work but are not portable.
1142 Avoid ``shell=True`` (and :attr:`Git.USE_SHELL`): this runs the command in
1143 a shell, which is generally unsafe. The shell interprets metacharacters
1144 such as ``;``, ``|``, ``&``, ``$(...)``, ``$VAR``, ``%VAR%``, and ``^``
1145 (depending on the platform) as syntax. Any untrusted text in the command
1146 can then execute arbitrary OS commands. See :attr:`Git.USE_SHELL`.
1148 Producing a sequence automatically by :func:`shlex.split` and passing it
1149 as the command is far safer than ``shell=True``. But :func:`shlex.split`
1150 parses POSIX shell syntax on all systems, and the result is still unsafe
1151 for anything but *fixed, fully trusted* strings. Do not use it on strings
1152 built by interpolating values: whitespace or quoting in an untrusted value
1153 can still inject arguments. For input derived in any way from untrusted
1154 data, build the argument sequence yourself, while ensuring each argument
1155 is fully sanitized.
1157 :param istream:
1158 Standard input filehandle passed to :class:`subprocess.Popen`.
1160 :param with_extended_output:
1161 Whether to return a (status, stdout, stderr) tuple.
1163 :param with_exceptions:
1164 Whether to raise an exception when git returns a non-zero status.
1166 :param as_process:
1167 Whether to return the created process instance directly from which
1168 streams can be read on demand. This will render `with_extended_output`
1169 and `with_exceptions` ineffective - the caller will have to deal with
1170 the details. It is important to note that the process will be placed
1171 into an :class:`AutoInterrupt` wrapper that will interrupt the process
1172 once it goes out of scope. If you use the command in iterators, you
1173 should pass the whole process instance instead of a single stream.
1175 :param output_stream:
1176 If set to a file-like object, data produced by the git command will be
1177 copied to the given stream instead of being returned as a string.
1178 This feature only has any effect if `as_process` is ``False``.
1180 :param stdout_as_string:
1181 If ``False``, the command's standard output will be bytes. Otherwise, it
1182 will be decoded into a string using the default encoding (usually UTF-8).
1183 The latter can fail, if the output contains binary data.
1185 :param kill_after_timeout:
1186 Specifies a timeout in seconds for the git command, after which the process
1187 should be killed. This will have no effect if `as_process` is set to
1188 ``True``. It is set to ``None`` by default and will let the process run
1189 until the timeout is explicitly specified. Uses of this feature should be
1190 carefully considered, due to the following limitations:
1192 1. This feature is not supported at all on Windows.
1193 2. Effectiveness may vary by operating system. ``ps --ppid`` is used to
1194 enumerate child processes, which is available on most GNU/Linux systems
1195 but not most others.
1196 3. Deeper descendants do not receive signals, though they may sometimes
1197 terminate as a consequence of their parent processes being killed.
1198 4. `kill_after_timeout` uses ``SIGKILL``, which can have negative side
1199 effects on a repository. For example, stale locks in case of
1200 :manpage:`git-gc(1)` could render the repository incapable of accepting
1201 changes until the lock is manually removed.
1203 :param with_stdout:
1204 If ``True``, default ``True``, we open stdout on the created process.
1206 :param universal_newlines:
1207 If ``True``, pipes will be opened as text, and lines are split at all known
1208 line endings.
1210 :param shell:
1211 Whether to invoke commands through a shell
1212 (see :class:`Popen(..., shell=True) <subprocess.Popen>`).
1213 If this is not ``None``, it overrides :attr:`USE_SHELL`.
1215 Passing ``shell=True`` to this or any other GitPython function should be
1216 avoided, as it is unsafe under most circumstances. This is because it is
1217 typically not feasible to fully consider and account for the effect of shell
1218 expansions, especially when passing ``shell=True`` to other methods that
1219 forward it to :meth:`Git.execute`. Passing ``shell=True`` is also no longer
1220 needed (nor useful) to work around any known operating system specific
1221 issues.
1223 On Unix-like systems, when migrating away from passing string commands with
1224 ``shell=True``, :func:`shlex.split` may serve as a transitional step in rare
1225 cases, with extreme care. (Drop ``shell=True`` and pass the resulting
1226 sequence as the command.) See the `command` parameter above on the risks.
1228 :param env:
1229 A dictionary of environment variables to be passed to
1230 :class:`subprocess.Popen`.
1232 :param max_chunk_size:
1233 Maximum number of bytes in one chunk of data passed to the `output_stream`
1234 in one invocation of its ``write()`` method. If the given number is not
1235 positive then the default value is used.
1237 :param strip_newline_in_stdout:
1238 Whether to strip the trailing ``\n`` of the command stdout.
1240 :param subprocess_kwargs:
1241 Keyword arguments to be passed to :class:`subprocess.Popen`. Please note
1242 that some of the valid kwargs are already set by this method; the ones you
1243 specify may not be the same ones.
1245 :return:
1246 * str(output), if `extended_output` is ``False`` (Default)
1247 * tuple(int(status), str(stdout), str(stderr)),
1248 if `extended_output` is ``True``
1250 If `output_stream` is ``True``, the stdout value will be your output stream:
1252 * output_stream, if `extended_output` is ``False``
1253 * tuple(int(status), output_stream, str(stderr)),
1254 if `extended_output` is ``True``
1256 Note that git is executed with ``LC_MESSAGES="C"`` to ensure consistent
1257 output regardless of system language.
1259 :raise git.exc.GitCommandError:
1261 :note:
1262 If you add additional keyword arguments to the signature of this method, you
1263 must update the ``execute_kwargs`` variable housed in this module.
1264 """
1265 # Remove password for the command if present.
1266 redacted_command = remove_password_if_present(command)
1267 if self.GIT_PYTHON_TRACE and (self.GIT_PYTHON_TRACE != "full" or as_process):
1268 _logger.info(" ".join(redacted_command))
1270 # Allow the user to have the command executed in their working dir.
1271 try:
1272 cwd = self._working_dir or os.getcwd() # type: Union[None, str]
1273 if not os.access(str(cwd), os.X_OK):
1274 cwd = None
1275 except FileNotFoundError:
1276 cwd = None
1278 # Start the process.
1279 inline_env = env
1280 env = os.environ.copy()
1281 # Attempt to force all output to plain ASCII English, which is what some parsing
1282 # code may expect.
1283 # According to https://askubuntu.com/a/311796, we are setting LANGUAGE as well
1284 # just to be sure.
1285 env["LANGUAGE"] = "C"
1286 env["LC_ALL"] = "C"
1287 env.update(self._environment)
1288 if inline_env is not None:
1289 env.update(inline_env)
1291 if sys.platform == "win32":
1292 if kill_after_timeout is not None:
1293 raise GitCommandError(
1294 redacted_command,
1295 '"kill_after_timeout" feature is not supported on Windows.',
1296 )
1297 cmd_not_found_exception = OSError
1298 else:
1299 cmd_not_found_exception = FileNotFoundError
1300 # END handle
1302 stdout_sink = PIPE if with_stdout else getattr(subprocess, "DEVNULL", None) or open(os.devnull, "wb")
1303 if shell is None:
1304 # Get the value of USE_SHELL with no deprecation warning. Do this without
1305 # warnings.catch_warnings, to avoid a race condition with application code
1306 # configuring warnings. The value could be looked up in type(self).__dict__
1307 # or Git.__dict__, but those can break under some circumstances. This works
1308 # the same as self.USE_SHELL in more situations; see Git.__getattribute__.
1309 shell = super().__getattribute__("USE_SHELL")
1310 _logger.debug(
1311 "Popen(%s, cwd=%s, stdin=%s, shell=%s, universal_newlines=%s)",
1312 redacted_command,
1313 cwd,
1314 "<valid stream>" if istream else "None",
1315 shell,
1316 universal_newlines,
1317 )
1318 try:
1319 proc = safer_popen(
1320 command,
1321 env=env,
1322 cwd=cwd,
1323 bufsize=-1,
1324 stdin=(istream or DEVNULL),
1325 stderr=PIPE,
1326 stdout=stdout_sink,
1327 shell=shell,
1328 universal_newlines=universal_newlines,
1329 encoding=defenc if universal_newlines else None,
1330 **subprocess_kwargs,
1331 )
1332 except cmd_not_found_exception as err:
1333 raise GitCommandNotFound(redacted_command, err) from err
1334 else:
1335 # Replace with a typeguard for Popen[bytes]?
1336 proc.stdout = cast(BinaryIO, proc.stdout)
1337 proc.stderr = cast(BinaryIO, proc.stderr)
1339 if as_process:
1340 return self.AutoInterrupt(proc, command)
1342 if sys.platform != "win32" and kill_after_timeout is not None:
1343 # Help mypy figure out this is not None even when used inside communicate().
1344 timeout = kill_after_timeout
1346 def kill_process(pid: int) -> None:
1347 """Callback to kill a process.
1349 This callback implementation would be ineffective and unsafe on Windows.
1350 """
1351 p = Popen(["ps", "--ppid", str(pid)], stdout=PIPE)
1352 child_pids = []
1353 if p.stdout is not None:
1354 for line in p.stdout:
1355 if len(line.split()) > 0:
1356 local_pid = (line.split())[0]
1357 if local_pid.isdigit():
1358 child_pids.append(int(local_pid))
1359 try:
1360 os.kill(pid, signal.SIGKILL)
1361 for child_pid in child_pids:
1362 try:
1363 os.kill(child_pid, signal.SIGKILL)
1364 except OSError:
1365 pass
1366 # Tell the main routine that the process was killed.
1367 kill_check.set()
1368 except OSError:
1369 # It is possible that the process gets completed in the duration
1370 # after timeout happens and before we try to kill the process.
1371 pass
1372 return
1374 def communicate() -> Tuple[AnyStr, AnyStr]:
1375 watchdog.start()
1376 out, err = proc.communicate()
1377 watchdog.cancel()
1378 if kill_check.is_set():
1379 err = 'Timeout: the command "%s" did not complete in %d secs.' % (
1380 " ".join(redacted_command),
1381 timeout,
1382 )
1383 if not universal_newlines:
1384 err = err.encode(defenc)
1385 return out, err
1387 # END helpers
1389 kill_check = threading.Event()
1390 watchdog = threading.Timer(timeout, kill_process, args=(proc.pid,))
1391 else:
1392 communicate = proc.communicate
1394 # Wait for the process to return.
1395 status = 0
1396 stdout_value: Union[str, bytes] = b""
1397 stderr_value: Union[str, bytes] = b""
1398 newline = "\n" if universal_newlines else b"\n"
1399 try:
1400 if output_stream is None:
1401 stdout_value, stderr_value = communicate()
1402 # Strip trailing "\n".
1403 if stdout_value is not None and stdout_value.endswith(newline) and strip_newline_in_stdout: # type: ignore[arg-type]
1404 stdout_value = stdout_value[:-1]
1405 if stderr_value is not None and stderr_value.endswith(newline): # type: ignore[arg-type]
1406 stderr_value = stderr_value[:-1]
1408 status = proc.returncode
1409 else:
1410 max_chunk_size = max_chunk_size if max_chunk_size and max_chunk_size > 0 else io.DEFAULT_BUFFER_SIZE
1411 if proc.stdout is not None:
1412 stream_copy(proc.stdout, output_stream, max_chunk_size)
1413 stdout_value = proc.stdout.read()
1414 if proc.stderr is not None:
1415 stderr_value = proc.stderr.read()
1416 # Strip trailing "\n".
1417 if stderr_value is not None and stderr_value.endswith(newline): # type: ignore[arg-type]
1418 stderr_value = stderr_value[:-1]
1419 status = proc.wait()
1420 # END stdout handling
1421 finally:
1422 if proc.stdout is not None:
1423 proc.stdout.close()
1424 if proc.stderr is not None:
1425 proc.stderr.close()
1427 if self.GIT_PYTHON_TRACE == "full":
1428 cmdstr = " ".join(redacted_command)
1430 def as_text(stdout_value: Union[bytes, str]) -> str:
1431 return not output_stream and safe_decode(stdout_value) or "<OUTPUT_STREAM>"
1433 # END as_text
1435 if stderr_value:
1436 _logger.info(
1437 "%s -> %d; stdout: '%s'; stderr: '%s'",
1438 cmdstr,
1439 status,
1440 as_text(stdout_value),
1441 safe_decode(stderr_value),
1442 )
1443 elif stdout_value:
1444 _logger.info("%s -> %d; stdout: '%s'", cmdstr, status, as_text(stdout_value))
1445 else:
1446 _logger.info("%s -> %d", cmdstr, status)
1447 # END handle debug printing
1449 if with_exceptions and status != 0:
1450 raise GitCommandError(redacted_command, status, stderr_value, stdout_value)
1452 if isinstance(stdout_value, bytes) and stdout_as_string: # Could also be output_stream.
1453 stdout_value = safe_decode(stdout_value)
1455 # Allow access to the command's status code.
1456 if with_extended_output:
1457 return (status, stdout_value, safe_decode(stderr_value))
1458 else:
1459 return stdout_value
1461 def environment(self) -> Dict[str, str]:
1462 return self._environment
1464 def update_environment(self, **kwargs: Any) -> Dict[str, Union[str, None]]:
1465 """Set environment variables for future git invocations. Return all changed
1466 values in a format that can be passed back into this function to revert the
1467 changes.
1469 Examples::
1471 old_env = self.update_environment(PWD='/tmp')
1472 self.update_environment(**old_env)
1474 :param kwargs:
1475 Environment variables to use for git processes.
1477 :return:
1478 Dict that maps environment variables to their old values
1479 """
1480 old_env = {}
1481 for key, value in kwargs.items():
1482 # Set value if it is None.
1483 if value is not None:
1484 old_env[key] = self._environment.get(key)
1485 self._environment[key] = value
1486 # Remove key from environment if its value is None.
1487 elif key in self._environment:
1488 old_env[key] = self._environment[key]
1489 del self._environment[key]
1490 return old_env
1492 @contextlib.contextmanager
1493 def custom_environment(self, **kwargs: Any) -> Iterator[None]:
1494 """A context manager around the above :meth:`update_environment` method to
1495 restore the environment back to its previous state after operation.
1497 Examples::
1499 with self.custom_environment(GIT_SSH='/bin/ssh_wrapper'):
1500 repo.remotes.origin.fetch()
1502 :param kwargs:
1503 See :meth:`update_environment`.
1504 """
1505 old_env = self.update_environment(**kwargs)
1506 try:
1507 yield
1508 finally:
1509 self.update_environment(**old_env)
1511 def transform_kwarg(self, name: str, value: Any, split_single_char_options: bool) -> List[str]:
1512 if len(name) == 1:
1513 if value is True:
1514 return ["-%s" % name]
1515 elif value not in (False, None):
1516 if split_single_char_options:
1517 return ["-%s" % name, "%s" % value]
1518 else:
1519 return ["-%s%s" % (name, value)]
1520 else:
1521 if value is True:
1522 return ["--%s" % dashify(name)]
1523 elif value is not False and value is not None:
1524 return ["--%s=%s" % (dashify(name), value)]
1525 return []
1527 def transform_kwargs(self, split_single_char_options: bool = True, **kwargs: Any) -> List[str]:
1528 """Transform Python-style kwargs into git command line options."""
1529 args = []
1530 for k, v in kwargs.items():
1531 if isinstance(v, (list, tuple)):
1532 for value in v:
1533 args += self.transform_kwarg(k, value, split_single_char_options)
1534 else:
1535 args += self.transform_kwarg(k, v, split_single_char_options)
1536 return args
1538 @classmethod
1539 def _unpack_args(cls, arg_list: Sequence[str]) -> List[str]:
1540 outlist = []
1541 if isinstance(arg_list, (list, tuple)):
1542 for arg in arg_list:
1543 outlist.extend(cls._unpack_args(arg))
1544 else:
1545 outlist.append(str(arg_list))
1547 return outlist
1549 def __call__(self, **kwargs: Any) -> "Git":
1550 """Specify command line options to the git executable for a subcommand call.
1552 :param kwargs:
1553 A dict of keyword arguments.
1554 These arguments are passed as in :meth:`_call_process`, but will be passed
1555 to the git command rather than the subcommand.
1557 Examples::
1559 git(work_tree='/tmp').difftool()
1560 """
1561 self._git_options = self.transform_kwargs(split_single_char_options=True, **kwargs)
1562 return self
1564 @overload
1565 def _call_process(
1566 self, method: str, *args: None, **kwargs: None
1567 ) -> str: ... # If no args were given, execute the call with all defaults.
1569 @overload
1570 def _call_process(
1571 self,
1572 method: str,
1573 istream: int,
1574 as_process: Literal[True],
1575 *args: Any,
1576 **kwargs: Any,
1577 ) -> "Git.AutoInterrupt": ...
1579 @overload
1580 def _call_process(
1581 self, method: str, *args: Any, **kwargs: Any
1582 ) -> Union[str, bytes, Tuple[int, Union[str, bytes], str], "Git.AutoInterrupt"]: ...
1584 def _call_process(
1585 self, method: str, *args: Any, **kwargs: Any
1586 ) -> Union[str, bytes, Tuple[int, Union[str, bytes], str], "Git.AutoInterrupt"]:
1587 """Run the given git command with the specified arguments and return the result
1588 as a string.
1590 :param method:
1591 The command. Contained ``_`` characters will be converted to hyphens, such
1592 as in ``ls_files`` to call ``ls-files``.
1594 :param args:
1595 The list of arguments. If ``None`` is included, it will be pruned.
1596 This allows your commands to call git more conveniently, as ``None`` is
1597 realized as non-existent.
1599 :param kwargs:
1600 Contains key-values for the following:
1602 - The :meth:`execute()` kwds, as listed in ``execute_kwargs``.
1603 - "Command options" to be converted by :meth:`transform_kwargs`.
1604 - The ``insert_kwargs_after`` key which its value must match one of
1605 ``*args``.
1607 It also contains any command options, to be appended after the matched arg.
1609 Examples::
1611 git.rev_list('master', max_count=10, header=True)
1613 turns into::
1615 git rev-list --max-count=10 --header master
1617 :return:
1618 Same as :meth:`execute`. If no args are given, used :meth:`execute`'s
1619 default (especially ``as_process = False``, ``stdout_as_string = True``) and
1620 return :class:`str`.
1621 """
1622 # Handle optional arguments prior to calling transform_kwargs.
1623 # Otherwise these'll end up in args, which is bad.
1624 exec_kwargs = {k: v for k, v in kwargs.items() if k in execute_kwargs}
1625 opts_kwargs = {k: v for k, v in kwargs.items() if k not in execute_kwargs}
1627 insert_after_this_arg = opts_kwargs.pop("insert_kwargs_after", None)
1629 # Prepare the argument list.
1631 opt_args = self.transform_kwargs(**opts_kwargs)
1632 ext_args = self._unpack_args([a for a in args if a is not None])
1634 if insert_after_this_arg is None:
1635 args_list = opt_args + ext_args
1636 else:
1637 try:
1638 index = ext_args.index(insert_after_this_arg)
1639 except ValueError as err:
1640 raise ValueError(
1641 "Couldn't find argument '%s' in args %s to insert cmd options after"
1642 % (insert_after_this_arg, str(ext_args))
1643 ) from err
1644 # END handle error
1645 args_list = ext_args[: index + 1] + opt_args + ext_args[index + 1 :]
1646 # END handle opts_kwargs
1648 call = [self.GIT_PYTHON_GIT_EXECUTABLE]
1650 # Add persistent git options.
1651 call.extend(self._persistent_git_options)
1653 # Add the git options, then reset to empty to avoid side effects.
1654 call.extend(self._git_options)
1655 self._git_options = ()
1657 call.append(dashify(method))
1658 call.extend(args_list)
1660 return self.execute(call, **exec_kwargs)
1662 def _parse_object_header(self, header_line: str) -> Tuple[str, str, int]:
1663 """
1664 :param header_line:
1665 A line of the form::
1667 <hex_sha> type_string size_as_int
1669 :return:
1670 (hex_sha, type_string, size_as_int)
1672 :raise ValueError:
1673 If the header contains indication for an error due to incorrect input sha.
1674 """
1675 tokens = header_line.split()
1676 if len(tokens) != 3:
1677 if not tokens:
1678 err_msg = (
1679 f"SHA is empty, possible dubious ownership in the repository "
1680 f"""at {self._working_dir}.\n If this is unintended run:\n\n """
1681 f""" "git config --global --add safe.directory {self._working_dir}" """
1682 )
1683 raise ValueError(err_msg)
1684 else:
1685 raise ValueError("SHA %s could not be resolved, git returned: %r" % (tokens[0], header_line.strip()))
1686 # END handle actual return value
1687 # END error handling
1689 if len(tokens[0]) != 40:
1690 raise ValueError("Failed to parse header: %r" % header_line)
1691 return (tokens[0], tokens[1], int(tokens[2]))
1693 def _prepare_ref(self, ref: AnyStr) -> bytes:
1694 # Required for command to separate refs on stdin, as bytes.
1695 if isinstance(ref, bytes):
1696 # Assume 40 bytes hexsha - bin-to-ascii for some reason returns bytes, not text.
1697 refstr: str = ref.decode("ascii")
1698 elif not isinstance(ref, str):
1699 refstr = str(ref) # Could be ref-object.
1700 else:
1701 refstr = ref
1703 if not refstr.endswith("\n"):
1704 refstr += "\n"
1705 return refstr.encode(defenc)
1707 def _get_persistent_cmd(self, attr_name: str, cmd_name: str, *args: Any, **kwargs: Any) -> "Git.AutoInterrupt":
1708 cur_val = getattr(self, attr_name)
1709 if cur_val is not None:
1710 return cur_val
1712 options = {"istream": PIPE, "as_process": True}
1713 options.update(kwargs)
1715 cmd = self._call_process(cmd_name, *args, **options)
1716 setattr(self, attr_name, cmd)
1717 cmd = cast("Git.AutoInterrupt", cmd)
1718 return cmd
1720 def __get_object_header(self, cmd: "Git.AutoInterrupt", ref: AnyStr) -> Tuple[str, str, int]:
1721 if cmd.stdin and cmd.stdout:
1722 cmd.stdin.write(self._prepare_ref(ref))
1723 cmd.stdin.flush()
1724 return self._parse_object_header(cmd.stdout.readline())
1725 else:
1726 raise ValueError("cmd stdin was empty")
1728 def get_object_header(self, ref: str) -> Tuple[str, str, int]:
1729 """Use this method to quickly examine the type and size of the object behind the
1730 given ref.
1732 :note:
1733 The method will only suffer from the costs of command invocation once and
1734 reuses the command in subsequent calls.
1736 :return:
1737 (hexsha, type_string, size_as_int)
1738 """
1739 cmd = self._get_persistent_cmd("cat_file_header", "cat_file", batch_check=True)
1740 return self.__get_object_header(cmd, ref)
1742 def get_object_data(self, ref: str) -> Tuple[str, str, int, bytes]:
1743 """Similar to :meth:`get_object_header`, but returns object data as well.
1745 :return:
1746 (hexsha, type_string, size_as_int, data_string)
1748 :note:
1749 Not threadsafe.
1750 """
1751 hexsha, typename, size, stream = self.stream_object_data(ref)
1752 data = stream.read(size)
1753 del stream
1754 return (hexsha, typename, size, data)
1756 def stream_object_data(self, ref: str) -> Tuple[str, str, int, "Git.CatFileContentStream"]:
1757 """Similar to :meth:`get_object_data`, but returns the data as a stream.
1759 :return:
1760 (hexsha, type_string, size_as_int, stream)
1762 :note:
1763 This method is not threadsafe. You need one independent :class:`Git`
1764 instance per thread to be safe!
1765 """
1766 cmd = self._get_persistent_cmd("cat_file_all", "cat_file", batch=True)
1767 hexsha, typename, size = self.__get_object_header(cmd, ref)
1768 cmd_stdout = cmd.stdout if cmd.stdout is not None else io.BytesIO()
1769 return (hexsha, typename, size, self.CatFileContentStream(size, cmd_stdout))
1771 def clear_cache(self) -> "Git":
1772 """Clear all kinds of internal caches to release resources.
1774 Currently persistent commands will be interrupted.
1776 :return:
1777 self
1778 """
1779 for cmd in (self.cat_file_all, self.cat_file_header):
1780 if cmd:
1781 cmd.__del__()
1783 self.cat_file_all = None
1784 self.cat_file_header = None
1785 return self