From b96b63448c898ff43308ab31136bf7e24a04654d Mon Sep 17 00:00:00 2001 From: ilmich Date: Mon, 29 Jan 2024 20:12:20 +0100 Subject: [PATCH] Update request urllib3 (#464) * requests: updated to version 2.27.1 * urllib3: updated to version 1.26.18 --- lib/requests/__init__.py | 77 +++-- lib/requests/__version__.py | 8 +- lib/requests/adapters.py | 9 +- lib/requests/api.py | 9 +- lib/requests/auth.py | 4 +- lib/requests/compat.py | 15 +- lib/requests/exceptions.py | 23 +- lib/requests/help.py | 20 +- lib/requests/models.py | 60 ++-- lib/requests/packages.py | 14 +- lib/requests/sessions.py | 75 ++-- lib/requests/status_codes.py | 15 +- lib/requests/structures.py | 4 +- lib/requests/utils.py | 117 ++++++- lib/urllib3/__init__.py | 36 +- lib/urllib3/_collections.py | 27 +- lib/urllib3/_version.py | 2 + lib/urllib3/connection.py | 322 ++++++++++++------ lib/urllib3/connectionpool.py | 302 ++++++++++------ lib/urllib3/contrib/_appengine_environ.py | 28 +- .../contrib/_securetransport/bindings.py | 65 ++-- .../contrib/_securetransport/low_level.py | 75 +++- lib/urllib3/contrib/appengine.py | 21 +- lib/urllib3/contrib/ntlmpool.py | 17 +- lib/urllib3/contrib/pyopenssl.py | 60 ++-- lib/urllib3/contrib/securetransport.py | 104 ++++-- lib/urllib3/contrib/socks.py | 22 +- lib/urllib3/exceptions.py | 146 +++++--- lib/urllib3/fields.py | 5 +- lib/urllib3/filepost.py | 4 +- lib/urllib3/packages/__init__.py | 5 - lib/urllib3/packages/backports/makefile.py | 1 - .../packages/backports/weakref_finalize.py | 155 +++++++++ lib/urllib3/packages/six.py | 125 +++++-- .../packages/ssl_match_hostname/__init__.py | 19 -- lib/urllib3/poolmanager.py | 116 +++++-- lib/urllib3/request.py | 32 +- lib/urllib3/response.py | 154 ++++++--- lib/urllib3/util/__init__.py | 17 +- lib/urllib3/util/connection.py | 21 +- lib/urllib3/util/proxy.py | 57 ++++ lib/urllib3/util/queue.py | 1 + lib/urllib3/util/request.py | 19 +- lib/urllib3/util/response.py | 31 +- lib/urllib3/util/retry.py | 246 ++++++++++--- lib/urllib3/util/ssl_.py | 172 +++++++--- .../ssl_match_hostname.py} | 21 +- lib/urllib3/util/ssltransport.py | 221 ++++++++++++ lib/urllib3/util/timeout.py | 63 ++-- lib/urllib3/util/url.py | 46 ++- lib/urllib3/util/wait.py | 7 +- 51 files changed, 2414 insertions(+), 801 deletions(-) create mode 100644 lib/urllib3/_version.py create mode 100644 lib/urllib3/packages/backports/weakref_finalize.py delete mode 100644 lib/urllib3/packages/ssl_match_hostname/__init__.py create mode 100644 lib/urllib3/util/proxy.py rename lib/urllib3/{packages/ssl_match_hostname/_implementation.py => util/ssl_match_hostname.py} (89%) create mode 100644 lib/urllib3/util/ssltransport.py diff --git a/lib/requests/__init__.py b/lib/requests/__init__.py index 9a899df6..53a5b42a 100644 --- a/lib/requests/__init__.py +++ b/lib/requests/__init__.py @@ -9,14 +9,14 @@ Requests HTTP Library ~~~~~~~~~~~~~~~~~~~~~ -Requests is an HTTP library, written in Python, for human beings. Basic GET -usage: +Requests is an HTTP library, written in Python, for human beings. +Basic GET usage: >>> import requests >>> r = requests.get('https://www.python.org') >>> r.status_code 200 - >>> 'Python is a programming language' in r.content + >>> b'Python is a programming language' in r.content True ... or POST: @@ -27,26 +27,34 @@ usage: { ... "form": { - "key2": "value2", - "key1": "value1" + "key1": "value1", + "key2": "value2" }, ... } The other HTTP methods are supported - see `requests.api`. Full documentation -is at . +is at . :copyright: (c) 2017 by Kenneth Reitz. :license: Apache 2.0, see LICENSE for more details. """ import urllib3 -import chardet import warnings from .exceptions import RequestsDependencyWarning +try: + from charset_normalizer import __version__ as charset_normalizer_version +except ImportError: + charset_normalizer_version = None -def check_compatibility(urllib3_version, chardet_version): +try: + from chardet import __version__ as chardet_version +except ImportError: + chardet_version = None + +def check_compatibility(urllib3_version, chardet_version, charset_normalizer_version): urllib3_version = urllib3_version.split('.') assert urllib3_version != ['dev'] # Verify urllib3 isn't installed from git. @@ -57,19 +65,24 @@ def check_compatibility(urllib3_version, chardet_version): # Check urllib3 for compatibility. major, minor, patch = urllib3_version # noqa: F811 major, minor, patch = int(major), int(minor), int(patch) - # urllib3 >= 1.21.1, <= 1.25 + # urllib3 >= 1.21.1, <= 1.26 assert major == 1 assert minor >= 21 - assert minor <= 25 - - # Check chardet for compatibility. - major, minor, patch = chardet_version.split('.')[:3] - major, minor, patch = int(major), int(minor), int(patch) - # chardet >= 3.0.2, < 3.1.0 - assert major == 3 - assert minor < 1 - assert patch >= 2 + assert minor <= 26 + # Check charset_normalizer for compatibility. + if chardet_version: + major, minor, patch = chardet_version.split('.')[:3] + major, minor, patch = int(major), int(minor), int(patch) + # chardet_version >= 3.0.2, < 5.0.0 + assert (3, 0, 2) <= (major, minor, patch) < (5, 0, 0) + elif charset_normalizer_version: + major, minor, patch = charset_normalizer_version.split('.')[:3] + major, minor, patch = int(major), int(minor), int(patch) + # charset_normalizer >= 2.0.0 < 3.0.0 + assert (2, 0, 0) <= (major, minor, patch) < (3, 0, 0) + else: + raise Exception("You need either charset_normalizer or chardet installed") def _check_cryptography(cryptography_version): # cryptography < 1.3.4 @@ -84,20 +97,28 @@ def _check_cryptography(cryptography_version): # Check imported dependencies for compatibility. try: - check_compatibility(urllib3.__version__, chardet.__version__) + check_compatibility(urllib3.__version__, chardet_version, charset_normalizer_version) except (AssertionError, ValueError): - warnings.warn("urllib3 ({}) or chardet ({}) doesn't match a supported " - "version!".format(urllib3.__version__, chardet.__version__), + warnings.warn("urllib3 ({}) or chardet ({})/charset_normalizer ({}) doesn't match a supported " + "version!".format(urllib3.__version__, chardet_version, charset_normalizer_version), RequestsDependencyWarning) -# Attempt to enable urllib3's SNI support, if possible +# Attempt to enable urllib3's fallback for SNI support +# if the standard library doesn't support SNI or the +# 'ssl' library isn't available. try: - from urllib3.contrib import pyopenssl - pyopenssl.inject_into_urllib3() + try: + import ssl + except ImportError: + ssl = None - # Check cryptography version - from cryptography import __version__ as cryptography_version - _check_cryptography(cryptography_version) + if not getattr(ssl, "HAS_SNI", False): + from urllib3.contrib import pyopenssl + pyopenssl.inject_into_urllib3() + + # Check cryptography version + from cryptography import __version__ as cryptography_version + _check_cryptography(cryptography_version) except ImportError: pass @@ -118,7 +139,7 @@ from .status_codes import codes from .exceptions import ( RequestException, Timeout, URLRequired, TooManyRedirects, HTTPError, ConnectionError, - FileModeWarning, ConnectTimeout, ReadTimeout + FileModeWarning, ConnectTimeout, ReadTimeout, JSONDecodeError ) # Set default logging handler to avoid "No handler found" warnings. diff --git a/lib/requests/__version__.py b/lib/requests/__version__.py index 9844f740..e973b03b 100644 --- a/lib/requests/__version__.py +++ b/lib/requests/__version__.py @@ -4,11 +4,11 @@ __title__ = 'requests' __description__ = 'Python HTTP for Humans.' -__url__ = 'http://python-requests.org' -__version__ = '2.22.0' -__build__ = 0x022200 +__url__ = 'https://requests.readthedocs.io' +__version__ = '2.27.1' +__build__ = 0x022701 __author__ = 'Kenneth Reitz' __author_email__ = 'me@kennethreitz.org' __license__ = 'Apache 2.0' -__copyright__ = 'Copyright 2019 Kenneth Reitz' +__copyright__ = 'Copyright 2022 Kenneth Reitz' __cake__ = u'\u2728 \U0001f370 \u2728' diff --git a/lib/requests/adapters.py b/lib/requests/adapters.py index fa4d9b3c..fe22ff45 100644 --- a/lib/requests/adapters.py +++ b/lib/requests/adapters.py @@ -19,6 +19,7 @@ from urllib3.util.retry import Retry from urllib3.exceptions import ClosedPoolError from urllib3.exceptions import ConnectTimeoutError from urllib3.exceptions import HTTPError as _HTTPError +from urllib3.exceptions import InvalidHeader as _InvalidHeader from urllib3.exceptions import MaxRetryError from urllib3.exceptions import NewConnectionError from urllib3.exceptions import ProxyError as _ProxyError @@ -37,7 +38,7 @@ from .structures import CaseInsensitiveDict from .cookies import extract_cookies_to_jar from .exceptions import (ConnectionError, ConnectTimeout, ReadTimeout, SSLError, ProxyError, RetryError, InvalidSchema, InvalidProxyURL, - InvalidURL) + InvalidURL, InvalidHeader) from .auth import _basic_auth_str try: @@ -457,9 +458,11 @@ class HTTPAdapter(BaseAdapter): low_conn = conn._get_conn(timeout=DEFAULT_POOL_TIMEOUT) try: + skip_host = 'Host' in request.headers low_conn.putrequest(request.method, url, - skip_accept_encoding=True) + skip_accept_encoding=True, + skip_host=skip_host) for header, value in request.headers.items(): low_conn.putheader(header, value) @@ -527,6 +530,8 @@ class HTTPAdapter(BaseAdapter): raise SSLError(e, request=request) elif isinstance(e, ReadTimeoutError): raise ReadTimeout(e, request=request) + elif isinstance(e, _InvalidHeader): + raise InvalidHeader(e, request=request) else: raise diff --git a/lib/requests/api.py b/lib/requests/api.py index ef71d075..4cba90ee 100644 --- a/lib/requests/api.py +++ b/lib/requests/api.py @@ -16,7 +16,7 @@ from . import sessions def request(method, url, **kwargs): """Constructs and sends a :class:`Request `. - :param method: method for the new :class:`Request` object. + :param method: method for the new :class:`Request` object: ``GET``, ``OPTIONS``, ``HEAD``, ``POST``, ``PUT``, ``PATCH``, or ``DELETE``. :param url: URL for the new :class:`Request` object. :param params: (optional) Dictionary, list of tuples or bytes to send in the query string for the :class:`Request`. @@ -50,6 +50,7 @@ def request(method, url, **kwargs): >>> import requests >>> req = requests.request('GET', 'https://httpbin.org/get') + >>> req """ @@ -71,7 +72,6 @@ def get(url, params=None, **kwargs): :rtype: requests.Response """ - kwargs.setdefault('allow_redirects', True) return request('get', url, params=params, **kwargs) @@ -84,7 +84,6 @@ def options(url, **kwargs): :rtype: requests.Response """ - kwargs.setdefault('allow_redirects', True) return request('options', url, **kwargs) @@ -92,7 +91,9 @@ def head(url, **kwargs): r"""Sends a HEAD request. :param url: URL for the new :class:`Request` object. - :param \*\*kwargs: Optional arguments that ``request`` takes. + :param \*\*kwargs: Optional arguments that ``request`` takes. If + `allow_redirects` is not provided, it will be set to `False` (as + opposed to the default :meth:`request` behavior). :return: :class:`Response ` object :rtype: requests.Response """ diff --git a/lib/requests/auth.py b/lib/requests/auth.py index bdde51c7..eeface39 100644 --- a/lib/requests/auth.py +++ b/lib/requests/auth.py @@ -50,7 +50,7 @@ def _basic_auth_str(username, password): "Non-string passwords will no longer be supported in Requests " "3.0.0. Please convert the object you've passed in ({!r}) to " "a string or bytes object in the near future to avoid " - "problems.".format(password), + "problems.".format(type(password)), category=DeprecationWarning, ) password = str(password) @@ -239,7 +239,7 @@ class HTTPDigestAuth(AuthBase): """ # If response is not 4xx, do not auth - # See https://github.com/requests/requests/issues/3772 + # See https://github.com/psf/requests/issues/3772 if not 400 <= r.status_code < 500: self._thread_local.num_401_calls = 1 return r diff --git a/lib/requests/compat.py b/lib/requests/compat.py index c44b35ef..029ae62a 100644 --- a/lib/requests/compat.py +++ b/lib/requests/compat.py @@ -8,7 +8,10 @@ This module handles import compatibility issues between Python 2 and Python 3. """ -import chardet +try: + import chardet +except ImportError: + import charset_normalizer as chardet import sys @@ -25,8 +28,10 @@ is_py2 = (_ver[0] == 2) #: Python 3.x? is_py3 = (_ver[0] == 3) +has_simplejson = False try: import simplejson as json + has_simplejson = True except ImportError: import json @@ -43,15 +48,16 @@ if is_py2: import cookielib from Cookie import Morsel from StringIO import StringIO + # Keep OrderedDict for backwards compatibility. from collections import Callable, Mapping, MutableMapping, OrderedDict - builtin_str = str bytes = str str = unicode basestring = basestring numeric_types = (int, long, float) integer_types = (int, long) + JSONDecodeError = ValueError elif is_py3: from urllib.parse import urlparse, urlunparse, urljoin, urlsplit, urlencode, quote, unquote, quote_plus, unquote_plus, urldefrag @@ -59,8 +65,13 @@ elif is_py3: from http import cookiejar as cookielib from http.cookies import Morsel from io import StringIO + # Keep OrderedDict for backwards compatibility. from collections import OrderedDict from collections.abc import Callable, Mapping, MutableMapping + if has_simplejson: + from simplejson import JSONDecodeError + else: + from json import JSONDecodeError builtin_str = str str = str diff --git a/lib/requests/exceptions.py b/lib/requests/exceptions.py index a80cad80..79697635 100644 --- a/lib/requests/exceptions.py +++ b/lib/requests/exceptions.py @@ -8,6 +8,8 @@ This module contains the set of Requests' exceptions. """ from urllib3.exceptions import HTTPError as BaseHTTPError +from .compat import JSONDecodeError as CompatJSONDecodeError + class RequestException(IOError): """There was an ambiguous exception that occurred while handling your @@ -25,6 +27,14 @@ class RequestException(IOError): super(RequestException, self).__init__(*args, **kwargs) +class InvalidJSONError(RequestException): + """A JSON error occurred.""" + + +class JSONDecodeError(InvalidJSONError, CompatJSONDecodeError): + """Couldn't decode the text into json""" + + class HTTPError(RequestException): """An HTTP error occurred.""" @@ -70,11 +80,11 @@ class TooManyRedirects(RequestException): class MissingSchema(RequestException, ValueError): - """The URL schema (e.g. http or https) is missing.""" + """The URL scheme (e.g. http or https) is missing.""" class InvalidSchema(RequestException, ValueError): - """See defaults.py for valid schemas.""" + """The URL scheme provided is either invalid or unsupported.""" class InvalidURL(RequestException, ValueError): @@ -94,11 +104,11 @@ class ChunkedEncodingError(RequestException): class ContentDecodingError(RequestException, BaseHTTPError): - """Failed to decode response content""" + """Failed to decode response content.""" class StreamConsumedError(RequestException, TypeError): - """The content for this response was already consumed""" + """The content for this response was already consumed.""" class RetryError(RequestException): @@ -106,21 +116,18 @@ class RetryError(RequestException): class UnrewindableBodyError(RequestException): - """Requests encountered an error when trying to rewind a body""" + """Requests encountered an error when trying to rewind a body.""" # Warnings class RequestsWarning(Warning): """Base warning for Requests.""" - pass class FileModeWarning(RequestsWarning, DeprecationWarning): """A file was opened in text mode, but Requests determined its binary length.""" - pass class RequestsDependencyWarning(RequestsWarning): """An imported dependency doesn't match the expected version range.""" - pass diff --git a/lib/requests/help.py b/lib/requests/help.py index e53d35ef..4cd6389f 100644 --- a/lib/requests/help.py +++ b/lib/requests/help.py @@ -8,10 +8,19 @@ import ssl import idna import urllib3 -import chardet from . import __version__ as requests_version +try: + import charset_normalizer +except ImportError: + charset_normalizer = None + +try: + import chardet +except ImportError: + chardet = None + try: from urllib3.contrib import pyopenssl except ImportError: @@ -71,7 +80,12 @@ def info(): implementation_info = _implementation() urllib3_info = {'version': urllib3.__version__} - chardet_info = {'version': chardet.__version__} + charset_normalizer_info = {'version': None} + chardet_info = {'version': None} + if charset_normalizer: + charset_normalizer_info = {'version': charset_normalizer.__version__} + if chardet: + chardet_info = {'version': chardet.__version__} pyopenssl_info = { 'version': None, @@ -99,9 +113,11 @@ def info(): 'implementation': implementation_info, 'system_ssl': system_ssl_info, 'using_pyopenssl': pyopenssl is not None, + 'using_charset_normalizer': chardet is None, 'pyOpenSSL': pyopenssl_info, 'urllib3': urllib3_info, 'chardet': chardet_info, + 'charset_normalizer': charset_normalizer_info, 'cryptography': cryptography_info, 'idna': idna_info, 'requests': { diff --git a/lib/requests/models.py b/lib/requests/models.py index 62dcd0b7..dfbea854 100644 --- a/lib/requests/models.py +++ b/lib/requests/models.py @@ -12,7 +12,7 @@ import sys # Import encoding now, to avoid implicit import later. # Implicit import within threads may cause LookupError when standard library is in a ZIP, -# such as in Embedded Python. See https://github.com/requests/requests/issues/3578. +# such as in Embedded Python. See https://github.com/psf/requests/issues/3578. import encodings.idna from urllib3.fields import RequestField @@ -29,7 +29,9 @@ from .auth import HTTPBasicAuth from .cookies import cookiejar_from_dict, get_cookie_header, _copy_cookie_jar from .exceptions import ( HTTPError, MissingSchema, InvalidURL, ChunkedEncodingError, - ContentDecodingError, ConnectionError, StreamConsumedError) + ContentDecodingError, ConnectionError, StreamConsumedError, + InvalidJSONError) +from .exceptions import JSONDecodeError as RequestsJSONDecodeError from ._internal_utils import to_native_string, unicode_is_ascii from .utils import ( guess_filename, get_auth_from_url, requote_uri, @@ -38,7 +40,7 @@ from .utils import ( from .compat import ( Callable, Mapping, cookielib, urlunparse, urlsplit, urlencode, str, bytes, - is_py2, chardet, builtin_str, basestring) + is_py2, chardet, builtin_str, basestring, JSONDecodeError) from .compat import json as complexjson from .status_codes import codes @@ -273,13 +275,16 @@ class PreparedRequest(RequestEncodingMixin, RequestHooksMixin): """The fully mutable :class:`PreparedRequest ` object, containing the exact bytes that will be sent to the server. - Generated from either a :class:`Request ` object or manually. + Instances are generated from a :class:`Request ` object, and + should not be instantiated manually; doing so may produce undesirable + effects. Usage:: >>> import requests >>> req = requests.Request('GET', 'https://httpbin.org/get') >>> r = req.prepare() + >>> r >>> s = requests.Session() @@ -358,7 +363,7 @@ class PreparedRequest(RequestEncodingMixin, RequestHooksMixin): #: We're unable to blindly call unicode/str functions #: as this will include the bytestring indicator (b'') #: on python 3.x. - #: https://github.com/requests/requests/pull/2238 + #: https://github.com/psf/requests/pull/2238 if isinstance(url, bytes): url = url.decode('utf8') else: @@ -381,7 +386,7 @@ class PreparedRequest(RequestEncodingMixin, RequestHooksMixin): raise InvalidURL(*e.args) if not scheme: - error = ("Invalid URL {0!r}: No schema supplied. Perhaps you meant http://{0}?") + error = ("Invalid URL {0!r}: No scheme supplied. Perhaps you meant http://{0}?") error = error.format(to_native_string(url, 'utf8')) raise MissingSchema(error) @@ -398,7 +403,7 @@ class PreparedRequest(RequestEncodingMixin, RequestHooksMixin): host = self._get_idna_encoded_host(host) except UnicodeError: raise InvalidURL('URL has an invalid label.') - elif host.startswith(u'*'): + elif host.startswith((u'*', u'.')): raise InvalidURL('URL has an invalid label.') # Carefully reconstruct the network location @@ -463,7 +468,12 @@ class PreparedRequest(RequestEncodingMixin, RequestHooksMixin): # urllib3 requires a bytes-like body. Python 2's json.dumps # provides this natively, but Python 3 gives a Unicode string. content_type = 'application/json' - body = complexjson.dumps(json) + + try: + body = complexjson.dumps(json, allow_nan=False) + except ValueError as ve: + raise InvalidJSONError(ve, request=self) + if not isinstance(body, bytes): body = body.encode('utf-8') @@ -472,12 +482,12 @@ class PreparedRequest(RequestEncodingMixin, RequestHooksMixin): not isinstance(data, (basestring, list, tuple, Mapping)) ]) - try: - length = super_len(data) - except (TypeError, AttributeError, UnsupportedOperation): - length = None - if is_stream: + try: + length = super_len(data) + except (TypeError, AttributeError, UnsupportedOperation): + length = None + body = data if getattr(body, 'tell', None) is not None: @@ -608,7 +618,7 @@ class Response(object): #: File-like object representation of response (for advanced usage). #: Use of ``raw`` requires that ``stream=True`` be set on the request. - # This requirement does not apply for use internally to Requests. + #: This requirement does not apply for use internally to Requests. self.raw = None #: Final URL location of Response. @@ -723,7 +733,7 @@ class Response(object): @property def apparent_encoding(self): - """The apparent encoding, provided by the chardet library.""" + """The apparent encoding, provided by the charset_normalizer or chardet libraries.""" return chardet.detect(self.content)['encoding'] def iter_content(self, chunk_size=1, decode_unicode=False): @@ -837,7 +847,7 @@ class Response(object): """Content of the response, in unicode. If Response.encoding is None, encoding will be guessed using - ``chardet``. + ``charset_normalizer`` or ``chardet``. The encoding of the response content is determined based solely on HTTP headers, following RFC 2616 to the letter. If you can take advantage of @@ -874,13 +884,14 @@ class Response(object): r"""Returns the json-encoded content of a response, if any. :param \*\*kwargs: Optional arguments that ``json.loads`` takes. - :raises ValueError: If the response body does not contain valid json. + :raises requests.exceptions.JSONDecodeError: If the response body does not + contain valid json. """ if not self.encoding and self.content and len(self.content) > 3: # No encoding set. JSON RFC 4627 section 3 states we should expect # UTF-8, -16 or -32. Detect which one to use; If the detection or - # decoding fails, fall back to `self.text` (using chardet to make + # decoding fails, fall back to `self.text` (using charset_normalizer to make # a best guess). encoding = guess_json_utf(self.content) if encoding is not None: @@ -894,7 +905,16 @@ class Response(object): # and the server didn't bother to tell us what codec *was* # used. pass - return complexjson.loads(self.text, **kwargs) + + try: + return complexjson.loads(self.text, **kwargs) + except JSONDecodeError as e: + # Catch JSON-related errors and raise as requests.JSONDecodeError + # This aliases json.JSONDecodeError and simplejson.JSONDecodeError + if is_py2: # e is a ValueError + raise RequestsJSONDecodeError(e.message) + else: + raise RequestsJSONDecodeError(e.msg, e.doc, e.pos) @property def links(self): @@ -915,7 +935,7 @@ class Response(object): return l def raise_for_status(self): - """Raises stored :class:`HTTPError`, if one occurred.""" + """Raises :class:`HTTPError`, if one occurred.""" http_error_msg = '' if isinstance(self.reason, bytes): diff --git a/lib/requests/packages.py b/lib/requests/packages.py index 7232fe0f..00196bff 100644 --- a/lib/requests/packages.py +++ b/lib/requests/packages.py @@ -1,9 +1,17 @@ import sys +try: + import chardet +except ImportError: + import charset_normalizer as chardet + import warnings + + warnings.filterwarnings('ignore', 'Trying to detect', module='charset_normalizer') + # This code exists for backwards compatibility reasons. # I don't like it either. Just look the other way. :) -for package in ('urllib3', 'idna', 'chardet'): +for package in ('urllib3', 'idna'): locals()[package] = __import__(package) # This traversal is apparently necessary such that the identities are # preserved (requests.packages.urllib3.* is urllib3.*) @@ -11,4 +19,8 @@ for package in ('urllib3', 'idna', 'chardet'): if mod == package or mod.startswith(package + '.'): sys.modules['requests.packages.' + mod] = sys.modules[mod] +target = chardet.__name__ +for mod in list(sys.modules): + if mod == target or mod.startswith(target + '.'): + sys.modules['requests.packages.' + target.replace(target, 'chardet')] = sys.modules[mod] # Kinda cool, though, right? diff --git a/lib/requests/sessions.py b/lib/requests/sessions.py index d73d700f..3f59cab9 100644 --- a/lib/requests/sessions.py +++ b/lib/requests/sessions.py @@ -1,8 +1,8 @@ # -*- coding: utf-8 -*- """ -requests.session -~~~~~~~~~~~~~~~~ +requests.sessions +~~~~~~~~~~~~~~~~~ This module provides a Session object to manage and persist settings across requests (cookies, auth, proxies). @@ -11,9 +11,10 @@ import os import sys import time from datetime import timedelta +from collections import OrderedDict from .auth import _basic_auth_str -from .compat import cookielib, is_py3, OrderedDict, urljoin, urlparse, Mapping +from .compat import cookielib, is_py3, urljoin, urlparse, Mapping from .cookies import ( cookiejar_from_dict, extract_cookies_to_jar, RequestsCookieJar, merge_cookies) from .models import Request, PreparedRequest, DEFAULT_REDIRECT_LIMIT @@ -28,7 +29,7 @@ from .adapters import HTTPAdapter from .utils import ( requote_uri, get_environ_proxies, get_netrc_auth, should_bypass_proxies, - get_auth_from_url, rewind_body + get_auth_from_url, rewind_body, resolve_proxies ) from .status_codes import codes @@ -162,7 +163,7 @@ class SessionRedirectMixin(object): resp.raw.read(decode_content=False) if len(resp.history) >= self.max_redirects: - raise TooManyRedirects('Exceeded %s redirects.' % self.max_redirects, response=resp) + raise TooManyRedirects('Exceeded {} redirects.'.format(self.max_redirects), response=resp) # Release the connection back into the pool. resp.close() @@ -170,7 +171,7 @@ class SessionRedirectMixin(object): # Handle redirection without scheme (see: RFC 1808 Section 4) if url.startswith('//'): parsed_rurl = urlparse(resp.url) - url = '%s:%s' % (to_native_string(parsed_rurl.scheme), url) + url = ':'.join([to_native_string(parsed_rurl.scheme), url]) # Normalize url case and attach previous fragment if needed (RFC 7231 7.1.2) parsed = urlparse(url) @@ -192,19 +193,16 @@ class SessionRedirectMixin(object): self.rebuild_method(prepared_request, resp) - # https://github.com/requests/requests/issues/1084 + # https://github.com/psf/requests/issues/1084 if resp.status_code not in (codes.temporary_redirect, codes.permanent_redirect): - # https://github.com/requests/requests/issues/3490 + # https://github.com/psf/requests/issues/3490 purged_headers = ('Content-Length', 'Content-Type', 'Transfer-Encoding') for header in purged_headers: prepared_request.headers.pop(header, None) prepared_request.body = None headers = prepared_request.headers - try: - del headers['Cookie'] - except KeyError: - pass + headers.pop('Cookie', None) # Extract any cookies sent on the response to the cookiejar # in the new request. Because we've mutated our copied prepared @@ -271,8 +269,6 @@ class SessionRedirectMixin(object): if new_auth is not None: prepared_request.prepare_auth(new_auth) - return - def rebuild_proxies(self, prepared_request, proxies): """This method re-evaluates the proxy configuration by considering the environment variables. If we are redirected to a URL covered by @@ -285,21 +281,9 @@ class SessionRedirectMixin(object): :rtype: dict """ - proxies = proxies if proxies is not None else {} headers = prepared_request.headers - url = prepared_request.url - scheme = urlparse(url).scheme - new_proxies = proxies.copy() - no_proxy = proxies.get('no_proxy') - - bypass_proxy = should_bypass_proxies(url, no_proxy=no_proxy) - if self.trust_env and not bypass_proxy: - environ_proxies = get_environ_proxies(url, no_proxy=no_proxy) - - proxy = environ_proxies.get(scheme, environ_proxies.get('all')) - - if proxy: - new_proxies.setdefault(scheme, proxy) + scheme = urlparse(prepared_request.url).scheme + new_proxies = resolve_proxies(prepared_request, proxies, self.trust_env) if 'Proxy-Authorization' in headers: del headers['Proxy-Authorization'] @@ -352,13 +336,13 @@ class Session(SessionRedirectMixin): Or as a context manager:: >>> with requests.Session() as s: - >>> s.get('https://httpbin.org/get') + ... s.get('https://httpbin.org/get') """ __attrs__ = [ 'headers', 'cookies', 'auth', 'proxies', 'hooks', 'params', 'verify', - 'cert', 'prefetch', 'adapters', 'stream', 'trust_env', + 'cert', 'adapters', 'stream', 'trust_env', 'max_redirects', ] @@ -390,6 +374,13 @@ class Session(SessionRedirectMixin): self.stream = False #: SSL Verification default. + #: Defaults to `True`, requiring requests to verify the TLS certificate at the + #: remote end. + #: If verify is set to `False`, requests will accept any TLS certificate + #: presented by the server, and will ignore hostname mismatches and/or + #: expired certificates, which will make your application vulnerable to + #: man-in-the-middle (MitM) attacks. + #: Only set this to `False` for testing. self.verify = True #: SSL client certificate default, if String, path to ssl client @@ -498,7 +489,12 @@ class Session(SessionRedirectMixin): content. Defaults to ``False``. :param verify: (optional) Either a boolean, in which case it controls whether we verify the server's TLS certificate, or a string, in which case it must be a path - to a CA bundle to use. Defaults to ``True``. + to a CA bundle to use. Defaults to ``True``. When set to + ``False``, requests will accept any TLS certificate presented by + the server, and will ignore hostname mismatches and/or expired + certificates, which will make your application vulnerable to + man-in-the-middle (MitM) attacks. Setting verify to ``False`` + may be useful during local development or testing. :param cert: (optional) if String, path to ssl client cert file (.pem). If Tuple, ('cert', 'key') pair. :rtype: requests.Response @@ -624,7 +620,10 @@ class Session(SessionRedirectMixin): kwargs.setdefault('stream', self.stream) kwargs.setdefault('verify', self.verify) kwargs.setdefault('cert', self.cert) - kwargs.setdefault('proxies', self.proxies) + if 'proxies' not in kwargs: + kwargs['proxies'] = resolve_proxies( + request, self.proxies, self.trust_env + ) # It's possible that users might accidentally send a Request object. # Guard against that specific failure case. @@ -661,11 +660,13 @@ class Session(SessionRedirectMixin): extract_cookies_to_jar(self.cookies, request, r.raw) - # Redirect resolving generator. - gen = self.resolve_redirects(r, request, **kwargs) - # Resolve redirects if allowed. - history = [resp for resp in gen] if allow_redirects else [] + if allow_redirects: + # Redirect resolving generator. + gen = self.resolve_redirects(r, request, **kwargs) + history = [resp for resp in gen] + else: + history = [] # Shuffle things around if there's history. if history: @@ -728,7 +729,7 @@ class Session(SessionRedirectMixin): return adapter # Nothing matches :-/ - raise InvalidSchema("No connection adapters were found for '%s'" % url) + raise InvalidSchema("No connection adapters were found for {!r}".format(url)) def close(self): """Closes all adapters and as such the session""" diff --git a/lib/requests/status_codes.py b/lib/requests/status_codes.py index 813e8c4e..d80a7cd4 100644 --- a/lib/requests/status_codes.py +++ b/lib/requests/status_codes.py @@ -5,12 +5,15 @@ The ``codes`` object defines a mapping from common names for HTTP statuses to their numerical codes, accessible either as attributes or as dictionary items. ->>> requests.codes['temporary_redirect'] -307 ->>> requests.codes.teapot -418 ->>> requests.codes['\o/'] -200 +Example:: + + >>> import requests + >>> requests.codes['temporary_redirect'] + 307 + >>> requests.codes.teapot + 418 + >>> requests.codes['\o/'] + 200 Some codes have multiple names, and both upper- and lower-case versions of the names are allowed. For example, ``codes.ok``, ``codes.OK``, and diff --git a/lib/requests/structures.py b/lib/requests/structures.py index da930e28..8ee0ba7a 100644 --- a/lib/requests/structures.py +++ b/lib/requests/structures.py @@ -7,7 +7,9 @@ requests.structures Data structures that power Requests. """ -from .compat import OrderedDict, Mapping, MutableMapping +from collections import OrderedDict + +from .compat import Mapping, MutableMapping class CaseInsensitiveDict(MutableMapping): diff --git a/lib/requests/utils.py b/lib/requests/utils.py index 8170a8d2..153776c7 100644 --- a/lib/requests/utils.py +++ b/lib/requests/utils.py @@ -19,6 +19,9 @@ import sys import tempfile import warnings import zipfile +from collections import OrderedDict +from urllib3.util import make_headers +from urllib3.util import parse_url from .__version__ import __version__ from . import certs @@ -26,7 +29,7 @@ from . import certs from ._internal_utils import to_native_string from .compat import parse_http_list as _parse_list_header from .compat import ( - quote, urlparse, bytes, str, OrderedDict, unquote, getproxies, + quote, urlparse, bytes, str, unquote, getproxies, proxy_bypass, urlunparse, basestring, integer_types, is_py3, proxy_bypass_environment, getproxies_environment, Mapping) from .cookies import cookiejar_from_dict @@ -40,6 +43,11 @@ DEFAULT_CA_BUNDLE_PATH = certs.where() DEFAULT_PORTS = {'http': 80, 'https': 443} +# Ensure that ', ' is used to preserve previous delimiter behavior. +DEFAULT_ACCEPT_ENCODING = ", ".join( + re.split(r",\s*", make_headers(accept_encoding=True)["accept-encoding"]) +) + if sys.platform == 'win32': # provide a proxy_bypass version on Windows without DNS lookups @@ -117,7 +125,10 @@ def super_len(o): elif hasattr(o, 'fileno'): try: fileno = o.fileno() - except io.UnsupportedOperation: + except (io.UnsupportedOperation, AttributeError): + # AttributeError is a surprising exception, seeing as how we've just checked + # that `hasattr(o, 'fileno')`. It happens for objects obtained via + # `Tarfile.extractfile()`, per issue 5229. pass else: total_length = os.fstat(fileno).st_size @@ -147,7 +158,7 @@ def super_len(o): current_position = total_length else: if hasattr(o, 'seek') and total_length is None: - # StringIO and BytesIO have seek but no useable fileno + # StringIO and BytesIO have seek but no usable fileno try: # seek to end of file o.seek(0, 2) @@ -168,18 +179,24 @@ def super_len(o): def get_netrc_auth(url, raise_errors=False): """Returns the Requests tuple auth for a given url from netrc.""" + netrc_file = os.environ.get('NETRC') + if netrc_file is not None: + netrc_locations = (netrc_file,) + else: + netrc_locations = ('~/{}'.format(f) for f in NETRC_FILES) + try: from netrc import netrc, NetrcParseError netrc_path = None - for f in NETRC_FILES: + for f in netrc_locations: try: - loc = os.path.expanduser('~/{}'.format(f)) + loc = os.path.expanduser(f) except KeyError: # os.path.expanduser can fail when $HOME is undefined and # getpwuid fails. See https://bugs.python.org/issue20164 & - # https://github.com/requests/requests/issues/1846 + # https://github.com/psf/requests/issues/1846 return if os.path.exists(loc): @@ -211,7 +228,7 @@ def get_netrc_auth(url, raise_errors=False): if raise_errors: raise - # AppEngine hackiness. + # App Engine hackiness. except (ImportError, AttributeError): pass @@ -238,6 +255,10 @@ def extract_zipped_paths(path): archive, member = os.path.split(path) while archive and not os.path.exists(archive): archive, prefix = os.path.split(archive) + if not prefix: + # If we don't check for an empty prefix after the split (in other words, archive remains unchanged after the split), + # we _can_ end up in an infinite loop on a rare corner case affecting a small number of users + break member = '/'.join([prefix, member]) if not zipfile.is_zipfile(archive): @@ -249,13 +270,28 @@ def extract_zipped_paths(path): # we have a valid zip archive and a valid member of that archive tmp = tempfile.gettempdir() - extracted_path = os.path.join(tmp, *member.split('/')) + extracted_path = os.path.join(tmp, member.split('/')[-1]) if not os.path.exists(extracted_path): - extracted_path = zip_file.extract(member, path=tmp) - + # use read + write to avoid the creating nested folders, we only want the file, avoids mkdir racing condition + with atomic_open(extracted_path) as file_handler: + file_handler.write(zip_file.read(member)) return extracted_path +@contextlib.contextmanager +def atomic_open(filename): + """Write a file to the disk in an atomic fashion""" + replacer = os.rename if sys.version_info[0] == 2 else os.replace + tmp_descriptor, tmp_name = tempfile.mkstemp(dir=os.path.dirname(filename)) + try: + with os.fdopen(tmp_descriptor, 'wb') as tmp_handler: + yield tmp_handler + replacer(tmp_name, filename) + except BaseException: + os.remove(tmp_name) + raise + + def from_key_val_list(value): """Take an object and test to see if it can be represented as a dictionary. Unless it can not be represented as such, return an @@ -266,6 +302,8 @@ def from_key_val_list(value): >>> from_key_val_list([('key', 'val')]) OrderedDict([('key', 'val')]) >>> from_key_val_list('string') + Traceback (most recent call last): + ... ValueError: cannot encode objects that are not 2-tuples >>> from_key_val_list({'key': 'val'}) OrderedDict([('key', 'val')]) @@ -292,7 +330,9 @@ def to_key_val_list(value): >>> to_key_val_list({'key': 'val'}) [('key', 'val')] >>> to_key_val_list('string') - ValueError: cannot encode objects that are not 2-tuples. + Traceback (most recent call last): + ... + ValueError: cannot encode objects that are not 2-tuples :rtype: list """ @@ -492,6 +532,10 @@ def get_encoding_from_headers(headers): if 'text' in content_type: return 'ISO-8859-1' + if 'application/json' in content_type: + # Assume UTF-8 based on RFC 4627: https://www.ietf.org/rfc/rfc4627.txt since the charset was unset + return 'utf-8' + def stream_decode_response_unicode(iterator, r): """Stream decodes a iterator.""" @@ -790,6 +834,33 @@ def select_proxy(url, proxies): return proxy +def resolve_proxies(request, proxies, trust_env=True): + """This method takes proxy information from a request and configuration + input to resolve a mapping of target proxies. This will consider settings + such a NO_PROXY to strip proxy configurations. + + :param request: Request or PreparedRequest + :param proxies: A dictionary of schemes or schemes and hosts to proxy URLs + :param trust_env: Boolean declaring whether to trust environment configs + + :rtype: dict + """ + proxies = proxies if proxies is not None else {} + url = request.url + scheme = urlparse(url).scheme + no_proxy = proxies.get('no_proxy') + new_proxies = proxies.copy() + + if trust_env and not should_bypass_proxies(url, no_proxy=no_proxy): + environ_proxies = get_environ_proxies(url, no_proxy=no_proxy) + + proxy = environ_proxies.get(scheme, environ_proxies.get('all')) + + if proxy: + new_proxies.setdefault(scheme, proxy) + return new_proxies + + def default_user_agent(name="python-requests"): """ Return a string representing the default user agent. @@ -805,7 +876,7 @@ def default_headers(): """ return CaseInsensitiveDict({ 'User-Agent': default_user_agent(), - 'Accept-Encoding': ', '.join(('gzip', 'deflate')), + 'Accept-Encoding': DEFAULT_ACCEPT_ENCODING, 'Accept': '*/*', 'Connection': 'keep-alive', }) @@ -892,15 +963,27 @@ def prepend_scheme_if_needed(url, new_scheme): :rtype: str """ - scheme, netloc, path, params, query, fragment = urlparse(url, new_scheme) + parsed = parse_url(url) + scheme, auth, host, port, path, query, fragment = parsed - # urlparse is a finicky beast, and sometimes decides that there isn't a - # netloc present. Assume that it's being over-cautious, and switch netloc - # and path if urlparse decided there was no netloc. + # A defect in urlparse determines that there isn't a netloc present in some + # urls. We previously assumed parsing was overly cautious, and swapped the + # netloc and path. Due to a lack of tests on the original defect, this is + # maintained with parse_url for backwards compatibility. + netloc = parsed.netloc if not netloc: netloc, path = path, netloc - return urlunparse((scheme, netloc, path, params, query, fragment)) + if auth: + # parse_url doesn't provide the netloc with auth + # so we'll add it ourselves. + netloc = '@'.join([auth, netloc]) + if scheme is None: + scheme = new_scheme + if path is None: + path = '' + + return urlunparse((scheme, netloc, path, '', query, fragment)) def get_auth_from_url(url): diff --git a/lib/urllib3/__init__.py b/lib/urllib3/__init__.py index 8f5a21f3..c6fa3821 100644 --- a/lib/urllib3/__init__.py +++ b/lib/urllib3/__init__.py @@ -1,28 +1,44 @@ """ -urllib3 - Thread-safe connection pooling and re-using. +Python HTTP library with thread-safe connection pooling, file post support, user friendly, and more """ from __future__ import absolute_import -import warnings -from .connectionpool import HTTPConnectionPool, HTTPSConnectionPool, connection_from_url +# Set default logging handler to avoid "No handler found" warnings. +import logging +import warnings +from logging import NullHandler from . import exceptions +from ._version import __version__ +from .connectionpool import HTTPConnectionPool, HTTPSConnectionPool, connection_from_url from .filepost import encode_multipart_formdata from .poolmanager import PoolManager, ProxyManager, proxy_from_url from .response import HTTPResponse from .util.request import make_headers -from .util.url import get_host -from .util.timeout import Timeout from .util.retry import Retry +from .util.timeout import Timeout +from .util.url import get_host - -# Set default logging handler to avoid "No handler found" warnings. -import logging -from logging import NullHandler +# === NOTE TO REPACKAGERS AND VENDORS === +# Please delete this block, this logic is only +# for urllib3 being distributed via PyPI. +# See: https://github.com/urllib3/urllib3/issues/2680 +try: + import urllib3_secure_extra # type: ignore # noqa: F401 +except ImportError: + pass +else: + warnings.warn( + "'urllib3[secure]' extra is deprecated and will be removed " + "in a future release of urllib3 2.x. Read more in this issue: " + "https://github.com/urllib3/urllib3/issues/2680", + category=DeprecationWarning, + stacklevel=2, + ) __author__ = "Andrey Petrov (andrey.petrov@shazow.net)" __license__ = "MIT" -__version__ = "1.25.6" +__version__ = __version__ __all__ = ( "HTTPConnectionPool", diff --git a/lib/urllib3/_collections.py b/lib/urllib3/_collections.py index 019d1511..bceb8451 100644 --- a/lib/urllib3/_collections.py +++ b/lib/urllib3/_collections.py @@ -17,9 +17,10 @@ except ImportError: # Platform-specific: No threads available from collections import OrderedDict -from .exceptions import InvalidHeader -from .packages.six import iterkeys, itervalues, PY3 +from .exceptions import InvalidHeader +from .packages import six +from .packages.six import iterkeys, itervalues __all__ = ["RecentlyUsedContainer", "HTTPHeaderDict"] @@ -174,7 +175,7 @@ class HTTPHeaderDict(MutableMapping): def __ne__(self, other): return not self.__eq__(other) - if not PY3: # Python 2 + if six.PY2: # Python 2 iterkeys = MutableMapping.iterkeys itervalues = MutableMapping.itervalues @@ -190,7 +191,7 @@ class HTTPHeaderDict(MutableMapping): def pop(self, key, default=__marker): """D.pop(k[,d]) -> v, remove specified key and return the corresponding value. - If key is not found, d is returned if given, otherwise KeyError is raised. + If key is not found, d is returned if given, otherwise KeyError is raised. """ # Using the MutableMapping function directly fails due to the private marker. # Using ordinary dict.pop would expose the internal structures. @@ -267,6 +268,24 @@ class HTTPHeaderDict(MutableMapping): else: return vals[1:] + def _prepare_for_method_change(self): + """ + Remove content-specific header fields before changing the request + method to GET or HEAD according to RFC 9110, Section 15.4. + """ + content_specific_headers = [ + "Content-Encoding", + "Content-Language", + "Content-Location", + "Content-Type", + "Content-Length", + "Digest", + "Last-Modified", + ] + for header in content_specific_headers: + self.discard(header) + return self + # Backwards compatibility for httplib getheaders = getlist getallmatchingheaders = getlist diff --git a/lib/urllib3/_version.py b/lib/urllib3/_version.py new file mode 100644 index 00000000..85e725ea --- /dev/null +++ b/lib/urllib3/_version.py @@ -0,0 +1,2 @@ +# This file is protected via CODEOWNERS +__version__ = "1.26.18" diff --git a/lib/urllib3/connection.py b/lib/urllib3/connection.py index 3eeb1af5..54b96b19 100644 --- a/lib/urllib3/connection.py +++ b/lib/urllib3/connection.py @@ -1,13 +1,18 @@ from __future__ import absolute_import + import datetime import logging import os +import re import socket -from socket import error as SocketError, timeout as SocketTimeout import warnings +from socket import error as SocketError +from socket import timeout as SocketTimeout + from .packages import six from .packages.six.moves.http_client import HTTPConnection as _HTTPConnection from .packages.six.moves.http_client import HTTPException # noqa: F401 +from .util.proxy import create_proxy_ssl_context try: # Compiled with SSL? import ssl @@ -29,26 +34,33 @@ except NameError: pass +try: # Python 3: + # Not a no-op, we're adding this to the namespace so it can be imported. + BrokenPipeError = BrokenPipeError +except NameError: # Python 2: + + class BrokenPipeError(Exception): + pass + + +from ._collections import HTTPHeaderDict # noqa (historical, removed in v2) +from ._version import __version__ from .exceptions import ( - NewConnectionError, ConnectTimeoutError, + NewConnectionError, SubjectAltNameWarning, SystemTimeWarning, ) -from .packages.ssl_match_hostname import match_hostname, CertificateError - +from .util import SKIP_HEADER, SKIPPABLE_HEADERS, connection from .util.ssl_ import ( - resolve_cert_reqs, - resolve_ssl_version, assert_fingerprint, create_urllib3_context, + is_ipaddress, + resolve_cert_reqs, + resolve_ssl_version, ssl_wrap_socket, ) - - -from .util import connection - -from ._collections import HTTPHeaderDict +from .util.ssl_match_hostname import CertificateError, match_hostname log = logging.getLogger(__name__) @@ -56,37 +68,35 @@ port_by_scheme = {"http": 80, "https": 443} # When it comes time to update this value as a part of regular maintenance # (ie test_recent_date is failing) update it to ~6 months before the current date. -RECENT_DATE = datetime.date(2019, 1, 1) +RECENT_DATE = datetime.date(2022, 1, 1) - -class DummyConnection(object): - """Used to detect a failed ConnectionCls import.""" - - pass +_CONTAINS_CONTROL_CHAR_RE = re.compile(r"[^-!#$%&'*+.^_`|~0-9a-zA-Z]") class HTTPConnection(_HTTPConnection, object): """ - Based on httplib.HTTPConnection but provides an extra constructor + Based on :class:`http.client.HTTPConnection` but provides an extra constructor backwards-compatibility layer between older and newer Pythons. Additional keyword parameters are used to configure attributes of the connection. Accepted parameters include: - - ``strict``: See the documentation on :class:`urllib3.connectionpool.HTTPConnectionPool` - - ``source_address``: Set the source address for the current connection. - - ``socket_options``: Set specific options on the underlying socket. If not specified, then - defaults are loaded from ``HTTPConnection.default_socket_options`` which includes disabling - Nagle's algorithm (sets TCP_NODELAY to 1) unless the connection is behind a proxy. + - ``strict``: See the documentation on :class:`urllib3.connectionpool.HTTPConnectionPool` + - ``source_address``: Set the source address for the current connection. + - ``socket_options``: Set specific options on the underlying socket. If not specified, then + defaults are loaded from ``HTTPConnection.default_socket_options`` which includes disabling + Nagle's algorithm (sets TCP_NODELAY to 1) unless the connection is behind a proxy. - For example, if you wish to enable TCP Keep Alive in addition to the defaults, - you might pass:: + For example, if you wish to enable TCP Keep Alive in addition to the defaults, + you might pass: - HTTPConnection.default_socket_options + [ - (socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1), - ] + .. code-block:: python - Or you may want to disable the defaults by passing an empty list (e.g., ``[]``). + HTTPConnection.default_socket_options + [ + (socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1), + ] + + Or you may want to disable the defaults by passing an empty list (e.g., ``[]``). """ default_port = port_by_scheme["http"] @@ -98,6 +108,10 @@ class HTTPConnection(_HTTPConnection, object): #: Whether this connection verifies the host's certificate. is_verified = False + #: Whether this proxy connection (if used) verifies the proxy host's + #: certificate. + proxy_is_verified = None + def __init__(self, *args, **kw): if not six.PY2: kw.pop("strict", None) @@ -109,6 +123,10 @@ class HTTPConnection(_HTTPConnection, object): #: provided, we use the default options. self.socket_options = kw.pop("socket_options", self.default_socket_options) + # Proxy options provided by the user. + self.proxy = kw.pop("proxy", None) + self.proxy_config = kw.pop("proxy_config", None) + _HTTPConnection.__init__(self, *args, **kw) @property @@ -141,7 +159,7 @@ class HTTPConnection(_HTTPConnection, object): self._dns_host = value def _new_conn(self): - """ Establish a socket connection and set nodelay settings on it. + """Establish a socket connection and set nodelay settings on it. :return: New socket connection. """ @@ -171,10 +189,13 @@ class HTTPConnection(_HTTPConnection, object): return conn + def _is_using_tunnel(self): + # Google App Engine's httplib does not define _tunnel_host + return getattr(self, "_tunnel_host", None) + def _prepare_conn(self, conn): self.sock = conn - # Google App Engine's httplib does not define _tunnel_host - if getattr(self, "_tunnel_host", None): + if self._is_using_tunnel(): # TODO: Fix tunnel so it doesn't depend on self.sock state. self._tunnel() # Mark this connection as not reusable @@ -184,20 +205,61 @@ class HTTPConnection(_HTTPConnection, object): conn = self._new_conn() self._prepare_conn(conn) + def putrequest(self, method, url, *args, **kwargs): + """ """ + # Empty docstring because the indentation of CPython's implementation + # is broken but we don't want this method in our documentation. + match = _CONTAINS_CONTROL_CHAR_RE.search(method) + if match: + raise ValueError( + "Method cannot contain non-token characters %r (found at least %r)" + % (method, match.group()) + ) + + return _HTTPConnection.putrequest(self, method, url, *args, **kwargs) + + def putheader(self, header, *values): + """ """ + if not any(isinstance(v, str) and v == SKIP_HEADER for v in values): + _HTTPConnection.putheader(self, header, *values) + elif six.ensure_str(header.lower()) not in SKIPPABLE_HEADERS: + raise ValueError( + "urllib3.util.SKIP_HEADER only supports '%s'" + % ("', '".join(map(str.title, sorted(SKIPPABLE_HEADERS))),) + ) + + def request(self, method, url, body=None, headers=None): + # Update the inner socket's timeout value to send the request. + # This only triggers if the connection is re-used. + if getattr(self, "sock", None) is not None: + self.sock.settimeout(self.timeout) + + if headers is None: + headers = {} + else: + # Avoid modifying the headers passed into .request() + headers = headers.copy() + if "user-agent" not in (six.ensure_str(k.lower()) for k in headers): + headers["User-Agent"] = _get_default_user_agent() + super(HTTPConnection, self).request(method, url, body=body, headers=headers) + def request_chunked(self, method, url, body=None, headers=None): """ Alternative to the common request method, which sends the body with chunked encoding and not as one block """ - headers = HTTPHeaderDict(headers if headers is not None else {}) - skip_accept_encoding = "accept-encoding" in headers - skip_host = "host" in headers + headers = headers or {} + header_keys = set([six.ensure_str(k.lower()) for k in headers]) + skip_accept_encoding = "accept-encoding" in header_keys + skip_host = "host" in header_keys self.putrequest( method, url, skip_accept_encoding=skip_accept_encoding, skip_host=skip_host ) + if "user-agent" not in header_keys: + self.putheader("User-Agent", _get_default_user_agent()) for header, value in headers.items(): self.putheader(header, value) - if "transfer-encoding" not in headers: + if "transfer-encoding" not in header_keys: self.putheader("Transfer-Encoding", "chunked") self.endheaders() @@ -211,19 +273,31 @@ class HTTPConnection(_HTTPConnection, object): if not isinstance(chunk, bytes): chunk = chunk.encode("utf8") len_str = hex(len(chunk))[2:] - self.send(len_str.encode("utf-8")) - self.send(b"\r\n") - self.send(chunk) - self.send(b"\r\n") + to_send = bytearray(len_str.encode()) + to_send += b"\r\n" + to_send += chunk + to_send += b"\r\n" + self.send(to_send) # After the if clause, to always have a closed body self.send(b"0\r\n\r\n") class HTTPSConnection(HTTPConnection): + """ + Many of the parameters to this constructor are passed to the underlying SSL + socket by means of :py:func:`urllib3.util.ssl_wrap_socket`. + """ + default_port = port_by_scheme["https"] + cert_reqs = None + ca_certs = None + ca_cert_dir = None + ca_cert_data = None ssl_version = None + assert_fingerprint = None + tls_in_tls_required = False def __init__( self, @@ -251,53 +325,6 @@ class HTTPSConnection(HTTPConnection): # HTTPS requests to go out as HTTP. (See Issue #356) self._protocol = "https" - def connect(self): - conn = self._new_conn() - self._prepare_conn(conn) - - # Wrap socket using verification with the root certs in - # trusted_root_certs - default_ssl_context = False - if self.ssl_context is None: - default_ssl_context = True - self.ssl_context = create_urllib3_context( - ssl_version=resolve_ssl_version(self.ssl_version), - cert_reqs=resolve_cert_reqs(self.cert_reqs), - ) - - # Try to load OS default certs if none are given. - # Works well on Windows (requires Python3.4+) - context = self.ssl_context - if ( - not self.ca_certs - and not self.ca_cert_dir - and default_ssl_context - and hasattr(context, "load_default_certs") - ): - context.load_default_certs() - - self.sock = ssl_wrap_socket( - sock=conn, - keyfile=self.key_file, - certfile=self.cert_file, - key_password=self.key_password, - ssl_context=self.ssl_context, - server_hostname=self.server_hostname, - ) - - -class VerifiedHTTPSConnection(HTTPSConnection): - """ - Based on httplib.HTTPSConnection but wraps the socket with - SSL certification. - """ - - cert_reqs = None - ca_certs = None - ca_cert_dir = None - ssl_version = None - assert_fingerprint = None - def set_cert( self, key_file=None, @@ -308,6 +335,7 @@ class VerifiedHTTPSConnection(HTTPSConnection): assert_hostname=None, assert_fingerprint=None, ca_cert_dir=None, + ca_cert_data=None, ): """ This method should only be called once, before the connection is used. @@ -328,15 +356,19 @@ class VerifiedHTTPSConnection(HTTPSConnection): self.assert_fingerprint = assert_fingerprint self.ca_certs = ca_certs and os.path.expanduser(ca_certs) self.ca_cert_dir = ca_cert_dir and os.path.expanduser(ca_cert_dir) + self.ca_cert_data = ca_cert_data def connect(self): # Add certificate verification - conn = self._new_conn() + self.sock = conn = self._new_conn() hostname = self.host + tls_in_tls = False + + if self._is_using_tunnel(): + if self.tls_in_tls_required: + self.sock = conn = self._connect_tls_proxy(hostname, conn) + tls_in_tls = True - # Google App Engine's httplib does not define _tunnel_host - if getattr(self, "_tunnel_host", None): - self.sock = conn # Calls self._set_hostport(), so self.host is # self._tunnel_host below. self._tunnel() @@ -378,6 +410,7 @@ class VerifiedHTTPSConnection(HTTPSConnection): if ( not self.ca_certs and not self.ca_cert_dir + and not self.ca_cert_data and default_ssl_context and hasattr(context, "load_default_certs") ): @@ -390,10 +423,29 @@ class VerifiedHTTPSConnection(HTTPSConnection): key_password=self.key_password, ca_certs=self.ca_certs, ca_cert_dir=self.ca_cert_dir, + ca_cert_data=self.ca_cert_data, server_hostname=server_hostname, ssl_context=context, + tls_in_tls=tls_in_tls, ) + # If we're using all defaults and the connection + # is TLSv1 or TLSv1.1 we throw a DeprecationWarning + # for the host. + if ( + default_ssl_context + and self.ssl_version is None + and hasattr(self.sock, "version") + and self.sock.version() in {"TLSv1", "TLSv1.1"} + ): + warnings.warn( + "Negotiating TLSv1/TLSv1.1 by default is deprecated " + "and will be disabled in urllib3 v2.0.0. Connecting to " + "'%s' with '%s' can be enabled by explicitly opting-in " + "with 'ssl_version'" % (self.host, self.sock.version()), + DeprecationWarning, + ) + if self.assert_fingerprint: assert_fingerprint( self.sock.getpeercert(binary_form=True), self.assert_fingerprint @@ -412,7 +464,7 @@ class VerifiedHTTPSConnection(HTTPSConnection): ( "Certificate for {0} has no `subjectAltName`, falling back to check for a " "`commonName` for now. This feature is being removed by major browsers and " - "deprecated by RFC 2818. (See https://github.com/shazow/urllib3/issues/497 " + "deprecated by RFC 2818. (See https://github.com/urllib3/urllib3/issues/497 " "for details.)".format(hostname) ), SubjectAltNameWarning, @@ -424,13 +476,76 @@ class VerifiedHTTPSConnection(HTTPSConnection): or self.assert_fingerprint is not None ) + def _connect_tls_proxy(self, hostname, conn): + """ + Establish a TLS connection to the proxy using the provided SSL context. + """ + proxy_config = self.proxy_config + ssl_context = proxy_config.ssl_context + if ssl_context: + # If the user provided a proxy context, we assume CA and client + # certificates have already been set + return ssl_wrap_socket( + sock=conn, + server_hostname=hostname, + ssl_context=ssl_context, + ) + + ssl_context = create_proxy_ssl_context( + self.ssl_version, + self.cert_reqs, + self.ca_certs, + self.ca_cert_dir, + self.ca_cert_data, + ) + + # If no cert was provided, use only the default options for server + # certificate validation + socket = ssl_wrap_socket( + sock=conn, + ca_certs=self.ca_certs, + ca_cert_dir=self.ca_cert_dir, + ca_cert_data=self.ca_cert_data, + server_hostname=hostname, + ssl_context=ssl_context, + ) + + if ssl_context.verify_mode != ssl.CERT_NONE and not getattr( + ssl_context, "check_hostname", False + ): + # While urllib3 attempts to always turn off hostname matching from + # the TLS library, this cannot always be done. So we check whether + # the TLS Library still thinks it's matching hostnames. + cert = socket.getpeercert() + if not cert.get("subjectAltName", ()): + warnings.warn( + ( + "Certificate for {0} has no `subjectAltName`, falling back to check for a " + "`commonName` for now. This feature is being removed by major browsers and " + "deprecated by RFC 2818. (See https://github.com/urllib3/urllib3/issues/497 " + "for details.)".format(hostname) + ), + SubjectAltNameWarning, + ) + _match_hostname(cert, hostname) + + self.proxy_is_verified = ssl_context.verify_mode == ssl.CERT_REQUIRED + return socket + def _match_hostname(cert, asserted_hostname): + # Our upstream implementation of ssl.match_hostname() + # only applies this normalization to IP addresses so it doesn't + # match DNS SANs so we do the same thing! + stripped_hostname = asserted_hostname.strip("u[]") + if is_ipaddress(stripped_hostname): + asserted_hostname = stripped_hostname + try: match_hostname(cert, asserted_hostname) except CertificateError as e: log.warning( - "Certificate did not match expected hostname: %s. " "Certificate: %s", + "Certificate did not match expected hostname: %s. Certificate: %s", asserted_hostname, cert, ) @@ -440,9 +555,18 @@ def _match_hostname(cert, asserted_hostname): raise -if ssl: - # Make a copy for testing. - UnverifiedHTTPSConnection = HTTPSConnection - HTTPSConnection = VerifiedHTTPSConnection -else: - HTTPSConnection = DummyConnection +def _get_default_user_agent(): + return "python-urllib3/%s" % __version__ + + +class DummyConnection(object): + """Used to detect a failed ConnectionCls import.""" + + pass + + +if not ssl: + HTTPSConnection = DummyConnection # noqa: F811 + + +VerifiedHTTPSConnection = HTTPSConnection diff --git a/lib/urllib3/connectionpool.py b/lib/urllib3/connectionpool.py index e73fa57a..5a6adcbd 100644 --- a/lib/urllib3/connectionpool.py +++ b/lib/urllib3/connectionpool.py @@ -1,57 +1,62 @@ from __future__ import absolute_import + import errno import logging +import re +import socket import sys import warnings +from socket import error as SocketError +from socket import timeout as SocketTimeout -from socket import error as SocketError, timeout as SocketTimeout -import socket - - +from ._collections import HTTPHeaderDict +from .connection import ( + BaseSSLError, + BrokenPipeError, + DummyConnection, + HTTPConnection, + HTTPException, + HTTPSConnection, + VerifiedHTTPSConnection, + port_by_scheme, +) from .exceptions import ( ClosedPoolError, - ProtocolError, EmptyPoolError, HeaderParsingError, HostChangedError, + InsecureRequestWarning, LocationValueError, MaxRetryError, + NewConnectionError, + ProtocolError, ProxyError, ReadTimeoutError, SSLError, TimeoutError, - InsecureRequestWarning, - NewConnectionError, ) -from .packages.ssl_match_hostname import CertificateError from .packages import six from .packages.six.moves import queue -from .connection import ( - port_by_scheme, - DummyConnection, - HTTPConnection, - HTTPSConnection, - VerifiedHTTPSConnection, - HTTPException, - BaseSSLError, -) from .request import RequestMethods from .response import HTTPResponse - from .util.connection import is_connection_dropped +from .util.proxy import connection_requires_http_tunnel +from .util.queue import LifoQueue from .util.request import set_file_position from .util.response import assert_header_parsing from .util.retry import Retry +from .util.ssl_match_hostname import CertificateError from .util.timeout import Timeout -from .util.url import ( - get_host, - parse_url, - Url, - _normalize_host as normalize_host, - _encode_target, -) -from .util.queue import LifoQueue +from .util.url import Url, _encode_target +from .util.url import _normalize_host as normalize_host +from .util.url import get_host, parse_url +try: # Platform-specific: Python 3 + import weakref + + weakref_finalize = weakref.finalize +except AttributeError: # Platform-specific: Python 2 + from .packages.backports.weakref_finalize import weakref_finalize xrange = six.moves.xrange @@ -65,6 +70,11 @@ class ConnectionPool(object): """ Base class for all connection pools, such as :class:`.HTTPConnectionPool` and :class:`.HTTPSConnectionPool`. + + .. note:: + ConnectionPool.urlopen() does not normalize or percent-encode target URIs + which is useful if your target server doesn't support percent-encoded + target URIs. """ scheme = None @@ -106,16 +116,16 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): :param host: Host used for this HTTP Connection (e.g. "localhost"), passed into - :class:`httplib.HTTPConnection`. + :class:`http.client.HTTPConnection`. :param port: Port used for this HTTP Connection (None is equivalent to 80), passed - into :class:`httplib.HTTPConnection`. + into :class:`http.client.HTTPConnection`. :param strict: Causes BadStatusLine to be raised if the status line can't be parsed as a valid HTTP/1.0 or 1.1 status line, passed into - :class:`httplib.HTTPConnection`. + :class:`http.client.HTTPConnection`. .. note:: Only works in Python 2. This parameter is ignored in Python 3. @@ -149,11 +159,11 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): :param _proxy: Parsed proxy URL, should not be used directly, instead, see - :class:`urllib3.connectionpool.ProxyManager`" + :class:`urllib3.ProxyManager` :param _proxy_headers: A dictionary with proxy headers, should not be used directly, - instead, see :class:`urllib3.connectionpool.ProxyManager`" + instead, see :class:`urllib3.ProxyManager` :param \\**conn_kw: Additional parameters are used to create fresh :class:`urllib3.connection.HTTPConnection`, @@ -176,6 +186,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): retries=None, _proxy=None, _proxy_headers=None, + _proxy_config=None, **conn_kw ): ConnectionPool.__init__(self, host, port) @@ -197,6 +208,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): self.proxy = _proxy self.proxy_headers = _proxy_headers or {} + self.proxy_config = _proxy_config # Fill the queue up so that doing get() on it will block properly for _ in xrange(maxsize): @@ -213,6 +225,19 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # list. self.conn_kw.setdefault("socket_options", []) + self.conn_kw["proxy"] = self.proxy + self.conn_kw["proxy_config"] = self.proxy_config + + # Do not pass 'self' as callback to 'finalize'. + # Then the 'finalize' would keep an endless living (leak) to self. + # By just passing a reference to the pool allows the garbage collector + # to free self if nobody else has a reference to it. + pool = self.pool + + # Close all the HTTPConnections in the pool before the + # HTTPConnectionPool object is garbage collected. + weakref_finalize(self, _close_pool_connections, pool) + def _new_conn(self): """ Return a fresh :class:`HTTPConnection`. @@ -257,7 +282,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): if self.block: raise EmptyPoolError( self, - "Pool reached maximum size and no more " "connections are allowed.", + "Pool reached maximum size and no more connections are allowed.", ) pass # Oh well, we'll create a new connection then @@ -267,7 +292,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): conn.close() if getattr(conn, "auto_open", 1) == 0: # This is a proxied connection that has been mutated by - # httplib._tunnel() and cannot be reused (since it would + # http.client._tunnel() and cannot be reused (since it would # attempt to bypass the proxy) conn = None @@ -295,8 +320,11 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): pass except queue.Full: # This should never happen if self.block == True - log.warning("Connection pool is full, discarding connection: %s", self.host) - + log.warning( + "Connection pool is full, discarding connection: %s. Connection pool size: %s", + self.host, + self.pool.qsize(), + ) # Connection never got put back into the pool, close it. if conn: conn.close() @@ -312,7 +340,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): pass def _get_timeout(self, timeout): - """ Helper that always returns a :class:`urllib3.util.Timeout` """ + """Helper that always returns a :class:`urllib3.util.Timeout`""" if timeout is _Default: return self.timeout.clone() @@ -369,7 +397,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): timeout_obj = self._get_timeout(timeout) timeout_obj.start_connect() - conn.timeout = timeout_obj.connect_timeout + conn.timeout = Timeout.resolve_default_timeout(timeout_obj.connect_timeout) # Trigger any extra validation we need to do. try: @@ -379,12 +407,30 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): self._raise_timeout(err=e, url=url, timeout_value=conn.timeout) raise - # conn.request() calls httplib.*.request, not the method in + # conn.request() calls http.client.*.request, not the method in # urllib3.request. It also calls makefile (recv) on the socket. - if chunked: - conn.request_chunked(method, url, **httplib_request_kw) - else: - conn.request(method, url, **httplib_request_kw) + try: + if chunked: + conn.request_chunked(method, url, **httplib_request_kw) + else: + conn.request(method, url, **httplib_request_kw) + + # We are swallowing BrokenPipeError (errno.EPIPE) since the server is + # legitimately able to close the connection after sending a valid response. + # With this behaviour, the received response is still readable. + except BrokenPipeError: + # Python 3 + pass + except IOError as e: + # Python 2 and macOS/Linux + # EPIPE and ESHUTDOWN are BrokenPipeError on Python 2, and EPROTOTYPE is needed on macOS + # https://erickt.github.io/blog/2014/11/19/adventures-in-debugging-a-potential-osx-kernel-bug/ + if e.errno not in { + errno.EPIPE, + errno.ESHUTDOWN, + errno.EPROTOTYPE, + }: + raise # Reset the timeout for the recv() on the socket read_timeout = timeout_obj.read_timeout @@ -461,14 +507,8 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # Disable access to the pool old_pool, self.pool = self.pool, None - try: - while True: - conn = old_pool.get(block=False) - if conn: - conn.close() - - except queue.Empty: - pass # Done. + # Close all the HTTPConnections in the pool. + _close_pool_connections(old_pool) def is_same_host(self, url): """ @@ -527,10 +567,12 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): :param method: HTTP request method (such as GET, POST, PUT, etc.) + :param url: + The URL to perform the request on. + :param body: - Data to send in the request body (useful for creating - POST requests, see HTTPConnectionPool.post_url for - more convenience). + Data to send in the request body, either :class:`str`, :class:`bytes`, + an iterable of :class:`str`/:class:`bytes`, or a file-like object. :param headers: Dictionary of custom headers to send, such as User-Agent, @@ -560,7 +602,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): :param assert_same_host: If ``True``, will make sure that the host of the pool requests is - consistent else will raise HostChangedError. When False, you can + consistent else will raise HostChangedError. When ``False``, you can use the pool on an HTTP proxy and request foreign hosts. :param timeout: @@ -597,6 +639,10 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): Additional parameters are passed to :meth:`urllib3.response.HTTPResponse.from_httplib` """ + + parsed_url = parse_url(url) + destination_scheme = parsed_url.scheme + if headers is None: headers = self.headers @@ -614,7 +660,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): if url.startswith("/"): url = six.ensure_str(_encode_target(url)) else: - url = six.ensure_str(parse_url(url).url) + url = six.ensure_str(parsed_url.url) conn = None @@ -626,13 +672,17 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # # See issue #651 [1] for details. # - # [1] + # [1] release_this_conn = release_conn - # Merge the proxy headers. Only do this in HTTP. We have to copy the - # headers dict so we can safely change it without those changes being - # reflected in anyone else's copy. - if self.scheme == "http": + http_tunnel_required = connection_requires_http_tunnel( + self.proxy, self.proxy_config, destination_scheme + ) + + # Merge the proxy headers. Only done when not using HTTP CONNECT. We + # have to copy the headers dict so we can safely change it without those + # changes being reflected in anyone else's copy. + if not http_tunnel_required: headers = headers.copy() headers.update(self.proxy_headers) @@ -658,7 +708,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): is_new_proxy_conn = self.proxy is not None and not getattr( conn, "sock", None ) - if is_new_proxy_conn: + if is_new_proxy_conn and http_tunnel_required: self._prepare_proxy(conn) # Make the request on the httplib connection object. @@ -693,9 +743,11 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # Everything went great! clean_exit = True - except queue.Empty: - # Timed out by queue. - raise EmptyPoolError(self, "No pool connections are available.") + except EmptyPoolError: + # Didn't get a connection from the pool, no need to clean up + clean_exit = True + release_this_conn = False + raise except ( TimeoutError, @@ -709,7 +761,35 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # Discard the connection for these exceptions. It will be # replaced during the next _get_conn() call. clean_exit = False - if isinstance(e, (BaseSSLError, CertificateError)): + + def _is_ssl_error_message_from_http_proxy(ssl_error): + # We're trying to detect the message 'WRONG_VERSION_NUMBER' but + # SSLErrors are kinda all over the place when it comes to the message, + # so we try to cover our bases here! + message = " ".join(re.split("[^a-z]", str(ssl_error).lower())) + return ( + "wrong version number" in message or "unknown protocol" in message + ) + + # Try to detect a common user error with proxies which is to + # set an HTTP proxy to be HTTPS when it should be 'http://' + # (ie {'http': 'http://proxy', 'https': 'https://proxy'}) + # Instead we add a nice error message and point to a URL. + if ( + isinstance(e, BaseSSLError) + and self.proxy + and _is_ssl_error_message_from_http_proxy(e) + and conn.proxy + and conn.proxy.scheme == "https" + ): + e = ProxyError( + "Your proxy appears to only use HTTP and not HTTPS, " + "try changing your proxy URL to be HTTP. See: " + "https://urllib3.readthedocs.io/en/1.26.x/advanced-usage.html" + "#https-proxy-error-http-proxy", + SSLError(e), + ) + elif isinstance(e, (BaseSSLError, CertificateError)): e = SSLError(e) elif isinstance(e, (SocketError, NewConnectionError)) and self.proxy: e = ProxyError("Cannot connect to proxy.", e) @@ -742,10 +822,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): if not conn: # Try again log.warning( - "Retrying (%r) after connection " "broken by '%r': %s", - retries, - err, - url, + "Retrying (%r) after connection broken by '%r': %s", retries, err, url ) return self.urlopen( method, @@ -758,44 +835,30 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): timeout=timeout, pool_timeout=pool_timeout, release_conn=release_conn, + chunked=chunked, body_pos=body_pos, **response_kw ) - def drain_and_release_conn(response): - try: - # discard any remaining response body, the connection will be - # released back to the pool once the entire response is read - response.read() - except ( - TimeoutError, - HTTPException, - SocketError, - ProtocolError, - BaseSSLError, - SSLError, - ): - pass - # Handle redirect? redirect_location = redirect and response.get_redirect_location() if redirect_location: if response.status == 303: + # Change the method according to RFC 9110, Section 15.4.4. method = "GET" + # And lose the body not to transfer anything sensitive. + body = None + headers = HTTPHeaderDict(headers)._prepare_for_method_change() try: retries = retries.increment(method, url, response=response, _pool=self) except MaxRetryError: if retries.raise_on_redirect: - # Drain and release the connection for this response, since - # we're not returning it to be released manually. - drain_and_release_conn(response) + response.drain_conn() raise return response - # drain and return the connection to the pool before recursing - drain_and_release_conn(response) - + response.drain_conn() retries.sleep_for_retry(response) log.debug("Redirecting %s -> %s", url, redirect_location) return self.urlopen( @@ -809,26 +872,23 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): timeout=timeout, pool_timeout=pool_timeout, release_conn=release_conn, + chunked=chunked, body_pos=body_pos, **response_kw ) # Check if we should retry the HTTP response. - has_retry_after = bool(response.getheader("Retry-After")) + has_retry_after = bool(response.headers.get("Retry-After")) if retries.is_retry(method, response.status, has_retry_after): try: retries = retries.increment(method, url, response=response, _pool=self) except MaxRetryError: if retries.raise_on_status: - # Drain and release the connection for this response, since - # we're not returning it to be released manually. - drain_and_release_conn(response) + response.drain_conn() raise return response - # drain and return the connection to the pool before recursing - drain_and_release_conn(response) - + response.drain_conn() retries.sleep(response) log.debug("Retry: %s", url) return self.urlopen( @@ -842,6 +902,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): timeout=timeout, pool_timeout=pool_timeout, release_conn=release_conn, + chunked=chunked, body_pos=body_pos, **response_kw ) @@ -853,11 +914,7 @@ class HTTPSConnectionPool(HTTPConnectionPool): """ Same as :class:`.HTTPConnectionPool`, but HTTPS. - When Python is compiled with the :mod:`ssl` module, then - :class:`.VerifiedHTTPSConnection` is used, which *can* verify certificates, - instead of :class:`.HTTPSConnection`. - - :class:`.VerifiedHTTPSConnection` uses one of ``assert_fingerprint``, + :class:`.HTTPSConnection` uses one of ``assert_fingerprint``, ``assert_hostname`` and ``host`` in this order to verify connections. If ``assert_hostname`` is False, no verification is done. @@ -941,15 +998,22 @@ class HTTPSConnectionPool(HTTPConnectionPool): def _prepare_proxy(self, conn): """ - Establish tunnel connection early, because otherwise httplib - would improperly set Host: header to proxy's IP:port. + Establishes a tunnel connection through HTTP CONNECT. + + Tunnel connection is established early because otherwise httplib would + improperly set Host: header to proxy's IP:port. """ + conn.set_tunnel(self._proxy_host, self.port, self.proxy_headers) + + if self.proxy.scheme == "https": + conn.tls_in_tls_required = True + conn.connect() def _new_conn(self): """ - Return a fresh :class:`httplib.HTTPSConnection`. + Return a fresh :class:`http.client.HTTPSConnection`. """ self.num_connections += 1 log.debug( @@ -961,7 +1025,7 @@ class HTTPSConnectionPool(HTTPConnectionPool): if not self.ConnectionCls or self.ConnectionCls is DummyConnection: raise SSLError( - "Can't connect to HTTPS URL because the SSL " "module is not available." + "Can't connect to HTTPS URL because the SSL module is not available." ) actual_host = self.host @@ -996,9 +1060,20 @@ class HTTPSConnectionPool(HTTPConnectionPool): if not conn.is_verified: warnings.warn( ( - "Unverified HTTPS request is being made. " + "Unverified HTTPS request is being made to host '%s'. " "Adding certificate verification is strongly advised. See: " - "https://urllib3.readthedocs.io/en/latest/advanced-usage.html" + "https://urllib3.readthedocs.io/en/1.26.x/advanced-usage.html" + "#ssl-warnings" % conn.host + ), + InsecureRequestWarning, + ) + + if getattr(conn, "proxy_is_verified", None) is False: + warnings.warn( + ( + "Unverified HTTPS connection done to an HTTPS proxy. " + "Adding certificate verification is strongly advised. See: " + "https://urllib3.readthedocs.io/en/1.26.x/advanced-usage.html" "#ssl-warnings" ), InsecureRequestWarning, @@ -1049,3 +1124,14 @@ def _normalize_host(host, scheme): if host.startswith("[") and host.endswith("]"): host = host[1:-1] return host + + +def _close_pool_connections(pool): + """Drains a queue of connections and closes each one.""" + try: + while True: + conn = pool.get(block=False) + if conn: + conn.close() + except queue.Empty: + pass # Done. diff --git a/lib/urllib3/contrib/_appengine_environ.py b/lib/urllib3/contrib/_appengine_environ.py index c909010b..8765b907 100644 --- a/lib/urllib3/contrib/_appengine_environ.py +++ b/lib/urllib3/contrib/_appengine_environ.py @@ -6,27 +6,31 @@ import os def is_appengine(): - return is_local_appengine() or is_prod_appengine() or is_prod_appengine_mvms() + return is_local_appengine() or is_prod_appengine() def is_appengine_sandbox(): - return is_appengine() and not is_prod_appengine_mvms() + """Reports if the app is running in the first generation sandbox. + + The second generation runtimes are technically still in a sandbox, but it + is much less restrictive, so generally you shouldn't need to check for it. + see https://cloud.google.com/appengine/docs/standard/runtimes + """ + return is_appengine() and os.environ["APPENGINE_RUNTIME"] == "python27" def is_local_appengine(): - return ( - "APPENGINE_RUNTIME" in os.environ - and "Development/" in os.environ["SERVER_SOFTWARE"] - ) + return "APPENGINE_RUNTIME" in os.environ and os.environ.get( + "SERVER_SOFTWARE", "" + ).startswith("Development/") def is_prod_appengine(): - return ( - "APPENGINE_RUNTIME" in os.environ - and "Google App Engine/" in os.environ["SERVER_SOFTWARE"] - and not is_prod_appengine_mvms() - ) + return "APPENGINE_RUNTIME" in os.environ and os.environ.get( + "SERVER_SOFTWARE", "" + ).startswith("Google App Engine/") def is_prod_appengine_mvms(): - return os.environ.get("GAE_VM", False) == "true" + """Deprecated.""" + return False diff --git a/lib/urllib3/contrib/_securetransport/bindings.py b/lib/urllib3/contrib/_securetransport/bindings.py index b46e1e3b..264d564d 100644 --- a/lib/urllib3/contrib/_securetransport/bindings.py +++ b/lib/urllib3/contrib/_securetransport/bindings.py @@ -32,30 +32,26 @@ license and by oscrypto's: from __future__ import absolute_import import platform -from ctypes.util import find_library from ctypes import ( - c_void_p, - c_int32, - c_char_p, - c_size_t, + CDLL, + CFUNCTYPE, + POINTER, + c_bool, c_byte, + c_char_p, + c_int32, + c_long, + c_size_t, c_uint32, c_ulong, - c_long, - c_bool, + c_void_p, ) -from ctypes import CDLL, POINTER, CFUNCTYPE +from ctypes.util import find_library +from ...packages.six import raise_from -security_path = find_library("Security") -if not security_path: - raise ImportError("The library Security could not be found") - - -core_foundation_path = find_library("CoreFoundation") -if not core_foundation_path: - raise ImportError("The library CoreFoundation could not be found") - +if platform.system() != "Darwin": + raise ImportError("Only macOS is supported") version = platform.mac_ver()[0] version_info = tuple(map(int, version.split("."))) @@ -65,8 +61,31 @@ if version_info < (10, 8): % (version_info[0], version_info[1]) ) -Security = CDLL(security_path, use_errno=True) -CoreFoundation = CDLL(core_foundation_path, use_errno=True) + +def load_cdll(name, macos10_16_path): + """Loads a CDLL by name, falling back to known path on 10.16+""" + try: + # Big Sur is technically 11 but we use 10.16 due to the Big Sur + # beta being labeled as 10.16. + if version_info >= (10, 16): + path = macos10_16_path + else: + path = find_library(name) + if not path: + raise OSError # Caught and reraised as 'ImportError' + return CDLL(path, use_errno=True) + except OSError: + raise_from(ImportError("The library %s failed to load" % name), None) + + +Security = load_cdll( + "Security", "/System/Library/Frameworks/Security.framework/Security" +) +CoreFoundation = load_cdll( + "CoreFoundation", + "/System/Library/Frameworks/CoreFoundation.framework/CoreFoundation", +) + Boolean = c_bool CFIndex = c_long @@ -276,6 +295,13 @@ try: Security.SSLSetProtocolVersionMax.argtypes = [SSLContextRef, SSLProtocol] Security.SSLSetProtocolVersionMax.restype = OSStatus + try: + Security.SSLSetALPNProtocols.argtypes = [SSLContextRef, CFArrayRef] + Security.SSLSetALPNProtocols.restype = OSStatus + except AttributeError: + # Supported only in 10.12+ + pass + Security.SecCopyErrorMessageString.argtypes = [OSStatus, c_void_p] Security.SecCopyErrorMessageString.restype = CFStringRef @@ -415,6 +441,7 @@ class SecurityConst(object): kTLSProtocol1 = 4 kTLSProtocol11 = 7 kTLSProtocol12 = 8 + # SecureTransport does not support TLS 1.3 even if there's a constant for it kTLSProtocol13 = 10 kTLSProtocolMaxSupported = 999 diff --git a/lib/urllib3/contrib/_securetransport/low_level.py b/lib/urllib3/contrib/_securetransport/low_level.py index e60168ca..fa0b245d 100644 --- a/lib/urllib3/contrib/_securetransport/low_level.py +++ b/lib/urllib3/contrib/_securetransport/low_level.py @@ -10,13 +10,13 @@ appropriate and useful assistance to the higher-level code. import base64 import ctypes import itertools -import re import os +import re import ssl +import struct import tempfile -from .bindings import Security, CoreFoundation, CFConst - +from .bindings import CFConst, CoreFoundation, Security # This regular expression is used to grab PEM data out of a PEM bundle. _PEM_CERTS_RE = re.compile( @@ -56,6 +56,51 @@ def _cf_dictionary_from_tuples(tuples): ) +def _cfstr(py_bstr): + """ + Given a Python binary data, create a CFString. + The string must be CFReleased by the caller. + """ + c_str = ctypes.c_char_p(py_bstr) + cf_str = CoreFoundation.CFStringCreateWithCString( + CoreFoundation.kCFAllocatorDefault, + c_str, + CFConst.kCFStringEncodingUTF8, + ) + return cf_str + + +def _create_cfstring_array(lst): + """ + Given a list of Python binary data, create an associated CFMutableArray. + The array must be CFReleased by the caller. + + Raises an ssl.SSLError on failure. + """ + cf_arr = None + try: + cf_arr = CoreFoundation.CFArrayCreateMutable( + CoreFoundation.kCFAllocatorDefault, + 0, + ctypes.byref(CoreFoundation.kCFTypeArrayCallBacks), + ) + if not cf_arr: + raise MemoryError("Unable to allocate memory!") + for item in lst: + cf_str = _cfstr(item) + if not cf_str: + raise MemoryError("Unable to allocate memory!") + try: + CoreFoundation.CFArrayAppendValue(cf_arr, cf_str) + finally: + CoreFoundation.CFRelease(cf_str) + except BaseException as e: + if cf_arr: + CoreFoundation.CFRelease(cf_arr) + raise ssl.SSLError("Unable to allocate array: %s" % (e,)) + return cf_arr + + def _cf_string_to_unicode(value): """ Creates a Unicode string from a CFString object. Used entirely for error @@ -143,6 +188,7 @@ def _cert_array_from_pem(pem_bundle): # We only want to do that if an error occurs: otherwise, the caller # should free. CoreFoundation.CFRelease(cert_array) + raise return cert_array @@ -326,3 +372,26 @@ def _load_client_cert_chain(keychain, *paths): finally: for obj in itertools.chain(identities, certificates): CoreFoundation.CFRelease(obj) + + +TLS_PROTOCOL_VERSIONS = { + "SSLv2": (0, 2), + "SSLv3": (3, 0), + "TLSv1": (3, 1), + "TLSv1.1": (3, 2), + "TLSv1.2": (3, 3), +} + + +def _build_tls_unknown_ca_alert(version): + """ + Builds a TLS alert record for an unknown CA. + """ + ver_maj, ver_min = TLS_PROTOCOL_VERSIONS[version] + severity_fatal = 0x02 + description_unknown_ca = 0x30 + msg = struct.pack(">BB", severity_fatal, description_unknown_ca) + msg_len = len(msg) + record_type_alert = 0x15 + record = struct.pack(">BBBH", record_type_alert, ver_maj, ver_min, msg_len) + msg + return record diff --git a/lib/urllib3/contrib/appengine.py b/lib/urllib3/contrib/appengine.py index 01c91409..a5a6d910 100644 --- a/lib/urllib3/contrib/appengine.py +++ b/lib/urllib3/contrib/appengine.py @@ -39,24 +39,24 @@ urllib3 on Google App Engine: """ from __future__ import absolute_import + import io import logging import warnings -from ..packages.six.moves.urllib.parse import urljoin from ..exceptions import ( HTTPError, HTTPWarning, MaxRetryError, ProtocolError, - TimeoutError, SSLError, + TimeoutError, ) - +from ..packages.six.moves.urllib.parse import urljoin from ..request import RequestMethods from ..response import HTTPResponse -from ..util.timeout import Timeout from ..util.retry import Retry +from ..util.timeout import Timeout from . import _appengine_environ try: @@ -90,7 +90,7 @@ class AppEngineManager(RequestMethods): * If you attempt to use this on App Engine Flexible, as full socket support is available. * If a request size is more than 10 megabytes. - * If a response size is more than 32 megabtyes. + * If a response size is more than 32 megabytes. * If you use an unsupported request method such as OPTIONS. Beyond those cases, it will raise normal urllib3 errors. @@ -108,17 +108,10 @@ class AppEngineManager(RequestMethods): "URLFetch is not available in this environment." ) - if is_prod_appengine_mvms(): - raise AppEnginePlatformError( - "Use normal urllib3.PoolManager instead of AppEngineManager" - "on Managed VMs, as using URLFetch is not necessary in " - "this environment." - ) - warnings.warn( "urllib3 is using URLFetch on Google App Engine sandbox instead " "of sockets. To use sockets directly instead of URLFetch see " - "https://urllib3.readthedocs.io/en/latest/reference/urllib3.contrib.html.", + "https://urllib3.readthedocs.io/en/1.26.x/reference/urllib3.contrib.html.", AppEnginePlatformWarning, ) @@ -231,7 +224,7 @@ class AppEngineManager(RequestMethods): ) # Check if we should retry the HTTP response. - has_retry_after = bool(http_response.getheader("Retry-After")) + has_retry_after = bool(http_response.headers.get("Retry-After")) if retries.is_retry(method, http_response.status, has_retry_after): retries = retries.increment(method, url, response=http_response, _pool=self) log.debug("Retry: %s", url) diff --git a/lib/urllib3/contrib/ntlmpool.py b/lib/urllib3/contrib/ntlmpool.py index 9c96be29..47166575 100644 --- a/lib/urllib3/contrib/ntlmpool.py +++ b/lib/urllib3/contrib/ntlmpool.py @@ -5,12 +5,21 @@ Issue #10, see: http://code.google.com/p/urllib3/issues/detail?id=10 """ from __future__ import absolute_import +import warnings from logging import getLogger + from ntlm import ntlm from .. import HTTPSConnectionPool from ..packages.six.moves.http_client import HTTPSConnection +warnings.warn( + "The 'urllib3.contrib.ntlmpool' module is deprecated and will be removed " + "in urllib3 v2.0 release, urllib3 is not able to support it properly due " + "to reasons listed in issue: https://github.com/urllib3/urllib3/issues/2282. " + "If you are a user of this module please comment in the mentioned issue.", + DeprecationWarning, +) log = getLogger(__name__) @@ -60,7 +69,7 @@ class NTLMConnectionPool(HTTPSConnectionPool): log.debug("Request headers: %s", headers) conn.request("GET", self.authurl, None, headers) res = conn.getresponse() - reshdr = dict(res.getheaders()) + reshdr = dict(res.headers) log.debug("Response status: %s %s", res.status, res.reason) log.debug("Response headers: %s", reshdr) log.debug("Response data: %s [...]", res.read(100)) @@ -92,13 +101,11 @@ class NTLMConnectionPool(HTTPSConnectionPool): conn.request("GET", self.authurl, None, headers) res = conn.getresponse() log.debug("Response status: %s %s", res.status, res.reason) - log.debug("Response headers: %s", dict(res.getheaders())) + log.debug("Response headers: %s", dict(res.headers)) log.debug("Response data: %s [...]", res.read()[:100]) if res.status != 200: if res.status == 401: - raise Exception( - "Server rejected request: wrong " "username or password" - ) + raise Exception("Server rejected request: wrong username or password") raise Exception("Wrong server response: %s %s" % (res.status, res.reason)) res.fp = None diff --git a/lib/urllib3/contrib/pyopenssl.py b/lib/urllib3/contrib/pyopenssl.py index 3051ef3a..1ed214b1 100644 --- a/lib/urllib3/contrib/pyopenssl.py +++ b/lib/urllib3/contrib/pyopenssl.py @@ -1,27 +1,31 @@ """ -SSL with SNI_-support for Python 2. Follow these instructions if you would -like to verify SSL certificates in Python 2. Note, the default libraries do +TLS with SNI_-support for Python 2. Follow these instructions if you would +like to verify TLS certificates in Python 2. Note, the default libraries do *not* do certificate checking; you need to do additional work to validate certificates yourself. This needs the following packages installed: -* pyOpenSSL (tested with 16.0.0) -* cryptography (minimum 1.3.4, from pyopenssl) -* idna (minimum 2.0, from cryptography) +* `pyOpenSSL`_ (tested with 16.0.0) +* `cryptography`_ (minimum 1.3.4, from pyopenssl) +* `idna`_ (minimum 2.0, from cryptography) However, pyopenssl depends on cryptography, which depends on idna, so while we use all three directly here we end up having relatively few packages required. You can install them with the following command: - pip install pyopenssl cryptography idna +.. code-block:: bash + + $ python -m pip install pyopenssl cryptography idna To activate certificate checking, call :func:`~urllib3.contrib.pyopenssl.inject_into_urllib3` from your Python code before you begin making HTTP requests. This can be done in a ``sitecustomize`` module, or at any other time before your application begins using ``urllib3``, -like this:: +like this: + +.. code-block:: python try: import urllib3.contrib.pyopenssl @@ -35,18 +39,18 @@ when the required modules are installed. Activating this module also has the positive side effect of disabling SSL/TLS compression in Python 2 (see `CRIME attack`_). -If you want to configure the default list of supported cipher suites, you can -set the ``urllib3.contrib.pyopenssl.DEFAULT_SSL_CIPHER_LIST`` variable. - .. _sni: https://en.wikipedia.org/wiki/Server_Name_Indication .. _crime attack: https://en.wikipedia.org/wiki/CRIME_(security_exploit) +.. _pyopenssl: https://www.pyopenssl.org +.. _cryptography: https://cryptography.io +.. _idna: https://github.com/kjd/idna """ from __future__ import absolute_import +import OpenSSL.crypto import OpenSSL.SSL from cryptography import x509 from cryptography.hazmat.backends.openssl import backend as openssl_backend -from cryptography.hazmat.backends.openssl.x509 import _Certificate try: from cryptography.x509 import UnsupportedExtension @@ -56,8 +60,9 @@ except ImportError: pass -from socket import timeout, error as SocketError from io import BytesIO +from socket import error as SocketError +from socket import timeout try: # Platform-specific: Python 2 from socket import _fileobject @@ -67,11 +72,20 @@ except ImportError: # Platform-specific: Python 3 import logging import ssl -from ..packages import six import sys +import warnings from .. import util +from ..packages import six +from ..util.ssl_ import PROTOCOL_TLS_CLIENT +warnings.warn( + "'urllib3.contrib.pyopenssl' module is deprecated and will be removed " + "in a future release of urllib3 2.x. Read more in this issue: " + "https://github.com/urllib3/urllib3/issues/2680", + category=DeprecationWarning, + stacklevel=2, +) __all__ = ["inject_into_urllib3", "extract_from_urllib3"] @@ -81,6 +95,7 @@ HAS_SNI = True # Map from urllib3 to PyOpenSSL compatible parameter-values. _openssl_versions = { util.PROTOCOL_TLS: OpenSSL.SSL.SSLv23_METHOD, + PROTOCOL_TLS_CLIENT: OpenSSL.SSL.SSLv23_METHOD, ssl.PROTOCOL_TLSv1: OpenSSL.SSL.TLSv1_METHOD, } @@ -213,9 +228,8 @@ def get_subj_alt_name(peer_cert): if hasattr(peer_cert, "to_cryptography"): cert = peer_cert.to_cryptography() else: - # This is technically using private APIs, but should work across all - # relevant versions before PyOpenSSL got a proper API for this. - cert = _Certificate(openssl_backend, peer_cert._x509) + der = OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_ASN1, peer_cert) + cert = x509.load_der_x509_certificate(der, openssl_backend) # We want to find the SAN extension. Ask Cryptography to locate it (it's # faster than looping in Python) @@ -400,7 +414,6 @@ if _fileobject: # Platform-specific: Python 2 self._makefile_refs += 1 return _fileobject(self, mode, bufsize, close=True) - else: # Platform-specific: Python 3 makefile = backport_makefile @@ -450,9 +463,12 @@ class PyOpenSSLContext(object): cafile = cafile.encode("utf-8") if capath is not None: capath = capath.encode("utf-8") - self._ctx.load_verify_locations(cafile, capath) - if cadata is not None: - self._ctx.load_verify_locations(BytesIO(cadata)) + try: + self._ctx.load_verify_locations(cafile, capath) + if cadata is not None: + self._ctx.load_verify_locations(BytesIO(cadata)) + except OpenSSL.SSL.Error as e: + raise ssl.SSLError("unable to load trusted certificates: %r" % e) def load_cert_chain(self, certfile, keyfile=None, password=None): self._ctx.use_certificate_chain_file(certfile) @@ -462,6 +478,10 @@ class PyOpenSSLContext(object): self._ctx.set_passwd_cb(lambda *_: password) self._ctx.use_privatekey_file(keyfile or certfile) + def set_alpn_protocols(self, protocols): + protocols = [six.ensure_binary(p) for p in protocols] + return self._ctx.set_alpn_protos(protocols) + def wrap_socket( self, sock, diff --git a/lib/urllib3/contrib/securetransport.py b/lib/urllib3/contrib/securetransport.py index 24e6b5c4..e311c0c8 100644 --- a/lib/urllib3/contrib/securetransport.py +++ b/lib/urllib3/contrib/securetransport.py @@ -29,6 +29,8 @@ library. An enormous debt is owed to him for blazing this trail for us. For that reason, this code should be considered to be covered both by urllib3's license and by oscrypto's: +.. code-block:: + Copyright (c) 2015-2016 Will Bond Permission is hereby granted, free of charge, to any person obtaining a @@ -58,16 +60,21 @@ import os.path import shutil import socket import ssl +import struct import threading import weakref from .. import util -from ._securetransport.bindings import Security, SecurityConst, CoreFoundation +from ..packages import six +from ..util.ssl_ import PROTOCOL_TLS_CLIENT +from ._securetransport.bindings import CoreFoundation, Security, SecurityConst from ._securetransport.low_level import ( _assert_no_error, + _build_tls_unknown_ca_alert, _cert_array_from_pem, - _temporary_keychain, + _create_cfstring_array, _load_client_cert_chain, + _temporary_keychain, ) try: # Platform-specific: Python 2 @@ -144,13 +151,11 @@ CIPHER_SUITES = [ ] # Basically this is simple: for PROTOCOL_SSLv23 we turn it into a low of -# TLSv1 and a high of TLSv1.3. For everything else, we pin to that version. -# TLSv1 to 1.2 are supported on macOS 10.8+ and TLSv1.3 is macOS 10.13+ +# TLSv1 and a high of TLSv1.2. For everything else, we pin to that version. +# TLSv1 to 1.2 are supported on macOS 10.8+ _protocol_to_min_max = { - util.PROTOCOL_TLS: ( - SecurityConst.kTLSProtocol1, - SecurityConst.kTLSProtocolMaxSupported, - ) + util.PROTOCOL_TLS: (SecurityConst.kTLSProtocol1, SecurityConst.kTLSProtocol12), + PROTOCOL_TLS_CLIENT: (SecurityConst.kTLSProtocol1, SecurityConst.kTLSProtocol12), } if hasattr(ssl, "PROTOCOL_SSLv2"): @@ -377,16 +382,55 @@ class WrappedSocket(object): ) _assert_no_error(result) + def _set_alpn_protocols(self, protocols): + """ + Sets up the ALPN protocols on the context. + """ + if not protocols: + return + protocols_arr = _create_cfstring_array(protocols) + try: + result = Security.SSLSetALPNProtocols(self.context, protocols_arr) + _assert_no_error(result) + finally: + CoreFoundation.CFRelease(protocols_arr) + def _custom_validate(self, verify, trust_bundle): """ Called when we have set custom validation. We do this in two cases: first, when cert validation is entirely disabled; and second, when using a custom trust DB. + Raises an SSLError if the connection is not trusted. """ # If we disabled cert validation, just say: cool. if not verify: return + successes = ( + SecurityConst.kSecTrustResultUnspecified, + SecurityConst.kSecTrustResultProceed, + ) + try: + trust_result = self._evaluate_trust(trust_bundle) + if trust_result in successes: + return + reason = "error code: %d" % (trust_result,) + except Exception as e: + # Do not trust on error + reason = "exception: %r" % (e,) + + # SecureTransport does not send an alert nor shuts down the connection. + rec = _build_tls_unknown_ca_alert(self.version()) + self.socket.sendall(rec) + # close the connection immediately + # l_onoff = 1, activate linger + # l_linger = 0, linger for 0 seoncds + opts = struct.pack("ii", 1, 0) + self.socket.setsockopt(socket.SOL_SOCKET, socket.SO_LINGER, opts) + self.close() + raise ssl.SSLError("certificate verify failed, %s" % reason) + + def _evaluate_trust(self, trust_bundle): # We want data in memory, so load it up. if os.path.isfile(trust_bundle): with open(trust_bundle, "rb") as f: @@ -424,15 +468,7 @@ class WrappedSocket(object): if cert_array is not None: CoreFoundation.CFRelease(cert_array) - # Ok, now we can look at what the result was. - successes = ( - SecurityConst.kSecTrustResultUnspecified, - SecurityConst.kSecTrustResultProceed, - ) - if trust_result.value not in successes: - raise ssl.SSLError( - "certificate verify failed, error code: %d" % trust_result.value - ) + return trust_result.value def handshake( self, @@ -444,6 +480,7 @@ class WrappedSocket(object): client_cert, client_key, client_key_passphrase, + alpn_protocols, ): """ Actually performs the TLS handshake. This is run automatically by @@ -484,19 +521,14 @@ class WrappedSocket(object): # Setup the ciphers. self._set_ciphers() + # Setup the ALPN protocols. + self._set_alpn_protocols(alpn_protocols) + # Set the minimum and maximum TLS versions. result = Security.SSLSetProtocolVersionMin(self.context, min_version) _assert_no_error(result) - # TLS 1.3 isn't necessarily enabled by the OS - # so we have to detect when we error out and try - # setting TLS 1.3 if it's allowed. kTLSProtocolMaxSupported - # was added in macOS 10.13 along with kTLSProtocol13. result = Security.SSLSetProtocolVersionMax(self.context, max_version) - if result != 0 and max_version == SecurityConst.kTLSProtocolMaxSupported: - result = Security.SSLSetProtocolVersionMax( - self.context, SecurityConst.kTLSProtocol12 - ) _assert_no_error(result) # If there's a trust DB, we need to use it. We do that by telling @@ -707,7 +739,7 @@ class WrappedSocket(object): ) _assert_no_error(result) if protocol.value == SecurityConst.kTLSProtocol13: - return "TLSv1.3" + raise ssl.SSLError("SecureTransport does not support TLS 1.3") elif protocol.value == SecurityConst.kTLSProtocol12: return "TLSv1.2" elif protocol.value == SecurityConst.kTLSProtocol11: @@ -737,7 +769,6 @@ if _fileobject: # Platform-specific: Python 2 self._makefile_refs += 1 return _fileobject(self, mode, bufsize, close=True) - else: # Platform-specific: Python 3 def makefile(self, mode="r", buffering=None, *args, **kwargs): @@ -765,6 +796,7 @@ class SecureTransportContext(object): self._client_cert = None self._client_key = None self._client_key_passphrase = None + self._alpn_protocols = None @property def check_hostname(self): @@ -830,6 +862,11 @@ class SecureTransportContext(object): if capath is not None: raise ValueError("SecureTransport does not support cert directories") + # Raise if cafile does not exist. + if cafile is not None: + with open(cafile): + pass + self._trust_bundle = cafile or cadata def load_cert_chain(self, certfile, keyfile=None, password=None): @@ -837,6 +874,18 @@ class SecureTransportContext(object): self._client_key = keyfile self._client_cert_passphrase = password + def set_alpn_protocols(self, protocols): + """ + Sets the ALPN protocols that will later be set on the context. + + Raises a NotImplementedError if ALPN is not supported. + """ + if not hasattr(Security, "SSLSetALPNProtocols"): + raise NotImplementedError( + "SecureTransport supports ALPN only in macOS 10.12+" + ) + self._alpn_protocols = [six.ensure_binary(p) for p in protocols] + def wrap_socket( self, sock, @@ -866,5 +915,6 @@ class SecureTransportContext(object): self._client_cert, self._client_key, self._client_key_passphrase, + self._alpn_protocols, ) return wrapped_socket diff --git a/lib/urllib3/contrib/socks.py b/lib/urllib3/contrib/socks.py index 9e97f7aa..c326e80d 100644 --- a/lib/urllib3/contrib/socks.py +++ b/lib/urllib3/contrib/socks.py @@ -14,22 +14,26 @@ supports the following SOCKS features: - SOCKS5 with local DNS (``proxy_url='socks5://...``) - Usernames and passwords for the SOCKS proxy - .. note:: - It is recommended to use ``socks5h://`` or ``socks4a://`` schemes in - your ``proxy_url`` to ensure that DNS resolution is done from the remote - server instead of client-side when connecting to a domain name. +.. note:: + It is recommended to use ``socks5h://`` or ``socks4a://`` schemes in + your ``proxy_url`` to ensure that DNS resolution is done from the remote + server instead of client-side when connecting to a domain name. SOCKS4 supports IPv4 and domain names with the SOCKS4A extension. SOCKS5 supports IPv4, IPv6, and domain names. When connecting to a SOCKS4 proxy the ``username`` portion of the ``proxy_url`` -will be sent as the ``userid`` section of the SOCKS request:: +will be sent as the ``userid`` section of the SOCKS request: + +.. code-block:: python proxy_url="socks4a://@proxy-host" When connecting to a SOCKS5 proxy the ``username`` and ``password`` portion of the ``proxy_url`` will be sent as the username/password to authenticate -with the proxy:: +with the proxy: + +.. code-block:: python proxy_url="socks5h://:@proxy-host" @@ -40,19 +44,21 @@ try: import socks except ImportError: import warnings + from ..exceptions import DependencyWarning warnings.warn( ( "SOCKS support in urllib3 requires the installation of optional " "dependencies: specifically, PySocks. For more information, see " - "https://urllib3.readthedocs.io/en/latest/contrib.html#socks-proxies" + "https://urllib3.readthedocs.io/en/1.26.x/contrib.html#socks-proxies" ), DependencyWarning, ) raise -from socket import error as SocketError, timeout as SocketTimeout +from socket import error as SocketError +from socket import timeout as SocketTimeout from ..connection import HTTPConnection, HTTPSConnection from ..connectionpool import HTTPConnectionPool, HTTPSConnectionPool diff --git a/lib/urllib3/exceptions.py b/lib/urllib3/exceptions.py index 93d93fba..cba6f3f5 100644 --- a/lib/urllib3/exceptions.py +++ b/lib/urllib3/exceptions.py @@ -1,21 +1,24 @@ from __future__ import absolute_import + from .packages.six.moves.http_client import IncompleteRead as httplib_IncompleteRead # Base Exceptions class HTTPError(Exception): - "Base exception used by this module." + """Base exception used by this module.""" + pass class HTTPWarning(Warning): - "Base warning used by this module." + """Base warning used by this module.""" + pass class PoolError(HTTPError): - "Base exception for errors caused within a pool." + """Base exception for errors caused within a pool.""" def __init__(self, pool, message): self.pool = pool @@ -27,7 +30,7 @@ class PoolError(HTTPError): class RequestError(PoolError): - "Base exception for PoolErrors that have associated URLs." + """Base exception for PoolErrors that have associated URLs.""" def __init__(self, pool, url, message): self.url = url @@ -39,22 +42,28 @@ class RequestError(PoolError): class SSLError(HTTPError): - "Raised when SSL certificate fails in an HTTPS connection." + """Raised when SSL certificate fails in an HTTPS connection.""" + pass class ProxyError(HTTPError): - "Raised when the connection to a proxy fails." - pass + """Raised when the connection to a proxy fails.""" + + def __init__(self, message, error, *args): + super(ProxyError, self).__init__(message, error, *args) + self.original_error = error class DecodeError(HTTPError): - "Raised when automatic decoding based on Content-Type fails." + """Raised when automatic decoding based on Content-Type fails.""" + pass class ProtocolError(HTTPError): - "Raised when something unexpected happens mid-request/response." + """Raised when something unexpected happens mid-request/response.""" + pass @@ -84,7 +93,7 @@ class MaxRetryError(RequestError): class HostChangedError(RequestError): - "Raised when an existing pool gets a request for a foreign host." + """Raised when an existing pool gets a request for a foreign host.""" def __init__(self, pool, url, retries=3): message = "Tried to open a foreign host with url: %s" % url @@ -93,13 +102,13 @@ class HostChangedError(RequestError): class TimeoutStateError(HTTPError): - """ Raised when passing an invalid state to a timeout """ + """Raised when passing an invalid state to a timeout""" pass class TimeoutError(HTTPError): - """ Raised when a socket timeout error occurs. + """Raised when a socket timeout error occurs. Catching this error will catch both :exc:`ReadTimeoutErrors ` and :exc:`ConnectTimeoutErrors `. @@ -109,39 +118,45 @@ class TimeoutError(HTTPError): class ReadTimeoutError(TimeoutError, RequestError): - "Raised when a socket timeout occurs while receiving data from a server" + """Raised when a socket timeout occurs while receiving data from a server""" + pass # This timeout error does not have a URL attached and needs to inherit from the # base HTTPError class ConnectTimeoutError(TimeoutError): - "Raised when a socket timeout occurs while connecting to a server" + """Raised when a socket timeout occurs while connecting to a server""" + pass class NewConnectionError(ConnectTimeoutError, PoolError): - "Raised when we fail to establish a new connection. Usually ECONNREFUSED." + """Raised when we fail to establish a new connection. Usually ECONNREFUSED.""" + pass class EmptyPoolError(PoolError): - "Raised when a pool runs out of connections and no more are allowed." + """Raised when a pool runs out of connections and no more are allowed.""" + pass class ClosedPoolError(PoolError): - "Raised when a request enters a pool after the pool has been closed." + """Raised when a request enters a pool after the pool has been closed.""" + pass class LocationValueError(ValueError, HTTPError): - "Raised when there is something wrong with a given URL input." + """Raised when there is something wrong with a given URL input.""" + pass class LocationParseError(LocationValueError): - "Raised when get_host or similar fails to parse the URL input." + """Raised when get_host or similar fails to parse the URL input.""" def __init__(self, location): message = "Failed to parse: %s" % location @@ -150,39 +165,56 @@ class LocationParseError(LocationValueError): self.location = location +class URLSchemeUnknown(LocationValueError): + """Raised when a URL input has an unsupported scheme.""" + + def __init__(self, scheme): + message = "Not supported URL scheme %s" % scheme + super(URLSchemeUnknown, self).__init__(message) + + self.scheme = scheme + + class ResponseError(HTTPError): - "Used as a container for an error reason supplied in a MaxRetryError." + """Used as a container for an error reason supplied in a MaxRetryError.""" + GENERIC_ERROR = "too many error responses" SPECIFIC_ERROR = "too many {status_code} error responses" class SecurityWarning(HTTPWarning): - "Warned when performing security reducing actions" + """Warned when performing security reducing actions""" + pass class SubjectAltNameWarning(SecurityWarning): - "Warned when connecting to a host with a certificate missing a SAN." + """Warned when connecting to a host with a certificate missing a SAN.""" + pass class InsecureRequestWarning(SecurityWarning): - "Warned when making an unverified HTTPS request." + """Warned when making an unverified HTTPS request.""" + pass class SystemTimeWarning(SecurityWarning): - "Warned when system time is suspected to be wrong" + """Warned when system time is suspected to be wrong""" + pass class InsecurePlatformWarning(SecurityWarning): - "Warned when certain SSL configuration is not available on a platform." + """Warned when certain TLS/SSL configuration is not available on a platform.""" + pass class SNIMissingWarning(HTTPWarning): - "Warned when making a HTTPS request without SNI available." + """Warned when making a HTTPS request without SNI available.""" + pass @@ -196,14 +228,15 @@ class DependencyWarning(HTTPWarning): class ResponseNotChunked(ProtocolError, ValueError): - "Response needs to be chunked in order to read it as chunks." + """Response needs to be chunked in order to read it as chunks.""" + pass class BodyNotHttplibCompatible(HTTPError): """ - Body should be httplib.HTTPResponse like (have an fp attribute which - returns raw chunks) for read_chunked(). + Body should be :class:`http.client.HTTPResponse` like + (have an fp attribute which returns raw chunks) for read_chunked(). """ pass @@ -213,37 +246,71 @@ class IncompleteRead(HTTPError, httplib_IncompleteRead): """ Response length doesn't match expected Content-Length - Subclass of http_client.IncompleteRead to allow int value - for `partial` to avoid creating large objects on streamed - reads. + Subclass of :class:`http.client.IncompleteRead` to allow int value + for ``partial`` to avoid creating large objects on streamed reads. """ def __init__(self, partial, expected): super(IncompleteRead, self).__init__(partial, expected) def __repr__(self): - return "IncompleteRead(%i bytes read, " "%i more expected)" % ( + return "IncompleteRead(%i bytes read, %i more expected)" % ( self.partial, self.expected, ) +class InvalidChunkLength(HTTPError, httplib_IncompleteRead): + """Invalid chunk length in a chunked response.""" + + def __init__(self, response, length): + super(InvalidChunkLength, self).__init__( + response.tell(), response.length_remaining + ) + self.response = response + self.length = length + + def __repr__(self): + return "InvalidChunkLength(got length %r, %i bytes read)" % ( + self.length, + self.partial, + ) + + class InvalidHeader(HTTPError): - "The header provided was somehow invalid." + """The header provided was somehow invalid.""" + pass -class ProxySchemeUnknown(AssertionError, ValueError): - "ProxyManager does not support the supplied scheme" +class ProxySchemeUnknown(AssertionError, URLSchemeUnknown): + """ProxyManager does not support the supplied scheme""" + # TODO(t-8ch): Stop inheriting from AssertionError in v2.0. def __init__(self, scheme): - message = "Not supported proxy scheme %s" % scheme + # 'localhost' is here because our URL parser parses + # localhost:8080 -> scheme=localhost, remove if we fix this. + if scheme == "localhost": + scheme = None + if scheme is None: + message = "Proxy URL had no scheme, should start with http:// or https://" + else: + message = ( + "Proxy URL had unsupported scheme %s, should use http:// or https://" + % scheme + ) super(ProxySchemeUnknown, self).__init__(message) +class ProxySchemeUnsupported(ValueError): + """Fetching HTTPS resources through HTTPS proxies is unsupported""" + + pass + + class HeaderParsingError(HTTPError): - "Raised by assert_header_parsing, but we convert it to a log.warning statement." + """Raised by assert_header_parsing, but we convert it to a log.warning statement.""" def __init__(self, defects, unparsed_data): message = "%s, unparsed data: %r" % (defects or "Unknown", unparsed_data) @@ -251,5 +318,6 @@ class HeaderParsingError(HTTPError): class UnrewindableBodyError(HTTPError): - "urllib3 encountered an error when trying to rewind a body" + """urllib3 encountered an error when trying to rewind a body""" + pass diff --git a/lib/urllib3/fields.py b/lib/urllib3/fields.py index 8715b220..9d630f49 100644 --- a/lib/urllib3/fields.py +++ b/lib/urllib3/fields.py @@ -1,4 +1,5 @@ from __future__ import absolute_import + import email.utils import mimetypes import re @@ -26,7 +27,8 @@ def format_header_param_rfc2231(name, value): strategy defined in RFC 2231. Particularly useful for header parameters which might contain - non-ASCII values, like file names. This follows RFC 2388 Section 4.4. + non-ASCII values, like file names. This follows + `RFC 2388 Section 4.4 `_. :param name: The name of the parameter, a string expected to be ASCII only. @@ -65,7 +67,6 @@ _HTML5_REPLACEMENTS = { u"\u0022": u"%22", # Replace "\" with "\\". u"\u005C": u"\u005C\u005C", - u"\u005C": u"\u005C\u005C", } # All control characters from 0x00 to 0x1F *except* 0x1B. diff --git a/lib/urllib3/filepost.py b/lib/urllib3/filepost.py index b7b00992..36c9252c 100644 --- a/lib/urllib3/filepost.py +++ b/lib/urllib3/filepost.py @@ -1,13 +1,13 @@ from __future__ import absolute_import + import binascii import codecs import os - from io import BytesIO +from .fields import RequestField from .packages import six from .packages.six import b -from .fields import RequestField writer = codecs.lookup("utf-8")[3] diff --git a/lib/urllib3/packages/__init__.py b/lib/urllib3/packages/__init__.py index fce4caa6..e69de29b 100644 --- a/lib/urllib3/packages/__init__.py +++ b/lib/urllib3/packages/__init__.py @@ -1,5 +0,0 @@ -from __future__ import absolute_import - -from . import ssl_match_hostname - -__all__ = ("ssl_match_hostname",) diff --git a/lib/urllib3/packages/backports/makefile.py b/lib/urllib3/packages/backports/makefile.py index a3156a69..b8fb2154 100644 --- a/lib/urllib3/packages/backports/makefile.py +++ b/lib/urllib3/packages/backports/makefile.py @@ -7,7 +7,6 @@ Backports the Python 3 ``socket.makefile`` method for use with anything that wants to create a "fake" socket object. """ import io - from socket import SocketIO diff --git a/lib/urllib3/packages/backports/weakref_finalize.py b/lib/urllib3/packages/backports/weakref_finalize.py new file mode 100644 index 00000000..a2f2966e --- /dev/null +++ b/lib/urllib3/packages/backports/weakref_finalize.py @@ -0,0 +1,155 @@ +# -*- coding: utf-8 -*- +""" +backports.weakref_finalize +~~~~~~~~~~~~~~~~~~ + +Backports the Python 3 ``weakref.finalize`` method. +""" +from __future__ import absolute_import + +import itertools +import sys +from weakref import ref + +__all__ = ["weakref_finalize"] + + +class weakref_finalize(object): + """Class for finalization of weakrefable objects + finalize(obj, func, *args, **kwargs) returns a callable finalizer + object which will be called when obj is garbage collected. The + first time the finalizer is called it evaluates func(*arg, **kwargs) + and returns the result. After this the finalizer is dead, and + calling it just returns None. + When the program exits any remaining finalizers for which the + atexit attribute is true will be run in reverse order of creation. + By default atexit is true. + """ + + # Finalizer objects don't have any state of their own. They are + # just used as keys to lookup _Info objects in the registry. This + # ensures that they cannot be part of a ref-cycle. + + __slots__ = () + _registry = {} + _shutdown = False + _index_iter = itertools.count() + _dirty = False + _registered_with_atexit = False + + class _Info(object): + __slots__ = ("weakref", "func", "args", "kwargs", "atexit", "index") + + def __init__(self, obj, func, *args, **kwargs): + if not self._registered_with_atexit: + # We may register the exit function more than once because + # of a thread race, but that is harmless + import atexit + + atexit.register(self._exitfunc) + weakref_finalize._registered_with_atexit = True + info = self._Info() + info.weakref = ref(obj, self) + info.func = func + info.args = args + info.kwargs = kwargs or None + info.atexit = True + info.index = next(self._index_iter) + self._registry[self] = info + weakref_finalize._dirty = True + + def __call__(self, _=None): + """If alive then mark as dead and return func(*args, **kwargs); + otherwise return None""" + info = self._registry.pop(self, None) + if info and not self._shutdown: + return info.func(*info.args, **(info.kwargs or {})) + + def detach(self): + """If alive then mark as dead and return (obj, func, args, kwargs); + otherwise return None""" + info = self._registry.get(self) + obj = info and info.weakref() + if obj is not None and self._registry.pop(self, None): + return (obj, info.func, info.args, info.kwargs or {}) + + def peek(self): + """If alive then return (obj, func, args, kwargs); + otherwise return None""" + info = self._registry.get(self) + obj = info and info.weakref() + if obj is not None: + return (obj, info.func, info.args, info.kwargs or {}) + + @property + def alive(self): + """Whether finalizer is alive""" + return self in self._registry + + @property + def atexit(self): + """Whether finalizer should be called at exit""" + info = self._registry.get(self) + return bool(info) and info.atexit + + @atexit.setter + def atexit(self, value): + info = self._registry.get(self) + if info: + info.atexit = bool(value) + + def __repr__(self): + info = self._registry.get(self) + obj = info and info.weakref() + if obj is None: + return "<%s object at %#x; dead>" % (type(self).__name__, id(self)) + else: + return "<%s object at %#x; for %r at %#x>" % ( + type(self).__name__, + id(self), + type(obj).__name__, + id(obj), + ) + + @classmethod + def _select_for_exit(cls): + # Return live finalizers marked for exit, oldest first + L = [(f, i) for (f, i) in cls._registry.items() if i.atexit] + L.sort(key=lambda item: item[1].index) + return [f for (f, i) in L] + + @classmethod + def _exitfunc(cls): + # At shutdown invoke finalizers for which atexit is true. + # This is called once all other non-daemonic threads have been + # joined. + reenable_gc = False + try: + if cls._registry: + import gc + + if gc.isenabled(): + reenable_gc = True + gc.disable() + pending = None + while True: + if pending is None or weakref_finalize._dirty: + pending = cls._select_for_exit() + weakref_finalize._dirty = False + if not pending: + break + f = pending.pop() + try: + # gc is disabled, so (assuming no daemonic + # threads) the following is the only line in + # this function which might trigger creation + # of a new finalizer + f() + except Exception: + sys.excepthook(*sys.exc_info()) + assert f not in cls._registry + finally: + # prevent any more finalizers from executing during shutdown + weakref_finalize._shutdown = True + if reenable_gc: + gc.enable() diff --git a/lib/urllib3/packages/six.py b/lib/urllib3/packages/six.py index 31442409..f099a3dc 100644 --- a/lib/urllib3/packages/six.py +++ b/lib/urllib3/packages/six.py @@ -1,4 +1,4 @@ -# Copyright (c) 2010-2019 Benjamin Peterson +# Copyright (c) 2010-2020 Benjamin Peterson # # Permission is hereby granted, free of charge, to any person obtaining a copy # of this software and associated documentation files (the "Software"), to deal @@ -29,7 +29,7 @@ import sys import types __author__ = "Benjamin Peterson " -__version__ = "1.12.0" +__version__ = "1.16.0" # Useful for very coarse version differentiation. @@ -71,6 +71,11 @@ else: MAXSIZE = int((1 << 63) - 1) del X +if PY34: + from importlib.util import spec_from_loader +else: + spec_from_loader = None + def _add_doc(func, doc): """Add documentation to a function.""" @@ -182,6 +187,11 @@ class _SixMetaPathImporter(object): return self return None + def find_spec(self, fullname, path, target=None): + if fullname in self.known_modules: + return spec_from_loader(fullname, self) + return None + def __get_module(self, fullname): try: return self.known_modules[fullname] @@ -220,6 +230,12 @@ class _SixMetaPathImporter(object): get_source = get_code # same as get_code + def create_module(self, spec): + return self.load_module(spec.name) + + def exec_module(self, module): + pass + _importer = _SixMetaPathImporter(__name__) @@ -260,9 +276,19 @@ _moved_attributes = [ ), MovedModule("builtins", "__builtin__"), MovedModule("configparser", "ConfigParser"), + MovedModule( + "collections_abc", + "collections", + "collections.abc" if sys.version_info >= (3, 3) else "collections", + ), MovedModule("copyreg", "copy_reg"), MovedModule("dbm_gnu", "gdbm", "dbm.gnu"), - MovedModule("_dummy_thread", "dummy_thread", "_dummy_thread"), + MovedModule("dbm_ndbm", "dbm", "dbm.ndbm"), + MovedModule( + "_dummy_thread", + "dummy_thread", + "_dummy_thread" if sys.version_info < (3, 9) else "_thread", + ), MovedModule("http_cookiejar", "cookielib", "http.cookiejar"), MovedModule("http_cookies", "Cookie", "http.cookies"), MovedModule("html_entities", "htmlentitydefs", "html.entities"), @@ -307,7 +333,9 @@ _moved_attributes = [ ] # Add windows specific modules. if sys.platform == "win32": - _moved_attributes += [MovedModule("winreg", "_winreg")] + _moved_attributes += [ + MovedModule("winreg", "_winreg"), + ] for attr in _moved_attributes: setattr(_MovedItems, attr.name, attr) @@ -476,7 +504,7 @@ class Module_six_moves_urllib_robotparser(_LazyModule): _urllib_robotparser_moved_attributes = [ - MovedAttribute("RobotFileParser", "robotparser", "urllib.robotparser") + MovedAttribute("RobotFileParser", "robotparser", "urllib.robotparser"), ] for attr in _urllib_robotparser_moved_attributes: setattr(Module_six_moves_urllib_robotparser, attr.name, attr) @@ -678,9 +706,11 @@ if PY3: if sys.version_info[1] <= 1: _assertRaisesRegex = "assertRaisesRegexp" _assertRegex = "assertRegexpMatches" + _assertNotRegex = "assertNotRegexpMatches" else: _assertRaisesRegex = "assertRaisesRegex" _assertRegex = "assertRegex" + _assertNotRegex = "assertNotRegex" else: def b(s): @@ -707,6 +737,7 @@ else: _assertCountEqual = "assertItemsEqual" _assertRaisesRegex = "assertRaisesRegexp" _assertRegex = "assertRegexpMatches" + _assertNotRegex = "assertNotRegexpMatches" _add_doc(b, """Byte literal""") _add_doc(u, """Text literal""") @@ -723,6 +754,10 @@ def assertRegex(self, *args, **kwargs): return getattr(self, _assertRegex)(*args, **kwargs) +def assertNotRegex(self, *args, **kwargs): + return getattr(self, _assertNotRegex)(*args, **kwargs) + + if PY3: exec_ = getattr(moves.builtins, "exec") @@ -737,7 +772,6 @@ if PY3: value = None tb = None - else: def exec_(_code_, _globs_=None, _locs_=None): @@ -750,7 +784,7 @@ else: del frame elif _locs_ is None: _locs_ = _globs_ - exec("""exec _code_ in _globs_, _locs_""") + exec ("""exec _code_ in _globs_, _locs_""") exec_( """def reraise(tp, value, tb=None): @@ -762,18 +796,7 @@ else: ) -if sys.version_info[:2] == (3, 2): - exec_( - """def raise_from(value, from_value): - try: - if from_value is None: - raise value - raise value from from_value - finally: - value = None -""" - ) -elif sys.version_info[:2] > (3, 2): +if sys.version_info[:2] > (3,): exec_( """def raise_from(value, from_value): try: @@ -863,19 +886,41 @@ if sys.version_info[:2] < (3, 3): _add_doc(reraise, """Reraise an exception.""") if sys.version_info[0:2] < (3, 4): + # This does exactly the same what the :func:`py3:functools.update_wrapper` + # function does on Python versions after 3.2. It sets the ``__wrapped__`` + # attribute on ``wrapper`` object and it doesn't raise an error if any of + # the attributes mentioned in ``assigned`` and ``updated`` are missing on + # ``wrapped`` object. + def _update_wrapper( + wrapper, + wrapped, + assigned=functools.WRAPPER_ASSIGNMENTS, + updated=functools.WRAPPER_UPDATES, + ): + for attr in assigned: + try: + value = getattr(wrapped, attr) + except AttributeError: + continue + else: + setattr(wrapper, attr, value) + for attr in updated: + getattr(wrapper, attr).update(getattr(wrapped, attr, {})) + wrapper.__wrapped__ = wrapped + return wrapper + + _update_wrapper.__doc__ = functools.update_wrapper.__doc__ def wraps( wrapped, assigned=functools.WRAPPER_ASSIGNMENTS, updated=functools.WRAPPER_UPDATES, ): - def wrapper(f): - f = functools.wraps(wrapped, assigned, updated)(f) - f.__wrapped__ = wrapped - return f - - return wrapper + return functools.partial( + _update_wrapper, wrapped=wrapped, assigned=assigned, updated=updated + ) + wraps.__doc__ = functools.wraps.__doc__ else: wraps = functools.wraps @@ -888,7 +933,15 @@ def with_metaclass(meta, *bases): # the actual metaclass. class metaclass(type): def __new__(cls, name, this_bases, d): - return meta(name, bases, d) + if sys.version_info[:2] >= (3, 7): + # This version introduced PEP 560 that requires a bit + # of extra care (we mimic what is done by __build_class__). + resolved_bases = types.resolve_bases(bases) + if resolved_bases is not bases: + d["__orig_bases__"] = bases + else: + resolved_bases = bases + return meta(name, resolved_bases, d) @classmethod def __prepare__(cls, name, this_bases): @@ -928,12 +981,11 @@ def ensure_binary(s, encoding="utf-8", errors="strict"): - `str` -> encoded to `bytes` - `bytes` -> `bytes` """ + if isinstance(s, binary_type): + return s if isinstance(s, text_type): return s.encode(encoding, errors) - elif isinstance(s, binary_type): - return s - else: - raise TypeError("not expecting type '%s'" % type(s)) + raise TypeError("not expecting type '%s'" % type(s)) def ensure_str(s, encoding="utf-8", errors="strict"): @@ -947,12 +999,15 @@ def ensure_str(s, encoding="utf-8", errors="strict"): - `str` -> `str` - `bytes` -> decoded to `str` """ - if not isinstance(s, (text_type, binary_type)): - raise TypeError("not expecting type '%s'" % type(s)) + # Optimization: Fast return for the common case. + if type(s) is str: + return s if PY2 and isinstance(s, text_type): - s = s.encode(encoding, errors) + return s.encode(encoding, errors) elif PY3 and isinstance(s, binary_type): - s = s.decode(encoding, errors) + return s.decode(encoding, errors) + elif not isinstance(s, (text_type, binary_type)): + raise TypeError("not expecting type '%s'" % type(s)) return s @@ -977,7 +1032,7 @@ def ensure_text(s, encoding="utf-8", errors="strict"): def python_2_unicode_compatible(klass): """ - A decorator that defines __unicode__ and __str__ methods under Python 2. + A class decorator that defines __unicode__ and __str__ methods under Python 2. Under Python 3 it does nothing. To support Python 2 and 3 with a single code base, define a __str__ method diff --git a/lib/urllib3/packages/ssl_match_hostname/__init__.py b/lib/urllib3/packages/ssl_match_hostname/__init__.py deleted file mode 100644 index 75b6bb1c..00000000 --- a/lib/urllib3/packages/ssl_match_hostname/__init__.py +++ /dev/null @@ -1,19 +0,0 @@ -import sys - -try: - # Our match_hostname function is the same as 3.5's, so we only want to - # import the match_hostname function if it's at least that good. - if sys.version_info < (3, 5): - raise ImportError("Fallback to vendored code") - - from ssl import CertificateError, match_hostname -except ImportError: - try: - # Backport of the function from a pypi module - from backports.ssl_match_hostname import CertificateError, match_hostname - except ImportError: - # Our vendored copy - from ._implementation import CertificateError, match_hostname - -# Not needed, but documenting what we provide. -__all__ = ("CertificateError", "match_hostname") diff --git a/lib/urllib3/poolmanager.py b/lib/urllib3/poolmanager.py index 242a2f82..fb51bf7d 100644 --- a/lib/urllib3/poolmanager.py +++ b/lib/urllib3/poolmanager.py @@ -1,18 +1,24 @@ from __future__ import absolute_import + import collections import functools import logging -from ._collections import RecentlyUsedContainer -from .connectionpool import HTTPConnectionPool, HTTPSConnectionPool -from .connectionpool import port_by_scheme -from .exceptions import LocationValueError, MaxRetryError, ProxySchemeUnknown +from ._collections import HTTPHeaderDict, RecentlyUsedContainer +from .connectionpool import HTTPConnectionPool, HTTPSConnectionPool, port_by_scheme +from .exceptions import ( + LocationValueError, + MaxRetryError, + ProxySchemeUnknown, + ProxySchemeUnsupported, + URLSchemeUnknown, +) from .packages import six from .packages.six.moves.urllib.parse import urljoin from .request import RequestMethods -from .util.url import parse_url +from .util.proxy import connection_requires_http_tunnel from .util.retry import Retry - +from .util.url import parse_url __all__ = ["PoolManager", "ProxyManager", "proxy_from_url"] @@ -28,6 +34,7 @@ SSL_KEYWORDS = ( "ca_cert_dir", "ssl_context", "key_password", + "server_hostname", ) # All known keyword arguments that could be provided to the pool manager, its @@ -53,6 +60,7 @@ _key_fields = ( "key_headers", # dict "key__proxy", # parsed proxy url "key__proxy_headers", # dict + "key__proxy_config", # class "key_socket_options", # list of (level (int), optname (int), value (int or str)) tuples "key__socks_options", # dict "key_assert_hostname", # bool or string @@ -64,6 +72,9 @@ _key_fields = ( #: All custom key schemes should include the fields in this key at a minimum. PoolKey = collections.namedtuple("PoolKey", _key_fields) +_proxy_config_fields = ("ssl_context", "use_forwarding_for_https") +ProxyConfig = collections.namedtuple("ProxyConfig", _proxy_config_fields) + def _default_key_normalizer(key_class, request_context): """ @@ -155,11 +166,12 @@ class PoolManager(RequestMethods): """ proxy = None + proxy_config = None def __init__(self, num_pools=10, headers=None, **connection_pool_kw): RequestMethods.__init__(self, headers) self.connection_pool_kw = connection_pool_kw - self.pools = RecentlyUsedContainer(num_pools, dispose_func=lambda p: p.close()) + self.pools = RecentlyUsedContainer(num_pools) # Locally set the pool classes and keys so other PoolManagers can # override them. @@ -176,7 +188,7 @@ class PoolManager(RequestMethods): def _new_pool(self, scheme, host, port, request_context=None): """ - Create a new :class:`ConnectionPool` based on host, port, scheme, and + Create a new :class:`urllib3.connectionpool.ConnectionPool` based on host, port, scheme, and any additional pool keyword arguments. If ``request_context`` is provided, it is provided as keyword arguments @@ -212,7 +224,7 @@ class PoolManager(RequestMethods): def connection_from_host(self, host, port=None, scheme="http", pool_kwargs=None): """ - Get a :class:`ConnectionPool` based on the host, port, and scheme. + Get a :class:`urllib3.connectionpool.ConnectionPool` based on the host, port, and scheme. If ``port`` isn't given, it will be derived from the ``scheme`` using ``urllib3.connectionpool.port_by_scheme``. If ``pool_kwargs`` is @@ -235,20 +247,22 @@ class PoolManager(RequestMethods): def connection_from_context(self, request_context): """ - Get a :class:`ConnectionPool` based on the request context. + Get a :class:`urllib3.connectionpool.ConnectionPool` based on the request context. ``request_context`` must at least contain the ``scheme`` key and its value must be a key in ``key_fn_by_scheme`` instance variable. """ scheme = request_context["scheme"].lower() - pool_key_constructor = self.key_fn_by_scheme[scheme] + pool_key_constructor = self.key_fn_by_scheme.get(scheme) + if not pool_key_constructor: + raise URLSchemeUnknown(scheme) pool_key = pool_key_constructor(request_context) return self.connection_from_pool_key(pool_key, request_context=request_context) def connection_from_pool_key(self, pool_key, request_context=None): """ - Get a :class:`ConnectionPool` based on the provided pool key. + Get a :class:`urllib3.connectionpool.ConnectionPool` based on the provided pool key. ``pool_key`` should be a namedtuple that only contains immutable objects. At a minimum it must have the ``scheme``, ``host``, and @@ -306,9 +320,39 @@ class PoolManager(RequestMethods): base_pool_kwargs[key] = value return base_pool_kwargs + def _proxy_requires_url_absolute_form(self, parsed_url): + """ + Indicates if the proxy requires the complete destination URL in the + request. Normally this is only needed when not using an HTTP CONNECT + tunnel. + """ + if self.proxy is None: + return False + + return not connection_requires_http_tunnel( + self.proxy, self.proxy_config, parsed_url.scheme + ) + + def _validate_proxy_scheme_url_selection(self, url_scheme): + """ + Validates that were not attempting to do TLS in TLS connections on + Python2 or with unsupported SSL implementations. + """ + if self.proxy is None or url_scheme != "https": + return + + if self.proxy.scheme != "https": + return + + if six.PY2 and not self.proxy_config.use_forwarding_for_https: + raise ProxySchemeUnsupported( + "Contacting HTTPS destinations through HTTPS proxies " + "'via CONNECT tunnels' is not supported in Python 2" + ) + def urlopen(self, method, url, redirect=True, **kw): """ - Same as :meth:`urllib3.connectionpool.HTTPConnectionPool.urlopen` + Same as :meth:`urllib3.HTTPConnectionPool.urlopen` with custom cross-host redirect logic and only sends the request-uri portion of the ``url``. @@ -316,6 +360,8 @@ class PoolManager(RequestMethods): :class:`urllib3.connectionpool.ConnectionPool` can be chosen for it. """ u = parse_url(url) + self._validate_proxy_scheme_url_selection(u.scheme) + conn = self.connection_from_host(u.host, port=u.port, scheme=u.scheme) kw["assert_same_host"] = False @@ -324,7 +370,7 @@ class PoolManager(RequestMethods): if "headers" not in kw: kw["headers"] = self.headers.copy() - if self.proxy is not None and u.scheme == "http": + if self._proxy_requires_url_absolute_form(u): response = conn.urlopen(method, url, **kw) else: response = conn.urlopen(method, u.request_uri, **kw) @@ -336,9 +382,12 @@ class PoolManager(RequestMethods): # Support relative URLs for redirecting. redirect_location = urljoin(url, redirect_location) - # RFC 7231, Section 6.4.4 if response.status == 303: + # Change the method according to RFC 9110, Section 15.4.4. method = "GET" + # And lose the body not to transfer anything sensitive. + kw["body"] = None + kw["headers"] = HTTPHeaderDict(kw["headers"])._prepare_for_method_change() retries = kw.get("retries") if not isinstance(retries, Retry): @@ -359,6 +408,7 @@ class PoolManager(RequestMethods): retries = retries.increment(method, url, response=response, _pool=conn) except MaxRetryError: if retries.raise_on_redirect: + response.drain_conn() raise return response @@ -366,6 +416,8 @@ class PoolManager(RequestMethods): kw["redirect"] = redirect log.info("Redirecting %s -> %s", url, redirect_location) + + response.drain_conn() return self.urlopen(method, redirect_location, **kw) @@ -383,6 +435,19 @@ class ProxyManager(PoolManager): HTTPS/CONNECT case they are sent only once. Could be used for proxy authentication. + :param proxy_ssl_context: + The proxy SSL context is used to establish the TLS connection to the + proxy when using HTTPS proxies. + + :param use_forwarding_for_https: + (Defaults to False) If set to True will forward requests to the HTTPS + proxy to be made on behalf of the client instead of creating a TLS + tunnel via the CONNECT method. **Enabling this flag means that request + and response headers and content will be visible from the HTTPS proxy** + whereas tunneling keeps request and response headers and content + private. IP address, target hostname, SNI, and port are always visible + to an HTTPS proxy even when this flag is disabled. + Example: >>> proxy = urllib3.ProxyManager('http://localhost:3128/') >>> r1 = proxy.request('GET', 'http://google.com/') @@ -402,6 +467,8 @@ class ProxyManager(PoolManager): num_pools=10, headers=None, proxy_headers=None, + proxy_ssl_context=None, + use_forwarding_for_https=False, **connection_pool_kw ): @@ -412,18 +479,22 @@ class ProxyManager(PoolManager): proxy_url.port, ) proxy = parse_url(proxy_url) - if not proxy.port: - port = port_by_scheme.get(proxy.scheme, 80) - proxy = proxy._replace(port=port) if proxy.scheme not in ("http", "https"): raise ProxySchemeUnknown(proxy.scheme) + if not proxy.port: + port = port_by_scheme.get(proxy.scheme, 80) + proxy = proxy._replace(port=port) + self.proxy = proxy self.proxy_headers = proxy_headers or {} + self.proxy_ssl_context = proxy_ssl_context + self.proxy_config = ProxyConfig(proxy_ssl_context, use_forwarding_for_https) connection_pool_kw["_proxy"] = self.proxy connection_pool_kw["_proxy_headers"] = self.proxy_headers + connection_pool_kw["_proxy_config"] = self.proxy_config super(ProxyManager, self).__init__(num_pools, headers, **connection_pool_kw) @@ -455,11 +526,10 @@ class ProxyManager(PoolManager): def urlopen(self, method, url, redirect=True, **kw): "Same as HTTP(S)ConnectionPool.urlopen, ``url`` must be absolute." u = parse_url(url) - - if u.scheme == "http": - # For proxied HTTPS requests, httplib sets the necessary headers - # on the CONNECT to the proxy. For HTTP, we'll definitely - # need to set 'Host' at the very least. + if not connection_requires_http_tunnel(self.proxy, self.proxy_config, u.scheme): + # For connections using HTTP CONNECT, httplib sets the necessary + # headers on the CONNECT to the proxy. If we're not using CONNECT, + # we'll definitely need to set 'Host' at the very least. headers = kw.get("headers", self.headers) kw["headers"] = self._set_proxy_headers(url, headers) diff --git a/lib/urllib3/request.py b/lib/urllib3/request.py index 55f160bb..3b4cf999 100644 --- a/lib/urllib3/request.py +++ b/lib/urllib3/request.py @@ -1,8 +1,10 @@ from __future__ import absolute_import -from .filepost import encode_multipart_formdata -from .packages.six.moves.urllib.parse import urlencode +import sys +from .filepost import encode_multipart_formdata +from .packages import six +from .packages.six.moves.urllib.parse import urlencode __all__ = ["RequestMethods"] @@ -10,8 +12,8 @@ __all__ = ["RequestMethods"] class RequestMethods(object): """ Convenience mixin for classes who implement a :meth:`urlopen` method, such - as :class:`~urllib3.connectionpool.HTTPConnectionPool` and - :class:`~urllib3.poolmanager.PoolManager`. + as :class:`urllib3.HTTPConnectionPool` and + :class:`urllib3.PoolManager`. Provides behavior for making common types of HTTP request methods and decides which type of request field encoding to use. @@ -111,9 +113,9 @@ class RequestMethods(object): the body. This is useful for request methods like POST, PUT, PATCH, etc. When ``encode_multipart=True`` (default), then - :meth:`urllib3.filepost.encode_multipart_formdata` is used to encode + :func:`urllib3.encode_multipart_formdata` is used to encode the payload with the appropriate content type. Otherwise - :meth:`urllib.urlencode` is used with the + :func:`urllib.parse.urlencode` is used with the 'application/x-www-form-urlencoded' content type. Multipart encoding must be used when posting files, and it's reasonably @@ -169,3 +171,21 @@ class RequestMethods(object): extra_kw.update(urlopen_kw) return self.urlopen(method, url, **extra_kw) + + +if not six.PY2: + + class RequestModule(sys.modules[__name__].__class__): + def __call__(self, *args, **kwargs): + """ + If user tries to call this module directly urllib3 v2.x style raise an error to the user + suggesting they may need urllib3 v2 + """ + raise TypeError( + "'module' object is not callable\n" + "urllib3.request() method is not supported in this release, " + "upgrade to urllib3 v2 to use it\n" + "see https://urllib3.readthedocs.io/en/stable/v2-migration-guide.html" + ) + + sys.modules[__name__].__class__ = RequestModule diff --git a/lib/urllib3/response.py b/lib/urllib3/response.py index adc321e7..0bd13d40 100644 --- a/lib/urllib3/response.py +++ b/lib/urllib3/response.py @@ -1,29 +1,38 @@ from __future__ import absolute_import -from contextlib import contextmanager -import zlib + import io import logging -from socket import timeout as SocketTimeout +import sys +import warnings +import zlib +from contextlib import contextmanager from socket import error as SocketError +from socket import timeout as SocketTimeout try: - import brotli + try: + import brotlicffi as brotli + except ImportError: + import brotli except ImportError: brotli = None +from . import util from ._collections import HTTPHeaderDict +from .connection import BaseSSLError, HTTPException from .exceptions import ( BodyNotHttplibCompatible, - ProtocolError, DecodeError, + HTTPError, + IncompleteRead, + InvalidChunkLength, + InvalidHeader, + ProtocolError, ReadTimeoutError, ResponseNotChunked, - IncompleteRead, - InvalidHeader, + SSLError, ) -from .packages.six import string_types as basestring, PY3 -from .packages.six.moves import http_client as httplib -from .connection import HTTPException, BaseSSLError +from .packages import six from .util.response import is_fp_closed, is_response_to_head log = logging.getLogger(__name__) @@ -106,11 +115,10 @@ if brotli is not None: # are for 'brotlipy' and bottom branches for 'Brotli' def __init__(self): self._obj = brotli.Decompressor() - - def decompress(self, data): if hasattr(self._obj, "decompress"): - return self._obj.decompress(data) - return self._obj.process(data) + self.decompress = self._obj.decompress + else: + self.decompress = self._obj.process def flush(self): if hasattr(self._obj, "flush"): @@ -156,13 +164,13 @@ class HTTPResponse(io.IOBase): """ HTTP Response container. - Backwards-compatible to httplib's HTTPResponse but the response ``body`` is + Backwards-compatible with :class:`http.client.HTTPResponse` but the response ``body`` is loaded and decoded on-demand when the ``data`` property is accessed. This class is also compatible with the Python standard library's :mod:`io` module, and can hence be treated as a readable object in the context of that framework. - Extra parameters for behaviour not present in httplib.HTTPResponse: + Extra parameters for behaviour not present in :class:`http.client.HTTPResponse`: :param preload_content: If True, the response's body will be preloaded during construction. @@ -172,7 +180,7 @@ class HTTPResponse(io.IOBase): 'content-encoding' header. :param original_response: - When this HTTPResponse wrapper is generated from an httplib.HTTPResponse + When this HTTPResponse wrapper is generated from an :class:`http.client.HTTPResponse` object, it's convenient to include the original for debug purposes. It's otherwise unused. @@ -232,7 +240,7 @@ class HTTPResponse(io.IOBase): self.msg = msg self._request_url = request_url - if body and isinstance(body, (basestring, bytes)): + if body and isinstance(body, (six.string_types, bytes)): self._body = body self._pool = pool @@ -277,9 +285,20 @@ class HTTPResponse(io.IOBase): self._pool._put_conn(self._connection) self._connection = None + def drain_conn(self): + """ + Read and discard any remaining HTTP response data in the response connection. + + Unread data in the HTTPResponse connection blocks the connection from being released back to the pool. + """ + try: + self.read() + except (HTTPError, SocketError, BaseSSLError, HTTPException): + pass + @property def data(self): - # For backwords-compat with earlier urllib3 0.4 and earlier. + # For backwards-compat with earlier urllib3 0.4 and earlier. if self._body: return self._body @@ -296,8 +315,8 @@ class HTTPResponse(io.IOBase): def tell(self): """ Obtain the number of bytes pulled over the wire so far. May differ from - the amount of content returned by :meth:``HTTPResponse.read`` if bytes - are encoded on the wire (e.g, compressed). + the amount of content returned by :meth:``urllib3.response.HTTPResponse.read`` + if bytes are encoded on the wire (e.g, compressed). """ return self._fp_bytes_read @@ -431,10 +450,9 @@ class HTTPResponse(io.IOBase): except BaseSSLError as e: # FIXME: Is there a better way to differentiate between SSLErrors? - if "read operation timed out" not in str(e): # Defensive: - # This shouldn't happen but just in case we're missing an edge - # case, let's avoid swallowing SSL errors. - raise + if "read operation timed out" not in str(e): + # SSL errors related to framing/MAC get wrapped and reraised here + raise SSLError(e) raise ReadTimeoutError(self._pool, None, "Read timed out.") @@ -466,9 +484,57 @@ class HTTPResponse(io.IOBase): if self._original_response and self._original_response.isclosed(): self.release_conn() + def _fp_read(self, amt): + """ + Read a response with the thought that reading the number of bytes + larger than can fit in a 32-bit int at a time via SSL in some + known cases leads to an overflow error that has to be prevented + if `amt` or `self.length_remaining` indicate that a problem may + happen. + + The known cases: + * 3.8 <= CPython < 3.9.7 because of a bug + https://github.com/urllib3/urllib3/issues/2513#issuecomment-1152559900. + * urllib3 injected with pyOpenSSL-backed SSL-support. + * CPython < 3.10 only when `amt` does not fit 32-bit int. + """ + assert self._fp + c_int_max = 2 ** 31 - 1 + if ( + ( + (amt and amt > c_int_max) + or (self.length_remaining and self.length_remaining > c_int_max) + ) + and not util.IS_SECURETRANSPORT + and (util.IS_PYOPENSSL or sys.version_info < (3, 10)) + ): + buffer = io.BytesIO() + # Besides `max_chunk_amt` being a maximum chunk size, it + # affects memory overhead of reading a response by this + # method in CPython. + # `c_int_max` equal to 2 GiB - 1 byte is the actual maximum + # chunk size that does not lead to an overflow error, but + # 256 MiB is a compromise. + max_chunk_amt = 2 ** 28 + while amt is None or amt != 0: + if amt is not None: + chunk_amt = min(amt, max_chunk_amt) + amt -= chunk_amt + else: + chunk_amt = max_chunk_amt + data = self._fp.read(chunk_amt) + if not data: + break + buffer.write(data) + del data # to reduce peak memory usage by `max_chunk_amt`. + return buffer.getvalue() + else: + # StringIO doesn't like amt=None + return self._fp.read(amt) if amt is not None else self._fp.read() + def read(self, amt=None, decode_content=None, cache_content=False): """ - Similar to :meth:`httplib.HTTPResponse.read`, but with two additional + Similar to :meth:`http.client.HTTPResponse.read`, but with two additional parameters: ``decode_content`` and ``cache_content``. :param amt: @@ -498,13 +564,11 @@ class HTTPResponse(io.IOBase): fp_closed = getattr(self._fp, "closed", False) with self._error_catcher(): + data = self._fp_read(amt) if not fp_closed else b"" if amt is None: - # cStringIO doesn't like amt=None - data = self._fp.read() if not fp_closed else b"" flush_decoder = True else: cache_content = False - data = self._fp.read(amt) if not fp_closed else b"" if ( amt != 0 and not data ): # Platform-specific: Buggy versions of Python. @@ -569,7 +633,7 @@ class HTTPResponse(io.IOBase): @classmethod def from_httplib(ResponseCls, r, **response_kw): """ - Given an :class:`httplib.HTTPResponse` instance ``r``, return a + Given an :class:`http.client.HTTPResponse` instance ``r``, return a corresponding :class:`urllib3.response.HTTPResponse` object. Remaining parameters are passed to the HTTPResponse constructor, along @@ -578,11 +642,11 @@ class HTTPResponse(io.IOBase): headers = r.msg if not isinstance(headers, HTTPHeaderDict): - if PY3: - headers = HTTPHeaderDict(headers.items()) - else: + if six.PY2: # Python 2.7 headers = HTTPHeaderDict.from_httplib(headers) + else: + headers = HTTPHeaderDict(headers.items()) # HTTPResponse objects in Python 3 don't have a .strict attribute strict = getattr(r, "strict", 0) @@ -598,11 +662,23 @@ class HTTPResponse(io.IOBase): ) return resp - # Backwards-compatibility methods for httplib.HTTPResponse + # Backwards-compatibility methods for http.client.HTTPResponse def getheaders(self): + warnings.warn( + "HTTPResponse.getheaders() is deprecated and will be removed " + "in urllib3 v2.1.0. Instead access HTTPResponse.headers directly.", + category=DeprecationWarning, + stacklevel=2, + ) return self.headers def getheader(self, name, default=None): + warnings.warn( + "HTTPResponse.getheader() is deprecated and will be removed " + "in urllib3 v2.1.0. Instead use HTTPResponse.headers.get(name, default).", + category=DeprecationWarning, + stacklevel=2, + ) return self.headers.get(name, default) # Backwards compatibility for http.cookiejar @@ -668,8 +744,8 @@ class HTTPResponse(io.IOBase): def supports_chunked_reads(self): """ Checks if the underlying file-like object looks like a - httplib.HTTPResponse object. We do this by testing for the fp - attribute. If it is present we assume it returns raw chunks as + :class:`http.client.HTTPResponse` object. We do this by testing for + the fp attribute. If it is present we assume it returns raw chunks as processed by read_chunked(). """ return hasattr(self._fp, "fp") @@ -686,7 +762,7 @@ class HTTPResponse(io.IOBase): except ValueError: # Invalid chunked protocol response, abort. self.close() - raise httplib.IncompleteRead(line) + raise InvalidChunkLength(self, line) def _handle_chunk(self, amt): returned_chunk = None @@ -733,7 +809,7 @@ class HTTPResponse(io.IOBase): ) if not self.supports_chunked_reads(): raise BodyNotHttplibCompatible( - "Body should be httplib.HTTPResponse like. " + "Body should be http.client.HTTPResponse like. " "It should have have an fp attribute which returns raw chunks." ) @@ -792,7 +868,7 @@ class HTTPResponse(io.IOBase): return self._request_url def __iter__(self): - buffer = [b""] + buffer = [] for chunk in self.stream(decode_content=True): if b"\n" in chunk: chunk = chunk.split(b"\n") diff --git a/lib/urllib3/util/__init__.py b/lib/urllib3/util/__init__.py index a96c73a9..4547fc52 100644 --- a/lib/urllib3/util/__init__.py +++ b/lib/urllib3/util/__init__.py @@ -2,23 +2,23 @@ from __future__ import absolute_import # For backwards compatibility, provide imports that used to be here. from .connection import is_connection_dropped -from .request import make_headers +from .request import SKIP_HEADER, SKIPPABLE_HEADERS, make_headers from .response import is_fp_closed +from .retry import Retry from .ssl_ import ( - SSLContext, + ALPN_PROTOCOLS, HAS_SNI, IS_PYOPENSSL, IS_SECURETRANSPORT, + PROTOCOL_TLS, + SSLContext, assert_fingerprint, resolve_cert_reqs, resolve_ssl_version, ssl_wrap_socket, - PROTOCOL_TLS, ) -from .timeout import current_time, Timeout - -from .retry import Retry -from .url import get_host, parse_url, split_first, Url +from .timeout import Timeout, current_time +from .url import Url, get_host, parse_url, split_first from .wait import wait_for_read, wait_for_write __all__ = ( @@ -27,6 +27,7 @@ __all__ = ( "IS_SECURETRANSPORT", "SSLContext", "PROTOCOL_TLS", + "ALPN_PROTOCOLS", "Retry", "Timeout", "Url", @@ -43,4 +44,6 @@ __all__ = ( "ssl_wrap_socket", "wait_for_read", "wait_for_write", + "SKIP_HEADER", + "SKIPPABLE_HEADERS", ) diff --git a/lib/urllib3/util/connection.py b/lib/urllib3/util/connection.py index 0e111262..6af1138f 100644 --- a/lib/urllib3/util/connection.py +++ b/lib/urllib3/util/connection.py @@ -1,7 +1,11 @@ from __future__ import absolute_import + import socket -from .wait import NoWayToWaitForSocketError, wait_for_read + from ..contrib import _appengine_environ +from ..exceptions import LocationParseError +from ..packages import six +from .wait import NoWayToWaitForSocketError, wait_for_read def is_connection_dropped(conn): # Platform-specific @@ -9,7 +13,7 @@ def is_connection_dropped(conn): # Platform-specific Returns True if the connection is dropped and should be closed. :param conn: - :class:`httplib.HTTPConnection` object. + :class:`http.client.HTTPConnection` object. Note: For platforms like AppEngine, this will always return ``False`` to let the platform handle connection recycling transparently for us. @@ -42,7 +46,7 @@ def create_connection( port)``) and return the socket object. Passing the optional *timeout* parameter will set the timeout on the socket instance before attempting to connect. If no *timeout* is supplied, the - global default timeout setting returned by :func:`getdefaulttimeout` + global default timeout setting returned by :func:`socket.getdefaulttimeout` is used. If *source_address* is set it must be a tuple of (host, port) for the socket to bind as a source address before making the connection. An host of '' or port 0 tells the OS to use the default. @@ -58,6 +62,13 @@ def create_connection( # The original create_connection function always returns all records. family = allowed_gai_family() + try: + host.encode("idna") + except UnicodeError: + return six.raise_from( + LocationParseError(u"'%s', label empty or too long" % host), None + ) + for res in socket.getaddrinfo(host, port, family, socket.SOCK_STREAM): af, socktype, proto, canonname, sa = res sock = None @@ -106,7 +117,7 @@ def allowed_gai_family(): def _has_ipv6(host): - """ Returns True if the system can bind an IPv6 address. """ + """Returns True if the system can bind an IPv6 address.""" sock = None has_ipv6 = False @@ -121,7 +132,7 @@ def _has_ipv6(host): # has_ipv6 returns true if cPython was compiled with IPv6 support. # It does not tell us if the system has IPv6 support enabled. To # determine that we must bind to an IPv6 address. - # https://github.com/shazow/urllib3/pull/611 + # https://github.com/urllib3/urllib3/pull/611 # https://bugs.python.org/issue658327 try: sock = socket.socket(socket.AF_INET6) diff --git a/lib/urllib3/util/proxy.py b/lib/urllib3/util/proxy.py new file mode 100644 index 00000000..2199cc7b --- /dev/null +++ b/lib/urllib3/util/proxy.py @@ -0,0 +1,57 @@ +from .ssl_ import create_urllib3_context, resolve_cert_reqs, resolve_ssl_version + + +def connection_requires_http_tunnel( + proxy_url=None, proxy_config=None, destination_scheme=None +): + """ + Returns True if the connection requires an HTTP CONNECT through the proxy. + + :param URL proxy_url: + URL of the proxy. + :param ProxyConfig proxy_config: + Proxy configuration from poolmanager.py + :param str destination_scheme: + The scheme of the destination. (i.e https, http, etc) + """ + # If we're not using a proxy, no way to use a tunnel. + if proxy_url is None: + return False + + # HTTP destinations never require tunneling, we always forward. + if destination_scheme == "http": + return False + + # Support for forwarding with HTTPS proxies and HTTPS destinations. + if ( + proxy_url.scheme == "https" + and proxy_config + and proxy_config.use_forwarding_for_https + ): + return False + + # Otherwise always use a tunnel. + return True + + +def create_proxy_ssl_context( + ssl_version, cert_reqs, ca_certs=None, ca_cert_dir=None, ca_cert_data=None +): + """ + Generates a default proxy ssl context if one hasn't been provided by the + user. + """ + ssl_context = create_urllib3_context( + ssl_version=resolve_ssl_version(ssl_version), + cert_reqs=resolve_cert_reqs(cert_reqs), + ) + + if ( + not ca_certs + and not ca_cert_dir + and not ca_cert_data + and hasattr(ssl_context, "load_default_certs") + ): + ssl_context.load_default_certs() + + return ssl_context diff --git a/lib/urllib3/util/queue.py b/lib/urllib3/util/queue.py index d3d379a1..41784104 100644 --- a/lib/urllib3/util/queue.py +++ b/lib/urllib3/util/queue.py @@ -1,4 +1,5 @@ import collections + from ..packages import six from ..packages.six.moves import queue diff --git a/lib/urllib3/util/request.py b/lib/urllib3/util/request.py index 262a6d61..b574b081 100644 --- a/lib/urllib3/util/request.py +++ b/lib/urllib3/util/request.py @@ -1,12 +1,23 @@ from __future__ import absolute_import + from base64 import b64encode -from ..packages.six import b, integer_types from ..exceptions import UnrewindableBodyError +from ..packages.six import b, integer_types + +# Pass as a value within ``headers`` to skip +# emitting some HTTP headers that are added automatically. +# The only headers that are supported are ``Accept-Encoding``, +# ``Host``, and ``User-Agent``. +SKIP_HEADER = "@@@SKIP_HEADER@@@" +SKIPPABLE_HEADERS = frozenset(["accept-encoding", "host", "user-agent"]) ACCEPT_ENCODING = "gzip,deflate" try: - import brotli as _unused_module_brotli # noqa: F401 + try: + import brotlicffi as _unused_module_brotli # noqa: F401 + except ImportError: + import brotli as _unused_module_brotli # noqa: F401 except ImportError: pass else: @@ -122,7 +133,7 @@ def rewind_body(body, body_pos): body_seek(body_pos) except (IOError, OSError): raise UnrewindableBodyError( - "An error occurred when rewinding request " "body for redirect/retry." + "An error occurred when rewinding request body for redirect/retry." ) elif body_pos is _FAILEDTELL: raise UnrewindableBodyError( @@ -131,5 +142,5 @@ def rewind_body(body, body_pos): ) else: raise ValueError( - "body_pos must be of type integer, " "instead it was %s." % type(body_pos) + "body_pos must be of type integer, instead it was %s." % type(body_pos) ) diff --git a/lib/urllib3/util/response.py b/lib/urllib3/util/response.py index 715868dd..5ea609cc 100644 --- a/lib/urllib3/util/response.py +++ b/lib/urllib3/util/response.py @@ -1,7 +1,9 @@ from __future__ import absolute_import -from ..packages.six.moves import http_client as httplib + +from email.errors import MultipartInvariantViolationDefect, StartBoundaryNotFoundDefect from ..exceptions import HeaderParsingError +from ..packages.six.moves import http_client as httplib def is_fp_closed(obj): @@ -42,8 +44,7 @@ def assert_header_parsing(headers): Only works on Python 3. - :param headers: Headers to verify. - :type headers: `httplib.HTTPMessage`. + :param http.client.HTTPMessage headers: Headers to verify. :raises urllib3.exceptions.HeaderParsingError: If parsing errors are found. @@ -66,6 +67,25 @@ def assert_header_parsing(headers): if isinstance(payload, (bytes, str)): unparsed_data = payload + if defects: + # httplib is assuming a response body is available + # when parsing headers even when httplib only sends + # header data to parse_headers() This results in + # defects on multipart responses in particular. + # See: https://github.com/urllib3/urllib3/issues/800 + + # So we ignore the following defects: + # - StartBoundaryNotFoundDefect: + # The claimed start boundary was never found. + # - MultipartInvariantViolationDefect: + # A message claimed to be a multipart but no subparts were found. + defects = [ + defect + for defect in defects + if not isinstance( + defect, (StartBoundaryNotFoundDefect, MultipartInvariantViolationDefect) + ) + ] if defects or unparsed_data: raise HeaderParsingError(defects=defects, unparsed_data=unparsed_data) @@ -76,8 +96,9 @@ def is_response_to_head(response): Checks whether the request of a response has been a HEAD-request. Handles the quirks of AppEngine. - :param conn: - :type conn: :class:`httplib.HTTPResponse` + :param http.client.HTTPResponse response: + Response to check if the originating request + used 'HEAD' as a method. """ # FIXME: Can we do this somehow without accessing private httplib _method? method = response._method diff --git a/lib/urllib3/util/retry.py b/lib/urllib3/util/retry.py index 5a049fe6..60ef6c4f 100644 --- a/lib/urllib3/util/retry.py +++ b/lib/urllib3/util/retry.py @@ -1,22 +1,24 @@ from __future__ import absolute_import -import time + +import email import logging +import re +import time +import warnings from collections import namedtuple from itertools import takewhile -import email -import re from ..exceptions import ( ConnectTimeoutError, + InvalidHeader, MaxRetryError, ProtocolError, + ProxyError, ReadTimeoutError, ResponseError, - InvalidHeader, ) from ..packages import six - log = logging.getLogger(__name__) @@ -26,8 +28,69 @@ RequestHistory = namedtuple( ) +# TODO: In v2 we can remove this sentinel and metaclass with deprecated options. +_Default = object() + + +class _RetryMeta(type): + @property + def DEFAULT_METHOD_WHITELIST(cls): + warnings.warn( + "Using 'Retry.DEFAULT_METHOD_WHITELIST' is deprecated and " + "will be removed in v2.0. Use 'Retry.DEFAULT_ALLOWED_METHODS' instead", + DeprecationWarning, + ) + return cls.DEFAULT_ALLOWED_METHODS + + @DEFAULT_METHOD_WHITELIST.setter + def DEFAULT_METHOD_WHITELIST(cls, value): + warnings.warn( + "Using 'Retry.DEFAULT_METHOD_WHITELIST' is deprecated and " + "will be removed in v2.0. Use 'Retry.DEFAULT_ALLOWED_METHODS' instead", + DeprecationWarning, + ) + cls.DEFAULT_ALLOWED_METHODS = value + + @property + def DEFAULT_REDIRECT_HEADERS_BLACKLIST(cls): + warnings.warn( + "Using 'Retry.DEFAULT_REDIRECT_HEADERS_BLACKLIST' is deprecated and " + "will be removed in v2.0. Use 'Retry.DEFAULT_REMOVE_HEADERS_ON_REDIRECT' instead", + DeprecationWarning, + ) + return cls.DEFAULT_REMOVE_HEADERS_ON_REDIRECT + + @DEFAULT_REDIRECT_HEADERS_BLACKLIST.setter + def DEFAULT_REDIRECT_HEADERS_BLACKLIST(cls, value): + warnings.warn( + "Using 'Retry.DEFAULT_REDIRECT_HEADERS_BLACKLIST' is deprecated and " + "will be removed in v2.0. Use 'Retry.DEFAULT_REMOVE_HEADERS_ON_REDIRECT' instead", + DeprecationWarning, + ) + cls.DEFAULT_REMOVE_HEADERS_ON_REDIRECT = value + + @property + def BACKOFF_MAX(cls): + warnings.warn( + "Using 'Retry.BACKOFF_MAX' is deprecated and " + "will be removed in v2.0. Use 'Retry.DEFAULT_BACKOFF_MAX' instead", + DeprecationWarning, + ) + return cls.DEFAULT_BACKOFF_MAX + + @BACKOFF_MAX.setter + def BACKOFF_MAX(cls, value): + warnings.warn( + "Using 'Retry.BACKOFF_MAX' is deprecated and " + "will be removed in v2.0. Use 'Retry.DEFAULT_BACKOFF_MAX' instead", + DeprecationWarning, + ) + cls.DEFAULT_BACKOFF_MAX = value + + +@six.add_metaclass(_RetryMeta) class Retry(object): - """ Retry configuration. + """Retry configuration. Each retry attempt will create a new Retry object with updated values, so they can be safely reused. @@ -53,8 +116,7 @@ class Retry(object): Total number of retries to allow. Takes precedence over other counts. Set to ``None`` to remove this constraint and fall back on other - counts. It's a good idea to set this to some sensibly-high value to - account for unexpected edge cases and avoid infinite retry loops. + counts. Set to ``0`` to fail on the first retry. @@ -95,18 +157,35 @@ class Retry(object): Set to ``0`` to fail on the first retry of this type. - :param iterable method_whitelist: + :param int other: + How many times to retry on other errors. + + Other errors are errors that are not connect, read, redirect or status errors. + These errors might be raised after the request was sent to the server, so the + request might have side-effects. + + Set to ``0`` to fail on the first retry of this type. + + If ``total`` is not set, it's a good idea to set this to 0 to account + for unexpected edge cases and avoid infinite retry loops. + + :param iterable allowed_methods: Set of uppercased HTTP method verbs that we should retry on. By default, we only retry on methods which are considered to be idempotent (multiple requests with the same parameters end with the - same state). See :attr:`Retry.DEFAULT_METHOD_WHITELIST`. + same state). See :attr:`Retry.DEFAULT_ALLOWED_METHODS`. Set to a ``False`` value to retry on any verb. + .. warning:: + + Previously this parameter was named ``method_whitelist``, that + usage is deprecated in v1.26.0 and will be removed in v2.0. + :param iterable status_forcelist: A set of integer HTTP status codes that we should force a retry on. - A retry is initiated if the request method is in ``method_whitelist`` + A retry is initiated if the request method is in ``allowed_methods`` and the response status code is in ``status_forcelist``. By default, this is disabled with ``None``. @@ -120,7 +199,7 @@ class Retry(object): seconds. If the backoff_factor is 0.1, then :func:`.sleep` will sleep for [0.0s, 0.2s, 0.4s, ...] between retries. It will never be longer - than :attr:`Retry.BACKOFF_MAX`. + than :attr:`Retry.DEFAULT_BACKOFF_MAX`. By default, backoff is disabled (set to 0). @@ -147,16 +226,19 @@ class Retry(object): request. """ - DEFAULT_METHOD_WHITELIST = frozenset( + #: Default methods to be used for ``allowed_methods`` + DEFAULT_ALLOWED_METHODS = frozenset( ["HEAD", "GET", "PUT", "DELETE", "OPTIONS", "TRACE"] ) + #: Default status codes to be used for ``status_forcelist`` RETRY_AFTER_STATUS_CODES = frozenset([413, 429, 503]) - DEFAULT_REDIRECT_HEADERS_BLACKLIST = frozenset(["Authorization"]) + #: Default headers to be used for ``remove_headers_on_redirect`` + DEFAULT_REMOVE_HEADERS_ON_REDIRECT = frozenset(["Cookie", "Authorization"]) #: Maximum backoff time. - BACKOFF_MAX = 120 + DEFAULT_BACKOFF_MAX = 120 def __init__( self, @@ -165,20 +247,43 @@ class Retry(object): read=None, redirect=None, status=None, - method_whitelist=DEFAULT_METHOD_WHITELIST, + other=None, + allowed_methods=_Default, status_forcelist=None, backoff_factor=0, raise_on_redirect=True, raise_on_status=True, history=None, respect_retry_after_header=True, - remove_headers_on_redirect=DEFAULT_REDIRECT_HEADERS_BLACKLIST, + remove_headers_on_redirect=_Default, + # TODO: Deprecated, remove in v2.0 + method_whitelist=_Default, ): + if method_whitelist is not _Default: + if allowed_methods is not _Default: + raise ValueError( + "Using both 'allowed_methods' and " + "'method_whitelist' together is not allowed. " + "Instead only use 'allowed_methods'" + ) + warnings.warn( + "Using 'method_whitelist' with Retry is deprecated and " + "will be removed in v2.0. Use 'allowed_methods' instead", + DeprecationWarning, + stacklevel=2, + ) + allowed_methods = method_whitelist + if allowed_methods is _Default: + allowed_methods = self.DEFAULT_ALLOWED_METHODS + if remove_headers_on_redirect is _Default: + remove_headers_on_redirect = self.DEFAULT_REMOVE_HEADERS_ON_REDIRECT + self.total = total self.connect = connect self.read = read self.status = status + self.other = other if redirect is False or total is False: redirect = 0 @@ -186,7 +291,7 @@ class Retry(object): self.redirect = redirect self.status_forcelist = status_forcelist or set() - self.method_whitelist = method_whitelist + self.allowed_methods = allowed_methods self.backoff_factor = backoff_factor self.raise_on_redirect = raise_on_redirect self.raise_on_status = raise_on_status @@ -203,7 +308,7 @@ class Retry(object): read=self.read, redirect=self.redirect, status=self.status, - method_whitelist=self.method_whitelist, + other=self.other, status_forcelist=self.status_forcelist, backoff_factor=self.backoff_factor, raise_on_redirect=self.raise_on_redirect, @@ -212,12 +317,29 @@ class Retry(object): remove_headers_on_redirect=self.remove_headers_on_redirect, respect_retry_after_header=self.respect_retry_after_header, ) + + # TODO: If already given in **kw we use what's given to us + # If not given we need to figure out what to pass. We decide + # based on whether our class has the 'method_whitelist' property + # and if so we pass the deprecated 'method_whitelist' otherwise + # we use 'allowed_methods'. Remove in v2.0 + if "method_whitelist" not in kw and "allowed_methods" not in kw: + if "method_whitelist" in self.__dict__: + warnings.warn( + "Using 'method_whitelist' with Retry is deprecated and " + "will be removed in v2.0. Use 'allowed_methods' instead", + DeprecationWarning, + ) + params["method_whitelist"] = self.allowed_methods + else: + params["allowed_methods"] = self.allowed_methods + params.update(kw) return type(self)(**params) @classmethod def from_int(cls, retries, redirect=True, default=None): - """ Backwards-compatibility for the old retries format.""" + """Backwards-compatibility for the old retries format.""" if retries is None: retries = default if default is not None else cls.DEFAULT @@ -230,7 +352,7 @@ class Retry(object): return new_retries def get_backoff_time(self): - """ Formula for computing the current backoff + """Formula for computing the current backoff :rtype: float """ @@ -244,17 +366,24 @@ class Retry(object): return 0 backoff_value = self.backoff_factor * (2 ** (consecutive_errors_len - 1)) - return min(self.BACKOFF_MAX, backoff_value) + return min(self.DEFAULT_BACKOFF_MAX, backoff_value) def parse_retry_after(self, retry_after): # Whitespace: https://tools.ietf.org/html/rfc7230#section-3.2.4 if re.match(r"^\s*[0-9]+\s*$", retry_after): seconds = int(retry_after) else: - retry_date_tuple = email.utils.parsedate(retry_after) + retry_date_tuple = email.utils.parsedate_tz(retry_after) if retry_date_tuple is None: raise InvalidHeader("Invalid Retry-After header: %s" % retry_after) - retry_date = time.mktime(retry_date_tuple) + if retry_date_tuple[9] is None: # Python 2 + # Assume UTC if no timezone was specified + # On Python2.7, parsedate_tz returns None for a timezone offset + # instead of 0 if no timezone is given, where mktime_tz treats + # a None timezone offset as local time. + retry_date_tuple = retry_date_tuple[:9] + (0,) + retry_date_tuple[10:] + + retry_date = email.utils.mktime_tz(retry_date_tuple) seconds = retry_date - time.time() if seconds < 0: @@ -263,9 +392,9 @@ class Retry(object): return seconds def get_retry_after(self, response): - """ Get the value of Retry-After in seconds. """ + """Get the value of Retry-After in seconds.""" - retry_after = response.getheader("Retry-After") + retry_after = response.headers.get("Retry-After") if retry_after is None: return None @@ -287,7 +416,7 @@ class Retry(object): time.sleep(backoff) def sleep(self, response=None): - """ Sleep between retry attempts. + """Sleep between retry attempts. This method will respect a server's ``Retry-After`` response header and sleep the duration of the time requested. If that is not present, it @@ -303,28 +432,41 @@ class Retry(object): self._sleep_backoff() def _is_connection_error(self, err): - """ Errors when we're fairly sure that the server did not receive the + """Errors when we're fairly sure that the server did not receive the request, so it should be safe to retry. """ + if isinstance(err, ProxyError): + err = err.original_error return isinstance(err, ConnectTimeoutError) def _is_read_error(self, err): - """ Errors that occur after the request has been started, so we should + """Errors that occur after the request has been started, so we should assume that the server began processing it. """ return isinstance(err, (ReadTimeoutError, ProtocolError)) def _is_method_retryable(self, method): - """ Checks if a given HTTP method should be retried upon, depending if - it is included on the method whitelist. + """Checks if a given HTTP method should be retried upon, depending if + it is included in the allowed_methods """ - if self.method_whitelist and method.upper() not in self.method_whitelist: - return False + # TODO: For now favor if the Retry implementation sets its own method_whitelist + # property outside of our constructor to avoid breaking custom implementations. + if "method_whitelist" in self.__dict__: + warnings.warn( + "Using 'method_whitelist' with Retry is deprecated and " + "will be removed in v2.0. Use 'allowed_methods' instead", + DeprecationWarning, + ) + allowed_methods = self.method_whitelist + else: + allowed_methods = self.allowed_methods + if allowed_methods and method.upper() not in allowed_methods: + return False return True def is_retry(self, method, status_code, has_retry_after=False): - """ Is this method/status code retryable? (Based on whitelists and control + """Is this method/status code retryable? (Based on allowlists and control variables such as the number of total retries to allow, whether to respect the Retry-After header, whether this header is present, and whether the returned status code is on the list of status codes to @@ -344,8 +486,15 @@ class Retry(object): ) def is_exhausted(self): - """ Are we out of retries? """ - retry_counts = (self.total, self.connect, self.read, self.redirect, self.status) + """Are we out of retries?""" + retry_counts = ( + self.total, + self.connect, + self.read, + self.redirect, + self.status, + self.other, + ) retry_counts = list(filter(None, retry_counts)) if not retry_counts: return False @@ -361,7 +510,7 @@ class Retry(object): _pool=None, _stacktrace=None, ): - """ Return a new Retry object with incremented retry counters. + """Return a new Retry object with incremented retry counters. :param response: A response object, or None, if the server did not return a response. @@ -383,6 +532,7 @@ class Retry(object): read = self.read redirect = self.redirect status_count = self.status + other = self.other cause = "unknown" status = None redirect_location = None @@ -401,6 +551,11 @@ class Retry(object): elif read is not None: read -= 1 + elif error: + # Other retry? + if other is not None: + other -= 1 + elif response and response.get_redirect_location(): # Redirect retry? if redirect is not None: @@ -411,7 +566,7 @@ class Retry(object): else: # Incrementing because of a server error like a 500 in - # status_forcelist and a the given method is in the whitelist + # status_forcelist and the given method is in the allowed_methods cause = ResponseError.GENERIC_ERROR if response and response.status: if status_count is not None: @@ -429,6 +584,7 @@ class Retry(object): read=read, redirect=redirect, status=status_count, + other=other, history=history, ) @@ -445,6 +601,20 @@ class Retry(object): "read={self.read}, redirect={self.redirect}, status={self.status})" ).format(cls=type(self), self=self) + def __getattr__(self, item): + if item == "method_whitelist": + # TODO: Remove this deprecated alias in v2.0 + warnings.warn( + "Using 'method_whitelist' with Retry is deprecated and " + "will be removed in v2.0. Use 'allowed_methods' instead", + DeprecationWarning, + ) + return self.allowed_methods + try: + return getattr(super(Retry, self), item) + except AttributeError: + return getattr(Retry, item) + # For backwards compatibility (equivalent to pre-v1.9): Retry.DEFAULT = Retry(3) diff --git a/lib/urllib3/util/ssl_.py b/lib/urllib3/util/ssl_.py index 8495b775..8f867812 100644 --- a/lib/urllib3/util/ssl_.py +++ b/lib/urllib3/util/ssl_.py @@ -1,21 +1,27 @@ from __future__ import absolute_import -import errno -import warnings -import hmac -import sys +import hmac +import os +import sys +import warnings from binascii import hexlify, unhexlify from hashlib import md5, sha1, sha256 -from .url import IPV4_RE, BRACELESS_IPV6_ADDRZ_RE -from ..exceptions import SSLError, InsecurePlatformWarning, SNIMissingWarning +from ..exceptions import ( + InsecurePlatformWarning, + ProxySchemeUnsupported, + SNIMissingWarning, + SSLError, +) from ..packages import six - +from .url import BRACELESS_IPV6_ADDRZ_RE, IPV4_RE SSLContext = None +SSLTransport = None HAS_SNI = False IS_PYOPENSSL = False IS_SECURETRANSPORT = False +ALPN_PROTOCOLS = ["http/1.1"] # Maps the length of a digest to a possible hash function producing this digest HASHFUNC_MAP = {32: md5, 40: sha1, 64: sha256} @@ -29,8 +35,8 @@ def _const_compare_digest_backport(a, b): Returns True if the digests match, and False otherwise. """ result = abs(len(a) - len(b)) - for l, r in zip(bytearray(a), bytearray(b)): - result |= l ^ r + for left, right in zip(bytearray(a), bytearray(b)): + result |= left ^ right return result == 0 @@ -38,11 +44,21 @@ _const_compare_digest = getattr(hmac, "compare_digest", _const_compare_digest_ba try: # Test for SSL features import ssl - from ssl import wrap_socket, CERT_REQUIRED + from ssl import CERT_REQUIRED, wrap_socket +except ImportError: + pass + +try: from ssl import HAS_SNI # Has SNI? except ImportError: pass +try: + from .ssltransport import SSLTransport +except ImportError: + pass + + try: # Platform-specific: Python 3.6 from ssl import PROTOCOL_TLS @@ -55,14 +71,25 @@ except ImportError: except ImportError: PROTOCOL_SSLv23 = PROTOCOL_TLS = 2 +try: + from ssl import PROTOCOL_TLS_CLIENT +except ImportError: + PROTOCOL_TLS_CLIENT = PROTOCOL_TLS + try: - from ssl import OP_NO_SSLv2, OP_NO_SSLv3, OP_NO_COMPRESSION + from ssl import OP_NO_COMPRESSION, OP_NO_SSLv2, OP_NO_SSLv3 except ImportError: OP_NO_SSLv2, OP_NO_SSLv3 = 0x1000000, 0x2000000 OP_NO_COMPRESSION = 0x20000 +try: # OP_NO_TICKET was added in Python 3.6 + from ssl import OP_NO_TICKET +except ImportError: + OP_NO_TICKET = 0x4000 + + # A secure default. # Sources for more information on TLS ciphers: # @@ -119,12 +146,15 @@ except ImportError: self.certfile = certfile self.keyfile = keyfile - def load_verify_locations(self, cafile=None, capath=None): + def load_verify_locations(self, cafile=None, capath=None, cadata=None): self.ca_certs = cafile if capath is not None: raise SSLError("CA directories not supported in older Pythons") + if cadata is not None: + raise SSLError("CA data not supported in older Pythons") + def set_ciphers(self, cipher_suite): self.ciphers = cipher_suite @@ -134,7 +164,7 @@ except ImportError: "urllib3 from configuring SSL appropriately and may cause " "certain SSL connections to fail. You can upgrade to a newer " "version of Python to solve this. For more information, see " - "https://urllib3.readthedocs.io/en/latest/advanced-usage.html" + "https://urllib3.readthedocs.io/en/1.26.x/advanced-usage.html" "#ssl-warnings", InsecurePlatformWarning, ) @@ -182,7 +212,7 @@ def resolve_cert_reqs(candidate): """ Resolves the argument to a numeric constant, which can be passed to the wrap_socket function/method from the ssl module. - Defaults to :data:`ssl.CERT_NONE`. + Defaults to :data:`ssl.CERT_REQUIRED`. If given a string it is assumed to be the name of the constant in the :mod:`ssl` module or its abbreviation. (So you can specify `REQUIRED` instead of `CERT_REQUIRED`. @@ -246,14 +276,18 @@ def create_urllib3_context( ``ssl.CERT_REQUIRED``. :param options: Specific OpenSSL options. These default to ``ssl.OP_NO_SSLv2``, - ``ssl.OP_NO_SSLv3``, ``ssl.OP_NO_COMPRESSION``. + ``ssl.OP_NO_SSLv3``, ``ssl.OP_NO_COMPRESSION``, and ``ssl.OP_NO_TICKET``. :param ciphers: Which cipher suites to allow the server to select. :returns: Constructed SSLContext object with specified options :rtype: SSLContext """ - context = SSLContext(ssl_version or PROTOCOL_TLS) + # PROTOCOL_TLS is deprecated in Python 3.10 + if not ssl_version or ssl_version == PROTOCOL_TLS: + ssl_version = PROTOCOL_TLS_CLIENT + + context = SSLContext(ssl_version) context.set_ciphers(ciphers or DEFAULT_CIPHERS) @@ -269,6 +303,11 @@ def create_urllib3_context( # Disable compression to prevent CRIME attacks for OpenSSL 1.0+ # (issue #309) options |= OP_NO_COMPRESSION + # TLSv1.2 only. Unless set explicitly, do not request tickets. + # This may save some bandwidth on wire, and although the ticket is encrypted, + # there is a risk associated with it being on wire, + # if the server is not rotating its ticketing keys properly. + options |= OP_NO_TICKET context.options |= options @@ -283,13 +322,33 @@ def create_urllib3_context( ) is not None: context.post_handshake_auth = True - context.verify_mode = cert_reqs - if ( - getattr(context, "check_hostname", None) is not None - ): # Platform-specific: Python 3.2 - # We do our own verification, including fingerprints and alternative - # hostnames. So disable it here - context.check_hostname = False + def disable_check_hostname(): + if ( + getattr(context, "check_hostname", None) is not None + ): # Platform-specific: Python 3.2 + # We do our own verification, including fingerprints and alternative + # hostnames. So disable it here + context.check_hostname = False + + # The order of the below lines setting verify_mode and check_hostname + # matter due to safe-guards SSLContext has to prevent an SSLContext with + # check_hostname=True, verify_mode=NONE/OPTIONAL. This is made even more + # complex because we don't know whether PROTOCOL_TLS_CLIENT will be used + # or not so we don't know the initial state of the freshly created SSLContext. + if cert_reqs == ssl.CERT_REQUIRED: + context.verify_mode = cert_reqs + disable_check_hostname() + else: + disable_check_hostname() + context.verify_mode = cert_reqs + + # Enable logging of TLS session keys via defacto standard environment variable + # 'SSLKEYLOGFILE', if the feature is available (Python 3.8+). Skip empty values. + if hasattr(context, "keylog_filename"): + sslkeylogfile = os.environ.get("SSLKEYLOGFILE") + if sslkeylogfile: + context.keylog_filename = sslkeylogfile + return context @@ -305,6 +364,8 @@ def ssl_wrap_socket( ssl_context=None, ca_cert_dir=None, key_password=None, + ca_cert_data=None, + tls_in_tls=False, ): """ All arguments except for server_hostname, ssl_context, and ca_cert_dir have @@ -323,6 +384,11 @@ def ssl_wrap_socket( SSLContext.load_verify_locations(). :param key_password: Optional password if the keyfile is encrypted. + :param ca_cert_data: + Optional string containing CA certificates in PEM format suitable for + passing as the cadata parameter to SSLContext.load_verify_locations() + :param tls_in_tls: + Use SSLTransport to wrap the existing socket. """ context = ssl_context if context is None: @@ -331,17 +397,11 @@ def ssl_wrap_socket( # this code. context = create_urllib3_context(ssl_version, cert_reqs, ciphers=ciphers) - if ca_certs or ca_cert_dir: + if ca_certs or ca_cert_dir or ca_cert_data: try: - context.load_verify_locations(ca_certs, ca_cert_dir) - except IOError as e: # Platform-specific: Python 2.7 + context.load_verify_locations(ca_certs, ca_cert_dir, ca_cert_data) + except (IOError, OSError) as e: raise SSLError(e) - # Py33 raises FileNotFoundError which subclasses OSError - # These are not equivalent unless we check the errno attribute - except OSError as e: # Platform-specific: Python 3.3 and beyond - if e.errno == errno.ENOENT: - raise SSLError(e) - raise elif ssl_context is None and hasattr(context, "load_default_certs"): # try to load OS default certs; works well on Windows (require Python3.4+) @@ -359,28 +419,39 @@ def ssl_wrap_socket( else: context.load_cert_chain(certfile, keyfile, key_password) + try: + if hasattr(context, "set_alpn_protocols"): + context.set_alpn_protocols(ALPN_PROTOCOLS) + except NotImplementedError: # Defensive: in CI, we always have set_alpn_protocols + pass + # If we detect server_hostname is an IP address then the SNI # extension should not be used according to RFC3546 Section 3.1 - # We shouldn't warn the user if SNI isn't available but we would - # not be using SNI anyways due to IP address for server_hostname. - if ( - server_hostname is not None and not is_ipaddress(server_hostname) - ) or IS_SECURETRANSPORT: - if HAS_SNI and server_hostname is not None: - return context.wrap_socket(sock, server_hostname=server_hostname) - + use_sni_hostname = server_hostname and not is_ipaddress(server_hostname) + # SecureTransport uses server_hostname in certificate verification. + send_sni = (use_sni_hostname and HAS_SNI) or ( + IS_SECURETRANSPORT and server_hostname + ) + # Do not warn the user if server_hostname is an invalid SNI hostname. + if not HAS_SNI and use_sni_hostname: warnings.warn( "An HTTPS request has been made, but the SNI (Server Name " "Indication) extension to TLS is not available on this platform. " "This may cause the server to present an incorrect TLS " "certificate, which can cause validation failures. You can upgrade to " "a newer version of Python to solve this. For more information, see " - "https://urllib3.readthedocs.io/en/latest/advanced-usage.html" + "https://urllib3.readthedocs.io/en/1.26.x/advanced-usage.html" "#ssl-warnings", SNIMissingWarning, ) - return context.wrap_socket(sock) + if send_sni: + ssl_sock = _ssl_wrap_socket_impl( + sock, context, tls_in_tls, server_hostname=server_hostname + ) + else: + ssl_sock = _ssl_wrap_socket_impl(sock, context, tls_in_tls) + return ssl_sock def is_ipaddress(hostname): @@ -405,3 +476,20 @@ def _is_key_file_encrypted(key_file): return True return False + + +def _ssl_wrap_socket_impl(sock, ssl_context, tls_in_tls, server_hostname=None): + if tls_in_tls: + if not SSLTransport: + # Import error, ssl is not available. + raise ProxySchemeUnsupported( + "TLS in TLS requires support for the 'ssl' module" + ) + + SSLTransport._validate_ssl_context_for_tls_in_tls(ssl_context) + return SSLTransport(sock, ssl_context, server_hostname) + + if server_hostname: + return ssl_context.wrap_socket(sock, server_hostname=server_hostname) + else: + return ssl_context.wrap_socket(sock) diff --git a/lib/urllib3/packages/ssl_match_hostname/_implementation.py b/lib/urllib3/util/ssl_match_hostname.py similarity index 89% rename from lib/urllib3/packages/ssl_match_hostname/_implementation.py rename to lib/urllib3/util/ssl_match_hostname.py index 2d8e7a17..1dd950c4 100644 --- a/lib/urllib3/packages/ssl_match_hostname/_implementation.py +++ b/lib/urllib3/util/ssl_match_hostname.py @@ -9,7 +9,7 @@ import sys # ipaddress has been backported to 2.6+ in pypi. If it is installed on the # system, use it to handle IPAddress ServerAltnames (this was added in # python-3.5) otherwise only do DNS matching. This allows -# backports.ssl_match_hostname to continue to be used in Python 2.7. +# util.ssl_match_hostname to continue to be used in Python 2.7. try: import ipaddress except ImportError: @@ -78,7 +78,8 @@ def _dnsname_match(dn, hostname, max_wildcards=1): def _to_unicode(obj): if isinstance(obj, str) and sys.version_info < (3,): - obj = unicode(obj, encoding="ascii", errors="strict") + # ignored flake8 # F821 to support python 2.7 function + obj = unicode(obj, encoding="ascii", errors="strict") # noqa: F821 return obj @@ -111,11 +112,9 @@ def match_hostname(cert, hostname): try: # Divergence from upstream: ipaddress can't handle byte str host_ip = ipaddress.ip_address(_to_unicode(hostname)) - except ValueError: - # Not an IP address (common case) - host_ip = None - except UnicodeError: - # Divergence from upstream: Have to deal with ipaddress not taking + except (UnicodeError, ValueError): + # ValueError: Not an IP address (common case) + # UnicodeError: Divergence from upstream: Have to deal with ipaddress not taking # byte strings. addresses should be all ascii, so we consider it not # an ipaddress in this case host_ip = None @@ -123,7 +122,7 @@ def match_hostname(cert, hostname): # Divergence from upstream: Make ipaddress library optional if ipaddress is None: host_ip = None - else: + else: # Defensive raise dnsnames = [] san = cert.get("subjectAltName", ()) @@ -153,10 +152,8 @@ def match_hostname(cert, hostname): "doesn't match either of %s" % (hostname, ", ".join(map(repr, dnsnames))) ) elif len(dnsnames) == 1: - raise CertificateError( - "hostname %r " "doesn't match %r" % (hostname, dnsnames[0]) - ) + raise CertificateError("hostname %r doesn't match %r" % (hostname, dnsnames[0])) else: raise CertificateError( - "no appropriate commonName or " "subjectAltName fields were found" + "no appropriate commonName or subjectAltName fields were found" ) diff --git a/lib/urllib3/util/ssltransport.py b/lib/urllib3/util/ssltransport.py new file mode 100644 index 00000000..4a7105d1 --- /dev/null +++ b/lib/urllib3/util/ssltransport.py @@ -0,0 +1,221 @@ +import io +import socket +import ssl + +from ..exceptions import ProxySchemeUnsupported +from ..packages import six + +SSL_BLOCKSIZE = 16384 + + +class SSLTransport: + """ + The SSLTransport wraps an existing socket and establishes an SSL connection. + + Contrary to Python's implementation of SSLSocket, it allows you to chain + multiple TLS connections together. It's particularly useful if you need to + implement TLS within TLS. + + The class supports most of the socket API operations. + """ + + @staticmethod + def _validate_ssl_context_for_tls_in_tls(ssl_context): + """ + Raises a ProxySchemeUnsupported if the provided ssl_context can't be used + for TLS in TLS. + + The only requirement is that the ssl_context provides the 'wrap_bio' + methods. + """ + + if not hasattr(ssl_context, "wrap_bio"): + if six.PY2: + raise ProxySchemeUnsupported( + "TLS in TLS requires SSLContext.wrap_bio() which isn't " + "supported on Python 2" + ) + else: + raise ProxySchemeUnsupported( + "TLS in TLS requires SSLContext.wrap_bio() which isn't " + "available on non-native SSLContext" + ) + + def __init__( + self, socket, ssl_context, server_hostname=None, suppress_ragged_eofs=True + ): + """ + Create an SSLTransport around socket using the provided ssl_context. + """ + self.incoming = ssl.MemoryBIO() + self.outgoing = ssl.MemoryBIO() + + self.suppress_ragged_eofs = suppress_ragged_eofs + self.socket = socket + + self.sslobj = ssl_context.wrap_bio( + self.incoming, self.outgoing, server_hostname=server_hostname + ) + + # Perform initial handshake. + self._ssl_io_loop(self.sslobj.do_handshake) + + def __enter__(self): + return self + + def __exit__(self, *_): + self.close() + + def fileno(self): + return self.socket.fileno() + + def read(self, len=1024, buffer=None): + return self._wrap_ssl_read(len, buffer) + + def recv(self, len=1024, flags=0): + if flags != 0: + raise ValueError("non-zero flags not allowed in calls to recv") + return self._wrap_ssl_read(len) + + def recv_into(self, buffer, nbytes=None, flags=0): + if flags != 0: + raise ValueError("non-zero flags not allowed in calls to recv_into") + if buffer and (nbytes is None): + nbytes = len(buffer) + elif nbytes is None: + nbytes = 1024 + return self.read(nbytes, buffer) + + def sendall(self, data, flags=0): + if flags != 0: + raise ValueError("non-zero flags not allowed in calls to sendall") + count = 0 + with memoryview(data) as view, view.cast("B") as byte_view: + amount = len(byte_view) + while count < amount: + v = self.send(byte_view[count:]) + count += v + + def send(self, data, flags=0): + if flags != 0: + raise ValueError("non-zero flags not allowed in calls to send") + response = self._ssl_io_loop(self.sslobj.write, data) + return response + + def makefile( + self, mode="r", buffering=None, encoding=None, errors=None, newline=None + ): + """ + Python's httpclient uses makefile and buffered io when reading HTTP + messages and we need to support it. + + This is unfortunately a copy and paste of socket.py makefile with small + changes to point to the socket directly. + """ + if not set(mode) <= {"r", "w", "b"}: + raise ValueError("invalid mode %r (only r, w, b allowed)" % (mode,)) + + writing = "w" in mode + reading = "r" in mode or not writing + assert reading or writing + binary = "b" in mode + rawmode = "" + if reading: + rawmode += "r" + if writing: + rawmode += "w" + raw = socket.SocketIO(self, rawmode) + self.socket._io_refs += 1 + if buffering is None: + buffering = -1 + if buffering < 0: + buffering = io.DEFAULT_BUFFER_SIZE + if buffering == 0: + if not binary: + raise ValueError("unbuffered streams must be binary") + return raw + if reading and writing: + buffer = io.BufferedRWPair(raw, raw, buffering) + elif reading: + buffer = io.BufferedReader(raw, buffering) + else: + assert writing + buffer = io.BufferedWriter(raw, buffering) + if binary: + return buffer + text = io.TextIOWrapper(buffer, encoding, errors, newline) + text.mode = mode + return text + + def unwrap(self): + self._ssl_io_loop(self.sslobj.unwrap) + + def close(self): + self.socket.close() + + def getpeercert(self, binary_form=False): + return self.sslobj.getpeercert(binary_form) + + def version(self): + return self.sslobj.version() + + def cipher(self): + return self.sslobj.cipher() + + def selected_alpn_protocol(self): + return self.sslobj.selected_alpn_protocol() + + def selected_npn_protocol(self): + return self.sslobj.selected_npn_protocol() + + def shared_ciphers(self): + return self.sslobj.shared_ciphers() + + def compression(self): + return self.sslobj.compression() + + def settimeout(self, value): + self.socket.settimeout(value) + + def gettimeout(self): + return self.socket.gettimeout() + + def _decref_socketios(self): + self.socket._decref_socketios() + + def _wrap_ssl_read(self, len, buffer=None): + try: + return self._ssl_io_loop(self.sslobj.read, len, buffer) + except ssl.SSLError as e: + if e.errno == ssl.SSL_ERROR_EOF and self.suppress_ragged_eofs: + return 0 # eof, return 0. + else: + raise + + def _ssl_io_loop(self, func, *args): + """Performs an I/O loop between incoming/outgoing and the socket.""" + should_loop = True + ret = None + + while should_loop: + errno = None + try: + ret = func(*args) + except ssl.SSLError as e: + if e.errno not in (ssl.SSL_ERROR_WANT_READ, ssl.SSL_ERROR_WANT_WRITE): + # WANT_READ, and WANT_WRITE are expected, others are not. + raise e + errno = e.errno + + buf = self.outgoing.read() + self.socket.sendall(buf) + + if errno is None: + should_loop = False + elif errno == ssl.SSL_ERROR_WANT_READ: + buf = self.socket.recv(SSL_BLOCKSIZE) + if buf: + self.incoming.write(buf) + else: + self.incoming.write_eof() + return ret diff --git a/lib/urllib3/util/timeout.py b/lib/urllib3/util/timeout.py index c1dc1e97..78e18a62 100644 --- a/lib/urllib3/util/timeout.py +++ b/lib/urllib3/util/timeout.py @@ -1,10 +1,10 @@ from __future__ import absolute_import -# The default socket timeout, used by httplib to indicate that no timeout was -# specified by the user -from socket import _GLOBAL_DEFAULT_TIMEOUT import time +# The default socket timeout, used by httplib to indicate that no timeout was; specified by the user +from socket import _GLOBAL_DEFAULT_TIMEOUT, getdefaulttimeout + from ..exceptions import TimeoutStateError # A sentinel value to indicate that no timeout was specified by the user in @@ -17,22 +17,28 @@ current_time = getattr(time, "monotonic", time.time) class Timeout(object): - """ Timeout configuration. + """Timeout configuration. - Timeouts can be defined as a default for a pool:: + Timeouts can be defined as a default for a pool: - timeout = Timeout(connect=2.0, read=7.0) - http = PoolManager(timeout=timeout) - response = http.request('GET', 'http://example.com/') + .. code-block:: python - Or per-request (which overrides the default for the pool):: + timeout = Timeout(connect=2.0, read=7.0) + http = PoolManager(timeout=timeout) + response = http.request('GET', 'http://example.com/') - response = http.request('GET', 'http://example.com/', timeout=Timeout(10)) + Or per-request (which overrides the default for the pool): - Timeouts can be disabled by setting all the parameters to ``None``:: + .. code-block:: python - no_timeout = Timeout(connect=None, read=None) - response = http.request('GET', 'http://example.com/, timeout=no_timeout) + response = http.request('GET', 'http://example.com/', timeout=Timeout(10)) + + Timeouts can be disabled by setting all the parameters to ``None``: + + .. code-block:: python + + no_timeout = Timeout(connect=None, read=None) + response = http.request('GET', 'http://example.com/, timeout=no_timeout) :param total: @@ -43,7 +49,7 @@ class Timeout(object): Defaults to None. - :type total: integer, float, or None + :type total: int, float, or None :param connect: The maximum amount of time (in seconds) to wait for a connection @@ -53,7 +59,7 @@ class Timeout(object): `_. None will set an infinite timeout for connection attempts. - :type connect: integer, float, or None + :type connect: int, float, or None :param read: The maximum amount of time (in seconds) to wait between consecutive @@ -63,7 +69,7 @@ class Timeout(object): `_. None will set an infinite timeout. - :type read: integer, float, or None + :type read: int, float, or None .. note:: @@ -98,7 +104,7 @@ class Timeout(object): self.total = self._validate_timeout(total, "total") self._start_connect = None - def __str__(self): + def __repr__(self): return "%s(connect=%r, read=%r, total=%r)" % ( type(self).__name__, self._connect, @@ -106,9 +112,16 @@ class Timeout(object): self.total, ) + # __str__ provided for backwards compatibility + __str__ = __repr__ + + @classmethod + def resolve_default_timeout(cls, timeout): + return getdefaulttimeout() if timeout is cls.DEFAULT_TIMEOUT else timeout + @classmethod def _validate_timeout(cls, value, name): - """ Check that a timeout attribute is valid. + """Check that a timeout attribute is valid. :param value: The timeout value to validate :param name: The name of the timeout attribute to validate. This is @@ -154,7 +167,7 @@ class Timeout(object): @classmethod def from_float(cls, timeout): - """ Create a new Timeout from a legacy timeout value. + """Create a new Timeout from a legacy timeout value. The timeout value used by httplib.py sets the same timeout on the connect(), and recv() socket requests. This creates a :class:`Timeout` @@ -169,7 +182,7 @@ class Timeout(object): return Timeout(read=timeout, connect=timeout) def clone(self): - """ Create a copy of the timeout object + """Create a copy of the timeout object Timeout properties are stored per-pool but each request needs a fresh Timeout object to ensure each one has its own start/stop configured. @@ -183,7 +196,7 @@ class Timeout(object): return Timeout(connect=self._connect, read=self._read, total=self.total) def start_connect(self): - """ Start the timeout clock, used during a connect() attempt + """Start the timeout clock, used during a connect() attempt :raises urllib3.exceptions.TimeoutStateError: if you attempt to start a timer that has been started already. @@ -194,7 +207,7 @@ class Timeout(object): return self._start_connect def get_connect_duration(self): - """ Gets the time elapsed since the call to :meth:`start_connect`. + """Gets the time elapsed since the call to :meth:`start_connect`. :return: Elapsed time in seconds. :rtype: float @@ -203,13 +216,13 @@ class Timeout(object): """ if self._start_connect is None: raise TimeoutStateError( - "Can't get connect duration for timer " "that has not started." + "Can't get connect duration for timer that has not started." ) return current_time() - self._start_connect @property def connect_timeout(self): - """ Get the value to use when setting a connection timeout. + """Get the value to use when setting a connection timeout. This will be a positive float or integer, the value None (never timeout), or the default system timeout. @@ -227,7 +240,7 @@ class Timeout(object): @property def read_timeout(self): - """ Get the value for the read timeout. + """Get the value for the read timeout. This assumes some time has elapsed in the connection timeout and computes the read timeout appropriately. diff --git a/lib/urllib3/util/url.py b/lib/urllib3/util/url.py index 9675f742..e5682d3b 100644 --- a/lib/urllib3/util/url.py +++ b/lib/urllib3/util/url.py @@ -1,11 +1,11 @@ from __future__ import absolute_import + import re from collections import namedtuple from ..exceptions import LocationParseError from ..packages import six - url_attrs = ["scheme", "auth", "host", "port", "path", "query", "fragment"] # We only want to normalize urls with an HTTP(S) scheme. @@ -18,7 +18,7 @@ PERCENT_RE = re.compile(r"%[a-fA-F0-9]{2}") SCHEME_RE = re.compile(r"^(?:[a-zA-Z][a-zA-Z0-9+-]*:|/)") URI_RE = re.compile( r"^(?:([a-zA-Z][a-zA-Z0-9+.-]*):)?" - r"(?://([^/?#]*))?" + r"(?://([^\\/?#]*))?" r"([^?#]*)" r"(?:\?([^#]*))?" r"(?:#(.*))?$", @@ -50,12 +50,12 @@ _variations = [ "(?:(?:%(hex)s:){0,6}%(hex)s)?::", ] -UNRESERVED_PAT = r"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789._!\-~" +UNRESERVED_PAT = r"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789._\-~" IPV6_PAT = "(?:" + "|".join([x % _subs for x in _variations]) + ")" ZONE_ID_PAT = "(?:%25|%)(?:[" + UNRESERVED_PAT + "]|%[a-fA-F0-9]{2})+" IPV6_ADDRZ_PAT = r"\[" + IPV6_PAT + r"(?:" + ZONE_ID_PAT + r")?\]" REG_NAME_PAT = r"(?:[^\[\]%:/?#]|%[a-fA-F0-9]{2})*" -TARGET_RE = re.compile(r"^(/[^?]*)(?:\?([^#]+))?(?:#(.*))?$") +TARGET_RE = re.compile(r"^(/[^?#]*)(?:\?([^#]*))?(?:#.*)?$") IPV4_RE = re.compile("^" + IPV4_PAT + "$") IPV6_RE = re.compile("^" + IPV6_PAT + "$") @@ -63,12 +63,12 @@ IPV6_ADDRZ_RE = re.compile("^" + IPV6_ADDRZ_PAT + "$") BRACELESS_IPV6_ADDRZ_RE = re.compile("^" + IPV6_ADDRZ_PAT[2:-2] + "$") ZONE_ID_RE = re.compile("(" + ZONE_ID_PAT + r")\]$") -SUBAUTHORITY_PAT = (u"^(?:(.*)@)?(%s|%s|%s)(?::([0-9]{0,5}))?$") % ( +_HOST_PORT_PAT = ("^(%s|%s|%s)(?::0*?(|0|[1-9][0-9]{0,4}))?$") % ( REG_NAME_PAT, IPV4_PAT, IPV6_ADDRZ_PAT, ) -SUBAUTHORITY_RE = re.compile(SUBAUTHORITY_PAT, re.UNICODE | re.DOTALL) +_HOST_PORT_RE = re.compile(_HOST_PORT_PAT, re.UNICODE | re.DOTALL) UNRESERVED_CHARS = set( "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789._-~" @@ -216,18 +216,15 @@ def _encode_invalid_chars(component, allowed_chars, encoding="utf-8"): component = six.ensure_text(component) + # Normalize existing percent-encoded bytes. # Try to see if the component we're encoding is already percent-encoded # so we can skip all '%' characters but still encode all others. - percent_encodings = PERCENT_RE.findall(component) - - # Normalize existing percent-encoded bytes. - for enc in percent_encodings: - if not enc.isupper(): - component = component.replace(enc, enc.upper()) + component, percent_encodings = PERCENT_RE.subn( + lambda match: match.group(0).upper(), component + ) uri_bytes = component.encode("utf-8", "surrogatepass") - is_percent_encoded = len(percent_encodings) == uri_bytes.count(b"%") - + is_percent_encoded = percent_encodings == uri_bytes.count(b"%") encoded_component = bytearray() for i in range(0, len(uri_bytes)): @@ -237,7 +234,7 @@ def _encode_invalid_chars(component, allowed_chars, encoding="utf-8"): if (is_percent_encoded and byte == b"%") or ( byte_ord < 128 and byte.decode() in allowed_chars ): - encoded_component.extend(byte) + encoded_component += byte continue encoded_component.extend(b"%" + (hex(byte_ord)[2:].encode().zfill(2).upper())) @@ -282,6 +279,9 @@ def _normalize_host(host, scheme): if scheme in NORMALIZABLE_SCHEMES: is_ipv6 = IPV6_ADDRZ_RE.match(host) if is_ipv6: + # IPv6 hosts of the form 'a::b%zone' are encoded in a URL as + # such per RFC 6874: 'a::b%25zone'. Unquote the ZoneID + # separator as necessary to return a valid RFC 4007 scoped IP. match = ZONE_ID_RE.search(host) if match: start, end = match.span(1) @@ -303,7 +303,7 @@ def _normalize_host(host, scheme): def _idna_encode(name): - if name and any([ord(x) > 128 for x in name]): + if name and any(ord(x) >= 128 for x in name): try: import idna except ImportError: @@ -322,17 +322,11 @@ def _idna_encode(name): def _encode_target(target): """Percent-encodes a request target so that there are no invalid characters""" - if not target.startswith("/"): - return target - - path, query, fragment = TARGET_RE.match(target).groups() + path, query = TARGET_RE.match(target).groups() target = _encode_invalid_chars(path, PATH_CHARS) query = _encode_invalid_chars(query, QUERY_CHARS) - fragment = _encode_invalid_chars(fragment, FRAGMENT_CHARS) if query is not None: target += "?" + query - if fragment is not None: - target += "#" + target return target @@ -340,7 +334,7 @@ def parse_url(url): """ Given a url, return a parsed :class:`.Url` namedtuple. Best-effort is performed to parse incomplete urls. Fields not provided will be None. - This parser is RFC 3986 compliant. + This parser is RFC 3986 and RFC 6874 compliant. The parser logic and helper functions are based heavily on work done in the ``rfc3986`` module. @@ -374,7 +368,9 @@ def parse_url(url): scheme = scheme.lower() if authority: - auth, host, port = SUBAUTHORITY_RE.match(authority).groups() + auth, _, host_port = authority.rpartition("@") + auth = auth or None + host, port = _HOST_PORT_RE.match(host_port).groups() if auth and normalize_uri: auth = _encode_invalid_chars(auth, USERINFO_CHARS) if port == "": diff --git a/lib/urllib3/util/wait.py b/lib/urllib3/util/wait.py index d71d2fd7..21b4590b 100644 --- a/lib/urllib3/util/wait.py +++ b/lib/urllib3/util/wait.py @@ -1,7 +1,7 @@ import errno -from functools import partial import select import sys +from functools import partial try: from time import monotonic @@ -42,7 +42,6 @@ if sys.version_info >= (3, 5): def _retry_on_intr(fn, timeout): return fn(timeout) - else: # Old and broken Pythons. def _retry_on_intr(fn, timeout): @@ -140,14 +139,14 @@ def wait_for_socket(*args, **kwargs): def wait_for_read(sock, timeout=None): - """ Waits for reading to be available on a given socket. + """Waits for reading to be available on a given socket. Returns True if the socket is readable, or False if the timeout expired. """ return wait_for_socket(sock, read=True, timeout=timeout) def wait_for_write(sock, timeout=None): - """ Waits for writing to be available on a given socket. + """Waits for writing to be available on a given socket. Returns True if the socket is readable, or False if the timeout expired. """ return wait_for_socket(sock, write=True, timeout=timeout)