From 07935763df07e266423e9a09e56231587c9c7103 Mon Sep 17 00:00:00 2001 From: JackDandy Date: Sun, 17 Sep 2023 20:52:08 +0100 Subject: [PATCH] =?UTF-8?q?Update=20urllib3=201.26.15=20(25cca389)=20?= =?UTF-8?q?=E2=86=92=202.0.4=20(af7c78fa).?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- CHANGES.md | 1 + lib/urllib3/__init__.py | 90 +- lib/urllib3/_base_connection.py | 173 +++ lib/urllib3/_collections.py | 402 +++--- .../{request.py => _request_methods.py} | 107 +- lib/urllib3/_version.py | 4 +- lib/urllib3/connection.py | 946 ++++++++++----- lib/urllib3/connectionpool.py | 780 ++++++------ lib/urllib3/contrib/__init__.pyi | 0 lib/urllib3/contrib/_appengine_environ.py | 36 - .../contrib/_securetransport/bindings.py | 111 +- .../contrib/_securetransport/low_level.py | 121 +- lib/urllib3/contrib/appengine.py | 314 ----- lib/urllib3/contrib/ntlmpool.py | 130 -- lib/urllib3/contrib/pyopenssl.py | 381 +++--- lib/urllib3/contrib/securetransport.py | 398 +++--- lib/urllib3/contrib/socks.py | 87 +- lib/urllib3/exceptions.py | 189 ++- lib/urllib3/exceptions.pyi | 55 - lib/urllib3/fields.py | 267 ++-- lib/urllib3/fields.pyi | 28 - lib/urllib3/filepost.py | 67 +- lib/urllib3/filepost.pyi | 16 - lib/urllib3/packages/__init__.py | 0 lib/urllib3/packages/__init__.pyi | 0 lib/urllib3/packages/backports/__init__.py | 0 lib/urllib3/packages/backports/makefile.py | 51 - .../packages/backports/weakref_finalize.py | 155 --- lib/urllib3/packages/six.py | 1076 ----------------- lib/urllib3/packages/six.pyi | 0 lib/urllib3/poolmanager.py | 345 ++++-- lib/urllib3/py.typed | 2 + lib/urllib3/response.py | 891 +++++++++----- lib/urllib3/util/__init__.py | 17 +- lib/urllib3/util/connection.py | 80 +- lib/urllib3/util/proxy.py | 38 +- lib/urllib3/util/queue.py | 22 - lib/urllib3/util/queue.pyi | 0 lib/urllib3/util/request.py | 170 ++- lib/urllib3/util/response.py | 78 +- lib/urllib3/util/retry.py | 357 ++---- lib/urllib3/util/ssl_.py | 534 ++++---- lib/urllib3/util/ssl_match_hostname.py | 94 +- lib/urllib3/util/ssltransport.py | 165 ++- lib/urllib3/util/timeout.py | 112 +- lib/urllib3/util/url.py | 396 +++--- lib/urllib3/util/url.pyi | 32 - lib/urllib3/util/util.py | 42 + lib/urllib3/util/wait.py | 88 +- 49 files changed, 4441 insertions(+), 5007 deletions(-) create mode 100644 lib/urllib3/_base_connection.py rename lib/urllib3/{request.py => _request_methods.py} (63%) delete mode 100644 lib/urllib3/contrib/__init__.pyi delete mode 100644 lib/urllib3/contrib/_appengine_environ.py delete mode 100644 lib/urllib3/contrib/appengine.py delete mode 100644 lib/urllib3/contrib/ntlmpool.py delete mode 100644 lib/urllib3/exceptions.pyi delete mode 100644 lib/urllib3/fields.pyi delete mode 100644 lib/urllib3/filepost.pyi delete mode 100644 lib/urllib3/packages/__init__.py delete mode 100644 lib/urllib3/packages/__init__.pyi delete mode 100644 lib/urllib3/packages/backports/__init__.py delete mode 100644 lib/urllib3/packages/backports/makefile.py delete mode 100644 lib/urllib3/packages/backports/weakref_finalize.py delete mode 100644 lib/urllib3/packages/six.py delete mode 100644 lib/urllib3/packages/six.pyi create mode 100644 lib/urllib3/py.typed delete mode 100644 lib/urllib3/util/queue.py delete mode 100644 lib/urllib3/util/queue.pyi delete mode 100644 lib/urllib3/util/url.pyi create mode 100644 lib/urllib3/util/util.py diff --git a/CHANGES.md b/CHANGES.md index 60377395..5822ba22 100644 --- a/CHANGES.md +++ b/CHANGES.md @@ -10,6 +10,7 @@ * Update package resource API 67.5.1 (f51eccd) to 68.1.2 (1ef36f2) * Update soupsieve 2.3.2.post1 (792d566) to 2.4.1 (2e66beb) * Update Tornado Web Server 6.3.2 (e3aa6c5) to 6.3.3 (e4d6984) +* Update urllib3 1.26.15 (25cca389) to 2.0.4 (af7c78fa) * Add thefuzz 0.19.0 (c2cd4f4) as a replacement with fallback to fuzzywuzzy 0.18.0 (2188520) * Fix regex that was not using py312 notation * Change sort backlog and manual segment search results episode number diff --git a/lib/urllib3/__init__.py b/lib/urllib3/__init__.py index c6fa3821..32c1f002 100644 --- a/lib/urllib3/__init__.py +++ b/lib/urllib3/__init__.py @@ -1,23 +1,48 @@ """ Python HTTP library with thread-safe connection pooling, file post support, user friendly, and more """ -from __future__ import absolute_import + +from __future__ import annotations # Set default logging handler to avoid "No handler found" warnings. import logging +import typing import warnings from logging import NullHandler from . import exceptions +from ._base_connection import _TYPE_BODY +from ._collections import HTTPHeaderDict from ._version import __version__ from .connectionpool import HTTPConnectionPool, HTTPSConnectionPool, connection_from_url -from .filepost import encode_multipart_formdata +from .filepost import _TYPE_FIELDS, encode_multipart_formdata from .poolmanager import PoolManager, ProxyManager, proxy_from_url -from .response import HTTPResponse +from .response import BaseHTTPResponse, HTTPResponse from .util.request import make_headers from .util.retry import Retry from .util.timeout import Timeout -from .util.url import get_host + +# Ensure that Python is compiled with OpenSSL 1.1.1+ +# If the 'ssl' module isn't available at all that's +# fine, we only care if the module is available. +try: + import ssl +except ImportError: + pass +else: + if not ssl.OPENSSL_VERSION.startswith("OpenSSL "): # Defensive: + warnings.warn( + "urllib3 v2.0 only supports OpenSSL 1.1.1+, currently " + f"the 'ssl' module is compiled with {ssl.OPENSSL_VERSION!r}. " + "See: https://github.com/urllib3/urllib3/issues/3020", + exceptions.NotOpenSSLWarning, + ) + elif ssl.OPENSSL_VERSION_INFO < (1, 1, 1): # Defensive: + raise ImportError( + "urllib3 v2.0 only supports OpenSSL 1.1.1+, currently " + f"the 'ssl' module is compiled with {ssl.OPENSSL_VERSION!r}. " + "See: https://github.com/urllib3/urllib3/issues/2168" + ) # === NOTE TO REPACKAGERS AND VENDORS === # Please delete this block, this logic is only @@ -25,12 +50,12 @@ from .util.url import get_host # See: https://github.com/urllib3/urllib3/issues/2680 try: import urllib3_secure_extra # type: ignore # noqa: F401 -except ImportError: +except ModuleNotFoundError: 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: " + "in urllib3 v2.1.0. Read more in this issue: " "https://github.com/urllib3/urllib3/issues/2680", category=DeprecationWarning, stacklevel=2, @@ -42,6 +67,7 @@ __version__ = __version__ __all__ = ( "HTTPConnectionPool", + "HTTPHeaderDict", "HTTPSConnectionPool", "PoolManager", "ProxyManager", @@ -52,15 +78,18 @@ __all__ = ( "connection_from_url", "disable_warnings", "encode_multipart_formdata", - "get_host", "make_headers", "proxy_from_url", + "request", + "BaseHTTPResponse", ) logging.getLogger(__name__).addHandler(NullHandler()) -def add_stderr_logger(level=logging.DEBUG): +def add_stderr_logger( + level: int = logging.DEBUG, +) -> logging.StreamHandler[typing.TextIO]: """ Helper for quickly adding a StreamHandler to the logger. Useful for debugging. @@ -87,16 +116,51 @@ del NullHandler # mechanisms to silence them. # SecurityWarning's always go off by default. warnings.simplefilter("always", exceptions.SecurityWarning, append=True) -# SubjectAltNameWarning's should go off once per host -warnings.simplefilter("default", exceptions.SubjectAltNameWarning, append=True) # InsecurePlatformWarning's don't vary between requests, so we keep it default. warnings.simplefilter("default", exceptions.InsecurePlatformWarning, append=True) -# SNIMissingWarnings should go off only once. -warnings.simplefilter("default", exceptions.SNIMissingWarning, append=True) -def disable_warnings(category=exceptions.HTTPWarning): +def disable_warnings(category: type[Warning] = exceptions.HTTPWarning) -> None: """ Helper for quickly disabling all urllib3 warnings. """ warnings.simplefilter("ignore", category) + + +_DEFAULT_POOL = PoolManager() + + +def request( + method: str, + url: str, + *, + body: _TYPE_BODY | None = None, + fields: _TYPE_FIELDS | None = None, + headers: typing.Mapping[str, str] | None = None, + preload_content: bool | None = True, + decode_content: bool | None = True, + redirect: bool | None = True, + retries: Retry | bool | int | None = None, + timeout: Timeout | float | int | None = 3, + json: typing.Any | None = None, +) -> BaseHTTPResponse: + """ + A convenience, top-level request method. It uses a module-global ``PoolManager`` instance. + Therefore, its side effects could be shared across dependencies relying on it. + To avoid side effects create a new ``PoolManager`` instance and use it instead. + The method does not accept low-level ``**urlopen_kw`` keyword arguments. + """ + + return _DEFAULT_POOL.request( + method, + url, + body=body, + fields=fields, + headers=headers, + preload_content=preload_content, + decode_content=decode_content, + redirect=redirect, + retries=retries, + timeout=timeout, + json=json, + ) diff --git a/lib/urllib3/_base_connection.py b/lib/urllib3/_base_connection.py new file mode 100644 index 00000000..25b633af --- /dev/null +++ b/lib/urllib3/_base_connection.py @@ -0,0 +1,173 @@ +from __future__ import annotations + +import typing + +from .util.connection import _TYPE_SOCKET_OPTIONS +from .util.timeout import _DEFAULT_TIMEOUT, _TYPE_TIMEOUT +from .util.url import Url + +_TYPE_BODY = typing.Union[bytes, typing.IO[typing.Any], typing.Iterable[bytes], str] + + +class ProxyConfig(typing.NamedTuple): + ssl_context: ssl.SSLContext | None + use_forwarding_for_https: bool + assert_hostname: None | str | Literal[False] + assert_fingerprint: str | None + + +class _ResponseOptions(typing.NamedTuple): + # TODO: Remove this in favor of a better + # HTTP request/response lifecycle tracking. + request_method: str + request_url: str + preload_content: bool + decode_content: bool + enforce_content_length: bool + + +if typing.TYPE_CHECKING: + import ssl + + from typing_extensions import Literal, Protocol + + from .response import BaseHTTPResponse + + class BaseHTTPConnection(Protocol): + default_port: typing.ClassVar[int] + default_socket_options: typing.ClassVar[_TYPE_SOCKET_OPTIONS] + + host: str + port: int + timeout: None | ( + float + ) # Instance doesn't store _DEFAULT_TIMEOUT, must be resolved. + blocksize: int + source_address: tuple[str, int] | None + socket_options: _TYPE_SOCKET_OPTIONS | None + + proxy: Url | None + proxy_config: ProxyConfig | None + + is_verified: bool + proxy_is_verified: bool | None + + def __init__( + self, + host: str, + port: int | None = None, + *, + timeout: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT, + source_address: tuple[str, int] | None = None, + blocksize: int = 8192, + socket_options: _TYPE_SOCKET_OPTIONS | None = ..., + proxy: Url | None = None, + proxy_config: ProxyConfig | None = None, + ) -> None: + ... + + def set_tunnel( + self, + host: str, + port: int | None = None, + headers: typing.Mapping[str, str] | None = None, + scheme: str = "http", + ) -> None: + ... + + def connect(self) -> None: + ... + + def request( + self, + method: str, + url: str, + body: _TYPE_BODY | None = None, + headers: typing.Mapping[str, str] | None = None, + # We know *at least* botocore is depending on the order of the + # first 3 parameters so to be safe we only mark the later ones + # as keyword-only to ensure we have space to extend. + *, + chunked: bool = False, + preload_content: bool = True, + decode_content: bool = True, + enforce_content_length: bool = True, + ) -> None: + ... + + def getresponse(self) -> BaseHTTPResponse: + ... + + def close(self) -> None: + ... + + @property + def is_closed(self) -> bool: + """Whether the connection either is brand new or has been previously closed. + If this property is True then both ``is_connected`` and ``has_connected_to_proxy`` + properties must be False. + """ + + @property + def is_connected(self) -> bool: + """Whether the connection is actively connected to any origin (proxy or target)""" + + @property + def has_connected_to_proxy(self) -> bool: + """Whether the connection has successfully connected to its proxy. + This returns False if no proxy is in use. Used to determine whether + errors are coming from the proxy layer or from tunnelling to the target origin. + """ + + class BaseHTTPSConnection(BaseHTTPConnection, Protocol): + default_port: typing.ClassVar[int] + default_socket_options: typing.ClassVar[_TYPE_SOCKET_OPTIONS] + + # Certificate verification methods + cert_reqs: int | str | None + assert_hostname: None | str | Literal[False] + assert_fingerprint: str | None + ssl_context: ssl.SSLContext | None + + # Trusted CAs + ca_certs: str | None + ca_cert_dir: str | None + ca_cert_data: None | str | bytes + + # TLS version + ssl_minimum_version: int | None + ssl_maximum_version: int | None + ssl_version: int | str | None # Deprecated + + # Client certificates + cert_file: str | None + key_file: str | None + key_password: str | None + + def __init__( + self, + host: str, + port: int | None = None, + *, + timeout: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT, + source_address: tuple[str, int] | None = None, + blocksize: int = 16384, + socket_options: _TYPE_SOCKET_OPTIONS | None = ..., + proxy: Url | None = None, + proxy_config: ProxyConfig | None = None, + cert_reqs: int | str | None = None, + assert_hostname: None | str | Literal[False] = None, + assert_fingerprint: str | None = None, + server_hostname: str | None = None, + ssl_context: ssl.SSLContext | None = None, + ca_certs: str | None = None, + ca_cert_dir: str | None = None, + ca_cert_data: None | str | bytes = None, + ssl_minimum_version: int | None = None, + ssl_maximum_version: int | None = None, + ssl_version: int | str | None = None, # Deprecated + cert_file: str | None = None, + key_file: str | None = None, + key_password: str | None = None, + ) -> None: + ... diff --git a/lib/urllib3/_collections.py b/lib/urllib3/_collections.py index da9857e9..7f9dca7f 100644 --- a/lib/urllib3/_collections.py +++ b/lib/urllib3/_collections.py @@ -1,34 +1,66 @@ -from __future__ import absolute_import - -try: - from collections.abc import Mapping, MutableMapping -except ImportError: - from collections import Mapping, MutableMapping -try: - from threading import RLock -except ImportError: # Platform-specific: No threads available - - class RLock: - def __enter__(self): - pass - - def __exit__(self, exc_type, exc_value, traceback): - pass - +from __future__ import annotations +import typing from collections import OrderedDict +from enum import Enum, auto +from threading import RLock + +if typing.TYPE_CHECKING: + # We can only import Protocol if TYPE_CHECKING because it's a development + # dependency, and is not available at runtime. + from typing_extensions import Protocol + + class HasGettableStringKeys(Protocol): + def keys(self) -> typing.Iterator[str]: + ... + + def __getitem__(self, key: str) -> str: + ... -from .exceptions import InvalidHeader -from .packages import six -from .packages.six import iterkeys, itervalues __all__ = ["RecentlyUsedContainer", "HTTPHeaderDict"] -_Null = object() +# Key type +_KT = typing.TypeVar("_KT") +# Value type +_VT = typing.TypeVar("_VT") +# Default type +_DT = typing.TypeVar("_DT") + +ValidHTTPHeaderSource = typing.Union[ + "HTTPHeaderDict", + typing.Mapping[str, str], + typing.Iterable[typing.Tuple[str, str]], + "HasGettableStringKeys", +] -class RecentlyUsedContainer(MutableMapping): +class _Sentinel(Enum): + not_passed = auto() + + +def ensure_can_construct_http_header_dict( + potential: object, +) -> ValidHTTPHeaderSource | None: + if isinstance(potential, HTTPHeaderDict): + return potential + elif isinstance(potential, typing.Mapping): + # Full runtime checking of the contents of a Mapping is expensive, so for the + # purposes of typechecking, we assume that any Mapping is the right shape. + return typing.cast(typing.Mapping[str, str], potential) + elif isinstance(potential, typing.Iterable): + # Similarly to Mapping, full runtime checking of the contents of an Iterable is + # expensive, so for the purposes of typechecking, we assume that any Iterable + # is the right shape. + return typing.cast(typing.Iterable[typing.Tuple[str, str]], potential) + elif hasattr(potential, "keys") and hasattr(potential, "__getitem__"): + return typing.cast("HasGettableStringKeys", potential) + else: + return None + + +class RecentlyUsedContainer(typing.Generic[_KT, _VT], typing.MutableMapping[_KT, _VT]): """ Provides a thread-safe dict-like container which maintains up to ``maxsize`` keys while throwing away the least-recently-used keys beyond @@ -42,69 +74,134 @@ class RecentlyUsedContainer(MutableMapping): ``dispose_func(value)`` is called. Callback which will get called """ - ContainerCls = OrderedDict + _container: typing.OrderedDict[_KT, _VT] + _maxsize: int + dispose_func: typing.Callable[[_VT], None] | None + lock: RLock - def __init__(self, maxsize=10, dispose_func=None): + def __init__( + self, + maxsize: int = 10, + dispose_func: typing.Callable[[_VT], None] | None = None, + ) -> None: + super().__init__() self._maxsize = maxsize self.dispose_func = dispose_func - - self._container = self.ContainerCls() + self._container = OrderedDict() self.lock = RLock() - def __getitem__(self, key): + def __getitem__(self, key: _KT) -> _VT: # Re-insert the item, moving it to the end of the eviction line. with self.lock: item = self._container.pop(key) self._container[key] = item return item - def __setitem__(self, key, value): - evicted_value = _Null + def __setitem__(self, key: _KT, value: _VT) -> None: + evicted_item = None with self.lock: # Possibly evict the existing value of 'key' - evicted_value = self._container.get(key, _Null) - self._container[key] = value + try: + # If the key exists, we'll overwrite it, which won't change the + # size of the pool. Because accessing a key should move it to + # the end of the eviction line, we pop it out first. + evicted_item = key, self._container.pop(key) + self._container[key] = value + except KeyError: + # When the key does not exist, we insert the value first so that + # evicting works in all cases, including when self._maxsize is 0 + self._container[key] = value + if len(self._container) > self._maxsize: + # If we didn't evict an existing value, and we've hit our maximum + # size, then we have to evict the least recently used item from + # the beginning of the container. + evicted_item = self._container.popitem(last=False) - # If we didn't evict an existing value, we might have to evict the - # least recently used item from the beginning of the container. - if len(self._container) > self._maxsize: - _key, evicted_value = self._container.popitem(last=False) - - if self.dispose_func and evicted_value is not _Null: + # After releasing the lock on the pool, dispose of any evicted value. + if evicted_item is not None and self.dispose_func: + _, evicted_value = evicted_item self.dispose_func(evicted_value) - def __delitem__(self, key): + def __delitem__(self, key: _KT) -> None: with self.lock: value = self._container.pop(key) if self.dispose_func: self.dispose_func(value) - def __len__(self): + def __len__(self) -> int: with self.lock: return len(self._container) - def __iter__(self): + def __iter__(self) -> typing.NoReturn: raise NotImplementedError( "Iteration over this class is unlikely to be threadsafe." ) - def clear(self): + def clear(self) -> None: with self.lock: # Copy pointers to all values, then wipe the mapping - values = list(itervalues(self._container)) + values = list(self._container.values()) self._container.clear() if self.dispose_func: for value in values: self.dispose_func(value) - def keys(self): + def keys(self) -> set[_KT]: # type: ignore[override] with self.lock: - return list(iterkeys(self._container)) + return set(self._container.keys()) -class HTTPHeaderDict(MutableMapping): +class HTTPHeaderDictItemView(typing.Set[typing.Tuple[str, str]]): + """ + HTTPHeaderDict is unusual for a Mapping[str, str] in that it has two modes of + address. + + If we directly try to get an item with a particular name, we will get a string + back that is the concatenated version of all the values: + + >>> d['X-Header-Name'] + 'Value1, Value2, Value3' + + However, if we iterate over an HTTPHeaderDict's items, we will optionally combine + these values based on whether combine=True was called when building up the dictionary + + >>> d = HTTPHeaderDict({"A": "1", "B": "foo"}) + >>> d.add("A", "2", combine=True) + >>> d.add("B", "bar") + >>> list(d.items()) + [ + ('A', '1, 2'), + ('B', 'foo'), + ('B', 'bar'), + ] + + This class conforms to the interface required by the MutableMapping ABC while + also giving us the nonstandard iteration behavior we want; items with duplicate + keys, ordered by time of first insertion. + """ + + _headers: HTTPHeaderDict + + def __init__(self, headers: HTTPHeaderDict) -> None: + self._headers = headers + + def __len__(self) -> int: + return len(list(self._headers.iteritems())) + + def __iter__(self) -> typing.Iterator[tuple[str, str]]: + return self._headers.iteritems() + + def __contains__(self, item: object) -> bool: + if isinstance(item, tuple) and len(item) == 2: + passed_key, passed_val = item + if isinstance(passed_key, str) and isinstance(passed_val, str): + return self._headers._has_value_for_header(passed_key, passed_val) + return False + + +class HTTPHeaderDict(typing.MutableMapping[str, str]): """ :param headers: An iterable of field-value pairs. Must not contain multiple field names @@ -138,9 +235,11 @@ class HTTPHeaderDict(MutableMapping): '7' """ - def __init__(self, headers=None, **kwargs): - super(HTTPHeaderDict, self).__init__() - self._container = OrderedDict() + _container: typing.MutableMapping[str, list[str]] + + def __init__(self, headers: ValidHTTPHeaderSource | None = None, **kwargs: str): + super().__init__() + self._container = {} # 'dict' is insert-ordered in Python 3.7+ if headers is not None: if isinstance(headers, HTTPHeaderDict): self._copy_from(headers) @@ -149,123 +248,147 @@ class HTTPHeaderDict(MutableMapping): if kwargs: self.extend(kwargs) - def __setitem__(self, key, val): + def __setitem__(self, key: str, val: str) -> None: + # avoid a bytes/str comparison by decoding before httplib + if isinstance(key, bytes): + key = key.decode("latin-1") self._container[key.lower()] = [key, val] - return self._container[key.lower()] - def __getitem__(self, key): + def __getitem__(self, key: str) -> str: val = self._container[key.lower()] return ", ".join(val[1:]) - def __delitem__(self, key): + def __delitem__(self, key: str) -> None: del self._container[key.lower()] - def __contains__(self, key): - return key.lower() in self._container + def __contains__(self, key: object) -> bool: + if isinstance(key, str): + return key.lower() in self._container + return False - def __eq__(self, other): - if not isinstance(other, Mapping) and not hasattr(other, "keys"): + def setdefault(self, key: str, default: str = "") -> str: + return super().setdefault(key, default) + + def __eq__(self, other: object) -> bool: + maybe_constructable = ensure_can_construct_http_header_dict(other) + if maybe_constructable is None: return False - if not isinstance(other, type(self)): - other = type(self)(other) - return dict((k.lower(), v) for k, v in self.itermerged()) == dict( - (k.lower(), v) for k, v in other.itermerged() - ) + else: + other_as_http_header_dict = type(self)(maybe_constructable) - def __ne__(self, other): + return {k.lower(): v for k, v in self.itermerged()} == { + k.lower(): v for k, v in other_as_http_header_dict.itermerged() + } + + def __ne__(self, other: object) -> bool: return not self.__eq__(other) - if six.PY2: # Python 2 - iterkeys = MutableMapping.iterkeys - itervalues = MutableMapping.itervalues - - __marker = object() - - def __len__(self): + def __len__(self) -> int: return len(self._container) - def __iter__(self): + def __iter__(self) -> typing.Iterator[str]: # Only provide the originally cased names for vals in self._container.values(): yield vals[0] - 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. - """ - # Using the MutableMapping function directly fails due to the private marker. - # Using ordinary dict.pop would expose the internal structures. - # So let's reinvent the wheel. - try: - value = self[key] - except KeyError: - if default is self.__marker: - raise - return default - else: - del self[key] - return value - - def discard(self, key): + def discard(self, key: str) -> None: try: del self[key] except KeyError: pass - def add(self, key, val): + def add(self, key: str, val: str, *, combine: bool = False) -> None: """Adds a (name, value) pair, doesn't overwrite the value if it already exists. + If this is called with combine=True, instead of adding a new header value + as a distinct item during iteration, this will instead append the value to + any existing header value with a comma. If no existing header value exists + for the key, then the value will simply be added, ignoring the combine parameter. + >>> headers = HTTPHeaderDict(foo='bar') >>> headers.add('Foo', 'baz') >>> headers['foo'] 'bar, baz' + >>> list(headers.items()) + [('foo', 'bar'), ('foo', 'baz')] + >>> headers.add('foo', 'quz', combine=True) + >>> list(headers.items()) + [('foo', 'bar, baz, quz')] """ + # avoid a bytes/str comparison by decoding before httplib + if isinstance(key, bytes): + key = key.decode("latin-1") key_lower = key.lower() new_vals = [key, val] # Keep the common case aka no item present as fast as possible vals = self._container.setdefault(key_lower, new_vals) if new_vals is not vals: - vals.append(val) + # if there are values here, then there is at least the initial + # key/value pair + assert len(vals) >= 2 + if combine: + vals[-1] = vals[-1] + ", " + val + else: + vals.append(val) - def extend(self, *args, **kwargs): + def extend(self, *args: ValidHTTPHeaderSource, **kwargs: str) -> None: """Generic import function for any type of header-like object. Adapted version of MutableMapping.update in order to insert items with self.add instead of self.__setitem__ """ if len(args) > 1: raise TypeError( - "extend() takes at most 1 positional " - "arguments ({0} given)".format(len(args)) + f"extend() takes at most 1 positional arguments ({len(args)} given)" ) other = args[0] if len(args) >= 1 else () if isinstance(other, HTTPHeaderDict): for key, val in other.iteritems(): self.add(key, val) - elif isinstance(other, Mapping): - for key in other: - self.add(key, other[key]) - elif hasattr(other, "keys"): - for key in other.keys(): - self.add(key, other[key]) - else: + elif isinstance(other, typing.Mapping): + for key, val in other.items(): + self.add(key, val) + elif isinstance(other, typing.Iterable): + other = typing.cast(typing.Iterable[typing.Tuple[str, str]], other) for key, value in other: self.add(key, value) + elif hasattr(other, "keys") and hasattr(other, "__getitem__"): + # THIS IS NOT A TYPESAFE BRANCH + # In this branch, the object has a `keys` attr but is not a Mapping or any of + # the other types indicated in the method signature. We do some stuff with + # it as though it partially implements the Mapping interface, but we're not + # doing that stuff safely AT ALL. + for key in other.keys(): + self.add(key, other[key]) for key, value in kwargs.items(): self.add(key, value) - def getlist(self, key, default=__marker): + @typing.overload + def getlist(self, key: str) -> list[str]: + ... + + @typing.overload + def getlist(self, key: str, default: _DT) -> list[str] | _DT: + ... + + def getlist( + self, key: str, default: _Sentinel | _DT = _Sentinel.not_passed + ) -> list[str] | _DT: """Returns a list of all the values for the named field. Returns an empty list if the key doesn't exist.""" try: vals = self._container[key.lower()] except KeyError: - if default is self.__marker: + if default is _Sentinel.not_passed: + # _DT is unbound; empty list is instance of List[str] return [] + # _DT is bound; default is instance of _DT return default else: + # _DT may or may not be bound; vals[1:] is instance of List[str], which + # meets our external interface requirement of `Union[List[str], _DT]`. return vals[1:] # Backwards compatibility for httplib @@ -276,62 +399,65 @@ class HTTPHeaderDict(MutableMapping): # Backwards compatibility for http.cookiejar get_all = getlist - def __repr__(self): - return "%s(%s)" % (type(self).__name__, dict(self.itermerged())) + def __repr__(self) -> str: + return f"{type(self).__name__}({dict(self.itermerged())})" - def _copy_from(self, other): + def _copy_from(self, other: HTTPHeaderDict) -> None: for key in other: val = other.getlist(key) - if isinstance(val, list): - # Don't need to convert tuples - val = list(val) - self._container[key.lower()] = [key] + val + self._container[key.lower()] = [key, *val] - def copy(self): + def copy(self) -> HTTPHeaderDict: clone = type(self)() clone._copy_from(self) return clone - def iteritems(self): + def iteritems(self) -> typing.Iterator[tuple[str, str]]: """Iterate over all header lines, including duplicate ones.""" for key in self: vals = self._container[key.lower()] for val in vals[1:]: yield vals[0], val - def itermerged(self): + def itermerged(self) -> typing.Iterator[tuple[str, str]]: """Iterate over all headers, merging duplicate ones together.""" for key in self: val = self._container[key.lower()] yield val[0], ", ".join(val[1:]) - def items(self): - return list(self.iteritems()) + def items(self) -> HTTPHeaderDictItemView: # type: ignore[override] + return HTTPHeaderDictItemView(self) - @classmethod - def from_httplib(cls, message): # Python 2 - """Read headers from a Python 2 httplib message object.""" - # python2.7 does not expose a proper API for exporting multiheaders - # efficiently. This function re-reads raw lines from the message - # object and extracts the multiheaders properly. - obs_fold_continued_leaders = (" ", "\t") - headers = [] + def _has_value_for_header(self, header_name: str, potential_value: str) -> bool: + if header_name in self: + return potential_value in self._container[header_name.lower()][1:] + return False - for line in message.headers: - if line.startswith(obs_fold_continued_leaders): - if not headers: - # We received a header line that starts with OWS as described - # in RFC-7230 S3.2.4. This indicates a multiline header, but - # there exists no previous header to which we can attach it. - raise InvalidHeader( - "Header continuation with no previous header: %s" % line - ) - else: - key, value = headers[-1] - headers[-1] = (key, value + " " + line.strip()) - continue + def __ior__(self, other: object) -> HTTPHeaderDict: + # Supports extending a header dict in-place using operator |= + # combining items with add instead of __setitem__ + maybe_constructable = ensure_can_construct_http_header_dict(other) + if maybe_constructable is None: + return NotImplemented + self.extend(maybe_constructable) + return self - key, value = line.split(":", 1) - headers.append((key, value.strip())) + def __or__(self, other: object) -> HTTPHeaderDict: + # Supports merging header dicts using operator | + # combining items with add instead of __setitem__ + maybe_constructable = ensure_can_construct_http_header_dict(other) + if maybe_constructable is None: + return NotImplemented + result = self.copy() + result.extend(maybe_constructable) + return result - return cls(headers) + def __ror__(self, other: object) -> HTTPHeaderDict: + # Supports merging header dicts using operator | when other is on left side + # combining items with add instead of __setitem__ + maybe_constructable = ensure_can_construct_http_header_dict(other) + if maybe_constructable is None: + return NotImplemented + result = type(self)(maybe_constructable) + result.extend(self) + return result diff --git a/lib/urllib3/request.py b/lib/urllib3/_request_methods.py similarity index 63% rename from lib/urllib3/request.py rename to lib/urllib3/_request_methods.py index 398386a5..1d0f3465 100644 --- a/lib/urllib3/request.py +++ b/lib/urllib3/_request_methods.py @@ -1,12 +1,23 @@ -from __future__ import absolute_import +from __future__ import annotations -from .filepost import encode_multipart_formdata -from .packages.six.moves.urllib.parse import urlencode +import json as _json +import typing +from urllib.parse import urlencode + +from ._base_connection import _TYPE_BODY +from ._collections import HTTPHeaderDict +from .filepost import _TYPE_FIELDS, encode_multipart_formdata +from .response import BaseHTTPResponse __all__ = ["RequestMethods"] +_TYPE_ENCODE_URL_FIELDS = typing.Union[ + typing.Sequence[typing.Tuple[str, typing.Union[str, bytes]]], + typing.Mapping[str, typing.Union[str, bytes]], +] -class RequestMethods(object): + +class RequestMethods: """ Convenience mixin for classes who implement a :meth:`urlopen` method, such as :class:`urllib3.HTTPConnectionPool` and @@ -37,25 +48,34 @@ class RequestMethods(object): _encode_url_methods = {"DELETE", "GET", "HEAD", "OPTIONS"} - def __init__(self, headers=None): + def __init__(self, headers: typing.Mapping[str, str] | None = None) -> None: self.headers = headers or {} def urlopen( self, - method, - url, - body=None, - headers=None, - encode_multipart=True, - multipart_boundary=None, - **kw - ): # Abstract + method: str, + url: str, + body: _TYPE_BODY | None = None, + headers: typing.Mapping[str, str] | None = None, + encode_multipart: bool = True, + multipart_boundary: str | None = None, + **kw: typing.Any, + ) -> BaseHTTPResponse: # Abstract raise NotImplementedError( "Classes extending RequestMethods must implement " "their own ``urlopen`` method." ) - def request(self, method, url, fields=None, headers=None, **urlopen_kw): + def request( + self, + method: str, + url: str, + body: _TYPE_BODY | None = None, + fields: _TYPE_FIELDS | None = None, + headers: typing.Mapping[str, str] | None = None, + json: typing.Any | None = None, + **urlopen_kw: typing.Any, + ) -> BaseHTTPResponse: """ Make a request using :meth:`urlopen` with the appropriate encoding of ``fields`` based on the ``method`` used. @@ -68,18 +88,45 @@ class RequestMethods(object): """ method = method.upper() - urlopen_kw["request_url"] = url + if json is not None and body is not None: + raise TypeError( + "request got values for both 'body' and 'json' parameters which are mutually exclusive" + ) + + if json is not None: + if headers is None: + headers = self.headers.copy() # type: ignore + if not ("content-type" in map(str.lower, headers.keys())): + headers["Content-Type"] = "application/json" # type: ignore + + body = _json.dumps(json, separators=(",", ":"), ensure_ascii=False).encode( + "utf-8" + ) + + if body is not None: + urlopen_kw["body"] = body if method in self._encode_url_methods: return self.request_encode_url( - method, url, fields=fields, headers=headers, **urlopen_kw + method, + url, + fields=fields, # type: ignore[arg-type] + headers=headers, + **urlopen_kw, ) else: return self.request_encode_body( method, url, fields=fields, headers=headers, **urlopen_kw ) - def request_encode_url(self, method, url, fields=None, headers=None, **urlopen_kw): + def request_encode_url( + self, + method: str, + url: str, + fields: _TYPE_ENCODE_URL_FIELDS | None = None, + headers: typing.Mapping[str, str] | None = None, + **urlopen_kw: str, + ) -> BaseHTTPResponse: """ Make a request using :meth:`urlopen` with the ``fields`` encoded in the url. This is useful for request methods like GET, HEAD, DELETE, etc. @@ -87,7 +134,7 @@ class RequestMethods(object): if headers is None: headers = self.headers - extra_kw = {"headers": headers} + extra_kw: dict[str, typing.Any] = {"headers": headers} extra_kw.update(urlopen_kw) if fields: @@ -97,14 +144,14 @@ class RequestMethods(object): def request_encode_body( self, - method, - url, - fields=None, - headers=None, - encode_multipart=True, - multipart_boundary=None, - **urlopen_kw - ): + method: str, + url: str, + fields: _TYPE_FIELDS | None = None, + headers: typing.Mapping[str, str] | None = None, + encode_multipart: bool = True, + multipart_boundary: str | None = None, + **urlopen_kw: str, + ) -> BaseHTTPResponse: """ Make a request using :meth:`urlopen` with the ``fields`` encoded in the body. This is useful for request methods like POST, PUT, PATCH, etc. @@ -143,7 +190,8 @@ class RequestMethods(object): if headers is None: headers = self.headers - extra_kw = {"headers": {}} + extra_kw: dict[str, typing.Any] = {"headers": HTTPHeaderDict(headers)} + body: bytes | str if fields: if "body" in urlopen_kw: @@ -157,14 +205,13 @@ class RequestMethods(object): ) else: body, content_type = ( - urlencode(fields), + urlencode(fields), # type: ignore[arg-type] "application/x-www-form-urlencoded", ) extra_kw["body"] = body - extra_kw["headers"] = {"Content-Type": content_type} + extra_kw["headers"].setdefault("Content-Type", content_type) - extra_kw["headers"].update(headers) extra_kw.update(urlopen_kw) return self.urlopen(method, url, **extra_kw) diff --git a/lib/urllib3/_version.py b/lib/urllib3/_version.py index e12dd0e7..782af433 100644 --- a/lib/urllib3/_version.py +++ b/lib/urllib3/_version.py @@ -1,2 +1,4 @@ # This file is protected via CODEOWNERS -__version__ = "1.26.15" +from __future__ import annotations + +__version__ = "2.0.4" diff --git a/lib/urllib3/connection.py b/lib/urllib3/connection.py index 54b96b19..4a71225c 100644 --- a/lib/urllib3/connection.py +++ b/lib/urllib3/connection.py @@ -1,59 +1,58 @@ -from __future__ import absolute_import +from __future__ import annotations import datetime import logging import os import re import socket +import sys +import typing import warnings -from socket import error as SocketError +from http.client import HTTPConnection as _HTTPConnection +from http.client import HTTPException as HTTPException # noqa: F401 +from http.client import ResponseNotReady 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 +if typing.TYPE_CHECKING: + from typing_extensions import Literal + + from .response import HTTPResponse + from .util.ssl_ import _TYPE_PEER_CERT_RET_DICT + from .util.ssltransport import SSLTransport + +from ._collections import HTTPHeaderDict +from .util.response import assert_header_parsing +from .util.timeout import _DEFAULT_TIMEOUT, _TYPE_TIMEOUT, Timeout +from .util.util import to_str +from .util.wait import wait_for_read try: # Compiled with SSL? import ssl BaseSSLError = ssl.SSLError -except (ImportError, AttributeError): # Platform-specific: No SSL. - ssl = None +except (ImportError, AttributeError): + ssl = None # type: ignore[assignment] - class BaseSSLError(BaseException): + class BaseSSLError(BaseException): # type: ignore[no-redef] pass -try: - # Python 3: not a no-op, we're adding this to the namespace so it can be imported. - ConnectionError = ConnectionError -except NameError: - # Python 2 - class ConnectionError(Exception): - 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 ._base_connection import _TYPE_BODY +from ._base_connection import ProxyConfig as ProxyConfig +from ._base_connection import _ResponseOptions as _ResponseOptions from ._version import __version__ from .exceptions import ( ConnectTimeoutError, + HeaderParsingError, + NameResolutionError, NewConnectionError, - SubjectAltNameWarning, + ProxyError, SystemTimeWarning, ) -from .util import SKIP_HEADER, SKIPPABLE_HEADERS, connection +from .util import SKIP_HEADER, SKIPPABLE_HEADERS, connection, ssl_ +from .util.request import body_to_chunks +from .util.ssl_ import assert_fingerprint as _assert_fingerprint from .util.ssl_ import ( - assert_fingerprint, create_urllib3_context, is_ipaddress, resolve_cert_reqs, @@ -61,6 +60,12 @@ from .util.ssl_ import ( ssl_wrap_socket, ) from .util.ssl_match_hostname import CertificateError, match_hostname +from .util.url import Url + +# Not a no-op, we're adding this to the namespace so it can be imported. +ConnectionError = ConnectionError +BrokenPipeError = BrokenPipeError + log = logging.getLogger(__name__) @@ -72,8 +77,10 @@ RECENT_DATE = datetime.date(2022, 1, 1) _CONTAINS_CONTROL_CHAR_RE = re.compile(r"[^-!#$%&'*+.^_`|~0-9a-zA-Z]") +_HAS_SYS_AUDIT = hasattr(sys, "audit") -class HTTPConnection(_HTTPConnection, object): + +class HTTPConnection(_HTTPConnection): """ Based on :class:`http.client.HTTPConnection` but provides an extra constructor backwards-compatibility layer between older and newer Pythons. @@ -81,7 +88,6 @@ class HTTPConnection(_HTTPConnection, object): 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 @@ -99,38 +105,68 @@ class HTTPConnection(_HTTPConnection, object): Or you may want to disable the defaults by passing an empty list (e.g., ``[]``). """ - default_port = port_by_scheme["http"] + default_port: typing.ClassVar[int] = port_by_scheme["http"] # type: ignore[misc] #: Disable Nagle's algorithm by default. #: ``[(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)]`` - default_socket_options = [(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)] + default_socket_options: typing.ClassVar[connection._TYPE_SOCKET_OPTIONS] = [ + (socket.IPPROTO_TCP, socket.TCP_NODELAY, 1) + ] #: Whether this connection verifies the host's certificate. - is_verified = False + is_verified: bool = False - #: Whether this proxy connection (if used) verifies the proxy host's - #: certificate. - proxy_is_verified = None + #: Whether this proxy connection verified the proxy host's certificate. + # If no proxy is currently connected to the value will be ``None``. + proxy_is_verified: bool | None = None - def __init__(self, *args, **kw): - if not six.PY2: - kw.pop("strict", None) + blocksize: int + source_address: tuple[str, int] | None + socket_options: connection._TYPE_SOCKET_OPTIONS | None - # Pre-set source_address. - self.source_address = kw.get("source_address") + _has_connected_to_proxy: bool + _response_options: _ResponseOptions | None + _tunnel_host: str | None + _tunnel_port: int | None + _tunnel_scheme: str | None - #: The socket options provided by the user. If no options are - #: provided, we use the default options. - self.socket_options = kw.pop("socket_options", self.default_socket_options) + def __init__( + self, + host: str, + port: int | None = None, + *, + timeout: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT, + source_address: tuple[str, int] | None = None, + blocksize: int = 16384, + socket_options: None + | (connection._TYPE_SOCKET_OPTIONS) = default_socket_options, + proxy: Url | None = None, + proxy_config: ProxyConfig | None = None, + ) -> None: + super().__init__( + host=host, + port=port, + timeout=Timeout.resolve_default_timeout(timeout), + source_address=source_address, + blocksize=blocksize, + ) + self.socket_options = socket_options + self.proxy = proxy + self.proxy_config = proxy_config - # 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) + self._has_connected_to_proxy = False + self._response_options = None + self._tunnel_host: str | None = None + self._tunnel_port: int | None = None + self._tunnel_scheme: str | None = None + # https://github.com/python/mypy/issues/4125 + # Mypy treats this as LSP violation, which is considered a bug. + # If `host` is made a property it violates LSP, because a writeable attribute is overridden with a read-only one. + # However, there is also a `host` setter so LSP is not violated. + # Potentially, a `@host.deleter` might be needed depending on how this issue will be fixed. @property - def host(self): + def host(self) -> str: """ Getter method to remove any trailing dots that indicate the hostname is an FQDN. @@ -149,7 +185,7 @@ class HTTPConnection(_HTTPConnection, object): return self._dns_host.rstrip(".") @host.setter - def host(self, value): + def host(self, value: str) -> None: """ Setter for the `host` property. @@ -158,129 +194,298 @@ class HTTPConnection(_HTTPConnection, object): """ self._dns_host = value - def _new_conn(self): + def _new_conn(self) -> socket.socket: """Establish a socket connection and set nodelay settings on it. :return: New socket connection. """ - extra_kw = {} - if self.source_address: - extra_kw["source_address"] = self.source_address - - if self.socket_options: - extra_kw["socket_options"] = self.socket_options - try: - conn = connection.create_connection( - (self._dns_host, self.port), self.timeout, **extra_kw + sock = connection.create_connection( + (self._dns_host, self.port), + self.timeout, + source_address=self.source_address, + socket_options=self.socket_options, ) - - except SocketTimeout: + except socket.gaierror as e: + raise NameResolutionError(self.host, self, e) from e + except SocketTimeout as e: raise ConnectTimeoutError( self, - "Connection to %s timed out. (connect timeout=%s)" - % (self.host, self.timeout), - ) + f"Connection to {self.host} timed out. (connect timeout={self.timeout})", + ) from e - except SocketError as e: + except OSError as e: raise NewConnectionError( - self, "Failed to establish a new connection: %s" % e + self, f"Failed to establish a new connection: {e}" + ) from e + + # Audit hooks are only available in Python 3.8+ + if _HAS_SYS_AUDIT: + sys.audit("http.client.connect", self, self.host, self.port) + + return sock + + def set_tunnel( + self, + host: str, + port: int | None = None, + headers: typing.Mapping[str, str] | None = None, + scheme: str = "http", + ) -> None: + if scheme not in ("http", "https"): + raise ValueError( + f"Invalid proxy scheme for tunneling: {scheme!r}, must be either 'http' or 'https'" ) + super().set_tunnel(host, port=port, headers=headers) + self._tunnel_scheme = scheme - return conn + def connect(self) -> None: + self.sock = self._new_conn() + if self._tunnel_host: + # If we're tunneling it means we're connected to our proxy. + self._has_connected_to_proxy = True - 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 - 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 - self.auto_open = 0 + self._tunnel() # type: ignore[attr-defined] - def connect(self): - conn = self._new_conn() - self._prepare_conn(conn) + # If there's a proxy to be connected to we are fully connected. + # This is set twice (once above and here) due to forwarding proxies + # not using tunnelling. + self._has_connected_to_proxy = bool(self.proxy) - def putrequest(self, method, url, *args, **kwargs): - """ """ + @property + def is_closed(self) -> bool: + return self.sock is None + + @property + def is_connected(self) -> bool: + if self.sock is None: + return False + return not wait_for_read(self.sock, timeout=0.0) + + @property + def has_connected_to_proxy(self) -> bool: + return self._has_connected_to_proxy + + def close(self) -> None: + try: + super().close() + finally: + # Reset all stateful properties so connection + # can be re-used without leaking prior configs. + self.sock = None + self.is_verified = False + self.proxy_is_verified = None + self._has_connected_to_proxy = False + self._response_options = None + self._tunnel_host = None + self._tunnel_port = None + self._tunnel_scheme = None + + def putrequest( + self, + method: str, + url: str, + skip_host: bool = False, + skip_accept_encoding: bool = False, + ) -> None: + """""" # 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()) + f"Method cannot contain non-token characters {method!r} (found at least {match.group()!r})" ) - return _HTTPConnection.putrequest(self, method, url, *args, **kwargs) + return super().putrequest( + method, url, skip_host=skip_host, skip_accept_encoding=skip_accept_encoding + ) - def putheader(self, header, *values): - """ """ + def putheader(self, header: str, *values: str) -> None: + """""" 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: + super().putheader(header, *values) + elif to_str(header.lower()) not in SKIPPABLE_HEADERS: + skippable_headers = "', '".join( + [str.title(header) for header in sorted(SKIPPABLE_HEADERS)] + ) raise ValueError( - "urllib3.util.SKIP_HEADER only supports '%s'" - % ("', '".join(map(str.title, sorted(SKIPPABLE_HEADERS))),) + f"urllib3.util.SKIP_HEADER only supports '{skippable_headers}'" ) - def request(self, method, url, body=None, headers=None): + # `request` method's signature intentionally violates LSP. + # urllib3's API is different from `http.client.HTTPConnection` and the subclassing is only incidental. + def request( # type: ignore[override] + self, + method: str, + url: str, + body: _TYPE_BODY | None = None, + headers: typing.Mapping[str, str] | None = None, + *, + chunked: bool = False, + preload_content: bool = True, + decode_content: bool = True, + enforce_content_length: bool = True, + ) -> 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: + if self.sock is not None: self.sock.settimeout(self.timeout) + # Store these values to be fed into the HTTPResponse + # object later. TODO: Remove this in favor of a real + # HTTP lifecycle mechanism. + + # We have to store these before we call .request() + # because sometimes we can still salvage a response + # off the wire even if we aren't able to completely + # send the request body. + self._response_options = _ResponseOptions( + request_method=method, + request_url=url, + preload_content=preload_content, + decode_content=decode_content, + enforce_content_length=enforce_content_length, + ) + 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 = headers or {} - header_keys = set([six.ensure_str(k.lower()) for k in headers]) + header_keys = frozenset(to_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 ) + + # Transform the body into an iterable of sendall()-able chunks + # and detect if an explicit Content-Length is doable. + chunks_and_cl = body_to_chunks(body, method=method, blocksize=self.blocksize) + chunks = chunks_and_cl.chunks + content_length = chunks_and_cl.content_length + + # When chunked is explicit set to 'True' we respect that. + if chunked: + if "transfer-encoding" not in header_keys: + self.putheader("Transfer-Encoding", "chunked") + else: + # Detect whether a framing mechanism is already in use. If so + # we respect that value, otherwise we pick chunked vs content-length + # depending on the type of 'body'. + if "content-length" in header_keys: + chunked = False + elif "transfer-encoding" in header_keys: + chunked = True + + # Otherwise we go off the recommendation of 'body_to_chunks()'. + else: + chunked = False + if content_length is None: + if chunks is not None: + chunked = True + self.putheader("Transfer-Encoding", "chunked") + else: + self.putheader("Content-Length", str(content_length)) + + # Now that framing headers are out of the way we send all the other headers. 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 header_keys: - self.putheader("Transfer-Encoding", "chunked") self.endheaders() - if body is not None: - stringish_types = six.string_types + (bytes,) - if isinstance(body, stringish_types): - body = (body,) - for chunk in body: + # If we're given a body we start sending that in chunks. + if chunks is not None: + for chunk in chunks: + # Sending empty chunks isn't allowed for TE: chunked + # as it indicates the end of the body. if not chunk: continue - if not isinstance(chunk, bytes): - chunk = chunk.encode("utf8") - len_str = hex(len(chunk))[2:] - to_send = bytearray(len_str.encode()) - to_send += b"\r\n" - to_send += chunk - to_send += b"\r\n" - self.send(to_send) + if isinstance(chunk, str): + chunk = chunk.encode("utf-8") + if chunked: + self.send(b"%x\r\n%b\r\n" % (len(chunk), chunk)) + else: + self.send(chunk) - # After the if clause, to always have a closed body - self.send(b"0\r\n\r\n") + # Regardless of whether we have a body or not, if we're in + # chunked mode we want to send an explicit empty chunk. + if chunked: + self.send(b"0\r\n\r\n") + + def request_chunked( + self, + method: str, + url: str, + body: _TYPE_BODY | None = None, + headers: typing.Mapping[str, str] | None = None, + ) -> None: + """ + Alternative to the common request method, which sends the + body with chunked encoding and not as one block + """ + warnings.warn( + "HTTPConnection.request_chunked() is deprecated and will be removed " + "in urllib3 v2.1.0. Instead use HTTPConnection.request(..., chunked=True).", + category=DeprecationWarning, + stacklevel=2, + ) + self.request(method, url, body=body, headers=headers, chunked=True) + + def getresponse( # type: ignore[override] + self, + ) -> HTTPResponse: + """ + Get the response from the server. + + If the HTTPConnection is in the correct state, returns an instance of HTTPResponse or of whatever object is returned by the response_class variable. + + If a request has not been sent or if a previous response has not be handled, ResponseNotReady is raised. If the HTTP response indicates that the connection should be closed, then it will be closed before the response is returned. When the connection is closed, the underlying socket is closed. + """ + # Raise the same error as http.client.HTTPConnection + if self._response_options is None: + raise ResponseNotReady() + + # Reset this attribute for being used again. + resp_options = self._response_options + self._response_options = None + + # Since the connection's timeout value may have been updated + # we need to set the timeout on the socket. + self.sock.settimeout(self.timeout) + + # This is needed here to avoid circular import errors + from .response import HTTPResponse + + # Get the response from http.client.HTTPConnection + httplib_response = super().getresponse() + + try: + assert_header_parsing(httplib_response.msg) + except (HeaderParsingError, TypeError) as hpe: + log.warning( + "Failed to parse headers (url=%s): %s", + _url_from_connection(self, resp_options.request_url), + hpe, + exc_info=True, + ) + + headers = HTTPHeaderDict(httplib_response.msg.items()) + + response = HTTPResponse( + body=httplib_response, + headers=headers, + status=httplib_response.status, + version=httplib_response.version, + reason=httplib_response.reason, + preload_content=resp_options.preload_content, + decode_content=resp_options.decode_content, + original_response=httplib_response, + enforce_content_length=resp_options.enforce_content_length, + request_method=resp_options.request_method, + request_url=resp_options.request_url, + ) + return response class HTTPSConnection(HTTPConnection): @@ -289,57 +494,100 @@ class HTTPSConnection(HTTPConnection): socket by means of :py:func:`urllib3.util.ssl_wrap_socket`. """ - default_port = port_by_scheme["https"] + default_port = port_by_scheme["https"] # type: ignore[misc] - 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 + cert_reqs: int | str | None = None + ca_certs: str | None = None + ca_cert_dir: str | None = None + ca_cert_data: None | str | bytes = None + ssl_version: int | str | None = None + ssl_minimum_version: int | None = None + ssl_maximum_version: int | None = None + assert_fingerprint: str | None = None def __init__( self, - host, - port=None, - key_file=None, - cert_file=None, - key_password=None, - strict=None, - timeout=socket._GLOBAL_DEFAULT_TIMEOUT, - ssl_context=None, - server_hostname=None, - **kw - ): - - HTTPConnection.__init__(self, host, port, strict=strict, timeout=timeout, **kw) + host: str, + port: int | None = None, + *, + timeout: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT, + source_address: tuple[str, int] | None = None, + blocksize: int = 16384, + socket_options: None + | (connection._TYPE_SOCKET_OPTIONS) = HTTPConnection.default_socket_options, + proxy: Url | None = None, + proxy_config: ProxyConfig | None = None, + cert_reqs: int | str | None = None, + assert_hostname: None | str | Literal[False] = None, + assert_fingerprint: str | None = None, + server_hostname: str | None = None, + ssl_context: ssl.SSLContext | None = None, + ca_certs: str | None = None, + ca_cert_dir: str | None = None, + ca_cert_data: None | str | bytes = None, + ssl_minimum_version: int | None = None, + ssl_maximum_version: int | None = None, + ssl_version: int | str | None = None, # Deprecated + cert_file: str | None = None, + key_file: str | None = None, + key_password: str | None = None, + ) -> None: + super().__init__( + host, + port=port, + timeout=timeout, + source_address=source_address, + blocksize=blocksize, + socket_options=socket_options, + proxy=proxy, + proxy_config=proxy_config, + ) self.key_file = key_file self.cert_file = cert_file self.key_password = key_password self.ssl_context = ssl_context self.server_hostname = server_hostname + self.assert_hostname = assert_hostname + self.assert_fingerprint = assert_fingerprint + self.ssl_version = ssl_version + self.ssl_minimum_version = ssl_minimum_version + self.ssl_maximum_version = ssl_maximum_version + 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 - # Required property for Google AppEngine 1.9.0 which otherwise causes - # HTTPS requests to go out as HTTP. (See Issue #356) - self._protocol = "https" + # cert_reqs depends on ssl_context so calculate last. + if cert_reqs is None: + if self.ssl_context is not None: + cert_reqs = self.ssl_context.verify_mode + else: + cert_reqs = resolve_cert_reqs(None) + self.cert_reqs = cert_reqs def set_cert( self, - key_file=None, - cert_file=None, - cert_reqs=None, - key_password=None, - ca_certs=None, - assert_hostname=None, - assert_fingerprint=None, - ca_cert_dir=None, - ca_cert_data=None, - ): + key_file: str | None = None, + cert_file: str | None = None, + cert_reqs: int | str | None = None, + key_password: str | None = None, + ca_certs: str | None = None, + assert_hostname: None | str | Literal[False] = None, + assert_fingerprint: str | None = None, + ca_cert_dir: str | None = None, + ca_cert_data: None | str | bytes = None, + ) -> None: """ This method should only be called once, before the connection is used. """ + warnings.warn( + "HTTPSConnection.set_cert() is deprecated and will be removed " + "in urllib3 v2.1.0. Instead provide the parameters to the " + "HTTPSConnection constructor.", + category=DeprecationWarning, + stacklevel=2, + ) + # If cert_reqs is not provided we'll assume CERT_REQUIRED unless we also # have an SSLContext object in which case we'll use its verify_mode. if cert_reqs is None: @@ -358,27 +606,26 @@ class HTTPSConnection(HTTPConnection): 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 - self.sock = conn = self._new_conn() - hostname = self.host + def connect(self) -> None: + sock: socket.socket | ssl.SSLSocket + self.sock = sock = self._new_conn() + server_hostname: str = 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) + # Do we need to establish a tunnel? + if self._tunnel_host is not None: + # We're tunneling to an HTTPS origin so need to do TLS-in-TLS. + if self._tunnel_scheme == "https": + self.sock = sock = self._connect_tls_proxy(self.host, sock) tls_in_tls = True - # Calls self._set_hostport(), so self.host is - # self._tunnel_host below. - self._tunnel() - # Mark this connection as not reusable - self.auto_open = 0 + # If we're tunneling it means we're connected to our proxy. + self._has_connected_to_proxy = True + self._tunnel() # type: ignore[attr-defined] # Override the host with the one we're requesting data from. - hostname = self._tunnel_host + server_hostname = self._tunnel_host - server_hostname = hostname if self.server_hostname is not None: server_hostname = self.server_hostname @@ -386,163 +633,218 @@ class HTTPSConnection(HTTPConnection): if is_time_off: warnings.warn( ( - "System time is way off (before {0}). This will probably " + f"System time is way off (before {RECENT_DATE}). This will probably " "lead to SSL verification errors" - ).format(RECENT_DATE), + ), SystemTimeWarning, ) - # 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), - ) - - context = self.ssl_context - context.verify_mode = resolve_cert_reqs(self.cert_reqs) - - # Try to load OS default certs if none are given. - # Works well on Windows (requires Python3.4+) - 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") - ): - context.load_default_certs() - - self.sock = ssl_wrap_socket( - sock=conn, - keyfile=self.key_file, - certfile=self.cert_file, - key_password=self.key_password, + sock_and_verified = _ssl_wrap_socket_and_match_hostname( + sock=sock, + cert_reqs=self.cert_reqs, + ssl_version=self.ssl_version, + ssl_minimum_version=self.ssl_minimum_version, + ssl_maximum_version=self.ssl_maximum_version, ca_certs=self.ca_certs, ca_cert_dir=self.ca_cert_dir, ca_cert_data=self.ca_cert_data, + cert_file=self.cert_file, + key_file=self.key_file, + key_password=self.key_password, server_hostname=server_hostname, - ssl_context=context, + ssl_context=self.ssl_context, tls_in_tls=tls_in_tls, + assert_hostname=self.assert_hostname, + assert_fingerprint=self.assert_fingerprint, ) + self.sock = sock_and_verified.socket + self.is_verified = sock_and_verified.is_verified - # 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 there's a proxy to be connected to we are fully connected. + # This is set twice (once above and here) due to forwarding proxies + # not using tunnelling. + self._has_connected_to_proxy = bool(self.proxy) - if self.assert_fingerprint: - assert_fingerprint( - self.sock.getpeercert(binary_form=True), self.assert_fingerprint - ) - elif ( - context.verify_mode != ssl.CERT_NONE - and not getattr(context, "check_hostname", False) - and self.assert_hostname is not 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 = self.sock.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, self.assert_hostname or server_hostname) - - self.is_verified = ( - context.verify_mode == ssl.CERT_REQUIRED - or self.assert_fingerprint is not None - ) - - def _connect_tls_proxy(self, hostname, conn): + def _connect_tls_proxy(self, hostname: str, sock: socket.socket) -> ssl.SSLSocket: """ Establish a TLS connection to the proxy using the provided SSL context. """ - proxy_config = self.proxy_config + # `_connect_tls_proxy` is called when self._tunnel_host is truthy. + proxy_config = typing.cast(ProxyConfig, 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, + sock_and_verified = _ssl_wrap_socket_and_match_hostname( + sock, + cert_reqs=self.cert_reqs, + ssl_version=self.ssl_version, + ssl_minimum_version=self.ssl_minimum_version, + ssl_maximum_version=self.ssl_maximum_version, 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, + assert_hostname=proxy_config.assert_hostname, + assert_fingerprint=proxy_config.assert_fingerprint, + # Features that aren't implemented for proxies yet: + cert_file=None, + key_file=None, + key_password=None, + tls_in_tls=False, ) + self.proxy_is_verified = sock_and_verified.is_verified + return sock_and_verified.socket # type: ignore[return-value] - if ssl_context.verify_mode != ssl.CERT_NONE and not getattr( - ssl_context, "check_hostname", False + +class _WrappedAndVerifiedSocket(typing.NamedTuple): + """ + Wrapped socket and whether the connection is + verified after the TLS handshake + """ + + socket: ssl.SSLSocket | SSLTransport + is_verified: bool + + +def _ssl_wrap_socket_and_match_hostname( + sock: socket.socket, + *, + cert_reqs: None | str | int, + ssl_version: None | str | int, + ssl_minimum_version: int | None, + ssl_maximum_version: int | None, + cert_file: str | None, + key_file: str | None, + key_password: str | None, + ca_certs: str | None, + ca_cert_dir: str | None, + ca_cert_data: None | str | bytes, + assert_hostname: None | str | Literal[False], + assert_fingerprint: str | None, + server_hostname: str | None, + ssl_context: ssl.SSLContext | None, + tls_in_tls: bool = False, +) -> _WrappedAndVerifiedSocket: + """Logic for constructing an SSLContext from all TLS parameters, passing + that down into ssl_wrap_socket, and then doing certificate verification + either via hostname or fingerprint. This function exists to guarantee + that both proxies and targets have the same behavior when connecting via TLS. + """ + default_ssl_context = False + if ssl_context is None: + default_ssl_context = True + context = create_urllib3_context( + ssl_version=resolve_ssl_version(ssl_version), + ssl_minimum_version=ssl_minimum_version, + ssl_maximum_version=ssl_maximum_version, + cert_reqs=resolve_cert_reqs(cert_reqs), + ) + else: + context = ssl_context + + context.verify_mode = resolve_cert_reqs(cert_reqs) + + # In some cases, we want to verify hostnames ourselves + if ( + # `ssl` can't verify fingerprints or alternate hostnames + assert_fingerprint + or assert_hostname + # assert_hostname can be set to False to disable hostname checking + or assert_hostname is False + # We still support OpenSSL 1.0.2, which prevents us from verifying + # hostnames easily: https://github.com/pyca/pyopenssl/pull/933 + or ssl_.IS_PYOPENSSL + or not ssl_.HAS_NEVER_CHECK_COMMON_NAME + ): + context.check_hostname = False + + # Try to load OS default certs if none are given. + # We need to do the hasattr() check for our custom + # pyOpenSSL and SecureTransport SSLContext objects + # because neither support load_default_certs(). + if ( + not ca_certs + and not ca_cert_dir + and not ca_cert_data + and default_ssl_context + and hasattr(context, "load_default_certs") + ): + context.load_default_certs() + + # Ensure that IPv6 addresses are in the proper format and don't have a + # scope ID. Python's SSL module fails to recognize scoped IPv6 addresses + # and interprets them as DNS hostnames. + if server_hostname is not None: + normalized = server_hostname.strip("[]") + if "%" in normalized: + normalized = normalized[: normalized.rfind("%")] + if is_ipaddress(normalized): + server_hostname = normalized + + ssl_sock = ssl_wrap_socket( + sock=sock, + keyfile=key_file, + certfile=cert_file, + key_password=key_password, + ca_certs=ca_certs, + ca_cert_dir=ca_cert_dir, + ca_cert_data=ca_cert_data, + server_hostname=server_hostname, + ssl_context=context, + tls_in_tls=tls_in_tls, + ) + + try: + if assert_fingerprint: + _assert_fingerprint( + ssl_sock.getpeercert(binary_form=True), assert_fingerprint + ) + elif ( + context.verify_mode != ssl.CERT_NONE + and not context.check_hostname + and assert_hostname is not 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, + cert: _TYPE_PEER_CERT_RET_DICT = ssl_sock.getpeercert() # type: ignore[assignment] + + # Need to signal to our match_hostname whether to use 'commonName' or not. + # If we're using our own constructed SSLContext we explicitly set 'False' + # because PyPy hard-codes 'True' from SSLContext.hostname_checks_common_name. + if default_ssl_context: + hostname_checks_common_name = False + else: + hostname_checks_common_name = ( + getattr(context, "hostname_checks_common_name", False) or False ) - _match_hostname(cert, hostname) - self.proxy_is_verified = ssl_context.verify_mode == ssl.CERT_REQUIRED - return socket + _match_hostname( + cert, + assert_hostname or server_hostname, # type: ignore[arg-type] + hostname_checks_common_name, + ) + + return _WrappedAndVerifiedSocket( + socket=ssl_sock, + is_verified=context.verify_mode == ssl.CERT_REQUIRED + or bool(assert_fingerprint), + ) + except BaseException: + ssl_sock.close() + raise -def _match_hostname(cert, asserted_hostname): +def _match_hostname( + cert: _TYPE_PEER_CERT_RET_DICT | None, + asserted_hostname: str, + hostname_checks_common_name: bool = False, +) -> None: # 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[]") + stripped_hostname = asserted_hostname.strip("[]") if is_ipaddress(stripped_hostname): asserted_hostname = stripped_hostname try: - match_hostname(cert, asserted_hostname) + match_hostname(cert, asserted_hostname, hostname_checks_common_name) except CertificateError as e: log.warning( "Certificate did not match expected hostname: %s. Certificate: %s", @@ -551,22 +853,54 @@ def _match_hostname(cert, asserted_hostname): ) # Add cert to exception and reraise so client code can inspect # the cert when catching the exception, if they want to - e._peer_cert = cert + e._peer_cert = cert # type: ignore[attr-defined] raise -def _get_default_user_agent(): - return "python-urllib3/%s" % __version__ +def _wrap_proxy_error(err: Exception, proxy_scheme: str | None) -> ProxyError: + # Look for the phrase 'wrong version number', if found + # then we should warn the user that we're very sure that + # this proxy is HTTP-only and they have a configuration issue. + error_normalized = " ".join(re.split("[^a-z]", str(err).lower())) + is_likely_http_proxy = ( + "wrong version number" in error_normalized + or "unknown protocol" in error_normalized + ) + http_proxy_warning = ( + ". Your proxy appears to only use HTTP and not HTTPS, " + "try changing your proxy URL to be HTTP. See: " + "https://urllib3.readthedocs.io/en/latest/advanced-usage.html" + "#https-proxy-error-http-proxy" + ) + new_err = ProxyError( + f"Unable to connect to proxy" + f"{http_proxy_warning if is_likely_http_proxy and proxy_scheme == 'https' else ''}", + err, + ) + new_err.__cause__ = err + return new_err -class DummyConnection(object): +def _get_default_user_agent() -> str: + return f"python-urllib3/{__version__}" + + +class DummyConnection: """Used to detect a failed ConnectionCls import.""" - pass - if not ssl: - HTTPSConnection = DummyConnection # noqa: F811 + HTTPSConnection = DummyConnection # type: ignore[misc, assignment] # noqa: F811 VerifiedHTTPSConnection = HTTPSConnection + + +def _url_from_connection( + conn: HTTPConnection | HTTPSConnection, path: str | None = None +) -> str: + """Returns the URL from a given connection. This is mainly used for testing and logging.""" + + scheme = "https" if isinstance(conn, HTTPSConnection) else "http" + + return Url(scheme=scheme, host=conn.host, port=conn.port, path=path).url diff --git a/lib/urllib3/connectionpool.py b/lib/urllib3/connectionpool.py index 96844d93..2479405b 100644 --- a/lib/urllib3/connectionpool.py +++ b/lib/urllib3/connectionpool.py @@ -1,14 +1,17 @@ -from __future__ import absolute_import +from __future__ import annotations import errno import logging -import re -import socket +import queue import sys +import typing import warnings -from socket import error as SocketError +import weakref from socket import timeout as SocketTimeout +from types import TracebackType +from ._base_connection import _TYPE_BODY +from ._request_methods import RequestMethods from .connection import ( BaseSSLError, BrokenPipeError, @@ -16,13 +19,14 @@ from .connection import ( HTTPConnection, HTTPException, HTTPSConnection, - VerifiedHTTPSConnection, - port_by_scheme, + ProxyConfig, + _wrap_proxy_error, ) +from .connection import port_by_scheme as port_by_scheme from .exceptions import ( ClosedPoolError, EmptyPoolError, - HeaderParsingError, + FullPoolError, HostChangedError, InsecureRequestWarning, LocationValueError, @@ -34,38 +38,34 @@ from .exceptions import ( SSLError, TimeoutError, ) -from .packages import six -from .packages.six.moves import queue -from .request import RequestMethods -from .response import HTTPResponse +from .response import BaseHTTPResponse 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.request import _TYPE_BODY_POSITION, set_file_position from .util.retry import Retry from .util.ssl_match_hostname import CertificateError -from .util.timeout import Timeout +from .util.timeout import _DEFAULT_TIMEOUT, _TYPE_DEFAULT, Timeout from .util.url import Url, _encode_target from .util.url import _normalize_host as normalize_host -from .util.url import get_host, parse_url +from .util.url import parse_url +from .util.util import to_str -try: # Platform-specific: Python 3 - import weakref +if typing.TYPE_CHECKING: + import ssl - weakref_finalize = weakref.finalize -except AttributeError: # Platform-specific: Python 2 - from .packages.backports.weakref_finalize import weakref_finalize + from typing_extensions import Literal -xrange = six.moves.xrange + from ._base_connection import BaseHTTPConnection, BaseHTTPSConnection log = logging.getLogger(__name__) -_Default = object() +_TYPE_TIMEOUT = typing.Union[Timeout, float, _TYPE_DEFAULT, None] + +_SelfT = typing.TypeVar("_SelfT") # Pool objects -class ConnectionPool(object): +class ConnectionPool: """ Base class for all connection pools, such as :class:`.HTTPConnectionPool` and :class:`.HTTPSConnectionPool`. @@ -76,33 +76,42 @@ class ConnectionPool(object): target URIs. """ - scheme = None - QueueCls = LifoQueue + scheme: str | None = None + QueueCls = queue.LifoQueue - def __init__(self, host, port=None): + def __init__(self, host: str, port: int | None = None) -> None: if not host: raise LocationValueError("No host specified.") self.host = _normalize_host(host, scheme=self.scheme) - self._proxy_host = host.lower() self.port = port - def __str__(self): - return "%s(host=%r, port=%r)" % (type(self).__name__, self.host, self.port) + # This property uses 'normalize_host()' (not '_normalize_host()') + # to avoid removing square braces around IPv6 addresses. + # This value is sent to `HTTPConnection.set_tunnel()` if called + # because square braces are required for HTTP CONNECT tunneling. + self._tunnel_host = normalize_host(host, scheme=self.scheme).lower() - def __enter__(self): + def __str__(self) -> str: + return f"{type(self).__name__}(host={self.host!r}, port={self.port!r})" + + def __enter__(self: _SelfT) -> _SelfT: return self - def __exit__(self, exc_type, exc_val, exc_tb): + def __exit__( + self, + exc_type: type[BaseException] | None, + exc_val: BaseException | None, + exc_tb: TracebackType | None, + ) -> Literal[False]: self.close() # Return False to re-raise any potential exceptions return False - def close(self): + def close(self) -> None: """ Close all pooled connections and disable the pool. """ - pass # This is taken from http://hg.python.org/cpython/file/7aaba721ebc0/Lib/socket.py#l252 @@ -121,14 +130,6 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): Port used for this HTTP Connection (None is equivalent to 80), passed 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:`http.client.HTTPConnection`. - - .. note:: - Only works in Python 2. This parameter is ignored in Python 3. - :param timeout: Socket timeout in seconds for each individual connection. This can be a float or integer, which sets the timeout for the HTTP request, @@ -170,29 +171,27 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): """ scheme = "http" - ConnectionCls = HTTPConnection - ResponseCls = HTTPResponse + ConnectionCls: ( + type[BaseHTTPConnection] | type[BaseHTTPSConnection] + ) = HTTPConnection def __init__( self, - host, - port=None, - strict=False, - timeout=Timeout.DEFAULT_TIMEOUT, - maxsize=1, - block=False, - headers=None, - retries=None, - _proxy=None, - _proxy_headers=None, - _proxy_config=None, - **conn_kw + host: str, + port: int | None = None, + timeout: _TYPE_TIMEOUT | None = _DEFAULT_TIMEOUT, + maxsize: int = 1, + block: bool = False, + headers: typing.Mapping[str, str] | None = None, + retries: Retry | bool | int | None = None, + _proxy: Url | None = None, + _proxy_headers: typing.Mapping[str, str] | None = None, + _proxy_config: ProxyConfig | None = None, + **conn_kw: typing.Any, ): ConnectionPool.__init__(self, host, port) RequestMethods.__init__(self, headers) - self.strict = strict - if not isinstance(timeout, Timeout): timeout = Timeout.from_float(timeout) @@ -202,7 +201,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): self.timeout = timeout self.retries = retries - self.pool = self.QueueCls(maxsize) + self.pool: queue.LifoQueue[typing.Any] | None = self.QueueCls(maxsize) self.block = block self.proxy = _proxy @@ -210,7 +209,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): self.proxy_config = _proxy_config # Fill the queue up so that doing get() on it will block properly - for _ in xrange(maxsize): + for _ in range(maxsize): self.pool.put(None) # These are mostly for testing and debugging purposes. @@ -235,9 +234,9 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # Close all the HTTPConnections in the pool before the # HTTPConnectionPool object is garbage collected. - weakref_finalize(self, _close_pool_connections, pool) + weakref.finalize(self, _close_pool_connections, pool) - def _new_conn(self): + def _new_conn(self) -> BaseHTTPConnection: """ Return a fresh :class:`HTTPConnection`. """ @@ -253,12 +252,11 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): host=self.host, port=self.port, timeout=self.timeout.connect_timeout, - strict=self.strict, - **self.conn_kw + **self.conn_kw, ) return conn - def _get_conn(self, timeout=None): + def _get_conn(self, timeout: float | None = None) -> BaseHTTPConnection: """ Get a connection. Will return a pooled connection if one is available. @@ -271,33 +269,32 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): :prop:`.block` is ``True``. """ conn = None + + if self.pool is None: + raise ClosedPoolError(self, "Pool is closed.") + try: conn = self.pool.get(block=self.block, timeout=timeout) except AttributeError: # self.pool is None - raise ClosedPoolError(self, "Pool is closed.") + raise ClosedPoolError(self, "Pool is closed.") from None # Defensive: except queue.Empty: if self.block: raise EmptyPoolError( self, - "Pool reached maximum size and no more connections are allowed.", - ) + "Pool is empty and a new connection can't be opened due to blocking mode.", + ) from None pass # Oh well, we'll create a new connection then # If this is a persistent connection, check if it got disconnected if conn and is_connection_dropped(conn): log.debug("Resetting dropped connection: %s", self.host) conn.close() - if getattr(conn, "auto_open", 1) == 0: - # This is a proxied connection that has been mutated by - # http.client._tunnel() and cannot be reused (since it would - # attempt to bypass the proxy) - conn = None return conn or self._new_conn() - def _put_conn(self, conn): + def _put_conn(self, conn: BaseHTTPConnection | None) -> None: """ Put a connection back into the pool. @@ -311,36 +308,47 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): If the pool is closed, then the connection will be closed and discarded. """ - try: - self.pool.put(conn, block=False) - return # Everything is dandy, done. - except AttributeError: - # self.pool is None. - pass - except queue.Full: - # This should never happen if self.block == True - log.warning( - "Connection pool is full, discarding connection: %s. Connection pool size: %s", - self.host, - self.pool.qsize(), - ) + if self.pool is not None: + try: + self.pool.put(conn, block=False) + return # Everything is dandy, done. + except AttributeError: + # self.pool is None. + pass + except queue.Full: + # Connection never got put back into the pool, close it. + if conn: + conn.close() + + if self.block: + # This should never happen if you got the conn from self._get_conn + raise FullPoolError( + self, + "Pool reached maximum size and no more connections are allowed.", + ) from None + + 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() - def _validate_conn(self, conn): + def _validate_conn(self, conn: BaseHTTPConnection) -> None: """ Called right before a request is made, after the socket is created. """ - pass - def _prepare_proxy(self, conn): + def _prepare_proxy(self, conn: BaseHTTPConnection) -> None: # Nothing to do for HTTP connections. pass - def _get_timeout(self, timeout): + def _get_timeout(self, timeout: _TYPE_TIMEOUT) -> Timeout: """Helper that always returns a :class:`urllib3.util.Timeout`""" - if timeout is _Default: + if timeout is _DEFAULT_TIMEOUT: return self.timeout.clone() if isinstance(timeout, Timeout): @@ -350,34 +358,40 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # can be removed later return Timeout.from_float(timeout) - def _raise_timeout(self, err, url, timeout_value): + def _raise_timeout( + self, + err: BaseSSLError | OSError | SocketTimeout, + url: str, + timeout_value: _TYPE_TIMEOUT | None, + ) -> None: """Is the error actually a timeout? Will raise a ReadTimeout or pass""" if isinstance(err, SocketTimeout): raise ReadTimeoutError( - self, url, "Read timed out. (read timeout=%s)" % timeout_value - ) + self, url, f"Read timed out. (read timeout={timeout_value})" + ) from err - # See the above comment about EAGAIN in Python 3. In Python 2 we have - # to specifically catch it and throw the timeout error + # See the above comment about EAGAIN in Python 3. if hasattr(err, "errno") and err.errno in _blocking_errnos: raise ReadTimeoutError( - self, url, "Read timed out. (read timeout=%s)" % timeout_value - ) - - # Catch possible read timeouts thrown as SSL errors. If not the - # case, rethrow the original. We need to do this because of: - # http://bugs.python.org/issue10272 - if "timed out" in str(err) or "did not complete (read)" in str( - err - ): # Python < 2.7.4 - raise ReadTimeoutError( - self, url, "Read timed out. (read timeout=%s)" % timeout_value - ) + self, url, f"Read timed out. (read timeout={timeout_value})" + ) from err def _make_request( - self, conn, method, url, timeout=_Default, chunked=False, **httplib_request_kw - ): + self, + conn: BaseHTTPConnection, + method: str, + url: str, + body: _TYPE_BODY | None = None, + headers: typing.Mapping[str, str] | None = None, + retries: Retry | None = None, + timeout: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT, + chunked: bool = False, + response_conn: BaseHTTPConnection | None = None, + preload_content: bool = True, + decode_content: bool = True, + enforce_content_length: bool = True, + ) -> BaseHTTPResponse: """ Perform a request on a given urllib connection object taken from our pool. @@ -385,12 +399,61 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): :param conn: a connection from one of our connection pools + :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, 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, + If-None-Match, etc. If None, pool headers are used. If provided, + these headers completely replace any pool-specific headers. + + :param retries: + Configure the number of retries to allow before raising a + :class:`~urllib3.exceptions.MaxRetryError` exception. + + Pass ``None`` to retry until you receive a response. Pass a + :class:`~urllib3.util.retry.Retry` object for fine-grained control + over different types of retries. + Pass an integer number to retry connection errors that many times, + but no other types of errors. Pass zero to never retry. + + If ``False``, then retries are disabled and any exception is raised + immediately. Also, instead of raising a MaxRetryError on redirects, + the redirect response will be returned. + + :type retries: :class:`~urllib3.util.retry.Retry`, False, or an int. + :param timeout: - Socket timeout in seconds for the request. This can be a - float or integer, which will set the same timeout value for - the socket connect and the socket read, or an instance of - :class:`urllib3.util.Timeout`, which gives you more fine-grained - control over your timeouts. + If specified, overrides the default timeout for this one + request. It may be a float (in seconds) or an instance of + :class:`urllib3.util.Timeout`. + + :param chunked: + If True, urllib3 will send the body using chunked transfer + encoding. Otherwise, urllib3 will send the body using the standard + content-length form. Defaults to False. + + :param response_conn: + Set this to ``None`` if you will handle releasing the connection or + set the connection to have the response release it. + + :param preload_content: + If True, the response's body will be preloaded during construction. + + :param decode_content: + If True, will attempt to decode the body based on the + 'content-encoding' header. + + :param enforce_content_length: + Enforce content length checking. Body returned by server must match + value of Content-Length header, if present. Otherwise, raise error. """ self.num_requests += 1 @@ -398,44 +461,65 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): timeout_obj.start_connect() conn.timeout = Timeout.resolve_default_timeout(timeout_obj.connect_timeout) - # Trigger any extra validation we need to do. try: - self._validate_conn(conn) - except (SocketTimeout, BaseSSLError) as e: - # Py2 raises this as a BaseSSLError, Py3 raises it as socket timeout. - self._raise_timeout(err=e, url=url, timeout_value=conn.timeout) - raise + # Trigger any extra validation we need to do. + try: + self._validate_conn(conn) + except (SocketTimeout, BaseSSLError) as e: + self._raise_timeout(err=e, url=url, timeout_value=conn.timeout) + raise + + # _validate_conn() starts the connection to an HTTPS proxy + # so we need to wrap errors with 'ProxyError' here too. + except ( + OSError, + NewConnectionError, + TimeoutError, + BaseSSLError, + CertificateError, + SSLError, + ) as e: + new_e: Exception = e + if isinstance(e, (BaseSSLError, CertificateError)): + new_e = SSLError(e) + # If the connection didn't successfully connect to it's proxy + # then there + if isinstance( + new_e, (OSError, NewConnectionError, TimeoutError, SSLError) + ) and (conn and conn.proxy and not conn.has_connected_to_proxy): + new_e = _wrap_proxy_error(new_e, conn.proxy.scheme) + raise new_e # conn.request() calls http.client.*.request, not the method in # urllib3.request. It also calls makefile (recv) on the socket. try: - if chunked: - conn.request_chunked(method, url, **httplib_request_kw) - else: - conn.request(method, url, **httplib_request_kw) + conn.request( + method, + url, + body=body, + headers=headers, + chunked=chunked, + preload_content=preload_content, + decode_content=decode_content, + enforce_content_length=enforce_content_length, + ) # 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 + except OSError as e: + # MacOS/Linux + # 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, - }: + if e.errno != errno.EPROTOTYPE: raise # Reset the timeout for the recv() on the socket read_timeout = timeout_obj.read_timeout - # App Engine doesn't have a sock attr - if getattr(conn, "sock", None): + if not conn.is_closed: # In Python 3 socket.py will catch EAGAIN and return None when you # try and read into the file pointer created by http.client, which # instead raises a BadStatusLine exception. Instead of catching @@ -443,33 +527,22 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # timeouts, check for a zero timeout before making the request. if read_timeout == 0: raise ReadTimeoutError( - self, url, "Read timed out. (read timeout=%s)" % read_timeout + self, url, f"Read timed out. (read timeout={read_timeout})" ) - if read_timeout is Timeout.DEFAULT_TIMEOUT: - conn.sock.settimeout(socket.getdefaulttimeout()) - else: # None or a value - conn.sock.settimeout(read_timeout) + conn.timeout = read_timeout # Receive the response from the server try: - try: - # Python 2.7, use buffering of HTTP responses - httplib_response = conn.getresponse(buffering=True) - except TypeError: - # Python 3 - try: - httplib_response = conn.getresponse() - except BaseException as e: - # Remove the TypeError from the exception chain in - # Python 3 (including for exceptions like SystemExit). - # Otherwise it looks like a bug in the code. - six.raise_from(e, None) - except (SocketTimeout, BaseSSLError, SocketError) as e: + response = conn.getresponse() + except (BaseSSLError, OSError) as e: self._raise_timeout(err=e, url=url, timeout_value=read_timeout) raise - # AppEngine doesn't have a version attr. - http_version = getattr(conn, "_http_vsn_str", "HTTP/?") + # Set properties that are used by the pooling layer. + response.retries = retries + response._connection = response_conn # type: ignore[attr-defined] + response._pool = self # type: ignore[attr-defined] + log.debug( '%s://%s:%s "%s %s %s" %s %s', self.scheme, @@ -477,27 +550,15 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): self.port, method, url, - http_version, - httplib_response.status, - httplib_response.length, + # HTTP version + conn._http_vsn_str, # type: ignore[attr-defined] + response.status, + response.length_remaining, # type: ignore[attr-defined] ) - try: - assert_header_parsing(httplib_response.msg) - except (HeaderParsingError, TypeError) as hpe: # Platform-specific: Python 3 - log.warning( - "Failed to parse headers (url=%s): %s", - self._absolute_url(url), - hpe, - exc_info=True, - ) + return response - return httplib_response - - def _absolute_url(self, path): - return Url(scheme=self.scheme, host=self.host, port=self.port, path=path).url - - def close(self): + def close(self) -> None: """ Close all pooled connections and disable the pool. """ @@ -509,7 +570,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # Close all the HTTPConnections in the pool. _close_pool_connections(old_pool) - def is_same_host(self, url): + def is_same_host(self, url: str) -> bool: """ Check if the given ``url`` is a member of the same host as this connection pool. @@ -518,7 +579,8 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): return True # TODO: Add optional support for socket.gethostbyname checking. - scheme, host, port = get_host(url) + scheme, _, host, port, *_ = parse_url(url) + scheme = scheme or "http" if host is not None: host = _normalize_host(host, scheme=scheme) @@ -530,22 +592,24 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): return (scheme, host, port) == (self.scheme, self.host, self.port) - def urlopen( + def urlopen( # type: ignore[override] self, - method, - url, - body=None, - headers=None, - retries=None, - redirect=True, - assert_same_host=True, - timeout=_Default, - pool_timeout=None, - release_conn=None, - chunked=False, - body_pos=None, - **response_kw - ): + method: str, + url: str, + body: _TYPE_BODY | None = None, + headers: typing.Mapping[str, str] | None = None, + retries: Retry | bool | int | None = None, + redirect: bool = True, + assert_same_host: bool = True, + timeout: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT, + pool_timeout: int | None = None, + release_conn: bool | None = None, + chunked: bool = False, + body_pos: _TYPE_BODY_POSITION | None = None, + preload_content: bool = True, + decode_content: bool = True, + **response_kw: typing.Any, + ) -> BaseHTTPResponse: """ Get a connection from the pool and perform an HTTP request. This is the lowest level call for making a request, so you'll need to specify all @@ -553,8 +617,8 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): .. note:: - More commonly, it's appropriate to use a convenience method provided - by :class:`.RequestMethods`, such as :meth:`request`. + More commonly, it's appropriate to use a convenience method + such as :meth:`request`. .. note:: @@ -614,6 +678,13 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): block for ``pool_timeout`` seconds and raise EmptyPoolError if no connection is available within the time period. + :param bool preload_content: + If True, the response's body will be preloaded into memory. + + :param bool decode_content: + If True, will attempt to decode the body based on the + 'content-encoding' header. + :param release_conn: If False, then the urlopen call will not release the connection back into the pool once a response is received (but will release if @@ -621,10 +692,10 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): `preload_content=True`). This is useful if you're not preloading the response's content immediately. You will need to call ``r.release_conn()`` on the response ``r`` to return the connection - back into the pool. If None, it takes the value of - ``response_kw.get('preload_content', True)``. + back into the pool. If None, it takes the value of ``preload_content`` + which defaults to ``True``. - :param chunked: + :param bool chunked: If True, urllib3 will send the body using chunked transfer encoding. Otherwise, urllib3 will send the body using the standard content-length form. Defaults to False. @@ -633,12 +704,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): Position to seek to in file-like body in the event of a retry or redirect. Typically this won't need to be set because urllib3 will auto-populate the value when needed. - - :param \\**response_kw: - Additional parameters are passed to - :meth:`urllib3.response.HTTPResponse.from_httplib` """ - parsed_url = parse_url(url) destination_scheme = parsed_url.scheme @@ -649,7 +715,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): retries = Retry.from_int(retries, redirect=redirect, default=self.retries) if release_conn is None: - release_conn = response_kw.get("preload_content", True) + release_conn = preload_content # Check host if assert_same_host and not self.is_same_host(url): @@ -657,9 +723,9 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # Ensure that the URL we're connecting to is properly encoded if url.startswith("/"): - url = six.ensure_str(_encode_target(url)) + url = to_str(_encode_target(url)) else: - url = six.ensure_str(parsed_url.url) + url = to_str(parsed_url.url) conn = None @@ -682,8 +748,8 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # 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) + headers = headers.copy() # type: ignore[attr-defined] + headers.update(self.proxy_headers) # type: ignore[union-attr] # Must keep the exception bound to a separate variable or else Python 3 # complains about UnboundLocalError. @@ -702,24 +768,17 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): timeout_obj = self._get_timeout(timeout) conn = self._get_conn(timeout=pool_timeout) - conn.timeout = timeout_obj.connect_timeout + conn.timeout = timeout_obj.connect_timeout # type: ignore[assignment] - is_new_proxy_conn = self.proxy is not None and not getattr( - conn, "sock", None - ) - if is_new_proxy_conn and http_tunnel_required: - self._prepare_proxy(conn) - - # Make the request on the httplib connection object. - httplib_response = self._make_request( - conn, - method, - url, - timeout=timeout_obj, - body=body, - headers=headers, - chunked=chunked, - ) + # Is this a closed/new connection that requires CONNECT tunnelling? + if self.proxy is not None and http_tunnel_required and conn.is_closed: + try: + self._prepare_proxy(conn) + except (BaseSSLError, OSError, SocketTimeout) as e: + self._raise_timeout( + err=e, url=self.proxy.url, timeout_value=conn.timeout + ) + raise # If we're going to release the connection in ``finally:``, then # the response doesn't need to know about the connection. Otherwise @@ -727,16 +786,20 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # mess. response_conn = conn if not release_conn else None - # Pass method to Response for length checking - response_kw["request_method"] = method - - # Import httplib's response into our own wrapper object - response = self.ResponseCls.from_httplib( - httplib_response, - pool=self, - connection=response_conn, + # Make the request on the HTTPConnection object + response = self._make_request( + conn, + method, + url, + timeout=timeout_obj, + body=body, + headers=headers, + chunked=chunked, retries=retries, - **response_kw + response_conn=response_conn, + preload_content=preload_content, + decode_content=decode_content, + **response_kw, ) # Everything went great! @@ -751,52 +814,35 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): except ( TimeoutError, HTTPException, - SocketError, + OSError, ProtocolError, BaseSSLError, SSLError, CertificateError, + ProxyError, ) as e: # Discard the connection for these exceptions. It will be # replaced during the next _get_conn() call. clean_exit = False - - 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) - elif isinstance(e, (SocketError, HTTPException)): - e = ProtocolError("Connection aborted.", e) + new_e: Exception = e + if isinstance(e, (BaseSSLError, CertificateError)): + new_e = SSLError(e) + if isinstance( + new_e, + ( + OSError, + NewConnectionError, + TimeoutError, + SSLError, + HTTPException, + ), + ) and (conn and conn.proxy and not conn.has_connected_to_proxy): + new_e = _wrap_proxy_error(new_e, conn.proxy.scheme) + elif isinstance(new_e, (OSError, HTTPException)): + new_e = ProtocolError("Connection aborted.", new_e) retries = retries.increment( - method, url, error=e, _pool=self, _stacktrace=sys.exc_info()[2] + method, url, error=new_e, _pool=self, _stacktrace=sys.exc_info()[2] ) retries.sleep() @@ -809,7 +855,9 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): # to throw the connection away unless explicitly told not to. # Close the connection, set the variable to None, and make sure # we put the None back in the pool to avoid leaking it. - conn = conn and conn.close() + if conn: + conn.close() + conn = None release_this_conn = True if release_this_conn: @@ -836,7 +884,9 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): release_conn=release_conn, chunked=chunked, body_pos=body_pos, - **response_kw + preload_content=preload_content, + decode_content=decode_content, + **response_kw, ) # Handle redirect? @@ -869,7 +919,9 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): release_conn=release_conn, chunked=chunked, body_pos=body_pos, - **response_kw + preload_content=preload_content, + decode_content=decode_content, + **response_kw, ) # Check if we should retry the HTTP response. @@ -899,7 +951,9 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods): release_conn=release_conn, chunked=chunked, body_pos=body_pos, - **response_kw + preload_content=preload_content, + decode_content=decode_content, + **response_kw, ) return response @@ -920,37 +974,35 @@ class HTTPSConnectionPool(HTTPConnectionPool): """ scheme = "https" - ConnectionCls = HTTPSConnection + ConnectionCls: type[BaseHTTPSConnection] = HTTPSConnection def __init__( self, - host, - port=None, - strict=False, - timeout=Timeout.DEFAULT_TIMEOUT, - maxsize=1, - block=False, - headers=None, - retries=None, - _proxy=None, - _proxy_headers=None, - key_file=None, - cert_file=None, - cert_reqs=None, - key_password=None, - ca_certs=None, - ssl_version=None, - assert_hostname=None, - assert_fingerprint=None, - ca_cert_dir=None, - **conn_kw - ): - - HTTPConnectionPool.__init__( - self, + host: str, + port: int | None = None, + timeout: _TYPE_TIMEOUT | None = _DEFAULT_TIMEOUT, + maxsize: int = 1, + block: bool = False, + headers: typing.Mapping[str, str] | None = None, + retries: Retry | bool | int | None = None, + _proxy: Url | None = None, + _proxy_headers: typing.Mapping[str, str] | None = None, + key_file: str | None = None, + cert_file: str | None = None, + cert_reqs: int | str | None = None, + key_password: str | None = None, + ca_certs: str | None = None, + ssl_version: int | str | None = None, + ssl_minimum_version: ssl.TLSVersion | None = None, + ssl_maximum_version: ssl.TLSVersion | None = None, + assert_hostname: str | Literal[False] | None = None, + assert_fingerprint: str | None = None, + ca_cert_dir: str | None = None, + **conn_kw: typing.Any, + ) -> None: + super().__init__( host, port, - strict, timeout, maxsize, block, @@ -958,7 +1010,7 @@ class HTTPSConnectionPool(HTTPConnectionPool): retries, _proxy, _proxy_headers, - **conn_kw + **conn_kw, ) self.key_file = key_file @@ -968,47 +1020,29 @@ class HTTPSConnectionPool(HTTPConnectionPool): self.ca_certs = ca_certs self.ca_cert_dir = ca_cert_dir self.ssl_version = ssl_version + self.ssl_minimum_version = ssl_minimum_version + self.ssl_maximum_version = ssl_maximum_version self.assert_hostname = assert_hostname self.assert_fingerprint = assert_fingerprint - def _prepare_conn(self, conn): - """ - Prepare the ``connection`` for :meth:`urllib3.util.ssl_wrap_socket` - and establish the tunnel if proxy is used. - """ - - if isinstance(conn, VerifiedHTTPSConnection): - conn.set_cert( - key_file=self.key_file, - key_password=self.key_password, - cert_file=self.cert_file, - cert_reqs=self.cert_reqs, - ca_certs=self.ca_certs, - ca_cert_dir=self.ca_cert_dir, - assert_hostname=self.assert_hostname, - assert_fingerprint=self.assert_fingerprint, - ) - conn.ssl_version = self.ssl_version - return conn - - def _prepare_proxy(self, conn): - """ - 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 + def _prepare_proxy(self, conn: HTTPSConnection) -> None: # type: ignore[override] + """Establishes a tunnel connection through HTTP CONNECT.""" + if self.proxy and self.proxy.scheme == "https": + tunnel_scheme = "https" + else: + tunnel_scheme = "http" + conn.set_tunnel( + scheme=tunnel_scheme, + host=self._tunnel_host, + port=self.port, + headers=self.proxy_headers, + ) conn.connect() - def _new_conn(self): + def _new_conn(self) -> BaseHTTPSConnection: """ - Return a fresh :class:`http.client.HTTPSConnection`. + Return a fresh :class:`urllib3.connection.HTTPConnection`. """ self.num_connections += 1 log.debug( @@ -1018,64 +1052,58 @@ class HTTPSConnectionPool(HTTPConnectionPool): self.port or "443", ) - if not self.ConnectionCls or self.ConnectionCls is DummyConnection: - raise SSLError( + if not self.ConnectionCls or self.ConnectionCls is DummyConnection: # type: ignore[comparison-overlap] + raise ImportError( "Can't connect to HTTPS URL because the SSL module is not available." ) - actual_host = self.host + actual_host: str = self.host actual_port = self.port - if self.proxy is not None: + if self.proxy is not None and self.proxy.host is not None: actual_host = self.proxy.host actual_port = self.proxy.port - conn = self.ConnectionCls( + return self.ConnectionCls( host=actual_host, port=actual_port, timeout=self.timeout.connect_timeout, - strict=self.strict, cert_file=self.cert_file, key_file=self.key_file, key_password=self.key_password, - **self.conn_kw + cert_reqs=self.cert_reqs, + ca_certs=self.ca_certs, + ca_cert_dir=self.ca_cert_dir, + assert_hostname=self.assert_hostname, + assert_fingerprint=self.assert_fingerprint, + ssl_version=self.ssl_version, + ssl_minimum_version=self.ssl_minimum_version, + ssl_maximum_version=self.ssl_maximum_version, + **self.conn_kw, ) - return self._prepare_conn(conn) - - def _validate_conn(self, conn): + def _validate_conn(self, conn: BaseHTTPConnection) -> None: """ Called right before a request is made, after the socket is created. """ - super(HTTPSConnectionPool, self)._validate_conn(conn) + super()._validate_conn(conn) # Force connect early to allow us to validate the connection. - if not getattr(conn, "sock", None): # AppEngine might not have `.sock` + if conn.is_closed: conn.connect() if not conn.is_verified: warnings.warn( ( - "Unverified HTTPS request is being made to host '%s'. " + f"Unverified HTTPS request is being made to host '{conn.host}'. " "Adding certificate verification is strongly advised. See: " - "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" + "https://urllib3.readthedocs.io/en/latest/advanced-usage.html" + "#tls-warnings" ), InsecureRequestWarning, ) -def connection_from_url(url, **kw): +def connection_from_url(url: str, **kw: typing.Any) -> HTTPConnectionPool: """ Given a url, return an :class:`.ConnectionPool` instance of its host. @@ -1095,15 +1123,26 @@ def connection_from_url(url, **kw): >>> conn = connection_from_url('http://google.com/') >>> r = conn.request('GET', '/') """ - scheme, host, port = get_host(url) + scheme, _, host, port, *_ = parse_url(url) + scheme = scheme or "http" port = port or port_by_scheme.get(scheme, 80) if scheme == "https": - return HTTPSConnectionPool(host, port=port, **kw) + return HTTPSConnectionPool(host, port=port, **kw) # type: ignore[arg-type] else: - return HTTPConnectionPool(host, port=port, **kw) + return HTTPConnectionPool(host, port=port, **kw) # type: ignore[arg-type] -def _normalize_host(host, scheme): +@typing.overload +def _normalize_host(host: None, scheme: str | None) -> None: + ... + + +@typing.overload +def _normalize_host(host: str, scheme: str | None) -> str: + ... + + +def _normalize_host(host: str | None, scheme: str | None) -> str | None: """ Normalize hosts for comparisons and use with sockets. """ @@ -1116,12 +1155,19 @@ def _normalize_host(host, scheme): # Instead, we need to make sure we never pass ``None`` as the port. # However, for backward compatibility reasons we can't actually # *assert* that. See http://bugs.python.org/issue28539 - if host.startswith("[") and host.endswith("]"): + if host and host.startswith("[") and host.endswith("]"): host = host[1:-1] return host -def _close_pool_connections(pool): +def _url_from_pool( + pool: HTTPConnectionPool | HTTPSConnectionPool, path: str | None = None +) -> str: + """Returns the URL from a given connection pool. This is mainly used for testing and logging.""" + return Url(scheme=pool.scheme, host=pool.host, port=pool.port, path=path).url + + +def _close_pool_connections(pool: queue.LifoQueue[typing.Any]) -> None: """Drains a queue of connections and closes each one.""" try: while True: diff --git a/lib/urllib3/contrib/__init__.pyi b/lib/urllib3/contrib/__init__.pyi deleted file mode 100644 index e69de29b..00000000 diff --git a/lib/urllib3/contrib/_appengine_environ.py b/lib/urllib3/contrib/_appengine_environ.py deleted file mode 100644 index 8765b907..00000000 --- a/lib/urllib3/contrib/_appengine_environ.py +++ /dev/null @@ -1,36 +0,0 @@ -""" -This module provides means to detect the App Engine environment. -""" - -import os - - -def is_appengine(): - return is_local_appengine() or is_prod_appengine() - - -def is_appengine_sandbox(): - """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 os.environ.get( - "SERVER_SOFTWARE", "" - ).startswith("Development/") - - -def is_prod_appengine(): - return "APPENGINE_RUNTIME" in os.environ and os.environ.get( - "SERVER_SOFTWARE", "" - ).startswith("Google App Engine/") - - -def is_prod_appengine_mvms(): - """Deprecated.""" - return False diff --git a/lib/urllib3/contrib/_securetransport/bindings.py b/lib/urllib3/contrib/_securetransport/bindings.py index 264d564d..3e4cd466 100644 --- a/lib/urllib3/contrib/_securetransport/bindings.py +++ b/lib/urllib3/contrib/_securetransport/bindings.py @@ -1,3 +1,5 @@ +# type: ignore + """ This module uses ctypes to bind a whole bunch of functions and constants from SecureTransport. The goal here is to provide the low-level API to @@ -29,7 +31,8 @@ license and by oscrypto's: FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. """ -from __future__ import absolute_import + +from __future__ import annotations import platform from ctypes import ( @@ -48,8 +51,6 @@ from ctypes import ( ) from ctypes.util import find_library -from ...packages.six import raise_from - if platform.system() != "Darwin": raise ImportError("Only macOS is supported") @@ -57,16 +58,16 @@ version = platform.mac_ver()[0] version_info = tuple(map(int, version.split("."))) if version_info < (10, 8): raise OSError( - "Only OS X 10.8 and newer are supported, not %s.%s" - % (version_info[0], version_info[1]) + f"Only OS X 10.8 and newer are supported, not {version_info[0]}.{version_info[1]}" ) -def load_cdll(name, macos10_16_path): +def load_cdll(name: str, macos10_16_path: str) -> CDLL: """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. + path: str | None if version_info >= (10, 16): path = macos10_16_path else: @@ -75,7 +76,7 @@ def load_cdll(name, macos10_16_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) + raise ImportError(f"The library {name} failed to load") from None Security = load_cdll( @@ -416,104 +417,14 @@ try: CoreFoundation.CFStringRef = CFStringRef CoreFoundation.CFDictionaryRef = CFDictionaryRef -except (AttributeError): - raise ImportError("Error initializing ctypes") +except AttributeError: + raise ImportError("Error initializing ctypes") from None -class CFConst(object): +class CFConst: """ A class object that acts as essentially a namespace for CoreFoundation constants. """ kCFStringEncodingUTF8 = CFStringEncoding(0x08000100) - - -class SecurityConst(object): - """ - A class object that acts as essentially a namespace for Security constants. - """ - - kSSLSessionOptionBreakOnServerAuth = 0 - - kSSLProtocol2 = 1 - kSSLProtocol3 = 2 - 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 - - kSSLClientSide = 1 - kSSLStreamType = 0 - - kSecFormatPEMSequence = 10 - - kSecTrustResultInvalid = 0 - kSecTrustResultProceed = 1 - # This gap is present on purpose: this was kSecTrustResultConfirm, which - # is deprecated. - kSecTrustResultDeny = 3 - kSecTrustResultUnspecified = 4 - kSecTrustResultRecoverableTrustFailure = 5 - kSecTrustResultFatalTrustFailure = 6 - kSecTrustResultOtherError = 7 - - errSSLProtocol = -9800 - errSSLWouldBlock = -9803 - errSSLClosedGraceful = -9805 - errSSLClosedNoNotify = -9816 - errSSLClosedAbort = -9806 - - errSSLXCertChainInvalid = -9807 - errSSLCrypto = -9809 - errSSLInternal = -9810 - errSSLCertExpired = -9814 - errSSLCertNotYetValid = -9815 - errSSLUnknownRootCert = -9812 - errSSLNoRootCert = -9813 - errSSLHostNameMismatch = -9843 - errSSLPeerHandshakeFail = -9824 - errSSLPeerUserCancelled = -9839 - errSSLWeakPeerEphemeralDHKey = -9850 - errSSLServerAuthCompleted = -9841 - errSSLRecordOverflow = -9847 - - errSecVerifyFailed = -67808 - errSecNoTrustSettings = -25263 - errSecItemNotFound = -25300 - errSecInvalidTrustSettings = -25262 - - # Cipher suites. We only pick the ones our default cipher string allows. - # Source: https://developer.apple.com/documentation/security/1550981-ssl_cipher_suite_values - TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384 = 0xC02C - TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384 = 0xC030 - TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256 = 0xC02B - TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256 = 0xC02F - TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256 = 0xCCA9 - TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256 = 0xCCA8 - TLS_DHE_RSA_WITH_AES_256_GCM_SHA384 = 0x009F - TLS_DHE_RSA_WITH_AES_128_GCM_SHA256 = 0x009E - TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA384 = 0xC024 - TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA384 = 0xC028 - TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA = 0xC00A - TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA = 0xC014 - TLS_DHE_RSA_WITH_AES_256_CBC_SHA256 = 0x006B - TLS_DHE_RSA_WITH_AES_256_CBC_SHA = 0x0039 - TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA256 = 0xC023 - TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA256 = 0xC027 - TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA = 0xC009 - TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA = 0xC013 - TLS_DHE_RSA_WITH_AES_128_CBC_SHA256 = 0x0067 - TLS_DHE_RSA_WITH_AES_128_CBC_SHA = 0x0033 - TLS_RSA_WITH_AES_256_GCM_SHA384 = 0x009D - TLS_RSA_WITH_AES_128_GCM_SHA256 = 0x009C - TLS_RSA_WITH_AES_256_CBC_SHA256 = 0x003D - TLS_RSA_WITH_AES_128_CBC_SHA256 = 0x003C - TLS_RSA_WITH_AES_256_CBC_SHA = 0x0035 - TLS_RSA_WITH_AES_128_CBC_SHA = 0x002F - TLS_AES_128_GCM_SHA256 = 0x1301 - TLS_AES_256_GCM_SHA384 = 0x1302 - TLS_AES_128_CCM_8_SHA256 = 0x1305 - TLS_AES_128_CCM_SHA256 = 0x1304 diff --git a/lib/urllib3/contrib/_securetransport/low_level.py b/lib/urllib3/contrib/_securetransport/low_level.py index fa0b245d..e2356997 100644 --- a/lib/urllib3/contrib/_securetransport/low_level.py +++ b/lib/urllib3/contrib/_securetransport/low_level.py @@ -7,6 +7,8 @@ CoreFoundation messing about and memory management. The concerns in this module are almost entirely about trying to avoid memory leaks and providing appropriate and useful assistance to the higher-level code. """ +from __future__ import annotations + import base64 import ctypes import itertools @@ -15,8 +17,20 @@ import re import ssl import struct import tempfile +import typing -from .bindings import CFConst, CoreFoundation, Security +from .bindings import ( # type: ignore[attr-defined] + CFArray, + CFConst, + CFData, + CFDictionary, + CFMutableArray, + CFString, + CFTypeRef, + CoreFoundation, + SecKeychainRef, + Security, +) # This regular expression is used to grab PEM data out of a PEM bundle. _PEM_CERTS_RE = re.compile( @@ -24,7 +38,7 @@ _PEM_CERTS_RE = re.compile( ) -def _cf_data_from_bytes(bytestring): +def _cf_data_from_bytes(bytestring: bytes) -> CFData: """ Given a bytestring, create a CFData object from it. This CFData object must be CFReleased by the caller. @@ -34,7 +48,9 @@ def _cf_data_from_bytes(bytestring): ) -def _cf_dictionary_from_tuples(tuples): +def _cf_dictionary_from_tuples( + tuples: list[tuple[typing.Any, typing.Any]] +) -> CFDictionary: """ Given a list of Python tuples, create an associated CFDictionary. """ @@ -56,7 +72,7 @@ def _cf_dictionary_from_tuples(tuples): ) -def _cfstr(py_bstr): +def _cfstr(py_bstr: bytes) -> CFString: """ Given a Python binary data, create a CFString. The string must be CFReleased by the caller. @@ -70,7 +86,7 @@ def _cfstr(py_bstr): return cf_str -def _create_cfstring_array(lst): +def _create_cfstring_array(lst: list[bytes]) -> CFMutableArray: """ Given a list of Python binary data, create an associated CFMutableArray. The array must be CFReleased by the caller. @@ -97,11 +113,11 @@ def _create_cfstring_array(lst): except BaseException as e: if cf_arr: CoreFoundation.CFRelease(cf_arr) - raise ssl.SSLError("Unable to allocate array: %s" % (e,)) + raise ssl.SSLError(f"Unable to allocate array: {e}") from None return cf_arr -def _cf_string_to_unicode(value): +def _cf_string_to_unicode(value: CFString) -> str | None: """ Creates a Unicode string from a CFString object. Used entirely for error reporting. @@ -123,10 +139,12 @@ def _cf_string_to_unicode(value): string = buffer.value if string is not None: string = string.decode("utf-8") - return string + return string # type: ignore[no-any-return] -def _assert_no_error(error, exception_class=None): +def _assert_no_error( + error: int, exception_class: type[BaseException] | None = None +) -> None: """ Checks the return code and throws an exception if there is an error to report @@ -138,8 +156,8 @@ def _assert_no_error(error, exception_class=None): output = _cf_string_to_unicode(cf_error_string) CoreFoundation.CFRelease(cf_error_string) - if output is None or output == u"": - output = u"OSStatus %s" % error + if output is None or output == "": + output = f"OSStatus {error}" if exception_class is None: exception_class = ssl.SSLError @@ -147,7 +165,7 @@ def _assert_no_error(error, exception_class=None): raise exception_class(output) -def _cert_array_from_pem(pem_bundle): +def _cert_array_from_pem(pem_bundle: bytes) -> CFArray: """ Given a bundle of certs in PEM format, turns them into a CFArray of certs that can be used to validate a cert chain. @@ -193,23 +211,23 @@ def _cert_array_from_pem(pem_bundle): return cert_array -def _is_cert(item): +def _is_cert(item: CFTypeRef) -> bool: """ Returns True if a given CFTypeRef is a certificate. """ expected = Security.SecCertificateGetTypeID() - return CoreFoundation.CFGetTypeID(item) == expected + return CoreFoundation.CFGetTypeID(item) == expected # type: ignore[no-any-return] -def _is_identity(item): +def _is_identity(item: CFTypeRef) -> bool: """ Returns True if a given CFTypeRef is an identity. """ expected = Security.SecIdentityGetTypeID() - return CoreFoundation.CFGetTypeID(item) == expected + return CoreFoundation.CFGetTypeID(item) == expected # type: ignore[no-any-return] -def _temporary_keychain(): +def _temporary_keychain() -> tuple[SecKeychainRef, str]: """ This function creates a temporary Mac keychain that we can use to work with credentials. This keychain uses a one-time password and a temporary file to @@ -244,7 +262,9 @@ def _temporary_keychain(): return keychain, tempdirectory -def _load_items_from_file(keychain, path): +def _load_items_from_file( + keychain: SecKeychainRef, path: str +) -> tuple[list[CFTypeRef], list[CFTypeRef]]: """ Given a single file, loads all the trust objects from it into arrays and the keychain. @@ -299,7 +319,7 @@ def _load_items_from_file(keychain, path): return (identities, certificates) -def _load_client_cert_chain(keychain, *paths): +def _load_client_cert_chain(keychain: SecKeychainRef, *paths: str | None) -> CFArray: """ Load certificates and maybe keys from a number of files. Has the end goal of returning a CFArray containing one SecIdentityRef, and then zero or more @@ -335,10 +355,10 @@ def _load_client_cert_chain(keychain, *paths): identities = [] # Filter out bad paths. - paths = (path for path in paths if path) + filtered_paths = (path for path in paths if path) try: - for file_path in paths: + for file_path in filtered_paths: new_identities, new_certs = _load_items_from_file(keychain, file_path) identities.extend(new_identities) certificates.extend(new_certs) @@ -383,7 +403,7 @@ TLS_PROTOCOL_VERSIONS = { } -def _build_tls_unknown_ca_alert(version): +def _build_tls_unknown_ca_alert(version: str) -> bytes: """ Builds a TLS alert record for an unknown CA. """ @@ -395,3 +415,60 @@ def _build_tls_unknown_ca_alert(version): record_type_alert = 0x15 record = struct.pack(">BBBH", record_type_alert, ver_maj, ver_min, msg_len) + msg return record + + +class SecurityConst: + """ + A class object that acts as essentially a namespace for Security constants. + """ + + kSSLSessionOptionBreakOnServerAuth = 0 + + kSSLProtocol2 = 1 + kSSLProtocol3 = 2 + 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 + + kSSLClientSide = 1 + kSSLStreamType = 0 + + kSecFormatPEMSequence = 10 + + kSecTrustResultInvalid = 0 + kSecTrustResultProceed = 1 + # This gap is present on purpose: this was kSecTrustResultConfirm, which + # is deprecated. + kSecTrustResultDeny = 3 + kSecTrustResultUnspecified = 4 + kSecTrustResultRecoverableTrustFailure = 5 + kSecTrustResultFatalTrustFailure = 6 + kSecTrustResultOtherError = 7 + + errSSLProtocol = -9800 + errSSLWouldBlock = -9803 + errSSLClosedGraceful = -9805 + errSSLClosedNoNotify = -9816 + errSSLClosedAbort = -9806 + + errSSLXCertChainInvalid = -9807 + errSSLCrypto = -9809 + errSSLInternal = -9810 + errSSLCertExpired = -9814 + errSSLCertNotYetValid = -9815 + errSSLUnknownRootCert = -9812 + errSSLNoRootCert = -9813 + errSSLHostNameMismatch = -9843 + errSSLPeerHandshakeFail = -9824 + errSSLPeerUserCancelled = -9839 + errSSLWeakPeerEphemeralDHKey = -9850 + errSSLServerAuthCompleted = -9841 + errSSLRecordOverflow = -9847 + + errSecVerifyFailed = -67808 + errSecNoTrustSettings = -25263 + errSecItemNotFound = -25300 + errSecInvalidTrustSettings = -25262 diff --git a/lib/urllib3/contrib/appengine.py b/lib/urllib3/contrib/appengine.py deleted file mode 100644 index a5a6d910..00000000 --- a/lib/urllib3/contrib/appengine.py +++ /dev/null @@ -1,314 +0,0 @@ -""" -This module provides a pool manager that uses Google App Engine's -`URLFetch Service `_. - -Example usage:: - - from urllib3 import PoolManager - from urllib3.contrib.appengine import AppEngineManager, is_appengine_sandbox - - if is_appengine_sandbox(): - # AppEngineManager uses AppEngine's URLFetch API behind the scenes - http = AppEngineManager() - else: - # PoolManager uses a socket-level API behind the scenes - http = PoolManager() - - r = http.request('GET', 'https://google.com/') - -There are `limitations `_ to the URLFetch service and it may not be -the best choice for your application. There are three options for using -urllib3 on Google App Engine: - -1. You can use :class:`AppEngineManager` with URLFetch. URLFetch is - cost-effective in many circumstances as long as your usage is within the - limitations. -2. You can use a normal :class:`~urllib3.PoolManager` by enabling sockets. - Sockets also have `limitations and restrictions - `_ and have a lower free quota than URLFetch. - To use sockets, be sure to specify the following in your ``app.yaml``:: - - env_variables: - GAE_USE_SOCKETS_HTTPLIB : 'true' - -3. If you are using `App Engine Flexible -`_, you can use the standard -:class:`PoolManager` without any configuration or special environment variables. -""" - -from __future__ import absolute_import - -import io -import logging -import warnings - -from ..exceptions import ( - HTTPError, - HTTPWarning, - MaxRetryError, - ProtocolError, - SSLError, - TimeoutError, -) -from ..packages.six.moves.urllib.parse import urljoin -from ..request import RequestMethods -from ..response import HTTPResponse -from ..util.retry import Retry -from ..util.timeout import Timeout -from . import _appengine_environ - -try: - from google.appengine.api import urlfetch -except ImportError: - urlfetch = None - - -log = logging.getLogger(__name__) - - -class AppEnginePlatformWarning(HTTPWarning): - pass - - -class AppEnginePlatformError(HTTPError): - pass - - -class AppEngineManager(RequestMethods): - """ - Connection manager for Google App Engine sandbox applications. - - This manager uses the URLFetch service directly instead of using the - emulated httplib, and is subject to URLFetch limitations as described in - the App Engine documentation `here - `_. - - Notably it will raise an :class:`AppEnginePlatformError` if: - * URLFetch is not available. - * 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 megabytes. - * If you use an unsupported request method such as OPTIONS. - - Beyond those cases, it will raise normal urllib3 errors. - """ - - def __init__( - self, - headers=None, - retries=None, - validate_certificate=True, - urlfetch_retries=True, - ): - if not urlfetch: - raise AppEnginePlatformError( - "URLFetch is not available 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/1.26.x/reference/urllib3.contrib.html.", - AppEnginePlatformWarning, - ) - - RequestMethods.__init__(self, headers) - self.validate_certificate = validate_certificate - self.urlfetch_retries = urlfetch_retries - - self.retries = retries or Retry.DEFAULT - - def __enter__(self): - return self - - def __exit__(self, exc_type, exc_val, exc_tb): - # Return False to re-raise any potential exceptions - return False - - def urlopen( - self, - method, - url, - body=None, - headers=None, - retries=None, - redirect=True, - timeout=Timeout.DEFAULT_TIMEOUT, - **response_kw - ): - - retries = self._get_retries(retries, redirect) - - try: - follow_redirects = redirect and retries.redirect != 0 and retries.total - response = urlfetch.fetch( - url, - payload=body, - method=method, - headers=headers or {}, - allow_truncated=False, - follow_redirects=self.urlfetch_retries and follow_redirects, - deadline=self._get_absolute_timeout(timeout), - validate_certificate=self.validate_certificate, - ) - except urlfetch.DeadlineExceededError as e: - raise TimeoutError(self, e) - - except urlfetch.InvalidURLError as e: - if "too large" in str(e): - raise AppEnginePlatformError( - "URLFetch request too large, URLFetch only " - "supports requests up to 10mb in size.", - e, - ) - raise ProtocolError(e) - - except urlfetch.DownloadError as e: - if "Too many redirects" in str(e): - raise MaxRetryError(self, url, reason=e) - raise ProtocolError(e) - - except urlfetch.ResponseTooLargeError as e: - raise AppEnginePlatformError( - "URLFetch response too large, URLFetch only supports" - "responses up to 32mb in size.", - e, - ) - - except urlfetch.SSLCertificateError as e: - raise SSLError(e) - - except urlfetch.InvalidMethodError as e: - raise AppEnginePlatformError( - "URLFetch does not support method: %s" % method, e - ) - - http_response = self._urlfetch_response_to_http_response( - response, retries=retries, **response_kw - ) - - # Handle redirect? - redirect_location = redirect and http_response.get_redirect_location() - if redirect_location: - # Check for redirect response - if self.urlfetch_retries and retries.raise_on_redirect: - raise MaxRetryError(self, url, "too many redirects") - else: - if http_response.status == 303: - method = "GET" - - try: - retries = retries.increment( - method, url, response=http_response, _pool=self - ) - except MaxRetryError: - if retries.raise_on_redirect: - raise MaxRetryError(self, url, "too many redirects") - return http_response - - retries.sleep_for_retry(http_response) - log.debug("Redirecting %s -> %s", url, redirect_location) - redirect_url = urljoin(url, redirect_location) - return self.urlopen( - method, - redirect_url, - body, - headers, - retries=retries, - redirect=redirect, - timeout=timeout, - **response_kw - ) - - # Check if we should retry the HTTP response. - 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) - retries.sleep(http_response) - return self.urlopen( - method, - url, - body=body, - headers=headers, - retries=retries, - redirect=redirect, - timeout=timeout, - **response_kw - ) - - return http_response - - def _urlfetch_response_to_http_response(self, urlfetch_resp, **response_kw): - - if is_prod_appengine(): - # Production GAE handles deflate encoding automatically, but does - # not remove the encoding header. - content_encoding = urlfetch_resp.headers.get("content-encoding") - - if content_encoding == "deflate": - del urlfetch_resp.headers["content-encoding"] - - transfer_encoding = urlfetch_resp.headers.get("transfer-encoding") - # We have a full response's content, - # so let's make sure we don't report ourselves as chunked data. - if transfer_encoding == "chunked": - encodings = transfer_encoding.split(",") - encodings.remove("chunked") - urlfetch_resp.headers["transfer-encoding"] = ",".join(encodings) - - original_response = HTTPResponse( - # In order for decoding to work, we must present the content as - # a file-like object. - body=io.BytesIO(urlfetch_resp.content), - msg=urlfetch_resp.header_msg, - headers=urlfetch_resp.headers, - status=urlfetch_resp.status_code, - **response_kw - ) - - return HTTPResponse( - body=io.BytesIO(urlfetch_resp.content), - headers=urlfetch_resp.headers, - status=urlfetch_resp.status_code, - original_response=original_response, - **response_kw - ) - - def _get_absolute_timeout(self, timeout): - if timeout is Timeout.DEFAULT_TIMEOUT: - return None # Defer to URLFetch's default. - if isinstance(timeout, Timeout): - if timeout._read is not None or timeout._connect is not None: - warnings.warn( - "URLFetch does not support granular timeout settings, " - "reverting to total or default URLFetch timeout.", - AppEnginePlatformWarning, - ) - return timeout.total - return timeout - - def _get_retries(self, retries, redirect): - if not isinstance(retries, Retry): - retries = Retry.from_int(retries, redirect=redirect, default=self.retries) - - if retries.connect or retries.read or retries.redirect: - warnings.warn( - "URLFetch only supports total retries and does not " - "recognize connect, read, or redirect retry parameters.", - AppEnginePlatformWarning, - ) - - return retries - - -# Alias methods from _appengine_environ to maintain public API interface. - -is_appengine = _appengine_environ.is_appengine -is_appengine_sandbox = _appengine_environ.is_appengine_sandbox -is_local_appengine = _appengine_environ.is_local_appengine -is_prod_appengine = _appengine_environ.is_prod_appengine -is_prod_appengine_mvms = _appengine_environ.is_prod_appengine_mvms diff --git a/lib/urllib3/contrib/ntlmpool.py b/lib/urllib3/contrib/ntlmpool.py deleted file mode 100644 index 47166575..00000000 --- a/lib/urllib3/contrib/ntlmpool.py +++ /dev/null @@ -1,130 +0,0 @@ -""" -NTLM authenticating pool, contributed by erikcederstran - -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__) - - -class NTLMConnectionPool(HTTPSConnectionPool): - """ - Implements an NTLM authentication version of an urllib3 connection pool - """ - - scheme = "https" - - def __init__(self, user, pw, authurl, *args, **kwargs): - """ - authurl is a random URL on the server that is protected by NTLM. - user is the Windows user, probably in the DOMAIN\\username format. - pw is the password for the user. - """ - super(NTLMConnectionPool, self).__init__(*args, **kwargs) - self.authurl = authurl - self.rawuser = user - user_parts = user.split("\\", 1) - self.domain = user_parts[0].upper() - self.user = user_parts[1] - self.pw = pw - - def _new_conn(self): - # Performs the NTLM handshake that secures the connection. The socket - # must be kept open while requests are performed. - self.num_connections += 1 - log.debug( - "Starting NTLM HTTPS connection no. %d: https://%s%s", - self.num_connections, - self.host, - self.authurl, - ) - - headers = {"Connection": "Keep-Alive"} - req_header = "Authorization" - resp_header = "www-authenticate" - - conn = HTTPSConnection(host=self.host, port=self.port) - - # Send negotiation message - headers[req_header] = "NTLM %s" % ntlm.create_NTLM_NEGOTIATE_MESSAGE( - self.rawuser - ) - log.debug("Request headers: %s", headers) - conn.request("GET", self.authurl, None, headers) - res = conn.getresponse() - 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)) - - # Remove the reference to the socket, so that it can not be closed by - # the response object (we want to keep the socket open) - res.fp = None - - # Server should respond with a challenge message - auth_header_values = reshdr[resp_header].split(", ") - auth_header_value = None - for s in auth_header_values: - if s[:5] == "NTLM ": - auth_header_value = s[5:] - if auth_header_value is None: - raise Exception( - "Unexpected %s response header: %s" % (resp_header, reshdr[resp_header]) - ) - - # Send authentication message - ServerChallenge, NegotiateFlags = ntlm.parse_NTLM_CHALLENGE_MESSAGE( - auth_header_value - ) - auth_msg = ntlm.create_NTLM_AUTHENTICATE_MESSAGE( - ServerChallenge, self.user, self.domain, self.pw, NegotiateFlags - ) - headers[req_header] = "NTLM %s" % auth_msg - log.debug("Request headers: %s", headers) - 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.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("Wrong server response: %s %s" % (res.status, res.reason)) - - res.fp = None - log.debug("Connection established") - return conn - - def urlopen( - self, - method, - url, - body=None, - headers=None, - retries=3, - redirect=True, - assert_same_host=True, - ): - if headers is None: - headers = {} - headers["Connection"] = "Keep-Alive" - return super(NTLMConnectionPool, self).urlopen( - method, url, body, headers, retries, redirect, assert_same_host - ) diff --git a/lib/urllib3/contrib/pyopenssl.py b/lib/urllib3/contrib/pyopenssl.py index 1ed214b1..0089cd27 100644 --- a/lib/urllib3/contrib/pyopenssl.py +++ b/lib/urllib3/contrib/pyopenssl.py @@ -1,8 +1,8 @@ """ -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. +Module for using pyOpenSSL as a TLS backend. This module was relevant before +the standard library ``ssl`` module supported SNI, but now that we've dropped +support for Python 2.7 all relevant Python versions support SNI so +**this module is no longer recommended**. This needs the following packages installed: @@ -10,7 +10,7 @@ This needs the following packages installed: * `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 +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: @@ -33,75 +33,55 @@ like this: except ImportError: pass -Now you can use :mod:`urllib3` as you normally would, and it will support SNI -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`_). - -.. _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 __future__ import annotations + +import OpenSSL.SSL # type: ignore[import] from cryptography import x509 -from cryptography.hazmat.backends.openssl import backend as openssl_backend try: - from cryptography.x509 import UnsupportedExtension + from cryptography.x509 import UnsupportedExtension # type: ignore[attr-defined] except ImportError: # UnsupportedExtension is gone in cryptography >= 2.1.0 - class UnsupportedExtension(Exception): + class UnsupportedExtension(Exception): # type: ignore[no-redef] pass -from io import BytesIO -from socket import error as SocketError -from socket import timeout - -try: # Platform-specific: Python 2 - from socket import _fileobject -except ImportError: # Platform-specific: Python 3 - _fileobject = None - from ..packages.backports.makefile import backport_makefile - import logging import ssl -import sys +import typing import warnings +from io import BytesIO +from socket import socket as socket_cls +from socket import timeout 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: " + "in urllib3 v2.1.0. Read more in this issue: " "https://github.com/urllib3/urllib3/issues/2680", category=DeprecationWarning, stacklevel=2, ) -__all__ = ["inject_into_urllib3", "extract_from_urllib3"] +if typing.TYPE_CHECKING: + from OpenSSL.crypto import X509 # type: ignore[import] -# SNI always works. -HAS_SNI = True + +__all__ = ["inject_into_urllib3", "extract_from_urllib3"] # Map from urllib3 to PyOpenSSL compatible parameter-values. _openssl_versions = { - util.PROTOCOL_TLS: OpenSSL.SSL.SSLv23_METHOD, - PROTOCOL_TLS_CLIENT: OpenSSL.SSL.SSLv23_METHOD, + util.ssl_.PROTOCOL_TLS: OpenSSL.SSL.SSLv23_METHOD, # type: ignore[attr-defined] + util.ssl_.PROTOCOL_TLS_CLIENT: OpenSSL.SSL.SSLv23_METHOD, # type: ignore[attr-defined] ssl.PROTOCOL_TLSv1: OpenSSL.SSL.TLSv1_METHOD, } -if hasattr(ssl, "PROTOCOL_SSLv3") and hasattr(OpenSSL.SSL, "SSLv3_METHOD"): - _openssl_versions[ssl.PROTOCOL_SSLv3] = OpenSSL.SSL.SSLv3_METHOD - if hasattr(ssl, "PROTOCOL_TLSv1_1") and hasattr(OpenSSL.SSL, "TLSv1_1_METHOD"): _openssl_versions[ssl.PROTOCOL_TLSv1_1] = OpenSSL.SSL.TLSv1_1_METHOD @@ -115,43 +95,77 @@ _stdlib_to_openssl_verify = { ssl.CERT_REQUIRED: OpenSSL.SSL.VERIFY_PEER + OpenSSL.SSL.VERIFY_FAIL_IF_NO_PEER_CERT, } -_openssl_to_stdlib_verify = dict((v, k) for k, v in _stdlib_to_openssl_verify.items()) +_openssl_to_stdlib_verify = {v: k for k, v in _stdlib_to_openssl_verify.items()} + +# The SSLvX values are the most likely to be missing in the future +# but we check them all just to be sure. +_OP_NO_SSLv2_OR_SSLv3: int = getattr(OpenSSL.SSL, "OP_NO_SSLv2", 0) | getattr( + OpenSSL.SSL, "OP_NO_SSLv3", 0 +) +_OP_NO_TLSv1: int = getattr(OpenSSL.SSL, "OP_NO_TLSv1", 0) +_OP_NO_TLSv1_1: int = getattr(OpenSSL.SSL, "OP_NO_TLSv1_1", 0) +_OP_NO_TLSv1_2: int = getattr(OpenSSL.SSL, "OP_NO_TLSv1_2", 0) +_OP_NO_TLSv1_3: int = getattr(OpenSSL.SSL, "OP_NO_TLSv1_3", 0) + +_openssl_to_ssl_minimum_version: dict[int, int] = { + ssl.TLSVersion.MINIMUM_SUPPORTED: _OP_NO_SSLv2_OR_SSLv3, + ssl.TLSVersion.TLSv1: _OP_NO_SSLv2_OR_SSLv3, + ssl.TLSVersion.TLSv1_1: _OP_NO_SSLv2_OR_SSLv3 | _OP_NO_TLSv1, + ssl.TLSVersion.TLSv1_2: _OP_NO_SSLv2_OR_SSLv3 | _OP_NO_TLSv1 | _OP_NO_TLSv1_1, + ssl.TLSVersion.TLSv1_3: ( + _OP_NO_SSLv2_OR_SSLv3 | _OP_NO_TLSv1 | _OP_NO_TLSv1_1 | _OP_NO_TLSv1_2 + ), + ssl.TLSVersion.MAXIMUM_SUPPORTED: ( + _OP_NO_SSLv2_OR_SSLv3 | _OP_NO_TLSv1 | _OP_NO_TLSv1_1 | _OP_NO_TLSv1_2 + ), +} +_openssl_to_ssl_maximum_version: dict[int, int] = { + ssl.TLSVersion.MINIMUM_SUPPORTED: ( + _OP_NO_SSLv2_OR_SSLv3 + | _OP_NO_TLSv1 + | _OP_NO_TLSv1_1 + | _OP_NO_TLSv1_2 + | _OP_NO_TLSv1_3 + ), + ssl.TLSVersion.TLSv1: ( + _OP_NO_SSLv2_OR_SSLv3 | _OP_NO_TLSv1_1 | _OP_NO_TLSv1_2 | _OP_NO_TLSv1_3 + ), + ssl.TLSVersion.TLSv1_1: _OP_NO_SSLv2_OR_SSLv3 | _OP_NO_TLSv1_2 | _OP_NO_TLSv1_3, + ssl.TLSVersion.TLSv1_2: _OP_NO_SSLv2_OR_SSLv3 | _OP_NO_TLSv1_3, + ssl.TLSVersion.TLSv1_3: _OP_NO_SSLv2_OR_SSLv3, + ssl.TLSVersion.MAXIMUM_SUPPORTED: _OP_NO_SSLv2_OR_SSLv3, +} # OpenSSL will only write 16K at a time SSL_WRITE_BLOCKSIZE = 16384 -orig_util_HAS_SNI = util.HAS_SNI orig_util_SSLContext = util.ssl_.SSLContext log = logging.getLogger(__name__) -def inject_into_urllib3(): +def inject_into_urllib3() -> None: "Monkey-patch urllib3 with PyOpenSSL-backed SSL-support." _validate_dependencies_met() - util.SSLContext = PyOpenSSLContext - util.ssl_.SSLContext = PyOpenSSLContext - util.HAS_SNI = HAS_SNI - util.ssl_.HAS_SNI = HAS_SNI + util.SSLContext = PyOpenSSLContext # type: ignore[assignment] + util.ssl_.SSLContext = PyOpenSSLContext # type: ignore[assignment] util.IS_PYOPENSSL = True util.ssl_.IS_PYOPENSSL = True -def extract_from_urllib3(): +def extract_from_urllib3() -> None: "Undo monkey-patching by :func:`inject_into_urllib3`." util.SSLContext = orig_util_SSLContext util.ssl_.SSLContext = orig_util_SSLContext - util.HAS_SNI = orig_util_HAS_SNI - util.ssl_.HAS_SNI = orig_util_HAS_SNI util.IS_PYOPENSSL = False util.ssl_.IS_PYOPENSSL = False -def _validate_dependencies_met(): +def _validate_dependencies_met() -> None: """ Verifies that PyOpenSSL's package-level dependencies have been met. Throws `ImportError` if they are not met. @@ -177,7 +191,7 @@ def _validate_dependencies_met(): ) -def _dnsname_to_stdlib(name): +def _dnsname_to_stdlib(name: str) -> str | None: """ Converts a dNSName SubjectAlternativeName field to the form used by the standard library on the given Python version. @@ -191,7 +205,7 @@ def _dnsname_to_stdlib(name): the name given should be skipped. """ - def idna_encode(name): + def idna_encode(name: str) -> bytes | None: """ Borrowed wholesale from the Python Cryptography Project. It turns out that we can't just safely call `idna.encode`: it can explode for @@ -200,7 +214,7 @@ def _dnsname_to_stdlib(name): import idna try: - for prefix in [u"*.", u"."]: + for prefix in ["*.", "."]: if name.startswith(prefix): name = name[len(prefix) :] return prefix.encode("ascii") + idna.encode(name) @@ -212,24 +226,17 @@ def _dnsname_to_stdlib(name): if ":" in name: return name - name = idna_encode(name) - if name is None: + encoded_name = idna_encode(name) + if encoded_name is None: return None - elif sys.version_info >= (3, 0): - name = name.decode("utf-8") - return name + return encoded_name.decode("utf-8") -def get_subj_alt_name(peer_cert): +def get_subj_alt_name(peer_cert: X509) -> list[tuple[str, str]]: """ Given an PyOpenSSL certificate, provides all the subject alternative names. """ - # Pass the cert to cryptography, which has much better APIs for this. - if hasattr(peer_cert, "to_cryptography"): - cert = peer_cert.to_cryptography() - else: - der = OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_ASN1, peer_cert) - cert = x509.load_der_x509_certificate(der, openssl_backend) + cert = peer_cert.to_cryptography() # We want to find the SAN extension. Ask Cryptography to locate it (it's # faster than looping in Python) @@ -273,93 +280,94 @@ def get_subj_alt_name(peer_cert): return names -class WrappedSocket(object): - """API-compatibility wrapper for Python OpenSSL's Connection-class. +class WrappedSocket: + """API-compatibility wrapper for Python OpenSSL's Connection-class.""" - Note: _makefile_refs, _drop() and _reuse() are needed for the garbage - collector of pypy. - """ - - def __init__(self, connection, socket, suppress_ragged_eofs=True): + def __init__( + self, + connection: OpenSSL.SSL.Connection, + socket: socket_cls, + suppress_ragged_eofs: bool = True, + ) -> None: self.connection = connection self.socket = socket self.suppress_ragged_eofs = suppress_ragged_eofs - self._makefile_refs = 0 + self._io_refs = 0 self._closed = False - def fileno(self): + def fileno(self) -> int: return self.socket.fileno() # Copy-pasted from Python 3.5 source code - def _decref_socketios(self): - if self._makefile_refs > 0: - self._makefile_refs -= 1 + def _decref_socketios(self) -> None: + if self._io_refs > 0: + self._io_refs -= 1 if self._closed: self.close() - def recv(self, *args, **kwargs): + def recv(self, *args: typing.Any, **kwargs: typing.Any) -> bytes: try: data = self.connection.recv(*args, **kwargs) except OpenSSL.SSL.SysCallError as e: if self.suppress_ragged_eofs and e.args == (-1, "Unexpected EOF"): return b"" else: - raise SocketError(str(e)) + raise OSError(e.args[0], str(e)) from e except OpenSSL.SSL.ZeroReturnError: if self.connection.get_shutdown() == OpenSSL.SSL.RECEIVED_SHUTDOWN: return b"" else: raise - except OpenSSL.SSL.WantReadError: + except OpenSSL.SSL.WantReadError as e: if not util.wait_for_read(self.socket, self.socket.gettimeout()): - raise timeout("The read operation timed out") + raise timeout("The read operation timed out") from e else: return self.recv(*args, **kwargs) # TLS 1.3 post-handshake authentication except OpenSSL.SSL.Error as e: - raise ssl.SSLError("read error: %r" % e) + raise ssl.SSLError(f"read error: {e!r}") from e else: - return data + return data # type: ignore[no-any-return] - def recv_into(self, *args, **kwargs): + def recv_into(self, *args: typing.Any, **kwargs: typing.Any) -> int: try: - return self.connection.recv_into(*args, **kwargs) + return self.connection.recv_into(*args, **kwargs) # type: ignore[no-any-return] except OpenSSL.SSL.SysCallError as e: if self.suppress_ragged_eofs and e.args == (-1, "Unexpected EOF"): return 0 else: - raise SocketError(str(e)) + raise OSError(e.args[0], str(e)) from e except OpenSSL.SSL.ZeroReturnError: if self.connection.get_shutdown() == OpenSSL.SSL.RECEIVED_SHUTDOWN: return 0 else: raise - except OpenSSL.SSL.WantReadError: + except OpenSSL.SSL.WantReadError as e: if not util.wait_for_read(self.socket, self.socket.gettimeout()): - raise timeout("The read operation timed out") + raise timeout("The read operation timed out") from e else: return self.recv_into(*args, **kwargs) # TLS 1.3 post-handshake authentication except OpenSSL.SSL.Error as e: - raise ssl.SSLError("read error: %r" % e) + raise ssl.SSLError(f"read error: {e!r}") from e - def settimeout(self, timeout): + def settimeout(self, timeout: float) -> None: return self.socket.settimeout(timeout) - def _send_until_done(self, data): + def _send_until_done(self, data: bytes) -> int: while True: try: - return self.connection.send(data) - except OpenSSL.SSL.WantWriteError: + return self.connection.send(data) # type: ignore[no-any-return] + except OpenSSL.SSL.WantWriteError as e: if not util.wait_for_write(self.socket, self.socket.gettimeout()): - raise timeout() + raise timeout() from e continue except OpenSSL.SSL.SysCallError as e: - raise SocketError(str(e)) + raise OSError(e.args[0], str(e)) from e - def sendall(self, data): + def sendall(self, data: bytes) -> None: total_sent = 0 while total_sent < len(data): sent = self._send_until_done( @@ -367,135 +375,135 @@ class WrappedSocket(object): ) total_sent += sent - def shutdown(self): + def shutdown(self) -> None: # FIXME rethrow compatible exceptions should we ever use this self.connection.shutdown() - def close(self): - if self._makefile_refs < 1: - try: - self._closed = True - return self.connection.close() - except OpenSSL.SSL.Error: - return - else: - self._makefile_refs -= 1 + def close(self) -> None: + self._closed = True + if self._io_refs <= 0: + self._real_close() - def getpeercert(self, binary_form=False): + def _real_close(self) -> None: + try: + return self.connection.close() # type: ignore[no-any-return] + except OpenSSL.SSL.Error: + return + + def getpeercert( + self, binary_form: bool = False + ) -> dict[str, list[typing.Any]] | None: x509 = self.connection.get_peer_certificate() if not x509: - return x509 + return x509 # type: ignore[no-any-return] if binary_form: - return OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_ASN1, x509) + return OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_ASN1, x509) # type: ignore[no-any-return] return { - "subject": ((("commonName", x509.get_subject().CN),),), + "subject": ((("commonName", x509.get_subject().CN),),), # type: ignore[dict-item] "subjectAltName": get_subj_alt_name(x509), } - def version(self): - return self.connection.get_protocol_version_name() - - def _reuse(self): - self._makefile_refs += 1 - - def _drop(self): - if self._makefile_refs < 1: - self.close() - else: - self._makefile_refs -= 1 + def version(self) -> str: + return self.connection.get_protocol_version_name() # type: ignore[no-any-return] -if _fileobject: # Platform-specific: Python 2 - - def makefile(self, mode, bufsize=-1): - self._makefile_refs += 1 - return _fileobject(self, mode, bufsize, close=True) - -else: # Platform-specific: Python 3 - makefile = backport_makefile - -WrappedSocket.makefile = makefile +WrappedSocket.makefile = socket_cls.makefile # type: ignore[attr-defined] -class PyOpenSSLContext(object): +class PyOpenSSLContext: """ I am a wrapper class for the PyOpenSSL ``Context`` object. I am responsible for translating the interface of the standard library ``SSLContext`` object to calls into PyOpenSSL. """ - def __init__(self, protocol): + def __init__(self, protocol: int) -> None: self.protocol = _openssl_versions[protocol] self._ctx = OpenSSL.SSL.Context(self.protocol) self._options = 0 self.check_hostname = False + self._minimum_version: int = ssl.TLSVersion.MINIMUM_SUPPORTED + self._maximum_version: int = ssl.TLSVersion.MAXIMUM_SUPPORTED @property - def options(self): + def options(self) -> int: return self._options @options.setter - def options(self, value): + def options(self, value: int) -> None: self._options = value - self._ctx.set_options(value) + self._set_ctx_options() @property - def verify_mode(self): + def verify_mode(self) -> int: return _openssl_to_stdlib_verify[self._ctx.get_verify_mode()] @verify_mode.setter - def verify_mode(self, value): + def verify_mode(self, value: ssl.VerifyMode) -> None: self._ctx.set_verify(_stdlib_to_openssl_verify[value], _verify_callback) - def set_default_verify_paths(self): + def set_default_verify_paths(self) -> None: self._ctx.set_default_verify_paths() - def set_ciphers(self, ciphers): - if isinstance(ciphers, six.text_type): + def set_ciphers(self, ciphers: bytes | str) -> None: + if isinstance(ciphers, str): ciphers = ciphers.encode("utf-8") self._ctx.set_cipher_list(ciphers) - def load_verify_locations(self, cafile=None, capath=None, cadata=None): + def load_verify_locations( + self, + cafile: str | None = None, + capath: str | None = None, + cadata: bytes | None = None, + ) -> None: if cafile is not None: - cafile = cafile.encode("utf-8") + cafile = cafile.encode("utf-8") # type: ignore[assignment] if capath is not None: - capath = capath.encode("utf-8") + capath = capath.encode("utf-8") # type: ignore[assignment] 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) + raise ssl.SSLError(f"unable to load trusted certificates: {e!r}") from e - def load_cert_chain(self, certfile, keyfile=None, password=None): - self._ctx.use_certificate_chain_file(certfile) - if password is not None: - if not isinstance(password, six.binary_type): - password = password.encode("utf-8") - self._ctx.set_passwd_cb(lambda *_: password) - self._ctx.use_privatekey_file(keyfile or certfile) + def load_cert_chain( + self, + certfile: str, + keyfile: str | None = None, + password: str | None = None, + ) -> None: + try: + self._ctx.use_certificate_chain_file(certfile) + if password is not None: + if not isinstance(password, bytes): + password = password.encode("utf-8") # type: ignore[assignment] + self._ctx.set_passwd_cb(lambda *_: password) + self._ctx.use_privatekey_file(keyfile or certfile) + except OpenSSL.SSL.Error as e: + raise ssl.SSLError(f"Unable to load certificate chain: {e!r}") from e - def set_alpn_protocols(self, protocols): - protocols = [six.ensure_binary(p) for p in protocols] - return self._ctx.set_alpn_protos(protocols) + def set_alpn_protocols(self, protocols: list[bytes | str]) -> None: + protocols = [util.util.to_bytes(p, "ascii") for p in protocols] + return self._ctx.set_alpn_protos(protocols) # type: ignore[no-any-return] def wrap_socket( self, - sock, - server_side=False, - do_handshake_on_connect=True, - suppress_ragged_eofs=True, - server_hostname=None, - ): + sock: socket_cls, + server_side: bool = False, + do_handshake_on_connect: bool = True, + suppress_ragged_eofs: bool = True, + server_hostname: bytes | str | None = None, + ) -> WrappedSocket: cnx = OpenSSL.SSL.Connection(self._ctx, sock) - if isinstance(server_hostname, six.text_type): # Platform-specific: Python 3 - server_hostname = server_hostname.encode("utf-8") - - if server_hostname is not None: + # If server_hostname is an IP, don't use it for SNI, per RFC6066 Section 3 + if server_hostname and not util.ssl_.is_ipaddress(server_hostname): + if isinstance(server_hostname, str): + server_hostname = server_hostname.encode("utf-8") cnx.set_tlsext_host_name(server_hostname) cnx.set_connect_state() @@ -503,16 +511,47 @@ class PyOpenSSLContext(object): while True: try: cnx.do_handshake() - except OpenSSL.SSL.WantReadError: + except OpenSSL.SSL.WantReadError as e: if not util.wait_for_read(sock, sock.gettimeout()): - raise timeout("select timed out") + raise timeout("select timed out") from e continue except OpenSSL.SSL.Error as e: - raise ssl.SSLError("bad handshake: %r" % e) + raise ssl.SSLError(f"bad handshake: {e!r}") from e break return WrappedSocket(cnx, sock) + def _set_ctx_options(self) -> None: + self._ctx.set_options( + self._options + | _openssl_to_ssl_minimum_version[self._minimum_version] + | _openssl_to_ssl_maximum_version[self._maximum_version] + ) -def _verify_callback(cnx, x509, err_no, err_depth, return_code): + @property + def minimum_version(self) -> int: + return self._minimum_version + + @minimum_version.setter + def minimum_version(self, minimum_version: int) -> None: + self._minimum_version = minimum_version + self._set_ctx_options() + + @property + def maximum_version(self) -> int: + return self._maximum_version + + @maximum_version.setter + def maximum_version(self, maximum_version: int) -> None: + self._maximum_version = maximum_version + self._set_ctx_options() + + +def _verify_callback( + cnx: OpenSSL.SSL.Connection, + x509: X509, + err_no: int, + err_depth: int, + return_code: int, +) -> bool: return err_no == 0 diff --git a/lib/urllib3/contrib/securetransport.py b/lib/urllib3/contrib/securetransport.py index 6c46a3b9..11beb3df 100644 --- a/lib/urllib3/contrib/securetransport.py +++ b/lib/urllib3/contrib/securetransport.py @@ -51,7 +51,8 @@ license and by oscrypto's: FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. """ -from __future__ import absolute_import + +from __future__ import annotations import contextlib import ctypes @@ -62,14 +63,18 @@ import socket import ssl import struct import threading +import typing +import warnings import weakref - -import six +from socket import socket as socket_cls from .. import util -from ..util.ssl_ import PROTOCOL_TLS_CLIENT -from ._securetransport.bindings import CoreFoundation, Security, SecurityConst +from ._securetransport.bindings import ( # type: ignore[attr-defined] + CoreFoundation, + Security, +) from ._securetransport.low_level import ( + SecurityConst, _assert_no_error, _build_tls_unknown_ca_alert, _cert_array_from_pem, @@ -78,18 +83,19 @@ from ._securetransport.low_level import ( _temporary_keychain, ) -try: # Platform-specific: Python 2 - from socket import _fileobject -except ImportError: # Platform-specific: Python 3 - _fileobject = None - from ..packages.backports.makefile import backport_makefile +warnings.warn( + "'urllib3.contrib.securetransport' module is deprecated and will be removed " + "in urllib3 v2.1.0. Read more in this issue: " + "https://github.com/urllib3/urllib3/issues/2681", + category=DeprecationWarning, + stacklevel=2, +) + +if typing.TYPE_CHECKING: + from typing_extensions import Literal __all__ = ["inject_into_urllib3", "extract_from_urllib3"] -# SNI always works -HAS_SNI = True - -orig_util_HAS_SNI = util.HAS_SNI orig_util_SSLContext = util.ssl_.SSLContext # This dictionary is used by the read callback to obtain a handle to the @@ -108,55 +114,24 @@ orig_util_SSLContext = util.ssl_.SSLContext # # This is good: if we had to lock in the callbacks we'd drastically slow down # the performance of this code. -_connection_refs = weakref.WeakValueDictionary() +_connection_refs: weakref.WeakValueDictionary[ + int, WrappedSocket +] = weakref.WeakValueDictionary() _connection_ref_lock = threading.Lock() # Limit writes to 16kB. This is OpenSSL's limit, but we'll cargo-cult it over # for no better reason than we need *a* limit, and this one is right there. SSL_WRITE_BLOCKSIZE = 16384 -# This is our equivalent of util.ssl_.DEFAULT_CIPHERS, but expanded out to -# individual cipher suites. We need to do this because this is how -# SecureTransport wants them. -CIPHER_SUITES = [ - SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384, - SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256, - SecurityConst.TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384, - SecurityConst.TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256, - SecurityConst.TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256, - SecurityConst.TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256, - SecurityConst.TLS_DHE_RSA_WITH_AES_256_GCM_SHA384, - SecurityConst.TLS_DHE_RSA_WITH_AES_128_GCM_SHA256, - SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA384, - SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA, - SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA256, - SecurityConst.TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA, - SecurityConst.TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA384, - SecurityConst.TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA, - SecurityConst.TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA256, - SecurityConst.TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA, - SecurityConst.TLS_DHE_RSA_WITH_AES_256_CBC_SHA256, - SecurityConst.TLS_DHE_RSA_WITH_AES_256_CBC_SHA, - SecurityConst.TLS_DHE_RSA_WITH_AES_128_CBC_SHA256, - SecurityConst.TLS_DHE_RSA_WITH_AES_128_CBC_SHA, - SecurityConst.TLS_AES_256_GCM_SHA384, - SecurityConst.TLS_AES_128_GCM_SHA256, - SecurityConst.TLS_RSA_WITH_AES_256_GCM_SHA384, - SecurityConst.TLS_RSA_WITH_AES_128_GCM_SHA256, - SecurityConst.TLS_AES_128_CCM_8_SHA256, - SecurityConst.TLS_AES_128_CCM_SHA256, - SecurityConst.TLS_RSA_WITH_AES_256_CBC_SHA256, - SecurityConst.TLS_RSA_WITH_AES_128_CBC_SHA256, - SecurityConst.TLS_RSA_WITH_AES_256_CBC_SHA, - SecurityConst.TLS_RSA_WITH_AES_128_CBC_SHA, -] - # Basically this is simple: for PROTOCOL_SSLv23 we turn it into a low of # 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.kTLSProtocol12), - PROTOCOL_TLS_CLIENT: (SecurityConst.kTLSProtocol1, SecurityConst.kTLSProtocol12), + util.ssl_.PROTOCOL_TLS: (SecurityConst.kTLSProtocol1, SecurityConst.kTLSProtocol12), # type: ignore[attr-defined] + util.ssl_.PROTOCOL_TLS_CLIENT: ( # type: ignore[attr-defined] + SecurityConst.kTLSProtocol1, + SecurityConst.kTLSProtocol12, + ), } if hasattr(ssl, "PROTOCOL_SSLv2"): @@ -186,31 +161,38 @@ if hasattr(ssl, "PROTOCOL_TLSv1_2"): ) -def inject_into_urllib3(): +_tls_version_to_st: dict[int, int] = { + ssl.TLSVersion.MINIMUM_SUPPORTED: SecurityConst.kTLSProtocol1, + ssl.TLSVersion.TLSv1: SecurityConst.kTLSProtocol1, + ssl.TLSVersion.TLSv1_1: SecurityConst.kTLSProtocol11, + ssl.TLSVersion.TLSv1_2: SecurityConst.kTLSProtocol12, + ssl.TLSVersion.MAXIMUM_SUPPORTED: SecurityConst.kTLSProtocol12, +} + + +def inject_into_urllib3() -> None: """ Monkey-patch urllib3 with SecureTransport-backed SSL-support. """ - util.SSLContext = SecureTransportContext - util.ssl_.SSLContext = SecureTransportContext - util.HAS_SNI = HAS_SNI - util.ssl_.HAS_SNI = HAS_SNI + util.SSLContext = SecureTransportContext # type: ignore[assignment] + util.ssl_.SSLContext = SecureTransportContext # type: ignore[assignment] util.IS_SECURETRANSPORT = True util.ssl_.IS_SECURETRANSPORT = True -def extract_from_urllib3(): +def extract_from_urllib3() -> None: """ Undo monkey-patching by :func:`inject_into_urllib3`. """ util.SSLContext = orig_util_SSLContext util.ssl_.SSLContext = orig_util_SSLContext - util.HAS_SNI = orig_util_HAS_SNI - util.ssl_.HAS_SNI = orig_util_HAS_SNI util.IS_SECURETRANSPORT = False util.ssl_.IS_SECURETRANSPORT = False -def _read_callback(connection_id, data_buffer, data_length_pointer): +def _read_callback( + connection_id: int, data_buffer: int, data_length_pointer: bytearray +) -> int: """ SecureTransport read callback. This is called by ST to request that data be returned from the socket. @@ -232,7 +214,7 @@ def _read_callback(connection_id, data_buffer, data_length_pointer): while read_count < requested_length: if timeout is None or timeout >= 0: if not util.wait_for_read(base_socket, timeout): - raise socket.error(errno.EAGAIN, "timed out") + raise OSError(errno.EAGAIN, "timed out") remaining = requested_length - read_count buffer = (ctypes.c_char * remaining).from_address( @@ -244,7 +226,7 @@ def _read_callback(connection_id, data_buffer, data_length_pointer): if not read_count: return SecurityConst.errSSLClosedGraceful break - except (socket.error) as e: + except OSError as e: error = e.errno if error is not None and error != errno.EAGAIN: @@ -265,7 +247,9 @@ def _read_callback(connection_id, data_buffer, data_length_pointer): return SecurityConst.errSSLInternal -def _write_callback(connection_id, data_buffer, data_length_pointer): +def _write_callback( + connection_id: int, data_buffer: int, data_length_pointer: bytearray +) -> int: """ SecureTransport write callback. This is called by ST to request that data actually be sent on the network. @@ -288,14 +272,14 @@ def _write_callback(connection_id, data_buffer, data_length_pointer): while sent < bytes_to_write: if timeout is None or timeout >= 0: if not util.wait_for_write(base_socket, timeout): - raise socket.error(errno.EAGAIN, "timed out") + raise OSError(errno.EAGAIN, "timed out") chunk_sent = base_socket.send(data) sent += chunk_sent # This has some needless copying here, but I'm not sure there's # much value in optimising this data path. data = data[chunk_sent:] - except (socket.error) as e: + except OSError as e: error = e.errno if error is not None and error != errno.EAGAIN: @@ -323,22 +307,20 @@ _read_callback_pointer = Security.SSLReadFunc(_read_callback) _write_callback_pointer = Security.SSLWriteFunc(_write_callback) -class WrappedSocket(object): +class WrappedSocket: """ API-compatibility wrapper for Python's OpenSSL wrapped socket object. - - Note: _makefile_refs, _drop(), and _reuse() are needed for the garbage - collector of PyPy. """ - def __init__(self, socket): + def __init__(self, socket: socket_cls) -> None: self.socket = socket self.context = None - self._makefile_refs = 0 + self._io_refs = 0 self._closed = False - self._exception = None + self._real_closed = False + self._exception: Exception | None = None self._keychain = None - self._keychain_dir = None + self._keychain_dir: str | None = None self._client_cert_chain = None # We save off the previously-configured timeout and then set it to @@ -350,7 +332,7 @@ class WrappedSocket(object): self.socket.settimeout(0) @contextlib.contextmanager - def _raise_on_error(self): + def _raise_on_error(self) -> typing.Generator[None, None, None]: """ A context manager that can be used to wrap calls that do I/O from SecureTransport. If any of the I/O callbacks hit an exception, this @@ -367,23 +349,10 @@ class WrappedSocket(object): yield if self._exception is not None: exception, self._exception = self._exception, None - self.close() + self._real_close() raise exception - def _set_ciphers(self): - """ - Sets up the allowed ciphers. By default this matches the set in - util.ssl_.DEFAULT_CIPHERS, at least as supported by macOS. This is done - custom and doesn't allow changing at this time, mostly because parsing - OpenSSL cipher strings is going to be a freaking nightmare. - """ - ciphers = (Security.SSLCipherSuite * len(CIPHER_SUITES))(*CIPHER_SUITES) - result = Security.SSLSetEnabledCiphers( - self.context, ciphers, len(CIPHER_SUITES) - ) - _assert_no_error(result) - - def _set_alpn_protocols(self, protocols): + def _set_alpn_protocols(self, protocols: list[bytes] | None) -> None: """ Sets up the ALPN protocols on the context. """ @@ -396,7 +365,7 @@ class WrappedSocket(object): finally: CoreFoundation.CFRelease(protocols_arr) - def _custom_validate(self, verify, trust_bundle): + def _custom_validate(self, verify: bool, trust_bundle: bytes | None) -> None: """ Called when we have set custom validation. We do this in two cases: first, when cert validation is entirely disabled; and second, when @@ -404,7 +373,7 @@ class WrappedSocket(object): Raises an SSLError if the connection is not trusted. """ # If we disabled cert validation, just say: cool. - if not verify: + if not verify or trust_bundle is None: return successes = ( @@ -415,10 +384,12 @@ class WrappedSocket(object): trust_result = self._evaluate_trust(trust_bundle) if trust_result in successes: return - reason = "error code: %d" % (trust_result,) + reason = f"error code: {int(trust_result)}" + exc = None except Exception as e: # Do not trust on error - reason = "exception: %r" % (e,) + reason = f"exception: {e!r}" + exc = e # SecureTransport does not send an alert nor shuts down the connection. rec = _build_tls_unknown_ca_alert(self.version()) @@ -428,10 +399,10 @@ class WrappedSocket(object): # 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) + self._real_close() + raise ssl.SSLError(f"certificate verify failed, {reason}") from exc - def _evaluate_trust(self, trust_bundle): + def _evaluate_trust(self, trust_bundle: bytes) -> int: # We want data in memory, so load it up. if os.path.isfile(trust_bundle): with open(trust_bundle, "rb") as f: @@ -469,20 +440,20 @@ class WrappedSocket(object): if cert_array is not None: CoreFoundation.CFRelease(cert_array) - return trust_result.value + return trust_result.value # type: ignore[no-any-return] def handshake( self, - server_hostname, - verify, - trust_bundle, - min_version, - max_version, - client_cert, - client_key, - client_key_passphrase, - alpn_protocols, - ): + server_hostname: bytes | str | None, + verify: bool, + trust_bundle: bytes | None, + min_version: int, + max_version: int, + client_cert: str | None, + client_key: str | None, + client_key_passphrase: typing.Any, + alpn_protocols: list[bytes] | None, + ) -> None: """ Actually performs the TLS handshake. This is run automatically by wrapped socket, and shouldn't be needed in user code. @@ -510,6 +481,8 @@ class WrappedSocket(object): _assert_no_error(result) # If we have a server hostname, we should set that too. + # RFC6066 Section 3 tells us not to use SNI when the host is an IP, but we have + # to do it anyway to match server_hostname against the server certificate if server_hostname: if not isinstance(server_hostname, bytes): server_hostname = server_hostname.encode("utf-8") @@ -519,9 +492,6 @@ class WrappedSocket(object): ) _assert_no_error(result) - # Setup the ciphers. - self._set_ciphers() - # Setup the ALPN protocols. self._set_alpn_protocols(alpn_protocols) @@ -564,25 +534,27 @@ class WrappedSocket(object): _assert_no_error(result) break - def fileno(self): + def fileno(self) -> int: return self.socket.fileno() # Copy-pasted from Python 3.5 source code - def _decref_socketios(self): - if self._makefile_refs > 0: - self._makefile_refs -= 1 + def _decref_socketios(self) -> None: + if self._io_refs > 0: + self._io_refs -= 1 if self._closed: self.close() - def recv(self, bufsiz): + def recv(self, bufsiz: int) -> bytes: buffer = ctypes.create_string_buffer(bufsiz) bytes_read = self.recv_into(buffer, bufsiz) data = buffer[:bytes_read] - return data + return typing.cast(bytes, data) - def recv_into(self, buffer, nbytes=None): + def recv_into( + self, buffer: ctypes.Array[ctypes.c_char], nbytes: int | None = None + ) -> int: # Read short on EOF. - if self._closed: + if self._real_closed: return 0 if nbytes is None: @@ -615,7 +587,7 @@ class WrappedSocket(object): # well. Note that we don't actually return here because in # principle this could actually be fired along with return data. # It's unlikely though. - self.close() + self._real_close() else: _assert_no_error(result) @@ -623,13 +595,13 @@ class WrappedSocket(object): # was actually read. return processed_bytes.value - def settimeout(self, timeout): + def settimeout(self, timeout: float) -> None: self._timeout = timeout - def gettimeout(self): + def gettimeout(self) -> float | None: return self._timeout - def send(self, data): + def send(self, data: bytes) -> int: processed_bytes = ctypes.c_size_t(0) with self._raise_on_error(): @@ -646,36 +618,38 @@ class WrappedSocket(object): # We sent, and probably succeeded. Tell them how much we sent. return processed_bytes.value - def sendall(self, data): + def sendall(self, data: bytes) -> None: total_sent = 0 while total_sent < len(data): sent = self.send(data[total_sent : total_sent + SSL_WRITE_BLOCKSIZE]) total_sent += sent - def shutdown(self): + def shutdown(self) -> None: with self._raise_on_error(): Security.SSLClose(self.context) - def close(self): + def close(self) -> None: + self._closed = True # TODO: should I do clean shutdown here? Do I have to? - if self._makefile_refs < 1: - self._closed = True - if self.context: - CoreFoundation.CFRelease(self.context) - self.context = None - if self._client_cert_chain: - CoreFoundation.CFRelease(self._client_cert_chain) - self._client_cert_chain = None - if self._keychain: - Security.SecKeychainDelete(self._keychain) - CoreFoundation.CFRelease(self._keychain) - shutil.rmtree(self._keychain_dir) - self._keychain = self._keychain_dir = None - return self.socket.close() - else: - self._makefile_refs -= 1 + if self._io_refs <= 0: + self._real_close() - def getpeercert(self, binary_form=False): + def _real_close(self) -> None: + self._real_closed = True + if self.context: + CoreFoundation.CFRelease(self.context) + self.context = None + if self._client_cert_chain: + CoreFoundation.CFRelease(self._client_cert_chain) + self._client_cert_chain = None + if self._keychain: + Security.SecKeychainDelete(self._keychain) + CoreFoundation.CFRelease(self._keychain) + shutil.rmtree(self._keychain_dir) + self._keychain = self._keychain_dir = None + return self.socket.close() + + def getpeercert(self, binary_form: bool = False) -> bytes | None: # Urgh, annoying. # # Here's how we do this: @@ -733,7 +707,7 @@ class WrappedSocket(object): return der_bytes - def version(self): + def version(self) -> str: protocol = Security.SSLProtocol() result = Security.SSLGetNegotiatedProtocolVersion( self.context, ctypes.byref(protocol) @@ -752,55 +726,50 @@ class WrappedSocket(object): elif protocol.value == SecurityConst.kSSLProtocol2: return "SSLv2" else: - raise ssl.SSLError("Unknown TLS version: %r" % protocol) - - def _reuse(self): - self._makefile_refs += 1 - - def _drop(self): - if self._makefile_refs < 1: - self.close() - else: - self._makefile_refs -= 1 + raise ssl.SSLError(f"Unknown TLS version: {protocol!r}") -if _fileobject: # Platform-specific: Python 2 - - def makefile(self, mode, bufsize=-1): - 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): - # We disable buffering with SecureTransport because it conflicts with - # the buffering that ST does internally (see issue #1153 for more). - buffering = 0 - return backport_makefile(self, mode, buffering, *args, **kwargs) +def makefile( + self: socket_cls, + mode: ( + Literal["r"] | Literal["w"] | Literal["rw"] | Literal["wr"] | Literal[""] + ) = "r", + buffering: int | None = None, + *args: typing.Any, + **kwargs: typing.Any, +) -> typing.BinaryIO | typing.TextIO: + # We disable buffering with SecureTransport because it conflicts with + # the buffering that ST does internally (see issue #1153 for more). + buffering = 0 + return socket_cls.makefile(self, mode, buffering, *args, **kwargs) -WrappedSocket.makefile = makefile +WrappedSocket.makefile = makefile # type: ignore[attr-defined] -class SecureTransportContext(object): +class SecureTransportContext: """ I am a wrapper class for the SecureTransport library, to translate the interface of the standard library ``SSLContext`` object to calls into SecureTransport. """ - def __init__(self, protocol): - self._min_version, self._max_version = _protocol_to_min_max[protocol] + def __init__(self, protocol: int) -> None: + self._minimum_version: int = ssl.TLSVersion.MINIMUM_SUPPORTED + self._maximum_version: int = ssl.TLSVersion.MAXIMUM_SUPPORTED + if protocol not in (None, ssl.PROTOCOL_TLS, ssl.PROTOCOL_TLS_CLIENT): + self._min_version, self._max_version = _protocol_to_min_max[protocol] + self._options = 0 self._verify = False - self._trust_bundle = None - self._client_cert = None - self._client_key = None + self._trust_bundle: bytes | None = None + self._client_cert: str | None = None + self._client_key: str | None = None self._client_key_passphrase = None - self._alpn_protocols = None + self._alpn_protocols: list[bytes] | None = None @property - def check_hostname(self): + def check_hostname(self) -> Literal[True]: """ SecureTransport cannot have its hostname checking disabled. For more, see the comment on getpeercert() in this file. @@ -808,15 +777,14 @@ class SecureTransportContext(object): return True @check_hostname.setter - def check_hostname(self, value): + def check_hostname(self, value: typing.Any) -> None: """ SecureTransport cannot have its hostname checking disabled. For more, see the comment on getpeercert() in this file. """ - pass @property - def options(self): + def options(self) -> int: # TODO: Well, crap. # # So this is the bit of the code that is the most likely to cause us @@ -826,19 +794,19 @@ class SecureTransportContext(object): return self._options @options.setter - def options(self, value): + def options(self, value: int) -> None: # TODO: Update in line with above. self._options = value @property - def verify_mode(self): + def verify_mode(self) -> int: return ssl.CERT_REQUIRED if self._verify else ssl.CERT_NONE @verify_mode.setter - def verify_mode(self, value): - self._verify = True if value == ssl.CERT_REQUIRED else False + def verify_mode(self, value: int) -> None: + self._verify = value == ssl.CERT_REQUIRED - def set_default_verify_paths(self): + def set_default_verify_paths(self) -> None: # So, this has to do something a bit weird. Specifically, what it does # is nothing. # @@ -850,15 +818,18 @@ class SecureTransportContext(object): # ignoring it. pass - def load_default_certs(self): + def load_default_certs(self) -> None: return self.set_default_verify_paths() - def set_ciphers(self, ciphers): - # For now, we just require the default cipher string. - if ciphers != util.ssl_.DEFAULT_CIPHERS: - raise ValueError("SecureTransport doesn't support custom cipher strings") + def set_ciphers(self, ciphers: typing.Any) -> None: + raise ValueError("SecureTransport doesn't support custom cipher strings") - def load_verify_locations(self, cafile=None, capath=None, cadata=None): + def load_verify_locations( + self, + cafile: str | None = None, + capath: str | None = None, + cadata: bytes | None = None, + ) -> None: # OK, we only really support cadata and cafile. if capath is not None: raise ValueError("SecureTransport does not support cert directories") @@ -868,14 +839,19 @@ class SecureTransportContext(object): with open(cafile): pass - self._trust_bundle = cafile or cadata + self._trust_bundle = cafile or cadata # type: ignore[assignment] - def load_cert_chain(self, certfile, keyfile=None, password=None): + def load_cert_chain( + self, + certfile: str, + keyfile: str | None = None, + password: str | None = None, + ) -> None: self._client_cert = certfile self._client_key = keyfile self._client_cert_passphrase = password - def set_alpn_protocols(self, protocols): + def set_alpn_protocols(self, protocols: list[str | bytes]) -> None: """ Sets the ALPN protocols that will later be set on the context. @@ -885,16 +861,16 @@ class SecureTransportContext(object): raise NotImplementedError( "SecureTransport supports ALPN only in macOS 10.12+" ) - self._alpn_protocols = [six.ensure_binary(p) for p in protocols] + self._alpn_protocols = [util.util.to_bytes(p, "ascii") for p in protocols] def wrap_socket( self, - sock, - server_side=False, - do_handshake_on_connect=True, - suppress_ragged_eofs=True, - server_hostname=None, - ): + sock: socket_cls, + server_side: bool = False, + do_handshake_on_connect: bool = True, + suppress_ragged_eofs: bool = True, + server_hostname: bytes | str | None = None, + ) -> WrappedSocket: # So, what do we do here? Firstly, we assert some properties. This is a # stripped down shim, so there is some functionality we don't support. # See PEP 543 for the real deal. @@ -911,11 +887,27 @@ class SecureTransportContext(object): server_hostname, self._verify, self._trust_bundle, - self._min_version, - self._max_version, + _tls_version_to_st[self._minimum_version], + _tls_version_to_st[self._maximum_version], self._client_cert, self._client_key, self._client_key_passphrase, self._alpn_protocols, ) return wrapped_socket + + @property + def minimum_version(self) -> int: + return self._minimum_version + + @minimum_version.setter + def minimum_version(self, minimum_version: int) -> None: + self._minimum_version = minimum_version + + @property + def maximum_version(self) -> int: + return self._maximum_version + + @maximum_version.setter + def maximum_version(self, maximum_version: int) -> None: + self._maximum_version = maximum_version diff --git a/lib/urllib3/contrib/socks.py b/lib/urllib3/contrib/socks.py index c326e80d..5e552dda 100644 --- a/lib/urllib3/contrib/socks.py +++ b/lib/urllib3/contrib/socks.py @@ -1,4 +1,3 @@ -# -*- coding: utf-8 -*- """ This module contains provisional support for SOCKS proxies from within urllib3. This module supports SOCKS4, SOCKS4A (an extension of SOCKS4), and @@ -38,10 +37,11 @@ with the proxy: proxy_url="socks5h://:@proxy-host" """ -from __future__ import absolute_import + +from __future__ import annotations try: - import socks + import socks # type: ignore[import] except ImportError: import warnings @@ -51,13 +51,13 @@ except ImportError: ( "SOCKS support in urllib3 requires the installation of optional " "dependencies: specifically, PySocks. For more information, see " - "https://urllib3.readthedocs.io/en/1.26.x/contrib.html#socks-proxies" + "https://urllib3.readthedocs.io/en/latest/contrib.html#socks-proxies" ), DependencyWarning, ) raise -from socket import error as SocketError +import typing from socket import timeout as SocketTimeout from ..connection import HTTPConnection, HTTPSConnection @@ -69,7 +69,21 @@ from ..util.url import parse_url try: import ssl except ImportError: - ssl = None + ssl = None # type: ignore[assignment] + +try: + from typing import TypedDict + + class _TYPE_SOCKS_OPTIONS(TypedDict): + socks_version: int + proxy_host: str | None + proxy_port: str | None + username: str | None + password: str | None + rdns: bool + +except ImportError: # Python 3.7 + _TYPE_SOCKS_OPTIONS = typing.Dict[str, typing.Any] # type: ignore[misc, assignment] class SOCKSConnection(HTTPConnection): @@ -77,15 +91,20 @@ class SOCKSConnection(HTTPConnection): A plain-text HTTP connection that connects via a SOCKS proxy. """ - def __init__(self, *args, **kwargs): - self._socks_options = kwargs.pop("_socks_options") - super(SOCKSConnection, self).__init__(*args, **kwargs) + def __init__( + self, + _socks_options: _TYPE_SOCKS_OPTIONS, + *args: typing.Any, + **kwargs: typing.Any, + ) -> None: + self._socks_options = _socks_options + super().__init__(*args, **kwargs) - def _new_conn(self): + def _new_conn(self) -> socks.socksocket: """ Establish a new connection via the SOCKS proxy. """ - extra_kw = {} + extra_kw: dict[str, typing.Any] = {} if self.source_address: extra_kw["source_address"] = self.source_address @@ -102,15 +121,14 @@ class SOCKSConnection(HTTPConnection): proxy_password=self._socks_options["password"], proxy_rdns=self._socks_options["rdns"], timeout=self.timeout, - **extra_kw + **extra_kw, ) - except SocketTimeout: + except SocketTimeout as e: raise ConnectTimeoutError( self, - "Connection to %s timed out. (connect timeout=%s)" - % (self.host, self.timeout), - ) + f"Connection to {self.host} timed out. (connect timeout={self.timeout})", + ) from e except socks.ProxyError as e: # This is fragile as hell, but it seems to be the only way to raise @@ -120,22 +138,23 @@ class SOCKSConnection(HTTPConnection): if isinstance(error, SocketTimeout): raise ConnectTimeoutError( self, - "Connection to %s timed out. (connect timeout=%s)" - % (self.host, self.timeout), - ) + f"Connection to {self.host} timed out. (connect timeout={self.timeout})", + ) from e else: + # Adding `from e` messes with coverage somehow, so it's omitted. + # See #2386. raise NewConnectionError( - self, "Failed to establish a new connection: %s" % error + self, f"Failed to establish a new connection: {error}" ) else: raise NewConnectionError( - self, "Failed to establish a new connection: %s" % e - ) + self, f"Failed to establish a new connection: {e}" + ) from e - except SocketError as e: # Defensive: PySocks should catch all these. + except OSError as e: # Defensive: PySocks should catch all these. raise NewConnectionError( - self, "Failed to establish a new connection: %s" % e - ) + self, f"Failed to establish a new connection: {e}" + ) from e return conn @@ -169,12 +188,12 @@ class SOCKSProxyManager(PoolManager): def __init__( self, - proxy_url, - username=None, - password=None, - num_pools=10, - headers=None, - **connection_pool_kw + proxy_url: str, + username: str | None = None, + password: str | None = None, + num_pools: int = 10, + headers: typing.Mapping[str, str] | None = None, + **connection_pool_kw: typing.Any, ): parsed = parse_url(proxy_url) @@ -195,7 +214,7 @@ class SOCKSProxyManager(PoolManager): socks_version = socks.PROXY_TYPE_SOCKS4 rdns = True else: - raise ValueError("Unable to determine SOCKS version from %s" % proxy_url) + raise ValueError(f"Unable to determine SOCKS version from {proxy_url}") self.proxy_url = proxy_url @@ -209,8 +228,6 @@ class SOCKSProxyManager(PoolManager): } connection_pool_kw["_socks_options"] = socks_options - super(SOCKSProxyManager, self).__init__( - num_pools, headers, **connection_pool_kw - ) + super().__init__(num_pools, headers, **connection_pool_kw) self.pool_classes_by_scheme = SOCKSProxyManager.pool_classes_by_scheme diff --git a/lib/urllib3/exceptions.py b/lib/urllib3/exceptions.py index cba6f3f5..5bb92369 100644 --- a/lib/urllib3/exceptions.py +++ b/lib/urllib3/exceptions.py @@ -1,6 +1,16 @@ -from __future__ import absolute_import +from __future__ import annotations -from .packages.six.moves.http_client import IncompleteRead as httplib_IncompleteRead +import socket +import typing +import warnings +from email.errors import MessageDefect +from http.client import IncompleteRead as httplib_IncompleteRead + +if typing.TYPE_CHECKING: + from .connection import HTTPConnection + from .connectionpool import ConnectionPool + from .response import HTTPResponse + from .util.retry import Retry # Base Exceptions @@ -8,23 +18,24 @@ from .packages.six.moves.http_client import IncompleteRead as httplib_Incomplete class HTTPError(Exception): """Base exception used by this module.""" - pass - class HTTPWarning(Warning): """Base warning used by this module.""" - pass + +_TYPE_REDUCE_RESULT = typing.Tuple[ + typing.Callable[..., object], typing.Tuple[object, ...] +] class PoolError(HTTPError): """Base exception for errors caused within a pool.""" - def __init__(self, pool, message): + def __init__(self, pool: ConnectionPool, message: str) -> None: self.pool = pool - HTTPError.__init__(self, "%s: %s" % (pool, message)) + super().__init__(f"{pool}: {message}") - def __reduce__(self): + def __reduce__(self) -> _TYPE_REDUCE_RESULT: # For pickling purposes. return self.__class__, (None, None) @@ -32,11 +43,11 @@ class PoolError(HTTPError): class RequestError(PoolError): """Base exception for PoolErrors that have associated URLs.""" - def __init__(self, pool, url, message): + def __init__(self, pool: ConnectionPool, url: str, message: str) -> None: self.url = url - PoolError.__init__(self, pool, message) + super().__init__(pool, message) - def __reduce__(self): + def __reduce__(self) -> _TYPE_REDUCE_RESULT: # For pickling purposes. return self.__class__, (None, self.url, None) @@ -44,28 +55,25 @@ class RequestError(PoolError): class SSLError(HTTPError): """Raised when SSL certificate fails in an HTTPS connection.""" - pass - class ProxyError(HTTPError): """Raised when the connection to a proxy fails.""" - def __init__(self, message, error, *args): - super(ProxyError, self).__init__(message, error, *args) + # The original error is also available as __cause__. + original_error: Exception + + def __init__(self, message: str, error: Exception) -> None: + super().__init__(message, error) self.original_error = error class DecodeError(HTTPError): """Raised when automatic decoding based on Content-Type fails.""" - pass - class ProtocolError(HTTPError): """Raised when something unexpected happens mid-request/response.""" - pass - #: Renamed to ProtocolError but aliased for backwards compatibility. ConnectionError = ProtocolError @@ -79,33 +87,36 @@ class MaxRetryError(RequestError): :param pool: The connection pool :type pool: :class:`~urllib3.connectionpool.HTTPConnectionPool` - :param string url: The requested Url - :param exceptions.Exception reason: The underlying error + :param str url: The requested Url + :param reason: The underlying error + :type reason: :class:`Exception` """ - def __init__(self, pool, url, reason=None): + def __init__( + self, pool: ConnectionPool, url: str, reason: Exception | None = None + ) -> None: self.reason = reason - message = "Max retries exceeded with url: %s (Caused by %r)" % (url, reason) + message = f"Max retries exceeded with url: {url} (Caused by {reason!r})" - RequestError.__init__(self, pool, url, message) + super().__init__(pool, url, message) class HostChangedError(RequestError): """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 - RequestError.__init__(self, pool, url, message) + def __init__( + self, pool: ConnectionPool, url: str, retries: Retry | int = 3 + ) -> None: + message = f"Tried to open a foreign host with url: {url}" + super().__init__(pool, url, message) self.retries = retries class TimeoutStateError(HTTPError): """Raised when passing an invalid state to a timeout""" - pass - class TimeoutError(HTTPError): """Raised when a socket timeout error occurs. @@ -114,53 +125,66 @@ class TimeoutError(HTTPError): ` and :exc:`ConnectTimeoutErrors `. """ - pass - class ReadTimeoutError(TimeoutError, RequestError): """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""" - pass - -class NewConnectionError(ConnectTimeoutError, PoolError): +class NewConnectionError(ConnectTimeoutError, HTTPError): """Raised when we fail to establish a new connection. Usually ECONNREFUSED.""" - pass + def __init__(self, conn: HTTPConnection, message: str) -> None: + self.conn = conn + super().__init__(f"{conn}: {message}") + + @property + def pool(self) -> HTTPConnection: + warnings.warn( + "The 'pool' property is deprecated and will be removed " + "in urllib3 v2.1.0. Use 'conn' instead.", + DeprecationWarning, + stacklevel=2, + ) + + return self.conn + + +class NameResolutionError(NewConnectionError): + """Raised when host name resolution fails.""" + + def __init__(self, host: str, conn: HTTPConnection, reason: socket.gaierror): + message = f"Failed to resolve '{host}' ({reason})" + super().__init__(conn, message) class EmptyPoolError(PoolError): """Raised when a pool runs out of connections and no more are allowed.""" - pass + +class FullPoolError(PoolError): + """Raised when we try to add a connection to a full pool in blocking mode.""" class ClosedPoolError(PoolError): """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.""" - pass - class LocationParseError(LocationValueError): """Raised when get_host or similar fails to parse the URL input.""" - def __init__(self, location): - message = "Failed to parse: %s" % location - HTTPError.__init__(self, message) + def __init__(self, location: str) -> None: + message = f"Failed to parse: {location}" + super().__init__(message) self.location = location @@ -168,9 +192,9 @@ class LocationParseError(LocationValueError): 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) + def __init__(self, scheme: str): + message = f"Not supported URL scheme {scheme}" + super().__init__(message) self.scheme = scheme @@ -185,38 +209,22 @@ class ResponseError(HTTPError): class SecurityWarning(HTTPWarning): """Warned when performing security reducing actions""" - pass - - -class SubjectAltNameWarning(SecurityWarning): - """Warned when connecting to a host with a certificate missing a SAN.""" - - pass - class InsecureRequestWarning(SecurityWarning): """Warned when making an unverified HTTPS request.""" - pass + +class NotOpenSSLWarning(SecurityWarning): + """Warned when using unsupported SSL library""" class SystemTimeWarning(SecurityWarning): """Warned when system time is suspected to be wrong""" - pass - class InsecurePlatformWarning(SecurityWarning): """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.""" - - pass - class DependencyWarning(HTTPWarning): """ @@ -224,14 +232,10 @@ class DependencyWarning(HTTPWarning): dependencies. """ - pass - class ResponseNotChunked(ProtocolError, ValueError): """Response needs to be chunked in order to read it as chunks.""" - pass - class BodyNotHttplibCompatible(HTTPError): """ @@ -239,8 +243,6 @@ class BodyNotHttplibCompatible(HTTPError): (have an fp attribute which returns raw chunks) for read_chunked(). """ - pass - class IncompleteRead(HTTPError, httplib_IncompleteRead): """ @@ -250,12 +252,13 @@ class IncompleteRead(HTTPError, httplib_IncompleteRead): for ``partial`` to avoid creating large objects on streamed reads. """ - def __init__(self, partial, expected): - super(IncompleteRead, self).__init__(partial, expected) + def __init__(self, partial: int, expected: int) -> None: + self.partial = partial # type: ignore[assignment] + self.expected = expected - def __repr__(self): + def __repr__(self) -> str: return "IncompleteRead(%i bytes read, %i more expected)" % ( - self.partial, + self.partial, # type: ignore[str-format] self.expected, ) @@ -263,14 +266,13 @@ class IncompleteRead(HTTPError, httplib_IncompleteRead): 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 - ) + def __init__(self, response: HTTPResponse, length: bytes) -> None: + self.partial: int = response.tell() # type: ignore[assignment] + self.expected: int | None = response.length_remaining self.response = response self.length = length - def __repr__(self): + def __repr__(self) -> str: return "InvalidChunkLength(got length %r, %i bytes read)" % ( self.length, self.partial, @@ -280,15 +282,13 @@ class InvalidChunkLength(HTTPError, httplib_IncompleteRead): class InvalidHeader(HTTPError): """The header provided was somehow invalid.""" - pass - 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): + def __init__(self, scheme: str | None) -> None: # 'localhost' is here because our URL parser parses # localhost:8080 -> scheme=localhost, remove if we fix this. if scheme == "localhost": @@ -296,28 +296,23 @@ class ProxySchemeUnknown(AssertionError, URLSchemeUnknown): 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) + message = f"Proxy URL had unsupported scheme {scheme}, should use http:// or https://" + super().__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.""" - def __init__(self, defects, unparsed_data): - message = "%s, unparsed data: %r" % (defects or "Unknown", unparsed_data) - super(HeaderParsingError, self).__init__(message) + def __init__( + self, defects: list[MessageDefect], unparsed_data: bytes | str | None + ) -> None: + message = f"{defects or 'Unknown'}, unparsed data: {unparsed_data!r}" + super().__init__(message) class UnrewindableBodyError(HTTPError): """urllib3 encountered an error when trying to rewind a body""" - - pass diff --git a/lib/urllib3/exceptions.pyi b/lib/urllib3/exceptions.pyi deleted file mode 100644 index ca528b09..00000000 --- a/lib/urllib3/exceptions.pyi +++ /dev/null @@ -1,55 +0,0 @@ -from typing import TYPE_CHECKING, Any, Optional, Tuple, Union - -if TYPE_CHECKING: - from urllib3.connectionpool import ConnectionPool - -class HTTPError(Exception): ... -class HTTPWarning(Warning): ... - -class PoolError(HTTPError): - pool: ConnectionPool - def __init__(self, pool: ConnectionPool, message: str) -> None: ... - def __reduce__(self) -> Union[str, Tuple[Any, ...]]: ... - -class RequestError(PoolError): - url: str - def __init__(self, pool: ConnectionPool, url: str, message: str) -> None: ... - def __reduce__(self) -> Union[str, Tuple[Any, ...]]: ... - -class SSLError(HTTPError): ... -class ProxyError(HTTPError): ... -class DecodeError(HTTPError): ... -class ProtocolError(HTTPError): ... - -ConnectionError: ProtocolError - -class MaxRetryError(RequestError): - reason: str - def __init__( - self, pool: ConnectionPool, url: str, reason: Optional[str] - ) -> None: ... - -class HostChangedError(RequestError): - retries: int - def __init__(self, pool: ConnectionPool, url: str, retries: int) -> None: ... - -class TimeoutStateError(HTTPError): ... -class TimeoutError(HTTPError): ... -class ReadTimeoutError(TimeoutError, RequestError): ... -class ConnectTimeoutError(TimeoutError): ... -class EmptyPoolError(PoolError): ... -class ClosedPoolError(PoolError): ... -class LocationValueError(ValueError, HTTPError): ... - -class LocationParseError(LocationValueError): - location: str - def __init__(self, location: str) -> None: ... - -class ResponseError(HTTPError): - GENERIC_ERROR: Any - SPECIFIC_ERROR: Any - -class SecurityWarning(HTTPWarning): ... -class InsecureRequestWarning(SecurityWarning): ... -class SystemTimeWarning(SecurityWarning): ... -class InsecurePlatformWarning(SecurityWarning): ... diff --git a/lib/urllib3/fields.py b/lib/urllib3/fields.py index 9d630f49..51d898e2 100644 --- a/lib/urllib3/fields.py +++ b/lib/urllib3/fields.py @@ -1,13 +1,20 @@ -from __future__ import absolute_import +from __future__ import annotations import email.utils import mimetypes -import re +import typing -from .packages import six +_TYPE_FIELD_VALUE = typing.Union[str, bytes] +_TYPE_FIELD_VALUE_TUPLE = typing.Union[ + _TYPE_FIELD_VALUE, + typing.Tuple[str, _TYPE_FIELD_VALUE], + typing.Tuple[str, _TYPE_FIELD_VALUE, str], +] -def guess_content_type(filename, default="application/octet-stream"): +def guess_content_type( + filename: str | None, default: str = "application/octet-stream" +) -> str: """ Guess the "Content-Type" of a file. @@ -21,7 +28,7 @@ def guess_content_type(filename, default="application/octet-stream"): return default -def format_header_param_rfc2231(name, value): +def format_header_param_rfc2231(name: str, value: _TYPE_FIELD_VALUE) -> str: """ Helper function to format and quote a single header parameter using the strategy defined in RFC 2231. @@ -34,14 +41,28 @@ def format_header_param_rfc2231(name, value): The name of the parameter, a string expected to be ASCII only. :param value: The value of the parameter, provided as ``bytes`` or `str``. - :ret: + :returns: An RFC-2231-formatted unicode string. + + .. deprecated:: 2.0.0 + Will be removed in urllib3 v2.1.0. This is not valid for + ``multipart/form-data`` header parameters. """ - if isinstance(value, six.binary_type): + import warnings + + warnings.warn( + "'format_header_param_rfc2231' is deprecated and will be " + "removed in urllib3 v2.1.0. This is not valid for " + "multipart/form-data header parameters.", + DeprecationWarning, + stacklevel=2, + ) + + if isinstance(value, bytes): value = value.decode("utf-8") if not any(ch in value for ch in '"\\\r\n'): - result = u'%s="%s"' % (name, value) + result = f'{name}="{value}"' try: result.encode("ascii") except (UnicodeEncodeError, UnicodeDecodeError): @@ -49,81 +70,87 @@ def format_header_param_rfc2231(name, value): else: return result - if six.PY2: # Python 2: - value = value.encode("utf-8") - - # encode_rfc2231 accepts an encoded string and returns an ascii-encoded - # string in Python 2 but accepts and returns unicode strings in Python 3 value = email.utils.encode_rfc2231(value, "utf-8") - value = "%s*=%s" % (name, value) - - if six.PY2: # Python 2: - value = value.decode("utf-8") + value = f"{name}*={value}" return value -_HTML5_REPLACEMENTS = { - u"\u0022": u"%22", - # Replace "\" with "\\". - u"\u005C": u"\u005C\u005C", -} - -# All control characters from 0x00 to 0x1F *except* 0x1B. -_HTML5_REPLACEMENTS.update( - { - six.unichr(cc): u"%{:02X}".format(cc) - for cc in range(0x00, 0x1F + 1) - if cc not in (0x1B,) - } -) - - -def _replace_multiple(value, needles_and_replacements): - def replacer(match): - return needles_and_replacements[match.group(0)] - - pattern = re.compile( - r"|".join([re.escape(needle) for needle in needles_and_replacements.keys()]) - ) - - result = pattern.sub(replacer, value) - - return result - - -def format_header_param_html5(name, value): +def format_multipart_header_param(name: str, value: _TYPE_FIELD_VALUE) -> str: """ - Helper function to format and quote a single header parameter using the - HTML5 strategy. + Format and quote a single multipart header parameter. - Particularly useful for header parameters which might contain - non-ASCII values, like file names. This follows the `HTML5 Working Draft - Section 4.10.22.7`_ and matches the behavior of curl and modern browsers. + This follows the `WHATWG HTML Standard`_ as of 2021/06/10, matching + the behavior of current browser and curl versions. Values are + assumed to be UTF-8. The ``\\n``, ``\\r``, and ``"`` characters are + percent encoded. - .. _HTML5 Working Draft Section 4.10.22.7: - https://w3c.github.io/html/sec-forms.html#multipart-form-data + .. _WHATWG HTML Standard: + https://html.spec.whatwg.org/multipage/ + form-control-infrastructure.html#multipart-form-data :param name: - The name of the parameter, a string expected to be ASCII only. + The name of the parameter, an ASCII-only ``str``. :param value: - The value of the parameter, provided as ``bytes`` or `str``. - :ret: - A unicode string, stripped of troublesome characters. + The value of the parameter, a ``str`` or UTF-8 encoded + ``bytes``. + :returns: + A string ``name="value"`` with the escaped value. + + .. versionchanged:: 2.0.0 + Matches the WHATWG HTML Standard as of 2021/06/10. Control + characters are no longer percent encoded. + + .. versionchanged:: 2.0.0 + Renamed from ``format_header_param_html5`` and + ``format_header_param``. The old names will be removed in + urllib3 v2.1.0. """ - if isinstance(value, six.binary_type): + if isinstance(value, bytes): value = value.decode("utf-8") - value = _replace_multiple(value, _HTML5_REPLACEMENTS) - - return u'%s="%s"' % (name, value) + # percent encode \n \r " + value = value.translate({10: "%0A", 13: "%0D", 34: "%22"}) + return f'{name}="{value}"' -# For backwards-compatibility. -format_header_param = format_header_param_html5 +def format_header_param_html5(name: str, value: _TYPE_FIELD_VALUE) -> str: + """ + .. deprecated:: 2.0.0 + Renamed to :func:`format_multipart_header_param`. Will be + removed in urllib3 v2.1.0. + """ + import warnings + + warnings.warn( + "'format_header_param_html5' has been renamed to " + "'format_multipart_header_param'. The old name will be " + "removed in urllib3 v2.1.0.", + DeprecationWarning, + stacklevel=2, + ) + return format_multipart_header_param(name, value) -class RequestField(object): +def format_header_param(name: str, value: _TYPE_FIELD_VALUE) -> str: + """ + .. deprecated:: 2.0.0 + Renamed to :func:`format_multipart_header_param`. Will be + removed in urllib3 v2.1.0. + """ + import warnings + + warnings.warn( + "'format_header_param' has been renamed to " + "'format_multipart_header_param'. The old name will be " + "removed in urllib3 v2.1.0.", + DeprecationWarning, + stacklevel=2, + ) + return format_multipart_header_param(name, value) + + +class RequestField: """ A data container for request body parameters. @@ -135,29 +162,47 @@ class RequestField(object): An optional filename of the request field. Must be unicode. :param headers: An optional dict-like object of headers to initially use for the field. - :param header_formatter: - An optional callable that is used to encode and format the headers. By - default, this is :func:`format_header_param_html5`. + + .. versionchanged:: 2.0.0 + The ``header_formatter`` parameter is deprecated and will + be removed in urllib3 v2.1.0. """ def __init__( self, - name, - data, - filename=None, - headers=None, - header_formatter=format_header_param_html5, + name: str, + data: _TYPE_FIELD_VALUE, + filename: str | None = None, + headers: typing.Mapping[str, str] | None = None, + header_formatter: typing.Callable[[str, _TYPE_FIELD_VALUE], str] | None = None, ): self._name = name self._filename = filename self.data = data - self.headers = {} + self.headers: dict[str, str | None] = {} if headers: self.headers = dict(headers) - self.header_formatter = header_formatter + + if header_formatter is not None: + import warnings + + warnings.warn( + "The 'header_formatter' parameter is deprecated and " + "will be removed in urllib3 v2.1.0.", + DeprecationWarning, + stacklevel=2, + ) + self.header_formatter = header_formatter + else: + self.header_formatter = format_multipart_header_param @classmethod - def from_tuples(cls, fieldname, value, header_formatter=format_header_param_html5): + def from_tuples( + cls, + fieldname: str, + value: _TYPE_FIELD_VALUE_TUPLE, + header_formatter: typing.Callable[[str, _TYPE_FIELD_VALUE], str] | None = None, + ) -> RequestField: """ A :class:`~urllib3.fields.RequestField` factory from old-style tuple parameters. @@ -174,11 +219,19 @@ class RequestField(object): Field names and filenames must be unicode. """ + filename: str | None + content_type: str | None + data: _TYPE_FIELD_VALUE + if isinstance(value, tuple): if len(value) == 3: - filename, data, content_type = value + filename, data, content_type = typing.cast( + typing.Tuple[str, _TYPE_FIELD_VALUE, str], value + ) else: - filename, data = value + filename, data = typing.cast( + typing.Tuple[str, _TYPE_FIELD_VALUE], value + ) content_type = guess_content_type(filename) else: filename = None @@ -192,20 +245,29 @@ class RequestField(object): return request_param - def _render_part(self, name, value): + def _render_part(self, name: str, value: _TYPE_FIELD_VALUE) -> str: """ - Overridable helper function to format a single header parameter. By - default, this calls ``self.header_formatter``. + Override this method to change how each multipart header + parameter is formatted. By default, this calls + :func:`format_multipart_header_param`. :param name: - The name of the parameter, a string expected to be ASCII only. + The name of the parameter, an ASCII-only ``str``. :param value: - The value of the parameter, provided as a unicode string. - """ + The value of the parameter, a ``str`` or UTF-8 encoded + ``bytes``. + :meta public: + """ return self.header_formatter(name, value) - def _render_parts(self, header_parts): + def _render_parts( + self, + header_parts: ( + dict[str, _TYPE_FIELD_VALUE | None] + | typing.Sequence[tuple[str, _TYPE_FIELD_VALUE | None]] + ), + ) -> str: """ Helper function to format and quote a single header. @@ -216,18 +278,21 @@ class RequestField(object): A sequence of (k, v) tuples or a :class:`dict` of (k, v) to format as `k1="v1"; k2="v2"; ...`. """ + iterable: typing.Iterable[tuple[str, _TYPE_FIELD_VALUE | None]] + parts = [] - iterable = header_parts if isinstance(header_parts, dict): iterable = header_parts.items() + else: + iterable = header_parts for name, value in iterable: if value is not None: parts.append(self._render_part(name, value)) - return u"; ".join(parts) + return "; ".join(parts) - def render_headers(self): + def render_headers(self) -> str: """ Renders the headers for this request field. """ @@ -236,39 +301,45 @@ class RequestField(object): sort_keys = ["Content-Disposition", "Content-Type", "Content-Location"] for sort_key in sort_keys: if self.headers.get(sort_key, False): - lines.append(u"%s: %s" % (sort_key, self.headers[sort_key])) + lines.append(f"{sort_key}: {self.headers[sort_key]}") for header_name, header_value in self.headers.items(): if header_name not in sort_keys: if header_value: - lines.append(u"%s: %s" % (header_name, header_value)) + lines.append(f"{header_name}: {header_value}") - lines.append(u"\r\n") - return u"\r\n".join(lines) + lines.append("\r\n") + return "\r\n".join(lines) def make_multipart( - self, content_disposition=None, content_type=None, content_location=None - ): + self, + content_disposition: str | None = None, + content_type: str | None = None, + content_location: str | None = None, + ) -> None: """ Makes this request field into a multipart request field. This method overrides "Content-Disposition", "Content-Type" and "Content-Location" headers to the request parameter. + :param content_disposition: + The 'Content-Disposition' of the request body. Defaults to 'form-data' :param content_type: The 'Content-Type' of the request body. :param content_location: The 'Content-Location' of the request body. """ - self.headers["Content-Disposition"] = content_disposition or u"form-data" - self.headers["Content-Disposition"] += u"; ".join( + content_disposition = (content_disposition or "form-data") + "; ".join( [ - u"", + "", self._render_parts( - ((u"name", self._name), (u"filename", self._filename)) + (("name", self._name), ("filename", self._filename)) ), ] ) + + self.headers["Content-Disposition"] = content_disposition self.headers["Content-Type"] = content_type self.headers["Content-Location"] = content_location diff --git a/lib/urllib3/fields.pyi b/lib/urllib3/fields.pyi deleted file mode 100644 index e1f3f905..00000000 --- a/lib/urllib3/fields.pyi +++ /dev/null @@ -1,28 +0,0 @@ -# Stubs for requests.packages.urllib3.fields (Python 3.4) - -from typing import Any, Callable, Mapping, Optional - -def guess_content_type(filename: str, default: str) -> str: ... -def format_header_param_rfc2231(name: str, value: str) -> str: ... -def format_header_param_html5(name: str, value: str) -> str: ... -def format_header_param(name: str, value: str) -> str: ... - -class RequestField: - data: Any - headers: Optional[Mapping[str, str]] - def __init__( - self, - name: str, - data: Any, - filename: Optional[str], - headers: Optional[Mapping[str, str]], - header_formatter: Callable[[str, str], str], - ) -> None: ... - @classmethod - def from_tuples( - cls, fieldname: str, value: str, header_formatter: Callable[[str, str], str] - ) -> RequestField: ... - def render_headers(self) -> str: ... - def make_multipart( - self, content_disposition: str, content_type: str, content_location: str - ) -> None: ... diff --git a/lib/urllib3/filepost.py b/lib/urllib3/filepost.py index 36c9252c..1c90a211 100644 --- a/lib/urllib3/filepost.py +++ b/lib/urllib3/filepost.py @@ -1,28 +1,32 @@ -from __future__ import absolute_import +from __future__ import annotations import binascii import codecs import os +import typing from io import BytesIO -from .fields import RequestField -from .packages import six -from .packages.six import b +from .fields import _TYPE_FIELD_VALUE_TUPLE, RequestField writer = codecs.lookup("utf-8")[3] +_TYPE_FIELDS_SEQUENCE = typing.Sequence[ + typing.Union[typing.Tuple[str, _TYPE_FIELD_VALUE_TUPLE], RequestField] +] +_TYPE_FIELDS = typing.Union[ + _TYPE_FIELDS_SEQUENCE, + typing.Mapping[str, _TYPE_FIELD_VALUE_TUPLE], +] -def choose_boundary(): + +def choose_boundary() -> str: """ Our embarrassingly-simple replacement for mimetools.choose_boundary. """ - boundary = binascii.hexlify(os.urandom(16)) - if not six.PY2: - boundary = boundary.decode("ascii") - return boundary + return binascii.hexlify(os.urandom(16)).decode() -def iter_field_objects(fields): +def iter_field_objects(fields: _TYPE_FIELDS) -> typing.Iterable[RequestField]: """ Iterate over fields. @@ -30,42 +34,29 @@ def iter_field_objects(fields): :class:`~urllib3.fields.RequestField`. """ - if isinstance(fields, dict): - i = six.iteritems(fields) - else: - i = iter(fields) + iterable: typing.Iterable[RequestField | tuple[str, _TYPE_FIELD_VALUE_TUPLE]] - for field in i: + if isinstance(fields, typing.Mapping): + iterable = fields.items() + else: + iterable = fields + + for field in iterable: if isinstance(field, RequestField): yield field else: yield RequestField.from_tuples(*field) -def iter_fields(fields): - """ - .. deprecated:: 1.6 - - Iterate over fields. - - The addition of :class:`~urllib3.fields.RequestField` makes this function - obsolete. Instead, use :func:`iter_field_objects`, which returns - :class:`~urllib3.fields.RequestField` objects. - - Supports list of (k, v) tuples and dicts. - """ - if isinstance(fields, dict): - return ((k, v) for k, v in six.iteritems(fields)) - - return ((k, v) for k, v in fields) - - -def encode_multipart_formdata(fields, boundary=None): +def encode_multipart_formdata( + fields: _TYPE_FIELDS, boundary: str | None = None +) -> tuple[bytes, str]: """ Encode a dictionary of ``fields`` using the multipart/form-data MIME format. :param fields: Dictionary of fields or list of (key, :class:`~urllib3.fields.RequestField`). + Values are processed by :func:`urllib3.fields.RequestField.from_tuples`. :param boundary: If not specified, then a random boundary will be generated using @@ -76,7 +67,7 @@ def encode_multipart_formdata(fields, boundary=None): boundary = choose_boundary() for field in iter_field_objects(fields): - body.write(b("--%s\r\n" % (boundary))) + body.write(f"--{boundary}\r\n".encode("latin-1")) writer(body).write(field.render_headers()) data = field.data @@ -84,15 +75,15 @@ def encode_multipart_formdata(fields, boundary=None): if isinstance(data, int): data = str(data) # Backwards compatibility - if isinstance(data, six.text_type): + if isinstance(data, str): writer(body).write(data) else: body.write(data) body.write(b"\r\n") - body.write(b("--%s--\r\n" % (boundary))) + body.write(f"--{boundary}--\r\n".encode("latin-1")) - content_type = str("multipart/form-data; boundary=%s" % boundary) + content_type = f"multipart/form-data; boundary={boundary}" return body.getvalue(), content_type diff --git a/lib/urllib3/filepost.pyi b/lib/urllib3/filepost.pyi deleted file mode 100644 index 54cf62a4..00000000 --- a/lib/urllib3/filepost.pyi +++ /dev/null @@ -1,16 +0,0 @@ -from typing import Any, Generator, List, Mapping, Optional, Tuple, Union - -from . import fields - -RequestField = fields.RequestField -Fields = Union[Mapping[str, str], List[Tuple[str]], List[RequestField]] -Iterator = Generator[Tuple[str], None, None] - -writer: Any - -def choose_boundary() -> str: ... -def iter_field_objects(fields: Fields) -> Iterator: ... -def iter_fields(fields: Fields) -> Iterator: ... -def encode_multipart_formdata( - fields: Fields, boundary: Optional[str] -) -> Tuple[str]: ... diff --git a/lib/urllib3/packages/__init__.py b/lib/urllib3/packages/__init__.py deleted file mode 100644 index e69de29b..00000000 diff --git a/lib/urllib3/packages/__init__.pyi b/lib/urllib3/packages/__init__.pyi deleted file mode 100644 index e69de29b..00000000 diff --git a/lib/urllib3/packages/backports/__init__.py b/lib/urllib3/packages/backports/__init__.py deleted file mode 100644 index e69de29b..00000000 diff --git a/lib/urllib3/packages/backports/makefile.py b/lib/urllib3/packages/backports/makefile.py deleted file mode 100644 index b8fb2154..00000000 --- a/lib/urllib3/packages/backports/makefile.py +++ /dev/null @@ -1,51 +0,0 @@ -# -*- coding: utf-8 -*- -""" -backports.makefile -~~~~~~~~~~~~~~~~~~ - -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 - - -def backport_makefile( - self, mode="r", buffering=None, encoding=None, errors=None, newline=None -): - """ - Backport of ``socket.makefile`` from Python 3.5. - """ - 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 = SocketIO(self, rawmode) - self._makefile_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 diff --git a/lib/urllib3/packages/backports/weakref_finalize.py b/lib/urllib3/packages/backports/weakref_finalize.py deleted file mode 100644 index a2f2966e..00000000 --- a/lib/urllib3/packages/backports/weakref_finalize.py +++ /dev/null @@ -1,155 +0,0 @@ -# -*- 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 deleted file mode 100644 index f099a3dc..00000000 --- a/lib/urllib3/packages/six.py +++ /dev/null @@ -1,1076 +0,0 @@ -# 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 -# in the Software without restriction, including without limitation the rights -# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -# copies of the Software, and to permit persons to whom the Software is -# furnished to do so, subject to the following conditions: -# -# The above copyright notice and this permission notice shall be included in all -# copies or substantial portions of the Software. -# -# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE -# SOFTWARE. - -"""Utilities for writing code that runs on Python 2 and 3""" - -from __future__ import absolute_import - -import functools -import itertools -import operator -import sys -import types - -__author__ = "Benjamin Peterson " -__version__ = "1.16.0" - - -# Useful for very coarse version differentiation. -PY2 = sys.version_info[0] == 2 -PY3 = sys.version_info[0] == 3 -PY34 = sys.version_info[0:2] >= (3, 4) - -if PY3: - string_types = (str,) - integer_types = (int,) - class_types = (type,) - text_type = str - binary_type = bytes - - MAXSIZE = sys.maxsize -else: - string_types = (basestring,) - integer_types = (int, long) - class_types = (type, types.ClassType) - text_type = unicode - binary_type = str - - if sys.platform.startswith("java"): - # Jython always uses 32 bits. - MAXSIZE = int((1 << 31) - 1) - else: - # It's possible to have sizeof(long) != sizeof(Py_ssize_t). - class X(object): - def __len__(self): - return 1 << 31 - - try: - len(X()) - except OverflowError: - # 32-bit - MAXSIZE = int((1 << 31) - 1) - else: - # 64-bit - 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.""" - func.__doc__ = doc - - -def _import_module(name): - """Import module, returning the module after the last dot.""" - __import__(name) - return sys.modules[name] - - -class _LazyDescr(object): - def __init__(self, name): - self.name = name - - def __get__(self, obj, tp): - result = self._resolve() - setattr(obj, self.name, result) # Invokes __set__. - try: - # This is a bit ugly, but it avoids running this again by - # removing this descriptor. - delattr(obj.__class__, self.name) - except AttributeError: - pass - return result - - -class MovedModule(_LazyDescr): - def __init__(self, name, old, new=None): - super(MovedModule, self).__init__(name) - if PY3: - if new is None: - new = name - self.mod = new - else: - self.mod = old - - def _resolve(self): - return _import_module(self.mod) - - def __getattr__(self, attr): - _module = self._resolve() - value = getattr(_module, attr) - setattr(self, attr, value) - return value - - -class _LazyModule(types.ModuleType): - def __init__(self, name): - super(_LazyModule, self).__init__(name) - self.__doc__ = self.__class__.__doc__ - - def __dir__(self): - attrs = ["__doc__", "__name__"] - attrs += [attr.name for attr in self._moved_attributes] - return attrs - - # Subclasses should override this - _moved_attributes = [] - - -class MovedAttribute(_LazyDescr): - def __init__(self, name, old_mod, new_mod, old_attr=None, new_attr=None): - super(MovedAttribute, self).__init__(name) - if PY3: - if new_mod is None: - new_mod = name - self.mod = new_mod - if new_attr is None: - if old_attr is None: - new_attr = name - else: - new_attr = old_attr - self.attr = new_attr - else: - self.mod = old_mod - if old_attr is None: - old_attr = name - self.attr = old_attr - - def _resolve(self): - module = _import_module(self.mod) - return getattr(module, self.attr) - - -class _SixMetaPathImporter(object): - - """ - A meta path importer to import six.moves and its submodules. - - This class implements a PEP302 finder and loader. It should be compatible - with Python 2.5 and all existing versions of Python3 - """ - - def __init__(self, six_module_name): - self.name = six_module_name - self.known_modules = {} - - def _add_module(self, mod, *fullnames): - for fullname in fullnames: - self.known_modules[self.name + "." + fullname] = mod - - def _get_module(self, fullname): - return self.known_modules[self.name + "." + fullname] - - def find_module(self, fullname, path=None): - if fullname in self.known_modules: - 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] - except KeyError: - raise ImportError("This loader does not know module " + fullname) - - def load_module(self, fullname): - try: - # in case of a reload - return sys.modules[fullname] - except KeyError: - pass - mod = self.__get_module(fullname) - if isinstance(mod, MovedModule): - mod = mod._resolve() - else: - mod.__loader__ = self - sys.modules[fullname] = mod - return mod - - def is_package(self, fullname): - """ - Return true, if the named module is a package. - - We need this method to get correct spec objects with - Python 3.4 (see PEP451) - """ - return hasattr(self.__get_module(fullname), "__path__") - - def get_code(self, fullname): - """Return None - - Required, if is_package is implemented""" - self.__get_module(fullname) # eventually raises ImportError - return None - - 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__) - - -class _MovedItems(_LazyModule): - - """Lazy loading of moved objects""" - - __path__ = [] # mark as package - - -_moved_attributes = [ - MovedAttribute("cStringIO", "cStringIO", "io", "StringIO"), - MovedAttribute("filter", "itertools", "builtins", "ifilter", "filter"), - MovedAttribute( - "filterfalse", "itertools", "itertools", "ifilterfalse", "filterfalse" - ), - MovedAttribute("input", "__builtin__", "builtins", "raw_input", "input"), - MovedAttribute("intern", "__builtin__", "sys"), - MovedAttribute("map", "itertools", "builtins", "imap", "map"), - MovedAttribute("getcwd", "os", "os", "getcwdu", "getcwd"), - MovedAttribute("getcwdb", "os", "os", "getcwd", "getcwdb"), - MovedAttribute("getoutput", "commands", "subprocess"), - MovedAttribute("range", "__builtin__", "builtins", "xrange", "range"), - MovedAttribute( - "reload_module", "__builtin__", "importlib" if PY34 else "imp", "reload" - ), - MovedAttribute("reduce", "__builtin__", "functools"), - MovedAttribute("shlex_quote", "pipes", "shlex", "quote"), - MovedAttribute("StringIO", "StringIO", "io"), - MovedAttribute("UserDict", "UserDict", "collections"), - MovedAttribute("UserList", "UserList", "collections"), - MovedAttribute("UserString", "UserString", "collections"), - MovedAttribute("xrange", "__builtin__", "builtins", "xrange", "range"), - MovedAttribute("zip", "itertools", "builtins", "izip", "zip"), - MovedAttribute( - "zip_longest", "itertools", "itertools", "izip_longest", "zip_longest" - ), - 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("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"), - MovedModule("html_parser", "HTMLParser", "html.parser"), - MovedModule("http_client", "httplib", "http.client"), - MovedModule("email_mime_base", "email.MIMEBase", "email.mime.base"), - MovedModule("email_mime_image", "email.MIMEImage", "email.mime.image"), - MovedModule("email_mime_multipart", "email.MIMEMultipart", "email.mime.multipart"), - MovedModule( - "email_mime_nonmultipart", "email.MIMENonMultipart", "email.mime.nonmultipart" - ), - MovedModule("email_mime_text", "email.MIMEText", "email.mime.text"), - MovedModule("BaseHTTPServer", "BaseHTTPServer", "http.server"), - MovedModule("CGIHTTPServer", "CGIHTTPServer", "http.server"), - MovedModule("SimpleHTTPServer", "SimpleHTTPServer", "http.server"), - MovedModule("cPickle", "cPickle", "pickle"), - MovedModule("queue", "Queue"), - MovedModule("reprlib", "repr"), - MovedModule("socketserver", "SocketServer"), - MovedModule("_thread", "thread", "_thread"), - MovedModule("tkinter", "Tkinter"), - MovedModule("tkinter_dialog", "Dialog", "tkinter.dialog"), - MovedModule("tkinter_filedialog", "FileDialog", "tkinter.filedialog"), - MovedModule("tkinter_scrolledtext", "ScrolledText", "tkinter.scrolledtext"), - MovedModule("tkinter_simpledialog", "SimpleDialog", "tkinter.simpledialog"), - MovedModule("tkinter_tix", "Tix", "tkinter.tix"), - MovedModule("tkinter_ttk", "ttk", "tkinter.ttk"), - MovedModule("tkinter_constants", "Tkconstants", "tkinter.constants"), - MovedModule("tkinter_dnd", "Tkdnd", "tkinter.dnd"), - MovedModule("tkinter_colorchooser", "tkColorChooser", "tkinter.colorchooser"), - MovedModule("tkinter_commondialog", "tkCommonDialog", "tkinter.commondialog"), - MovedModule("tkinter_tkfiledialog", "tkFileDialog", "tkinter.filedialog"), - MovedModule("tkinter_font", "tkFont", "tkinter.font"), - MovedModule("tkinter_messagebox", "tkMessageBox", "tkinter.messagebox"), - MovedModule("tkinter_tksimpledialog", "tkSimpleDialog", "tkinter.simpledialog"), - MovedModule("urllib_parse", __name__ + ".moves.urllib_parse", "urllib.parse"), - MovedModule("urllib_error", __name__ + ".moves.urllib_error", "urllib.error"), - MovedModule("urllib", __name__ + ".moves.urllib", __name__ + ".moves.urllib"), - MovedModule("urllib_robotparser", "robotparser", "urllib.robotparser"), - MovedModule("xmlrpc_client", "xmlrpclib", "xmlrpc.client"), - MovedModule("xmlrpc_server", "SimpleXMLRPCServer", "xmlrpc.server"), -] -# Add windows specific modules. -if sys.platform == "win32": - _moved_attributes += [ - MovedModule("winreg", "_winreg"), - ] - -for attr in _moved_attributes: - setattr(_MovedItems, attr.name, attr) - if isinstance(attr, MovedModule): - _importer._add_module(attr, "moves." + attr.name) -del attr - -_MovedItems._moved_attributes = _moved_attributes - -moves = _MovedItems(__name__ + ".moves") -_importer._add_module(moves, "moves") - - -class Module_six_moves_urllib_parse(_LazyModule): - - """Lazy loading of moved objects in six.moves.urllib_parse""" - - -_urllib_parse_moved_attributes = [ - MovedAttribute("ParseResult", "urlparse", "urllib.parse"), - MovedAttribute("SplitResult", "urlparse", "urllib.parse"), - MovedAttribute("parse_qs", "urlparse", "urllib.parse"), - MovedAttribute("parse_qsl", "urlparse", "urllib.parse"), - MovedAttribute("urldefrag", "urlparse", "urllib.parse"), - MovedAttribute("urljoin", "urlparse", "urllib.parse"), - MovedAttribute("urlparse", "urlparse", "urllib.parse"), - MovedAttribute("urlsplit", "urlparse", "urllib.parse"), - MovedAttribute("urlunparse", "urlparse", "urllib.parse"), - MovedAttribute("urlunsplit", "urlparse", "urllib.parse"), - MovedAttribute("quote", "urllib", "urllib.parse"), - MovedAttribute("quote_plus", "urllib", "urllib.parse"), - MovedAttribute("unquote", "urllib", "urllib.parse"), - MovedAttribute("unquote_plus", "urllib", "urllib.parse"), - MovedAttribute( - "unquote_to_bytes", "urllib", "urllib.parse", "unquote", "unquote_to_bytes" - ), - MovedAttribute("urlencode", "urllib", "urllib.parse"), - MovedAttribute("splitquery", "urllib", "urllib.parse"), - MovedAttribute("splittag", "urllib", "urllib.parse"), - MovedAttribute("splituser", "urllib", "urllib.parse"), - MovedAttribute("splitvalue", "urllib", "urllib.parse"), - MovedAttribute("uses_fragment", "urlparse", "urllib.parse"), - MovedAttribute("uses_netloc", "urlparse", "urllib.parse"), - MovedAttribute("uses_params", "urlparse", "urllib.parse"), - MovedAttribute("uses_query", "urlparse", "urllib.parse"), - MovedAttribute("uses_relative", "urlparse", "urllib.parse"), -] -for attr in _urllib_parse_moved_attributes: - setattr(Module_six_moves_urllib_parse, attr.name, attr) -del attr - -Module_six_moves_urllib_parse._moved_attributes = _urllib_parse_moved_attributes - -_importer._add_module( - Module_six_moves_urllib_parse(__name__ + ".moves.urllib_parse"), - "moves.urllib_parse", - "moves.urllib.parse", -) - - -class Module_six_moves_urllib_error(_LazyModule): - - """Lazy loading of moved objects in six.moves.urllib_error""" - - -_urllib_error_moved_attributes = [ - MovedAttribute("URLError", "urllib2", "urllib.error"), - MovedAttribute("HTTPError", "urllib2", "urllib.error"), - MovedAttribute("ContentTooShortError", "urllib", "urllib.error"), -] -for attr in _urllib_error_moved_attributes: - setattr(Module_six_moves_urllib_error, attr.name, attr) -del attr - -Module_six_moves_urllib_error._moved_attributes = _urllib_error_moved_attributes - -_importer._add_module( - Module_six_moves_urllib_error(__name__ + ".moves.urllib.error"), - "moves.urllib_error", - "moves.urllib.error", -) - - -class Module_six_moves_urllib_request(_LazyModule): - - """Lazy loading of moved objects in six.moves.urllib_request""" - - -_urllib_request_moved_attributes = [ - MovedAttribute("urlopen", "urllib2", "urllib.request"), - MovedAttribute("install_opener", "urllib2", "urllib.request"), - MovedAttribute("build_opener", "urllib2", "urllib.request"), - MovedAttribute("pathname2url", "urllib", "urllib.request"), - MovedAttribute("url2pathname", "urllib", "urllib.request"), - MovedAttribute("getproxies", "urllib", "urllib.request"), - MovedAttribute("Request", "urllib2", "urllib.request"), - MovedAttribute("OpenerDirector", "urllib2", "urllib.request"), - MovedAttribute("HTTPDefaultErrorHandler", "urllib2", "urllib.request"), - MovedAttribute("HTTPRedirectHandler", "urllib2", "urllib.request"), - MovedAttribute("HTTPCookieProcessor", "urllib2", "urllib.request"), - MovedAttribute("ProxyHandler", "urllib2", "urllib.request"), - MovedAttribute("BaseHandler", "urllib2", "urllib.request"), - MovedAttribute("HTTPPasswordMgr", "urllib2", "urllib.request"), - MovedAttribute("HTTPPasswordMgrWithDefaultRealm", "urllib2", "urllib.request"), - MovedAttribute("AbstractBasicAuthHandler", "urllib2", "urllib.request"), - MovedAttribute("HTTPBasicAuthHandler", "urllib2", "urllib.request"), - MovedAttribute("ProxyBasicAuthHandler", "urllib2", "urllib.request"), - MovedAttribute("AbstractDigestAuthHandler", "urllib2", "urllib.request"), - MovedAttribute("HTTPDigestAuthHandler", "urllib2", "urllib.request"), - MovedAttribute("ProxyDigestAuthHandler", "urllib2", "urllib.request"), - MovedAttribute("HTTPHandler", "urllib2", "urllib.request"), - MovedAttribute("HTTPSHandler", "urllib2", "urllib.request"), - MovedAttribute("FileHandler", "urllib2", "urllib.request"), - MovedAttribute("FTPHandler", "urllib2", "urllib.request"), - MovedAttribute("CacheFTPHandler", "urllib2", "urllib.request"), - MovedAttribute("UnknownHandler", "urllib2", "urllib.request"), - MovedAttribute("HTTPErrorProcessor", "urllib2", "urllib.request"), - MovedAttribute("urlretrieve", "urllib", "urllib.request"), - MovedAttribute("urlcleanup", "urllib", "urllib.request"), - MovedAttribute("URLopener", "urllib", "urllib.request"), - MovedAttribute("FancyURLopener", "urllib", "urllib.request"), - MovedAttribute("proxy_bypass", "urllib", "urllib.request"), - MovedAttribute("parse_http_list", "urllib2", "urllib.request"), - MovedAttribute("parse_keqv_list", "urllib2", "urllib.request"), -] -for attr in _urllib_request_moved_attributes: - setattr(Module_six_moves_urllib_request, attr.name, attr) -del attr - -Module_six_moves_urllib_request._moved_attributes = _urllib_request_moved_attributes - -_importer._add_module( - Module_six_moves_urllib_request(__name__ + ".moves.urllib.request"), - "moves.urllib_request", - "moves.urllib.request", -) - - -class Module_six_moves_urllib_response(_LazyModule): - - """Lazy loading of moved objects in six.moves.urllib_response""" - - -_urllib_response_moved_attributes = [ - MovedAttribute("addbase", "urllib", "urllib.response"), - MovedAttribute("addclosehook", "urllib", "urllib.response"), - MovedAttribute("addinfo", "urllib", "urllib.response"), - MovedAttribute("addinfourl", "urllib", "urllib.response"), -] -for attr in _urllib_response_moved_attributes: - setattr(Module_six_moves_urllib_response, attr.name, attr) -del attr - -Module_six_moves_urllib_response._moved_attributes = _urllib_response_moved_attributes - -_importer._add_module( - Module_six_moves_urllib_response(__name__ + ".moves.urllib.response"), - "moves.urllib_response", - "moves.urllib.response", -) - - -class Module_six_moves_urllib_robotparser(_LazyModule): - - """Lazy loading of moved objects in six.moves.urllib_robotparser""" - - -_urllib_robotparser_moved_attributes = [ - MovedAttribute("RobotFileParser", "robotparser", "urllib.robotparser"), -] -for attr in _urllib_robotparser_moved_attributes: - setattr(Module_six_moves_urllib_robotparser, attr.name, attr) -del attr - -Module_six_moves_urllib_robotparser._moved_attributes = ( - _urllib_robotparser_moved_attributes -) - -_importer._add_module( - Module_six_moves_urllib_robotparser(__name__ + ".moves.urllib.robotparser"), - "moves.urllib_robotparser", - "moves.urllib.robotparser", -) - - -class Module_six_moves_urllib(types.ModuleType): - - """Create a six.moves.urllib namespace that resembles the Python 3 namespace""" - - __path__ = [] # mark as package - parse = _importer._get_module("moves.urllib_parse") - error = _importer._get_module("moves.urllib_error") - request = _importer._get_module("moves.urllib_request") - response = _importer._get_module("moves.urllib_response") - robotparser = _importer._get_module("moves.urllib_robotparser") - - def __dir__(self): - return ["parse", "error", "request", "response", "robotparser"] - - -_importer._add_module( - Module_six_moves_urllib(__name__ + ".moves.urllib"), "moves.urllib" -) - - -def add_move(move): - """Add an item to six.moves.""" - setattr(_MovedItems, move.name, move) - - -def remove_move(name): - """Remove item from six.moves.""" - try: - delattr(_MovedItems, name) - except AttributeError: - try: - del moves.__dict__[name] - except KeyError: - raise AttributeError("no such move, %r" % (name,)) - - -if PY3: - _meth_func = "__func__" - _meth_self = "__self__" - - _func_closure = "__closure__" - _func_code = "__code__" - _func_defaults = "__defaults__" - _func_globals = "__globals__" -else: - _meth_func = "im_func" - _meth_self = "im_self" - - _func_closure = "func_closure" - _func_code = "func_code" - _func_defaults = "func_defaults" - _func_globals = "func_globals" - - -try: - advance_iterator = next -except NameError: - - def advance_iterator(it): - return it.next() - - -next = advance_iterator - - -try: - callable = callable -except NameError: - - def callable(obj): - return any("__call__" in klass.__dict__ for klass in type(obj).__mro__) - - -if PY3: - - def get_unbound_function(unbound): - return unbound - - create_bound_method = types.MethodType - - def create_unbound_method(func, cls): - return func - - Iterator = object -else: - - def get_unbound_function(unbound): - return unbound.im_func - - def create_bound_method(func, obj): - return types.MethodType(func, obj, obj.__class__) - - def create_unbound_method(func, cls): - return types.MethodType(func, None, cls) - - class Iterator(object): - def next(self): - return type(self).__next__(self) - - callable = callable -_add_doc( - get_unbound_function, """Get the function out of a possibly unbound function""" -) - - -get_method_function = operator.attrgetter(_meth_func) -get_method_self = operator.attrgetter(_meth_self) -get_function_closure = operator.attrgetter(_func_closure) -get_function_code = operator.attrgetter(_func_code) -get_function_defaults = operator.attrgetter(_func_defaults) -get_function_globals = operator.attrgetter(_func_globals) - - -if PY3: - - def iterkeys(d, **kw): - return iter(d.keys(**kw)) - - def itervalues(d, **kw): - return iter(d.values(**kw)) - - def iteritems(d, **kw): - return iter(d.items(**kw)) - - def iterlists(d, **kw): - return iter(d.lists(**kw)) - - viewkeys = operator.methodcaller("keys") - - viewvalues = operator.methodcaller("values") - - viewitems = operator.methodcaller("items") -else: - - def iterkeys(d, **kw): - return d.iterkeys(**kw) - - def itervalues(d, **kw): - return d.itervalues(**kw) - - def iteritems(d, **kw): - return d.iteritems(**kw) - - def iterlists(d, **kw): - return d.iterlists(**kw) - - viewkeys = operator.methodcaller("viewkeys") - - viewvalues = operator.methodcaller("viewvalues") - - viewitems = operator.methodcaller("viewitems") - -_add_doc(iterkeys, "Return an iterator over the keys of a dictionary.") -_add_doc(itervalues, "Return an iterator over the values of a dictionary.") -_add_doc(iteritems, "Return an iterator over the (key, value) pairs of a dictionary.") -_add_doc( - iterlists, "Return an iterator over the (key, [values]) pairs of a dictionary." -) - - -if PY3: - - def b(s): - return s.encode("latin-1") - - def u(s): - return s - - unichr = chr - import struct - - int2byte = struct.Struct(">B").pack - del struct - byte2int = operator.itemgetter(0) - indexbytes = operator.getitem - iterbytes = iter - import io - - StringIO = io.StringIO - BytesIO = io.BytesIO - del io - _assertCountEqual = "assertCountEqual" - if sys.version_info[1] <= 1: - _assertRaisesRegex = "assertRaisesRegexp" - _assertRegex = "assertRegexpMatches" - _assertNotRegex = "assertNotRegexpMatches" - else: - _assertRaisesRegex = "assertRaisesRegex" - _assertRegex = "assertRegex" - _assertNotRegex = "assertNotRegex" -else: - - def b(s): - return s - - # Workaround for standalone backslash - - def u(s): - return unicode(s.replace(r"\\", r"\\\\"), "unicode_escape") - - unichr = unichr - int2byte = chr - - def byte2int(bs): - return ord(bs[0]) - - def indexbytes(buf, i): - return ord(buf[i]) - - iterbytes = functools.partial(itertools.imap, ord) - import StringIO - - StringIO = BytesIO = StringIO.StringIO - _assertCountEqual = "assertItemsEqual" - _assertRaisesRegex = "assertRaisesRegexp" - _assertRegex = "assertRegexpMatches" - _assertNotRegex = "assertNotRegexpMatches" -_add_doc(b, """Byte literal""") -_add_doc(u, """Text literal""") - - -def assertCountEqual(self, *args, **kwargs): - return getattr(self, _assertCountEqual)(*args, **kwargs) - - -def assertRaisesRegex(self, *args, **kwargs): - return getattr(self, _assertRaisesRegex)(*args, **kwargs) - - -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") - - def reraise(tp, value, tb=None): - try: - if value is None: - value = tp() - if value.__traceback__ is not tb: - raise value.with_traceback(tb) - raise value - finally: - value = None - tb = None - -else: - - def exec_(_code_, _globs_=None, _locs_=None): - """Execute code in a namespace.""" - if _globs_ is None: - frame = sys._getframe(1) - _globs_ = frame.f_globals - if _locs_ is None: - _locs_ = frame.f_locals - del frame - elif _locs_ is None: - _locs_ = _globs_ - exec ("""exec _code_ in _globs_, _locs_""") - - exec_( - """def reraise(tp, value, tb=None): - try: - raise tp, value, tb - finally: - tb = None -""" - ) - - -if sys.version_info[:2] > (3,): - exec_( - """def raise_from(value, from_value): - try: - raise value from from_value - finally: - value = None -""" - ) -else: - - def raise_from(value, from_value): - raise value - - -print_ = getattr(moves.builtins, "print", None) -if print_ is None: - - def print_(*args, **kwargs): - """The new-style print function for Python 2.4 and 2.5.""" - fp = kwargs.pop("file", sys.stdout) - if fp is None: - return - - def write(data): - if not isinstance(data, basestring): - data = str(data) - # If the file has an encoding, encode unicode with it. - if ( - isinstance(fp, file) - and isinstance(data, unicode) - and fp.encoding is not None - ): - errors = getattr(fp, "errors", None) - if errors is None: - errors = "strict" - data = data.encode(fp.encoding, errors) - fp.write(data) - - want_unicode = False - sep = kwargs.pop("sep", None) - if sep is not None: - if isinstance(sep, unicode): - want_unicode = True - elif not isinstance(sep, str): - raise TypeError("sep must be None or a string") - end = kwargs.pop("end", None) - if end is not None: - if isinstance(end, unicode): - want_unicode = True - elif not isinstance(end, str): - raise TypeError("end must be None or a string") - if kwargs: - raise TypeError("invalid keyword arguments to print()") - if not want_unicode: - for arg in args: - if isinstance(arg, unicode): - want_unicode = True - break - if want_unicode: - newline = unicode("\n") - space = unicode(" ") - else: - newline = "\n" - space = " " - if sep is None: - sep = space - if end is None: - end = newline - for i, arg in enumerate(args): - if i: - write(sep) - write(arg) - write(end) - - -if sys.version_info[:2] < (3, 3): - _print = print_ - - def print_(*args, **kwargs): - fp = kwargs.get("file", sys.stdout) - flush = kwargs.pop("flush", False) - _print(*args, **kwargs) - if flush and fp is not None: - fp.flush() - - -_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, - ): - return functools.partial( - _update_wrapper, wrapped=wrapped, assigned=assigned, updated=updated - ) - - wraps.__doc__ = functools.wraps.__doc__ - -else: - wraps = functools.wraps - - -def with_metaclass(meta, *bases): - """Create a base class with a metaclass.""" - # This requires a bit of explanation: the basic idea is to make a dummy - # metaclass for one level of class instantiation that replaces itself with - # the actual metaclass. - class metaclass(type): - def __new__(cls, name, this_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): - return meta.__prepare__(name, bases) - - return type.__new__(metaclass, "temporary_class", (), {}) - - -def add_metaclass(metaclass): - """Class decorator for creating a class with a metaclass.""" - - def wrapper(cls): - orig_vars = cls.__dict__.copy() - slots = orig_vars.get("__slots__") - if slots is not None: - if isinstance(slots, str): - slots = [slots] - for slots_var in slots: - orig_vars.pop(slots_var) - orig_vars.pop("__dict__", None) - orig_vars.pop("__weakref__", None) - if hasattr(cls, "__qualname__"): - orig_vars["__qualname__"] = cls.__qualname__ - return metaclass(cls.__name__, cls.__bases__, orig_vars) - - return wrapper - - -def ensure_binary(s, encoding="utf-8", errors="strict"): - """Coerce **s** to six.binary_type. - - For Python 2: - - `unicode` -> encoded to `str` - - `str` -> `str` - - For Python 3: - - `str` -> encoded to `bytes` - - `bytes` -> `bytes` - """ - if isinstance(s, binary_type): - return s - if isinstance(s, text_type): - return s.encode(encoding, errors) - raise TypeError("not expecting type '%s'" % type(s)) - - -def ensure_str(s, encoding="utf-8", errors="strict"): - """Coerce *s* to `str`. - - For Python 2: - - `unicode` -> encoded to `str` - - `str` -> `str` - - For Python 3: - - `str` -> `str` - - `bytes` -> decoded to `str` - """ - # Optimization: Fast return for the common case. - if type(s) is str: - return s - if PY2 and isinstance(s, text_type): - return s.encode(encoding, errors) - elif PY3 and isinstance(s, binary_type): - return s.decode(encoding, errors) - elif not isinstance(s, (text_type, binary_type)): - raise TypeError("not expecting type '%s'" % type(s)) - return s - - -def ensure_text(s, encoding="utf-8", errors="strict"): - """Coerce *s* to six.text_type. - - For Python 2: - - `unicode` -> `unicode` - - `str` -> `unicode` - - For Python 3: - - `str` -> `str` - - `bytes` -> decoded to `str` - """ - if isinstance(s, binary_type): - return s.decode(encoding, errors) - elif isinstance(s, text_type): - return s - else: - raise TypeError("not expecting type '%s'" % type(s)) - - -def python_2_unicode_compatible(klass): - """ - 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 - returning text and apply this decorator to the class. - """ - if PY2: - if "__str__" not in klass.__dict__: - raise ValueError( - "@python_2_unicode_compatible cannot be applied " - "to %s because it doesn't define __str__()." % klass.__name__ - ) - klass.__unicode__ = klass.__str__ - klass.__str__ = lambda self: self.__unicode__().encode("utf-8") - return klass - - -# Complete the moves implementation. -# This code is at the end of this module to speed up module loading. -# Turn this module into a package. -__path__ = [] # required for PEP 302 and PEP 451 -__package__ = __name__ # see PEP 366 @ReservedAssignment -if globals().get("__spec__") is not None: - __spec__.submodule_search_locations = [] # PEP 451 @UndefinedVariable -# Remove other six meta path importers, since they cause problems. This can -# happen if six is removed from sys.modules and then reloaded. (Setuptools does -# this for some reason.) -if sys.meta_path: - for i, importer in enumerate(sys.meta_path): - # Here's some real nastiness: Another "instance" of the six module might - # be floating around. Therefore, we can't use isinstance() to check for - # the six meta path importer, since the other six instance will have - # inserted an importer with different class. - if ( - type(importer).__name__ == "_SixMetaPathImporter" - and importer.name == __name__ - ): - del sys.meta_path[i] - break - del i, importer -# Finally, add the importer to the meta path import hook. -sys.meta_path.append(_importer) diff --git a/lib/urllib3/packages/six.pyi b/lib/urllib3/packages/six.pyi deleted file mode 100644 index e69de29b..00000000 diff --git a/lib/urllib3/poolmanager.py b/lib/urllib3/poolmanager.py index 14b10daf..02b2f622 100644 --- a/lib/urllib3/poolmanager.py +++ b/lib/urllib3/poolmanager.py @@ -1,24 +1,33 @@ -from __future__ import absolute_import +from __future__ import annotations -import collections import functools import logging +import typing +import warnings +from types import TracebackType +from urllib.parse import urljoin from ._collections import RecentlyUsedContainer +from ._request_methods import RequestMethods +from .connection import ProxyConfig 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 .response import BaseHTTPResponse +from .util.connection import _TYPE_SOCKET_OPTIONS from .util.proxy import connection_requires_http_tunnel from .util.retry import Retry -from .util.url import parse_url +from .util.timeout import Timeout +from .util.url import Url, parse_url + +if typing.TYPE_CHECKING: + import ssl + + from typing_extensions import Literal __all__ = ["PoolManager", "ProxyManager", "proxy_from_url"] @@ -31,52 +40,61 @@ SSL_KEYWORDS = ( "cert_reqs", "ca_certs", "ssl_version", + "ssl_minimum_version", + "ssl_maximum_version", "ca_cert_dir", "ssl_context", "key_password", "server_hostname", ) +# Default value for `blocksize` - a new parameter introduced to +# http.client.HTTPConnection & http.client.HTTPSConnection in Python 3.7 +_DEFAULT_BLOCKSIZE = 16384 -# All known keyword arguments that could be provided to the pool manager, its -# pools, or the underlying connections. This is used to construct a pool key. -_key_fields = ( - "key_scheme", # str - "key_host", # str - "key_port", # int - "key_timeout", # int or float or Timeout - "key_retries", # int or Retry - "key_strict", # bool - "key_block", # bool - "key_source_address", # str - "key_key_file", # str - "key_key_password", # str - "key_cert_file", # str - "key_cert_reqs", # str - "key_ca_certs", # str - "key_ssl_version", # str - "key_ca_cert_dir", # str - "key_ssl_context", # instance of ssl.SSLContext or urllib3.util.ssl_.SSLContext - "key_maxsize", # int - "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 - "key_assert_fingerprint", # str - "key_server_hostname", # str -) - -#: The namedtuple class used to construct keys for the connection pool. -#: 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) +_SelfT = typing.TypeVar("_SelfT") -def _default_key_normalizer(key_class, request_context): +class PoolKey(typing.NamedTuple): + """ + All known keyword arguments that could be provided to the pool manager, its + pools, or the underlying connections. + + All custom key schemes should include the fields in this key at a minimum. + """ + + key_scheme: str + key_host: str + key_port: int | None + key_timeout: Timeout | float | int | None + key_retries: Retry | bool | int | None + key_block: bool | None + key_source_address: tuple[str, int] | None + key_key_file: str | None + key_key_password: str | None + key_cert_file: str | None + key_cert_reqs: str | None + key_ca_certs: str | None + key_ssl_version: int | str | None + key_ssl_minimum_version: ssl.TLSVersion | None + key_ssl_maximum_version: ssl.TLSVersion | None + key_ca_cert_dir: str | None + key_ssl_context: ssl.SSLContext | None + key_maxsize: int | None + key_headers: frozenset[tuple[str, str]] | None + key__proxy: Url | None + key__proxy_headers: frozenset[tuple[str, str]] | None + key__proxy_config: ProxyConfig | None + key_socket_options: _TYPE_SOCKET_OPTIONS | None + key__socks_options: frozenset[tuple[str, str]] | None + key_assert_hostname: bool | str | None + key_assert_fingerprint: str | None + key_server_hostname: str | None + key_blocksize: int | None + + +def _default_key_normalizer( + key_class: type[PoolKey], request_context: dict[str, typing.Any] +) -> PoolKey: """ Create a pool key out of a request context dictionary. @@ -122,6 +140,10 @@ def _default_key_normalizer(key_class, request_context): if field not in context: context[field] = None + # Default key_blocksize to _DEFAULT_BLOCKSIZE if missing from the context + if context.get("key_blocksize") is None: + context["key_blocksize"] = _DEFAULT_BLOCKSIZE + return key_class(**context) @@ -154,23 +176,36 @@ class PoolManager(RequestMethods): Additional parameters are used to create fresh :class:`urllib3.connectionpool.ConnectionPool` instances. - Example:: + Example: - >>> manager = PoolManager(num_pools=2) - >>> r = manager.request('GET', 'http://google.com/') - >>> r = manager.request('GET', 'http://google.com/mail') - >>> r = manager.request('GET', 'http://yahoo.com/') - >>> len(manager.pools) - 2 + .. code-block:: python + + import urllib3 + + http = urllib3.PoolManager(num_pools=2) + + resp1 = http.request("GET", "https://google.com/") + resp2 = http.request("GET", "https://google.com/mail") + resp3 = http.request("GET", "https://yahoo.com/") + + print(len(http.pools)) + # 2 """ - proxy = None - proxy_config = None + proxy: Url | None = None + proxy_config: ProxyConfig | None = None - def __init__(self, num_pools=10, headers=None, **connection_pool_kw): - RequestMethods.__init__(self, headers) + def __init__( + self, + num_pools: int = 10, + headers: typing.Mapping[str, str] | None = None, + **connection_pool_kw: typing.Any, + ) -> None: + super().__init__(headers) self.connection_pool_kw = connection_pool_kw + + self.pools: RecentlyUsedContainer[PoolKey, HTTPConnectionPool] self.pools = RecentlyUsedContainer(num_pools) # Locally set the pool classes and keys so other PoolManagers can @@ -178,15 +213,26 @@ class PoolManager(RequestMethods): self.pool_classes_by_scheme = pool_classes_by_scheme self.key_fn_by_scheme = key_fn_by_scheme.copy() - def __enter__(self): + def __enter__(self: _SelfT) -> _SelfT: return self - def __exit__(self, exc_type, exc_val, exc_tb): + def __exit__( + self, + exc_type: type[BaseException] | None, + exc_val: BaseException | None, + exc_tb: TracebackType | None, + ) -> Literal[False]: self.clear() # Return False to re-raise any potential exceptions return False - def _new_pool(self, scheme, host, port, request_context=None): + def _new_pool( + self, + scheme: str, + host: str, + port: int, + request_context: dict[str, typing.Any] | None = None, + ) -> HTTPConnectionPool: """ Create a new :class:`urllib3.connectionpool.ConnectionPool` based on host, port, scheme, and any additional pool keyword arguments. @@ -196,10 +242,15 @@ class PoolManager(RequestMethods): connection pools handed out by :meth:`connection_from_url` and companion methods. It is intended to be overridden for customization. """ - pool_cls = self.pool_classes_by_scheme[scheme] + pool_cls: type[HTTPConnectionPool] = self.pool_classes_by_scheme[scheme] if request_context is None: request_context = self.connection_pool_kw.copy() + # Default blocksize to _DEFAULT_BLOCKSIZE if missing or explicitly + # set to 'None' in the request_context. + if request_context.get("blocksize") is None: + request_context["blocksize"] = _DEFAULT_BLOCKSIZE + # Although the context has everything necessary to create the pool, # this function has historically only used the scheme, host, and port # in the positional args. When an API change is acceptable these can @@ -213,7 +264,7 @@ class PoolManager(RequestMethods): return pool_cls(host, port, **request_context) - def clear(self): + def clear(self) -> None: """ Empty our store of pools and direct them all to close. @@ -222,7 +273,13 @@ class PoolManager(RequestMethods): """ self.pools.clear() - def connection_from_host(self, host, port=None, scheme="http", pool_kwargs=None): + def connection_from_host( + self, + host: str | None, + port: int | None = None, + scheme: str | None = "http", + pool_kwargs: dict[str, typing.Any] | None = None, + ) -> HTTPConnectionPool: """ Get a :class:`urllib3.connectionpool.ConnectionPool` based on the host, port, and scheme. @@ -245,13 +302,23 @@ class PoolManager(RequestMethods): return self.connection_from_context(request_context) - def connection_from_context(self, request_context): + def connection_from_context( + self, request_context: dict[str, typing.Any] + ) -> HTTPConnectionPool: """ 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. """ + if "strict" in request_context: + warnings.warn( + "The 'strict' parameter is no longer needed on Python 3+. " + "This will raise an error in urllib3 v2.1.0.", + DeprecationWarning, + ) + request_context.pop("strict") + scheme = request_context["scheme"].lower() pool_key_constructor = self.key_fn_by_scheme.get(scheme) if not pool_key_constructor: @@ -260,7 +327,9 @@ class PoolManager(RequestMethods): return self.connection_from_pool_key(pool_key, request_context=request_context) - def connection_from_pool_key(self, pool_key, request_context=None): + def connection_from_pool_key( + self, pool_key: PoolKey, request_context: dict[str, typing.Any] + ) -> HTTPConnectionPool: """ Get a :class:`urllib3.connectionpool.ConnectionPool` based on the provided pool key. @@ -284,7 +353,9 @@ class PoolManager(RequestMethods): return pool - def connection_from_url(self, url, pool_kwargs=None): + def connection_from_url( + self, url: str, pool_kwargs: dict[str, typing.Any] | None = None + ) -> HTTPConnectionPool: """ Similar to :func:`urllib3.connectionpool.connection_from_url`. @@ -300,7 +371,9 @@ class PoolManager(RequestMethods): u.host, port=u.port, scheme=u.scheme, pool_kwargs=pool_kwargs ) - def _merge_pool_kwargs(self, override): + def _merge_pool_kwargs( + self, override: dict[str, typing.Any] | None + ) -> dict[str, typing.Any]: """ Merge a dictionary of override values for self.connection_pool_kw. @@ -320,7 +393,7 @@ class PoolManager(RequestMethods): base_pool_kwargs[key] = value return base_pool_kwargs - def _proxy_requires_url_absolute_form(self, parsed_url): + def _proxy_requires_url_absolute_form(self, parsed_url: Url) -> bool: """ Indicates if the proxy requires the complete destination URL in the request. Normally this is only needed when not using an HTTP CONNECT @@ -333,24 +406,9 @@ class PoolManager(RequestMethods): 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): + def urlopen( # type: ignore[override] + self, method: str, url: str, redirect: bool = True, **kw: typing.Any + ) -> BaseHTTPResponse: """ Same as :meth:`urllib3.HTTPConnectionPool.urlopen` with custom cross-host redirect logic and only sends the request-uri @@ -360,7 +418,16 @@ class PoolManager(RequestMethods): :class:`urllib3.connectionpool.ConnectionPool` can be chosen for it. """ u = parse_url(url) - self._validate_proxy_scheme_url_selection(u.scheme) + + if u.scheme is None: + warnings.warn( + "URLs without a scheme (ie 'https://') are deprecated and will raise an error " + "in a future version of urllib3. To avoid this DeprecationWarning ensure all URLs " + "start with 'https://' or 'http://'. Read more in this issue: " + "https://github.com/urllib3/urllib3/issues/2920", + category=DeprecationWarning, + stacklevel=2, + ) conn = self.connection_from_host(u.host, port=u.port, scheme=u.scheme) @@ -368,7 +435,7 @@ class PoolManager(RequestMethods): kw["redirect"] = False if "headers" not in kw: - kw["headers"] = self.headers.copy() + kw["headers"] = self.headers if self._proxy_requires_url_absolute_form(u): response = conn.urlopen(method, url, **kw) @@ -396,10 +463,11 @@ class PoolManager(RequestMethods): if retries.remove_headers_on_redirect and not conn.is_same_host( redirect_location ): - headers = list(six.iterkeys(kw["headers"])) - for header in headers: + new_headers = kw["headers"].copy() + for header in kw["headers"]: if header.lower() in retries.remove_headers_on_redirect: - kw["headers"].pop(header, None) + new_headers.pop(header, None) + kw["headers"] = new_headers try: retries = retries.increment(method, url, response=response, _pool=conn) @@ -445,37 +513,51 @@ class ProxyManager(PoolManager): private. IP address, target hostname, SNI, and port are always visible to an HTTPS proxy even when this flag is disabled. + :param proxy_assert_hostname: + The hostname of the certificate to verify against. + + :param proxy_assert_fingerprint: + The fingerprint of the certificate to verify against. + Example: - >>> proxy = urllib3.ProxyManager('http://localhost:3128/') - >>> r1 = proxy.request('GET', 'http://google.com/') - >>> r2 = proxy.request('GET', 'http://httpbin.org/') - >>> len(proxy.pools) - 1 - >>> r3 = proxy.request('GET', 'https://httpbin.org/') - >>> r4 = proxy.request('GET', 'https://twitter.com/') - >>> len(proxy.pools) - 3 + + .. code-block:: python + + import urllib3 + + proxy = urllib3.ProxyManager("https://localhost:3128/") + + resp1 = proxy.request("GET", "https://google.com/") + resp2 = proxy.request("GET", "https://httpbin.org/") + + print(len(proxy.pools)) + # 1 + + resp3 = proxy.request("GET", "https://httpbin.org/") + resp4 = proxy.request("GET", "https://twitter.com/") + + print(len(proxy.pools)) + # 3 """ def __init__( self, - proxy_url, - num_pools=10, - headers=None, - proxy_headers=None, - proxy_ssl_context=None, - use_forwarding_for_https=False, - **connection_pool_kw - ): - + proxy_url: str, + num_pools: int = 10, + headers: typing.Mapping[str, str] | None = None, + proxy_headers: typing.Mapping[str, str] | None = None, + proxy_ssl_context: ssl.SSLContext | None = None, + use_forwarding_for_https: bool = False, + proxy_assert_hostname: None | str | Literal[False] = None, + proxy_assert_fingerprint: str | None = None, + **connection_pool_kw: typing.Any, + ) -> None: if isinstance(proxy_url, HTTPConnectionPool): - proxy_url = "%s://%s:%i" % ( - proxy_url.scheme, - proxy_url.host, - proxy_url.port, - ) - proxy = parse_url(proxy_url) + str_proxy_url = f"{proxy_url.scheme}://{proxy_url.host}:{proxy_url.port}" + else: + str_proxy_url = proxy_url + proxy = parse_url(str_proxy_url) if proxy.scheme not in ("http", "https"): raise ProxySchemeUnknown(proxy.scheme) @@ -487,25 +569,38 @@ class ProxyManager(PoolManager): 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) + self.proxy_config = ProxyConfig( + proxy_ssl_context, + use_forwarding_for_https, + proxy_assert_hostname, + proxy_assert_fingerprint, + ) 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) + super().__init__(num_pools, headers, **connection_pool_kw) - def connection_from_host(self, host, port=None, scheme="http", pool_kwargs=None): + def connection_from_host( + self, + host: str | None, + port: int | None = None, + scheme: str | None = "http", + pool_kwargs: dict[str, typing.Any] | None = None, + ) -> HTTPConnectionPool: if scheme == "https": - return super(ProxyManager, self).connection_from_host( + return super().connection_from_host( host, port, scheme, pool_kwargs=pool_kwargs ) - return super(ProxyManager, self).connection_from_host( - self.proxy.host, self.proxy.port, self.proxy.scheme, pool_kwargs=pool_kwargs + return super().connection_from_host( + self.proxy.host, self.proxy.port, self.proxy.scheme, pool_kwargs=pool_kwargs # type: ignore[union-attr] ) - def _set_proxy_headers(self, url, headers=None): + def _set_proxy_headers( + self, url: str, headers: typing.Mapping[str, str] | None = None + ) -> typing.Mapping[str, str]: """ Sets headers needed by proxies: specifically, the Accept and Host headers. Only sets headers not provided by the user. @@ -520,7 +615,9 @@ class ProxyManager(PoolManager): headers_.update(headers) return headers_ - def urlopen(self, method, url, redirect=True, **kw): + def urlopen( # type: ignore[override] + self, method: str, url: str, redirect: bool = True, **kw: typing.Any + ) -> BaseHTTPResponse: "Same as HTTP(S)ConnectionPool.urlopen, ``url`` must be absolute." u = parse_url(url) if not connection_requires_http_tunnel(self.proxy, self.proxy_config, u.scheme): @@ -530,8 +627,8 @@ class ProxyManager(PoolManager): headers = kw.get("headers", self.headers) kw["headers"] = self._set_proxy_headers(url, headers) - return super(ProxyManager, self).urlopen(method, url, redirect=redirect, **kw) + return super().urlopen(method, url, redirect=redirect, **kw) -def proxy_from_url(url, **kw): +def proxy_from_url(url: str, **kw: typing.Any) -> ProxyManager: return ProxyManager(proxy_url=url, **kw) diff --git a/lib/urllib3/py.typed b/lib/urllib3/py.typed new file mode 100644 index 00000000..5f3ea3d9 --- /dev/null +++ b/lib/urllib3/py.typed @@ -0,0 +1,2 @@ +# Instruct type checkers to look for inline type annotations in this package. +# See PEP 561. diff --git a/lib/urllib3/response.py b/lib/urllib3/response.py index 0bd13d40..12097ea9 100644 --- a/lib/urllib3/response.py +++ b/lib/urllib3/response.py @@ -1,25 +1,47 @@ -from __future__ import absolute_import +from __future__ import annotations +import collections import io +import json as _json import logging +import re import sys +import typing import warnings import zlib from contextlib import contextmanager -from socket import error as SocketError +from http.client import HTTPMessage as _HttplibHTTPMessage +from http.client import HTTPResponse as _HttplibHTTPResponse from socket import timeout as SocketTimeout try: try: - import brotlicffi as brotli + import brotlicffi as brotli # type: ignore[import] except ImportError: - import brotli + import brotli # type: ignore[import] except ImportError: brotli = None +try: + import zstandard as zstd # type: ignore[import] + + # The package 'zstandard' added the 'eof' property starting + # in v0.18.0 which we require to ensure a complete and + # valid zstd stream was fed into the ZstdDecoder. + # See: https://github.com/urllib3/urllib3/pull/2624 + _zstd_version = _zstd_version = tuple( + map(int, re.search(r"^([0-9]+)\.([0-9]+)", zstd.__version__).groups()) # type: ignore[union-attr] + ) + if _zstd_version < (0, 18): # Defensive: + zstd = None + +except (AttributeError, ImportError, ValueError): # Defensive: + zstd = None + from . import util +from ._base_connection import _TYPE_BODY from ._collections import HTTPHeaderDict -from .connection import BaseSSLError, HTTPException +from .connection import BaseSSLError, HTTPConnection, HTTPException from .exceptions import ( BodyNotHttplibCompatible, DecodeError, @@ -32,22 +54,32 @@ from .exceptions import ( ResponseNotChunked, SSLError, ) -from .packages import six from .util.response import is_fp_closed, is_response_to_head +from .util.retry import Retry + +if typing.TYPE_CHECKING: + from typing_extensions import Literal + + from .connectionpool import HTTPConnectionPool log = logging.getLogger(__name__) -class DeflateDecoder(object): - def __init__(self): +class ContentDecoder: + def decompress(self, data: bytes) -> bytes: + raise NotImplementedError() + + def flush(self) -> bytes: + raise NotImplementedError() + + +class DeflateDecoder(ContentDecoder): + def __init__(self) -> None: self._first_try = True self._data = b"" self._obj = zlib.decompressobj() - def __getattr__(self, name): - return getattr(self._obj, name) - - def decompress(self, data): + def decompress(self, data: bytes) -> bytes: if not data: return data @@ -59,7 +91,7 @@ class DeflateDecoder(object): decompressed = self._obj.decompress(data) if decompressed: self._first_try = False - self._data = None + self._data = None # type: ignore[assignment] return decompressed except zlib.error: self._first_try = False @@ -67,25 +99,24 @@ class DeflateDecoder(object): try: return self.decompress(self._data) finally: - self._data = None + self._data = None # type: ignore[assignment] + + def flush(self) -> bytes: + return self._obj.flush() -class GzipDecoderState(object): - +class GzipDecoderState: FIRST_MEMBER = 0 OTHER_MEMBERS = 1 SWALLOW_DATA = 2 -class GzipDecoder(object): - def __init__(self): +class GzipDecoder(ContentDecoder): + def __init__(self) -> None: self._obj = zlib.decompressobj(16 + zlib.MAX_WBITS) self._state = GzipDecoderState.FIRST_MEMBER - def __getattr__(self, name): - return getattr(self._obj, name) - - def decompress(self, data): + def decompress(self, data: bytes) -> bytes: ret = bytearray() if self._state == GzipDecoderState.SWALLOW_DATA or not data: return bytes(ret) @@ -106,27 +137,53 @@ class GzipDecoder(object): self._state = GzipDecoderState.OTHER_MEMBERS self._obj = zlib.decompressobj(16 + zlib.MAX_WBITS) + def flush(self) -> bytes: + return self._obj.flush() + if brotli is not None: - class BrotliDecoder(object): + class BrotliDecoder(ContentDecoder): # Supports both 'brotlipy' and 'Brotli' packages # since they share an import name. The top branches # are for 'brotlipy' and bottom branches for 'Brotli' - def __init__(self): + def __init__(self) -> None: self._obj = brotli.Decompressor() if hasattr(self._obj, "decompress"): - self.decompress = self._obj.decompress + setattr(self, "decompress", self._obj.decompress) else: - self.decompress = self._obj.process + setattr(self, "decompress", self._obj.process) - def flush(self): + def flush(self) -> bytes: if hasattr(self._obj, "flush"): - return self._obj.flush() + return self._obj.flush() # type: ignore[no-any-return] return b"" -class MultiDecoder(object): +if zstd is not None: + + class ZstdDecoder(ContentDecoder): + def __init__(self) -> None: + self._obj = zstd.ZstdDecompressor().decompressobj() + + def decompress(self, data: bytes) -> bytes: + if not data: + return b"" + data_parts = [self._obj.decompress(data)] + while self._obj.eof and self._obj.unused_data: + unused_data = self._obj.unused_data + self._obj = zstd.ZstdDecompressor().decompressobj() + data_parts.append(self._obj.decompress(unused_data)) + return b"".join(data_parts) + + def flush(self) -> bytes: + ret = self._obj.flush() # note: this is a no-op + if not self._obj.eof: + raise DecodeError("Zstandard data is incomplete") + return ret # type: ignore[no-any-return] + + +class MultiDecoder(ContentDecoder): """ From RFC7231: If one or more encodings have been applied to a representation, the @@ -135,19 +192,19 @@ class MultiDecoder(object): they were applied. """ - def __init__(self, modes): + def __init__(self, modes: str) -> None: self._decoders = [_get_decoder(m.strip()) for m in modes.split(",")] - def flush(self): + def flush(self) -> bytes: return self._decoders[0].flush() - def decompress(self, data): + def decompress(self, data: bytes) -> bytes: for d in reversed(self._decoders): data = d.decompress(data) return data -def _get_decoder(mode): +def _get_decoder(mode: str) -> ContentDecoder: if "," in mode: return MultiDecoder(mode) @@ -157,10 +214,294 @@ def _get_decoder(mode): if brotli is not None and mode == "br": return BrotliDecoder() + if zstd is not None and mode == "zstd": + return ZstdDecoder() + return DeflateDecoder() -class HTTPResponse(io.IOBase): +class BytesQueueBuffer: + """Memory-efficient bytes buffer + + To return decoded data in read() and still follow the BufferedIOBase API, we need a + buffer to always return the correct amount of bytes. + + This buffer should be filled using calls to put() + + Our maximum memory usage is determined by the sum of the size of: + + * self.buffer, which contains the full data + * the largest chunk that we will copy in get() + + The worst case scenario is a single chunk, in which case we'll make a full copy of + the data inside get(). + """ + + def __init__(self) -> None: + self.buffer: typing.Deque[bytes] = collections.deque() + self._size: int = 0 + + def __len__(self) -> int: + return self._size + + def put(self, data: bytes) -> None: + self.buffer.append(data) + self._size += len(data) + + def get(self, n: int) -> bytes: + if n == 0: + return b"" + elif not self.buffer: + raise RuntimeError("buffer is empty") + elif n < 0: + raise ValueError("n should be > 0") + + fetched = 0 + ret = io.BytesIO() + while fetched < n: + remaining = n - fetched + chunk = self.buffer.popleft() + chunk_length = len(chunk) + if remaining < chunk_length: + left_chunk, right_chunk = chunk[:remaining], chunk[remaining:] + ret.write(left_chunk) + self.buffer.appendleft(right_chunk) + self._size -= remaining + break + else: + ret.write(chunk) + self._size -= chunk_length + fetched += chunk_length + + if not self.buffer: + break + + return ret.getvalue() + + +class BaseHTTPResponse(io.IOBase): + CONTENT_DECODERS = ["gzip", "deflate"] + if brotli is not None: + CONTENT_DECODERS += ["br"] + if zstd is not None: + CONTENT_DECODERS += ["zstd"] + REDIRECT_STATUSES = [301, 302, 303, 307, 308] + + DECODER_ERROR_CLASSES: tuple[type[Exception], ...] = (IOError, zlib.error) + if brotli is not None: + DECODER_ERROR_CLASSES += (brotli.error,) + + if zstd is not None: + DECODER_ERROR_CLASSES += (zstd.ZstdError,) + + def __init__( + self, + *, + headers: typing.Mapping[str, str] | typing.Mapping[bytes, bytes] | None = None, + status: int, + version: int, + reason: str | None, + decode_content: bool, + request_url: str | None, + retries: Retry | None = None, + ) -> None: + if isinstance(headers, HTTPHeaderDict): + self.headers = headers + else: + self.headers = HTTPHeaderDict(headers) # type: ignore[arg-type] + self.status = status + self.version = version + self.reason = reason + self.decode_content = decode_content + self._has_decoded_content = False + self._request_url: str | None = request_url + self.retries = retries + + self.chunked = False + tr_enc = self.headers.get("transfer-encoding", "").lower() + # Don't incur the penalty of creating a list and then discarding it + encodings = (enc.strip() for enc in tr_enc.split(",")) + if "chunked" in encodings: + self.chunked = True + + self._decoder: ContentDecoder | None = None + + def get_redirect_location(self) -> str | None | Literal[False]: + """ + Should we redirect and where to? + + :returns: Truthy redirect location string if we got a redirect status + code and valid location. ``None`` if redirect status and no + location. ``False`` if not a redirect status code. + """ + if self.status in self.REDIRECT_STATUSES: + return self.headers.get("location") + return False + + @property + def data(self) -> bytes: + raise NotImplementedError() + + def json(self) -> typing.Any: + """ + Parses the body of the HTTP response as JSON. + + To use a custom JSON decoder pass the result of :attr:`HTTPResponse.data` to the decoder. + + This method can raise either `UnicodeDecodeError` or `json.JSONDecodeError`. + + Read more :ref:`here `. + """ + data = self.data.decode("utf-8") + return _json.loads(data) + + @property + def url(self) -> str | None: + raise NotImplementedError() + + @url.setter + def url(self, url: str | None) -> None: + raise NotImplementedError() + + @property + def connection(self) -> HTTPConnection | None: + raise NotImplementedError() + + @property + def retries(self) -> Retry | None: + return self._retries + + @retries.setter + def retries(self, retries: Retry | None) -> None: + # Override the request_url if retries has a redirect location. + if retries is not None and retries.history: + self.url = retries.history[-1].redirect_location + self._retries = retries + + def stream( + self, amt: int | None = 2**16, decode_content: bool | None = None + ) -> typing.Iterator[bytes]: + raise NotImplementedError() + + def read( + self, + amt: int | None = None, + decode_content: bool | None = None, + cache_content: bool = False, + ) -> bytes: + raise NotImplementedError() + + def read_chunked( + self, + amt: int | None = None, + decode_content: bool | None = None, + ) -> typing.Iterator[bytes]: + raise NotImplementedError() + + def release_conn(self) -> None: + raise NotImplementedError() + + def drain_conn(self) -> None: + raise NotImplementedError() + + def close(self) -> None: + raise NotImplementedError() + + def _init_decoder(self) -> None: + """ + Set-up the _decoder attribute if necessary. + """ + # Note: content-encoding value should be case-insensitive, per RFC 7230 + # Section 3.2 + content_encoding = self.headers.get("content-encoding", "").lower() + if self._decoder is None: + if content_encoding in self.CONTENT_DECODERS: + self._decoder = _get_decoder(content_encoding) + elif "," in content_encoding: + encodings = [ + e.strip() + for e in content_encoding.split(",") + if e.strip() in self.CONTENT_DECODERS + ] + if encodings: + self._decoder = _get_decoder(content_encoding) + + def _decode( + self, data: bytes, decode_content: bool | None, flush_decoder: bool + ) -> bytes: + """ + Decode the data passed in and potentially flush the decoder. + """ + if not decode_content: + if self._has_decoded_content: + raise RuntimeError( + "Calling read(decode_content=False) is not supported after " + "read(decode_content=True) was called." + ) + return data + + try: + if self._decoder: + data = self._decoder.decompress(data) + self._has_decoded_content = True + except self.DECODER_ERROR_CLASSES as e: + content_encoding = self.headers.get("content-encoding", "").lower() + raise DecodeError( + "Received response with content-encoding: %s, but " + "failed to decode it." % content_encoding, + e, + ) from e + if flush_decoder: + data += self._flush_decoder() + + return data + + def _flush_decoder(self) -> bytes: + """ + Flushes the decoder. Should only be called if the decoder is actually + being used. + """ + if self._decoder: + return self._decoder.decompress(b"") + self._decoder.flush() + return b"" + + # Compatibility methods for `io` module + def readinto(self, b: bytearray) -> int: + temp = self.read(len(b)) + if len(temp) == 0: + return 0 + else: + b[: len(temp)] = temp + return len(temp) + + # Compatibility methods for http.client.HTTPResponse + def getheaders(self) -> HTTPHeaderDict: + 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: str, default: str | None = None) -> str | 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) + + # Compatibility method for http.cookiejar + def info(self) -> HTTPHeaderDict: + return self.headers + + def geturl(self) -> str | None: + return self.url + + +class HTTPResponse(BaseHTTPResponse): """ HTTP Response container. @@ -193,99 +534,74 @@ class HTTPResponse(io.IOBase): value of Content-Length header, if present. Otherwise, raise error. """ - CONTENT_DECODERS = ["gzip", "deflate"] - if brotli is not None: - CONTENT_DECODERS += ["br"] - REDIRECT_STATUSES = [301, 302, 303, 307, 308] - def __init__( self, - body="", - headers=None, - status=0, - version=0, - reason=None, - strict=0, - preload_content=True, - decode_content=True, - original_response=None, - pool=None, - connection=None, - msg=None, - retries=None, - enforce_content_length=False, - request_method=None, - request_url=None, - auto_close=True, - ): + body: _TYPE_BODY = "", + headers: typing.Mapping[str, str] | typing.Mapping[bytes, bytes] | None = None, + status: int = 0, + version: int = 0, + reason: str | None = None, + preload_content: bool = True, + decode_content: bool = True, + original_response: _HttplibHTTPResponse | None = None, + pool: HTTPConnectionPool | None = None, + connection: HTTPConnection | None = None, + msg: _HttplibHTTPMessage | None = None, + retries: Retry | None = None, + enforce_content_length: bool = True, + request_method: str | None = None, + request_url: str | None = None, + auto_close: bool = True, + ) -> None: + super().__init__( + headers=headers, + status=status, + version=version, + reason=reason, + decode_content=decode_content, + request_url=request_url, + retries=retries, + ) - if isinstance(headers, HTTPHeaderDict): - self.headers = headers - else: - self.headers = HTTPHeaderDict(headers) - self.status = status - self.version = version - self.reason = reason - self.strict = strict - self.decode_content = decode_content - self.retries = retries self.enforce_content_length = enforce_content_length self.auto_close = auto_close - self._decoder = None self._body = None - self._fp = None + self._fp: _HttplibHTTPResponse | None = None self._original_response = original_response self._fp_bytes_read = 0 self.msg = msg - self._request_url = request_url - if body and isinstance(body, (six.string_types, bytes)): + if body and isinstance(body, (str, bytes)): self._body = body self._pool = pool self._connection = connection if hasattr(body, "read"): - self._fp = body + self._fp = body # type: ignore[assignment] # Are we using the chunked-style of transfer encoding? - self.chunked = False - self.chunk_left = None - tr_enc = self.headers.get("transfer-encoding", "").lower() - # Don't incur the penalty of creating a list and then discarding it - encodings = (enc.strip() for enc in tr_enc.split(",")) - if "chunked" in encodings: - self.chunked = True + self.chunk_left: int | None = None # Determine length of response self.length_remaining = self._init_length(request_method) + # Used to return the correct amount of bytes for partial read()s + self._decoded_buffer = BytesQueueBuffer() + # If requested, preload the body. if preload_content and not self._body: self._body = self.read(decode_content=decode_content) - def get_redirect_location(self): - """ - Should we redirect and where to? - - :returns: Truthy redirect location string if we got a redirect status - code and valid location. ``None`` if redirect status and no - location. ``False`` if not a redirect status code. - """ - if self.status in self.REDIRECT_STATUSES: - return self.headers.get("location") - - return False - - def release_conn(self): + def release_conn(self) -> None: if not self._pool or not self._connection: - return + return None self._pool._put_conn(self._connection) self._connection = None - def drain_conn(self): + def drain_conn(self) -> None: """ Read and discard any remaining HTTP response data in the response connection. @@ -293,26 +609,28 @@ class HTTPResponse(io.IOBase): """ try: self.read() - except (HTTPError, SocketError, BaseSSLError, HTTPException): + except (HTTPError, OSError, BaseSSLError, HTTPException): pass @property - def data(self): + def data(self) -> bytes: # For backwards-compat with earlier urllib3 0.4 and earlier. if self._body: - return self._body + return self._body # type: ignore[return-value] if self._fp: return self.read(cache_content=True) + return None # type: ignore[return-value] + @property - def connection(self): + def connection(self) -> HTTPConnection | None: return self._connection - def isclosed(self): + def isclosed(self) -> bool: return is_fp_closed(self._fp) - def tell(self): + def tell(self) -> int: """ Obtain the number of bytes pulled over the wire so far. May differ from the amount of content returned by :meth:``urllib3.response.HTTPResponse.read`` @@ -320,13 +638,14 @@ class HTTPResponse(io.IOBase): """ return self._fp_bytes_read - def _init_length(self, request_method): + def _init_length(self, request_method: str | None) -> int | None: """ Set initial length value for Response content if available. """ - length = self.headers.get("content-length") + length: int | None + content_length: str | None = self.headers.get("content-length") - if length is not None: + if content_length is not None: if self.chunked: # This Response will fail with an IncompleteRead if it can't be # received as chunked. This method falls back to attempt reading @@ -346,11 +665,11 @@ class HTTPResponse(io.IOBase): # (e.g. Content-Length: 42, 42). This line ensures the values # are all valid ints and that as long as the `set` length is 1, # all values are the same. Otherwise, the header is invalid. - lengths = set([int(val) for val in length.split(",")]) + lengths = {int(val) for val in content_length.split(",")} if len(lengths) > 1: raise InvalidHeader( "Content-Length contained multiple " - "unmatching values (%s)" % length + "unmatching values (%s)" % content_length ) length = lengths.pop() except ValueError: @@ -359,6 +678,9 @@ class HTTPResponse(io.IOBase): if length < 0: length = None + else: # if content_length is None + length = None + # Convert status to int for comparison # In some cases, httplib returns a status of "_UNKNOWN" try: @@ -372,64 +694,8 @@ class HTTPResponse(io.IOBase): return length - def _init_decoder(self): - """ - Set-up the _decoder attribute if necessary. - """ - # Note: content-encoding value should be case-insensitive, per RFC 7230 - # Section 3.2 - content_encoding = self.headers.get("content-encoding", "").lower() - if self._decoder is None: - if content_encoding in self.CONTENT_DECODERS: - self._decoder = _get_decoder(content_encoding) - elif "," in content_encoding: - encodings = [ - e.strip() - for e in content_encoding.split(",") - if e.strip() in self.CONTENT_DECODERS - ] - if len(encodings): - self._decoder = _get_decoder(content_encoding) - - DECODER_ERROR_CLASSES = (IOError, zlib.error) - if brotli is not None: - DECODER_ERROR_CLASSES += (brotli.error,) - - def _decode(self, data, decode_content, flush_decoder): - """ - Decode the data passed in and potentially flush the decoder. - """ - if not decode_content: - return data - - try: - if self._decoder: - data = self._decoder.decompress(data) - except self.DECODER_ERROR_CLASSES as e: - content_encoding = self.headers.get("content-encoding", "").lower() - raise DecodeError( - "Received response with content-encoding: %s, but " - "failed to decode it." % content_encoding, - e, - ) - if flush_decoder: - data += self._flush_decoder() - - return data - - def _flush_decoder(self): - """ - Flushes the decoder. Should only be called if the decoder is actually - being used. - """ - if self._decoder: - buf = self._decoder.decompress(b"") - return buf + self._decoder.flush() - - return b"" - @contextmanager - def _error_catcher(self): + def _error_catcher(self) -> typing.Generator[None, None, None]: """ Catch low-level python exceptions, instead re-raising urllib3 variants, so that low-level exceptions are not leaked in the @@ -443,22 +709,22 @@ class HTTPResponse(io.IOBase): try: yield - except SocketTimeout: + except SocketTimeout as e: # FIXME: Ideally we'd like to include the url in the ReadTimeoutError but # there is yet no clean way to get at it from this context. - raise ReadTimeoutError(self._pool, None, "Read timed out.") + raise ReadTimeoutError(self._pool, None, "Read timed out.") from e # type: ignore[arg-type] except BaseSSLError as e: # FIXME: Is there a better way to differentiate between SSLErrors? if "read operation timed out" not in str(e): # SSL errors related to framing/MAC get wrapped and reraised here - raise SSLError(e) + raise SSLError(e) from e - raise ReadTimeoutError(self._pool, None, "Read timed out.") + raise ReadTimeoutError(self._pool, None, "Read timed out.") from e # type: ignore[arg-type] - except (HTTPException, SocketError) as e: + except (HTTPException, OSError) as e: # This includes IncompleteRead. - raise ProtocolError("Connection broken: %r" % e, e) + raise ProtocolError(f"Connection broken: {e!r}", e) from e # If no exception is thrown, we should avoid cleaning up # unnecessarily. @@ -484,7 +750,7 @@ class HTTPResponse(io.IOBase): if self._original_response and self._original_response.isclosed(): self.release_conn() - def _fp_read(self, amt): + def _fp_read(self, amt: int | None = None) -> bytes: """ 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 @@ -499,7 +765,7 @@ class HTTPResponse(io.IOBase): * CPython < 3.10 only when `amt` does not fit 32-bit int. """ assert self._fp - c_int_max = 2 ** 31 - 1 + c_int_max = 2**31 - 1 if ( ( (amt and amt > c_int_max) @@ -515,7 +781,7 @@ class HTTPResponse(io.IOBase): # `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 + max_chunk_amt = 2**28 while amt is None or amt != 0: if amt is not None: chunk_amt = min(amt, max_chunk_amt) @@ -532,7 +798,54 @@ class HTTPResponse(io.IOBase): # 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): + def _raw_read( + self, + amt: int | None = None, + ) -> bytes: + """ + Reads `amt` of bytes from the socket. + """ + if self._fp is None: + return None # type: ignore[return-value] + + 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 not None and amt != 0 and not data: + # Platform-specific: Buggy versions of Python. + # Close the connection when no data is returned + # + # This is redundant to what httplib/http.client _should_ + # already do. However, versions of python released before + # December 15, 2012 (http://bugs.python.org/issue16298) do + # not properly close the connection in all cases. There is + # no harm in redundantly calling close. + self._fp.close() + if ( + self.enforce_content_length + and self.length_remaining is not None + and self.length_remaining != 0 + ): + # This is an edge case that httplib failed to cover due + # to concerns of backward compatibility. We're + # addressing it here to make sure IncompleteRead is + # raised during streaming, so all calls with incorrect + # Content-Length are caught. + raise IncompleteRead(self._fp_bytes_read, self.length_remaining) + + if data: + self._fp_bytes_read += len(data) + if self.length_remaining is not None: + self.length_remaining -= len(data) + return data + + def read( + self, + amt: int | None = None, + decode_content: bool | None = None, + cache_content: bool = False, + ) -> bytes: """ Similar to :meth:`http.client.HTTPResponse.read`, but with two additional parameters: ``decode_content`` and ``cache_content``. @@ -557,54 +870,50 @@ class HTTPResponse(io.IOBase): if decode_content is None: decode_content = self.decode_content - if self._fp is None: - return + if amt is not None: + cache_content = False - flush_decoder = False - fp_closed = getattr(self._fp, "closed", False) + if len(self._decoded_buffer) >= amt: + return self._decoded_buffer.get(amt) - with self._error_catcher(): - data = self._fp_read(amt) if not fp_closed else b"" - if amt is None: - flush_decoder = True - else: - cache_content = False - if ( - amt != 0 and not data - ): # Platform-specific: Buggy versions of Python. - # Close the connection when no data is returned - # - # This is redundant to what httplib/http.client _should_ - # already do. However, versions of python released before - # December 15, 2012 (http://bugs.python.org/issue16298) do - # not properly close the connection in all cases. There is - # no harm in redundantly calling close. - self._fp.close() - flush_decoder = True - if self.enforce_content_length and self.length_remaining not in ( - 0, - None, - ): - # This is an edge case that httplib failed to cover due - # to concerns of backward compatibility. We're - # addressing it here to make sure IncompleteRead is - # raised during streaming, so all calls with incorrect - # Content-Length are caught. - raise IncompleteRead(self._fp_bytes_read, self.length_remaining) + data = self._raw_read(amt) - if data: - self._fp_bytes_read += len(data) - if self.length_remaining is not None: - self.length_remaining -= len(data) + flush_decoder = amt is None or (amt != 0 and not data) + if not data and len(self._decoded_buffer) == 0: + return data + + if amt is None: data = self._decode(data, decode_content, flush_decoder) - if cache_content: self._body = data + else: + # do not waste memory on buffer when not decoding + if not decode_content: + if self._has_decoded_content: + raise RuntimeError( + "Calling read(decode_content=False) is not supported after " + "read(decode_content=True) was called." + ) + return data + + decoded_data = self._decode(data, decode_content, flush_decoder) + self._decoded_buffer.put(decoded_data) + + while len(self._decoded_buffer) < amt and data: + # TODO make sure to initially read enough data to get past the headers + # For example, the GZ file header takes 10 bytes, we don't want to read + # it one byte at a time + data = self._raw_read(amt) + decoded_data = self._decode(data, decode_content, flush_decoder) + self._decoded_buffer.put(decoded_data) + data = self._decoded_buffer.get(amt) return data - def stream(self, amt=2 ** 16, decode_content=None): + def stream( + self, amt: int | None = 2**16, decode_content: bool | None = None + ) -> typing.Generator[bytes, None, None]: """ A generator wrapper for the read() method. A call will block until ``amt`` bytes have been read from the connection or until the @@ -621,73 +930,20 @@ class HTTPResponse(io.IOBase): 'content-encoding' header. """ if self.chunked and self.supports_chunked_reads(): - for line in self.read_chunked(amt, decode_content=decode_content): - yield line + yield from self.read_chunked(amt, decode_content=decode_content) else: - while not is_fp_closed(self._fp): + while not is_fp_closed(self._fp) or len(self._decoded_buffer) > 0: data = self.read(amt=amt, decode_content=decode_content) if data: yield data - @classmethod - def from_httplib(ResponseCls, r, **response_kw): - """ - 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 - with ``original_response=r``. - """ - headers = r.msg - - if not isinstance(headers, HTTPHeaderDict): - 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) - resp = ResponseCls( - body=r, - headers=headers, - status=r.status, - version=r.version, - reason=r.reason, - strict=strict, - original_response=r, - **response_kw - ) - return resp - - # 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 - def info(self): - return self.headers - # Overrides from io.IOBase - def close(self): - if not self.closed: + def readable(self) -> bool: + return True + + def close(self) -> None: + if not self.closed and self._fp: self._fp.close() if self._connection: @@ -697,9 +953,9 @@ class HTTPResponse(io.IOBase): io.IOBase.close(self) @property - def closed(self): + def closed(self) -> bool: if not self.auto_close: - return io.IOBase.closed.__get__(self) + return io.IOBase.closed.__get__(self) # type: ignore[no-any-return] elif self._fp is None: return True elif hasattr(self._fp, "isclosed"): @@ -709,18 +965,18 @@ class HTTPResponse(io.IOBase): else: return True - def fileno(self): + def fileno(self) -> int: if self._fp is None: - raise IOError("HTTPResponse has no file to get a fileno from") + raise OSError("HTTPResponse has no file to get a fileno from") elif hasattr(self._fp, "fileno"): return self._fp.fileno() else: - raise IOError( + raise OSError( "The file-like object this HTTPResponse is wrapped " "around has no file descriptor" ) - def flush(self): + def flush(self) -> None: if ( self._fp is not None and hasattr(self._fp, "flush") @@ -728,20 +984,7 @@ class HTTPResponse(io.IOBase): ): return self._fp.flush() - def readable(self): - # This method is required for `io` module compatibility. - return True - - def readinto(self, b): - # This method is required for `io` module compatibility. - temp = self.read(len(b)) - if len(temp) == 0: - return 0 - else: - b[: len(temp)] = temp - return len(temp) - - def supports_chunked_reads(self): + def supports_chunked_reads(self) -> bool: """ Checks if the underlying file-like object looks like a :class:`http.client.HTTPResponse` object. We do this by testing for @@ -750,43 +993,45 @@ class HTTPResponse(io.IOBase): """ return hasattr(self._fp, "fp") - def _update_chunk_length(self): + def _update_chunk_length(self) -> None: # First, we'll figure out length of a chunk and then # we'll try to read it from socket. if self.chunk_left is not None: - return - line = self._fp.fp.readline() + return None + line = self._fp.fp.readline() # type: ignore[union-attr] line = line.split(b";", 1)[0] try: self.chunk_left = int(line, 16) except ValueError: # Invalid chunked protocol response, abort. self.close() - raise InvalidChunkLength(self, line) + raise InvalidChunkLength(self, line) from None - def _handle_chunk(self, amt): + def _handle_chunk(self, amt: int | None) -> bytes: returned_chunk = None if amt is None: - chunk = self._fp._safe_read(self.chunk_left) + chunk = self._fp._safe_read(self.chunk_left) # type: ignore[union-attr] returned_chunk = chunk - self._fp._safe_read(2) # Toss the CRLF at the end of the chunk. + self._fp._safe_read(2) # type: ignore[union-attr] # Toss the CRLF at the end of the chunk. self.chunk_left = None - elif amt < self.chunk_left: - value = self._fp._safe_read(amt) + elif self.chunk_left is not None and amt < self.chunk_left: + value = self._fp._safe_read(amt) # type: ignore[union-attr] self.chunk_left = self.chunk_left - amt returned_chunk = value elif amt == self.chunk_left: - value = self._fp._safe_read(amt) - self._fp._safe_read(2) # Toss the CRLF at the end of the chunk. + value = self._fp._safe_read(amt) # type: ignore[union-attr] + self._fp._safe_read(2) # type: ignore[union-attr] # Toss the CRLF at the end of the chunk. self.chunk_left = None returned_chunk = value else: # amt > self.chunk_left - returned_chunk = self._fp._safe_read(self.chunk_left) - self._fp._safe_read(2) # Toss the CRLF at the end of the chunk. + returned_chunk = self._fp._safe_read(self.chunk_left) # type: ignore[union-attr] + self._fp._safe_read(2) # type: ignore[union-attr] # Toss the CRLF at the end of the chunk. self.chunk_left = None - return returned_chunk + return returned_chunk # type: ignore[no-any-return] - def read_chunked(self, amt=None, decode_content=None): + def read_chunked( + self, amt: int | None = None, decode_content: bool | None = None + ) -> typing.Generator[bytes, None, None]: """ Similar to :meth:`HTTPResponse.read`, but with an additional parameter: ``decode_content``. @@ -817,12 +1062,12 @@ class HTTPResponse(io.IOBase): # Don't bother reading the body of a HEAD request. if self._original_response and is_response_to_head(self._original_response): self._original_response.close() - return + return None # If a response is already read and closed # then return immediately. - if self._fp.fp is None: - return + if self._fp.fp is None: # type: ignore[union-attr] + return None while True: self._update_chunk_length() @@ -844,7 +1089,7 @@ class HTTPResponse(io.IOBase): yield decoded # Chunk content ends with \r\n: discard it. - while True: + while self._fp is not None: line = self._fp.fp.readline() if not line: # Some sites may not end with '\r\n'. @@ -856,27 +1101,29 @@ class HTTPResponse(io.IOBase): if self._original_response: self._original_response.close() - def geturl(self): + @property + def url(self) -> str | None: """ Returns the URL that was the source of this response. If the request that generated this response redirected, this method will return the final redirect location. """ - if self.retries is not None and len(self.retries.history): - return self.retries.history[-1].redirect_location - else: - return self._request_url + return self._request_url - def __iter__(self): - buffer = [] + @url.setter + def url(self, url: str) -> None: + self._request_url = url + + def __iter__(self) -> typing.Iterator[bytes]: + buffer: list[bytes] = [] for chunk in self.stream(decode_content=True): if b"\n" in chunk: - chunk = chunk.split(b"\n") - yield b"".join(buffer) + chunk[0] + b"\n" - for x in chunk[1:-1]: + chunks = chunk.split(b"\n") + yield b"".join(buffer) + chunks[0] + b"\n" + for x in chunks[1:-1]: yield x + b"\n" - if chunk[-1]: - buffer = [chunk[-1]] + if chunks[-1]: + buffer = [chunks[-1]] else: buffer = [] else: diff --git a/lib/urllib3/util/__init__.py b/lib/urllib3/util/__init__.py index 4547fc52..ff56c55b 100644 --- a/lib/urllib3/util/__init__.py +++ b/lib/urllib3/util/__init__.py @@ -1,46 +1,41 @@ -from __future__ import absolute_import - # For backwards compatibility, provide imports that used to be here. +from __future__ import annotations + from .connection import is_connection_dropped from .request import SKIP_HEADER, SKIPPABLE_HEADERS, make_headers from .response import is_fp_closed from .retry import Retry from .ssl_ import ( ALPN_PROTOCOLS, - HAS_SNI, IS_PYOPENSSL, IS_SECURETRANSPORT, - PROTOCOL_TLS, SSLContext, assert_fingerprint, + create_urllib3_context, resolve_cert_reqs, resolve_ssl_version, ssl_wrap_socket, ) -from .timeout import Timeout, current_time -from .url import Url, get_host, parse_url, split_first +from .timeout import Timeout +from .url import Url, parse_url from .wait import wait_for_read, wait_for_write __all__ = ( - "HAS_SNI", "IS_PYOPENSSL", "IS_SECURETRANSPORT", "SSLContext", - "PROTOCOL_TLS", "ALPN_PROTOCOLS", "Retry", "Timeout", "Url", "assert_fingerprint", - "current_time", + "create_urllib3_context", "is_connection_dropped", "is_fp_closed", - "get_host", "parse_url", "make_headers", "resolve_cert_reqs", "resolve_ssl_version", - "split_first", "ssl_wrap_socket", "wait_for_read", "wait_for_write", diff --git a/lib/urllib3/util/connection.py b/lib/urllib3/util/connection.py index 6af1138f..5c7da73f 100644 --- a/lib/urllib3/util/connection.py +++ b/lib/urllib3/util/connection.py @@ -1,33 +1,23 @@ -from __future__ import absolute_import +from __future__ import annotations import socket +import typing -from ..contrib import _appengine_environ from ..exceptions import LocationParseError -from ..packages import six -from .wait import NoWayToWaitForSocketError, wait_for_read +from .timeout import _DEFAULT_TIMEOUT, _TYPE_TIMEOUT + +_TYPE_SOCKET_OPTIONS = typing.Sequence[typing.Tuple[int, int, typing.Union[int, bytes]]] + +if typing.TYPE_CHECKING: + from .._base_connection import BaseHTTPConnection -def is_connection_dropped(conn): # Platform-specific +def is_connection_dropped(conn: BaseHTTPConnection) -> bool: # Platform-specific """ Returns True if the connection is dropped and should be closed. - - :param conn: - :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. + :param conn: :class:`urllib3.connection.HTTPConnection` object. """ - sock = getattr(conn, "sock", False) - if sock is False: # Platform-specific: AppEngine - return False - if sock is None: # Connection already closed (such as by httplib). - return True - try: - # Returns True if readable, which here means it's been dropped - return wait_for_read(sock, timeout=0.0) - except NoWayToWaitForSocketError: # Platform-specific: AppEngine - return False + return not conn.is_connected # This function is copied from socket.py in the Python 2.7 standard @@ -35,11 +25,11 @@ def is_connection_dropped(conn): # Platform-specific # One additional modification is that we avoid binding to IPv6 servers # discovered in DNS if the system doesn't have IPv6 functionality. def create_connection( - address, - timeout=socket._GLOBAL_DEFAULT_TIMEOUT, - source_address=None, - socket_options=None, -): + address: tuple[str, int], + timeout: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT, + source_address: tuple[str, int] | None = None, + socket_options: _TYPE_SOCKET_OPTIONS | None = None, +) -> socket.socket: """Connect to *address* and return the socket object. Convenience function. Connect to *address* (a 2-tuple ``(host, @@ -65,9 +55,7 @@ def create_connection( try: host.encode("idna") except UnicodeError: - return six.raise_from( - LocationParseError(u"'%s', label empty or too long" % host), None - ) + raise LocationParseError(f"'{host}', label empty or too long") from None for res in socket.getaddrinfo(host, port, family, socket.SOCK_STREAM): af, socktype, proto, canonname, sa = res @@ -78,26 +66,33 @@ def create_connection( # If provided, set socket level options before connecting. _set_socket_options(sock, socket_options) - if timeout is not socket._GLOBAL_DEFAULT_TIMEOUT: + if timeout is not _DEFAULT_TIMEOUT: sock.settimeout(timeout) if source_address: sock.bind(source_address) sock.connect(sa) + # Break explicitly a reference cycle + err = None return sock - except socket.error as e: - err = e + except OSError as _: + err = _ if sock is not None: sock.close() - sock = None if err is not None: - raise err - - raise socket.error("getaddrinfo returns an empty list") + try: + raise err + finally: + # Break explicitly a reference cycle + err = None + else: + raise OSError("getaddrinfo returns an empty list") -def _set_socket_options(sock, options): +def _set_socket_options( + sock: socket.socket, options: _TYPE_SOCKET_OPTIONS | None +) -> None: if options is None: return @@ -105,7 +100,7 @@ def _set_socket_options(sock, options): sock.setsockopt(*opt) -def allowed_gai_family(): +def allowed_gai_family() -> socket.AddressFamily: """This function is designed to work in the context of getaddrinfo, where family=socket.AF_UNSPEC is the default and will perform a DNS search for both IPv6 and IPv4 records.""" @@ -116,18 +111,11 @@ def allowed_gai_family(): return family -def _has_ipv6(host): +def _has_ipv6(host: str) -> bool: """Returns True if the system can bind an IPv6 address.""" sock = None has_ipv6 = False - # App Engine doesn't support IPV6 sockets and actually has a quota on the - # number of sockets that can be used, so just early out here instead of - # creating a socket needlessly. - # See https://github.com/urllib3/urllib3/issues/1446 - if _appengine_environ.is_appengine_sandbox(): - return False - if socket.has_ipv6: # has_ipv6 returns true if cPython was compiled with IPv6 support. # It does not tell us if the system has IPv6 support enabled. To diff --git a/lib/urllib3/util/proxy.py b/lib/urllib3/util/proxy.py index 2199cc7b..908fc662 100644 --- a/lib/urllib3/util/proxy.py +++ b/lib/urllib3/util/proxy.py @@ -1,9 +1,18 @@ -from .ssl_ import create_urllib3_context, resolve_cert_reqs, resolve_ssl_version +from __future__ import annotations + +import typing + +from .url import Url + +if typing.TYPE_CHECKING: + from ..connection import ProxyConfig def connection_requires_http_tunnel( - proxy_url=None, proxy_config=None, destination_scheme=None -): + proxy_url: Url | None = None, + proxy_config: ProxyConfig | None = None, + destination_scheme: str | None = None, +) -> bool: """ Returns True if the connection requires an HTTP CONNECT through the proxy. @@ -32,26 +41,3 @@ def connection_requires_http_tunnel( # 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 deleted file mode 100644 index 41784104..00000000 --- a/lib/urllib3/util/queue.py +++ /dev/null @@ -1,22 +0,0 @@ -import collections - -from ..packages import six -from ..packages.six.moves import queue - -if six.PY2: - # Queue is imported for side effects on MS Windows. See issue #229. - import Queue as _unused_module_Queue # noqa: F401 - - -class LifoQueue(queue.Queue): - def _init(self, _): - self.queue = collections.deque() - - def _qsize(self, len=len): - return len(self.queue) - - def _put(self, item): - self.queue.append(item) - - def _get(self): - return self.queue.pop() diff --git a/lib/urllib3/util/queue.pyi b/lib/urllib3/util/queue.pyi deleted file mode 100644 index e69de29b..00000000 diff --git a/lib/urllib3/util/request.py b/lib/urllib3/util/request.py index b574b081..7d6866f3 100644 --- a/lib/urllib3/util/request.py +++ b/lib/urllib3/util/request.py @@ -1,9 +1,15 @@ -from __future__ import absolute_import +from __future__ import annotations +import io +import typing from base64 import b64encode +from enum import Enum from ..exceptions import UnrewindableBodyError -from ..packages.six import b, integer_types +from .util import to_bytes + +if typing.TYPE_CHECKING: + from typing_extensions import Final # Pass as a value within ``headers`` to skip # emitting some HTTP headers that are added automatically. @@ -15,25 +21,45 @@ SKIPPABLE_HEADERS = frozenset(["accept-encoding", "host", "user-agent"]) ACCEPT_ENCODING = "gzip,deflate" try: try: - import brotlicffi as _unused_module_brotli # noqa: F401 + import brotlicffi as _unused_module_brotli # type: ignore[import] # noqa: F401 except ImportError: - import brotli as _unused_module_brotli # noqa: F401 + import brotli as _unused_module_brotli # type: ignore[import] # noqa: F401 except ImportError: pass else: ACCEPT_ENCODING += ",br" +try: + import zstandard as _unused_module_zstd # type: ignore[import] # noqa: F401 +except ImportError: + pass +else: + ACCEPT_ENCODING += ",zstd" -_FAILEDTELL = object() + +class _TYPE_FAILEDTELL(Enum): + token = 0 + + +_FAILEDTELL: Final[_TYPE_FAILEDTELL] = _TYPE_FAILEDTELL.token + +_TYPE_BODY_POSITION = typing.Union[int, _TYPE_FAILEDTELL] + +# When sending a request with these methods we aren't expecting +# a body so don't need to set an explicit 'Content-Length: 0' +# The reason we do this in the negative instead of tracking methods +# which 'should' have a body is because unknown methods should be +# treated as if they were 'POST' which *does* expect a body. +_METHODS_NOT_EXPECTING_BODY = {"GET", "HEAD", "DELETE", "TRACE", "OPTIONS", "CONNECT"} def make_headers( - keep_alive=None, - accept_encoding=None, - user_agent=None, - basic_auth=None, - proxy_basic_auth=None, - disable_cache=None, -): + keep_alive: bool | None = None, + accept_encoding: bool | list[str] | str | None = None, + user_agent: str | None = None, + basic_auth: str | None = None, + proxy_basic_auth: str | None = None, + disable_cache: bool | None = None, +) -> dict[str, str]: """ Shortcuts for generating request headers. @@ -42,7 +68,8 @@ def make_headers( :param accept_encoding: Can be a boolean, list, or string. - ``True`` translates to 'gzip,deflate'. + ``True`` translates to 'gzip,deflate'. If either the ``brotli`` or + ``brotlicffi`` package is installed 'gzip,deflate,br' is used instead. List will get joined by comma. String will be used as provided. @@ -61,14 +88,18 @@ def make_headers( :param disable_cache: If ``True``, adds 'cache-control: no-cache' header. - Example:: + Example: - >>> make_headers(keep_alive=True, user_agent="Batman/1.0") - {'connection': 'keep-alive', 'user-agent': 'Batman/1.0'} - >>> make_headers(accept_encoding=True) - {'accept-encoding': 'gzip,deflate'} + .. code-block:: python + + import urllib3 + + print(urllib3.util.make_headers(keep_alive=True, user_agent="Batman/1.0")) + # {'connection': 'keep-alive', 'user-agent': 'Batman/1.0'} + print(urllib3.util.make_headers(accept_encoding=True)) + # {'accept-encoding': 'gzip,deflate'} """ - headers = {} + headers: dict[str, str] = {} if accept_encoding: if isinstance(accept_encoding, str): pass @@ -85,12 +116,14 @@ def make_headers( headers["connection"] = "keep-alive" if basic_auth: - headers["authorization"] = "Basic " + b64encode(b(basic_auth)).decode("utf-8") + headers[ + "authorization" + ] = f"Basic {b64encode(basic_auth.encode('latin-1')).decode()}" if proxy_basic_auth: - headers["proxy-authorization"] = "Basic " + b64encode( - b(proxy_basic_auth) - ).decode("utf-8") + headers[ + "proxy-authorization" + ] = f"Basic {b64encode(proxy_basic_auth.encode('latin-1')).decode()}" if disable_cache: headers["cache-control"] = "no-cache" @@ -98,7 +131,9 @@ def make_headers( return headers -def set_file_position(body, pos): +def set_file_position( + body: typing.Any, pos: _TYPE_BODY_POSITION | None +) -> _TYPE_BODY_POSITION | None: """ If a position is provided, move file to that point. Otherwise, we'll attempt to record a position for future use. @@ -108,7 +143,7 @@ def set_file_position(body, pos): elif getattr(body, "tell", None) is not None: try: pos = body.tell() - except (IOError, OSError): + except OSError: # This differentiates from None, allowing us to catch # a failed `tell()` later when trying to rewind the body. pos = _FAILEDTELL @@ -116,7 +151,7 @@ def set_file_position(body, pos): return pos -def rewind_body(body, body_pos): +def rewind_body(body: typing.IO[typing.AnyStr], body_pos: _TYPE_BODY_POSITION) -> None: """ Attempt to rewind body to a certain position. Primarily used for request redirects and retries. @@ -128,13 +163,13 @@ def rewind_body(body, body_pos): Position to seek to in file. """ body_seek = getattr(body, "seek", None) - if body_seek is not None and isinstance(body_pos, integer_types): + if body_seek is not None and isinstance(body_pos, int): try: body_seek(body_pos) - except (IOError, OSError): + except OSError as e: raise UnrewindableBodyError( "An error occurred when rewinding request body for redirect/retry." - ) + ) from e elif body_pos is _FAILEDTELL: raise UnrewindableBodyError( "Unable to record file position for rewinding " @@ -142,5 +177,80 @@ def rewind_body(body, body_pos): ) else: raise ValueError( - "body_pos must be of type integer, instead it was %s." % type(body_pos) + f"body_pos must be of type integer, instead it was {type(body_pos)}." ) + + +class ChunksAndContentLength(typing.NamedTuple): + chunks: typing.Iterable[bytes] | None + content_length: int | None + + +def body_to_chunks( + body: typing.Any | None, method: str, blocksize: int +) -> ChunksAndContentLength: + """Takes the HTTP request method, body, and blocksize and + transforms them into an iterable of chunks to pass to + socket.sendall() and an optional 'Content-Length' header. + + A 'Content-Length' of 'None' indicates the length of the body + can't be determined so should use 'Transfer-Encoding: chunked' + for framing instead. + """ + + chunks: typing.Iterable[bytes] | None + content_length: int | None + + # No body, we need to make a recommendation on 'Content-Length' + # based on whether that request method is expected to have + # a body or not. + if body is None: + chunks = None + if method.upper() not in _METHODS_NOT_EXPECTING_BODY: + content_length = 0 + else: + content_length = None + + # Bytes or strings become bytes + elif isinstance(body, (str, bytes)): + chunks = (to_bytes(body),) + content_length = len(chunks[0]) + + # File-like object, TODO: use seek() and tell() for length? + elif hasattr(body, "read"): + + def chunk_readable() -> typing.Iterable[bytes]: + nonlocal body, blocksize + encode = isinstance(body, io.TextIOBase) + while True: + datablock = body.read(blocksize) + if not datablock: + break + if encode: + datablock = datablock.encode("iso-8859-1") + yield datablock + + chunks = chunk_readable() + content_length = None + + # Otherwise we need to start checking via duck-typing. + else: + try: + # Check if the body implements the buffer API. + mv = memoryview(body) + except TypeError: + try: + # Check if the body is an iterable + chunks = iter(body) + content_length = None + except TypeError: + raise TypeError( + f"'body' must be a bytes-like object, file-like " + f"object, or iterable. Instead was {body!r}" + ) from None + else: + # Since it implements the buffer API can be passed directly to socket.sendall() + chunks = (body,) + content_length = mv.nbytes + + return ChunksAndContentLength(chunks=chunks, content_length=content_length) diff --git a/lib/urllib3/util/response.py b/lib/urllib3/util/response.py index 5ea609cc..0f457869 100644 --- a/lib/urllib3/util/response.py +++ b/lib/urllib3/util/response.py @@ -1,12 +1,12 @@ -from __future__ import absolute_import +from __future__ import annotations +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): +def is_fp_closed(obj: object) -> bool: """ Checks whether a given file-like object is closed. @@ -17,27 +17,27 @@ def is_fp_closed(obj): try: # Check `isclosed()` first, in case Python3 doesn't set `closed`. # GH Issue #928 - return obj.isclosed() + return obj.isclosed() # type: ignore[no-any-return, attr-defined] except AttributeError: pass try: # Check via the official file-like-object way. - return obj.closed + return obj.closed # type: ignore[no-any-return, attr-defined] except AttributeError: pass try: # Check if the object is a container for another file-like object that # gets released on exhaustion (e.g. HTTPResponse). - return obj.fp is None + return obj.fp is None # type: ignore[attr-defined] except AttributeError: pass raise ValueError("Unable to determine whether fp is closed.") -def assert_header_parsing(headers): +def assert_header_parsing(headers: httplib.HTTPMessage) -> None: """ Asserts whether all headers have been successfully parsed. Extracts encountered errors from the result of parsing headers. @@ -53,55 +53,49 @@ def assert_header_parsing(headers): # This will fail silently if we pass in the wrong kind of parameter. # To make debugging easier add an explicit check. if not isinstance(headers, httplib.HTTPMessage): - raise TypeError("expected httplib.Message, got {0}.".format(type(headers))) - - defects = getattr(headers, "defects", None) - get_payload = getattr(headers, "get_payload", None) + raise TypeError(f"expected httplib.Message, got {type(headers)}.") unparsed_data = None - if get_payload: - # get_payload is actually email.message.Message.get_payload; - # we're only interested in the result if it's not a multipart message - if not headers.is_multipart(): - payload = get_payload() - 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 + # get_payload is actually email.message.Message.get_payload; + # we're only interested in the result if it's not a multipart message + if not headers.is_multipart(): + payload = headers.get_payload() - # 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 isinstance(payload, (bytes, str)): + unparsed_data = payload + + # 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 headers.defects + if not isinstance( + defect, (StartBoundaryNotFoundDefect, MultipartInvariantViolationDefect) + ) + ] if defects or unparsed_data: raise HeaderParsingError(defects=defects, unparsed_data=unparsed_data) -def is_response_to_head(response): +def is_response_to_head(response: httplib.HTTPResponse) -> bool: """ Checks whether the request of a response has been a HEAD-request. - Handles the quirks of AppEngine. :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 - if isinstance(method, int): # Platform-specific: Appengine - return method == 3 - return method.upper() == "HEAD" + method_str = response._method # type: str # type: ignore[attr-defined] + return method_str.upper() == "HEAD" diff --git a/lib/urllib3/util/retry.py b/lib/urllib3/util/retry.py index 2490d5e5..ea48afe3 100644 --- a/lib/urllib3/util/retry.py +++ b/lib/urllib3/util/retry.py @@ -1,12 +1,13 @@ -from __future__ import absolute_import +from __future__ import annotations import email import logging +import random import re import time -import warnings -from collections import namedtuple +import typing from itertools import takewhile +from types import TracebackType from ..exceptions import ( ConnectTimeoutError, @@ -17,97 +18,49 @@ from ..exceptions import ( ReadTimeoutError, ResponseError, ) -from ..packages import six +from .util import reraise + +if typing.TYPE_CHECKING: + from ..connectionpool import ConnectionPool + from ..response import BaseHTTPResponse log = logging.getLogger(__name__) # Data structure for representing the metadata of requests that result in a retry. -RequestHistory = namedtuple( - "RequestHistory", ["method", "url", "error", "status", "redirect_location"] -) +class RequestHistory(typing.NamedTuple): + method: str | None + url: str | None + error: Exception | None + status: int | None + redirect_location: str | None -# 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): +class Retry: """Retry configuration. Each retry attempt will create a new Retry object with updated values, so they can be safely reused. - Retries can be defined as a default for a pool:: + Retries can be defined as a default for a pool: + + .. code-block:: python retries = Retry(connect=5, read=2, redirect=5) http = PoolManager(retries=retries) - response = http.request('GET', 'http://example.com/') + response = http.request("GET", "https://example.com/") - Or per-request (which overrides the default for the pool):: + Or per-request (which overrides the default for the pool): - response = http.request('GET', 'http://example.com/', retries=Retry(10)) + .. code-block:: python - Retries can be disabled by passing ``False``:: + response = http.request("GET", "https://example.com/", retries=Retry(10)) - response = http.request('GET', 'http://example.com/', retries=False) + Retries can be disabled by passing ``False``: + + .. code-block:: python + + response = http.request("GET", "https://example.com/", retries=False) Errors will be wrapped in :class:`~urllib3.exceptions.MaxRetryError` unless retries are disabled, in which case the causing exception will be raised. @@ -169,21 +122,16 @@ class Retry(object): 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: + :param Collection 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_ALLOWED_METHODS`. - Set to a ``False`` value to retry on any verb. + Set to a ``None`` 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: + :param Collection 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 ``allowed_methods`` and the response status code is in ``status_forcelist``. @@ -195,13 +143,17 @@ class Retry(object): (most errors are resolved immediately by a second try without a delay). urllib3 will sleep for:: - {backoff factor} * (2 ** ({number of total retries} - 1)) + {backoff factor} * (2 ** ({number of previous retries})) - 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.DEFAULT_BACKOFF_MAX`. + seconds. If `backoff_jitter` is non-zero, this sleep is extended by:: - By default, backoff is disabled (set to 0). + random.uniform(0, {backoff jitter}) + + seconds. For example, if the backoff_factor is 0.1, then :func:`Retry.sleep` will + sleep for [0.0s, 0.2s, 0.4s, 0.8s, ...] between retries. No backoff will ever + be longer than `backoff_max`. + + By default, backoff is disabled (factor set to 0). :param bool raise_on_redirect: Whether, if the number of redirects is exhausted, to raise a MaxRetryError, or to return a response with a @@ -220,7 +172,7 @@ class Retry(object): Whether to respect Retry-After header on status codes defined as :attr:`Retry.RETRY_AFTER_STATUS_CODES` or not. - :param iterable remove_headers_on_redirect: + :param Collection remove_headers_on_redirect: Sequence of headers to remove from the request when a response indicating a redirect is returned before firing off the redirected request. @@ -237,48 +189,33 @@ class Retry(object): #: Default headers to be used for ``remove_headers_on_redirect`` DEFAULT_REMOVE_HEADERS_ON_REDIRECT = frozenset(["Authorization"]) - #: Maximum backoff time. + #: Default maximum backoff time. DEFAULT_BACKOFF_MAX = 120 + # Backward compatibility; assigned outside of the class. + DEFAULT: typing.ClassVar[Retry] + def __init__( self, - total=10, - connect=None, - read=None, - redirect=None, - status=None, - 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, - # 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 - + total: bool | int | None = 10, + connect: int | None = None, + read: int | None = None, + redirect: bool | int | None = None, + status: int | None = None, + other: int | None = None, + allowed_methods: typing.Collection[str] | None = DEFAULT_ALLOWED_METHODS, + status_forcelist: typing.Collection[int] | None = None, + backoff_factor: float = 0, + backoff_max: float = DEFAULT_BACKOFF_MAX, + raise_on_redirect: bool = True, + raise_on_status: bool = True, + history: tuple[RequestHistory, ...] | None = None, + respect_retry_after_header: bool = True, + remove_headers_on_redirect: typing.Collection[ + str + ] = DEFAULT_REMOVE_HEADERS_ON_REDIRECT, + backoff_jitter: float = 0.0, + ) -> None: self.total = total self.connect = connect self.read = read @@ -293,15 +230,17 @@ class Retry(object): self.status_forcelist = status_forcelist or set() self.allowed_methods = allowed_methods self.backoff_factor = backoff_factor + self.backoff_max = backoff_max self.raise_on_redirect = raise_on_redirect self.raise_on_status = raise_on_status - self.history = history or tuple() + self.history = history or () self.respect_retry_after_header = respect_retry_after_header self.remove_headers_on_redirect = frozenset( - [h.lower() for h in remove_headers_on_redirect] + h.lower() for h in remove_headers_on_redirect ) + self.backoff_jitter = backoff_jitter - def new(self, **kw): + def new(self, **kw: typing.Any) -> Retry: params = dict( total=self.total, connect=self.connect, @@ -309,36 +248,28 @@ class Retry(object): redirect=self.redirect, status=self.status, other=self.other, + allowed_methods=self.allowed_methods, status_forcelist=self.status_forcelist, backoff_factor=self.backoff_factor, + backoff_max=self.backoff_max, raise_on_redirect=self.raise_on_redirect, raise_on_status=self.raise_on_status, history=self.history, remove_headers_on_redirect=self.remove_headers_on_redirect, respect_retry_after_header=self.respect_retry_after_header, + backoff_jitter=self.backoff_jitter, ) - # 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) + return type(self)(**params) # type: ignore[arg-type] @classmethod - def from_int(cls, retries, redirect=True, default=None): + def from_int( + cls, + retries: Retry | bool | int | None, + redirect: bool | int | None = True, + default: Retry | bool | int | None = None, + ) -> Retry: """Backwards-compatibility for the old retries format.""" if retries is None: retries = default if default is not None else cls.DEFAULT @@ -351,7 +282,7 @@ class Retry(object): log.debug("Converted retries value: %r -> %r", retries, new_retries) return new_retries - def get_backoff_time(self): + def get_backoff_time(self) -> float: """Formula for computing the current backoff :rtype: float @@ -366,32 +297,28 @@ class Retry(object): return 0 backoff_value = self.backoff_factor * (2 ** (consecutive_errors_len - 1)) - return min(self.DEFAULT_BACKOFF_MAX, backoff_value) + if self.backoff_jitter != 0.0: + backoff_value += random.random() * self.backoff_jitter + return float(max(0, min(self.backoff_max, backoff_value))) - def parse_retry_after(self, retry_after): + def parse_retry_after(self, retry_after: str) -> float: + seconds: float # 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_tz(retry_after) if retry_date_tuple is None: - raise InvalidHeader("Invalid Retry-After header: %s" % retry_after) - 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:] + raise InvalidHeader(f"Invalid Retry-After header: {retry_after}") retry_date = email.utils.mktime_tz(retry_date_tuple) seconds = retry_date - time.time() - if seconds < 0: - seconds = 0 + seconds = max(seconds, 0) return seconds - def get_retry_after(self, response): + def get_retry_after(self, response: BaseHTTPResponse) -> float | None: """Get the value of Retry-After in seconds.""" retry_after = response.headers.get("Retry-After") @@ -401,7 +328,7 @@ class Retry(object): return self.parse_retry_after(retry_after) - def sleep_for_retry(self, response=None): + def sleep_for_retry(self, response: BaseHTTPResponse) -> bool: retry_after = self.get_retry_after(response) if retry_after: time.sleep(retry_after) @@ -409,13 +336,13 @@ class Retry(object): return False - def _sleep_backoff(self): + def _sleep_backoff(self) -> None: backoff = self.get_backoff_time() if backoff <= 0: return time.sleep(backoff) - def sleep(self, response=None): + def sleep(self, response: BaseHTTPResponse | None = None) -> None: """Sleep between retry attempts. This method will respect a server's ``Retry-After`` response header @@ -431,7 +358,7 @@ class Retry(object): self._sleep_backoff() - def _is_connection_error(self, err): + def _is_connection_error(self, err: Exception) -> bool: """Errors when we're fairly sure that the server did not receive the request, so it should be safe to retry. """ @@ -439,33 +366,23 @@ class Retry(object): err = err.original_error return isinstance(err, ConnectTimeoutError) - def _is_read_error(self, err): + def _is_read_error(self, err: Exception) -> bool: """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): + def _is_method_retryable(self, method: str) -> bool: """Checks if a given HTTP method should be retried upon, depending if it is included in the allowed_methods """ - # 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: + if self.allowed_methods and method.upper() not in self.allowed_methods: return False return True - def is_retry(self, method, status_code, has_retry_after=False): + def is_retry( + self, method: str, status_code: int, has_retry_after: bool = False + ) -> bool: """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 @@ -478,24 +395,27 @@ class Retry(object): if self.status_forcelist and status_code in self.status_forcelist: return True - return ( + return bool( self.total and self.respect_retry_after_header and has_retry_after and (status_code in self.RETRY_AFTER_STATUS_CODES) ) - def is_exhausted(self): + def is_exhausted(self) -> bool: """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)) + retry_counts = [ + x + for x in ( + self.total, + self.connect, + self.read, + self.redirect, + self.status, + self.other, + ) + if x + ] if not retry_counts: return False @@ -503,18 +423,18 @@ class Retry(object): def increment( self, - method=None, - url=None, - response=None, - error=None, - _pool=None, - _stacktrace=None, - ): + method: str | None = None, + url: str | None = None, + response: BaseHTTPResponse | None = None, + error: Exception | None = None, + _pool: ConnectionPool | None = None, + _stacktrace: TracebackType | None = None, + ) -> Retry: """Return a new Retry object with incremented retry counters. :param response: A response object, or None, if the server did not return a response. - :type response: :class:`~urllib3.response.HTTPResponse` + :type response: :class:`~urllib3.response.BaseHTTPResponse` :param Exception error: An error encountered during the request, or None if the response was received successfully. @@ -522,7 +442,7 @@ class Retry(object): """ if self.total is False and error: # Disabled, indicate to re-raise the error. - raise six.reraise(type(error), error, _stacktrace) + raise reraise(type(error), error, _stacktrace) total = self.total if total is not None: @@ -540,14 +460,14 @@ class Retry(object): if error and self._is_connection_error(error): # Connect retry? if connect is False: - raise six.reraise(type(error), error, _stacktrace) + raise reraise(type(error), error, _stacktrace) elif connect is not None: connect -= 1 elif error and self._is_read_error(error): # Read retry? - if read is False or not self._is_method_retryable(method): - raise six.reraise(type(error), error, _stacktrace) + if read is False or method is None or not self._is_method_retryable(method): + raise reraise(type(error), error, _stacktrace) elif read is not None: read -= 1 @@ -561,7 +481,9 @@ class Retry(object): if redirect is not None: redirect -= 1 cause = "too many redirects" - redirect_location = response.get_redirect_location() + response_redirect_location = response.get_redirect_location() + if response_redirect_location: + redirect_location = response_redirect_location status = response.status else: @@ -589,31 +511,18 @@ class Retry(object): ) if new_retry.is_exhausted(): - raise MaxRetryError(_pool, url, error or ResponseError(cause)) + reason = error or ResponseError(cause) + raise MaxRetryError(_pool, url, reason) from reason # type: ignore[arg-type] log.debug("Incremented Retry for (url='%s'): %r", url, new_retry) return new_retry - def __repr__(self): + def __repr__(self) -> str: return ( - "{cls.__name__}(total={self.total}, connect={self.connect}, " - "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) + f"{type(self).__name__}(total={self.total}, connect={self.connect}, " + f"read={self.read}, redirect={self.redirect}, status={self.status})" + ) # For backwards compatibility (equivalent to pre-v1.9): diff --git a/lib/urllib3/util/ssl_.py b/lib/urllib3/util/ssl_.py index 8f867812..77628032 100644 --- a/lib/urllib3/util/ssl_.py +++ b/lib/urllib3/util/ssl_.py @@ -1,185 +1,152 @@ -from __future__ import absolute_import +from __future__ import annotations import hmac import os +import socket import sys +import typing import warnings -from binascii import hexlify, unhexlify +from binascii import unhexlify from hashlib import md5, sha1, sha256 -from ..exceptions import ( - InsecurePlatformWarning, - ProxySchemeUnsupported, - SNIMissingWarning, - SSLError, -) -from ..packages import six -from .url import BRACELESS_IPV6_ADDRZ_RE, IPV4_RE +from ..exceptions import ProxySchemeUnsupported, SSLError +from .url import _BRACELESS_IPV6_ADDRZ_RE, _IPV4_RE SSLContext = None SSLTransport = None -HAS_SNI = False +HAS_NEVER_CHECK_COMMON_NAME = False IS_PYOPENSSL = False IS_SECURETRANSPORT = False ALPN_PROTOCOLS = ["http/1.1"] +_TYPE_VERSION_INFO = typing.Tuple[int, int, int, str, int] + # Maps the length of a digest to a possible hash function producing this digest HASHFUNC_MAP = {32: md5, 40: sha1, 64: sha256} -def _const_compare_digest_backport(a, b): +def _is_bpo_43522_fixed( + implementation_name: str, + version_info: _TYPE_VERSION_INFO, + pypy_version_info: _TYPE_VERSION_INFO | None, +) -> bool: + """Return True for CPython 3.8.9+, 3.9.3+ or 3.10+ and PyPy 7.3.8+ where + setting SSLContext.hostname_checks_common_name to False works. + + Outside of CPython and PyPy we don't know which implementations work + or not so we conservatively use our hostname matching as we know that works + on all implementations. + + https://github.com/urllib3/urllib3/issues/2192#issuecomment-821832963 + https://foss.heptapod.net/pypy/pypy/-/issues/3539 """ - Compare two digests of equal length in constant time. - - The digests must be of type str/bytes. - Returns True if the digests match, and False otherwise. - """ - result = abs(len(a) - len(b)) - for left, right in zip(bytearray(a), bytearray(b)): - result |= left ^ right - return result == 0 + if implementation_name == "pypy": + # https://foss.heptapod.net/pypy/pypy/-/issues/3129 + return pypy_version_info >= (7, 3, 8) and version_info >= (3, 8) # type: ignore[operator] + elif implementation_name == "cpython": + major_minor = version_info[:2] + micro = version_info[2] + return ( + (major_minor == (3, 8) and micro >= 9) + or (major_minor == (3, 9) and micro >= 3) + or major_minor >= (3, 10) + ) + else: # Defensive: + return False -_const_compare_digest = getattr(hmac, "compare_digest", _const_compare_digest_backport) +def _is_has_never_check_common_name_reliable( + openssl_version: str, + openssl_version_number: int, + implementation_name: str, + version_info: _TYPE_VERSION_INFO, + pypy_version_info: _TYPE_VERSION_INFO | None, +) -> bool: + # As of May 2023, all released versions of LibreSSL fail to reject certificates with + # only common names, see https://github.com/urllib3/urllib3/pull/3024 + is_openssl = openssl_version.startswith("OpenSSL ") + # Before fixing OpenSSL issue #14579, the SSL_new() API was not copying hostflags + # like X509_CHECK_FLAG_NEVER_CHECK_SUBJECT, which tripped up CPython. + # https://github.com/openssl/openssl/issues/14579 + # This was released in OpenSSL 1.1.1l+ (>=0x101010cf) + is_openssl_issue_14579_fixed = openssl_version_number >= 0x101010CF -try: # Test for SSL features + return is_openssl and ( + is_openssl_issue_14579_fixed + or _is_bpo_43522_fixed(implementation_name, version_info, pypy_version_info) + ) + + +if typing.TYPE_CHECKING: + from ssl import VerifyMode + + from typing_extensions import Literal, TypedDict + + from .ssltransport import SSLTransport as SSLTransportType + + class _TYPE_PEER_CERT_RET_DICT(TypedDict, total=False): + subjectAltName: tuple[tuple[str, str], ...] + subject: tuple[tuple[tuple[str, str], ...], ...] + serialNumber: str + + +# Mapping from 'ssl.PROTOCOL_TLSX' to 'TLSVersion.X' +_SSL_VERSION_TO_TLS_VERSION: dict[int, int] = {} + +try: # Do we have ssl at all? import ssl - 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 + from ssl import ( # type: ignore[assignment] + CERT_REQUIRED, + HAS_NEVER_CHECK_COMMON_NAME, + OP_NO_COMPRESSION, + OP_NO_TICKET, + OPENSSL_VERSION, + OPENSSL_VERSION_NUMBER, + PROTOCOL_TLS, + PROTOCOL_TLS_CLIENT, + OP_NO_SSLv2, + OP_NO_SSLv3, + SSLContext, + TLSVersion, + ) PROTOCOL_SSLv23 = PROTOCOL_TLS -except ImportError: - try: - from ssl import PROTOCOL_SSLv23 as PROTOCOL_TLS - PROTOCOL_SSLv23 = PROTOCOL_TLS - except ImportError: - PROTOCOL_SSLv23 = PROTOCOL_TLS = 2 + # Setting SSLContext.hostname_checks_common_name = False didn't work before CPython + # 3.8.9, 3.9.3, and 3.10 (but OK on PyPy) or OpenSSL 1.1.1l+ + if HAS_NEVER_CHECK_COMMON_NAME and not _is_has_never_check_common_name_reliable( + OPENSSL_VERSION, + OPENSSL_VERSION_NUMBER, + sys.implementation.name, + sys.version_info, + sys.pypy_version_info if sys.implementation.name == "pypy" else None, # type: ignore[attr-defined] + ): + HAS_NEVER_CHECK_COMMON_NAME = False -try: - from ssl import PROTOCOL_TLS_CLIENT -except ImportError: - PROTOCOL_TLS_CLIENT = PROTOCOL_TLS - - -try: - 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: -# -# - https://wiki.mozilla.org/Security/Server_Side_TLS -# - https://www.ssllabs.com/projects/best-practices/index.html -# - https://hynek.me/articles/hardening-your-web-servers-ssl-ciphers/ -# -# The general intent is: -# - prefer cipher suites that offer perfect forward secrecy (DHE/ECDHE), -# - prefer ECDHE over DHE for better performance, -# - prefer any AES-GCM and ChaCha20 over any AES-CBC for better performance and -# security, -# - prefer AES-GCM over ChaCha20 because hardware-accelerated AES is common, -# - disable NULL authentication, MD5 MACs, DSS, and other -# insecure ciphers for security reasons. -# - NOTE: TLS 1.3 cipher suites are managed through a different interface -# not exposed by CPython (yet!) and are enabled by default if they're available. -DEFAULT_CIPHERS = ":".join( - [ - "ECDHE+AESGCM", - "ECDHE+CHACHA20", - "DHE+AESGCM", - "DHE+CHACHA20", - "ECDH+AESGCM", - "DH+AESGCM", - "ECDH+AES", - "DH+AES", - "RSA+AESGCM", - "RSA+AES", - "!aNULL", - "!eNULL", - "!MD5", - "!DSS", - ] -) - -try: - from ssl import SSLContext # Modern SSL? -except ImportError: - - class SSLContext(object): # Platform-specific: Python 2 - def __init__(self, protocol_version): - self.protocol = protocol_version - # Use default values from a real SSLContext - self.check_hostname = False - self.verify_mode = ssl.CERT_NONE - self.ca_certs = None - self.options = 0 - self.certfile = None - self.keyfile = None - self.ciphers = None - - def load_cert_chain(self, certfile, keyfile): - self.certfile = certfile - self.keyfile = keyfile - - 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 - - def wrap_socket(self, socket, server_hostname=None, server_side=False): - warnings.warn( - "A true SSLContext object is not available. This prevents " - "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/1.26.x/advanced-usage.html" - "#ssl-warnings", - InsecurePlatformWarning, + # Need to be careful here in case old TLS versions get + # removed in future 'ssl' module implementations. + for attr in ("TLSv1", "TLSv1_1", "TLSv1_2"): + try: + _SSL_VERSION_TO_TLS_VERSION[getattr(ssl, f"PROTOCOL_{attr}")] = getattr( + TLSVersion, attr ) - kwargs = { - "keyfile": self.keyfile, - "certfile": self.certfile, - "ca_certs": self.ca_certs, - "cert_reqs": self.verify_mode, - "ssl_version": self.protocol, - "server_side": server_side, - } - return wrap_socket(socket, ciphers=self.ciphers, **kwargs) + except AttributeError: # Defensive: + continue + + from .ssltransport import SSLTransport # type: ignore[assignment] +except ImportError: + OP_NO_COMPRESSION = 0x20000 # type: ignore[assignment] + OP_NO_TICKET = 0x4000 # type: ignore[assignment] + OP_NO_SSLv2 = 0x1000000 # type: ignore[assignment] + OP_NO_SSLv3 = 0x2000000 # type: ignore[assignment] + PROTOCOL_SSLv23 = PROTOCOL_TLS = 2 # type: ignore[assignment] + PROTOCOL_TLS_CLIENT = 16 # type: ignore[assignment] -def assert_fingerprint(cert, fingerprint): +_TYPE_PEER_CERT_RET = typing.Union["_TYPE_PEER_CERT_RET_DICT", bytes, None] + + +def assert_fingerprint(cert: bytes | None, fingerprint: str) -> None: """ Checks if given fingerprint matches the supplied certificate. @@ -189,26 +156,27 @@ def assert_fingerprint(cert, fingerprint): Fingerprint as string of hexdigits, can be interspersed by colons. """ + if cert is None: + raise SSLError("No certificate for the peer.") + fingerprint = fingerprint.replace(":", "").lower() digest_length = len(fingerprint) hashfunc = HASHFUNC_MAP.get(digest_length) if not hashfunc: - raise SSLError("Fingerprint of invalid length: {0}".format(fingerprint)) + raise SSLError(f"Fingerprint of invalid length: {fingerprint}") # We need encode() here for py32; works on py2 and p33. fingerprint_bytes = unhexlify(fingerprint.encode()) cert_digest = hashfunc(cert).digest() - if not _const_compare_digest(cert_digest, fingerprint_bytes): + if not hmac.compare_digest(cert_digest, fingerprint_bytes): raise SSLError( - 'Fingerprints did not match. Expected "{0}", got "{1}".'.format( - fingerprint, hexlify(cert_digest) - ) + f'Fingerprints did not match. Expected "{fingerprint}", got "{cert_digest.hex()}"' ) -def resolve_cert_reqs(candidate): +def resolve_cert_reqs(candidate: None | int | str) -> VerifyMode: """ Resolves the argument to a numeric constant, which can be passed to the wrap_socket function/method from the ssl module. @@ -226,12 +194,12 @@ def resolve_cert_reqs(candidate): res = getattr(ssl, candidate, None) if res is None: res = getattr(ssl, "CERT_" + candidate) - return res + return res # type: ignore[no-any-return] - return candidate + return candidate # type: ignore[return-value] -def resolve_ssl_version(candidate): +def resolve_ssl_version(candidate: None | int | str) -> int: """ like resolve_cert_reqs """ @@ -242,35 +210,33 @@ def resolve_ssl_version(candidate): res = getattr(ssl, candidate, None) if res is None: res = getattr(ssl, "PROTOCOL_" + candidate) - return res + return typing.cast(int, res) return candidate def create_urllib3_context( - ssl_version=None, cert_reqs=None, options=None, ciphers=None -): - """All arguments have the same meaning as ``ssl_wrap_socket``. - - By default, this function does a lot of the same work that - ``ssl.create_default_context`` does on Python 3.4+. It: - - - Disables SSLv2, SSLv3, and compression - - Sets a restricted set of server ciphers - - If you wish to enable SSLv3, you can do:: - - from urllib3.util import ssl_ - context = ssl_.create_urllib3_context() - context.options &= ~ssl_.OP_NO_SSLv3 - - You can do the same to enable compression (substituting ``COMPRESSION`` - for ``SSLv3`` in the last line above). + ssl_version: int | None = None, + cert_reqs: int | None = None, + options: int | None = None, + ciphers: str | None = None, + ssl_minimum_version: int | None = None, + ssl_maximum_version: int | None = None, +) -> ssl.SSLContext: + """Creates and configures an :class:`ssl.SSLContext` instance for use with urllib3. :param ssl_version: The desired protocol version to use. This will default to PROTOCOL_SSLv23 which will negotiate the highest protocol that both the server and your installation of OpenSSL support. + + This parameter is deprecated instead use 'ssl_minimum_version'. + :param ssl_minimum_version: + The minimum version of TLS to be used. Use the 'ssl.TLSVersion' enum for specifying the value. + :param ssl_maximum_version: + The maximum version of TLS to be used. Use the 'ssl.TLSVersion' enum for specifying the value. + Not recommended to set to anything other than 'ssl.TLSVersion.MAXIMUM_SUPPORTED' which is the + default value. :param cert_reqs: Whether to require the certificate verification. This defaults to ``ssl.CERT_REQUIRED``. @@ -278,18 +244,60 @@ def create_urllib3_context( Specific OpenSSL options. These default to ``ssl.OP_NO_SSLv2``, ``ssl.OP_NO_SSLv3``, ``ssl.OP_NO_COMPRESSION``, and ``ssl.OP_NO_TICKET``. :param ciphers: - Which cipher suites to allow the server to select. + Which cipher suites to allow the server to select. Defaults to either system configured + ciphers if OpenSSL 1.1.1+, otherwise uses a secure default set of ciphers. :returns: Constructed SSLContext object with specified options :rtype: SSLContext """ - # PROTOCOL_TLS is deprecated in Python 3.10 - if not ssl_version or ssl_version == PROTOCOL_TLS: - ssl_version = PROTOCOL_TLS_CLIENT + if SSLContext is None: + raise TypeError("Can't create an SSLContext object without an ssl module") - context = SSLContext(ssl_version) + # This means 'ssl_version' was specified as an exact value. + if ssl_version not in (None, PROTOCOL_TLS, PROTOCOL_TLS_CLIENT): + # Disallow setting 'ssl_version' and 'ssl_minimum|maximum_version' + # to avoid conflicts. + if ssl_minimum_version is not None or ssl_maximum_version is not None: + raise ValueError( + "Can't specify both 'ssl_version' and either " + "'ssl_minimum_version' or 'ssl_maximum_version'" + ) - context.set_ciphers(ciphers or DEFAULT_CIPHERS) + # 'ssl_version' is deprecated and will be removed in the future. + else: + # Use 'ssl_minimum_version' and 'ssl_maximum_version' instead. + ssl_minimum_version = _SSL_VERSION_TO_TLS_VERSION.get( + ssl_version, TLSVersion.MINIMUM_SUPPORTED + ) + ssl_maximum_version = _SSL_VERSION_TO_TLS_VERSION.get( + ssl_version, TLSVersion.MAXIMUM_SUPPORTED + ) + + # This warning message is pushing users to use 'ssl_minimum_version' + # instead of both min/max. Best practice is to only set the minimum version and + # keep the maximum version to be it's default value: 'TLSVersion.MAXIMUM_SUPPORTED' + warnings.warn( + "'ssl_version' option is deprecated and will be " + "removed in urllib3 v2.1.0. Instead use 'ssl_minimum_version'", + category=DeprecationWarning, + stacklevel=2, + ) + + # PROTOCOL_TLS is deprecated in Python 3.10 so we always use PROTOCOL_TLS_CLIENT + context = SSLContext(PROTOCOL_TLS_CLIENT) + + if ssl_minimum_version is not None: + context.minimum_version = ssl_minimum_version + else: # Python <3.10 defaults to 'MINIMUM_SUPPORTED' so explicitly set TLSv1.2 here + context.minimum_version = TLSVersion.TLSv1_2 + + if ssl_maximum_version is not None: + context.maximum_version = ssl_maximum_version + + # Unless we're given ciphers defer to either system ciphers in + # the case of OpenSSL 1.1.1+ or use our own secure default ciphers. + if ciphers: + context.set_ciphers(ciphers) # Setting the default here, as we may have no ssl module on import cert_reqs = ssl.CERT_REQUIRED if cert_reqs is None else cert_reqs @@ -322,26 +330,23 @@ def create_urllib3_context( ) is not None: context.post_handshake_auth = True - 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: + # check_hostname=True, verify_mode=NONE/OPTIONAL. + # We always set 'check_hostname=False' for pyOpenSSL so we rely on our own + # 'ssl.match_hostname()' implementation. + if cert_reqs == ssl.CERT_REQUIRED and not IS_PYOPENSSL: context.verify_mode = cert_reqs - disable_check_hostname() + context.check_hostname = True else: - disable_check_hostname() + context.check_hostname = False context.verify_mode = cert_reqs + try: + context.hostname_checks_common_name = False + except AttributeError: # Defensive: for CPython < 3.8.9 and 3.9.3; for PyPy < 7.3.8 + pass + # 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"): @@ -352,21 +357,59 @@ def create_urllib3_context( return context +@typing.overload def ssl_wrap_socket( - sock, - keyfile=None, - certfile=None, - cert_reqs=None, - ca_certs=None, - server_hostname=None, - ssl_version=None, - ciphers=None, - ssl_context=None, - ca_cert_dir=None, - key_password=None, - ca_cert_data=None, - tls_in_tls=False, -): + sock: socket.socket, + keyfile: str | None = ..., + certfile: str | None = ..., + cert_reqs: int | None = ..., + ca_certs: str | None = ..., + server_hostname: str | None = ..., + ssl_version: int | None = ..., + ciphers: str | None = ..., + ssl_context: ssl.SSLContext | None = ..., + ca_cert_dir: str | None = ..., + key_password: str | None = ..., + ca_cert_data: None | str | bytes = ..., + tls_in_tls: Literal[False] = ..., +) -> ssl.SSLSocket: + ... + + +@typing.overload +def ssl_wrap_socket( + sock: socket.socket, + keyfile: str | None = ..., + certfile: str | None = ..., + cert_reqs: int | None = ..., + ca_certs: str | None = ..., + server_hostname: str | None = ..., + ssl_version: int | None = ..., + ciphers: str | None = ..., + ssl_context: ssl.SSLContext | None = ..., + ca_cert_dir: str | None = ..., + key_password: str | None = ..., + ca_cert_data: None | str | bytes = ..., + tls_in_tls: bool = ..., +) -> ssl.SSLSocket | SSLTransportType: + ... + + +def ssl_wrap_socket( + sock: socket.socket, + keyfile: str | None = None, + certfile: str | None = None, + cert_reqs: int | None = None, + ca_certs: str | None = None, + server_hostname: str | None = None, + ssl_version: int | None = None, + ciphers: str | None = None, + ssl_context: ssl.SSLContext | None = None, + ca_cert_dir: str | None = None, + key_password: str | None = None, + ca_cert_data: None | str | bytes = None, + tls_in_tls: bool = False, +) -> ssl.SSLSocket | SSLTransportType: """ All arguments except for server_hostname, ssl_context, and ca_cert_dir have the same meaning as they do when using :func:`ssl.wrap_socket`. @@ -392,19 +435,18 @@ def ssl_wrap_socket( """ context = ssl_context if context is None: - # Note: This branch of code and all the variables in it are no longer - # used by urllib3 itself. We should consider deprecating and removing - # this code. + # Note: This branch of code and all the variables in it are only used in tests. + # We should consider deprecating and removing this code. context = create_urllib3_context(ssl_version, cert_reqs, ciphers=ciphers) if ca_certs or ca_cert_dir or ca_cert_data: try: context.load_verify_locations(ca_certs, ca_cert_dir, ca_cert_data) - except (IOError, OSError) as e: - raise SSLError(e) + except OSError as e: + raise SSLError(e) from e elif ssl_context is None and hasattr(context, "load_default_certs"): - # try to load OS default certs; works well on Windows (require Python3.4+) + # try to load OS default certs; works well on Windows. context.load_default_certs() # Attempt to detect if we get the goofy behavior of the @@ -420,56 +462,30 @@ def ssl_wrap_socket( context.load_cert_chain(certfile, keyfile, key_password) try: - if hasattr(context, "set_alpn_protocols"): - context.set_alpn_protocols(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 - 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/1.26.x/advanced-usage.html" - "#ssl-warnings", - SNIMissingWarning, - ) - - 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) + ssl_sock = _ssl_wrap_socket_impl(sock, context, tls_in_tls, server_hostname) return ssl_sock -def is_ipaddress(hostname): +def is_ipaddress(hostname: str | bytes) -> bool: """Detects whether the hostname given is an IPv4 or IPv6 address. Also detects IPv6 addresses with Zone IDs. :param str hostname: Hostname to examine. :return: True if the hostname is an IP address, False otherwise. """ - if not six.PY2 and isinstance(hostname, bytes): + if isinstance(hostname, bytes): # IDN A-label bytes are ASCII compatible. hostname = hostname.decode("ascii") - return bool(IPV4_RE.match(hostname) or BRACELESS_IPV6_ADDRZ_RE.match(hostname)) + return bool(_IPV4_RE.match(hostname) or _BRACELESS_IPV6_ADDRZ_RE.match(hostname)) -def _is_key_file_encrypted(key_file): +def _is_key_file_encrypted(key_file: str) -> bool: """Detects if a key file is encrypted or not.""" - with open(key_file, "r") as f: + with open(key_file) as f: for line in f: # Look for Proc-Type: 4,ENCRYPTED if "ENCRYPTED" in line: @@ -478,7 +494,12 @@ def _is_key_file_encrypted(key_file): return False -def _ssl_wrap_socket_impl(sock, ssl_context, tls_in_tls, server_hostname=None): +def _ssl_wrap_socket_impl( + sock: socket.socket, + ssl_context: ssl.SSLContext, + tls_in_tls: bool, + server_hostname: str | None = None, +) -> ssl.SSLSocket | SSLTransportType: if tls_in_tls: if not SSLTransport: # Import error, ssl is not available. @@ -489,7 +510,4 @@ def _ssl_wrap_socket_impl(sock, ssl_context, tls_in_tls, server_hostname=None): 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) + return ssl_context.wrap_socket(sock, server_hostname=server_hostname) diff --git a/lib/urllib3/util/ssl_match_hostname.py b/lib/urllib3/util/ssl_match_hostname.py index 1dd950c4..453cfd42 100644 --- a/lib/urllib3/util/ssl_match_hostname.py +++ b/lib/urllib3/util/ssl_match_hostname.py @@ -1,19 +1,18 @@ -"""The match_hostname() function from Python 3.3.3, essential when using SSL.""" +"""The match_hostname() function from Python 3.5, essential when using SSL.""" # Note: This file is under the PSF license as the code comes from the python # stdlib. http://docs.python.org/3/license.html +# It is modified to remove commonName support. +from __future__ import annotations + +import ipaddress import re -import sys +import typing +from ipaddress import IPv4Address, IPv6Address -# 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 -# util.ssl_match_hostname to continue to be used in Python 2.7. -try: - import ipaddress -except ImportError: - ipaddress = None +if typing.TYPE_CHECKING: + from .ssl_ import _TYPE_PEER_CERT_RET_DICT __version__ = "3.5.0.1" @@ -22,7 +21,9 @@ class CertificateError(ValueError): pass -def _dnsname_match(dn, hostname, max_wildcards=1): +def _dnsname_match( + dn: typing.Any, hostname: str, max_wildcards: int = 1 +) -> typing.Match[str] | None | bool: """Matching according to RFC 6125, section 6.4.3 http://tools.ietf.org/html/rfc6125#section-6.4.3 @@ -49,7 +50,7 @@ def _dnsname_match(dn, hostname, max_wildcards=1): # speed up common case w/o wildcards if not wildcards: - return dn.lower() == hostname.lower() + return bool(dn.lower() == hostname.lower()) # RFC 6125, section 6.4.3, subitem 1. # The client SHOULD NOT attempt to match a presented identifier in which @@ -76,26 +77,26 @@ def _dnsname_match(dn, hostname, max_wildcards=1): return pat.match(hostname) -def _to_unicode(obj): - if isinstance(obj, str) and sys.version_info < (3,): - # ignored flake8 # F821 to support python 2.7 function - obj = unicode(obj, encoding="ascii", errors="strict") # noqa: F821 - return obj - - -def _ipaddress_match(ipname, host_ip): +def _ipaddress_match(ipname: str, host_ip: IPv4Address | IPv6Address) -> bool: """Exact matching of IP addresses. - RFC 6125 explicitly doesn't define an algorithm for this - (section 1.7.2 - "Out of Scope"). + RFC 9110 section 4.3.5: "A reference identity of IP-ID contains the decoded + bytes of the IP address. An IP version 4 address is 4 octets, and an IP + version 6 address is 16 octets. [...] A reference identity of type IP-ID + matches if the address is identical to an iPAddress value of the + subjectAltName extension of the certificate." """ # OpenSSL may add a trailing newline to a subjectAltName's IP address # Divergence from upstream: ipaddress can't handle byte str - ip = ipaddress.ip_address(_to_unicode(ipname).rstrip()) - return ip == host_ip + ip = ipaddress.ip_address(ipname.rstrip()) + return bool(ip.packed == host_ip.packed) -def match_hostname(cert, hostname): +def match_hostname( + cert: _TYPE_PEER_CERT_RET_DICT | None, + hostname: str, + hostname_checks_common_name: bool = False, +) -> None: """Verify that *cert* (in decoded format as returned by SSLSocket.getpeercert()) matches the *hostname*. RFC 2818 and RFC 6125 rules are followed, but IP addresses are not accepted for *hostname*. @@ -111,21 +112,22 @@ def match_hostname(cert, hostname): ) try: # Divergence from upstream: ipaddress can't handle byte str - host_ip = ipaddress.ip_address(_to_unicode(hostname)) - 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 + # + # The ipaddress module shipped with Python < 3.9 does not support + # scoped IPv6 addresses so we unconditionally strip the Zone IDs for + # now. Once we drop support for Python 3.9 we can remove this branch. + if "%" in hostname: + host_ip = ipaddress.ip_address(hostname[: hostname.rfind("%")]) + else: + host_ip = ipaddress.ip_address(hostname) + + except ValueError: + # Not an IP address (common case) host_ip = None - except AttributeError: - # Divergence from upstream: Make ipaddress library optional - if ipaddress is None: - host_ip = None - else: # Defensive - raise dnsnames = [] - san = cert.get("subjectAltName", ()) + san: tuple[tuple[str, str], ...] = cert.get("subjectAltName", ()) + key: str + value: str for key, value in san: if key == "DNS": if host_ip is None and _dnsname_match(value, hostname): @@ -135,25 +137,23 @@ def match_hostname(cert, hostname): if host_ip is not None and _ipaddress_match(value, host_ip): return dnsnames.append(value) - if not dnsnames: - # The subject is only checked when there is no dNSName entry - # in subjectAltName + + # We only check 'commonName' if it's enabled and we're not verifying + # an IP address. IP addresses aren't valid within 'commonName'. + if hostname_checks_common_name and host_ip is None and not dnsnames: for sub in cert.get("subject", ()): for key, value in sub: - # XXX according to RFC 2818, the most specific Common Name - # must be used. if key == "commonName": if _dnsname_match(value, hostname): return dnsnames.append(value) + if len(dnsnames) > 1: raise CertificateError( "hostname %r " "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(f"hostname {hostname!r} doesn't match {dnsnames[0]!r}") else: - raise CertificateError( - "no appropriate commonName or subjectAltName fields were found" - ) + raise CertificateError("no appropriate subjectAltName fields were found") diff --git a/lib/urllib3/util/ssltransport.py b/lib/urllib3/util/ssltransport.py index 4a7105d1..5ec86473 100644 --- a/lib/urllib3/util/ssltransport.py +++ b/lib/urllib3/util/ssltransport.py @@ -1,9 +1,21 @@ +from __future__ import annotations + import io import socket import ssl +import typing from ..exceptions import ProxySchemeUnsupported -from ..packages import six + +if typing.TYPE_CHECKING: + from typing_extensions import Literal + + from .ssl_ import _TYPE_PEER_CERT_RET, _TYPE_PEER_CERT_RET_DICT + + +_SelfT = typing.TypeVar("_SelfT", bound="SSLTransport") +_WriteBuffer = typing.Union[bytearray, memoryview] +_ReturnValue = typing.TypeVar("_ReturnValue") SSL_BLOCKSIZE = 16384 @@ -20,7 +32,7 @@ class SSLTransport: """ @staticmethod - def _validate_ssl_context_for_tls_in_tls(ssl_context): + def _validate_ssl_context_for_tls_in_tls(ssl_context: ssl.SSLContext) -> None: """ Raises a ProxySchemeUnsupported if the provided ssl_context can't be used for TLS in TLS. @@ -30,20 +42,18 @@ class SSLTransport: """ 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" - ) + 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 - ): + self, + socket: socket.socket, + ssl_context: ssl.SSLContext, + server_hostname: str | None = None, + suppress_ragged_eofs: bool = True, + ) -> None: """ Create an SSLTransport around socket using the provided ssl_context. """ @@ -60,33 +70,36 @@ class SSLTransport: # Perform initial handshake. self._ssl_io_loop(self.sslobj.do_handshake) - def __enter__(self): + def __enter__(self: _SelfT) -> _SelfT: return self - def __exit__(self, *_): + def __exit__(self, *_: typing.Any) -> None: self.close() - def fileno(self): + def fileno(self) -> int: return self.socket.fileno() - def read(self, len=1024, buffer=None): + def read(self, len: int = 1024, buffer: typing.Any | None = None) -> int | bytes: return self._wrap_ssl_read(len, buffer) - def recv(self, len=1024, flags=0): + def recv(self, buflen: int = 1024, flags: int = 0) -> int | bytes: if flags != 0: raise ValueError("non-zero flags not allowed in calls to recv") - return self._wrap_ssl_read(len) + return self._wrap_ssl_read(buflen) - def recv_into(self, buffer, nbytes=None, flags=0): + def recv_into( + self, + buffer: _WriteBuffer, + nbytes: int | None = None, + flags: int = 0, + ) -> None | int | bytes: if flags != 0: raise ValueError("non-zero flags not allowed in calls to recv_into") - if buffer and (nbytes is None): + if nbytes is None: nbytes = len(buffer) - elif nbytes is None: - nbytes = 1024 return self.read(nbytes, buffer) - def sendall(self, data, flags=0): + def sendall(self, data: bytes, flags: int = 0) -> None: if flags != 0: raise ValueError("non-zero flags not allowed in calls to sendall") count = 0 @@ -96,15 +109,20 @@ class SSLTransport: v = self.send(byte_view[count:]) count += v - def send(self, data, flags=0): + def send(self, data: bytes, flags: int = 0) -> int: 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 + return self._ssl_io_loop(self.sslobj.write, data) def makefile( - self, mode="r", buffering=None, encoding=None, errors=None, newline=None - ): + self, + mode: str, + buffering: int | None = None, + *, + encoding: str | None = None, + errors: str | None = None, + newline: str | None = None, + ) -> typing.BinaryIO | typing.TextIO | socket.SocketIO: """ Python's httpclient uses makefile and buffered io when reading HTTP messages and we need to support it. @@ -113,7 +131,7 @@ class SSLTransport: 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,)) + raise ValueError(f"invalid mode {mode!r} (only r, w, b allowed)") writing = "w" in mode reading = "r" in mode or not writing @@ -124,8 +142,8 @@ class SSLTransport: rawmode += "r" if writing: rawmode += "w" - raw = socket.SocketIO(self, rawmode) - self.socket._io_refs += 1 + raw = socket.SocketIO(self, rawmode) # type: ignore[arg-type] + self.socket._io_refs += 1 # type: ignore[attr-defined] if buffering is None: buffering = -1 if buffering < 0: @@ -134,8 +152,9 @@ class SSLTransport: if not binary: raise ValueError("unbuffered streams must be binary") return raw + buffer: typing.BinaryIO if reading and writing: - buffer = io.BufferedRWPair(raw, raw, buffering) + buffer = io.BufferedRWPair(raw, raw, buffering) # type: ignore[assignment] elif reading: buffer = io.BufferedReader(raw, buffering) else: @@ -144,46 +163,56 @@ class SSLTransport: if binary: return buffer text = io.TextIOWrapper(buffer, encoding, errors, newline) - text.mode = mode + text.mode = mode # type: ignore[misc] return text - def unwrap(self): + def unwrap(self) -> None: self._ssl_io_loop(self.sslobj.unwrap) - def close(self): + def close(self) -> None: self.socket.close() - def getpeercert(self, binary_form=False): - return self.sslobj.getpeercert(binary_form) + @typing.overload + def getpeercert( + self, binary_form: Literal[False] = ... + ) -> _TYPE_PEER_CERT_RET_DICT | None: + ... - def version(self): + @typing.overload + def getpeercert(self, binary_form: Literal[True]) -> bytes | None: + ... + + def getpeercert(self, binary_form: bool = False) -> _TYPE_PEER_CERT_RET: + return self.sslobj.getpeercert(binary_form) # type: ignore[return-value] + + def version(self) -> str | None: return self.sslobj.version() - def cipher(self): + def cipher(self) -> tuple[str, str, int] | None: return self.sslobj.cipher() - def selected_alpn_protocol(self): + def selected_alpn_protocol(self) -> str | None: return self.sslobj.selected_alpn_protocol() - def selected_npn_protocol(self): + def selected_npn_protocol(self) -> str | None: return self.sslobj.selected_npn_protocol() - def shared_ciphers(self): + def shared_ciphers(self) -> list[tuple[str, str, int]] | None: return self.sslobj.shared_ciphers() - def compression(self): + def compression(self) -> str | None: return self.sslobj.compression() - def settimeout(self, value): + def settimeout(self, value: float | None) -> None: self.socket.settimeout(value) - def gettimeout(self): + def gettimeout(self) -> float | None: return self.socket.gettimeout() - def _decref_socketios(self): - self.socket._decref_socketios() + def _decref_socketios(self) -> None: + self.socket._decref_socketios() # type: ignore[attr-defined] - def _wrap_ssl_read(self, len, buffer=None): + def _wrap_ssl_read(self, len: int, buffer: bytearray | None = None) -> int | bytes: try: return self._ssl_io_loop(self.sslobj.read, len, buffer) except ssl.SSLError as e: @@ -192,7 +221,32 @@ class SSLTransport: else: raise - def _ssl_io_loop(self, func, *args): + # func is sslobj.do_handshake or sslobj.unwrap + @typing.overload + def _ssl_io_loop(self, func: typing.Callable[[], None]) -> None: + ... + + # func is sslobj.write, arg1 is data + @typing.overload + def _ssl_io_loop(self, func: typing.Callable[[bytes], int], arg1: bytes) -> int: + ... + + # func is sslobj.read, arg1 is len, arg2 is buffer + @typing.overload + def _ssl_io_loop( + self, + func: typing.Callable[[int, bytearray | None], bytes], + arg1: int, + arg2: bytearray | None, + ) -> bytes: + ... + + def _ssl_io_loop( + self, + func: typing.Callable[..., _ReturnValue], + arg1: None | bytes | int = None, + arg2: bytearray | None = None, + ) -> _ReturnValue: """Performs an I/O loop between incoming/outgoing and the socket.""" should_loop = True ret = None @@ -200,7 +254,12 @@ class SSLTransport: while should_loop: errno = None try: - ret = func(*args) + if arg1 is None and arg2 is None: + ret = func() + elif arg2 is None: + ret = func(arg1) + else: + ret = func(arg1, arg2) 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. @@ -218,4 +277,4 @@ class SSLTransport: self.incoming.write(buf) else: self.incoming.write_eof() - return ret + return typing.cast(_ReturnValue, ret) diff --git a/lib/urllib3/util/timeout.py b/lib/urllib3/util/timeout.py index 78e18a62..ec090f69 100644 --- a/lib/urllib3/util/timeout.py +++ b/lib/urllib3/util/timeout.py @@ -1,44 +1,56 @@ -from __future__ import absolute_import +from __future__ import annotations 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 +import typing +from enum import Enum +from socket import getdefaulttimeout from ..exceptions import TimeoutStateError -# A sentinel value to indicate that no timeout was specified by the user in -# urllib3 -_Default = object() +if typing.TYPE_CHECKING: + from typing_extensions import Final -# Use time.monotonic if available. -current_time = getattr(time, "monotonic", time.time) +class _TYPE_DEFAULT(Enum): + # This value should never be passed to socket.settimeout() so for safety we use a -1. + # socket.settimout() raises a ValueError for negative values. + token = -1 -class Timeout(object): +_DEFAULT_TIMEOUT: Final[_TYPE_DEFAULT] = _TYPE_DEFAULT.token + +_TYPE_TIMEOUT = typing.Optional[typing.Union[float, _TYPE_DEFAULT]] + + +class Timeout: """Timeout configuration. Timeouts can be defined as a default for a pool: .. code-block:: python - timeout = Timeout(connect=2.0, read=7.0) - http = PoolManager(timeout=timeout) - response = http.request('GET', 'http://example.com/') + import urllib3 + + timeout = urllib3.util.Timeout(connect=2.0, read=7.0) + + http = urllib3.PoolManager(timeout=timeout) + + resp = http.request("GET", "https://example.com/") + + print(resp.status) Or per-request (which overrides the default for the pool): .. code-block:: python - response = http.request('GET', 'http://example.com/', timeout=Timeout(10)) + response = http.request("GET", "https://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) + response = http.request("GET", "https://example.com/", timeout=no_timeout) :param total: @@ -96,31 +108,31 @@ class Timeout(object): """ #: A sentinel object representing the default timeout value - DEFAULT_TIMEOUT = _GLOBAL_DEFAULT_TIMEOUT + DEFAULT_TIMEOUT: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT - def __init__(self, total=None, connect=_Default, read=_Default): + def __init__( + self, + total: _TYPE_TIMEOUT = None, + connect: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT, + read: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT, + ) -> None: self._connect = self._validate_timeout(connect, "connect") self._read = self._validate_timeout(read, "read") self.total = self._validate_timeout(total, "total") - self._start_connect = None + self._start_connect: float | None = None - def __repr__(self): - return "%s(connect=%r, read=%r, total=%r)" % ( - type(self).__name__, - self._connect, - self._read, - self.total, - ) + def __repr__(self) -> str: + return f"{type(self).__name__}(connect={self._connect!r}, read={self._read!r}, total={self.total!r})" # __str__ provided for backwards compatibility __str__ = __repr__ - @classmethod - def resolve_default_timeout(cls, timeout): - return getdefaulttimeout() if timeout is cls.DEFAULT_TIMEOUT else timeout + @staticmethod + def resolve_default_timeout(timeout: _TYPE_TIMEOUT) -> float | None: + return getdefaulttimeout() if timeout is _DEFAULT_TIMEOUT else timeout @classmethod - def _validate_timeout(cls, value, name): + def _validate_timeout(cls, value: _TYPE_TIMEOUT, name: str) -> _TYPE_TIMEOUT: """Check that a timeout attribute is valid. :param value: The timeout value to validate @@ -130,10 +142,7 @@ class Timeout(object): :raises ValueError: If it is a numeric value less than or equal to zero, or the type is not an integer, float, or None. """ - if value is _Default: - return cls.DEFAULT_TIMEOUT - - if value is None or value is cls.DEFAULT_TIMEOUT: + if value is None or value is _DEFAULT_TIMEOUT: return value if isinstance(value, bool): @@ -147,7 +156,7 @@ class Timeout(object): raise ValueError( "Timeout value %s was %s, but it must be an " "int, float or None." % (name, value) - ) + ) from None try: if value <= 0: @@ -157,16 +166,15 @@ class Timeout(object): "than or equal to 0." % (name, value) ) except TypeError: - # Python 3 raise ValueError( "Timeout value %s was %s, but it must be an " "int, float or None." % (name, value) - ) + ) from None return value @classmethod - def from_float(cls, timeout): + def from_float(cls, timeout: _TYPE_TIMEOUT) -> Timeout: """Create a new Timeout from a legacy timeout value. The timeout value used by httplib.py sets the same timeout on the @@ -175,13 +183,13 @@ class Timeout(object): passed to this function. :param timeout: The legacy timeout value. - :type timeout: integer, float, sentinel default object, or None + :type timeout: integer, float, :attr:`urllib3.util.Timeout.DEFAULT_TIMEOUT`, or None :return: Timeout object :rtype: :class:`Timeout` """ return Timeout(read=timeout, connect=timeout) - def clone(self): + def clone(self) -> Timeout: """Create a copy of the timeout object Timeout properties are stored per-pool but each request needs a fresh @@ -195,7 +203,7 @@ class Timeout(object): # detect the user default. return Timeout(connect=self._connect, read=self._read, total=self.total) - def start_connect(self): + def start_connect(self) -> float: """Start the timeout clock, used during a connect() attempt :raises urllib3.exceptions.TimeoutStateError: if you attempt @@ -203,10 +211,10 @@ class Timeout(object): """ if self._start_connect is not None: raise TimeoutStateError("Timeout timer has already been started.") - self._start_connect = current_time() + self._start_connect = time.monotonic() return self._start_connect - def get_connect_duration(self): + def get_connect_duration(self) -> float: """Gets the time elapsed since the call to :meth:`start_connect`. :return: Elapsed time in seconds. @@ -218,10 +226,10 @@ class Timeout(object): raise TimeoutStateError( "Can't get connect duration for timer that has not started." ) - return current_time() - self._start_connect + return time.monotonic() - self._start_connect @property - def connect_timeout(self): + def connect_timeout(self) -> _TYPE_TIMEOUT: """Get the value to use when setting a connection timeout. This will be a positive float or integer, the value None @@ -233,13 +241,13 @@ class Timeout(object): if self.total is None: return self._connect - if self._connect is None or self._connect is self.DEFAULT_TIMEOUT: + if self._connect is None or self._connect is _DEFAULT_TIMEOUT: return self.total - return min(self._connect, self.total) + return min(self._connect, self.total) # type: ignore[type-var] @property - def read_timeout(self): + def read_timeout(self) -> float | None: """Get the value for the read timeout. This assumes some time has elapsed in the connection timeout and @@ -251,21 +259,21 @@ class Timeout(object): raised. :return: Value to use for the read timeout. - :rtype: int, float, :attr:`Timeout.DEFAULT_TIMEOUT` or None + :rtype: int, float or None :raises urllib3.exceptions.TimeoutStateError: If :meth:`start_connect` has not yet been called on this object. """ if ( self.total is not None - and self.total is not self.DEFAULT_TIMEOUT + and self.total is not _DEFAULT_TIMEOUT and self._read is not None - and self._read is not self.DEFAULT_TIMEOUT + and self._read is not _DEFAULT_TIMEOUT ): # In case the connect timeout has not yet been established. if self._start_connect is None: return self._read return max(0, min(self.total - self.get_connect_duration(), self._read)) - elif self.total is not None and self.total is not self.DEFAULT_TIMEOUT: + elif self.total is not None and self.total is not _DEFAULT_TIMEOUT: return max(0, self.total - self.get_connect_duration()) else: - return self._read + return self.resolve_default_timeout(self._read) diff --git a/lib/urllib3/util/url.py b/lib/urllib3/util/url.py index e5682d3b..d53ea932 100644 --- a/lib/urllib3/util/url.py +++ b/lib/urllib3/util/url.py @@ -1,22 +1,20 @@ -from __future__ import absolute_import +from __future__ import annotations import re -from collections import namedtuple +import typing from ..exceptions import LocationParseError -from ..packages import six - -url_attrs = ["scheme", "auth", "host", "port", "path", "query", "fragment"] +from .util import to_str # We only want to normalize urls with an HTTP(S) scheme. # urllib3 infers URLs without a scheme (None) to be http. -NORMALIZABLE_SCHEMES = ("http", "https", None) +_NORMALIZABLE_SCHEMES = ("http", "https", None) # Almost all of these patterns were derived from the # 'rfc3986' module: https://github.com/python-hyper/rfc3986 -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( +_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"([^?#]*)" @@ -25,10 +23,10 @@ URI_RE = re.compile( re.UNICODE | re.DOTALL, ) -IPV4_PAT = r"(?:[0-9]{1,3}\.){3}[0-9]{1,3}" -HEX_PAT = "[0-9A-Fa-f]{1,4}" -LS32_PAT = "(?:{hex}:{hex}|{ipv4})".format(hex=HEX_PAT, ipv4=IPV4_PAT) -_subs = {"hex": HEX_PAT, "ls32": LS32_PAT} +_IPV4_PAT = r"(?:[0-9]{1,3}\.){3}[0-9]{1,3}" +_HEX_PAT = "[0-9A-Fa-f]{1,4}" +_LS32_PAT = "(?:{hex}:{hex}|{ipv4})".format(hex=_HEX_PAT, ipv4=_IPV4_PAT) +_subs = {"hex": _HEX_PAT, "ls32": _LS32_PAT} _variations = [ # 6( h16 ":" ) ls32 "(?:%(hex)s:){6}%(ls32)s", @@ -50,69 +48,78 @@ _variations = [ "(?:(?:%(hex)s:){0,6}%(hex)s)?::", ] -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"^(/[^?#]*)(?:\?([^#]*))?(?:#.*)?$") +_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"^(/[^?#]*)(?:\?([^#]*))?(?:#.*)?$") -IPV4_RE = re.compile("^" + IPV4_PAT + "$") -IPV6_RE = re.compile("^" + IPV6_PAT + "$") -IPV6_ADDRZ_RE = re.compile("^" + IPV6_ADDRZ_PAT + "$") -BRACELESS_IPV6_ADDRZ_RE = re.compile("^" + IPV6_ADDRZ_PAT[2:-2] + "$") -ZONE_ID_RE = re.compile("(" + ZONE_ID_PAT + r")\]$") +_IPV4_RE = re.compile("^" + _IPV4_PAT + "$") +_IPV6_RE = re.compile("^" + _IPV6_PAT + "$") +_IPV6_ADDRZ_RE = re.compile("^" + _IPV6_ADDRZ_PAT + "$") +_BRACELESS_IPV6_ADDRZ_RE = re.compile("^" + _IPV6_ADDRZ_PAT[2:-2] + "$") +_ZONE_ID_RE = re.compile("(" + _ZONE_ID_PAT + r")\]$") _HOST_PORT_PAT = ("^(%s|%s|%s)(?::0*?(|0|[1-9][0-9]{0,4}))?$") % ( - REG_NAME_PAT, - IPV4_PAT, - IPV6_ADDRZ_PAT, + _REG_NAME_PAT, + _IPV4_PAT, + _IPV6_ADDRZ_PAT, ) _HOST_PORT_RE = re.compile(_HOST_PORT_PAT, re.UNICODE | re.DOTALL) -UNRESERVED_CHARS = set( +_UNRESERVED_CHARS = set( "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789._-~" ) -SUB_DELIM_CHARS = set("!$&'()*+,;=") -USERINFO_CHARS = UNRESERVED_CHARS | SUB_DELIM_CHARS | {":"} -PATH_CHARS = USERINFO_CHARS | {"@", "/"} -QUERY_CHARS = FRAGMENT_CHARS = PATH_CHARS | {"?"} +_SUB_DELIM_CHARS = set("!$&'()*+,;=") +_USERINFO_CHARS = _UNRESERVED_CHARS | _SUB_DELIM_CHARS | {":"} +_PATH_CHARS = _USERINFO_CHARS | {"@", "/"} +_QUERY_CHARS = _FRAGMENT_CHARS = _PATH_CHARS | {"?"} -class Url(namedtuple("Url", url_attrs)): +class Url( + typing.NamedTuple( + "Url", + [ + ("scheme", typing.Optional[str]), + ("auth", typing.Optional[str]), + ("host", typing.Optional[str]), + ("port", typing.Optional[int]), + ("path", typing.Optional[str]), + ("query", typing.Optional[str]), + ("fragment", typing.Optional[str]), + ], + ) +): """ Data structure for representing an HTTP URL. Used as a return value for :func:`parse_url`. Both the scheme and host are normalized as they are both case-insensitive according to RFC 3986. """ - __slots__ = () - - def __new__( + def __new__( # type: ignore[no-untyped-def] cls, - scheme=None, - auth=None, - host=None, - port=None, - path=None, - query=None, - fragment=None, + scheme: str | None = None, + auth: str | None = None, + host: str | None = None, + port: int | None = None, + path: str | None = None, + query: str | None = None, + fragment: str | None = None, ): if path and not path.startswith("/"): path = "/" + path if scheme is not None: scheme = scheme.lower() - return super(Url, cls).__new__( - cls, scheme, auth, host, port, path, query, fragment - ) + return super().__new__(cls, scheme, auth, host, port, path, query, fragment) @property - def hostname(self): + def hostname(self) -> str | None: """For backwards-compatibility with urlparse. We're nice like that.""" return self.host @property - def request_uri(self): + def request_uri(self) -> str: """Absolute path including the query string.""" uri = self.path or "/" @@ -122,14 +129,37 @@ class Url(namedtuple("Url", url_attrs)): return uri @property - def netloc(self): - """Network location including host and port""" + def authority(self) -> str | None: + """ + Authority component as defined in RFC 3986 3.2. + This includes userinfo (auth), host and port. + + i.e. + userinfo@host:port + """ + userinfo = self.auth + netloc = self.netloc + if netloc is None or userinfo is None: + return netloc + else: + return f"{userinfo}@{netloc}" + + @property + def netloc(self) -> str | None: + """ + Network location including host and port. + + If you need the equivalent of urllib.parse's ``netloc``, + use the ``authority`` property instead. + """ + if self.host is None: + return None if self.port: - return "%s:%d" % (self.host, self.port) + return f"{self.host}:{self.port}" return self.host @property - def url(self): + def url(self) -> str: """ Convert self into a url @@ -138,88 +168,77 @@ class Url(namedtuple("Url", url_attrs)): :func:`.parse_url`, but it should be equivalent by the RFC (e.g., urls with a blank port will have : removed). - Example: :: + Example: - >>> U = parse_url('http://google.com/mail/') - >>> U.url - 'http://google.com/mail/' - >>> Url('http', 'username:password', 'host.com', 80, - ... '/path', 'query', 'fragment').url - 'http://username:password@host.com:80/path?query#fragment' + .. code-block:: python + + import urllib3 + + U = urllib3.util.parse_url("https://google.com/mail/") + + print(U.url) + # "https://google.com/mail/" + + print( urllib3.util.Url("https", "username:password", + "host.com", 80, "/path", "query", "fragment" + ).url + ) + # "https://username:password@host.com:80/path?query#fragment" """ scheme, auth, host, port, path, query, fragment = self - url = u"" + url = "" # We use "is not None" we want things to happen with empty strings (or 0 port) if scheme is not None: - url += scheme + u"://" + url += scheme + "://" if auth is not None: - url += auth + u"@" + url += auth + "@" if host is not None: url += host if port is not None: - url += u":" + str(port) + url += ":" + str(port) if path is not None: url += path if query is not None: - url += u"?" + query + url += "?" + query if fragment is not None: - url += u"#" + fragment + url += "#" + fragment return url - def __str__(self): + def __str__(self) -> str: return self.url -def split_first(s, delims): - """ - .. deprecated:: 1.25 - - Given a string and an iterable of delimiters, split on the first found - delimiter. Return two split parts and the matched delimiter. - - If not found, then the first part is the full input string. - - Example:: - - >>> split_first('foo/bar?baz', '?/=') - ('foo', 'bar?baz', '/') - >>> split_first('foo/bar?baz', '123') - ('foo/bar?baz', '', None) - - Scales linearly with number of delims. Not ideal for large number of delims. - """ - min_idx = None - min_delim = None - for d in delims: - idx = s.find(d) - if idx < 0: - continue - - if min_idx is None or idx < min_idx: - min_idx = idx - min_delim = d - - if min_idx is None or min_idx < 0: - return s, "", None - - return s[:min_idx], s[min_idx + 1 :], min_delim +@typing.overload +def _encode_invalid_chars( + component: str, allowed_chars: typing.Container[str] +) -> str: # Abstract + ... -def _encode_invalid_chars(component, allowed_chars, encoding="utf-8"): +@typing.overload +def _encode_invalid_chars( + component: None, allowed_chars: typing.Container[str] +) -> None: # Abstract + ... + + +def _encode_invalid_chars( + component: str | None, allowed_chars: typing.Container[str] +) -> str | None: """Percent-encodes a URI component without reapplying onto an already percent-encoded component. """ if component is None: return component - component = six.ensure_text(component) + component = to_str(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. - component, percent_encodings = PERCENT_RE.subn( + component, percent_encodings = _PERCENT_RE.subn( lambda match: match.group(0).upper(), component ) @@ -228,7 +247,7 @@ def _encode_invalid_chars(component, allowed_chars, encoding="utf-8"): encoded_component = bytearray() for i in range(0, len(uri_bytes)): - # Will return a single character bytestring on both Python 2 & 3 + # Will return a single character bytestring byte = uri_bytes[i : i + 1] byte_ord = ord(byte) if (is_percent_encoded and byte == b"%") or ( @@ -238,10 +257,10 @@ def _encode_invalid_chars(component, allowed_chars, encoding="utf-8"): continue encoded_component.extend(b"%" + (hex(byte_ord)[2:].encode().zfill(2).upper())) - return encoded_component.decode(encoding) + return encoded_component.decode() -def _remove_path_dot_segments(path): +def _remove_path_dot_segments(path: str) -> str: # See http://tools.ietf.org/html/rfc3986#section-5.2.4 for pseudo-code segments = path.split("/") # Turn the path into a list of segments output = [] # Initialize the variable to use to store output @@ -251,7 +270,7 @@ def _remove_path_dot_segments(path): if segment == ".": continue # Anything other than '..', should be appended to the output - elif segment != "..": + if segment != "..": output.append(segment) # In this case segment == '..', if we can, we should pop the last # element @@ -271,18 +290,25 @@ def _remove_path_dot_segments(path): return "/".join(output) -def _normalize_host(host, scheme): - if host: - if isinstance(host, six.binary_type): - host = six.ensure_str(host) +@typing.overload +def _normalize_host(host: None, scheme: str | None) -> None: + ... - if scheme in NORMALIZABLE_SCHEMES: - is_ipv6 = IPV6_ADDRZ_RE.match(host) + +@typing.overload +def _normalize_host(host: str, scheme: str | None) -> str: + ... + + +def _normalize_host(host: str | None, scheme: str | None) -> str | None: + if host: + 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) + match = _ZONE_ID_RE.search(host) if match: start, end = match.span(1) zone_id = host[start:end] @@ -291,46 +317,56 @@ def _normalize_host(host, scheme): zone_id = zone_id[3:] else: zone_id = zone_id[1:] - zone_id = "%" + _encode_invalid_chars(zone_id, UNRESERVED_CHARS) - return host[:start].lower() + zone_id + host[end:] + zone_id = _encode_invalid_chars(zone_id, _UNRESERVED_CHARS) + return f"{host[:start].lower()}%{zone_id}{host[end:]}" else: return host.lower() - elif not IPV4_RE.match(host): - return six.ensure_str( - b".".join([_idna_encode(label) for label in host.split(".")]) + elif not _IPV4_RE.match(host): + return to_str( + b".".join([_idna_encode(label) for label in host.split(".")]), + "ascii", ) return host -def _idna_encode(name): - if name and any(ord(x) >= 128 for x in name): +def _idna_encode(name: str) -> bytes: + if not name.isascii(): try: import idna except ImportError: - six.raise_from( - LocationParseError("Unable to parse URL without the 'idna' module"), - None, - ) + raise LocationParseError( + "Unable to parse URL without the 'idna' module" + ) from None + try: return idna.encode(name.lower(), strict=True, std3_rules=True) except idna.IDNAError: - six.raise_from( - LocationParseError(u"Name '%s' is not a valid IDNA label" % name), None - ) + raise LocationParseError( + f"Name '{name}' is not a valid IDNA label" + ) from None + return name.lower().encode("ascii") -def _encode_target(target): - """Percent-encodes a request target so that there are no invalid characters""" - path, query = TARGET_RE.match(target).groups() - target = _encode_invalid_chars(path, PATH_CHARS) - query = _encode_invalid_chars(query, QUERY_CHARS) +def _encode_target(target: str) -> str: + """Percent-encodes a request target so that there are no invalid characters + + Pre-condition for this function is that 'target' must start with '/'. + If that is the case then _TARGET_RE will always produce a match. + """ + match = _TARGET_RE.match(target) + if not match: # Defensive: + raise LocationParseError(f"{target!r} is not a valid request URI") + + path, query = match.groups() + encoded_target = _encode_invalid_chars(path, _PATH_CHARS) if query is not None: - target += "?" + query - return target + query = _encode_invalid_chars(query, _QUERY_CHARS) + encoded_target += "?" + query + return encoded_target -def parse_url(url): +def parse_url(url: str) -> Url: """ Given a url, return a parsed :class:`.Url` namedtuple. Best-effort is performed to parse incomplete urls. Fields not provided will be None. @@ -341,28 +377,44 @@ def parse_url(url): :param str url: URL to parse into a :class:`.Url` namedtuple. - Partly backwards-compatible with :mod:`urlparse`. + Partly backwards-compatible with :mod:`urllib.parse`. - Example:: + Example: - >>> parse_url('http://google.com/mail/') - Url(scheme='http', host='google.com', port=None, path='/mail/', ...) - >>> parse_url('google.com:80') - Url(scheme=None, host='google.com', port=80, path=None, ...) - >>> parse_url('/foo?bar') - Url(scheme=None, host=None, port=None, path='/foo', query='bar', ...) + .. code-block:: python + + import urllib3 + + print( urllib3.util.parse_url('http://google.com/mail/')) + # Url(scheme='http', host='google.com', port=None, path='/mail/', ...) + + print( urllib3.util.parse_url('google.com:80')) + # Url(scheme=None, host='google.com', port=80, path=None, ...) + + print( urllib3.util.parse_url('/foo?bar')) + # Url(scheme=None, host=None, port=None, path='/foo', query='bar', ...) """ if not url: # Empty return Url() source_url = url - if not SCHEME_RE.search(url): + if not _SCHEME_RE.search(url): url = "//" + url + scheme: str | None + authority: str | None + auth: str | None + host: str | None + port: str | None + port_int: int | None + path: str | None + query: str | None + fragment: str | None + try: - scheme, authority, path, query, fragment = URI_RE.match(url).groups() - normalize_uri = scheme is None or scheme.lower() in NORMALIZABLE_SCHEMES + scheme, authority, path, query, fragment = _URI_RE.match(url).groups() # type: ignore[union-attr] + normalize_uri = scheme is None or scheme.lower() in _NORMALIZABLE_SCHEMES if scheme: scheme = scheme.lower() @@ -370,31 +422,33 @@ def parse_url(url): if authority: auth, _, host_port = authority.rpartition("@") auth = auth or None - host, port = _HOST_PORT_RE.match(host_port).groups() + host, port = _HOST_PORT_RE.match(host_port).groups() # type: ignore[union-attr] if auth and normalize_uri: - auth = _encode_invalid_chars(auth, USERINFO_CHARS) + auth = _encode_invalid_chars(auth, _USERINFO_CHARS) if port == "": port = None else: auth, host, port = None, None, None if port is not None: - port = int(port) - if not (0 <= port <= 65535): + port_int = int(port) + if not (0 <= port_int <= 65535): raise LocationParseError(url) + else: + port_int = None host = _normalize_host(host, scheme) if normalize_uri and path: path = _remove_path_dot_segments(path) - path = _encode_invalid_chars(path, PATH_CHARS) + path = _encode_invalid_chars(path, _PATH_CHARS) if normalize_uri and query: - query = _encode_invalid_chars(query, QUERY_CHARS) + query = _encode_invalid_chars(query, _QUERY_CHARS) if normalize_uri and fragment: - fragment = _encode_invalid_chars(fragment, FRAGMENT_CHARS) + fragment = _encode_invalid_chars(fragment, _FRAGMENT_CHARS) - except (ValueError, AttributeError): - return six.raise_from(LocationParseError(source_url), None) + except (ValueError, AttributeError) as e: + raise LocationParseError(source_url) from e # For the sake of backwards compatibility we put empty # string values for path if there are any defined values @@ -406,30 +460,12 @@ def parse_url(url): else: path = None - # Ensure that each part of the URL is a `str` for - # backwards compatibility. - if isinstance(url, six.text_type): - ensure_func = six.ensure_text - else: - ensure_func = six.ensure_str - - def ensure_type(x): - return x if x is None else ensure_func(x) - return Url( - scheme=ensure_type(scheme), - auth=ensure_type(auth), - host=ensure_type(host), - port=port, - path=ensure_type(path), - query=ensure_type(query), - fragment=ensure_type(fragment), + scheme=scheme, + auth=auth, + host=host, + port=port_int, + path=path, + query=query, + fragment=fragment, ) - - -def get_host(url): - """ - Deprecated. Use :func:`parse_url` instead. - """ - p = parse_url(url) - return p.scheme or "http", p.hostname, p.port diff --git a/lib/urllib3/util/url.pyi b/lib/urllib3/util/url.pyi deleted file mode 100644 index 67e308fe..00000000 --- a/lib/urllib3/util/url.pyi +++ /dev/null @@ -1,32 +0,0 @@ -from typing import Any, List, Optional, Tuple, Union - -from .. import exceptions - -LocationParseError = exceptions.LocationParseError - -url_attrs: List[str] - -class Url: - slots: Any - def __new__( - cls, - scheme: Optional[str], - auth: Optional[str], - host: Optional[str], - port: Optional[str], - path: Optional[str], - query: Optional[str], - fragment: Optional[str], - ) -> Url: ... - @property - def hostname(self) -> str: ... - @property - def request_uri(self) -> str: ... - @property - def netloc(self) -> str: ... - @property - def url(self) -> str: ... - -def split_first(s: str, delims: str) -> Tuple[str, str, Optional[str]]: ... -def parse_url(url: str) -> Url: ... -def get_host(url: str) -> Union[str, Tuple[str]]: ... diff --git a/lib/urllib3/util/util.py b/lib/urllib3/util/util.py new file mode 100644 index 00000000..35c77e40 --- /dev/null +++ b/lib/urllib3/util/util.py @@ -0,0 +1,42 @@ +from __future__ import annotations + +import typing +from types import TracebackType + + +def to_bytes( + x: str | bytes, encoding: str | None = None, errors: str | None = None +) -> bytes: + if isinstance(x, bytes): + return x + elif not isinstance(x, str): + raise TypeError(f"not expecting type {type(x).__name__}") + if encoding or errors: + return x.encode(encoding or "utf-8", errors=errors or "strict") + return x.encode() + + +def to_str( + x: str | bytes, encoding: str | None = None, errors: str | None = None +) -> str: + if isinstance(x, str): + return x + elif not isinstance(x, bytes): + raise TypeError(f"not expecting type {type(x).__name__}") + if encoding or errors: + return x.decode(encoding or "utf-8", errors=errors or "strict") + return x.decode() + + +def reraise( + tp: type[BaseException] | None, + value: BaseException, + tb: TracebackType | None = None, +) -> typing.NoReturn: + try: + if value.__traceback__ is not tb: + raise value.with_traceback(tb) + raise value + finally: + value = None # type: ignore[assignment] + tb = None diff --git a/lib/urllib3/util/wait.py b/lib/urllib3/util/wait.py index 21b4590b..aeca0c7a 100644 --- a/lib/urllib3/util/wait.py +++ b/lib/urllib3/util/wait.py @@ -1,18 +1,10 @@ -import errno +from __future__ import annotations + import select -import sys +import socket from functools import partial -try: - from time import monotonic -except ImportError: - from time import time as monotonic - -__all__ = ["NoWayToWaitForSocketError", "wait_for_read", "wait_for_write"] - - -class NoWayToWaitForSocketError(Exception): - pass +__all__ = ["wait_for_read", "wait_for_write"] # How should we wait on sockets? @@ -37,37 +29,13 @@ class NoWayToWaitForSocketError(Exception): # So: on Windows we use select(), and everywhere else we use poll(). We also # fall back to select() in case poll() is somehow broken or missing. -if sys.version_info >= (3, 5): - # Modern Python, that retries syscalls by default - def _retry_on_intr(fn, timeout): - return fn(timeout) -else: - # Old and broken Pythons. - def _retry_on_intr(fn, timeout): - if timeout is None: - deadline = float("inf") - else: - deadline = monotonic() + timeout - - while True: - try: - return fn(timeout) - # OSError for 3 <= pyver < 3.5, select.error for pyver <= 2.7 - except (OSError, select.error) as e: - # 'e.args[0]' incantation works for both OSError and select.error - if e.args[0] != errno.EINTR: - raise - else: - timeout = deadline - monotonic() - if timeout < 0: - timeout = 0 - if timeout == float("inf"): - timeout = None - continue - - -def select_wait_for_socket(sock, read=False, write=False, timeout=None): +def select_wait_for_socket( + sock: socket.socket, + read: bool = False, + write: bool = False, + timeout: float | None = None, +) -> bool: if not read and not write: raise RuntimeError("must specify at least one of read=True, write=True") rcheck = [] @@ -82,11 +50,16 @@ def select_wait_for_socket(sock, read=False, write=False, timeout=None): # sockets for both conditions. (The stdlib selectors module does the same # thing.) fn = partial(select.select, rcheck, wcheck, wcheck) - rready, wready, xready = _retry_on_intr(fn, timeout) + rready, wready, xready = fn(timeout) return bool(rready or wready or xready) -def poll_wait_for_socket(sock, read=False, write=False, timeout=None): +def poll_wait_for_socket( + sock: socket.socket, + read: bool = False, + write: bool = False, + timeout: float | None = None, +) -> bool: if not read and not write: raise RuntimeError("must specify at least one of read=True, write=True") mask = 0 @@ -98,32 +71,33 @@ def poll_wait_for_socket(sock, read=False, write=False, timeout=None): poll_obj.register(sock, mask) # For some reason, poll() takes timeout in milliseconds - def do_poll(t): + def do_poll(t: float | None) -> list[tuple[int, int]]: if t is not None: t *= 1000 return poll_obj.poll(t) - return bool(_retry_on_intr(do_poll, timeout)) + return bool(do_poll(timeout)) -def null_wait_for_socket(*args, **kwargs): - raise NoWayToWaitForSocketError("no select-equivalent available") - - -def _have_working_poll(): +def _have_working_poll() -> bool: # Apparently some systems have a select.poll that fails as soon as you try # to use it, either due to strange configuration or broken monkeypatching # from libraries like eventlet/greenlet. try: poll_obj = select.poll() - _retry_on_intr(poll_obj.poll, 0) + poll_obj.poll(0) except (AttributeError, OSError): return False else: return True -def wait_for_socket(*args, **kwargs): +def wait_for_socket( + sock: socket.socket, + read: bool = False, + write: bool = False, + timeout: float | None = None, +) -> bool: # We delay choosing which implementation to use until the first time we're # called. We could do it at import time, but then we might make the wrong # decision if someone goes wild with monkeypatching select.poll after @@ -133,19 +107,17 @@ def wait_for_socket(*args, **kwargs): wait_for_socket = poll_wait_for_socket elif hasattr(select, "select"): wait_for_socket = select_wait_for_socket - else: # Platform-specific: Appengine. - wait_for_socket = null_wait_for_socket - return wait_for_socket(*args, **kwargs) + return wait_for_socket(sock, read, write, timeout) -def wait_for_read(sock, timeout=None): +def wait_for_read(sock: socket.socket, timeout: float | None = None) -> bool: """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): +def wait_for_write(sock: socket.socket, timeout: float | None = None) -> bool: """Waits for writing to be available on a given socket. Returns True if the socket is readable, or False if the timeout expired. """