aboutsummaryrefslogtreecommitdiff
path: root/urllib3
diff options
context:
space:
mode:
Diffstat (limited to 'urllib3')
-rw-r--r--urllib3/__init__.py9
-rw-r--r--urllib3/_collections.py229
-rw-r--r--urllib3/connection.py2
-rw-r--r--urllib3/connectionpool.py31
-rw-r--r--urllib3/contrib/pyopenssl.py23
-rw-r--r--urllib3/exceptions.py10
-rw-r--r--urllib3/poolmanager.py19
-rw-r--r--urllib3/response.py203
-rw-r--r--urllib3/util/connection.py1
-rw-r--r--urllib3/util/ssl_.py58
-rw-r--r--urllib3/util/url.py2
11 files changed, 455 insertions, 132 deletions
diff --git a/urllib3/__init__.py b/urllib3/__init__.py
index 4f9d4a7..f48ac4a 100644
--- a/urllib3/__init__.py
+++ b/urllib3/__init__.py
@@ -4,7 +4,7 @@ urllib3 - Thread-safe connection pooling and re-using.
__author__ = 'Andrey Petrov (andrey.petrov@shazow.net)'
__license__ = 'MIT'
-__version__ = '1.10'
+__version__ = '1.10.4'
from .connectionpool import (
@@ -55,9 +55,12 @@ def add_stderr_logger(level=logging.DEBUG):
del NullHandler
-# Set security warning to always go off by default.
import warnings
-warnings.simplefilter('always', exceptions.SecurityWarning)
+# SecurityWarning's always go off by default.
+warnings.simplefilter('always', exceptions.SecurityWarning, append=True)
+# InsecurePlatformWarning's don't vary between requests, so we keep it default.
+warnings.simplefilter('default', exceptions.InsecurePlatformWarning,
+ append=True)
def disable_warnings(category=exceptions.HTTPWarning):
"""
diff --git a/urllib3/_collections.py b/urllib3/_collections.py
index 784342a..279416c 100644
--- a/urllib3/_collections.py
+++ b/urllib3/_collections.py
@@ -1,7 +1,7 @@
from collections import Mapping, MutableMapping
try:
from threading import RLock
-except ImportError: # Platform-specific: No threads available
+except ImportError: # Platform-specific: No threads available
class RLock:
def __enter__(self):
pass
@@ -10,11 +10,11 @@ except ImportError: # Platform-specific: No threads available
pass
-try: # Python 2.7+
+try: # Python 2.7+
from collections import OrderedDict
except ImportError:
from .packages.ordered_dict import OrderedDict
-from .packages.six import iterkeys, itervalues
+from .packages.six import iterkeys, itervalues, PY3
__all__ = ['RecentlyUsedContainer', 'HTTPHeaderDict']
@@ -97,7 +97,14 @@ class RecentlyUsedContainer(MutableMapping):
return list(iterkeys(self._container))
-class HTTPHeaderDict(MutableMapping):
+_dict_setitem = dict.__setitem__
+_dict_getitem = dict.__getitem__
+_dict_delitem = dict.__delitem__
+_dict_contains = dict.__contains__
+_dict_setdefault = dict.setdefault
+
+
+class HTTPHeaderDict(dict):
"""
:param headers:
An iterable of field-value pairs. Must not contain multiple field names
@@ -129,25 +136,75 @@ class HTTPHeaderDict(MutableMapping):
'foo=bar, baz=quxx'
>>> headers['Content-Length']
'7'
-
- If you want to access the raw headers with their original casing
- for debugging purposes you can access the private ``._data`` attribute
- which is a normal python ``dict`` that maps the case-insensitive key to a
- list of tuples stored as (case-sensitive-original-name, value). Using the
- structure from above as our example:
-
- >>> headers._data
- {'set-cookie': [('Set-Cookie', 'foo=bar'), ('set-cookie', 'baz=quxx')],
- 'content-length': [('content-length', '7')]}
"""
def __init__(self, headers=None, **kwargs):
- self._data = {}
- if headers is None:
- headers = {}
- self.update(headers, **kwargs)
+ dict.__init__(self)
+ if headers is not None:
+ if isinstance(headers, HTTPHeaderDict):
+ self._copy_from(headers)
+ else:
+ self.extend(headers)
+ if kwargs:
+ self.extend(kwargs)
+
+ def __setitem__(self, key, val):
+ return _dict_setitem(self, key.lower(), (key, val))
+
+ def __getitem__(self, key):
+ val = _dict_getitem(self, key.lower())
+ return ', '.join(val[1:])
+
+ def __delitem__(self, key):
+ return _dict_delitem(self, key.lower())
- def add(self, key, value):
+ def __contains__(self, key):
+ return _dict_contains(self, key.lower())
+
+ def __eq__(self, other):
+ if not isinstance(other, Mapping) and not hasattr(other, 'keys'):
+ return False
+ if not isinstance(other, type(self)):
+ other = type(self)(other)
+ return dict((k1, self[k1]) for k1 in self) == dict((k2, other[k2]) for k2 in other)
+
+ def __ne__(self, other):
+ return not self.__eq__(other)
+
+ values = MutableMapping.values
+ get = MutableMapping.get
+ update = MutableMapping.update
+
+ if not PY3: # Python 2
+ iterkeys = MutableMapping.iterkeys
+ itervalues = MutableMapping.itervalues
+
+ __marker = object()
+
+ 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):
+ try:
+ del self[key]
+ except KeyError:
+ pass
+
+ def add(self, key, val):
"""Adds a (name, value) pair, doesn't overwrite the value if it already
exists.
@@ -156,43 +213,111 @@ class HTTPHeaderDict(MutableMapping):
>>> headers['foo']
'bar, baz'
"""
- self._data.setdefault(key.lower(), []).append((key, value))
+ key_lower = key.lower()
+ new_vals = key, val
+ # Keep the common case aka no item present as fast as possible
+ vals = _dict_setdefault(self, key_lower, new_vals)
+ if new_vals is not vals:
+ # new_vals was not inserted, as there was a previous one
+ if isinstance(vals, list):
+ # If already several items got inserted, we have a list
+ vals.append(val)
+ else:
+ # vals should be a tuple then, i.e. only one item so far
+ # Need to convert the tuple to list for further extension
+ _dict_setitem(self, key_lower, [vals[0], vals[1], val])
+
+ def extend(self, *args, **kwargs):
+ """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 ({} given)".format(len(args)))
+ 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:
+ for key, value in other:
+ self.add(key, value)
+
+ for key, value in kwargs.items():
+ self.add(key, value)
def getlist(self, key):
"""Returns a list of all the values for the named field. Returns an
empty list if the key doesn't exist."""
- return self[key].split(', ') if key in self else []
-
- def copy(self):
- h = HTTPHeaderDict()
- for key in self._data:
- for rawkey, value in self._data[key]:
- h.add(rawkey, value)
- return h
-
- def __eq__(self, other):
- if not isinstance(other, Mapping):
- return False
- other = HTTPHeaderDict(other)
- return dict((k1, self[k1]) for k1 in self._data) == \
- dict((k2, other[k2]) for k2 in other._data)
-
- def __getitem__(self, key):
- values = self._data[key.lower()]
- return ', '.join(value[1] for value in values)
-
- def __setitem__(self, key, value):
- self._data[key.lower()] = [(key, value)]
+ try:
+ vals = _dict_getitem(self, key.lower())
+ except KeyError:
+ return []
+ else:
+ if isinstance(vals, tuple):
+ return [vals[1]]
+ else:
+ return vals[1:]
+
+ # Backwards compatibility for httplib
+ getheaders = getlist
+ getallmatchingheaders = getlist
+ iget = getlist
- def __delitem__(self, key):
- del self._data[key.lower()]
+ def __repr__(self):
+ return "%s(%s)" % (type(self).__name__, dict(self.itermerged()))
- def __len__(self):
- return len(self._data)
+ def _copy_from(self, other):
+ for key in other:
+ val = _dict_getitem(other, key)
+ if isinstance(val, list):
+ # Don't need to convert tuples
+ val = list(val)
+ _dict_setitem(self, key, val)
- def __iter__(self):
- for headers in itervalues(self._data):
- yield headers[0][0]
-
- def __repr__(self):
- return '%s(%r)' % (self.__class__.__name__, dict(self.items()))
+ def copy(self):
+ clone = type(self)()
+ clone._copy_from(self)
+ return clone
+
+ def iteritems(self):
+ """Iterate over all header lines, including duplicate ones."""
+ for key in self:
+ vals = _dict_getitem(self, key)
+ for val in vals[1:]:
+ yield vals[0], val
+
+ def itermerged(self):
+ """Iterate over all headers, merging duplicate ones together."""
+ for key in self:
+ val = _dict_getitem(self, key)
+ yield val[0], ', '.join(val[1:])
+
+ def items(self):
+ return list(self.iteritems())
+
+ @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.
+ headers = []
+
+ for line in message.headers:
+ if line.startswith((' ', '\t')):
+ key, value = headers[-1]
+ headers[-1] = (key, value + '\r\n' + line.rstrip())
+ continue
+
+ key, value = line.split(':', 1)
+ headers.append((key, value.strip()))
+
+ return cls(headers)
diff --git a/urllib3/connection.py b/urllib3/connection.py
index e5de769..2a8c359 100644
--- a/urllib3/connection.py
+++ b/urllib3/connection.py
@@ -260,3 +260,5 @@ if ssl:
# Make a copy for testing.
UnverifiedHTTPSConnection = HTTPSConnection
HTTPSConnection = VerifiedHTTPSConnection
+else:
+ HTTPSConnection = DummyConnection
diff --git a/urllib3/connectionpool.py b/urllib3/connectionpool.py
index 8bdf228..117269a 100644
--- a/urllib3/connectionpool.py
+++ b/urllib3/connectionpool.py
@@ -72,6 +72,21 @@ class ConnectionPool(object):
return '%s(host=%r, port=%r)' % (type(self).__name__,
self.host, self.port)
+ def __enter__(self):
+ return self
+
+ def __exit__(self, exc_type, exc_val, exc_tb):
+ self.close()
+ # Return False to re-raise any potential exceptions
+ return False
+
+ def close():
+ """
+ Close all pooled connections and disable the pool.
+ """
+ pass
+
+
# This is taken from http://hg.python.org/cpython/file/7aaba721ebc0/Lib/socket.py#l252
_blocking_errnos = set([errno.EAGAIN, errno.EWOULDBLOCK])
@@ -353,7 +368,7 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods):
# Receive the response from the server
try:
- try: # Python 2.7+, use buffering of HTTP responses
+ try: # Python 2.7, use buffering of HTTP responses
httplib_response = conn.getresponse(buffering=True)
except TypeError: # Python 2.6 and older
httplib_response = conn.getresponse()
@@ -558,6 +573,14 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods):
conn = None
raise SSLError(e)
+ except SSLError:
+ # Treat SSLError separately from BaseSSLError to preserve
+ # traceback.
+ if conn:
+ conn.close()
+ conn = None
+ raise
+
except (TimeoutError, HTTPException, SocketError, ConnectionError) as e:
if conn:
# Discard the connection for these exceptions. It will be
@@ -565,14 +588,13 @@ class HTTPConnectionPool(ConnectionPool, RequestMethods):
conn.close()
conn = None
- stacktrace = sys.exc_info()[2]
if isinstance(e, SocketError) and self.proxy:
e = ProxyError('Cannot connect to proxy.', e)
elif isinstance(e, (SocketError, HTTPException)):
e = ProtocolError('Connection aborted.', e)
- retries = retries.increment(method, url, error=e,
- _pool=self, _stacktrace=stacktrace)
+ retries = retries.increment(method, url, error=e, _pool=self,
+ _stacktrace=sys.exc_info()[2])
retries.sleep()
# Keep track of the error for the retry warning.
@@ -713,7 +735,6 @@ class HTTPSConnectionPool(HTTPConnectionPool):
% (self.num_connections, self.host))
if not self.ConnectionCls or self.ConnectionCls is DummyConnection:
- # Platform-specific: Python without ssl
raise SSLError("Can't connect to HTTPS URL because the SSL "
"module is not available.")
diff --git a/urllib3/contrib/pyopenssl.py b/urllib3/contrib/pyopenssl.py
index ee657fb..b2c34a8 100644
--- a/urllib3/contrib/pyopenssl.py
+++ b/urllib3/contrib/pyopenssl.py
@@ -38,8 +38,6 @@ Module Variables
----------------
:var DEFAULT_SSL_CIPHER_LIST: The list of supported SSL/TLS cipher suites.
- Default: ``ECDH+AESGCM:DH+AESGCM:ECDH+AES256:DH+AES256:ECDH+AES128:DH+AES:
- ECDH+3DES:DH+3DES:RSA+AESGCM:RSA+AES:RSA+3DES:!aNULL:!MD5:!DSS``
.. _sni: https://en.wikipedia.org/wiki/Server_Name_Indication
.. _crime attack: https://en.wikipedia.org/wiki/CRIME_(security_exploit)
@@ -85,22 +83,7 @@ _openssl_verify = {
+ OpenSSL.SSL.VERIFY_FAIL_IF_NO_PEER_CERT,
}
-# 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 over any AES-CBC for better performance and security,
-# - use 3DES as fallback which is secure but slow,
-# - disable NULL authentication, MD5 MACs and DSS for security reasons.
-DEFAULT_SSL_CIPHER_LIST = "ECDH+AESGCM:DH+AESGCM:ECDH+AES256:DH+AES256:" + \
- "ECDH+AES128:DH+AES:ECDH+3DES:DH+3DES:RSA+AESGCM:RSA+AES:RSA+3DES:" + \
- "!aNULL:!MD5:!DSS"
+DEFAULT_SSL_CIPHER_LIST = util.ssl_.DEFAULT_CIPHERS
orig_util_HAS_SNI = util.HAS_SNI
@@ -299,7 +282,9 @@ def ssl_wrap_socket(sock, keyfile=None, certfile=None, cert_reqs=None,
try:
cnx.do_handshake()
except OpenSSL.SSL.WantReadError:
- select.select([sock], [], [])
+ rd, _, _ = select.select([sock], [], [], sock.gettimeout())
+ if not rd:
+ raise timeout('select timed out')
continue
except OpenSSL.SSL.Error as e:
raise ssl.SSLError('bad handshake', e)
diff --git a/urllib3/exceptions.py b/urllib3/exceptions.py
index 0c6fd3c..31bda1c 100644
--- a/urllib3/exceptions.py
+++ b/urllib3/exceptions.py
@@ -157,3 +157,13 @@ class InsecureRequestWarning(SecurityWarning):
class SystemTimeWarning(SecurityWarning):
"Warned when system time is suspected to be wrong"
pass
+
+
+class InsecurePlatformWarning(SecurityWarning):
+ "Warned when certain SSL configuration is not available on a platform."
+ pass
+
+
+class ResponseNotChunked(ProtocolError, ValueError):
+ "Response needs to be chunked in order to read it as chunks."
+ pass
diff --git a/urllib3/poolmanager.py b/urllib3/poolmanager.py
index 515dc96..b8d1e74 100644
--- a/urllib3/poolmanager.py
+++ b/urllib3/poolmanager.py
@@ -8,7 +8,7 @@ except ImportError:
from ._collections import RecentlyUsedContainer
from .connectionpool import HTTPConnectionPool, HTTPSConnectionPool
from .connectionpool import port_by_scheme
-from .exceptions import LocationValueError
+from .exceptions import LocationValueError, MaxRetryError
from .request import RequestMethods
from .util.url import parse_url
from .util.retry import Retry
@@ -64,6 +64,14 @@ class PoolManager(RequestMethods):
self.pools = RecentlyUsedContainer(num_pools,
dispose_func=lambda p: p.close())
+ def __enter__(self):
+ return self
+
+ def __exit__(self, exc_type, exc_val, exc_tb):
+ self.clear()
+ # Return False to re-raise any potential exceptions
+ return False
+
def _new_pool(self, scheme, host, port):
"""
Create a new :class:`ConnectionPool` based on host, port and scheme.
@@ -167,7 +175,14 @@ class PoolManager(RequestMethods):
if not isinstance(retries, Retry):
retries = Retry.from_int(retries, redirect=redirect)
- kw['retries'] = retries.increment(method, redirect_location)
+ try:
+ retries = retries.increment(method, url, response=response, _pool=conn)
+ except MaxRetryError:
+ if retries.raise_on_redirect:
+ raise
+ return response
+
+ kw['retries'] = retries
kw['redirect'] = redirect
log.info("Redirecting %s -> %s" % (url, redirect_location))
diff --git a/urllib3/response.py b/urllib3/response.py
index e69de95..24140c4 100644
--- a/urllib3/response.py
+++ b/urllib3/response.py
@@ -1,15 +1,20 @@
+try:
+ import http.client as httplib
+except ImportError:
+ import httplib
import zlib
import io
from socket import timeout as SocketTimeout
from ._collections import HTTPHeaderDict
-from .exceptions import ProtocolError, DecodeError, ReadTimeoutError
-from .packages.six import string_types as basestring, binary_type
+from .exceptions import (
+ ProtocolError, DecodeError, ReadTimeoutError, ResponseNotChunked
+)
+from .packages.six import string_types as basestring, binary_type, PY3
from .connection import HTTPException, BaseSSLError
from .util.response import is_fp_closed
-
class DeflateDecoder(object):
def __init__(self):
@@ -21,6 +26,9 @@ class DeflateDecoder(object):
return getattr(self._obj, name)
def decompress(self, data):
+ if not data:
+ return data
+
if not self._first_try:
return self._obj.decompress(data)
@@ -36,9 +44,23 @@ class DeflateDecoder(object):
self._data = None
+class GzipDecoder(object):
+
+ def __init__(self):
+ self._obj = zlib.decompressobj(16 + zlib.MAX_WBITS)
+
+ def __getattr__(self, name):
+ return getattr(self._obj, name)
+
+ def decompress(self, data):
+ if not data:
+ return data
+ return self._obj.decompress(data)
+
+
def _get_decoder(mode):
if mode == 'gzip':
- return zlib.decompressobj(16 + zlib.MAX_WBITS)
+ return GzipDecoder()
return DeflateDecoder()
@@ -76,9 +98,10 @@ class HTTPResponse(io.IOBase):
strict=0, preload_content=True, decode_content=True,
original_response=None, pool=None, connection=None):
- self.headers = HTTPHeaderDict()
- if headers:
- self.headers.update(headers)
+ if isinstance(headers, HTTPHeaderDict):
+ self.headers = headers
+ else:
+ self.headers = HTTPHeaderDict(headers)
self.status = status
self.version = version
self.reason = reason
@@ -100,7 +123,17 @@ class HTTPResponse(io.IOBase):
if hasattr(body, 'read'):
self._fp = body
- if preload_content and not self._body:
+ # 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
+
+ # We certainly don't want to preload content when the response is chunked.
+ if not self.chunked and preload_content and not self._body:
self._body = self.read(decode_content=decode_content)
def get_redirect_location(self):
@@ -140,6 +173,35 @@ class HTTPResponse(io.IOBase):
"""
return self._fp_bytes_read
+ def _init_decoder(self):
+ """
+ Set-up the _decoder attribute if necessar.
+ """
+ # 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 and content_encoding in self.CONTENT_DECODERS:
+ self._decoder = _get_decoder(content_encoding)
+
+ def _decode(self, data, decode_content, flush_decoder):
+ """
+ Decode the data passed in and potentially flush the decoder.
+ """
+ try:
+ if decode_content and self._decoder:
+ data = self._decoder.decompress(data)
+ except (IOError, zlib.error) 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 and decode_content and self._decoder:
+ buf = self._decoder.decompress(binary_type())
+ data += buf + self._decoder.flush()
+
+ return data
+
def read(self, amt=None, decode_content=None, cache_content=False):
"""
Similar to :meth:`httplib.HTTPResponse.read`, but with two additional
@@ -161,12 +223,7 @@ class HTTPResponse(io.IOBase):
after having ``.read()`` the file object. (Overridden if ``amt`` is
set.)
"""
- # 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)
+ self._init_decoder()
if decode_content is None:
decode_content = self.decode_content
@@ -202,7 +259,7 @@ class HTTPResponse(io.IOBase):
except BaseSSLError as e:
# FIXME: Is there a better way to differentiate between SSLErrors?
- if not 'read operation timed out' in str(e): # Defensive:
+ if 'read operation timed out' not in str(e): # Defensive:
# This shouldn't happen but just in case we're missing an edge
# case, let's avoid swallowing SSL errors.
raise
@@ -215,17 +272,7 @@ class HTTPResponse(io.IOBase):
self._fp_bytes_read += len(data)
- try:
- if decode_content and self._decoder:
- data = self._decoder.decompress(data)
- except (IOError, zlib.error) as e:
- raise DecodeError(
- "Received response with content-encoding: %s, but "
- "failed to decode it." % content_encoding, e)
-
- if flush_decoder and decode_content and self._decoder:
- buf = self._decoder.decompress(binary_type())
- data += buf + self._decoder.flush()
+ data = self._decode(data, decode_content, flush_decoder)
if cache_content:
self._body = data
@@ -252,11 +299,15 @@ class HTTPResponse(io.IOBase):
If True, will attempt to decode the body based on the
'content-encoding' header.
"""
- while not is_fp_closed(self._fp):
- data = self.read(amt=amt, decode_content=decode_content)
+ if self.chunked:
+ for line in self.read_chunked(amt, decode_content=decode_content):
+ yield line
+ else:
+ while not is_fp_closed(self._fp):
+ data = self.read(amt=amt, decode_content=decode_content)
- if data:
- yield data
+ if data:
+ yield data
@classmethod
def from_httplib(ResponseCls, r, **response_kw):
@@ -267,14 +318,16 @@ class HTTPResponse(io.IOBase):
Remaining parameters are passed to the HTTPResponse constructor, along
with ``original_response=r``.
"""
-
- headers = HTTPHeaderDict()
- for k, v in r.getheaders():
- headers.add(k, v)
+ headers = r.msg
+ if not isinstance(headers, HTTPHeaderDict):
+ if PY3: # Python 3
+ headers = HTTPHeaderDict(headers.items())
+ else: # Python 2
+ headers = HTTPHeaderDict.from_httplib(headers)
# HTTPResponse objects in Python 3 don't have a .strict attribute
strict = getattr(r, 'strict', 0)
- return ResponseCls(body=r,
+ resp = ResponseCls(body=r,
headers=headers,
status=r.status,
version=r.version,
@@ -282,6 +335,7 @@ class HTTPResponse(io.IOBase):
strict=strict,
original_response=r,
**response_kw)
+ return resp
# Backwards-compatibility methods for httplib.HTTPResponse
def getheaders(self):
@@ -331,3 +385,82 @@ class HTTPResponse(io.IOBase):
else:
b[:len(temp)] = temp
return len(temp)
+
+ def _update_chunk_length(self):
+ # 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()
+ line = line.split(b';', 1)[0]
+ try:
+ self.chunk_left = int(line, 16)
+ except ValueError:
+ # Invalid chunked protocol response, abort.
+ self.close()
+ raise httplib.IncompleteRead(line)
+
+ def _handle_chunk(self, amt):
+ returned_chunk = None
+ if amt is None:
+ chunk = self._fp._safe_read(self.chunk_left)
+ returned_chunk = chunk
+ self._fp._safe_read(2) # Toss the CRLF at the end of the chunk.
+ self.chunk_left = None
+ elif amt < self.chunk_left:
+ value = self._fp._safe_read(amt)
+ 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.
+ 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.
+ self.chunk_left = None
+ return returned_chunk
+
+ def read_chunked(self, amt=None, decode_content=None):
+ """
+ Similar to :meth:`HTTPResponse.read`, but with an additional
+ parameter: ``decode_content``.
+
+ :param decode_content:
+ If True, will attempt to decode the body based on the
+ 'content-encoding' header.
+ """
+ self._init_decoder()
+ # FIXME: Rewrite this method and make it a class with a better structured logic.
+ if not self.chunked:
+ raise ResponseNotChunked("Response is not chunked. "
+ "Header 'transfer-encoding: chunked' is missing.")
+
+ if self._original_response and self._original_response._method.upper() == 'HEAD':
+ # Don't bother reading the body of a HEAD request.
+ # FIXME: Can we do this somehow without accessing private httplib _method?
+ self._original_response.close()
+ return
+
+ while True:
+ self._update_chunk_length()
+ if self.chunk_left == 0:
+ break
+ chunk = self._handle_chunk(amt)
+ yield self._decode(chunk, decode_content=decode_content,
+ flush_decoder=True)
+
+ # Chunk content ends with \r\n: discard it.
+ while True:
+ line = self._fp.fp.readline()
+ if not line:
+ # Some sites may not end with '\r\n'.
+ break
+ if line == b'\r\n':
+ break
+
+ # We read everything; close the "file".
+ if self._original_response:
+ self._original_response.close()
+ self.release_conn()
diff --git a/urllib3/util/connection.py b/urllib3/util/connection.py
index 2156993..859aec6 100644
--- a/urllib3/util/connection.py
+++ b/urllib3/util/connection.py
@@ -82,6 +82,7 @@ def create_connection(address, timeout=socket._GLOBAL_DEFAULT_TIMEOUT,
err = _
if sock is not None:
sock.close()
+ sock = None
if err is not None:
raise err
diff --git a/urllib3/util/ssl_.py b/urllib3/util/ssl_.py
index a788b1b..b846d42 100644
--- a/urllib3/util/ssl_.py
+++ b/urllib3/util/ssl_.py
@@ -1,7 +1,7 @@
from binascii import hexlify, unhexlify
-from hashlib import md5, sha1
+from hashlib import md5, sha1, sha256
-from ..exceptions import SSLError
+from ..exceptions import SSLError, InsecurePlatformWarning
SSLContext = None
@@ -9,9 +9,10 @@ HAS_SNI = False
create_default_context = None
import errno
-import ssl
+import warnings
try: # Test for SSL features
+ import ssl
from ssl import wrap_socket, CERT_NONE, PROTOCOL_SSLv23
from ssl import HAS_SNI # Has SNI?
except ImportError:
@@ -24,14 +25,24 @@ except ImportError:
OP_NO_SSLv2, OP_NO_SSLv3 = 0x1000000, 0x2000000
OP_NO_COMPRESSION = 0x20000
-try:
- from ssl import _DEFAULT_CIPHERS
-except ImportError:
- _DEFAULT_CIPHERS = (
- 'ECDH+AESGCM:DH+AESGCM:ECDH+AES256:DH+AES256:ECDH+AES128:DH+AES:ECDH+HIGH:'
- 'DH+HIGH:ECDH+3DES:DH+3DES:RSA+AESGCM:RSA+AES:RSA+HIGH:RSA+3DES:ECDH+RC4:'
- 'DH+RC4:RSA+RC4:!aNULL:!eNULL:!MD5'
- )
+# 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 over any AES-CBC for better performance and security,
+# - use 3DES as fallback which is secure but slow,
+# - disable NULL authentication, MD5 MACs and DSS for security reasons.
+DEFAULT_CIPHERS = (
+ 'ECDH+AESGCM:DH+AESGCM:ECDH+AES256:DH+AES256:ECDH+AES128:DH+AES:ECDH+HIGH:'
+ 'DH+HIGH:ECDH+3DES:DH+3DES:RSA+AESGCM:RSA+AES:RSA+HIGH:RSA+3DES:!aNULL:'
+ '!eNULL:!MD5'
+)
try:
from ssl import SSLContext # Modern SSL?
@@ -39,7 +50,8 @@ except ImportError:
import sys
class SSLContext(object): # Platform-specific: Python 2 & 3.1
- supports_set_ciphers = sys.version_info >= (2, 7)
+ supports_set_ciphers = ((2, 7) <= sys.version_info < (3,) or
+ (3, 2) <= sys.version_info)
def __init__(self, protocol_version):
self.protocol = protocol_version
@@ -69,6 +81,14 @@ except ImportError:
self.ciphers = cipher_suite
def wrap_socket(self, socket, server_hostname=None):
+ warnings.warn(
+ 'A true SSLContext object is not available. This prevents '
+ 'urllib3 from configuring SSL appropriately and may cause '
+ 'certain SSL connections to fail. For more information, see '
+ 'https://urllib3.readthedocs.org/en/latest/security.html'
+ '#insecureplatformwarning.',
+ InsecurePlatformWarning
+ )
kwargs = {
'keyfile': self.keyfile,
'certfile': self.certfile,
@@ -96,7 +116,8 @@ def assert_fingerprint(cert, fingerprint):
# this digest.
hashfunc_map = {
16: md5,
- 20: sha1
+ 20: sha1,
+ 32: sha256,
}
fingerprint = fingerprint.replace(':', '').lower()
@@ -157,7 +178,7 @@ def resolve_ssl_version(candidate):
return candidate
-def create_urllib3_context(ssl_version=None, cert_reqs=ssl.CERT_REQUIRED,
+def create_urllib3_context(ssl_version=None, cert_reqs=None,
options=None, ciphers=None):
"""All arguments have the same meaning as ``ssl_wrap_socket``.
@@ -194,6 +215,9 @@ def create_urllib3_context(ssl_version=None, cert_reqs=ssl.CERT_REQUIRED,
"""
context = SSLContext(ssl_version or ssl.PROTOCOL_SSLv23)
+ # 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
+
if options is None:
options = 0
# SSLv2 is easily broken and is considered harmful and dangerous
@@ -207,11 +231,13 @@ def create_urllib3_context(ssl_version=None, cert_reqs=ssl.CERT_REQUIRED,
context.options |= options
if getattr(context, 'supports_set_ciphers', True): # Platform-specific: Python 2.6
- context.set_ciphers(ciphers or _DEFAULT_CIPHERS)
+ context.set_ciphers(ciphers or DEFAULT_CIPHERS)
context.verify_mode = cert_reqs
if getattr(context, 'check_hostname', None) is not None: # Platform-specific: Python 3.2
- context.check_hostname = (context.verify_mode == ssl.CERT_REQUIRED)
+ # We do our own verification, including fingerprints and alternative
+ # hostnames. So disable it here
+ context.check_hostname = False
return context
diff --git a/urllib3/util/url.py b/urllib3/util/url.py
index b2ec834..e58050c 100644
--- a/urllib3/util/url.py
+++ b/urllib3/util/url.py
@@ -15,6 +15,8 @@ class Url(namedtuple('Url', url_attrs)):
def __new__(cls, scheme=None, auth=None, host=None, port=None, path=None,
query=None, fragment=None):
+ if path and not path.startswith('/'):
+ path = '/' + path
return super(Url, cls).__new__(cls, scheme, auth, host, port, path,
query, fragment)