bytes-like objects with socket.sendall(), SSL, and http.client

[vadmium]

BPO	27340
Nosy	@tiran, @vadmium
PRs	bpo-27340: Use memoryview in SSLSocket.sendall() #3384 [3.6] bpo-27340: Use memoryview in SSLSocket.sendall() (GH-3384) #3434

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2017-09-08.00:05:19.262>
created_at = <Date 2016-06-17.04:25:20.448>
labels = ['3.7', 'expert-SSL', 'type-bug', 'library', 'docs']
title = 'bytes-like objects with socket.sendall(), SSL, and http.client'
updated_at = <Date 2017-09-08.00:05:19.261>
user = 'https://github.com/vadmium'

bugs.python.org fields:

activity = <Date 2017-09-08.00:05:19.261>
actor = 'christian.heimes'
assignee = 'docs@python'
closed = True
closed_date = <Date 2017-09-08.00:05:19.262>
closer = 'christian.heimes'
components = ['Documentation', 'Library (Lib)', 'SSL']
creation = <Date 2016-06-17.04:25:20.448>
creator = 'martin.panter'
dependencies = []
files = []
hgrepos = []
issue_num = 27340
keywords = ['patch']
message_count = 4.0
messages = ['268707', '301639', '301652', '301653']
nosy_count = 3.0
nosy_names = ['christian.heimes', 'docs@python', 'martin.panter']
pr_nums = ['3384', '3434']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue27340'
versions = ['Python 3.6', 'Python 3.7']

[vadmium]

According to the documentation, HTTPSConnection.request() should accept arbitrary bytes-like objects, but this is not the case. Currently (since bpo-23756), a “bytes-like object” is defined to be anything that works with Python’s buffer API, as long as it is C-contiguous. These objects can be passed to socket.sendall():

>> byteslike = (ctypes.c_ubyte * 6).from_buffer_copy(b"DATA\r\n")
>> s = socket.create_connection(("localhost", 80))
>> s.sendall(byteslike) # Server receives b"DATA\r\n"

This is not explicitly documented for socket objects. But since bpo-23539 (3.4+), HTTPConnection.request() does document support for bytes-like objects:

>> h = HTTPConnection("localhost", 80)
>> # Send Content-Length: 6 and body b"DATA\r\n"
>> h.request("POST", "/", body=byteslike)

On its own, there is no problem with Python relying on its own undocumented behaviour. But Python’s “ssl” module does not support arbitrary bytes-like objects, and as a result neither does HTTPSConnection:

>>> s = ssl.wrap_socket(socket.create_connection(("localhost", 443)))
>>> s.sendall(byteslike)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.5/ssl.py", line 886, in sendall
    v = self.send(data[count:])
  File "/usr/lib/python3.5/ssl.py", line 856, in send
    return self._sslobj.write(data)
  File "/usr/lib/python3.5/ssl.py", line 581, in write
    return self._sslobj.write(data)
TypeError: a bytes-like object is required, not 'list'
>>> c = ssl.create_default_context(cafile="/lib/python3.5/test/keycert.pem")
>>> h = HTTPSConnection("localhost", 443, context=c)
>>> h.request("POST", "/", body=byteslike)
Traceback (most recent call last):
  File "/usr/lib/python3.5/http/client.py", line 885, in send
    self.sock.sendall(data)
  File "/usr/lib/python3.5/ssl.py", line 886, in sendall
    v = self.send(data[count:])
  File "/usr/lib/python3.5/ssl.py", line 856, in send
    return self._sslobj.write(data)
  File "/usr/lib/python3.5/ssl.py", line 581, in write
    return self._sslobj.write(data)
TypeError: a bytes-like object is required, not 'list'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.5/http/client.py", line 1083, in request
    self._send_request(method, url, body, headers)
  File "/usr/lib/python3.5/http/client.py", line 1128, in _send_request
    self.endheaders(body)
  File "/usr/lib/python3.5/http/client.py", line 1079, in endheaders
    self._send_output(message_body)
  File "/usr/lib/python3.5/http/client.py", line 913, in _send_output
    self.send(message_body)
  File "/usr/lib/python3.5/http/client.py", line 892, in send
    "or an iterable, got %r" % type(data))
TypeError: data should be a bytes-like object or an iterable, got <class '__main__.c_ubyte_Array_6'>

This could be fixed in the implementation of SSLSocket.sendall(). But I am not sure if that is an appropriate change for 3.5. Another option would be to adjust the documentation of HTTP(S)Connection in 3.5, either not mentioning bytes-like objects at all, or clarifying that they don’t work with SSL.

[tiran]

New changeset 888bbdc by Christian Heimes in branch 'master':
bpo-27340: Use memoryview in SSLSocket.sendall() (bpo-3384)
888bbdc

[tiran]

New changeset 9423f5d by Christian Heimes in branch '3.6':
[3.6] bpo-27340: Use memoryview in SSLSocket.sendall() (GH-3384) (bpo-3434)
9423f5d

[tiran]

In Python 3.6 and master, SSLSocket.sendall() now accepts any bytes-like object that can be cast to bytes. Since it uses a sliced memoryview instead of sliced bytes, it avoids extra copies of the payload, too.