Connection timeout doesn't work for chunked-encoded request

[psushin]

Sample code:

def gen():
    for i in xrange(1, 1000):
        yield 'x'
import requests, datetime

print datetime.datetime.now()

try: requests.post(url='http://1.1.1.1', timeout=5, data=gen())
except Exception as e: print e

print datetime.datetime.now()

try: requests.post(url='http://1.1.1.1', timeout=5, data='a')
except Exception as e: print e

print datetime.datetime.now()

gives the following output

2013-06-14 00:33:03.079389
[Errno 110] Connection timed out
2013-06-14 00:34:06.236742
HTTPConnectionPool(host='1.1.1.1', port=80): Request timed out. (timeout=5)
2013-06-14 00:34:11.242869

The first request blocked for more than a minute, before a timeout has been thrown, while the second one is ok.

Requests 1.2.3
Python 2.7.3

[Lukasa]

Thanks for opening this issue!

Looking at the Requests code, I suspect you might be right. I'm at work at the moment and won't have internet at home until the 18th, but when I can get time I'll investigate this.

To begin with, I think the problem is around here. For chunked uploads, we do this:

low_conn = conn._get_conn(timeout=timeout)
low_conn.putrequest(request.method, url, skip_accept_encoding=True)

for header, value in request.headers.items():
    low_conn.putheader(header, value)

low_conn.endheaders()

for i in request.body:
    low_conn.send(hex(len(i))[2:].encode('utf-8'))
    low_conn.send(b'\r\n')
    low_conn.send(i)
    low_conn.send(b'\r\n')

low_conn.send(b'0\r\n\r\n')

r = low_conn.getresponse()
resp = HTTPResponse.from_httplib(r,
    pool=conn,
    connection=low_conn,
    preload_content=False,
    decode_content=False
 )

The only place we set the timeout is in the call to urllib3.HTTPConnectionPool._get_conn(), which passes down into its LifoQueue. This timeout affects how long we wait to obtain a connection object, but does not set the timeout on that connection.

I need to think about what's best to do here, but I think the answer is to set low_conn.timeout = timeout before the initial call to putrequest(). Feel free to try that out if I don't get around to testing it first.

[sigmavirus24]

I wonder if our new Timeout system will fix this.

[isobit]

Seems this is still happening with 2.18.4. I was able to fix it using low_conn.timeout = timeout.connect_timeout, like @Lukasa mentioned. Should I write a test and submit a pull request or just create a new issue?

[sethmlarson]

Raise a new issue, link this one, and include your test case that can reproduce the problem consistently. Then you can submit a PR and link to your new issue if you have time :) Thanks @isobit!