Repository navigation
Inspect headers for correct syntax #128
Description
In a similar vein as #119 there is another difference between WebTest and a real WSGI server: WebTest passes HTTP headers through as-is, but HTTP has syntactical rules that need to be obeyed. I noticed this when I tried to base64-encode a (somewhat lengthy) header value -- the result can contain newlines, which (when not escaped) do not survive a transport through HTTP.
Here is an example:
from wsgiref.simple_server import WSGIServer, WSGIRequestHandler
import json
import requests
import threading
import unittest
import webtest
class Test(unittest.TestCase):
HEADERS = {'X-Foo': 'Bar\nBaz\n'}
def test_real_http_truncates_header_on_newline(self):
server = WSGIServer(('localhost', 0), WSGIRequestHandler)
port = server.server_port
server.set_app(echo_header_app)
thread = threading.Thread(target=server.serve_forever)
thread.daemon = True
thread.start()
r = requests.get('http://localhost:%s/' % port, headers=self.HEADERS)
# This is the expected truncated result.
self.assertEqual('Bar', r.json()['HTTP_X_FOO'])
server.shutdown()
thread.join()
def test_webtest_should_not_pass_header_through_unchecked(self):
app = webtest.TestApp(echo_header_app)
r = app.get('/', headers=self.HEADERS)
# This fails, actual value is 'Bar\nBaz\n'
self.assertEqual('Bar', json.loads(r.body)['HTTP_X_FOO'])
def echo_header_app(env, start_response):
status = '200 OK'
headers = [('Content-type', 'application/json')]
start_response(status, headers)
return [json.dumps(
{key: value for key, value in env.items() if key.startswith('HTTP_')})]
Activity