mirror of
https://github.com/ytdl-org/youtube-dl
synced 2024-11-18 11:30:11 +09:00
[core] Update redirect handling from yt-dlp
* Thx coletdjnz: https://github.com/yt-dlp/yt-dlp/pull/7094 * add test that redirected `POST` loses its `Content-Type`
This commit is contained in:
parent
648dc5304c
commit
46fde7caee
@ -8,33 +8,160 @@ import sys
|
|||||||
import unittest
|
import unittest
|
||||||
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
|
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
|
||||||
|
|
||||||
|
import gzip
|
||||||
|
import io
|
||||||
|
import ssl
|
||||||
|
import tempfile
|
||||||
|
import threading
|
||||||
|
import zlib
|
||||||
|
|
||||||
|
# avoid deprecated alias assertRaisesRegexp
|
||||||
|
if hasattr(unittest.TestCase, 'assertRaisesRegex'):
|
||||||
|
unittest.TestCase.assertRaisesRegexp = unittest.TestCase.assertRaisesRegex
|
||||||
|
|
||||||
|
try:
|
||||||
|
import brotli
|
||||||
|
except ImportError:
|
||||||
|
brotli = None
|
||||||
|
try:
|
||||||
|
from urllib.request import pathname2url
|
||||||
|
except ImportError:
|
||||||
|
from urllib import pathname2url
|
||||||
|
|
||||||
|
from youtube_dl.compat import (
|
||||||
|
compat_http_cookiejar_Cookie,
|
||||||
|
compat_http_server,
|
||||||
|
compat_str as str,
|
||||||
|
compat_urllib_error,
|
||||||
|
compat_urllib_HTTPError,
|
||||||
|
compat_urllib_parse,
|
||||||
|
compat_urllib_request,
|
||||||
|
)
|
||||||
|
|
||||||
|
from youtube_dl.utils import (
|
||||||
|
sanitized_Request,
|
||||||
|
urlencode_postdata,
|
||||||
|
)
|
||||||
|
|
||||||
from test.helper import (
|
from test.helper import (
|
||||||
|
FakeYDL,
|
||||||
FakeLogger,
|
FakeLogger,
|
||||||
http_server_port,
|
http_server_port,
|
||||||
)
|
)
|
||||||
from youtube_dl import YoutubeDL
|
from youtube_dl import YoutubeDL
|
||||||
from youtube_dl.compat import compat_http_server, compat_urllib_request
|
|
||||||
import ssl
|
|
||||||
import threading
|
|
||||||
|
|
||||||
TEST_DIR = os.path.dirname(os.path.abspath(__file__))
|
TEST_DIR = os.path.dirname(os.path.abspath(__file__))
|
||||||
|
|
||||||
|
|
||||||
class HTTPTestRequestHandler(compat_http_server.BaseHTTPRequestHandler):
|
class HTTPTestRequestHandler(compat_http_server.BaseHTTPRequestHandler):
|
||||||
|
protocol_version = 'HTTP/1.1'
|
||||||
|
|
||||||
|
# work-around old/new -style class inheritance
|
||||||
|
def super(self, meth_name, *args, **kwargs):
|
||||||
|
from types import MethodType
|
||||||
|
try:
|
||||||
|
super()
|
||||||
|
fn = lambda s, m, *a, **k: getattr(super(), m)(*a, **k)
|
||||||
|
except TypeError:
|
||||||
|
fn = lambda s, m, *a, **k: getattr(compat_http_server.BaseHTTPRequestHandler, m)(s, *a, **k)
|
||||||
|
self.super = MethodType(fn, self)
|
||||||
|
return self.super(meth_name, *args, **kwargs)
|
||||||
|
|
||||||
def log_message(self, format, *args):
|
def log_message(self, format, *args):
|
||||||
pass
|
pass
|
||||||
|
|
||||||
|
def _headers(self):
|
||||||
|
payload = str(self.headers).encode('utf-8')
|
||||||
|
self.send_response(200)
|
||||||
|
self.send_header('Content-Type', 'application/json')
|
||||||
|
self.send_header('Content-Length', str(len(payload)))
|
||||||
|
self.end_headers()
|
||||||
|
self.wfile.write(payload)
|
||||||
|
|
||||||
|
def _redirect(self):
|
||||||
|
self.send_response(int(self.path[len('/redirect_'):]))
|
||||||
|
self.send_header('Location', '/method')
|
||||||
|
self.send_header('Content-Length', '0')
|
||||||
|
self.end_headers()
|
||||||
|
|
||||||
|
def _method(self, method, payload=None):
|
||||||
|
self.send_response(200)
|
||||||
|
self.send_header('Content-Length', str(len(payload or '')))
|
||||||
|
self.send_header('Method', method)
|
||||||
|
self.end_headers()
|
||||||
|
if payload:
|
||||||
|
self.wfile.write(payload)
|
||||||
|
|
||||||
|
def _status(self, status):
|
||||||
|
payload = '<html>{0} NOT FOUND</html>'.format(status).encode('utf-8')
|
||||||
|
self.send_response(int(status))
|
||||||
|
self.send_header('Content-Type', 'text/html; charset=utf-8')
|
||||||
|
self.send_header('Content-Length', str(len(payload)))
|
||||||
|
self.end_headers()
|
||||||
|
self.wfile.write(payload)
|
||||||
|
|
||||||
|
def _read_data(self):
|
||||||
|
if 'Content-Length' in self.headers:
|
||||||
|
return self.rfile.read(int(self.headers['Content-Length']))
|
||||||
|
|
||||||
|
def _test_url(self, path, host='127.0.0.1', scheme='http', port=None):
|
||||||
|
return '{0}://{1}:{2}/{3}'.format(
|
||||||
|
scheme, host,
|
||||||
|
port if port is not None
|
||||||
|
else http_server_port(self.server), path)
|
||||||
|
|
||||||
|
def do_POST(self):
|
||||||
|
data = self._read_data()
|
||||||
|
if self.path.startswith('/redirect_'):
|
||||||
|
self._redirect()
|
||||||
|
elif self.path.startswith('/method'):
|
||||||
|
self._method('POST', data)
|
||||||
|
elif self.path.startswith('/headers'):
|
||||||
|
self._headers()
|
||||||
|
else:
|
||||||
|
self._status(404)
|
||||||
|
|
||||||
|
def do_HEAD(self):
|
||||||
|
if self.path.startswith('/redirect_'):
|
||||||
|
self._redirect()
|
||||||
|
elif self.path.startswith('/method'):
|
||||||
|
self._method('HEAD')
|
||||||
|
else:
|
||||||
|
self._status(404)
|
||||||
|
|
||||||
|
def do_PUT(self):
|
||||||
|
data = self._read_data()
|
||||||
|
if self.path.startswith('/redirect_'):
|
||||||
|
self._redirect()
|
||||||
|
elif self.path.startswith('/method'):
|
||||||
|
self._method('PUT', data)
|
||||||
|
else:
|
||||||
|
self._status(404)
|
||||||
|
|
||||||
def do_GET(self):
|
def do_GET(self):
|
||||||
|
|
||||||
|
def respond(payload=b'<html><video src="/vid.mp4" /></html>',
|
||||||
|
payload_type='text/html; charset=utf-8',
|
||||||
|
payload_encoding=None,
|
||||||
|
resp_code=200):
|
||||||
|
self.send_response(resp_code)
|
||||||
|
self.send_header('Content-Type', payload_type)
|
||||||
|
if payload_encoding:
|
||||||
|
self.send_header('Content-Encoding', payload_encoding)
|
||||||
|
self.send_header('Content-Length', str(len(payload))) # required for persistent connections
|
||||||
|
self.end_headers()
|
||||||
|
self.wfile.write(payload)
|
||||||
|
|
||||||
|
def gzip_compress(p):
|
||||||
|
buf = io.BytesIO()
|
||||||
|
with gzip.GzipFile(fileobj=buf, mode='wb') as f:
|
||||||
|
f.write(p)
|
||||||
|
return buf.getvalue()
|
||||||
|
|
||||||
if self.path == '/video.html':
|
if self.path == '/video.html':
|
||||||
self.send_response(200)
|
respond()
|
||||||
self.send_header('Content-Type', 'text/html; charset=utf-8')
|
|
||||||
self.end_headers()
|
|
||||||
self.wfile.write(b'<html><video src="/vid.mp4" /></html>')
|
|
||||||
elif self.path == '/vid.mp4':
|
elif self.path == '/vid.mp4':
|
||||||
self.send_response(200)
|
respond(b'\x00\x00\x00\x00\x20\x66\x74[video]', 'video/mp4')
|
||||||
self.send_header('Content-Type', 'video/mp4')
|
|
||||||
self.end_headers()
|
|
||||||
self.wfile.write(b'\x00\x00\x00\x00\x20\x66\x74[video]')
|
|
||||||
elif self.path == '/302':
|
elif self.path == '/302':
|
||||||
if sys.version_info[0] == 3:
|
if sys.version_info[0] == 3:
|
||||||
# XXX: Python 3 http server does not allow non-ASCII header values
|
# XXX: Python 3 http server does not allow non-ASCII header values
|
||||||
@ -42,60 +169,284 @@ class HTTPTestRequestHandler(compat_http_server.BaseHTTPRequestHandler):
|
|||||||
self.end_headers()
|
self.end_headers()
|
||||||
return
|
return
|
||||||
|
|
||||||
new_url = 'http://127.0.0.1:%d/中文.html' % http_server_port(self.server)
|
new_url = self._test_url('中文.html')
|
||||||
self.send_response(302)
|
self.send_response(302)
|
||||||
self.send_header(b'Location', new_url.encode('utf-8'))
|
self.send_header(b'Location', new_url.encode('utf-8'))
|
||||||
self.end_headers()
|
self.end_headers()
|
||||||
elif self.path == '/%E4%B8%AD%E6%96%87.html':
|
elif self.path == '/%E4%B8%AD%E6%96%87.html':
|
||||||
self.send_response(200)
|
respond()
|
||||||
self.send_header('Content-Type', 'text/html; charset=utf-8')
|
elif self.path == '/%c7%9f':
|
||||||
|
respond()
|
||||||
|
elif self.path.startswith('/redirect_'):
|
||||||
|
self._redirect()
|
||||||
|
elif self.path.startswith('/method'):
|
||||||
|
self._method('GET')
|
||||||
|
elif self.path.startswith('/headers'):
|
||||||
|
self._headers()
|
||||||
|
elif self.path == '/trailing_garbage':
|
||||||
|
payload = b'<html><video src="/vid.mp4" /></html>'
|
||||||
|
compressed = gzip_compress(payload) + b'trailing garbage'
|
||||||
|
respond(compressed, payload_encoding='gzip')
|
||||||
|
elif self.path == '/302-non-ascii-redirect':
|
||||||
|
new_url = self._test_url('中文.html')
|
||||||
|
# actually respond with permanent redirect
|
||||||
|
self.send_response(301)
|
||||||
|
self.send_header('Location', new_url)
|
||||||
|
self.send_header('Content-Length', '0')
|
||||||
self.end_headers()
|
self.end_headers()
|
||||||
self.wfile.write(b'<html><video src="/vid.mp4" /></html>')
|
elif self.path == '/content-encoding':
|
||||||
|
encodings = self.headers.get('ytdl-encoding', '')
|
||||||
|
payload = b'<html><video src="/vid.mp4" /></html>'
|
||||||
|
for encoding in filter(None, (e.strip() for e in encodings.split(','))):
|
||||||
|
if encoding == 'br' and brotli:
|
||||||
|
payload = brotli.compress(payload)
|
||||||
|
elif encoding == 'gzip':
|
||||||
|
payload = gzip_compress(payload)
|
||||||
|
elif encoding == 'deflate':
|
||||||
|
payload = zlib.compress(payload)
|
||||||
|
elif encoding == 'unsupported':
|
||||||
|
payload = b'raw'
|
||||||
|
break
|
||||||
|
else:
|
||||||
|
self._status(415)
|
||||||
|
return
|
||||||
|
respond(payload, payload_encoding=encodings)
|
||||||
|
|
||||||
else:
|
else:
|
||||||
assert False
|
self._status(404)
|
||||||
|
|
||||||
|
def send_header(self, keyword, value):
|
||||||
|
"""
|
||||||
|
Forcibly allow HTTP server to send non percent-encoded non-ASCII characters in headers.
|
||||||
|
This is against what is defined in RFC 3986: but we need to test that we support this
|
||||||
|
since some sites incorrectly do this.
|
||||||
|
"""
|
||||||
|
if keyword.lower() == 'connection':
|
||||||
|
return self.super('send_header', keyword, value)
|
||||||
|
|
||||||
|
if not hasattr(self, '_headers_buffer'):
|
||||||
|
self._headers_buffer = []
|
||||||
|
|
||||||
|
self._headers_buffer.append('{0}: {1}\r\n'.format(keyword, value).encode('utf-8'))
|
||||||
|
|
||||||
|
def end_headers(self):
|
||||||
|
if hasattr(self, '_headers_buffer'):
|
||||||
|
self.wfile.write(b''.join(self._headers_buffer))
|
||||||
|
self._headers_buffer = []
|
||||||
|
self.super('end_headers')
|
||||||
|
|
||||||
|
|
||||||
class TestHTTP(unittest.TestCase):
|
class TestHTTP(unittest.TestCase):
|
||||||
def setUp(self):
|
def setUp(self):
|
||||||
self.httpd = compat_http_server.HTTPServer(
|
# HTTP server
|
||||||
|
self.http_httpd = compat_http_server.HTTPServer(
|
||||||
('127.0.0.1', 0), HTTPTestRequestHandler)
|
('127.0.0.1', 0), HTTPTestRequestHandler)
|
||||||
self.port = http_server_port(self.httpd)
|
self.http_port = http_server_port(self.http_httpd)
|
||||||
self.server_thread = threading.Thread(target=self.httpd.serve_forever)
|
|
||||||
self.server_thread.daemon = True
|
|
||||||
self.server_thread.start()
|
|
||||||
|
|
||||||
def test_unicode_path_redirection(self):
|
self.http_server_thread = threading.Thread(target=self.http_httpd.serve_forever)
|
||||||
# XXX: Python 3 http server does not allow non-ASCII header values
|
self.http_server_thread.daemon = True
|
||||||
if sys.version_info[0] == 3:
|
self.http_server_thread.start()
|
||||||
return
|
|
||||||
|
|
||||||
ydl = YoutubeDL({'logger': FakeLogger()})
|
try:
|
||||||
r = ydl.extract_info('http://127.0.0.1:%d/302' % self.port)
|
from http.server import ThreadingHTTPServer
|
||||||
self.assertEqual(r['entries'][0]['url'], 'http://127.0.0.1:%d/vid.mp4' % self.port)
|
except ImportError:
|
||||||
|
try:
|
||||||
|
from socketserver import ThreadingMixIn
|
||||||
|
except ImportError:
|
||||||
|
from SocketServer import ThreadingMixIn
|
||||||
|
|
||||||
|
class ThreadingHTTPServer(ThreadingMixIn, compat_http_server.HTTPServer):
|
||||||
|
pass
|
||||||
|
|
||||||
class TestHTTPS(unittest.TestCase):
|
# HTTPS server
|
||||||
def setUp(self):
|
|
||||||
certfn = os.path.join(TEST_DIR, 'testcert.pem')
|
certfn = os.path.join(TEST_DIR, 'testcert.pem')
|
||||||
self.httpd = compat_http_server.HTTPServer(
|
self.https_httpd = ThreadingHTTPServer(
|
||||||
('127.0.0.1', 0), HTTPTestRequestHandler)
|
('127.0.0.1', 0), HTTPTestRequestHandler)
|
||||||
self.httpd.socket = ssl.wrap_socket(
|
try:
|
||||||
self.httpd.socket, certfile=certfn, server_side=True)
|
sslctx = ssl.SSLContext(ssl.PROTOCOL_SSLv23)
|
||||||
self.port = http_server_port(self.httpd)
|
sslctx.verify_mode = ssl.CERT_NONE
|
||||||
self.server_thread = threading.Thread(target=self.httpd.serve_forever)
|
sslctx.check_hostname = False
|
||||||
self.server_thread.daemon = True
|
sslctx.load_cert_chain(certfn, None)
|
||||||
self.server_thread.start()
|
self.https_httpd.socket = sslctx.wrap_socket(
|
||||||
|
self.https_httpd.socket, server_side=True)
|
||||||
|
except AttributeError:
|
||||||
|
self.https_httpd.socket = ssl.wrap_socket(
|
||||||
|
self.https_httpd.socket, certfile=certfn, server_side=True)
|
||||||
|
|
||||||
|
self.https_port = http_server_port(self.https_httpd)
|
||||||
|
self.https_server_thread = threading.Thread(target=self.https_httpd.serve_forever)
|
||||||
|
self.https_server_thread.daemon = True
|
||||||
|
self.https_server_thread.start()
|
||||||
|
|
||||||
|
def tearDown(self):
|
||||||
|
|
||||||
|
def closer(svr):
|
||||||
|
def _closer():
|
||||||
|
svr.shutdown()
|
||||||
|
svr.server_close()
|
||||||
|
return _closer
|
||||||
|
|
||||||
|
shutdown_thread = threading.Thread(target=closer(self.http_httpd))
|
||||||
|
shutdown_thread.start()
|
||||||
|
self.http_server_thread.join(2.0)
|
||||||
|
|
||||||
|
shutdown_thread = threading.Thread(target=closer(self.https_httpd))
|
||||||
|
shutdown_thread.start()
|
||||||
|
self.https_server_thread.join(2.0)
|
||||||
|
|
||||||
|
def _test_url(self, path, host='127.0.0.1', scheme='http', port=None):
|
||||||
|
return '{0}://{1}:{2}/{3}'.format(
|
||||||
|
scheme, host,
|
||||||
|
port if port is not None
|
||||||
|
else self.https_port if scheme == 'https'
|
||||||
|
else self.http_port, path)
|
||||||
|
|
||||||
def test_nocheckcertificate(self):
|
def test_nocheckcertificate(self):
|
||||||
if sys.version_info >= (2, 7, 9): # No certificate checking anyways
|
with FakeYDL({'logger': FakeLogger()}) as ydl:
|
||||||
ydl = YoutubeDL({'logger': FakeLogger()})
|
with self.assertRaises(compat_urllib_error.URLError):
|
||||||
self.assertRaises(
|
ydl.urlopen(sanitized_Request(self._test_url('headers', scheme='https')))
|
||||||
Exception,
|
|
||||||
ydl.extract_info, 'https://127.0.0.1:%d/video.html' % self.port)
|
|
||||||
|
|
||||||
ydl = YoutubeDL({'logger': FakeLogger(), 'nocheckcertificate': True})
|
with FakeYDL({'logger': FakeLogger(), 'nocheckcertificate': True}) as ydl:
|
||||||
r = ydl.extract_info('https://127.0.0.1:%d/video.html' % self.port)
|
r = ydl.urlopen(sanitized_Request(self._test_url('headers', scheme='https')))
|
||||||
self.assertEqual(r['entries'][0]['url'], 'https://127.0.0.1:%d/vid.mp4' % self.port)
|
self.assertEqual(r.getcode(), 200)
|
||||||
|
r.close()
|
||||||
|
|
||||||
|
def test_percent_encode(self):
|
||||||
|
with FakeYDL() as ydl:
|
||||||
|
# Unicode characters should be encoded with uppercase percent-encoding
|
||||||
|
res = ydl.urlopen(sanitized_Request(self._test_url('中文.html')))
|
||||||
|
self.assertEqual(res.getcode(), 200)
|
||||||
|
res.close()
|
||||||
|
# don't normalize existing percent encodings
|
||||||
|
res = ydl.urlopen(sanitized_Request(self._test_url('%c7%9f')))
|
||||||
|
self.assertEqual(res.getcode(), 200)
|
||||||
|
res.close()
|
||||||
|
|
||||||
|
def test_unicode_path_redirection(self):
|
||||||
|
with FakeYDL() as ydl:
|
||||||
|
r = ydl.urlopen(sanitized_Request(self._test_url('302-non-ascii-redirect')))
|
||||||
|
self.assertEqual(r.url, self._test_url('%E4%B8%AD%E6%96%87.html'))
|
||||||
|
r.close()
|
||||||
|
|
||||||
|
def test_redirect(self):
|
||||||
|
with FakeYDL() as ydl:
|
||||||
|
def do_req(redirect_status, method, check_no_content=False):
|
||||||
|
data = b'testdata' if method in ('POST', 'PUT') else None
|
||||||
|
res = ydl.urlopen(sanitized_Request(
|
||||||
|
self._test_url('redirect_{0}'.format(redirect_status)),
|
||||||
|
method=method, data=data))
|
||||||
|
if check_no_content:
|
||||||
|
self.assertNotIn('Content-Type', res.headers)
|
||||||
|
return res.read().decode('utf-8'), res.headers.get('method', '')
|
||||||
|
# A 303 must either use GET or HEAD for subsequent request
|
||||||
|
self.assertEqual(do_req(303, 'POST'), ('', 'GET'))
|
||||||
|
self.assertEqual(do_req(303, 'HEAD'), ('', 'HEAD'))
|
||||||
|
|
||||||
|
self.assertEqual(do_req(303, 'PUT'), ('', 'GET'))
|
||||||
|
|
||||||
|
# 301 and 302 turn POST only into a GET, with no Content-Type
|
||||||
|
self.assertEqual(do_req(301, 'POST', True), ('', 'GET'))
|
||||||
|
self.assertEqual(do_req(301, 'HEAD'), ('', 'HEAD'))
|
||||||
|
self.assertEqual(do_req(302, 'POST', True), ('', 'GET'))
|
||||||
|
self.assertEqual(do_req(302, 'HEAD'), ('', 'HEAD'))
|
||||||
|
|
||||||
|
self.assertEqual(do_req(301, 'PUT'), ('testdata', 'PUT'))
|
||||||
|
self.assertEqual(do_req(302, 'PUT'), ('testdata', 'PUT'))
|
||||||
|
|
||||||
|
# 307 and 308 should not change method
|
||||||
|
for m in ('POST', 'PUT'):
|
||||||
|
self.assertEqual(do_req(307, m), ('testdata', m))
|
||||||
|
self.assertEqual(do_req(308, m), ('testdata', m))
|
||||||
|
|
||||||
|
self.assertEqual(do_req(307, 'HEAD'), ('', 'HEAD'))
|
||||||
|
self.assertEqual(do_req(308, 'HEAD'), ('', 'HEAD'))
|
||||||
|
|
||||||
|
# These should not redirect and instead raise an HTTPError
|
||||||
|
for code in (300, 304, 305, 306):
|
||||||
|
with self.assertRaises(compat_urllib_HTTPError):
|
||||||
|
do_req(code, 'GET')
|
||||||
|
|
||||||
|
def test_content_type(self):
|
||||||
|
# https://github.com/yt-dlp/yt-dlp/commit/379a4f161d4ad3e40932dcf5aca6e6fb9715ab28
|
||||||
|
with FakeYDL({'nocheckcertificate': True}) as ydl:
|
||||||
|
# method should be auto-detected as POST
|
||||||
|
r = sanitized_Request(self._test_url('headers', scheme='https'), data=urlencode_postdata({'test': 'test'}))
|
||||||
|
|
||||||
|
headers = ydl.urlopen(r).read().decode('utf-8')
|
||||||
|
self.assertIn('Content-Type: application/x-www-form-urlencoded', headers)
|
||||||
|
|
||||||
|
# test http
|
||||||
|
r = sanitized_Request(self._test_url('headers'), data=urlencode_postdata({'test': 'test'}))
|
||||||
|
headers = ydl.urlopen(r).read().decode('utf-8')
|
||||||
|
self.assertIn('Content-Type: application/x-www-form-urlencoded', headers)
|
||||||
|
|
||||||
|
def test_cookiejar(self):
|
||||||
|
with FakeYDL() as ydl:
|
||||||
|
ydl.cookiejar.set_cookie(compat_http_cookiejar_Cookie(
|
||||||
|
0, 'test', 'ytdl', None, False, '127.0.0.1', True,
|
||||||
|
False, '/headers', True, False, None, False, None, None, {}))
|
||||||
|
data = ydl.urlopen(sanitized_Request(self._test_url('headers'))).read()
|
||||||
|
self.assertIn(b'Cookie: test=ytdl', data)
|
||||||
|
|
||||||
|
def test_no_compression_compat_header(self):
|
||||||
|
with FakeYDL() as ydl:
|
||||||
|
data = ydl.urlopen(
|
||||||
|
sanitized_Request(
|
||||||
|
self._test_url('headers'),
|
||||||
|
headers={'Youtubedl-no-compression': True})).read()
|
||||||
|
self.assertIn(b'Accept-Encoding: identity', data)
|
||||||
|
self.assertNotIn(b'youtubedl-no-compression', data.lower())
|
||||||
|
|
||||||
|
def test_gzip_trailing_garbage(self):
|
||||||
|
# https://github.com/ytdl-org/youtube-dl/commit/aa3e950764337ef9800c936f4de89b31c00dfcf5
|
||||||
|
# https://github.com/ytdl-org/youtube-dl/commit/6f2ec15cee79d35dba065677cad9da7491ec6e6f
|
||||||
|
with FakeYDL() as ydl:
|
||||||
|
data = ydl.urlopen(sanitized_Request(self._test_url('trailing_garbage'))).read().decode('utf-8')
|
||||||
|
self.assertEqual(data, '<html><video src="/vid.mp4" /></html>')
|
||||||
|
|
||||||
|
def __test_compression(self, encoding):
|
||||||
|
with FakeYDL() as ydl:
|
||||||
|
res = ydl.urlopen(
|
||||||
|
sanitized_Request(
|
||||||
|
self._test_url('content-encoding'),
|
||||||
|
headers={'ytdl-encoding': encoding}))
|
||||||
|
self.assertEqual(res.headers.get('Content-Encoding'), encoding)
|
||||||
|
self.assertEqual(res.read(), b'<html><video src="/vid.mp4" /></html>')
|
||||||
|
|
||||||
|
@unittest.skipUnless(brotli, 'brotli support is not installed')
|
||||||
|
@unittest.expectedFailure
|
||||||
|
def test_brotli(self):
|
||||||
|
self.__test_compression('br')
|
||||||
|
|
||||||
|
@unittest.expectedFailure
|
||||||
|
def test_deflate(self):
|
||||||
|
self.__test_compression('deflate')
|
||||||
|
|
||||||
|
@unittest.expectedFailure
|
||||||
|
def test_gzip(self):
|
||||||
|
self.__test_compression('gzip')
|
||||||
|
|
||||||
|
@unittest.expectedFailure # not yet implemented
|
||||||
|
def test_multiple_encodings(self):
|
||||||
|
# https://www.rfc-editor.org/rfc/rfc9110.html#section-8.4
|
||||||
|
with FakeYDL() as ydl:
|
||||||
|
for pair in ('gzip,deflate', 'deflate, gzip', 'gzip, gzip', 'deflate, deflate'):
|
||||||
|
res = ydl.urlopen(
|
||||||
|
sanitized_Request(
|
||||||
|
self._test_url('content-encoding'),
|
||||||
|
headers={'ytdl-encoding': pair}))
|
||||||
|
self.assertEqual(res.headers.get('Content-Encoding'), pair)
|
||||||
|
self.assertEqual(res.read(), b'<html><video src="/vid.mp4" /></html>')
|
||||||
|
|
||||||
|
def test_unsupported_encoding(self):
|
||||||
|
# it should return the raw content
|
||||||
|
with FakeYDL() as ydl:
|
||||||
|
res = ydl.urlopen(
|
||||||
|
sanitized_Request(
|
||||||
|
self._test_url('content-encoding'),
|
||||||
|
headers={'ytdl-encoding': 'unsupported'}))
|
||||||
|
self.assertEqual(res.headers.get('Content-Encoding'), 'unsupported')
|
||||||
|
self.assertEqual(res.read(), b'raw')
|
||||||
|
|
||||||
|
|
||||||
def _build_proxy_handler(name):
|
def _build_proxy_handler(name):
|
||||||
@ -109,7 +460,7 @@ def _build_proxy_handler(name):
|
|||||||
self.send_response(200)
|
self.send_response(200)
|
||||||
self.send_header('Content-Type', 'text/plain; charset=utf-8')
|
self.send_header('Content-Type', 'text/plain; charset=utf-8')
|
||||||
self.end_headers()
|
self.end_headers()
|
||||||
self.wfile.write('{self.proxy_name}: {self.path}'.format(self=self).encode('utf-8'))
|
self.wfile.write('{0}: {1}'.format(self.proxy_name, self.path).encode('utf-8'))
|
||||||
return HTTPTestRequestHandler
|
return HTTPTestRequestHandler
|
||||||
|
|
||||||
|
|
||||||
@ -129,10 +480,30 @@ class TestProxy(unittest.TestCase):
|
|||||||
self.geo_proxy_thread.daemon = True
|
self.geo_proxy_thread.daemon = True
|
||||||
self.geo_proxy_thread.start()
|
self.geo_proxy_thread.start()
|
||||||
|
|
||||||
|
def tearDown(self):
|
||||||
|
|
||||||
|
def closer(svr):
|
||||||
|
def _closer():
|
||||||
|
svr.shutdown()
|
||||||
|
svr.server_close()
|
||||||
|
return _closer
|
||||||
|
|
||||||
|
shutdown_thread = threading.Thread(target=closer(self.proxy))
|
||||||
|
shutdown_thread.start()
|
||||||
|
self.proxy_thread.join(2.0)
|
||||||
|
|
||||||
|
shutdown_thread = threading.Thread(target=closer(self.geo_proxy))
|
||||||
|
shutdown_thread.start()
|
||||||
|
self.geo_proxy_thread.join(2.0)
|
||||||
|
|
||||||
|
def _test_proxy(self, host='127.0.0.1', port=None):
|
||||||
|
return '{0}:{1}'.format(
|
||||||
|
host, port if port is not None else self.port)
|
||||||
|
|
||||||
def test_proxy(self):
|
def test_proxy(self):
|
||||||
geo_proxy = '127.0.0.1:{0}'.format(self.geo_port)
|
geo_proxy = self._test_proxy(port=self.geo_port)
|
||||||
ydl = YoutubeDL({
|
ydl = YoutubeDL({
|
||||||
'proxy': '127.0.0.1:{0}'.format(self.port),
|
'proxy': self._test_proxy(),
|
||||||
'geo_verification_proxy': geo_proxy,
|
'geo_verification_proxy': geo_proxy,
|
||||||
})
|
})
|
||||||
url = 'http://foo.com/bar'
|
url = 'http://foo.com/bar'
|
||||||
@ -146,7 +517,7 @@ class TestProxy(unittest.TestCase):
|
|||||||
|
|
||||||
def test_proxy_with_idn(self):
|
def test_proxy_with_idn(self):
|
||||||
ydl = YoutubeDL({
|
ydl = YoutubeDL({
|
||||||
'proxy': '127.0.0.1:{0}'.format(self.port),
|
'proxy': self._test_proxy(),
|
||||||
})
|
})
|
||||||
url = 'http://中文.tw/'
|
url = 'http://中文.tw/'
|
||||||
response = ydl.urlopen(url).read().decode('utf-8')
|
response = ydl.urlopen(url).read().decode('utf-8')
|
||||||
@ -154,5 +525,25 @@ class TestProxy(unittest.TestCase):
|
|||||||
self.assertEqual(response, 'normal: http://xn--fiq228c.tw/')
|
self.assertEqual(response, 'normal: http://xn--fiq228c.tw/')
|
||||||
|
|
||||||
|
|
||||||
|
class TestFileURL(unittest.TestCase):
|
||||||
|
# See https://github.com/ytdl-org/youtube-dl/issues/8227
|
||||||
|
def test_file_urls(self):
|
||||||
|
tf = tempfile.NamedTemporaryFile(delete=False)
|
||||||
|
tf.write(b'foobar')
|
||||||
|
tf.close()
|
||||||
|
url = compat_urllib_parse.urljoin('file://', pathname2url(tf.name))
|
||||||
|
with FakeYDL() as ydl:
|
||||||
|
self.assertRaisesRegexp(
|
||||||
|
compat_urllib_error.URLError, 'file:// scheme is explicitly disabled in youtube-dl for security reasons', ydl.urlopen, url)
|
||||||
|
# not yet implemented
|
||||||
|
"""
|
||||||
|
with FakeYDL({'enable_file_urls': True}) as ydl:
|
||||||
|
res = ydl.urlopen(url)
|
||||||
|
self.assertEqual(res.read(), b'foobar')
|
||||||
|
res.close()
|
||||||
|
"""
|
||||||
|
os.unlink(tf.name)
|
||||||
|
|
||||||
|
|
||||||
if __name__ == '__main__':
|
if __name__ == '__main__':
|
||||||
unittest.main()
|
unittest.main()
|
||||||
|
@ -41,7 +41,6 @@ import zlib
|
|||||||
from .compat import (
|
from .compat import (
|
||||||
compat_HTMLParseError,
|
compat_HTMLParseError,
|
||||||
compat_HTMLParser,
|
compat_HTMLParser,
|
||||||
compat_HTTPError,
|
|
||||||
compat_basestring,
|
compat_basestring,
|
||||||
compat_casefold,
|
compat_casefold,
|
||||||
compat_chr,
|
compat_chr,
|
||||||
@ -64,6 +63,7 @@ from .compat import (
|
|||||||
compat_struct_pack,
|
compat_struct_pack,
|
||||||
compat_struct_unpack,
|
compat_struct_unpack,
|
||||||
compat_urllib_error,
|
compat_urllib_error,
|
||||||
|
compat_urllib_HTTPError,
|
||||||
compat_urllib_parse,
|
compat_urllib_parse,
|
||||||
compat_urllib_parse_parse_qs as compat_parse_qs,
|
compat_urllib_parse_parse_qs as compat_parse_qs,
|
||||||
compat_urllib_parse_urlencode,
|
compat_urllib_parse_urlencode,
|
||||||
@ -2614,7 +2614,8 @@ class YoutubeDLHandler(compat_urllib_request.HTTPHandler):
|
|||||||
|
|
||||||
Part of this code was copied from:
|
Part of this code was copied from:
|
||||||
|
|
||||||
http://techknack.net/python-urllib2-handlers/
|
http://techknack.net/python-urllib2-handlers/, archived at
|
||||||
|
https://web.archive.org/web/20130527205558/http://techknack.net/python-urllib2-handlers/
|
||||||
|
|
||||||
Andrew Rowls, the author of that code, agreed to release it to the
|
Andrew Rowls, the author of that code, agreed to release it to the
|
||||||
public domain.
|
public domain.
|
||||||
@ -2672,7 +2673,9 @@ class YoutubeDLHandler(compat_urllib_request.HTTPHandler):
|
|||||||
req._Request__original = req._Request__original.partition('#')[0]
|
req._Request__original = req._Request__original.partition('#')[0]
|
||||||
req._Request__r_type = req._Request__r_type.partition('#')[0]
|
req._Request__r_type = req._Request__r_type.partition('#')[0]
|
||||||
|
|
||||||
return req
|
# Use the totally undocumented AbstractHTTPHandler per
|
||||||
|
# https://github.com/yt-dlp/yt-dlp/pull/4158
|
||||||
|
return compat_urllib_request.AbstractHTTPHandler.do_request_(self, req)
|
||||||
|
|
||||||
def http_response(self, req, resp):
|
def http_response(self, req, resp):
|
||||||
old_resp = resp
|
old_resp = resp
|
||||||
@ -2683,7 +2686,7 @@ class YoutubeDLHandler(compat_urllib_request.HTTPHandler):
|
|||||||
try:
|
try:
|
||||||
uncompressed = io.BytesIO(gz.read())
|
uncompressed = io.BytesIO(gz.read())
|
||||||
except IOError as original_ioerror:
|
except IOError as original_ioerror:
|
||||||
# There may be junk add the end of the file
|
# There may be junk at the end of the file
|
||||||
# See http://stackoverflow.com/q/4928560/35070 for details
|
# See http://stackoverflow.com/q/4928560/35070 for details
|
||||||
for i in range(1, 1024):
|
for i in range(1, 1024):
|
||||||
try:
|
try:
|
||||||
@ -2710,9 +2713,8 @@ class YoutubeDLHandler(compat_urllib_request.HTTPHandler):
|
|||||||
if location:
|
if location:
|
||||||
# As of RFC 2616 default charset is iso-8859-1 that is respected by python 3
|
# As of RFC 2616 default charset is iso-8859-1 that is respected by python 3
|
||||||
if sys.version_info >= (3, 0):
|
if sys.version_info >= (3, 0):
|
||||||
location = location.encode('iso-8859-1').decode('utf-8')
|
location = location.encode('iso-8859-1')
|
||||||
else:
|
location = location.decode('utf-8')
|
||||||
location = location.decode('utf-8')
|
|
||||||
location_escaped = escape_url(location)
|
location_escaped = escape_url(location)
|
||||||
if location != location_escaped:
|
if location != location_escaped:
|
||||||
del resp.headers['Location']
|
del resp.headers['Location']
|
||||||
@ -2940,17 +2942,16 @@ class YoutubeDLRedirectHandler(compat_urllib_request.HTTPRedirectHandler):
|
|||||||
|
|
||||||
The code is based on HTTPRedirectHandler implementation from CPython [1].
|
The code is based on HTTPRedirectHandler implementation from CPython [1].
|
||||||
|
|
||||||
This redirect handler solves two issues:
|
This redirect handler fixes and improves the logic to better align with RFC7261
|
||||||
- ensures redirect URL is always unicode under python 2
|
and what browsers tend to do [2][3]
|
||||||
- introduces support for experimental HTTP response status code
|
|
||||||
308 Permanent Redirect [2] used by some sites [3]
|
|
||||||
|
|
||||||
1. https://github.com/python/cpython/blob/master/Lib/urllib/request.py
|
1. https://github.com/python/cpython/blob/master/Lib/urllib/request.py
|
||||||
2. https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/308
|
2. https://datatracker.ietf.org/doc/html/rfc7231
|
||||||
3. https://github.com/ytdl-org/youtube-dl/issues/28768
|
3. https://github.com/python/cpython/issues/91306
|
||||||
"""
|
"""
|
||||||
|
|
||||||
http_error_301 = http_error_303 = http_error_307 = http_error_308 = compat_urllib_request.HTTPRedirectHandler.http_error_302
|
# Supply possibly missing alias
|
||||||
|
http_error_308 = compat_urllib_request.HTTPRedirectHandler.http_error_302
|
||||||
|
|
||||||
def redirect_request(self, req, fp, code, msg, headers, newurl):
|
def redirect_request(self, req, fp, code, msg, headers, newurl):
|
||||||
"""Return a Request or None in response to a redirect.
|
"""Return a Request or None in response to a redirect.
|
||||||
@ -2962,19 +2963,16 @@ class YoutubeDLRedirectHandler(compat_urllib_request.HTTPRedirectHandler):
|
|||||||
else should try to handle this url. Return None if you can't
|
else should try to handle this url. Return None if you can't
|
||||||
but another Handler might.
|
but another Handler might.
|
||||||
"""
|
"""
|
||||||
m = req.get_method()
|
if code not in (301, 302, 303, 307, 308):
|
||||||
if (not (code in (301, 302, 303, 307, 308) and m in ("GET", "HEAD")
|
raise compat_urllib_HTTPError(req.full_url, code, msg, headers, fp)
|
||||||
or code in (301, 302, 303) and m == "POST")):
|
|
||||||
raise compat_HTTPError(req.full_url, code, msg, headers, fp)
|
new_method = req.get_method()
|
||||||
# Strictly (according to RFC 2616), 301 or 302 in response to
|
new_data = req.data
|
||||||
# a POST MUST NOT cause a redirection without confirmation
|
remove_headers = []
|
||||||
# from the user (of urllib.request, in this case). In practice,
|
|
||||||
# essentially all clients do redirect in this case, so we do
|
|
||||||
# the same.
|
|
||||||
|
|
||||||
# On python 2 urlh.geturl() may sometimes return redirect URL
|
# On python 2 urlh.geturl() may sometimes return redirect URL
|
||||||
# as byte string instead of unicode. This workaround allows
|
# as a byte string instead of unicode. This workaround forces
|
||||||
# to force it always return unicode.
|
# it to return unicode.
|
||||||
if sys.version_info[0] < 3:
|
if sys.version_info[0] < 3:
|
||||||
newurl = compat_str(newurl)
|
newurl = compat_str(newurl)
|
||||||
|
|
||||||
@ -2983,13 +2981,29 @@ class YoutubeDLRedirectHandler(compat_urllib_request.HTTPRedirectHandler):
|
|||||||
# but it is kept for compatibility with other callers.
|
# but it is kept for compatibility with other callers.
|
||||||
newurl = newurl.replace(' ', '%20')
|
newurl = newurl.replace(' ', '%20')
|
||||||
|
|
||||||
CONTENT_HEADERS = ("content-length", "content-type")
|
# A 303 must either use GET or HEAD for subsequent request
|
||||||
|
# https://datatracker.ietf.org/doc/html/rfc7231#section-6.4.4
|
||||||
|
if code == 303 and req.get_method() != 'HEAD':
|
||||||
|
new_method = 'GET'
|
||||||
|
# 301 and 302 redirects are commonly turned into a GET from a POST
|
||||||
|
# for subsequent requests by browsers, so we'll do the same.
|
||||||
|
# https://datatracker.ietf.org/doc/html/rfc7231#section-6.4.2
|
||||||
|
# https://datatracker.ietf.org/doc/html/rfc7231#section-6.4.3
|
||||||
|
elif code in (301, 302) and req.get_method() == 'POST':
|
||||||
|
new_method = 'GET'
|
||||||
|
|
||||||
|
# only remove payload if method changed (e.g. POST to GET)
|
||||||
|
if new_method != req.get_method():
|
||||||
|
new_data = None
|
||||||
|
remove_headers.extend(['Content-Length', 'Content-Type'])
|
||||||
|
|
||||||
# NB: don't use dict comprehension for python 2.6 compatibility
|
# NB: don't use dict comprehension for python 2.6 compatibility
|
||||||
newheaders = dict((k, v) for k, v in req.headers.items()
|
new_headers = dict((k, v) for k, v in req.header_items()
|
||||||
if k.lower() not in CONTENT_HEADERS)
|
if k.lower() not in remove_headers)
|
||||||
|
|
||||||
return compat_urllib_request.Request(
|
return compat_urllib_request.Request(
|
||||||
newurl, headers=newheaders, origin_req_host=req.origin_req_host,
|
newurl, headers=new_headers, origin_req_host=req.origin_req_host,
|
||||||
unverifiable=True)
|
unverifiable=True, method=new_method, data=new_data)
|
||||||
|
|
||||||
|
|
||||||
def extract_timezone(date_str):
|
def extract_timezone(date_str):
|
||||||
|
Loading…
Reference in New Issue
Block a user