1
# Copyright (C) 2006-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23
import SimpleHTTPServer
27
from bzrlib import urlutils
28
from bzrlib.tests import test_server
31
class BadWebserverPath(ValueError):
33
return 'path %s is not in %s' % self.args
36
class TestingHTTPRequestHandler(SimpleHTTPServer.SimpleHTTPRequestHandler):
37
"""Handles one request.
39
A TestingHTTPRequestHandler is instantiated for every request received by
40
the associated server. Note that 'request' here is inherited from the base
41
TCPServer class, for the HTTP server it is really a connection which itself
42
will handle one or several HTTP requests.
44
# Default protocol version
45
protocol_version = 'HTTP/1.1'
47
# The Message-like class used to parse the request headers
48
MessageClass = httplib.HTTPMessage
51
SimpleHTTPServer.SimpleHTTPRequestHandler.setup(self)
52
self._cwd = self.server._home_dir
53
tcs = self.server.test_case_server
54
if tcs.protocol_version is not None:
55
# If the test server forced a protocol version, use it
56
self.protocol_version = tcs.protocol_version
58
def log_message(self, format, *args):
59
tcs = self.server.test_case_server
60
tcs.log('webserver - %s - - [%s] %s "%s" "%s"',
61
self.address_string(),
62
self.log_date_time_string(),
64
self.headers.get('referer', '-'),
65
self.headers.get('user-agent', '-'))
68
SimpleHTTPServer.SimpleHTTPRequestHandler.handle(self)
69
# Some client (pycurl, I'm looking at you) are more picky than others
70
# and require that the socket itself is closed
71
# (SocketServer.StreamRequestHandler only close the two associated
73
self.connection.close()
75
def handle_one_request(self):
76
"""Handle a single HTTP request.
78
We catch all socket errors occurring when the client close the
79
connection early to avoid polluting the test results.
82
self._handle_one_request()
83
except socket.error, e:
84
# Any socket error should close the connection, but some errors are
85
# due to the client closing early and we don't want to pollute test
86
# results, so we raise only the others.
87
self.close_connection = 1
89
or e.args[0] not in (errno.EPIPE, errno.ECONNRESET,
90
errno.ECONNABORTED, errno.EBADF)):
93
error_content_type = 'text/plain'
94
error_message_format = '''\
99
def send_error(self, code, message=None):
100
"""Send and log an error reply.
102
We redefine the python-provided version to be able to set a
103
``Content-Length`` header as some http/1.1 clients complain otherwise
106
:param code: The HTTP error code.
108
:param message: The explanation of the error code, Defaults to a short
114
message = self.responses[code][0]
117
self.log_error("code %d, message %s", code, message)
118
content = (self.error_message_format %
119
{'code': code, 'message': message})
120
self.send_response(code, message)
121
self.send_header("Content-Type", self.error_content_type)
122
self.send_header("Content-Length", "%d" % len(content))
123
self.send_header('Connection', 'close')
125
if self.command != 'HEAD' and code >= 200 and code not in (204, 304):
126
self.wfile.write(content)
128
def _handle_one_request(self):
129
SimpleHTTPServer.SimpleHTTPRequestHandler.handle_one_request(self)
131
_range_regexp = re.compile(r'^(?P<start>\d+)-(?P<end>\d+)?$')
132
_tail_regexp = re.compile(r'^-(?P<tail>\d+)$')
134
def _parse_ranges(self, ranges_header, file_size):
135
"""Parse the range header value and returns ranges.
137
RFC2616 14.35 says that syntactically invalid range specifiers MUST be
138
ignored. In that case, we return None instead of a range list.
140
:param ranges_header: The 'Range' header value.
142
:param file_size: The size of the requested file.
144
:return: A list of (start, end) tuples or None if some invalid range
145
specifier is encountered.
147
if not ranges_header.startswith('bytes='):
148
# Syntactically invalid header
153
ranges_header = ranges_header[len('bytes='):]
154
for range_str in ranges_header.split(','):
155
range_match = self._range_regexp.match(range_str)
156
if range_match is not None:
157
start = int(range_match.group('start'))
158
end_match = range_match.group('end')
159
if end_match is None:
160
# RFC2616 says end is optional and default to file_size
165
# Syntactically invalid range
167
ranges.append((start, end))
169
tail_match = self._tail_regexp.match(range_str)
170
if tail_match is not None:
171
tail = int(tail_match.group('tail'))
173
# Syntactically invalid range
176
# Normalize tail into ranges
177
ranges.append((max(0, file_size - tail), file_size))
180
for start, end in ranges:
181
if start >= file_size:
182
# RFC2616 14.35, ranges are invalid if start >= file_size
184
# RFC2616 14.35, end values should be truncated
185
# to file_size -1 if they exceed it
186
end = min(end, file_size - 1)
187
checked_ranges.append((start, end))
188
return checked_ranges
190
def _header_line_length(self, keyword, value):
191
header_line = '%s: %s\r\n' % (keyword, value)
192
return len(header_line)
195
"""Overrides base implementation to work around a bug in python2.5."""
196
path = self.translate_path(self.path)
197
if os.path.isdir(path) and not self.path.endswith('/'):
198
# redirect browser - doing basically what apache does when
199
# DirectorySlash option is On which is quite common (braindead, but
201
self.send_response(301)
202
self.send_header("Location", self.path + "/")
203
# Indicates that the body is empty for HTTP/1.1 clients
204
self.send_header('Content-Length', '0')
208
return SimpleHTTPServer.SimpleHTTPRequestHandler.send_head(self)
210
def send_range_content(self, file, start, length):
212
self.wfile.write(file.read(length))
214
def get_single_range(self, file, file_size, start, end):
215
self.send_response(206)
216
length = end - start + 1
217
self.send_header('Accept-Ranges', 'bytes')
218
self.send_header("Content-Length", "%d" % length)
220
self.send_header("Content-Type", 'application/octet-stream')
221
self.send_header("Content-Range", "bytes %d-%d/%d" % (start,
225
self.send_range_content(file, start, length)
227
def get_multiple_ranges(self, file, file_size, ranges):
228
self.send_response(206)
229
self.send_header('Accept-Ranges', 'bytes')
230
boundary = '%d' % random.randint(0,0x7FFFFFFF)
231
self.send_header('Content-Type',
232
'multipart/byteranges; boundary=%s' % boundary)
233
boundary_line = '--%s\r\n' % boundary
234
# Calculate the Content-Length
236
for (start, end) in ranges:
237
content_length += len(boundary_line)
238
content_length += self._header_line_length(
239
'Content-type', 'application/octet-stream')
240
content_length += self._header_line_length(
241
'Content-Range', 'bytes %d-%d/%d' % (start, end, file_size))
242
content_length += len('\r\n') # end headers
243
content_length += end - start + 1
244
content_length += len(boundary_line)
245
self.send_header('Content-length', content_length)
248
# Send the multipart body
249
for (start, end) in ranges:
250
self.wfile.write(boundary_line)
251
self.send_header('Content-type', 'application/octet-stream')
252
self.send_header('Content-Range', 'bytes %d-%d/%d'
253
% (start, end, file_size))
255
self.send_range_content(file, start, end - start + 1)
257
self.wfile.write(boundary_line)
260
"""Serve a GET request.
262
Handles the Range header.
265
self.server.test_case_server.GET_request_nb += 1
267
path = self.translate_path(self.path)
268
ranges_header_value = self.headers.get('Range')
269
if ranges_header_value is None or os.path.isdir(path):
270
# Let the mother class handle most cases
271
return SimpleHTTPServer.SimpleHTTPRequestHandler.do_GET(self)
274
# Always read in binary mode. Opening files in text
275
# mode may cause newline translations, making the
276
# actual size of the content transmitted *less* than
277
# the content-length!
280
self.send_error(404, "File not found")
283
file_size = os.fstat(f.fileno())[6]
284
ranges = self._parse_ranges(ranges_header_value, file_size)
286
# RFC2616 14.16 and 14.35 says that when a server
287
# encounters unsatisfiable range specifiers, it
288
# SHOULD return a 416.
290
# FIXME: We SHOULD send a Content-Range header too,
291
# but the implementation of send_error does not
292
# allows that. So far.
293
self.send_error(416, "Requested range not satisfiable")
297
(start, end) = ranges[0]
298
self.get_single_range(f, file_size, start, end)
300
self.get_multiple_ranges(f, file_size, ranges)
303
def translate_path(self, path):
304
"""Translate a /-separated PATH to the local filename syntax.
306
If the server requires it, proxy the path before the usual translation
308
if self.server.test_case_server.proxy_requests:
309
# We need to act as a proxy and accept absolute urls,
310
# which SimpleHTTPRequestHandler (parent) is not
311
# ready for. So we just drop the protocol://host:port
312
# part in front of the request-url (because we know
313
# we would not forward the request to *another*
316
# So we do what SimpleHTTPRequestHandler.translate_path
317
# do beginning with python 2.4.3: abandon query
318
# parameters, scheme, host port, etc (which ensure we
319
# provide the right behaviour on all python versions).
320
path = urlparse.urlparse(path)[2]
321
# And now, we can apply *our* trick to proxy files
324
return self._translate_path(path)
326
def _translate_path(self, path):
327
"""Translate a /-separated PATH to the local filename syntax.
329
Note that we're translating http URLs here, not file URLs.
330
The URL root location is the server's startup directory.
331
Components that mean special things to the local file system
332
(e.g. drive or directory names) are ignored. (XXX They should
333
probably be diagnosed.)
335
Override from python standard library to stop it calling os.getcwd()
337
# abandon query parameters
338
path = urlparse.urlparse(path)[2]
339
path = posixpath.normpath(urlutils.unquote(path))
340
path = path.decode('utf-8')
341
words = path.split('/')
342
words = filter(None, words)
344
for num, word in enumerate(words):
346
drive, word = os.path.splitdrive(word)
347
head, word = os.path.split(word)
348
if word in (os.curdir, os.pardir): continue
349
path = os.path.join(path, word)
353
class TestingHTTPServerMixin:
355
def __init__(self, test_case_server):
356
# test_case_server can be used to communicate between the
357
# tests and the server (or the request handler and the
358
# server), allowing dynamic behaviors to be defined from
360
self.test_case_server = test_case_server
361
self._home_dir = test_case_server._home_dir
364
class TestingHTTPServer(test_server.TestingTCPServer, TestingHTTPServerMixin):
366
def __init__(self, server_address, request_handler_class,
368
test_server.TestingTCPServer.__init__(self, server_address,
369
request_handler_class)
370
TestingHTTPServerMixin.__init__(self, test_case_server)
373
class TestingThreadingHTTPServer(test_server.TestingThreadingTCPServer,
374
TestingHTTPServerMixin):
375
"""A threading HTTP test server for HTTP 1.1.
377
Since tests can initiate several concurrent connections to the same http
378
server, we need an independent connection for each of them. We achieve that
379
by spawning a new thread for each connection.
381
def __init__(self, server_address, request_handler_class,
383
test_server.TestingThreadingTCPServer.__init__(self, server_address,
384
request_handler_class)
385
TestingHTTPServerMixin.__init__(self, test_case_server)
388
class HttpServer(test_server.TestingTCPServerInAThread):
389
"""A test server for http transports.
391
Subclasses can provide a specific request handler.
394
# The real servers depending on the protocol
395
http_server_class = {'HTTP/1.0': TestingHTTPServer,
396
'HTTP/1.1': TestingThreadingHTTPServer,
399
# Whether or not we proxy the requests (see
400
# TestingHTTPRequestHandler.translate_path).
401
proxy_requests = False
403
# used to form the url that connects to this server
404
_url_protocol = 'http'
406
def __init__(self, request_handler=TestingHTTPRequestHandler,
407
protocol_version=None):
410
:param request_handler: a class that will be instantiated to handle an
411
http connection (one or several requests).
413
:param protocol_version: if specified, will override the protocol
414
version of the request handler.
416
# Depending on the protocol version, we will create the approriate
418
if protocol_version is None:
419
# Use the request handler one
420
proto_vers = request_handler.protocol_version
422
# Use our own, it will be used to override the request handler
424
proto_vers = protocol_version
425
# Get the appropriate server class for the required protocol
426
serv_cls = self.http_server_class.get(proto_vers, None)
428
raise httplib.UnknownProtocol(proto_vers)
429
self.host = 'localhost'
431
super(HttpServer, self).__init__((self.host, self.port),
434
self.protocol_version = proto_vers
435
# Allows tests to verify number of GET requests issued
436
self.GET_request_nb = 0
437
self._http_base_url = None
440
def create_server(self):
441
return self.server_class(
442
(self.host, self.port), self.request_handler_class, self)
444
def _get_remote_url(self, path):
445
path_parts = path.split(os.path.sep)
446
if os.path.isabs(path):
447
if path_parts[:len(self._local_path_parts)] != \
448
self._local_path_parts:
449
raise BadWebserverPath(path, self.test_dir)
450
remote_path = '/'.join(path_parts[len(self._local_path_parts):])
452
remote_path = '/'.join(path_parts)
454
return self._http_base_url + remote_path
456
def log(self, format, *args):
457
"""Capture Server log output."""
458
self.logs.append(format % args)
460
def start_server(self, backing_transport_server=None):
461
"""See bzrlib.transport.Server.start_server.
463
:param backing_transport_server: The transport that requests over this
464
protocol should be forwarded to. Note that this is currently not
467
# XXX: TODO: make the server back onto vfs_server rather than local
469
if not (backing_transport_server is None
470
or isinstance(backing_transport_server,
471
test_server.LocalURLServer)):
472
raise AssertionError(
473
"HTTPServer currently assumes local transport, got %s" %
474
backing_transport_server)
475
self._home_dir = os.getcwdu()
476
self._local_path_parts = self._home_dir.split(os.path.sep)
479
super(HttpServer, self).start_server()
480
self._http_base_url = '%s://%s:%s/' % (
481
self._url_protocol, self.host, self.port)
484
"""See bzrlib.transport.Server.get_url."""
485
return self._get_remote_url(self._home_dir)
487
def get_bogus_url(self):
488
"""See bzrlib.transport.Server.get_bogus_url."""
489
# this is chosen to try to prevent trouble with proxies, weird dns,
491
return self._url_protocol + '://127.0.0.1:1/'
494
class HttpServer_urllib(HttpServer):
495
"""Subclass of HttpServer that gives http+urllib urls.
497
This is for use in testing: connections to this server will always go
498
through urllib where possible.
501
# urls returned by this server should require the urllib client impl
502
_url_protocol = 'http+urllib'
505
class HttpServer_PyCurl(HttpServer):
506
"""Subclass of HttpServer that gives http+pycurl urls.
508
This is for use in testing: connections to this server will always go
509
through pycurl where possible.
512
# We don't care about checking the pycurl availability as
513
# this server will be required only when pycurl is present
515
# urls returned by this server should require the pycurl client impl
516
_url_protocol = 'http+pycurl'
added
removed
bzrlib/tests/http_server.py
