64
52
self.headers.get('referer', '-'),
65
53
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
55
def handle_one_request(self):
76
56
"""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.
58
You normally don't need to override this method; see the class
59
__doc__ string for information on how to handle specific HTTP
60
commands such as GET and POST.
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.
63
for i in xrange(1,11): # Don't try more than 10 times
65
self.raw_requestline = self.rfile.readline()
66
except socket.error, e:
67
if e.args[0] in (errno.EAGAIN, errno.EWOULDBLOCK):
68
# omitted for now because some tests look at the log of
69
# the server and expect to see no errors. see recent
70
# email thread. -- mbp 20051021.
71
## self.log_message('EAGAIN (%d) while reading from raw_requestline' % i)
77
if not self.raw_requestline:
87
78
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+)?$')
80
if not self.parse_request(): # An error code has been sent, just exit
82
mname = 'do_' + self.command
83
if getattr(self, mname, None) is None:
84
self.send_error(501, "Unsupported method (%r)" % self.command)
86
method = getattr(self, mname)
89
_range_regexp = re.compile(r'^(?P<start>\d+)-(?P<end>\d+)$')
132
90
_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.
92
def parse_ranges(self, ranges_header):
93
"""Parse the range header value and returns ranges and tail.
95
RFC2616 14.35 says that syntactically invalid range
96
specifiers MUST be ignored. In that case, we return 0 for
97
tail and [] for ranges.
147
101
if not ranges_header.startswith('bytes='):
148
102
# Syntactically invalid header
153
105
ranges_header = ranges_header[len('bytes='):]
154
106
for range_str in ranges_header.split(','):
107
# FIXME: RFC2616 says end is optional and default to file_size
155
108
range_match = self._range_regexp.match(range_str)
156
109
if range_match is not None:
157
110
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
111
end = int(range_match.group('end'))
165
113
# Syntactically invalid range
167
115
ranges.append((start, end))
169
117
tail_match = self._tail_regexp.match(range_str)
171
119
tail = int(tail_match.group('tail'))
173
121
# 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
125
def send_range_content(self, file, start, length):
227
142
def get_multiple_ranges(self, file, file_size, ranges):
228
143
self.send_response(206)
229
144
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)
145
boundary = "%d" % random.randint(0,0x7FFFFFFF)
146
self.send_header("Content-Type",
147
"multipart/byteranges; boundary=%s" % boundary)
246
148
self.end_headers()
248
# Send the multipart body
249
149
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))
150
self.wfile.write("--%s\r\n" % boundary)
151
self.send_header("Content-type", 'application/octet-stream')
152
self.send_header("Content-Range", "bytes %d-%d/%d" % (start,
254
155
self.end_headers()
255
156
self.send_range_content(file, start, end - start + 1)
257
self.wfile.write(boundary_line)
157
self.wfile.write("--%s\r\n" % boundary)
259
159
def do_GET(self):
260
160
"""Serve a GET request.
262
162
Handles the Range header.
265
self.server.test_case_server.GET_request_nb += 1
267
165
path = self.translate_path(self.path)
268
166
ranges_header_value = self.headers.get('Range')
269
167
if ranges_header_value is None or os.path.isdir(path):
270
168
# Let the mother class handle most cases
271
return SimpleHTTPServer.SimpleHTTPRequestHandler.do_GET(self)
169
return SimpleHTTPRequestHandler.do_GET(self)
274
172
# Always read in binary mode. Opening files in text
275
173
# mode may cause newline translations, making the
276
174
# actual size of the content transmitted *less* than
277
175
# the content-length!
176
file = open(path, 'rb')
280
178
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)
181
file_size = os.fstat(file.fileno())[6]
182
tail, ranges = self.parse_ranges(ranges_header_value)
183
# Normalize tail into ranges
185
ranges.append((file_size - tail, file_size))
187
self._satisfiable_ranges = True
189
self._satisfiable_ranges = False
191
def check_range(range_specifier):
192
start, end = range_specifier
193
# RFC2616 14.35, ranges are invalid if start >= file_size
194
if start >= file_size:
195
self._satisfiable_ranges = False # Side-effect !
197
# RFC2616 14.35, end values should be truncated
198
# to file_size -1 if they exceed it
199
end = min(end, file_size - 1)
202
ranges = map(check_range, ranges)
204
if not self._satisfiable_ranges:
286
205
# RFC2616 14.16 and 14.35 says that when a server
287
206
# encounters unsatisfiable range specifiers, it
288
207
# SHOULD return a 416.
290
209
# FIXME: We SHOULD send a Content-Range header too,
291
210
# but the implementation of send_error does not
292
211
# allows that. So far.
296
215
if len(ranges) == 1:
297
216
(start, end) = ranges[0]
298
self.get_single_range(f, file_size, start, end)
217
self.get_single_range(file, 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).
219
self.get_multiple_ranges(file, file_size, ranges)
222
if sys.platform == 'win32':
223
# On win32 you cannot access non-ascii filenames without
224
# decoding them into unicode first.
225
# However, under Linux, you can access bytestream paths
226
# without any problems. If this function was always active
227
# it would probably break tests when LANG=C was set
228
def translate_path(self, path):
229
"""Translate a /-separated PATH to the local filename syntax.
231
For bzr, all url paths are considered to be utf8 paths.
232
On Linux, you can access these paths directly over the bytestream
233
request, but on win32, you must decode them, and access them
236
# abandon query parameters
320
237
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(urllib.unquote(path))
340
path = path.decode('utf-8')
341
words = path.split('/')
342
words = filter(None, words)
344
for num, word in enumerate(words):
238
path = posixpath.normpath(urllib.unquote(path))
239
path = path.decode('utf-8')
240
words = path.split('/')
241
words = filter(None, words)
346
244
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):
245
head, word = os.path.split(word)
246
if word in (os.curdir, os.pardir): continue
247
path = os.path.join(path, word)
251
class TestingHTTPServer(BaseHTTPServer.HTTPServer):
253
def __init__(self, server_address, RequestHandlerClass,
255
BaseHTTPServer.HTTPServer.__init__(self, server_address,
356
257
# test_case_server can be used to communicate between the
357
258
# tests and the server (or the request handler and the
358
259
# server), allowing dynamic behaviors to be defined from
359
260
# the tests cases.
360
261
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):
264
class HttpServer(Server):
389
265
"""A test server for http transports.
391
267
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
270
# used to form the url that connects to this server
404
271
_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)
273
# Subclasses can provide a specific request handler
274
def __init__(self, request_handler=TestingHTTPRequestHandler):
275
Server.__init__(self)
276
self.request_handler = request_handler
429
277
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)
281
def _get_httpd(self):
282
if self._httpd is None:
283
self._httpd = TestingHTTPServer((self.host, self.port),
284
self.request_handler,
286
host, self.port = self._httpd.socket.getsockname()
289
def _http_start(self):
290
httpd = self._get_httpd()
291
self._http_base_url = '%s://%s:%s/' % (self._url_protocol,
294
self._http_starting.release()
295
httpd.socket.settimeout(0.1)
297
while self._http_running:
299
httpd.handle_request()
300
except socket.timeout:
444
303
def _get_remote_url(self, path):
445
304
path_parts = path.split(os.path.sep)
457
316
"""Capture Server log output."""
458
317
self.logs.append(format % args)
460
def start_server(self, backing_transport_server=None):
461
"""See bzrlib.transport.Server.start_server.
319
def setUp(self, backing_transport_server=None):
320
"""See bzrlib.transport.Server.setUp.
463
322
:param backing_transport_server: The transport that requests over this
464
323
protocol should be forwarded to. Note that this is currently not
465
324
supported for HTTP.
467
326
# 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)
328
assert backing_transport_server is None or \
329
isinstance(backing_transport_server, LocalURLServer), \
330
"HTTPServer currently assumes local transport, got %s" % \
331
backing_transport_server
475
332
self._home_dir = os.getcwdu()
476
333
self._local_path_parts = self._home_dir.split(os.path.sep)
334
self._http_starting = threading.Lock()
335
self._http_starting.acquire()
336
self._http_running = True
337
self._http_base_url = None
338
self._http_thread = threading.Thread(target=self._http_start)
339
self._http_thread.setDaemon(True)
340
self._http_thread.start()
341
# Wait for the server thread to start (i.e release the lock)
342
self._http_starting.acquire()
343
self._http_starting.release()
479
super(HttpServer, self).start_server()
480
self._http_base_url = '%s://%s:%s/' % (
481
self._url_protocol, self.host, self.port)
347
"""See bzrlib.transport.Server.tearDown."""
348
self._http_running = False
349
self._http_thread.join()
483
351
def get_url(self):
484
352
"""See bzrlib.transport.Server.get_url."""
added
removed
bzrlib/tests/HttpServer.py
