41
33
return 'path %s is not in %s' % self.args
44
class TestingHTTPRequestHandler(SimpleHTTPRequestHandler):
36
class TestingHTTPRequestHandler(SimpleHTTPServer.SimpleHTTPRequestHandler):
45
37
"""Handles one request.
47
A TestingHTTPRequestHandler is instantiated for every request
48
received by the associated server.
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
51
58
def log_message(self, format, *args):
52
59
tcs = self.server.test_case_server
57
64
self.headers.get('referer', '-'),
58
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()
60
75
def handle_one_request(self):
61
76
"""Handle a single HTTP request.
63
You normally don't need to override this method; see the class
64
__doc__ string for information on how to handle specific HTTP
65
commands such as GET and POST.
78
We catch all socket errors occurring when the client close the
79
connection early to avoid polluting the test results.
68
for i in xrange(1,11): # Don't try more than 10 times
70
self.raw_requestline = self.rfile.readline()
71
except socket.error, e:
72
if e.args[0] in (errno.EAGAIN, errno.EWOULDBLOCK):
73
# omitted for now because some tests look at the log of
74
# the server and expect to see no errors. see recent
75
# email thread. -- mbp 20051021.
76
## self.log_message('EAGAIN (%d) while reading from raw_requestline' % i)
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)):
82
if not self.raw_requestline:
83
self.close_connection = 1
85
if not self.parse_request(): # An error code has been sent, just exit
87
mname = 'do_' + self.command
88
if getattr(self, mname, None) is None:
89
self.send_error(501, "Unsupported method (%r)" % self.command)
91
method = getattr(self, mname)
94
_range_regexp = re.compile(r'^(?P<start>\d+)-(?P<end>\d+)$')
93
error_content_type = 'text/plain'
94
error_message_format = '''\
99
def send_error(self, code, message=None):
100
"""Send and log an error reply.
102
We redefine the python-provided version to be able to set a
103
``Content-Length`` header as some http/1.1 clients complain otherwise
106
:param code: The HTTP error code.
108
:param message: The explanation of the error code, Defaults to a short
114
message = self.responses[code][0]
117
self.log_error("code %d, message %s", code, message)
118
content = (self.error_message_format %
119
{'code': code, 'message': message})
120
self.send_response(code, message)
121
self.send_header("Content-Type", self.error_content_type)
122
self.send_header("Content-Length", "%d" % len(content))
123
self.send_header('Connection', 'close')
125
if self.command != 'HEAD' and code >= 200 and code not in (204, 304):
126
self.wfile.write(content)
128
def _handle_one_request(self):
129
SimpleHTTPServer.SimpleHTTPRequestHandler.handle_one_request(self)
131
_range_regexp = re.compile(r'^(?P<start>\d+)-(?P<end>\d+)?$')
95
132
_tail_regexp = re.compile(r'^-(?P<tail>\d+)$')
97
def parse_ranges(self, ranges_header):
98
"""Parse the range header value and returns ranges and tail.
100
RFC2616 14.35 says that syntactically invalid range
101
specifiers MUST be ignored. In that case, we return 0 for
102
tail and [] for ranges.
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.
106
147
if not ranges_header.startswith('bytes='):
107
148
# Syntactically invalid header
110
153
ranges_header = ranges_header[len('bytes='):]
111
154
for range_str in ranges_header.split(','):
112
# FIXME: RFC2616 says end is optional and default to file_size
113
155
range_match = self._range_regexp.match(range_str)
114
156
if range_match is not None:
115
157
start = int(range_match.group('start'))
116
end = int(range_match.group('end'))
158
end_match = range_match.group('end')
159
if end_match is None:
160
# RFC2616 says end is optional and default to file_size
118
165
# Syntactically invalid range
120
167
ranges.append((start, end))
122
169
tail_match = self._tail_regexp.match(range_str)
124
171
tail = int(tail_match.group('tail'))
126
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)
130
210
def send_range_content(self, file, start, length):
147
227
def get_multiple_ranges(self, file, file_size, ranges):
148
228
self.send_response(206)
149
229
self.send_header('Accept-Ranges', 'bytes')
150
boundary = "%d" % random.randint(0,0x7FFFFFFF)
151
self.send_header("Content-Type",
152
"multipart/byteranges; boundary=%s" % boundary)
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)
153
246
self.end_headers()
248
# Send the multipart body
154
249
for (start, end) in ranges:
155
self.wfile.write("--%s\r\n" % boundary)
156
self.send_header("Content-type", 'application/octet-stream')
157
self.send_header("Content-Range", "bytes %d-%d/%d" % (start,
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))
160
254
self.end_headers()
161
255
self.send_range_content(file, start, end - start + 1)
162
self.wfile.write("--%s\r\n" % boundary)
257
self.wfile.write(boundary_line)
164
259
def do_GET(self):
165
260
"""Serve a GET request.
167
262
Handles the Range header.
265
self.server.test_case_server.GET_request_nb += 1
170
267
path = self.translate_path(self.path)
171
268
ranges_header_value = self.headers.get('Range')
172
269
if ranges_header_value is None or os.path.isdir(path):
173
270
# Let the mother class handle most cases
174
return SimpleHTTPRequestHandler.do_GET(self)
271
return SimpleHTTPServer.SimpleHTTPRequestHandler.do_GET(self)
177
274
# Always read in binary mode. Opening files in text
178
275
# mode may cause newline translations, making the
179
276
# actual size of the content transmitted *less* than
180
277
# the content-length!
181
file = open(path, 'rb')
183
280
self.send_error(404, "File not found")
186
file_size = os.fstat(file.fileno())[6]
187
tail, ranges = self.parse_ranges(ranges_header_value)
188
# Normalize tail into ranges
190
ranges.append((file_size - tail, file_size))
192
self._satisfiable_ranges = True
194
self._satisfiable_ranges = False
196
def check_range(range_specifier):
197
start, end = range_specifier
198
# RFC2616 14.35, ranges are invalid if start >= file_size
199
if start >= file_size:
200
self._satisfiable_ranges = False # Side-effect !
202
# RFC2616 14.35, end values should be truncated
203
# to file_size -1 if they exceed it
204
end = min(end, file_size - 1)
207
ranges = map(check_range, ranges)
209
if not self._satisfiable_ranges:
283
file_size = os.fstat(f.fileno())[6]
284
ranges = self._parse_ranges(ranges_header_value, file_size)
210
286
# RFC2616 14.16 and 14.35 says that when a server
211
287
# encounters unsatisfiable range specifiers, it
212
288
# SHOULD return a 416.
214
290
# FIXME: We SHOULD send a Content-Range header too,
215
291
# but the implementation of send_error does not
216
292
# allows that. So far.
248
324
return self._translate_path(path)
250
326
def _translate_path(self, path):
251
return SimpleHTTPRequestHandler.translate_path(self, path)
253
if sys.platform == 'win32':
254
# On win32 you cannot access non-ascii filenames without
255
# decoding them into unicode first.
256
# However, under Linux, you can access bytestream paths
257
# without any problems. If this function was always active
258
# it would probably break tests when LANG=C was set
259
def _translate_path(self, path):
260
"""Translate a /-separated PATH to the local filename syntax.
262
For bzr, all url paths are considered to be utf8 paths.
263
On Linux, you can access these paths directly over the bytestream
264
request, but on win32, you must decode them, and access them
267
# abandon query parameters
268
path = urlparse.urlparse(path)[2]
269
path = posixpath.normpath(urllib.unquote(path))
270
path = path.decode('utf-8')
271
words = path.split('/')
272
words = filter(None, words)
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):
275
346
drive, word = os.path.splitdrive(word)
276
head, word = os.path.split(word)
277
if word in (os.curdir, os.pardir): continue
278
path = os.path.join(path, word)
282
class TestingHTTPServer(BaseHTTPServer.HTTPServer):
284
def __init__(self, server_address, RequestHandlerClass,
286
BaseHTTPServer.HTTPServer.__init__(self, server_address,
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):
288
356
# test_case_server can be used to communicate between the
289
357
# tests and the server (or the request handler and the
290
358
# server), allowing dynamic behaviors to be defined from
291
359
# the tests cases.
292
360
self.test_case_server = test_case_server
295
class HttpServer(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):
296
389
"""A test server for http transports.
298
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,
301
399
# Whether or not we proxy the requests (see
302
400
# TestingHTTPRequestHandler.translate_path).
303
401
proxy_requests = False
305
403
# used to form the url that connects to this server
306
404
_url_protocol = 'http'
308
# Subclasses can provide a specific request handler
309
def __init__(self, request_handler=TestingHTTPRequestHandler):
310
Server.__init__(self)
311
self.request_handler = request_handler
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)
312
429
self.host = 'localhost'
316
def _get_httpd(self):
317
if self._httpd is None:
318
self._httpd = TestingHTTPServer((self.host, self.port),
319
self.request_handler,
321
host, self.port = self._httpd.socket.getsockname()
324
def _http_start(self):
325
httpd = self._get_httpd()
326
self._http_base_url = '%s://%s:%s/' % (self._url_protocol,
329
self._http_starting.release()
330
httpd.socket.settimeout(0.1)
332
while self._http_running:
334
httpd.handle_request()
335
except socket.timeout:
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)
338
444
def _get_remote_url(self, path):
339
445
path_parts = path.split(os.path.sep)
351
457
"""Capture Server log output."""
352
458
self.logs.append(format % args)
354
def setUp(self, backing_transport_server=None):
355
"""See bzrlib.transport.Server.setUp.
460
def start_server(self, backing_transport_server=None):
461
"""See bzrlib.transport.Server.start_server.
357
463
:param backing_transport_server: The transport that requests over this
358
464
protocol should be forwarded to. Note that this is currently not
359
465
supported for HTTP.
361
467
# XXX: TODO: make the server back onto vfs_server rather than local
363
assert backing_transport_server is None or \
364
isinstance(backing_transport_server, LocalURLServer), \
365
"HTTPServer currently assumes local transport, got %s" % \
366
backing_transport_server
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)
367
475
self._home_dir = os.getcwdu()
368
476
self._local_path_parts = self._home_dir.split(os.path.sep)
369
self._http_starting = threading.Lock()
370
self._http_starting.acquire()
371
self._http_running = True
372
self._http_base_url = None
373
self._http_thread = threading.Thread(target=self._http_start)
374
self._http_thread.setDaemon(True)
375
self._http_thread.start()
376
# Wait for the server thread to start (i.e release the lock)
377
self._http_starting.acquire()
378
self._http_starting.release()
382
"""See bzrlib.transport.Server.tearDown."""
383
self._http_running = False
384
self._http_thread.join()
479
super(HttpServer, self).start_server()
480
self._http_base_url = '%s://%s:%s/' % (
481
self._url_protocol, self.host, self.port)
386
483
def get_url(self):
387
484
"""See bzrlib.transport.Server.get_url."""