13
13
# You should have received a copy of the GNU General Public License
14
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
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20
from SimpleHTTPServer import SimpleHTTPRequestHandler
23
import SimpleHTTPServer
27
from bzrlib import urlutils
28
from bzrlib.tests import test_server
31
from bzrlib.transport import Server
32
from bzrlib.transport.local import LocalURLServer
35
class WebserverNotAvailable(Exception):
31
39
class BadWebserverPath(ValueError):
33
41
return 'path %s is not in %s' % self.args
36
class TestingHTTPRequestHandler(SimpleHTTPServer.SimpleHTTPRequestHandler):
44
class TestingHTTPRequestHandler(SimpleHTTPRequestHandler):
37
45
"""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.
47
A TestingHTTPRequestHandler is instantiated for every request
48
received by the associated server.
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
51
def log_message(self, format, *args):
59
52
tcs = self.server.test_case_server
79
64
connection early to avoid polluting the test results.
82
self._handle_one_request()
67
SimpleHTTPRequestHandler.handle_one_request(self)
83
68
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)):
70
and e.args[0] in (errno.EPIPE, errno.ECONNRESET,
71
errno.ECONNABORTED,)):
72
self.close_connection = 1
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+)?$')
77
_range_regexp = re.compile(r'^(?P<start>\d+)-(?P<end>\d+)$')
132
78
_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.
80
def parse_ranges(self, ranges_header):
81
"""Parse the range header value and returns ranges and tail.
83
RFC2616 14.35 says that syntactically invalid range
84
specifiers MUST be ignored. In that case, we return 0 for
85
tail and [] for ranges.
147
89
if not ranges_header.startswith('bytes='):
148
90
# Syntactically invalid header
153
93
ranges_header = ranges_header[len('bytes='):]
154
94
for range_str in ranges_header.split(','):
95
# FIXME: RFC2616 says end is optional and default to file_size
155
96
range_match = self._range_regexp.match(range_str)
156
97
if range_match is not None:
157
98
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
99
end = int(range_match.group('end'))
165
101
# Syntactically invalid range
167
103
ranges.append((start, end))
169
105
tail_match = self._tail_regexp.match(range_str)
171
107
tail = int(tail_match.group('tail'))
173
109
# 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
113
def send_range_content(self, file, start, length):
227
130
def get_multiple_ranges(self, file, file_size, ranges):
228
131
self.send_response(206)
229
132
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)
133
boundary = "%d" % random.randint(0,0x7FFFFFFF)
134
self.send_header("Content-Type",
135
"multipart/byteranges; boundary=%s" % boundary)
246
136
self.end_headers()
248
# Send the multipart body
249
137
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))
138
self.wfile.write("--%s\r\n" % boundary)
139
self.send_header("Content-type", 'application/octet-stream')
140
self.send_header("Content-Range", "bytes %d-%d/%d" % (start,
254
143
self.end_headers()
255
144
self.send_range_content(file, start, end - start + 1)
257
self.wfile.write(boundary_line)
146
self.wfile.write("--%s\r\n" % boundary)
259
148
def do_GET(self):
260
149
"""Serve a GET request.
268
157
ranges_header_value = self.headers.get('Range')
269
158
if ranges_header_value is None or os.path.isdir(path):
270
159
# Let the mother class handle most cases
271
return SimpleHTTPServer.SimpleHTTPRequestHandler.do_GET(self)
160
return SimpleHTTPRequestHandler.do_GET(self)
274
163
# Always read in binary mode. Opening files in text
275
164
# mode may cause newline translations, making the
276
165
# actual size of the content transmitted *less* than
277
166
# the content-length!
167
file = open(path, 'rb')
280
169
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)
172
file_size = os.fstat(file.fileno())[6]
173
tail, ranges = self.parse_ranges(ranges_header_value)
174
# Normalize tail into ranges
176
ranges.append((file_size - tail, file_size))
178
self._satisfiable_ranges = True
180
self._satisfiable_ranges = False
182
def check_range(range_specifier):
183
start, end = range_specifier
184
# RFC2616 14.35, ranges are invalid if start >= file_size
185
if start >= file_size:
186
self._satisfiable_ranges = False # Side-effect !
188
# RFC2616 14.35, end values should be truncated
189
# to file_size -1 if they exceed it
190
end = min(end, file_size - 1)
193
ranges = map(check_range, ranges)
195
if not self._satisfiable_ranges:
286
196
# RFC2616 14.16 and 14.35 says that when a server
287
197
# encounters unsatisfiable range specifiers, it
288
198
# SHOULD return a 416.
290
200
# FIXME: We SHOULD send a Content-Range header too,
291
201
# but the implementation of send_error does not
292
202
# allows that. So far.
324
234
return self._translate_path(path)
326
236
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):
237
return SimpleHTTPRequestHandler.translate_path(self, path)
239
if sys.platform == 'win32':
240
# On win32 you cannot access non-ascii filenames without
241
# decoding them into unicode first.
242
# However, under Linux, you can access bytestream paths
243
# without any problems. If this function was always active
244
# it would probably break tests when LANG=C was set
245
def _translate_path(self, path):
246
"""Translate a /-separated PATH to the local filename syntax.
248
For bzr, all url paths are considered to be utf8 paths.
249
On Linux, you can access these paths directly over the bytestream
250
request, but on win32, you must decode them, and access them
253
# abandon query parameters
254
path = urlparse.urlparse(path)[2]
255
path = posixpath.normpath(urllib.unquote(path))
256
path = path.decode('utf-8')
257
words = path.split('/')
258
words = filter(None, words)
346
261
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):
262
head, word = os.path.split(word)
263
if word in (os.curdir, os.pardir): continue
264
path = os.path.join(path, word)
268
class TestingHTTPServer(BaseHTTPServer.HTTPServer):
270
def __init__(self, server_address, RequestHandlerClass,
272
BaseHTTPServer.HTTPServer.__init__(self, server_address,
356
274
# test_case_server can be used to communicate between the
357
275
# tests and the server (or the request handler and the
358
276
# server), allowing dynamic behaviors to be defined from
359
277
# the tests cases.
360
278
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):
280
def server_close(self):
281
"""Called to clean-up the server.
283
Since the server may be in a blocking read, we shutdown the socket
286
self.socket.shutdown(socket.SHUT_RDWR)
287
BaseHTTPServer.HTTPServer.server_close(self)
290
class HttpServer(Server):
389
291
"""A test server for http transports.
391
293
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
296
# Whether or not we proxy the requests (see
400
297
# TestingHTTPRequestHandler.translate_path).
401
298
proxy_requests = False
403
300
# used to form the url that connects to this server
404
301
_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)
303
# Subclasses can provide a specific request handler
304
def __init__(self, request_handler=TestingHTTPRequestHandler):
305
Server.__init__(self)
306
self.request_handler = request_handler
429
307
self.host = 'localhost'
431
super(HttpServer, self).__init__((self.host, self.port),
434
self.protocol_version = proto_vers
435
310
# Allows tests to verify number of GET requests issued
436
311
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)
313
def _get_httpd(self):
314
if self._httpd is None:
315
self._httpd = TestingHTTPServer((self.host, self.port),
316
self.request_handler,
318
host, self.port = self._httpd.socket.getsockname()
321
def _http_start(self):
322
httpd = self._get_httpd()
323
self._http_base_url = '%s://%s:%s/' % (self._url_protocol,
326
self._http_starting.release()
328
while self._http_running:
330
httpd.handle_request()
331
except socket.timeout:
444
334
def _get_remote_url(self, path):
445
335
path_parts = path.split(os.path.sep)
457
347
"""Capture Server log output."""
458
348
self.logs.append(format % args)
460
def start_server(self, backing_transport_server=None):
461
"""See bzrlib.transport.Server.start_server.
350
def setUp(self, backing_transport_server=None):
351
"""See bzrlib.transport.Server.setUp.
463
353
:param backing_transport_server: The transport that requests over this
464
354
protocol should be forwarded to. Note that this is currently not
465
355
supported for HTTP.
467
357
# 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)
359
assert backing_transport_server is None or \
360
isinstance(backing_transport_server, LocalURLServer), \
361
"HTTPServer currently assumes local transport, got %s" % \
362
backing_transport_server
475
363
self._home_dir = os.getcwdu()
476
364
self._local_path_parts = self._home_dir.split(os.path.sep)
365
self._http_starting = threading.Lock()
366
self._http_starting.acquire()
367
self._http_running = True
368
self._http_base_url = None
369
self._http_thread = threading.Thread(target=self._http_start)
370
self._http_thread.setDaemon(True)
371
self._http_thread.start()
372
# Wait for the server thread to start (i.e release the lock)
373
self._http_starting.acquire()
374
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)
378
"""See bzrlib.transport.Server.tearDown."""
379
self._httpd.server_close()
380
self._http_running = False
381
self._http_thread.join()
483
383
def get_url(self):
484
384
"""See bzrlib.transport.Server.get_url."""
added
removed
bzrlib/tests/HttpServer.py
