Show More
@@ -1,574 +1,575 b'' | |||
|
1 | 1 | # hgweb/request.py - An http request from either CGI or the standalone server. |
|
2 | 2 | # |
|
3 | 3 | # Copyright 21 May 2005 - (c) 2005 Jake Edge <jake@edge2.net> |
|
4 | 4 | # Copyright 2005, 2006 Matt Mackall <mpm@selenic.com> |
|
5 | 5 | # |
|
6 | 6 | # This software may be used and distributed according to the terms of the |
|
7 | 7 | # GNU General Public License version 2 or any later version. |
|
8 | 8 | |
|
9 | 9 | from __future__ import absolute_import |
|
10 | 10 | |
|
11 | 11 | #import wsgiref.validate |
|
12 | 12 | |
|
13 | 13 | from ..thirdparty import ( |
|
14 | 14 | attr, |
|
15 | 15 | ) |
|
16 | 16 | from .. import ( |
|
17 | 17 | error, |
|
18 | 18 | pycompat, |
|
19 | 19 | util, |
|
20 | 20 | ) |
|
21 | 21 | |
|
22 | 22 | class multidict(object): |
|
23 | 23 | """A dict like object that can store multiple values for a key. |
|
24 | 24 | |
|
25 | 25 | Used to store parsed request parameters. |
|
26 | 26 | |
|
27 | 27 | This is inspired by WebOb's class of the same name. |
|
28 | 28 | """ |
|
29 | 29 | def __init__(self): |
|
30 | 30 | self._items = {} |
|
31 | 31 | |
|
32 | 32 | def __getitem__(self, key): |
|
33 | 33 | """Returns the last set value for a key.""" |
|
34 | 34 | return self._items[key][-1] |
|
35 | 35 | |
|
36 | 36 | def __setitem__(self, key, value): |
|
37 | 37 | """Replace a values for a key with a new value.""" |
|
38 | 38 | self._items[key] = [value] |
|
39 | 39 | |
|
40 | 40 | def __delitem__(self, key): |
|
41 | 41 | """Delete all values for a key.""" |
|
42 | 42 | del self._items[key] |
|
43 | 43 | |
|
44 | 44 | def __contains__(self, key): |
|
45 | 45 | return key in self._items |
|
46 | 46 | |
|
47 | 47 | def __len__(self): |
|
48 | 48 | return len(self._items) |
|
49 | 49 | |
|
50 | 50 | def get(self, key, default=None): |
|
51 | 51 | try: |
|
52 | 52 | return self.__getitem__(key) |
|
53 | 53 | except KeyError: |
|
54 | 54 | return default |
|
55 | 55 | |
|
56 | 56 | def add(self, key, value): |
|
57 | 57 | """Add a new value for a key. Does not replace existing values.""" |
|
58 | 58 | self._items.setdefault(key, []).append(value) |
|
59 | 59 | |
|
60 | 60 | def getall(self, key): |
|
61 | 61 | """Obtains all values for a key.""" |
|
62 | 62 | return self._items.get(key, []) |
|
63 | 63 | |
|
64 | 64 | def getone(self, key): |
|
65 | 65 | """Obtain a single value for a key. |
|
66 | 66 | |
|
67 | 67 | Raises KeyError if key not defined or it has multiple values set. |
|
68 | 68 | """ |
|
69 | 69 | vals = self._items[key] |
|
70 | 70 | |
|
71 | 71 | if len(vals) > 1: |
|
72 | 72 | raise KeyError('multiple values for %r' % key) |
|
73 | 73 | |
|
74 | 74 | return vals[0] |
|
75 | 75 | |
|
76 | 76 | def asdictoflists(self): |
|
77 | 77 | return {k: list(v) for k, v in self._items.iteritems()} |
|
78 | 78 | |
|
79 | 79 | @attr.s(frozen=True) |
|
80 | 80 | class parsedrequest(object): |
|
81 | 81 | """Represents a parsed WSGI request. |
|
82 | 82 | |
|
83 | 83 | Contains both parsed parameters as well as a handle on the input stream. |
|
84 | 84 | """ |
|
85 | 85 | |
|
86 | 86 | # Request method. |
|
87 | 87 | method = attr.ib() |
|
88 | 88 | # Full URL for this request. |
|
89 | 89 | url = attr.ib() |
|
90 | 90 | # URL without any path components. Just <proto>://<host><port>. |
|
91 | 91 | baseurl = attr.ib() |
|
92 | 92 | # Advertised URL. Like ``url`` and ``baseurl`` but uses SERVER_NAME instead |
|
93 | 93 | # of HTTP: Host header for hostname. This is likely what clients used. |
|
94 | 94 | advertisedurl = attr.ib() |
|
95 | 95 | advertisedbaseurl = attr.ib() |
|
96 | 96 | # URL scheme (part before ``://``). e.g. ``http`` or ``https``. |
|
97 | 97 | urlscheme = attr.ib() |
|
98 | 98 | # Value of REMOTE_USER, if set, or None. |
|
99 | 99 | remoteuser = attr.ib() |
|
100 | 100 | # Value of REMOTE_HOST, if set, or None. |
|
101 | 101 | remotehost = attr.ib() |
|
102 | 102 | # Relative WSGI application path. If defined, will begin with a |
|
103 | 103 | # ``/``. |
|
104 | 104 | apppath = attr.ib() |
|
105 | 105 | # List of path parts to be used for dispatch. |
|
106 | 106 | dispatchparts = attr.ib() |
|
107 | 107 | # URL path component (no query string) used for dispatch. Can be |
|
108 | 108 | # ``None`` to signal no path component given to the request, an |
|
109 | 109 | # empty string to signal a request to the application's root URL, |
|
110 | 110 | # or a string not beginning with ``/`` containing the requested |
|
111 | 111 | # path under the application. |
|
112 | 112 | dispatchpath = attr.ib() |
|
113 | 113 | # The name of the repository being accessed. |
|
114 | 114 | reponame = attr.ib() |
|
115 | 115 | # Raw query string (part after "?" in URL). |
|
116 | 116 | querystring = attr.ib() |
|
117 | 117 | # multidict of query string parameters. |
|
118 | 118 | qsparams = attr.ib() |
|
119 | 119 | # wsgiref.headers.Headers instance. Operates like a dict with case |
|
120 | 120 | # insensitive keys. |
|
121 | 121 | headers = attr.ib() |
|
122 | 122 | # Request body input stream. |
|
123 | 123 | bodyfh = attr.ib() |
|
124 | 124 | # WSGI environment dict, unmodified. |
|
125 | 125 | rawenv = attr.ib() |
|
126 | 126 | |
|
127 | 127 | def parserequestfromenv(env, reponame=None, altbaseurl=None, bodyfh=None): |
|
128 | 128 | """Parse URL components from environment variables. |
|
129 | 129 | |
|
130 | 130 | WSGI defines request attributes via environment variables. This function |
|
131 | 131 | parses the environment variables into a data structure. |
|
132 | 132 | |
|
133 | 133 | If ``reponame`` is defined, the leading path components matching that |
|
134 | 134 | string are effectively shifted from ``PATH_INFO`` to ``SCRIPT_NAME``. |
|
135 | 135 | This simulates the world view of a WSGI application that processes |
|
136 | 136 | requests from the base URL of a repo. |
|
137 | 137 | |
|
138 | 138 | If ``altbaseurl`` (typically comes from ``web.baseurl`` config option) |
|
139 | 139 | is defined, it is used - instead of the WSGI environment variables - for |
|
140 | 140 | constructing URL components up to and including the WSGI application path. |
|
141 | 141 | For example, if the current WSGI application is at ``/repo`` and a request |
|
142 | 142 | is made to ``/rev/@`` with this argument set to |
|
143 | 143 | ``http://myserver:9000/prefix``, the URL and path components will resolve as |
|
144 | 144 | if the request were to ``http://myserver:9000/prefix/rev/@``. In other |
|
145 | 145 | words, ``wsgi.url_scheme``, ``SERVER_NAME``, ``SERVER_PORT``, and |
|
146 | 146 | ``SCRIPT_NAME`` are all effectively replaced by components from this URL. |
|
147 | 147 | |
|
148 | 148 | ``bodyfh`` can be used to specify a file object to read the request body |
|
149 | 149 | from. If not defined, ``wsgi.input`` from the environment dict is used. |
|
150 | 150 | """ |
|
151 | 151 | # PEP 3333 defines the WSGI spec and is a useful reference for this code. |
|
152 | 152 | |
|
153 | 153 | # We first validate that the incoming object conforms with the WSGI spec. |
|
154 | 154 | # We only want to be dealing with spec-conforming WSGI implementations. |
|
155 | 155 | # TODO enable this once we fix internal violations. |
|
156 | 156 | #wsgiref.validate.check_environ(env) |
|
157 | 157 | |
|
158 | 158 | # PEP-0333 states that environment keys and values are native strings |
|
159 | 159 | # (bytes on Python 2 and str on Python 3). The code points for the Unicode |
|
160 | 160 | # strings on Python 3 must be between \00000-\000FF. We deal with bytes |
|
161 | 161 | # in Mercurial, so mass convert string keys and values to bytes. |
|
162 | 162 | if pycompat.ispy3: |
|
163 | 163 | env = {k.encode('latin-1'): v for k, v in env.iteritems()} |
|
164 | 164 | env = {k: v.encode('latin-1') if isinstance(v, str) else v |
|
165 | 165 | for k, v in env.iteritems()} |
|
166 | 166 | |
|
167 | 167 | # Some hosting solutions are emulating hgwebdir, and dispatching directly |
|
168 | 168 | # to an hgweb instance using this environment variable. This was always |
|
169 | 169 | # checked prior to d7fd203e36cc; keep doing so to avoid breaking them. |
|
170 | 170 | if not reponame: |
|
171 | 171 | reponame = env.get('REPO_NAME') |
|
172 | 172 | |
|
173 | 173 | if altbaseurl: |
|
174 | 174 | altbaseurl = util.url(altbaseurl) |
|
175 | 175 | |
|
176 | 176 | # https://www.python.org/dev/peps/pep-0333/#environ-variables defines |
|
177 | 177 | # the environment variables. |
|
178 | 178 | # https://www.python.org/dev/peps/pep-0333/#url-reconstruction defines |
|
179 | 179 | # how URLs are reconstructed. |
|
180 | 180 | fullurl = env['wsgi.url_scheme'] + '://' |
|
181 | 181 | |
|
182 | 182 | if altbaseurl and altbaseurl.scheme: |
|
183 | 183 | advertisedfullurl = altbaseurl.scheme + '://' |
|
184 | 184 | else: |
|
185 | 185 | advertisedfullurl = fullurl |
|
186 | 186 | |
|
187 | 187 | def addport(s, port): |
|
188 | 188 | if s.startswith('https://'): |
|
189 | 189 | if port != '443': |
|
190 | 190 | s += ':' + port |
|
191 | 191 | else: |
|
192 | 192 | if port != '80': |
|
193 | 193 | s += ':' + port |
|
194 | 194 | |
|
195 | 195 | return s |
|
196 | 196 | |
|
197 | 197 | if env.get('HTTP_HOST'): |
|
198 | 198 | fullurl += env['HTTP_HOST'] |
|
199 | 199 | else: |
|
200 | 200 | fullurl += env['SERVER_NAME'] |
|
201 | 201 | fullurl = addport(fullurl, env['SERVER_PORT']) |
|
202 | 202 | |
|
203 | 203 | if altbaseurl and altbaseurl.host: |
|
204 | 204 | advertisedfullurl += altbaseurl.host |
|
205 | 205 | |
|
206 | 206 | if altbaseurl.port: |
|
207 | 207 | port = altbaseurl.port |
|
208 | 208 | elif altbaseurl.scheme == 'http' and not altbaseurl.port: |
|
209 | 209 | port = '80' |
|
210 | 210 | elif altbaseurl.scheme == 'https' and not altbaseurl.port: |
|
211 | 211 | port = '443' |
|
212 | 212 | else: |
|
213 | 213 | port = env['SERVER_PORT'] |
|
214 | 214 | |
|
215 | 215 | advertisedfullurl = addport(advertisedfullurl, port) |
|
216 | 216 | else: |
|
217 | 217 | advertisedfullurl += env['SERVER_NAME'] |
|
218 | 218 | advertisedfullurl = addport(advertisedfullurl, env['SERVER_PORT']) |
|
219 | 219 | |
|
220 | 220 | baseurl = fullurl |
|
221 | 221 | advertisedbaseurl = advertisedfullurl |
|
222 | 222 | |
|
223 | 223 | fullurl += util.urlreq.quote(env.get('SCRIPT_NAME', '')) |
|
224 | 224 | fullurl += util.urlreq.quote(env.get('PATH_INFO', '')) |
|
225 | 225 | |
|
226 | 226 | if altbaseurl: |
|
227 | 227 | path = altbaseurl.path or '' |
|
228 | 228 | if path and not path.startswith('/'): |
|
229 | 229 | path = '/' + path |
|
230 | 230 | advertisedfullurl += util.urlreq.quote(path) |
|
231 | 231 | else: |
|
232 | 232 | advertisedfullurl += util.urlreq.quote(env.get('SCRIPT_NAME', '')) |
|
233 | 233 | |
|
234 | 234 | advertisedfullurl += util.urlreq.quote(env.get('PATH_INFO', '')) |
|
235 | 235 | |
|
236 | 236 | if env.get('QUERY_STRING'): |
|
237 | 237 | fullurl += '?' + env['QUERY_STRING'] |
|
238 | 238 | advertisedfullurl += '?' + env['QUERY_STRING'] |
|
239 | 239 | |
|
240 | 240 | # If ``reponame`` is defined, that must be a prefix on PATH_INFO |
|
241 | 241 | # that represents the repository being dispatched to. When computing |
|
242 | 242 | # the dispatch info, we ignore these leading path components. |
|
243 | 243 | |
|
244 | 244 | if altbaseurl: |
|
245 | 245 | apppath = altbaseurl.path or '' |
|
246 | 246 | if apppath and not apppath.startswith('/'): |
|
247 | 247 | apppath = '/' + apppath |
|
248 | 248 | else: |
|
249 | 249 | apppath = env.get('SCRIPT_NAME', '') |
|
250 | 250 | |
|
251 | 251 | if reponame: |
|
252 | 252 | repoprefix = '/' + reponame.strip('/') |
|
253 | 253 | |
|
254 | 254 | if not env.get('PATH_INFO'): |
|
255 | 255 | raise error.ProgrammingError('reponame requires PATH_INFO') |
|
256 | 256 | |
|
257 | 257 | if not env['PATH_INFO'].startswith(repoprefix): |
|
258 | 258 | raise error.ProgrammingError('PATH_INFO does not begin with repo ' |
|
259 | 259 | 'name: %s (%s)' % (env['PATH_INFO'], |
|
260 | 260 | reponame)) |
|
261 | 261 | |
|
262 | 262 | dispatchpath = env['PATH_INFO'][len(repoprefix):] |
|
263 | 263 | |
|
264 | 264 | if dispatchpath and not dispatchpath.startswith('/'): |
|
265 | 265 | raise error.ProgrammingError('reponame prefix of PATH_INFO does ' |
|
266 | 266 | 'not end at path delimiter: %s (%s)' % |
|
267 | 267 | (env['PATH_INFO'], reponame)) |
|
268 | 268 | |
|
269 | 269 | apppath = apppath.rstrip('/') + repoprefix |
|
270 | 270 | dispatchparts = dispatchpath.strip('/').split('/') |
|
271 | 271 | dispatchpath = '/'.join(dispatchparts) |
|
272 | 272 | |
|
273 | 273 | elif 'PATH_INFO' in env: |
|
274 | 274 | if env['PATH_INFO'].strip('/'): |
|
275 | 275 | dispatchparts = env['PATH_INFO'].strip('/').split('/') |
|
276 | 276 | dispatchpath = '/'.join(dispatchparts) |
|
277 | 277 | else: |
|
278 | 278 | dispatchparts = [] |
|
279 | 279 | dispatchpath = '' |
|
280 | 280 | else: |
|
281 | 281 | dispatchparts = [] |
|
282 | 282 | dispatchpath = None |
|
283 | 283 | |
|
284 | 284 | querystring = env.get('QUERY_STRING', '') |
|
285 | 285 | |
|
286 | 286 | # We store as a list so we have ordering information. We also store as |
|
287 | 287 | # a dict to facilitate fast lookup. |
|
288 | 288 | qsparams = multidict() |
|
289 | 289 | for k, v in util.urlreq.parseqsl(querystring, keep_blank_values=True): |
|
290 | 290 | qsparams.add(k, v) |
|
291 | 291 | |
|
292 | 292 | # HTTP_* keys contain HTTP request headers. The Headers structure should |
|
293 | 293 | # perform case normalization for us. We just rewrite underscore to dash |
|
294 | 294 | # so keys match what likely went over the wire. |
|
295 | 295 | headers = [] |
|
296 | 296 | for k, v in env.iteritems(): |
|
297 | 297 | if k.startswith('HTTP_'): |
|
298 | 298 | headers.append((k[len('HTTP_'):].replace('_', '-'), v)) |
|
299 | 299 | |
|
300 | 300 | from . import wsgiheaders # avoid cycle |
|
301 | 301 | headers = wsgiheaders.Headers(headers) |
|
302 | 302 | |
|
303 | 303 | # This is kind of a lie because the HTTP header wasn't explicitly |
|
304 | 304 | # sent. But for all intents and purposes it should be OK to lie about |
|
305 | 305 | # this, since a consumer will either either value to determine how many |
|
306 | 306 | # bytes are available to read. |
|
307 | 307 | if 'CONTENT_LENGTH' in env and 'HTTP_CONTENT_LENGTH' not in env: |
|
308 | 308 | headers['Content-Length'] = env['CONTENT_LENGTH'] |
|
309 | 309 | |
|
310 | 310 | if 'CONTENT_TYPE' in env and 'HTTP_CONTENT_TYPE' not in env: |
|
311 | 311 | headers['Content-Type'] = env['CONTENT_TYPE'] |
|
312 | 312 | |
|
313 | 313 | if bodyfh is None: |
|
314 | 314 | bodyfh = env['wsgi.input'] |
|
315 | 315 | if 'Content-Length' in headers: |
|
316 |
bodyfh = util.cappedreader(bodyfh, |
|
|
316 | bodyfh = util.cappedreader(bodyfh, | |
|
317 | int(headers['Content-Length'] or '0')) | |
|
317 | 318 | |
|
318 | 319 | return parsedrequest(method=env['REQUEST_METHOD'], |
|
319 | 320 | url=fullurl, baseurl=baseurl, |
|
320 | 321 | advertisedurl=advertisedfullurl, |
|
321 | 322 | advertisedbaseurl=advertisedbaseurl, |
|
322 | 323 | urlscheme=env['wsgi.url_scheme'], |
|
323 | 324 | remoteuser=env.get('REMOTE_USER'), |
|
324 | 325 | remotehost=env.get('REMOTE_HOST'), |
|
325 | 326 | apppath=apppath, |
|
326 | 327 | dispatchparts=dispatchparts, dispatchpath=dispatchpath, |
|
327 | 328 | reponame=reponame, |
|
328 | 329 | querystring=querystring, |
|
329 | 330 | qsparams=qsparams, |
|
330 | 331 | headers=headers, |
|
331 | 332 | bodyfh=bodyfh, |
|
332 | 333 | rawenv=env) |
|
333 | 334 | |
|
334 | 335 | class offsettrackingwriter(object): |
|
335 | 336 | """A file object like object that is append only and tracks write count. |
|
336 | 337 | |
|
337 | 338 | Instances are bound to a callable. This callable is called with data |
|
338 | 339 | whenever a ``write()`` is attempted. |
|
339 | 340 | |
|
340 | 341 | Instances track the amount of written data so they can answer ``tell()`` |
|
341 | 342 | requests. |
|
342 | 343 | |
|
343 | 344 | The intent of this class is to wrap the ``write()`` function returned by |
|
344 | 345 | a WSGI ``start_response()`` function. Since ``write()`` is a callable and |
|
345 | 346 | not a file object, it doesn't implement other file object methods. |
|
346 | 347 | """ |
|
347 | 348 | def __init__(self, writefn): |
|
348 | 349 | self._write = writefn |
|
349 | 350 | self._offset = 0 |
|
350 | 351 | |
|
351 | 352 | def write(self, s): |
|
352 | 353 | res = self._write(s) |
|
353 | 354 | # Some Python objects don't report the number of bytes written. |
|
354 | 355 | if res is None: |
|
355 | 356 | self._offset += len(s) |
|
356 | 357 | else: |
|
357 | 358 | self._offset += res |
|
358 | 359 | |
|
359 | 360 | def flush(self): |
|
360 | 361 | pass |
|
361 | 362 | |
|
362 | 363 | def tell(self): |
|
363 | 364 | return self._offset |
|
364 | 365 | |
|
365 | 366 | class wsgiresponse(object): |
|
366 | 367 | """Represents a response to a WSGI request. |
|
367 | 368 | |
|
368 | 369 | A response consists of a status line, headers, and a body. |
|
369 | 370 | |
|
370 | 371 | Consumers must populate the ``status`` and ``headers`` fields and |
|
371 | 372 | make a call to a ``setbody*()`` method before the response can be |
|
372 | 373 | issued. |
|
373 | 374 | |
|
374 | 375 | When it is time to start sending the response over the wire, |
|
375 | 376 | ``sendresponse()`` is called. It handles emitting the header portion |
|
376 | 377 | of the response message. It then yields chunks of body data to be |
|
377 | 378 | written to the peer. Typically, the WSGI application itself calls |
|
378 | 379 | and returns the value from ``sendresponse()``. |
|
379 | 380 | """ |
|
380 | 381 | |
|
381 | 382 | def __init__(self, req, startresponse): |
|
382 | 383 | """Create an empty response tied to a specific request. |
|
383 | 384 | |
|
384 | 385 | ``req`` is a ``parsedrequest``. ``startresponse`` is the |
|
385 | 386 | ``start_response`` function passed to the WSGI application. |
|
386 | 387 | """ |
|
387 | 388 | self._req = req |
|
388 | 389 | self._startresponse = startresponse |
|
389 | 390 | |
|
390 | 391 | self.status = None |
|
391 | 392 | from . import wsgiheaders # avoid cycle |
|
392 | 393 | self.headers = wsgiheaders.Headers([]) |
|
393 | 394 | |
|
394 | 395 | self._bodybytes = None |
|
395 | 396 | self._bodygen = None |
|
396 | 397 | self._bodywillwrite = False |
|
397 | 398 | self._started = False |
|
398 | 399 | self._bodywritefn = None |
|
399 | 400 | |
|
400 | 401 | def _verifybody(self): |
|
401 | 402 | if (self._bodybytes is not None or self._bodygen is not None |
|
402 | 403 | or self._bodywillwrite): |
|
403 | 404 | raise error.ProgrammingError('cannot define body multiple times') |
|
404 | 405 | |
|
405 | 406 | def setbodybytes(self, b): |
|
406 | 407 | """Define the response body as static bytes. |
|
407 | 408 | |
|
408 | 409 | The empty string signals that there is no response body. |
|
409 | 410 | """ |
|
410 | 411 | self._verifybody() |
|
411 | 412 | self._bodybytes = b |
|
412 | 413 | self.headers['Content-Length'] = '%d' % len(b) |
|
413 | 414 | |
|
414 | 415 | def setbodygen(self, gen): |
|
415 | 416 | """Define the response body as a generator of bytes.""" |
|
416 | 417 | self._verifybody() |
|
417 | 418 | self._bodygen = gen |
|
418 | 419 | |
|
419 | 420 | def setbodywillwrite(self): |
|
420 | 421 | """Signal an intent to use write() to emit the response body. |
|
421 | 422 | |
|
422 | 423 | **This is the least preferred way to send a body.** |
|
423 | 424 | |
|
424 | 425 | It is preferred for WSGI applications to emit a generator of chunks |
|
425 | 426 | constituting the response body. However, some consumers can't emit |
|
426 | 427 | data this way. So, WSGI provides a way to obtain a ``write(data)`` |
|
427 | 428 | function that can be used to synchronously perform an unbuffered |
|
428 | 429 | write. |
|
429 | 430 | |
|
430 | 431 | Calling this function signals an intent to produce the body in this |
|
431 | 432 | manner. |
|
432 | 433 | """ |
|
433 | 434 | self._verifybody() |
|
434 | 435 | self._bodywillwrite = True |
|
435 | 436 | |
|
436 | 437 | def sendresponse(self): |
|
437 | 438 | """Send the generated response to the client. |
|
438 | 439 | |
|
439 | 440 | Before this is called, ``status`` must be set and one of |
|
440 | 441 | ``setbodybytes()`` or ``setbodygen()`` must be called. |
|
441 | 442 | |
|
442 | 443 | Calling this method multiple times is not allowed. |
|
443 | 444 | """ |
|
444 | 445 | if self._started: |
|
445 | 446 | raise error.ProgrammingError('sendresponse() called multiple times') |
|
446 | 447 | |
|
447 | 448 | self._started = True |
|
448 | 449 | |
|
449 | 450 | if not self.status: |
|
450 | 451 | raise error.ProgrammingError('status line not defined') |
|
451 | 452 | |
|
452 | 453 | if (self._bodybytes is None and self._bodygen is None |
|
453 | 454 | and not self._bodywillwrite): |
|
454 | 455 | raise error.ProgrammingError('response body not defined') |
|
455 | 456 | |
|
456 | 457 | # RFC 7232 Section 4.1 states that a 304 MUST generate one of |
|
457 | 458 | # {Cache-Control, Content-Location, Date, ETag, Expires, Vary} |
|
458 | 459 | # and SHOULD NOT generate other headers unless they could be used |
|
459 | 460 | # to guide cache updates. Furthermore, RFC 7230 Section 3.3.2 |
|
460 | 461 | # states that no response body can be issued. Content-Length can |
|
461 | 462 | # be sent. But if it is present, it should be the size of the response |
|
462 | 463 | # that wasn't transferred. |
|
463 | 464 | if self.status.startswith('304 '): |
|
464 | 465 | # setbodybytes('') will set C-L to 0. This doesn't conform with the |
|
465 | 466 | # spec. So remove it. |
|
466 | 467 | if self.headers.get('Content-Length') == '0': |
|
467 | 468 | del self.headers['Content-Length'] |
|
468 | 469 | |
|
469 | 470 | # Strictly speaking, this is too strict. But until it causes |
|
470 | 471 | # problems, let's be strict. |
|
471 | 472 | badheaders = {k for k in self.headers.keys() |
|
472 | 473 | if k.lower() not in ('date', 'etag', 'expires', |
|
473 | 474 | 'cache-control', |
|
474 | 475 | 'content-location', |
|
475 | 476 | 'vary')} |
|
476 | 477 | if badheaders: |
|
477 | 478 | raise error.ProgrammingError( |
|
478 | 479 | 'illegal header on 304 response: %s' % |
|
479 | 480 | ', '.join(sorted(badheaders))) |
|
480 | 481 | |
|
481 | 482 | if self._bodygen is not None or self._bodywillwrite: |
|
482 | 483 | raise error.ProgrammingError("must use setbodybytes('') with " |
|
483 | 484 | "304 responses") |
|
484 | 485 | |
|
485 | 486 | # Various HTTP clients (notably httplib) won't read the HTTP response |
|
486 | 487 | # until the HTTP request has been sent in full. If servers (us) send a |
|
487 | 488 | # response before the HTTP request has been fully sent, the connection |
|
488 | 489 | # may deadlock because neither end is reading. |
|
489 | 490 | # |
|
490 | 491 | # We work around this by "draining" the request data before |
|
491 | 492 | # sending any response in some conditions. |
|
492 | 493 | drain = False |
|
493 | 494 | close = False |
|
494 | 495 | |
|
495 | 496 | # If the client sent Expect: 100-continue, we assume it is smart enough |
|
496 | 497 | # to deal with the server sending a response before reading the request. |
|
497 | 498 | # (httplib doesn't do this.) |
|
498 | 499 | if self._req.headers.get('Expect', '').lower() == '100-continue': |
|
499 | 500 | pass |
|
500 | 501 | # Only tend to request methods that have bodies. Strictly speaking, |
|
501 | 502 | # we should sniff for a body. But this is fine for our existing |
|
502 | 503 | # WSGI applications. |
|
503 | 504 | elif self._req.method not in ('POST', 'PUT'): |
|
504 | 505 | pass |
|
505 | 506 | else: |
|
506 | 507 | # If we don't know how much data to read, there's no guarantee |
|
507 | 508 | # that we can drain the request responsibly. The WSGI |
|
508 | 509 | # specification only says that servers *should* ensure the |
|
509 | 510 | # input stream doesn't overrun the actual request. So there's |
|
510 | 511 | # no guarantee that reading until EOF won't corrupt the stream |
|
511 | 512 | # state. |
|
512 | 513 | if not isinstance(self._req.bodyfh, util.cappedreader): |
|
513 | 514 | close = True |
|
514 | 515 | else: |
|
515 | 516 | # We /could/ only drain certain HTTP response codes. But 200 and |
|
516 | 517 | # non-200 wire protocol responses both require draining. Since |
|
517 | 518 | # we have a capped reader in place for all situations where we |
|
518 | 519 | # drain, it is safe to read from that stream. We'll either do |
|
519 | 520 | # a drain or no-op if we're already at EOF. |
|
520 | 521 | drain = True |
|
521 | 522 | |
|
522 | 523 | if close: |
|
523 | 524 | self.headers['Connection'] = 'Close' |
|
524 | 525 | |
|
525 | 526 | if drain: |
|
526 | 527 | assert isinstance(self._req.bodyfh, util.cappedreader) |
|
527 | 528 | while True: |
|
528 | 529 | chunk = self._req.bodyfh.read(32768) |
|
529 | 530 | if not chunk: |
|
530 | 531 | break |
|
531 | 532 | |
|
532 | 533 | strheaders = [(pycompat.strurl(k), pycompat.strurl(v)) for |
|
533 | 534 | k, v in self.headers.items()] |
|
534 | 535 | write = self._startresponse(pycompat.sysstr(self.status), |
|
535 | 536 | strheaders) |
|
536 | 537 | |
|
537 | 538 | if self._bodybytes: |
|
538 | 539 | yield self._bodybytes |
|
539 | 540 | elif self._bodygen: |
|
540 | 541 | for chunk in self._bodygen: |
|
541 | 542 | yield chunk |
|
542 | 543 | elif self._bodywillwrite: |
|
543 | 544 | self._bodywritefn = write |
|
544 | 545 | else: |
|
545 | 546 | error.ProgrammingError('do not know how to send body') |
|
546 | 547 | |
|
547 | 548 | def getbodyfile(self): |
|
548 | 549 | """Obtain a file object like object representing the response body. |
|
549 | 550 | |
|
550 | 551 | For this to work, you must call ``setbodywillwrite()`` and then |
|
551 | 552 | ``sendresponse()`` first. ``sendresponse()`` is a generator and the |
|
552 | 553 | function won't run to completion unless the generator is advanced. The |
|
553 | 554 | generator yields not items. The easiest way to consume it is with |
|
554 | 555 | ``list(res.sendresponse())``, which should resolve to an empty list - |
|
555 | 556 | ``[]``. |
|
556 | 557 | """ |
|
557 | 558 | if not self._bodywillwrite: |
|
558 | 559 | raise error.ProgrammingError('must call setbodywillwrite() first') |
|
559 | 560 | |
|
560 | 561 | if not self._started: |
|
561 | 562 | raise error.ProgrammingError('must call sendresponse() first; did ' |
|
562 | 563 | 'you remember to consume it since it ' |
|
563 | 564 | 'is a generator?') |
|
564 | 565 | |
|
565 | 566 | assert self._bodywritefn |
|
566 | 567 | return offsettrackingwriter(self._bodywritefn) |
|
567 | 568 | |
|
568 | 569 | def wsgiapplication(app_maker): |
|
569 | 570 | '''For compatibility with old CGI scripts. A plain hgweb() or hgwebdir() |
|
570 | 571 | can and should now be used as a WSGI application.''' |
|
571 | 572 | application = app_maker() |
|
572 | 573 | def run_wsgi(env, respond): |
|
573 | 574 | return application(env, respond) |
|
574 | 575 | return run_wsgi |
General Comments 0
You need to be logged in to leave comments.
Login now