2 Copyright (C) 1995, 1996, 1997, 1998, 2000 Free Software Foundation, Inc.
4 This file is part of Wget.
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or
9 (at your option) any later version.
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with this program; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
24 #include <sys/types.h>
36 #if TIME_WITH_SYS_TIME
37 # include <sys/time.h>
41 # include <sys/time.h>
65 extern char *version_string;
75 #define TEXTHTML_S "text/html"
76 #define HTTP_ACCEPT "*/*"
78 /* Some status code validation macros: */
79 #define H_20X(x) (((x) >= 200) && ((x) < 300))
80 #define H_PARTIAL(x) ((x) == HTTP_STATUS_PARTIAL_CONTENTS)
81 #define H_REDIRECTED(x) (((x) == HTTP_STATUS_MOVED_PERMANENTLY) \
82 || ((x) == HTTP_STATUS_MOVED_TEMPORARILY))
84 /* HTTP/1.0 status codes from RFC1945, provided for reference. */
86 #define HTTP_STATUS_OK 200
87 #define HTTP_STATUS_CREATED 201
88 #define HTTP_STATUS_ACCEPTED 202
89 #define HTTP_STATUS_NO_CONTENT 204
90 #define HTTP_STATUS_PARTIAL_CONTENTS 206
92 /* Redirection 3xx. */
93 #define HTTP_STATUS_MULTIPLE_CHOICES 300
94 #define HTTP_STATUS_MOVED_PERMANENTLY 301
95 #define HTTP_STATUS_MOVED_TEMPORARILY 302
96 #define HTTP_STATUS_NOT_MODIFIED 304
98 /* Client error 4xx. */
99 #define HTTP_STATUS_BAD_REQUEST 400
100 #define HTTP_STATUS_UNAUTHORIZED 401
101 #define HTTP_STATUS_FORBIDDEN 403
102 #define HTTP_STATUS_NOT_FOUND 404
104 /* Server errors 5xx. */
105 #define HTTP_STATUS_INTERNAL 500
106 #define HTTP_STATUS_NOT_IMPLEMENTED 501
107 #define HTTP_STATUS_BAD_GATEWAY 502
108 #define HTTP_STATUS_UNAVAILABLE 503
111 /* Parse the HTTP status line, which is of format:
113 HTTP-Version SP Status-Code SP Reason-Phrase
115 The function returns the status-code, or -1 if the status line is
116 malformed. The pointer to reason-phrase is returned in RP. */
118 parse_http_status_line (const char *line, const char **reason_phrase_ptr)
120 /* (the variables must not be named `major' and `minor', because
121 that breaks compilation with SunOS4 cc.) */
122 int mjr, mnr, statcode;
125 *reason_phrase_ptr = NULL;
127 /* The standard format of HTTP-Version is: `HTTP/X.Y', where X is
128 major version, and Y is minor version. */
129 if (strncmp (line, "HTTP/", 5) != 0)
133 /* Calculate major HTTP version. */
135 for (mjr = 0; ISDIGIT (*line); line++)
136 mjr = 10 * mjr + (*line - '0');
137 if (*line != '.' || p == line)
141 /* Calculate minor HTTP version. */
143 for (mnr = 0; ISDIGIT (*line); line++)
144 mnr = 10 * mnr + (*line - '0');
145 if (*line != ' ' || p == line)
147 /* Wget will accept only 1.0 and higher HTTP-versions. The value of
148 minor version can be safely ignored. */
153 /* Calculate status code. */
154 if (!(ISDIGIT (*line) && ISDIGIT (line[1]) && ISDIGIT (line[2])))
156 statcode = 100 * (*line - '0') + 10 * (line[1] - '0') + (line[2] - '0');
158 /* Set up the reason phrase pointer. */
160 /* RFC2068 requires SPC here, but we allow the string to finish
161 here, in case no reason-phrase is present. */
165 *reason_phrase_ptr = line;
170 *reason_phrase_ptr = line + 1;
175 /* Functions to be used as arguments to header_process(): */
177 struct http_process_range_closure {
183 /* Parse the `Content-Range' header and extract the information it
184 contains. Returns 1 if successful, -1 otherwise. */
186 http_process_range (const char *hdr, void *arg)
188 struct http_process_range_closure *closure
189 = (struct http_process_range_closure *)arg;
192 /* Certain versions of Nutscape proxy server send out
193 `Content-Length' without "bytes" specifier, which is a breach of
194 RFC2068 (as well as the HTTP/1.1 draft which was current at the
195 time). But hell, I must support it... */
196 if (!strncasecmp (hdr, "bytes", 5))
199 hdr += skip_lws (hdr);
205 for (num = 0; ISDIGIT (*hdr); hdr++)
206 num = 10 * num + (*hdr - '0');
207 if (*hdr != '-' || !ISDIGIT (*(hdr + 1)))
209 closure->first_byte_pos = num;
211 for (num = 0; ISDIGIT (*hdr); hdr++)
212 num = 10 * num + (*hdr - '0');
213 if (*hdr != '/' || !ISDIGIT (*(hdr + 1)))
215 closure->last_byte_pos = num;
217 for (num = 0; ISDIGIT (*hdr); hdr++)
218 num = 10 * num + (*hdr - '0');
219 closure->entity_length = num;
223 /* Place 1 to ARG if the HDR contains the word "none", 0 otherwise.
224 Used for `Accept-Ranges'. */
226 http_process_none (const char *hdr, void *arg)
228 int *where = (int *)arg;
230 if (strstr (hdr, "none"))
237 /* Place the malloc-ed copy of HDR hdr, to the first `;' to ARG. */
239 http_process_type (const char *hdr, void *arg)
241 char **result = (char **)arg;
244 p = strrchr (hdr, ';');
248 *result = (char *)xmalloc (len + 1);
249 memcpy (*result, hdr, len);
250 (*result)[len] = '\0';
253 *result = xstrdup (hdr);
257 /* Check whether the `Connection' header is set to "keep-alive". */
259 http_process_connection (const char *hdr, void *arg)
261 int *flag = (int *)arg;
262 if (!strcasecmp (hdr, "Keep-Alive"))
267 /* Persistent connections (pc). Currently, we cache the most recently
268 used connection as persistent, provided that the HTTP server agrees
269 to make it such. The persistence data is stored in the variables
270 below. Ideally, it would be in a structure, and it should be
271 possible to cache an arbitrary fixed number of these connections.
273 I think the code is quite easy to extend in that direction. */
275 /* Whether the persistent connection is active. */
276 static int pc_active_p;
278 /* Host and port of the last persistent connection. */
279 static unsigned char pc_last_host[4];
280 static unsigned short pc_last_port;
282 /* File descriptor of the last persistent connection. */
283 static int pc_last_fd;
285 /* Mark the persistent connection as invalid. This is used by the
286 CLOSE_* macros after they forcefully close a registered persistent
290 invalidate_persistent (void)
293 DEBUGP (("Invalidating fd %d from further reuse.\n", pc_last_fd));
296 /* Register FD, which should be a TCP/IP connection to HOST:PORT, as
297 persistent. This will enable someone to use the same connection
298 later. In the context of HTTP, this must be called only AFTER the
299 response has been received and the server has promised that the
300 connection will remain alive.
302 If a previous connection was persistent, it is closed. */
305 register_persistent (const char *host, unsigned short port, int fd)
311 if (pc_last_fd == fd)
313 /* The connection FD is already registered. Nothing to
319 /* The old persistent connection is still active; let's
320 close it first. This situation arises whenever a
321 persistent connection exists, but we then connect to a
322 different host, and try to register a persistent
323 connection to that one. */
325 invalidate_persistent ();
329 /* This store_hostaddress may not fail, because it has the results
331 success = store_hostaddress (pc_last_host, host);
336 DEBUGP (("Registered fd %d for persistent reuse.\n", fd));
339 /* Return non-zero if a persistent connection is available for
340 connecting to HOST:PORT. */
343 persistent_available_p (const char *host, unsigned short port)
345 unsigned char this_host[4];
348 if (port != pc_last_port)
350 if (!store_hostaddress (this_host, host))
352 if (memcmp (pc_last_host, this_host, 4))
354 if (!test_socket_open (pc_last_fd))
357 invalidate_persistent ();
363 /* The idea behind these two CLOSE macros is to distinguish between
364 two cases: one when the job we've been doing is finished, and we
365 want to close the connection and leave, and two when something is
366 seriously wrong and we're closing the connection as part of
369 In case of keep_alive, CLOSE_FINISH should leave the connection
370 open, while CLOSE_INVALIDATE should still close it.
372 Note that the semantics of the flag `keep_alive' is "this
373 connection *will* be reused (the server has promised not to close
374 the connection once we're done)", while the semantics of
375 `pc_active_p && (fd) == pc_last_fd' is "we're *now* using an
376 active, registered connection". */
378 #define CLOSE_FINISH(fd) do { \
382 if (pc_active_p && (fd) == pc_last_fd) \
383 invalidate_persistent (); \
387 #define CLOSE_INVALIDATE(fd) do { \
389 if (pc_active_p && (fd) == pc_last_fd) \
390 invalidate_persistent (); \
396 long len; /* received length */
397 long contlen; /* expected length */
398 long restval; /* the restart value */
399 int res; /* the result of last read */
400 char *newloc; /* new location (redirection) */
401 char *remote_time; /* remote time-stamp string */
402 char *error; /* textual HTTP error */
403 int statcode; /* status code */
404 long dltime; /* time of the download */
407 /* Free the elements of hstat X. */
408 #define FREEHSTAT(x) do \
410 FREE_MAYBE ((x).newloc); \
411 FREE_MAYBE ((x).remote_time); \
412 FREE_MAYBE ((x).error); \
413 (x).newloc = (x).remote_time = (x).error = NULL; \
416 static char *create_authorization_line PARAMS ((const char *, const char *,
417 const char *, const char *,
419 static char *basic_authentication_encode PARAMS ((const char *, const char *,
421 static int known_authentication_scheme_p PARAMS ((const char *));
423 static time_t http_atotm PARAMS ((char *));
425 #define BEGINS_WITH(line, string_constant) \
426 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
427 && (ISSPACE (line[sizeof (string_constant) - 1]) \
428 || !line[sizeof (string_constant) - 1]))
430 /* Retrieve a document through HTTP protocol. It recognizes status
431 code, and correctly handles redirections. It closes the network
432 socket. If it receives an error from the functions below it, it
433 will print it if there is enough information to do so (almost
434 always), returning the error to the caller (i.e. http_loop).
436 Various HTTP parameters are stored to hs. Although it parses the
437 response code correctly, it is not used in a sane way. The caller
440 If u->proxy is non-NULL, the URL u will be taken as a proxy URL,
441 and u->proxy->url will be given to the proxy server (bad naming,
444 gethttp (struct urlinfo *u, struct http_stat *hs, int *dt)
446 char *request, *type, *command, *path;
448 char *pragma_h, *referer, *useragent, *range, *wwwauth, *remhost;
449 char *authenticate_h;
453 char *request_keep_alive;
454 int sock, hcount, num_written, all_length, remport, statcode;
455 long contlen, contrange;
459 int auth_tried_already;
462 /* Whether this connection will be kept alive after the HTTP request
466 /* Flags that detect the two ways of specifying HTTP keep-alive
468 int http_keep_alive_1, http_keep_alive_2;
470 /* Whether keep-alive should be inhibited. */
471 int inhibit_keep_alive;
473 if (!(*dt & HEAD_ONLY))
474 /* If we're doing a GET on the URL, as opposed to just a HEAD, we need to
475 know the local filename so we can save to it. */
476 assert (u->local != NULL);
479 auth_tried_already = 0;
481 inhibit_keep_alive = (!opt.http_keep_alive || u->proxy != NULL);
484 /* We need to come back here when the initial attempt to retrieve
485 without authorization header fails. (Expected to happen at least
486 for the Digest authorization scheme.) */
489 http_keep_alive_1 = http_keep_alive_2 = 0;
491 /* Initialize certain elements of struct http_stat. */
496 hs->remote_time = NULL;
499 /* Which structure to use to retrieve the original URL data. */
505 /* First: establish the connection. */
506 if (inhibit_keep_alive
507 || !persistent_available_p (u->host, u->port))
509 logprintf (LOG_VERBOSE, _("Connecting to %s:%hu... "), u->host, u->port);
510 err = make_connection (&sock, u->host, u->port);
514 logputs (LOG_VERBOSE, "\n");
515 logprintf (LOG_NOTQUIET, "%s: %s.\n", u->host, herrmsg (h_errno));
519 logputs (LOG_VERBOSE, "\n");
520 logprintf (LOG_NOTQUIET, "socket: %s\n", strerror (errno));
524 logputs (LOG_VERBOSE, "\n");
525 logprintf (LOG_NOTQUIET,
526 _("Connection to %s:%hu refused.\n"), u->host, u->port);
530 logputs (LOG_VERBOSE, "\n");
531 logprintf (LOG_NOTQUIET, "connect: %s\n", strerror (errno));
536 /* Everything is fine! */
537 logputs (LOG_VERBOSE, _("connected!\n"));
546 logprintf (LOG_VERBOSE, _("Reusing connection to %s:%hu.\n"), u->host, u->port);
547 /* #### pc_last_fd should be accessed through an accessor
550 DEBUGP (("Reusing fd %d.\n", sock));
554 path = u->proxy->url;
558 command = (*dt & HEAD_ONLY) ? "HEAD" : "GET";
562 referer = (char *)alloca (9 + strlen (ou->referer) + 3);
563 sprintf (referer, "Referer: %s\r\n", ou->referer);
565 if (*dt & SEND_NOCACHE)
566 pragma_h = "Pragma: no-cache\r\n";
571 range = (char *)alloca (13 + numdigit (hs->restval) + 4);
572 /* Gag me! Some servers (e.g. WebSitePro) have been known to
573 respond to the following `Range' format by generating a
574 multipart/x-byte-ranges MIME document! This MIME type was
575 present in an old draft of the byteranges specification.
576 HTTP/1.1 specifies a multipart/byte-ranges MIME type, but
577 only if multiple non-overlapping ranges are requested --
578 which Wget never does. */
579 sprintf (range, "Range: bytes=%ld-\r\n", hs->restval);
584 STRDUP_ALLOCA (useragent, opt.useragent);
587 useragent = (char *)alloca (10 + strlen (version_string));
588 sprintf (useragent, "Wget/%s", version_string);
590 /* Construct the authentication, if userid is present. */
593 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
594 user = user ? user : opt.http_user;
595 passwd = passwd ? passwd : opt.http_passwd;
602 /* We have the username and the password, but haven't tried
603 any authorization yet. Let's see if the "Basic" method
604 works. If not, we'll come back here and construct a
605 proper authorization method with the right challenges.
607 If we didn't employ this kind of logic, every URL that
608 requires authorization would have to be processed twice,
609 which is very suboptimal and generates a bunch of false
610 "unauthorized" errors in the server log.
612 #### But this logic also has a serious problem when used
613 with stronger authentications: we *first* transmit the
614 username and the password in clear text, and *then*
615 attempt a stronger authentication scheme. That cannot be
616 right! We are only fortunate that almost everyone still
617 uses the `Basic' scheme anyway.
619 There should be an option to prevent this from happening,
620 for those who use strong authentication schemes and value
622 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
626 wwwauth = create_authorization_line (authenticate_h, user, passwd,
634 char *proxy_user, *proxy_passwd;
635 /* For normal username and password, URL components override
636 command-line/wgetrc parameters. With proxy authentication,
637 it's the reverse, because proxy URLs are normally the
638 "permanent" ones, so command-line args should take
640 if (opt.proxy_user && opt.proxy_passwd)
642 proxy_user = opt.proxy_user;
643 proxy_passwd = opt.proxy_passwd;
647 proxy_user = u->user;
648 proxy_passwd = u->passwd;
650 /* #### This is junky. Can't the proxy request, say, `Digest'
652 if (proxy_user && proxy_passwd)
653 proxyauth = basic_authentication_encode (proxy_user, proxy_passwd,
654 "Proxy-Authorization");
659 /* String of the form :PORT. Used only for non-standard ports. */
663 port_maybe = (char *)alloca (numdigit (remport) + 2);
664 sprintf (port_maybe, ":%d", remport);
667 if (!inhibit_keep_alive)
668 request_keep_alive = "Connection: Keep-Alive\r\n";
670 request_keep_alive = NULL;
672 /* Allocate the memory for the request. */
673 request = (char *)alloca (strlen (command) + strlen (path)
676 + (port_maybe ? strlen (port_maybe) : 0)
677 + strlen (HTTP_ACCEPT)
678 + (request_keep_alive
679 ? strlen (request_keep_alive) : 0)
680 + (referer ? strlen (referer) : 0)
681 + (wwwauth ? strlen (wwwauth) : 0)
682 + (proxyauth ? strlen (proxyauth) : 0)
683 + (range ? strlen (range) : 0)
685 + (opt.user_header ? strlen (opt.user_header) : 0)
687 /* Construct the request. */
694 command, path, useragent, remhost,
695 port_maybe ? port_maybe : "",
697 request_keep_alive ? request_keep_alive : "",
698 referer ? referer : "",
699 wwwauth ? wwwauth : "",
700 proxyauth ? proxyauth : "",
703 opt.user_header ? opt.user_header : "");
704 DEBUGP (("---request begin---\n%s---request end---\n", request));
705 /* Free the temporary memory. */
706 FREE_MAYBE (wwwauth);
707 FREE_MAYBE (proxyauth);
709 /* Send the request to server. */
710 num_written = iwrite (sock, request, strlen (request));
713 logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"),
715 CLOSE_INVALIDATE (sock);
718 logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
719 u->proxy ? "Proxy" : "HTTP");
720 contlen = contrange = -1;
725 /* Before reading anything, initialize the rbuf. */
726 rbuf_initialize (&rbuf, sock);
730 /* Header-fetching loop. */
738 /* Get the header. */
739 status = header_get (&rbuf, &hdr,
740 /* Disallow continuations for status line. */
741 (hcount == 1 ? HG_NO_CONTINUATIONS : HG_NONE));
743 /* Check for errors. */
744 if (status == HG_EOF && *hdr)
746 /* This used to be an unconditional error, but that was
747 somewhat controversial, because of a large number of
748 broken CGI's that happily "forget" to send the second EOL
749 before closing the connection of a HEAD request.
751 So, the deal is to check whether the header is empty
752 (*hdr is zero if it is); if yes, it means that the
753 previous header was fully retrieved, and that -- most
754 probably -- the request is complete. "...be liberal in
755 what you accept." Oh boy. */
756 logputs (LOG_VERBOSE, "\n");
757 logputs (LOG_NOTQUIET, _("End of file while parsing headers.\n"));
760 FREE_MAYBE (hs->newloc);
761 FREE_MAYBE (all_headers);
762 CLOSE_INVALIDATE (sock);
765 else if (status == HG_ERROR)
767 logputs (LOG_VERBOSE, "\n");
768 logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
772 FREE_MAYBE (hs->newloc);
773 FREE_MAYBE (all_headers);
774 CLOSE_INVALIDATE (sock);
778 /* If the headers are to be saved to a file later, save them to
780 if (opt.save_headers)
782 int lh = strlen (hdr);
783 all_headers = (char *)xrealloc (all_headers, all_length + lh + 2);
784 memcpy (all_headers + all_length, hdr, lh);
786 all_headers[all_length++] = '\n';
787 all_headers[all_length] = '\0';
790 /* Print the header if requested. */
791 if (opt.server_response && hcount != 1)
792 logprintf (LOG_VERBOSE, "\n%d %s", hcount, hdr);
794 /* Check for status line. */
798 /* Parse the first line of server response. */
799 statcode = parse_http_status_line (hdr, &error);
800 hs->statcode = statcode;
801 /* Store the descriptive response. */
802 if (statcode == -1) /* malformed response */
804 /* A common reason for "malformed response" error is the
805 case when no data was actually received. Handle this
808 hs->error = xstrdup (_("No data received"));
810 hs->error = xstrdup (_("Malformed status line"));
815 hs->error = xstrdup (_("(no description)"));
817 hs->error = xstrdup (error);
824 logprintf (LOG_VERBOSE, "%d %s", statcode, error);
829 /* Exit on empty header. */
836 /* Try getting content-length. */
837 if (contlen == -1 && !opt.ignore_length)
838 if (header_process (hdr, "Content-Length", header_extract_number,
841 /* Try getting content-type. */
843 if (header_process (hdr, "Content-Type", http_process_type, &type))
845 /* Try getting location. */
847 if (header_process (hdr, "Location", header_strdup, &hs->newloc))
849 /* Try getting last-modified. */
850 if (!hs->remote_time)
851 if (header_process (hdr, "Last-Modified", header_strdup,
854 /* Try getting www-authentication. */
856 if (header_process (hdr, "WWW-Authenticate", header_strdup,
859 /* Check for accept-ranges header. If it contains the word
860 `none', disable the ranges. */
861 if (*dt & ACCEPTRANGES)
864 if (header_process (hdr, "Accept-Ranges", http_process_none, &nonep))
867 *dt &= ~ACCEPTRANGES;
871 /* Try getting content-range. */
874 struct http_process_range_closure closure;
875 if (header_process (hdr, "Content-Range", http_process_range, &closure))
877 contrange = closure.first_byte_pos;
881 /* Check for keep-alive related responses. */
882 if (!inhibit_keep_alive)
884 /* Check for the `Keep-Alive' header. */
885 if (!http_keep_alive_1)
887 if (header_process (hdr, "Keep-Alive", header_exists,
891 /* Check for `Connection: Keep-Alive'. */
892 if (!http_keep_alive_2)
894 if (header_process (hdr, "Connection", http_process_connection,
903 logputs (LOG_VERBOSE, "\n");
906 && (http_keep_alive_1 || http_keep_alive_2))
908 assert (inhibit_keep_alive == 0);
912 /* The server has promised that it will not close the connection
913 when we're done. This means that we can register it. */
914 register_persistent (u->host, u->port, sock);
916 if ((statcode == HTTP_STATUS_UNAUTHORIZED)
919 /* Authorization is required. */
924 if (auth_tried_already)
926 /* If we have tried it already, then there is not point
929 logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
930 free (authenticate_h);
933 else if (!known_authentication_scheme_p (authenticate_h))
935 free (authenticate_h);
936 logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
939 else if (BEGINS_WITH (authenticate_h, "Basic"))
941 /* The authentication scheme is basic, the one we try by
942 default, and it failed. There's no sense in trying
948 auth_tried_already = 1;
952 /* We do not need this anymore. */
955 free (authenticate_h);
956 authenticate_h = NULL;
959 /* 20x responses are counted among successful by default. */
960 if (H_20X (statcode))
963 if (type && !strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S)))
966 /* We don't assume text/html by default. */
969 if (opt.html_extension && (*dt & TEXTHTML))
970 /* -E / --html-extension / html_extension = on was specified, and this is a
971 text/html file. If some case-insensitive variation on ".htm[l]" isn't
972 already the file's suffix, tack on ".html". */
974 char* last_period_in_local_filename = strrchr(u->local, '.');
976 if (last_period_in_local_filename == NULL ||
977 !(strcasecmp(last_period_in_local_filename, ".htm") == EQ ||
978 strcasecmp(last_period_in_local_filename, ".html") == EQ))
980 size_t local_filename_len = strlen(u->local);
982 u->local = xrealloc(u->local, local_filename_len + sizeof(".html"));
983 strcpy(u->local + local_filename_len, ".html");
985 *dt |= ADDED_HTML_EXTENSION;
991 else if (contrange != hs->restval ||
992 (H_PARTIAL (statcode) && contrange == -1))
994 /* This means the whole request was somehow misunderstood by the
997 FREE_MAYBE (hs->newloc);
998 FREE_MAYBE (all_headers);
999 CLOSE_INVALIDATE (sock);
1006 contlen += contrange;
1008 contrange = -1; /* If conent-length was not sent,
1009 content-range will be ignored. */
1011 hs->contlen = contlen;
1013 /* Return if redirected. */
1014 if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES)
1016 /* RFC2068 says that in case of the 300 (multiple choices)
1017 response, the server can output a preferred URL through
1018 `Location' header; otherwise, the request should be treated
1019 like GET. So, if the location is set, it will be a
1020 redirection; otherwise, just proceed normally. */
1021 if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc)
1025 logprintf (LOG_VERBOSE,
1026 _("Location: %s%s\n"),
1027 hs->newloc ? hs->newloc : _("unspecified"),
1028 hs->newloc ? _(" [following]") : "");
1029 CLOSE_FINISH (sock);
1031 FREE_MAYBE (all_headers);
1037 if ((*dt & RETROKF) && !opt.server_response)
1039 /* No need to print this output if the body won't be
1040 downloaded at all, or if the original server response is
1042 logputs (LOG_VERBOSE, _("Length: "));
1045 logputs (LOG_VERBOSE, legible (contlen));
1046 if (contrange != -1)
1047 logprintf (LOG_VERBOSE, _(" (%s to go)"),
1048 legible (contlen - contrange));
1051 logputs (LOG_VERBOSE,
1052 opt.ignore_length ? _("ignored") : _("unspecified"));
1054 logprintf (LOG_VERBOSE, " [%s]\n", type);
1056 logputs (LOG_VERBOSE, "\n");
1060 type = NULL; /* We don't need it any more. */
1062 /* Return if we have no intention of further downloading. */
1063 if (!(*dt & RETROKF) || (*dt & HEAD_ONLY))
1065 /* In case someone cares to look... */
1069 FREE_MAYBE (all_headers);
1070 CLOSE_FINISH (sock);
1071 return RETRFINISHED;
1074 /* Open the local file. */
1077 mkalldirs (u->local);
1079 rotate_backups (u->local);
1080 fp = fopen (u->local, hs->restval ? "ab" : "wb");
1083 logprintf (LOG_NOTQUIET, "%s: %s\n", u->local, strerror (errno));
1084 CLOSE_FINISH (sock);
1085 FREE_MAYBE (all_headers);
1094 /* This will silently fail for streams that don't correspond
1095 to regular files, but that's OK. */
1101 /* #### This confuses the code that checks for file size. There
1102 should be some overhead information. */
1103 if (opt.save_headers)
1104 fwrite (all_headers, 1, all_length, fp);
1106 /* Get the contents of the document. */
1107 hs->res = get_contents (sock, fp, &hs->len, hs->restval,
1108 (contlen != -1 ? contlen : 0),
1110 hs->dltime = elapsed_time ();
1112 /* Close or flush the file. We have to be careful to check for
1113 error here. Checking the result of fwrite() is not enough --
1114 errors could go unnoticed! */
1117 flush_res = fclose (fp);
1119 flush_res = fflush (fp);
1120 if (flush_res == EOF)
1123 FREE_MAYBE (all_headers);
1124 CLOSE_FINISH (sock);
1127 return RETRFINISHED;
1130 /* The genuine HTTP loop! This is the part where the retrieval is
1131 retried, and retried, and retried, and... */
1133 http_loop (struct urlinfo *u, char **newloc, int *dt)
1135 static int first_retrieval = 1;
1138 int use_ts, got_head = 0; /* time-stamping info */
1139 char *filename_plus_orig_suffix;
1140 char *local_filename = NULL;
1141 char *tms, *suf, *locf, *tmrate;
1143 time_t tml = -1, tmr = -1; /* local and remote time-stamps */
1144 long local_size = 0; /* the size of the local file */
1145 size_t filename_len;
1146 struct http_stat hstat; /* HTTP status */
1151 /* Warn on (likely bogus) wildcard usage in HTTP. Don't use
1152 has_wildcards_p because it would also warn on `?', and we know that
1153 shows up in CGI paths a *lot*. */
1154 if (strchr (u->url, '*'))
1155 logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
1157 /* Determine the local filename. */
1159 u->local = url_filename (u->proxy ? u->proxy : u);
1161 if (!opt.output_document)
1164 locf = opt.output_document;
1166 /* Yuck. Multiple returns suck. We need to remember to free() the space we
1167 xmalloc() here before EACH return. This is one reason it's better to set
1168 flags that influence flow control and then return once at the end. */
1169 filename_len = strlen(u->local);
1170 filename_plus_orig_suffix = xmalloc(filename_len + sizeof(".orig"));
1172 if (opt.noclobber && file_exists_p (u->local))
1174 /* If opt.noclobber is turned on and file already exists, do not
1175 retrieve the file */
1176 logprintf (LOG_VERBOSE, _("\
1177 File `%s' already there, will not retrieve.\n"), u->local);
1178 /* If the file is there, we suppose it's retrieved OK. */
1181 /* #### Bogusness alert. */
1182 /* If its suffix is "html" or (yuck!) "htm", we suppose it's
1183 text/html, a harmless lie. */
1184 if (((suf = suffix (u->local)) != NULL)
1185 && (!strcmp (suf, "html") || !strcmp (suf, "htm")))
1188 free(filename_plus_orig_suffix); /* must precede every return! */
1189 /* Another harmless lie: */
1194 if (opt.timestamping)
1196 boolean local_dot_orig_file_exists = FALSE;
1198 if (opt.backup_converted)
1199 /* If -K is specified, we'll act on the assumption that it was specified
1200 last time these files were downloaded as well, and instead of just
1201 comparing local file X against server file X, we'll compare local
1202 file X.orig (if extant, else X) against server file X. If -K
1203 _wasn't_ specified last time, or the server contains files called
1204 *.orig, -N will be back to not operating correctly with -k. */
1206 /* Would a single s[n]printf() call be faster? --dan
1208 It wouldn't. sprintf() is horribly slow. At one point I
1209 profiled Wget, and found that a measurable and
1210 non-negligible amount of time was lost calling sprintf()
1211 in url.c. Replacing sprintf with inline calls to
1212 strcpy() and long_to_string() made a difference.
1214 strcpy(filename_plus_orig_suffix, u->local);
1215 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
1217 /* Try to stat() the .orig file. */
1218 if (stat(filename_plus_orig_suffix, &st) == 0)
1220 local_dot_orig_file_exists = TRUE;
1221 local_filename = filename_plus_orig_suffix;
1225 if (!local_dot_orig_file_exists)
1226 /* Couldn't stat() <file>.orig, so try to stat() <file>. */
1227 if (stat (u->local, &st) == 0)
1228 local_filename = u->local;
1230 if (local_filename != NULL)
1231 /* There was a local file, so we'll check later to see if the version
1232 the server has is the same version we already have, allowing us to
1237 local_size = st.st_size;
1241 /* Reset the counter. */
1243 *dt = 0 | ACCEPTRANGES;
1247 /* Increment the pass counter. */
1249 /* Wait before the retrieval (unless this is the very first
1251 Check if we are retrying or not, wait accordingly - HEH */
1252 if (!first_retrieval && (opt.wait || (count && opt.waitretry)))
1256 if (count<opt.waitretry)
1259 sleep(opt.waitretry);
1264 if (first_retrieval)
1265 first_retrieval = 0;
1266 /* Get the current time string. */
1267 tms = time_str (NULL);
1268 /* Print fetch message, if opt.verbose. */
1271 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
1275 sprintf (tmp, _("(try:%2d)"), count);
1276 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
1277 tms, hurl, tmp, locf);
1279 ws_changetitle (hurl, 1);
1284 /* Default document type is empty. However, if spider mode is
1285 on or time-stamping is employed, HEAD_ONLY commands is
1286 encoded within *dt. */
1287 if (opt.spider || (use_ts && !got_head))
1291 /* Assume no restarting. */
1293 /* Decide whether or not to restart. */
1294 if (((count > 1 && (*dt & ACCEPTRANGES)) || opt.always_rest)
1295 && file_exists_p (u->local))
1296 if (stat (u->local, &st) == 0)
1297 hstat.restval = st.st_size;
1298 /* Decide whether to send the no-cache directive. */
1299 if (u->proxy && (count > 1 || (opt.proxy_cache == 0)))
1300 *dt |= SEND_NOCACHE;
1302 *dt &= ~SEND_NOCACHE;
1304 /* Try fetching the document, or at least its head. :-) */
1305 err = gethttp (u, &hstat, dt);
1307 /* It's unfortunate that wget determines the local filename before finding
1308 out the Content-Type of the file. Barring a major restructuring of the
1309 code, we need to re-set locf here, since gethttp() may have xrealloc()d
1310 u->local to tack on ".html". */
1311 if (!opt.output_document)
1314 locf = opt.output_document;
1317 tms = time_str (NULL);
1318 /* Get the new location (with or without the redirection). */
1320 *newloc = xstrdup (hstat.newloc);
1323 case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
1324 case CONERROR: case READERR: case WRITEFAILED:
1326 /* Non-fatal errors continue executing the loop, which will
1327 bring them to "while" statement at the end, to judge
1328 whether the number of tries was exceeded. */
1330 printwhat (count, opt.ntry);
1333 case HOSTERR: case CONREFUSED: case PROXERR: case AUTHFAILED:
1334 /* Fatal errors just return from the function. */
1336 free(filename_plus_orig_suffix); /* must precede every return! */
1339 case FWRITEERR: case FOPENERR:
1340 /* Another fatal error. */
1341 logputs (LOG_VERBOSE, "\n");
1342 logprintf (LOG_NOTQUIET, _("Cannot write to `%s' (%s).\n"),
1343 u->local, strerror (errno));
1345 free(filename_plus_orig_suffix); /* must precede every return! */
1349 /* Return the new location to the caller. */
1352 logprintf (LOG_NOTQUIET,
1353 _("ERROR: Redirection (%d) without location.\n"),
1355 free(filename_plus_orig_suffix); /* must precede every return! */
1359 free(filename_plus_orig_suffix); /* must precede every return! */
1363 /* Deal with you later. */
1366 /* All possibilities should have been exhausted. */
1369 if (!(*dt & RETROKF))
1373 /* #### Ugly ugly ugly! */
1374 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
1375 logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
1378 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
1379 tms, hstat.statcode, hstat.error);
1380 logputs (LOG_VERBOSE, "\n");
1382 free(filename_plus_orig_suffix); /* must precede every return! */
1386 /* Did we get the time-stamp? */
1389 if (opt.timestamping && !hstat.remote_time)
1391 logputs (LOG_NOTQUIET, _("\
1392 Last-modified header missing -- time-stamps turned off.\n"));
1394 else if (hstat.remote_time)
1396 /* Convert the date-string into struct tm. */
1397 tmr = http_atotm (hstat.remote_time);
1398 if (tmr == (time_t) (-1))
1399 logputs (LOG_VERBOSE, _("\
1400 Last-modified header invalid -- time-stamp ignored.\n"));
1404 /* The time-stamping section. */
1409 use_ts = 0; /* no more time-stamping */
1410 count = 0; /* the retrieve count for HEAD is
1412 if (hstat.remote_time && tmr != (time_t) (-1))
1414 /* Now time-stamping can be used validly. Time-stamping
1415 means that if the sizes of the local and remote file
1416 match, and local file is newer than the remote file,
1417 it will not be retrieved. Otherwise, the normal
1418 download procedure is resumed. */
1420 (hstat.contlen == -1 || local_size == hstat.contlen))
1422 logprintf (LOG_VERBOSE, _("\
1423 Server file no newer than local file `%s' -- not retrieving.\n\n"),
1426 free(filename_plus_orig_suffix);/*must precede every return!*/
1429 else if (tml >= tmr)
1430 logprintf (LOG_VERBOSE, _("\
1431 The sizes do not match (local %ld) -- retrieving.\n"), local_size);
1433 logputs (LOG_VERBOSE,
1434 _("Remote file is newer, retrieving.\n"));
1440 && (tmr != (time_t) (-1))
1442 && ((hstat.len == hstat.contlen) ||
1443 ((hstat.res == 0) &&
1444 ((hstat.contlen == -1) ||
1445 (hstat.len >= hstat.contlen && !opt.kill_longer)))))
1447 touch (u->local, tmr);
1449 /* End of time-stamping section. */
1453 logprintf (LOG_NOTQUIET, "%d %s\n\n", hstat.statcode, hstat.error);
1454 free(filename_plus_orig_suffix); /* must precede every return! */
1458 /* It is now safe to free the remainder of hstat, since the
1459 strings within it will no longer be used. */
1462 tmrate = rate (hstat.len - hstat.restval, hstat.dltime);
1464 if (hstat.len == hstat.contlen)
1468 logprintf (LOG_VERBOSE,
1469 _("%s (%s) - `%s' saved [%ld/%ld]\n\n"),
1470 tms, tmrate, locf, hstat.len, hstat.contlen);
1471 logprintf (LOG_NONVERBOSE,
1472 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1473 tms, u->url, hstat.len, hstat.contlen, locf, count);
1476 downloaded_increase (hstat.len);
1478 /* Remember that we downloaded the file for later ".orig" code. */
1479 if (*dt & ADDED_HTML_EXTENSION)
1480 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1482 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1484 free(filename_plus_orig_suffix); /* must precede every return! */
1487 else if (hstat.res == 0) /* No read error */
1489 if (hstat.contlen == -1) /* We don't know how much we were supposed
1490 to get, so assume we succeeded. */
1494 logprintf (LOG_VERBOSE,
1495 _("%s (%s) - `%s' saved [%ld]\n\n"),
1496 tms, tmrate, locf, hstat.len);
1497 logprintf (LOG_NONVERBOSE,
1498 "%s URL:%s [%ld] -> \"%s\" [%d]\n",
1499 tms, u->url, hstat.len, locf, count);
1502 downloaded_increase (hstat.len);
1504 /* Remember that we downloaded the file for later ".orig" code. */
1505 if (*dt & ADDED_HTML_EXTENSION)
1506 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1508 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1510 free(filename_plus_orig_suffix); /* must precede every return! */
1513 else if (hstat.len < hstat.contlen) /* meaning we lost the
1514 connection too soon */
1516 logprintf (LOG_VERBOSE,
1517 _("%s (%s) - Connection closed at byte %ld. "),
1518 tms, tmrate, hstat.len);
1519 printwhat (count, opt.ntry);
1522 else if (!opt.kill_longer) /* meaning we got more than expected */
1524 logprintf (LOG_VERBOSE,
1525 _("%s (%s) - `%s' saved [%ld/%ld])\n\n"),
1526 tms, tmrate, locf, hstat.len, hstat.contlen);
1527 logprintf (LOG_NONVERBOSE,
1528 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1529 tms, u->url, hstat.len, hstat.contlen, locf, count);
1531 downloaded_increase (hstat.len);
1533 /* Remember that we downloaded the file for later ".orig" code. */
1534 if (*dt & ADDED_HTML_EXTENSION)
1535 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1537 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1539 free(filename_plus_orig_suffix); /* must precede every return! */
1542 else /* the same, but not accepted */
1544 logprintf (LOG_VERBOSE,
1545 _("%s (%s) - Connection closed at byte %ld/%ld. "),
1546 tms, tmrate, hstat.len, hstat.contlen);
1547 printwhat (count, opt.ntry);
1551 else /* now hstat.res can only be -1 */
1553 if (hstat.contlen == -1)
1555 logprintf (LOG_VERBOSE,
1556 _("%s (%s) - Read error at byte %ld (%s)."),
1557 tms, tmrate, hstat.len, strerror (errno));
1558 printwhat (count, opt.ntry);
1561 else /* hstat.res == -1 and contlen is given */
1563 logprintf (LOG_VERBOSE,
1564 _("%s (%s) - Read error at byte %ld/%ld (%s). "),
1565 tms, tmrate, hstat.len, hstat.contlen,
1567 printwhat (count, opt.ntry);
1574 while (!opt.ntry || (count < opt.ntry));
1575 free(filename_plus_orig_suffix); /* must precede every return! */
1579 /* Converts struct tm to time_t, assuming the data in tm is UTC rather
1580 than local timezone (mktime assumes the latter).
1582 Contributed by Roger Beeman <beeman@cisco.com>, with the help of
1583 Mark Baushke <mdb@cisco.com> and the rest of the Gurus at CISCO. */
1585 mktime_from_utc (struct tm *t)
1592 tb = mktime (gmtime (&tl));
1593 return (tl <= tb ? (tl + (tl - tb)) : (tl - (tb - tl)));
1596 /* Check whether the result of strptime() indicates success.
1597 strptime() returns the pointer to how far it got to in the string.
1598 The processing has been successful if the string is at `GMT' or
1599 `+X', or at the end of the string.
1601 In extended regexp parlance, the function returns 1 if P matches
1602 "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (a valid result of
1603 strptime()) is considered a failure and 0 is returned. */
1609 while (ISSPACE (*p))
1612 || (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
1613 || ((p[0] == '+' || p[1] == '-') && ISDIGIT (p[1])))
1619 /* Convert TIME_STRING time to time_t. TIME_STRING can be in any of
1620 the three formats RFC2068 allows the HTTP servers to emit --
1621 RFC1123-date, RFC850-date or asctime-date. Timezones are ignored,
1624 We use strptime() to recognize various dates, which makes it a
1625 little bit slacker than the RFC1123/RFC850/asctime (e.g. it always
1626 allows shortened dates and months, one-digit days, etc.). It also
1627 allows more than one space anywhere where the specs require one SP.
1628 The routine should probably be even more forgiving (as recommended
1629 by RFC2068), but I do not have the time to write one.
1631 Return the computed time_t representation, or -1 if all the
1634 Needless to say, what we *really* need here is something like
1635 Marcus Hennecke's atotm(), which is forgiving, fast, to-the-point,
1636 and does not use strptime(). atotm() is to be found in the sources
1637 of `phttpd', a little-known HTTP server written by Peter Erikson. */
1639 http_atotm (char *time_string)
1643 /* Roger Beeman says: "This function dynamically allocates struct tm
1644 t, but does no initialization. The only field that actually
1645 needs initialization is tm_isdst, since the others will be set by
1646 strptime. Since strptime does not set tm_isdst, it will return
1647 the data structure with whatever data was in tm_isdst to begin
1648 with. For those of us in timezones where DST can occur, there
1649 can be a one hour shift depending on the previous contents of the
1650 data area where the data structure is allocated." */
1653 /* Note that under foreign locales Solaris strptime() fails to
1654 recognize English dates, which renders this function useless. I
1655 assume that other non-GNU strptime's are plagued by the same
1656 disease. We solve this by setting only LC_MESSAGES in
1657 i18n_initialize(), instead of LC_ALL.
1659 Another solution could be to temporarily set locale to C, invoke
1660 strptime(), and restore it back. This is slow and dirty,
1661 however, and locale support other than LC_MESSAGES can mess other
1662 things, so I rather chose to stick with just setting LC_MESSAGES.
1664 Also note that none of this is necessary under GNU strptime(),
1665 because it recognizes both international and local dates. */
1667 /* NOTE: We don't use `%n' for white space, as OSF's strptime uses
1668 it to eat all white space up to (and including) a newline, and
1669 the function fails if there is no newline (!).
1671 Let's hope all strptime() implementations use ` ' to skip *all*
1672 whitespace instead of just one (it works that way on all the
1673 systems I've tested it on). */
1675 /* RFC1123: Thu, 29 Jan 1998 22:12:57 */
1676 if (check_end (strptime (time_string, "%a, %d %b %Y %T", &t)))
1677 return mktime_from_utc (&t);
1678 /* RFC850: Thu, 29-Jan-98 22:12:57 */
1679 if (check_end (strptime (time_string, "%a, %d-%b-%y %T", &t)))
1680 return mktime_from_utc (&t);
1681 /* asctime: Thu Jan 29 22:12:57 1998 */
1682 if (check_end (strptime (time_string, "%a %b %d %T %Y", &t)))
1683 return mktime_from_utc (&t);
1688 /* Authorization support: We support two authorization schemes:
1690 * `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
1692 * `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
1693 consisting of answering to the server's challenge with the proper
1696 /* How many bytes it will take to store LEN bytes in base64. */
1697 #define BASE64_LENGTH(len) (4 * (((len) + 2) / 3))
1699 /* Encode the string S of length LENGTH to base64 format and place it
1700 to STORE. STORE will be 0-terminated, and must point to a writable
1701 buffer of at least 1+BASE64_LENGTH(length) bytes. */
1703 base64_encode (const char *s, char *store, int length)
1705 /* Conversion table. */
1706 static char tbl[64] = {
1707 'A','B','C','D','E','F','G','H',
1708 'I','J','K','L','M','N','O','P',
1709 'Q','R','S','T','U','V','W','X',
1710 'Y','Z','a','b','c','d','e','f',
1711 'g','h','i','j','k','l','m','n',
1712 'o','p','q','r','s','t','u','v',
1713 'w','x','y','z','0','1','2','3',
1714 '4','5','6','7','8','9','+','/'
1717 unsigned char *p = (unsigned char *)store;
1719 /* Transform the 3x8 bits to 4x6 bits, as required by base64. */
1720 for (i = 0; i < length; i += 3)
1722 *p++ = tbl[s[0] >> 2];
1723 *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
1724 *p++ = tbl[((s[1] & 0xf) << 2) + (s[2] >> 6)];
1725 *p++ = tbl[s[2] & 0x3f];
1728 /* Pad the result if necessary... */
1729 if (i == length + 1)
1731 else if (i == length + 2)
1732 *(p - 1) = *(p - 2) = '=';
1733 /* ...and zero-terminate it. */
1737 /* Create the authentication header contents for the `Basic' scheme.
1738 This is done by encoding the string `USER:PASS' in base64 and
1739 prepending `HEADER: Basic ' to it. */
1741 basic_authentication_encode (const char *user, const char *passwd,
1744 char *t1, *t2, *res;
1745 int len1 = strlen (user) + 1 + strlen (passwd);
1746 int len2 = BASE64_LENGTH (len1);
1748 t1 = (char *)alloca (len1 + 1);
1749 sprintf (t1, "%s:%s", user, passwd);
1750 t2 = (char *)alloca (1 + len2);
1751 base64_encode (t1, t2, len1);
1752 res = (char *)malloc (len2 + 11 + strlen (header));
1753 sprintf (res, "%s: Basic %s\r\n", header, t2);
1759 /* Parse HTTP `WWW-Authenticate:' header. AU points to the beginning
1760 of a field in such a header. If the field is the one specified by
1761 ATTR_NAME ("realm", "opaque", and "nonce" are used by the current
1762 digest authorization code), extract its value in the (char*)
1763 variable pointed by RET. Returns negative on a malformed header,
1764 or number of bytes that have been parsed by this call. */
1766 extract_header_attr (const char *au, const char *attr_name, char **ret)
1768 const char *cp, *ep;
1772 if (strncmp (cp, attr_name, strlen (attr_name)) == 0)
1774 cp += strlen (attr_name);
1777 cp += skip_lws (cp);
1782 cp += skip_lws (cp);
1787 for (ep = cp; *ep && *ep != '\"'; ep++)
1792 *ret = strdupdelim (cp, ep);
1799 /* Response value needs to be in lowercase, so we cannot use HEXD2ASC
1800 from url.h. See RFC 2069 2.1.2 for the syntax of response-digest. */
1801 #define HEXD2asc(x) (((x) < 10) ? ((x) + '0') : ((x) - 10 + 'a'))
1803 /* Dump the hexadecimal representation of HASH to BUF. HASH should be
1804 an array of 16 bytes containing the hash keys, and BUF should be a
1805 buffer of 33 writable characters (32 for hex digits plus one for
1806 zero termination). */
1808 dump_hash (unsigned char *buf, const unsigned char *hash)
1812 for (i = 0; i < MD5_HASHLEN; i++, hash++)
1814 *buf++ = HEXD2asc (*hash >> 4);
1815 *buf++ = HEXD2asc (*hash & 0xf);
1820 /* Take the line apart to find the challenge, and compose a digest
1821 authorization header. See RFC2069 section 2.1.2. */
1823 digest_authentication_encode (const char *au, const char *user,
1824 const char *passwd, const char *method,
1827 static char *realm, *opaque, *nonce;
1832 { "realm", &realm },
1833 { "opaque", &opaque },
1838 realm = opaque = nonce = NULL;
1840 au += 6; /* skip over `Digest' */
1845 au += skip_lws (au);
1846 for (i = 0; i < ARRAY_SIZE (options); i++)
1848 int skip = extract_header_attr (au, options[i].name,
1849 options[i].variable);
1853 FREE_MAYBE (opaque);
1863 if (i == ARRAY_SIZE (options))
1865 while (*au && *au != '=')
1869 au += skip_lws (au);
1873 while (*au && *au != '\"')
1880 while (*au && *au != ',')
1885 if (!realm || !nonce || !user || !passwd || !path || !method)
1888 FREE_MAYBE (opaque);
1893 /* Calculate the digest value. */
1896 unsigned char hash[MD5_HASHLEN];
1897 unsigned char a1buf[MD5_HASHLEN * 2 + 1], a2buf[MD5_HASHLEN * 2 + 1];
1898 unsigned char response_digest[MD5_HASHLEN * 2 + 1];
1900 /* A1BUF = H(user ":" realm ":" password) */
1901 md5_init_ctx (&ctx);
1902 md5_process_bytes (user, strlen (user), &ctx);
1903 md5_process_bytes (":", 1, &ctx);
1904 md5_process_bytes (realm, strlen (realm), &ctx);
1905 md5_process_bytes (":", 1, &ctx);
1906 md5_process_bytes (passwd, strlen (passwd), &ctx);
1907 md5_finish_ctx (&ctx, hash);
1908 dump_hash (a1buf, hash);
1910 /* A2BUF = H(method ":" path) */
1911 md5_init_ctx (&ctx);
1912 md5_process_bytes (method, strlen (method), &ctx);
1913 md5_process_bytes (":", 1, &ctx);
1914 md5_process_bytes (path, strlen (path), &ctx);
1915 md5_finish_ctx (&ctx, hash);
1916 dump_hash (a2buf, hash);
1918 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */
1919 md5_init_ctx (&ctx);
1920 md5_process_bytes (a1buf, MD5_HASHLEN * 2, &ctx);
1921 md5_process_bytes (":", 1, &ctx);
1922 md5_process_bytes (nonce, strlen (nonce), &ctx);
1923 md5_process_bytes (":", 1, &ctx);
1924 md5_process_bytes (a2buf, MD5_HASHLEN * 2, &ctx);
1925 md5_finish_ctx (&ctx, hash);
1926 dump_hash (response_digest, hash);
1928 res = (char*) xmalloc (strlen (user)
1933 + 2 * MD5_HASHLEN /*strlen (response_digest)*/
1934 + (opaque ? strlen (opaque) : 0)
1936 sprintf (res, "Authorization: Digest \
1937 username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
1938 user, realm, nonce, path, response_digest);
1941 char *p = res + strlen (res);
1942 strcat (p, ", opaque=\"");
1946 strcat (res, "\r\n");
1950 #endif /* USE_DIGEST */
1953 #define BEGINS_WITH(line, string_constant) \
1954 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
1955 && (ISSPACE (line[sizeof (string_constant) - 1]) \
1956 || !line[sizeof (string_constant) - 1]))
1959 known_authentication_scheme_p (const char *au)
1961 return BEGINS_WITH (au, "Basic")
1962 || BEGINS_WITH (au, "Digest")
1963 || BEGINS_WITH (au, "NTLM");
1968 /* Create the HTTP authorization request header. When the
1969 `WWW-Authenticate' response header is seen, according to the
1970 authorization scheme specified in that header (`Basic' and `Digest'
1971 are supported by the current implementation), produce an
1972 appropriate HTTP authorization request header. */
1974 create_authorization_line (const char *au, const char *user,
1975 const char *passwd, const char *method,
1978 char *wwwauth = NULL;
1980 if (!strncasecmp (au, "Basic", 5))
1981 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
1982 if (!strncasecmp (au, "NTLM", 4))
1983 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
1985 else if (!strncasecmp (au, "Digest", 6))
1986 wwwauth = digest_authentication_encode (au, user, passwd, method, path);
1987 #endif /* USE_DIGEST */