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 /* Retrieve a document through HTTP protocol. It recognizes status
426 code, and correctly handles redirections. It closes the network
427 socket. If it receives an error from the functions below it, it
428 will print it if there is enough information to do so (almost
429 always), returning the error to the caller (i.e. http_loop).
431 Various HTTP parameters are stored to hs. Although it parses the
432 response code correctly, it is not used in a sane way. The caller
435 If u->proxy is non-NULL, the URL u will be taken as a proxy URL,
436 and u->proxy->url will be given to the proxy server (bad naming,
439 gethttp (struct urlinfo *u, struct http_stat *hs, int *dt)
441 char *request, *type, *command, *path;
443 char *pragma_h, *referer, *useragent, *range, *wwwauth, *remhost;
444 char *authenticate_h;
448 char *request_keep_alive;
450 int sock, hcount, num_written, all_length, remport, statcode;
451 long contlen, contrange;
455 int auth_tried_already;
458 /* Whether this connection will be kept alive after the HTTP request
462 /* Flags that detect the two ways of specifying HTTP keep-alive
464 int http_keep_alive_1, http_keep_alive_2;
466 /* Whether keep-alive should be inhibited. */
467 int inhibit_keep_alive;
469 if (!(*dt & HEAD_ONLY))
470 /* If we're doing a GET on the URL, as opposed to just a HEAD, we need to
471 know the local filename so we can save to it. */
472 assert (u->local != NULL);
475 auth_tried_already = 0;
477 inhibit_keep_alive = (u->proxy != NULL);
480 /* We need to come back here when the initial attempt to retrieve
481 without authorization header fails. */
483 http_keep_alive_1 = http_keep_alive_2 = 0;
485 /* Initialize certain elements of struct http_stat. */
490 hs->remote_time = NULL;
493 /* Which structure to use to retrieve the original URL data. */
499 /* First: establish the connection. */
500 if (inhibit_keep_alive
501 || !persistent_available_p (u->host, u->port))
503 logprintf (LOG_VERBOSE, _("Connecting to %s:%hu... "), u->host, u->port);
504 err = make_connection (&sock, u->host, u->port);
508 logputs (LOG_VERBOSE, "\n");
509 logprintf (LOG_NOTQUIET, "%s: %s.\n", u->host, herrmsg (h_errno));
513 logputs (LOG_VERBOSE, "\n");
514 logprintf (LOG_NOTQUIET, "socket: %s\n", strerror (errno));
518 logputs (LOG_VERBOSE, "\n");
519 logprintf (LOG_NOTQUIET,
520 _("Connection to %s:%hu refused.\n"), u->host, u->port);
524 logputs (LOG_VERBOSE, "\n");
525 logprintf (LOG_NOTQUIET, "connect: %s\n", strerror (errno));
530 /* Everything is fine! */
531 logputs (LOG_VERBOSE, _("connected!\n"));
540 logprintf (LOG_VERBOSE, _("Reusing connection to %s:%hu.\n"), u->host, u->port);
541 /* #### pc_last_fd should be accessed through an accessor
544 DEBUGP (("Reusing fd %d.\n", sock));
548 path = u->proxy->url;
552 command = (*dt & HEAD_ONLY) ? "HEAD" : "GET";
556 referer = (char *)alloca (9 + strlen (ou->referer) + 3);
557 sprintf (referer, "Referer: %s\r\n", ou->referer);
559 if (*dt & SEND_NOCACHE)
560 pragma_h = "Pragma: no-cache\r\n";
565 range = (char *)alloca (13 + numdigit (hs->restval) + 4);
566 /* Gag me! Some servers (e.g. WebSitePro) have been known to
567 respond to the following `Range' format by generating a
568 multipart/x-byte-ranges MIME document! This MIME type was
569 present in an old draft of the byteranges specification.
570 HTTP/1.1 specifies a multipart/byte-ranges MIME type, but
571 only if multiple non-overlapping ranges are requested --
572 which Wget never does. */
573 sprintf (range, "Range: bytes=%ld-\r\n", hs->restval);
578 STRDUP_ALLOCA (useragent, opt.useragent);
581 useragent = (char *)alloca (10 + strlen (version_string));
582 sprintf (useragent, "Wget/%s", version_string);
584 /* Construct the authentication, if userid is present. */
587 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
588 user = user ? user : opt.http_user;
589 passwd = passwd ? passwd : opt.http_passwd;
592 if (authenticate_h && user && passwd)
594 wwwauth = create_authorization_line (authenticate_h, user, passwd,
601 char *proxy_user, *proxy_passwd;
602 /* For normal username and password, URL components override
603 command-line/wgetrc parameters. With proxy authentication,
604 it's the reverse, because proxy URLs are normally the
605 "permanent" ones, so command-line args should take
607 if (opt.proxy_user && opt.proxy_passwd)
609 proxy_user = opt.proxy_user;
610 proxy_passwd = opt.proxy_passwd;
614 proxy_user = u->user;
615 proxy_passwd = u->passwd;
617 /* #### This is junky. Can't the proxy request, say, `Digest'
619 if (proxy_user && proxy_passwd)
620 proxyauth = basic_authentication_encode (proxy_user, proxy_passwd,
621 "Proxy-Authorization");
633 host_port = (char *)alloca (numdigit (remport) + 2);
634 host_port_len = sprintf (host_port, ":%d", remport);
637 if (!inhibit_keep_alive)
638 request_keep_alive = "Connection: Keep-Alive\r\n";
640 request_keep_alive = NULL;
642 /* Allocate the memory for the request. */
643 request = (char *)alloca (strlen (command) + strlen (path)
645 + strlen (remhost) + host_port_len
646 + strlen (HTTP_ACCEPT)
647 + (request_keep_alive
648 ? strlen (request_keep_alive) : 0)
649 + (referer ? strlen (referer) : 0)
650 + (wwwauth ? strlen (wwwauth) : 0)
651 + (proxyauth ? strlen (proxyauth) : 0)
652 + (range ? strlen (range) : 0)
654 + (opt.user_header ? strlen (opt.user_header) : 0)
656 /* Construct the request. */
663 command, path, useragent, remhost,
664 host_port ? host_port : "",
666 request_keep_alive ? request_keep_alive : "",
667 referer ? referer : "",
668 wwwauth ? wwwauth : "",
669 proxyauth ? proxyauth : "",
672 opt.user_header ? opt.user_header : "");
673 DEBUGP (("---request begin---\n%s---request end---\n", request));
674 /* Free the temporary memory. */
675 FREE_MAYBE (wwwauth);
676 FREE_MAYBE (proxyauth);
678 /* Send the request to server. */
679 num_written = iwrite (sock, request, strlen (request));
682 logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"),
684 CLOSE_INVALIDATE (sock);
687 logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
688 u->proxy ? "Proxy" : "HTTP");
689 contlen = contrange = -1;
694 /* Before reading anything, initialize the rbuf. */
695 rbuf_initialize (&rbuf, sock);
699 /* Header-fetching loop. */
707 /* Get the header. */
708 status = header_get (&rbuf, &hdr,
709 /* Disallow continuations for status line. */
710 (hcount == 1 ? HG_NO_CONTINUATIONS : HG_NONE));
712 /* Check for errors. */
713 if (status == HG_EOF && *hdr)
715 /* This used to be an unconditional error, but that was
716 somewhat controversial, because of a large number of
717 broken CGI's that happily "forget" to send the second EOL
718 before closing the connection of a HEAD request.
720 So, the deal is to check whether the header is empty
721 (*hdr is zero if it is); if yes, it means that the
722 previous header was fully retrieved, and that -- most
723 probably -- the request is complete. "...be liberal in
724 what you accept." Oh boy. */
725 logputs (LOG_VERBOSE, "\n");
726 logputs (LOG_NOTQUIET, _("End of file while parsing headers.\n"));
729 FREE_MAYBE (hs->newloc);
730 FREE_MAYBE (all_headers);
731 CLOSE_INVALIDATE (sock);
734 else if (status == HG_ERROR)
736 logputs (LOG_VERBOSE, "\n");
737 logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
741 FREE_MAYBE (hs->newloc);
742 FREE_MAYBE (all_headers);
743 CLOSE_INVALIDATE (sock);
747 /* If the headers are to be saved to a file later, save them to
749 if (opt.save_headers)
751 int lh = strlen (hdr);
752 all_headers = (char *)xrealloc (all_headers, all_length + lh + 2);
753 memcpy (all_headers + all_length, hdr, lh);
755 all_headers[all_length++] = '\n';
756 all_headers[all_length] = '\0';
759 /* Print the header if requested. */
760 if (opt.server_response && hcount != 1)
761 logprintf (LOG_VERBOSE, "\n%d %s", hcount, hdr);
763 /* Check for status line. */
767 /* Parse the first line of server response. */
768 statcode = parse_http_status_line (hdr, &error);
769 hs->statcode = statcode;
770 /* Store the descriptive response. */
771 if (statcode == -1) /* malformed response */
773 /* A common reason for "malformed response" error is the
774 case when no data was actually received. Handle this
777 hs->error = xstrdup (_("No data received"));
779 hs->error = xstrdup (_("Malformed status line"));
784 hs->error = xstrdup (_("(no description)"));
786 hs->error = xstrdup (error);
793 logprintf (LOG_VERBOSE, "%d %s", statcode, error);
798 /* Exit on empty header. */
805 /* Try getting content-length. */
806 if (contlen == -1 && !opt.ignore_length)
807 if (header_process (hdr, "Content-Length", header_extract_number,
810 /* Try getting content-type. */
812 if (header_process (hdr, "Content-Type", http_process_type, &type))
814 /* Try getting location. */
816 if (header_process (hdr, "Location", header_strdup, &hs->newloc))
818 /* Try getting last-modified. */
819 if (!hs->remote_time)
820 if (header_process (hdr, "Last-Modified", header_strdup,
823 /* Try getting www-authentication. */
825 if (header_process (hdr, "WWW-Authenticate", header_strdup,
828 /* Check for accept-ranges header. If it contains the word
829 `none', disable the ranges. */
830 if (*dt & ACCEPTRANGES)
833 if (header_process (hdr, "Accept-Ranges", http_process_none, &nonep))
836 *dt &= ~ACCEPTRANGES;
840 /* Try getting content-range. */
843 struct http_process_range_closure closure;
844 if (header_process (hdr, "Content-Range", http_process_range, &closure))
846 contrange = closure.first_byte_pos;
850 /* Check for keep-alive related responses. */
851 if (!inhibit_keep_alive)
853 /* Check for the `Keep-Alive' header. */
854 if (!http_keep_alive_1)
856 if (header_process (hdr, "Keep-Alive", header_exists,
860 /* Check for `Connection: Keep-Alive'. */
861 if (!http_keep_alive_2)
863 if (header_process (hdr, "Connection", http_process_connection,
872 logputs (LOG_VERBOSE, "\n");
875 && (http_keep_alive_1 || http_keep_alive_2))
877 assert (inhibit_keep_alive == 0);
881 /* The server has promised that it will not close the connection
882 when we're done. This means that we can register it. */
883 register_persistent (u->host, u->port, sock);
885 if ((statcode == HTTP_STATUS_UNAUTHORIZED)
888 /* Authorization is required. */
893 if (auth_tried_already)
895 /* If we have tried it already, then there is not point
897 logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
898 free (authenticate_h);
901 else if (!known_authentication_scheme_p (authenticate_h))
903 free (authenticate_h);
904 logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
909 auth_tried_already = 1;
913 /* We do not need this anymore. */
916 free (authenticate_h);
917 authenticate_h = NULL;
920 /* 20x responses are counted among successful by default. */
921 if (H_20X (statcode))
924 if (type && !strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S)))
927 /* We don't assume text/html by default. */
930 if (opt.html_extension && (*dt & TEXTHTML))
931 /* -E / --html-extension / html_extension = on was specified, and this is a
932 text/html file. If some case-insensitive variation on ".htm[l]" isn't
933 already the file's suffix, tack on ".html". */
935 char* last_period_in_local_filename = strrchr(u->local, '.');
937 if (last_period_in_local_filename == NULL ||
938 !(strcasecmp(last_period_in_local_filename, ".htm") == EQ ||
939 strcasecmp(last_period_in_local_filename, ".html") == EQ))
941 size_t local_filename_len = strlen(u->local);
943 u->local = xrealloc(u->local, local_filename_len + sizeof(".html"));
944 strcpy(u->local + local_filename_len, ".html");
946 *dt |= ADDED_HTML_EXTENSION;
952 else if (contrange != hs->restval ||
953 (H_PARTIAL (statcode) && contrange == -1))
955 /* This means the whole request was somehow misunderstood by the
958 FREE_MAYBE (hs->newloc);
959 FREE_MAYBE (all_headers);
960 CLOSE_INVALIDATE (sock);
967 contlen += contrange;
969 contrange = -1; /* If conent-length was not sent,
970 content-range will be ignored. */
972 hs->contlen = contlen;
974 /* Return if redirected. */
975 if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES)
977 /* RFC2068 says that in case of the 300 (multiple choices)
978 response, the server can output a preferred URL through
979 `Location' header; otherwise, the request should be treated
980 like GET. So, if the location is set, it will be a
981 redirection; otherwise, just proceed normally. */
982 if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc)
986 logprintf (LOG_VERBOSE,
987 _("Location: %s%s\n"),
988 hs->newloc ? hs->newloc : _("unspecified"),
989 hs->newloc ? _(" [following]") : "");
992 FREE_MAYBE (all_headers);
998 if ((*dt & RETROKF) && !opt.server_response)
1000 /* No need to print this output if the body won't be
1001 downloaded at all, or if the original server response is
1003 logputs (LOG_VERBOSE, _("Length: "));
1006 logputs (LOG_VERBOSE, legible (contlen));
1007 if (contrange != -1)
1008 logprintf (LOG_VERBOSE, _(" (%s to go)"),
1009 legible (contlen - contrange));
1012 logputs (LOG_VERBOSE,
1013 opt.ignore_length ? _("ignored") : _("unspecified"));
1015 logprintf (LOG_VERBOSE, " [%s]\n", type);
1017 logputs (LOG_VERBOSE, "\n");
1021 type = NULL; /* We don't need it any more. */
1023 /* Return if we have no intention of further downloading. */
1024 if (!(*dt & RETROKF) || (*dt & HEAD_ONLY))
1026 /* In case someone cares to look... */
1030 FREE_MAYBE (all_headers);
1031 CLOSE_FINISH (sock);
1032 return RETRFINISHED;
1035 /* Open the local file. */
1038 mkalldirs (u->local);
1040 rotate_backups (u->local);
1041 fp = fopen (u->local, hs->restval ? "ab" : "wb");
1044 logprintf (LOG_NOTQUIET, "%s: %s\n", u->local, strerror (errno));
1045 CLOSE_FINISH (sock);
1046 FREE_MAYBE (all_headers);
1055 /* This will silently fail for streams that don't correspond
1056 to regular files, but that's OK. */
1062 /* #### This confuses the code that checks for file size. There
1063 should be some overhead information. */
1064 if (opt.save_headers)
1065 fwrite (all_headers, 1, all_length, fp);
1067 /* Get the contents of the document. */
1068 hs->res = get_contents (sock, fp, &hs->len, hs->restval,
1069 (contlen != -1 ? contlen : 0),
1071 hs->dltime = elapsed_time ();
1073 /* Close or flush the file. We have to be careful to check for
1074 error here. Checking the result of fwrite() is not enough --
1075 errors could go unnoticed! */
1078 flush_res = fclose (fp);
1080 flush_res = fflush (fp);
1081 if (flush_res == EOF)
1084 FREE_MAYBE (all_headers);
1085 CLOSE_FINISH (sock);
1088 return RETRFINISHED;
1091 /* The genuine HTTP loop! This is the part where the retrieval is
1092 retried, and retried, and retried, and... */
1094 http_loop (struct urlinfo *u, char **newloc, int *dt)
1096 static int first_retrieval = 1;
1099 int use_ts, got_head = 0; /* time-stamping info */
1100 char *filename_plus_orig_suffix;
1101 char *local_filename = NULL;
1102 char *tms, *suf, *locf, *tmrate;
1104 time_t tml = -1, tmr = -1; /* local and remote time-stamps */
1105 long local_size = 0; /* the size of the local file */
1106 size_t filename_len;
1107 struct http_stat hstat; /* HTTP status */
1112 /* Warn on (likely bogus) wildcard usage in HTTP. Don't use
1113 has_wildcards_p because it would also warn on `?', and we know that
1114 shows up in CGI paths a *lot*. */
1115 if (strchr (u->url, '*'))
1116 logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
1118 /* Determine the local filename. */
1120 u->local = url_filename (u->proxy ? u->proxy : u);
1122 if (!opt.output_document)
1125 locf = opt.output_document;
1127 /* Yuck. Multiple returns suck. We need to remember to free() the space we
1128 xmalloc() here before EACH return. This is one reason it's better to set
1129 flags that influence flow control and then return once at the end. */
1130 filename_len = strlen(u->local);
1131 filename_plus_orig_suffix = xmalloc(filename_len + sizeof(".orig"));
1133 if (opt.noclobber && file_exists_p (u->local))
1135 /* If opt.noclobber is turned on and file already exists, do not
1136 retrieve the file */
1137 logprintf (LOG_VERBOSE, _("\
1138 File `%s' already there, will not retrieve.\n"), u->local);
1139 /* If the file is there, we suppose it's retrieved OK. */
1142 /* #### Bogusness alert. */
1143 /* If its suffix is "html" or (yuck!) "htm", we suppose it's
1144 text/html, a harmless lie. */
1145 if (((suf = suffix (u->local)) != NULL)
1146 && (!strcmp (suf, "html") || !strcmp (suf, "htm")))
1149 free(filename_plus_orig_suffix); /* must precede every return! */
1150 /* Another harmless lie: */
1155 if (opt.timestamping)
1157 boolean local_dot_orig_file_exists = FALSE;
1159 if (opt.backup_converted)
1160 /* If -K is specified, we'll act on the assumption that it was specified
1161 last time these files were downloaded as well, and instead of just
1162 comparing local file X against server file X, we'll compare local
1163 file X.orig (if extant, else X) against server file X. If -K
1164 _wasn't_ specified last time, or the server contains files called
1165 *.orig, -N will be back to not operating correctly with -k. */
1167 /* Would a single s[n]printf() call be faster? --dan
1169 It wouldn't. sprintf() is horribly slow. At one point I
1170 profiled Wget, and found that a measurable and
1171 non-negligible amount of time was lost calling sprintf()
1172 in url.c. Replacing sprintf with inline calls to
1173 strcpy() and long_to_string() made a difference.
1175 strcpy(filename_plus_orig_suffix, u->local);
1176 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
1178 /* Try to stat() the .orig file. */
1179 if (stat(filename_plus_orig_suffix, &st) == 0)
1181 local_dot_orig_file_exists = TRUE;
1182 local_filename = filename_plus_orig_suffix;
1186 if (!local_dot_orig_file_exists)
1187 /* Couldn't stat() <file>.orig, so try to stat() <file>. */
1188 if (stat (u->local, &st) == 0)
1189 local_filename = u->local;
1191 if (local_filename != NULL)
1192 /* There was a local file, so we'll check later to see if the version
1193 the server has is the same version we already have, allowing us to
1198 local_size = st.st_size;
1202 /* Reset the counter. */
1204 *dt = 0 | ACCEPTRANGES;
1208 /* Increment the pass counter. */
1210 /* Wait before the retrieval (unless this is the very first
1212 Check if we are retrying or not, wait accordingly - HEH */
1213 if (!first_retrieval && (opt.wait || (count && opt.waitretry)))
1217 if (count<opt.waitretry)
1220 sleep(opt.waitretry);
1225 if (first_retrieval)
1226 first_retrieval = 0;
1227 /* Get the current time string. */
1228 tms = time_str (NULL);
1229 /* Print fetch message, if opt.verbose. */
1232 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
1236 sprintf (tmp, _("(try:%2d)"), count);
1237 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
1238 tms, hurl, tmp, locf);
1240 ws_changetitle (hurl, 1);
1245 /* Default document type is empty. However, if spider mode is
1246 on or time-stamping is employed, HEAD_ONLY commands is
1247 encoded within *dt. */
1248 if (opt.spider || (use_ts && !got_head))
1252 /* Assume no restarting. */
1254 /* Decide whether or not to restart. */
1255 if (((count > 1 && (*dt & ACCEPTRANGES)) || opt.always_rest)
1256 && file_exists_p (u->local))
1257 if (stat (u->local, &st) == 0)
1258 hstat.restval = st.st_size;
1259 /* Decide whether to send the no-cache directive. */
1260 if (u->proxy && (count > 1 || (opt.proxy_cache == 0)))
1261 *dt |= SEND_NOCACHE;
1263 *dt &= ~SEND_NOCACHE;
1265 /* Try fetching the document, or at least its head. :-) */
1266 err = gethttp (u, &hstat, dt);
1268 /* It's unfortunate that wget determines the local filename before finding
1269 out the Content-Type of the file. Barring a major restructuring of the
1270 code, we need to re-set locf here, since gethttp() may have xrealloc()d
1271 u->local to tack on ".html". */
1272 if (!opt.output_document)
1275 locf = opt.output_document;
1278 tms = time_str (NULL);
1279 /* Get the new location (with or without the redirection). */
1281 *newloc = xstrdup (hstat.newloc);
1284 case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
1285 case CONERROR: case READERR: case WRITEFAILED:
1287 /* Non-fatal errors continue executing the loop, which will
1288 bring them to "while" statement at the end, to judge
1289 whether the number of tries was exceeded. */
1291 printwhat (count, opt.ntry);
1294 case HOSTERR: case CONREFUSED: case PROXERR: case AUTHFAILED:
1295 /* Fatal errors just return from the function. */
1297 free(filename_plus_orig_suffix); /* must precede every return! */
1300 case FWRITEERR: case FOPENERR:
1301 /* Another fatal error. */
1302 logputs (LOG_VERBOSE, "\n");
1303 logprintf (LOG_NOTQUIET, _("Cannot write to `%s' (%s).\n"),
1304 u->local, strerror (errno));
1306 free(filename_plus_orig_suffix); /* must precede every return! */
1310 /* Return the new location to the caller. */
1313 logprintf (LOG_NOTQUIET,
1314 _("ERROR: Redirection (%d) without location.\n"),
1316 free(filename_plus_orig_suffix); /* must precede every return! */
1320 free(filename_plus_orig_suffix); /* must precede every return! */
1324 /* Deal with you later. */
1327 /* All possibilities should have been exhausted. */
1330 if (!(*dt & RETROKF))
1334 /* #### Ugly ugly ugly! */
1335 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
1336 logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
1339 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
1340 tms, hstat.statcode, hstat.error);
1341 logputs (LOG_VERBOSE, "\n");
1343 free(filename_plus_orig_suffix); /* must precede every return! */
1347 /* Did we get the time-stamp? */
1350 if (opt.timestamping && !hstat.remote_time)
1352 logputs (LOG_NOTQUIET, _("\
1353 Last-modified header missing -- time-stamps turned off.\n"));
1355 else if (hstat.remote_time)
1357 /* Convert the date-string into struct tm. */
1358 tmr = http_atotm (hstat.remote_time);
1359 if (tmr == (time_t) (-1))
1360 logputs (LOG_VERBOSE, _("\
1361 Last-modified header invalid -- time-stamp ignored.\n"));
1365 /* The time-stamping section. */
1370 use_ts = 0; /* no more time-stamping */
1371 count = 0; /* the retrieve count for HEAD is
1373 if (hstat.remote_time && tmr != (time_t) (-1))
1375 /* Now time-stamping can be used validly. Time-stamping
1376 means that if the sizes of the local and remote file
1377 match, and local file is newer than the remote file,
1378 it will not be retrieved. Otherwise, the normal
1379 download procedure is resumed. */
1381 (hstat.contlen == -1 || local_size == hstat.contlen))
1383 logprintf (LOG_VERBOSE, _("\
1384 Server file no newer than local file `%s' -- not retrieving.\n\n"),
1387 free(filename_plus_orig_suffix);/*must precede every return!*/
1390 else if (tml >= tmr)
1391 logprintf (LOG_VERBOSE, _("\
1392 The sizes do not match (local %ld) -- retrieving.\n"), local_size);
1394 logputs (LOG_VERBOSE,
1395 _("Remote file is newer, retrieving.\n"));
1401 && (tmr != (time_t) (-1))
1403 && ((hstat.len == hstat.contlen) ||
1404 ((hstat.res == 0) &&
1405 ((hstat.contlen == -1) ||
1406 (hstat.len >= hstat.contlen && !opt.kill_longer)))))
1408 touch (u->local, tmr);
1410 /* End of time-stamping section. */
1414 logprintf (LOG_NOTQUIET, "%d %s\n\n", hstat.statcode, hstat.error);
1415 free(filename_plus_orig_suffix); /* must precede every return! */
1419 /* It is now safe to free the remainder of hstat, since the
1420 strings within it will no longer be used. */
1423 tmrate = rate (hstat.len - hstat.restval, hstat.dltime);
1425 if (hstat.len == hstat.contlen)
1429 logprintf (LOG_VERBOSE,
1430 _("%s (%s) - `%s' saved [%ld/%ld]\n\n"),
1431 tms, tmrate, locf, hstat.len, hstat.contlen);
1432 logprintf (LOG_NONVERBOSE,
1433 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1434 tms, u->url, hstat.len, hstat.contlen, locf, count);
1437 downloaded_increase (hstat.len);
1439 /* Remember that we downloaded the file for later ".orig" code. */
1440 if (*dt & ADDED_HTML_EXTENSION)
1441 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1443 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1445 free(filename_plus_orig_suffix); /* must precede every return! */
1448 else if (hstat.res == 0) /* No read error */
1450 if (hstat.contlen == -1) /* We don't know how much we were supposed
1451 to get, so assume we succeeded. */
1455 logprintf (LOG_VERBOSE,
1456 _("%s (%s) - `%s' saved [%ld]\n\n"),
1457 tms, tmrate, locf, hstat.len);
1458 logprintf (LOG_NONVERBOSE,
1459 "%s URL:%s [%ld] -> \"%s\" [%d]\n",
1460 tms, u->url, hstat.len, locf, count);
1463 downloaded_increase (hstat.len);
1465 /* Remember that we downloaded the file for later ".orig" code. */
1466 if (*dt & ADDED_HTML_EXTENSION)
1467 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1469 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1471 free(filename_plus_orig_suffix); /* must precede every return! */
1474 else if (hstat.len < hstat.contlen) /* meaning we lost the
1475 connection too soon */
1477 logprintf (LOG_VERBOSE,
1478 _("%s (%s) - Connection closed at byte %ld. "),
1479 tms, tmrate, hstat.len);
1480 printwhat (count, opt.ntry);
1483 else if (!opt.kill_longer) /* meaning we got more than expected */
1485 logprintf (LOG_VERBOSE,
1486 _("%s (%s) - `%s' saved [%ld/%ld])\n\n"),
1487 tms, tmrate, locf, hstat.len, hstat.contlen);
1488 logprintf (LOG_NONVERBOSE,
1489 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1490 tms, u->url, hstat.len, hstat.contlen, locf, count);
1492 downloaded_increase (hstat.len);
1494 /* Remember that we downloaded the file for later ".orig" code. */
1495 if (*dt & ADDED_HTML_EXTENSION)
1496 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1498 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1500 free(filename_plus_orig_suffix); /* must precede every return! */
1503 else /* the same, but not accepted */
1505 logprintf (LOG_VERBOSE,
1506 _("%s (%s) - Connection closed at byte %ld/%ld. "),
1507 tms, tmrate, hstat.len, hstat.contlen);
1508 printwhat (count, opt.ntry);
1512 else /* now hstat.res can only be -1 */
1514 if (hstat.contlen == -1)
1516 logprintf (LOG_VERBOSE,
1517 _("%s (%s) - Read error at byte %ld (%s)."),
1518 tms, tmrate, hstat.len, strerror (errno));
1519 printwhat (count, opt.ntry);
1522 else /* hstat.res == -1 and contlen is given */
1524 logprintf (LOG_VERBOSE,
1525 _("%s (%s) - Read error at byte %ld/%ld (%s). "),
1526 tms, tmrate, hstat.len, hstat.contlen,
1528 printwhat (count, opt.ntry);
1535 while (!opt.ntry || (count < opt.ntry));
1536 free(filename_plus_orig_suffix); /* must precede every return! */
1540 /* Converts struct tm to time_t, assuming the data in tm is UTC rather
1541 than local timezone (mktime assumes the latter).
1543 Contributed by Roger Beeman <beeman@cisco.com>, with the help of
1544 Mark Baushke <mdb@cisco.com> and the rest of the Gurus at CISCO. */
1546 mktime_from_utc (struct tm *t)
1553 tb = mktime (gmtime (&tl));
1554 return (tl <= tb ? (tl + (tl - tb)) : (tl - (tb - tl)));
1557 /* Check whether the result of strptime() indicates success.
1558 strptime() returns the pointer to how far it got to in the string.
1559 The processing has been successful if the string is at `GMT' or
1560 `+X', or at the end of the string.
1562 In extended regexp parlance, the function returns 1 if P matches
1563 "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (a valid result of
1564 strptime()) is considered a failure and 0 is returned. */
1570 while (ISSPACE (*p))
1573 || (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
1574 || ((p[0] == '+' || p[1] == '-') && ISDIGIT (p[1])))
1580 /* Convert TIME_STRING time to time_t. TIME_STRING can be in any of
1581 the three formats RFC2068 allows the HTTP servers to emit --
1582 RFC1123-date, RFC850-date or asctime-date. Timezones are ignored,
1585 We use strptime() to recognize various dates, which makes it a
1586 little bit slacker than the RFC1123/RFC850/asctime (e.g. it always
1587 allows shortened dates and months, one-digit days, etc.). It also
1588 allows more than one space anywhere where the specs require one SP.
1589 The routine should probably be even more forgiving (as recommended
1590 by RFC2068), but I do not have the time to write one.
1592 Return the computed time_t representation, or -1 if all the
1595 Needless to say, what we *really* need here is something like
1596 Marcus Hennecke's atotm(), which is forgiving, fast, to-the-point,
1597 and does not use strptime(). atotm() is to be found in the sources
1598 of `phttpd', a little-known HTTP server written by Peter Erikson. */
1600 http_atotm (char *time_string)
1604 /* Roger Beeman says: "This function dynamically allocates struct tm
1605 t, but does no initialization. The only field that actually
1606 needs initialization is tm_isdst, since the others will be set by
1607 strptime. Since strptime does not set tm_isdst, it will return
1608 the data structure with whatever data was in tm_isdst to begin
1609 with. For those of us in timezones where DST can occur, there
1610 can be a one hour shift depending on the previous contents of the
1611 data area where the data structure is allocated." */
1614 /* Note that under foreign locales Solaris strptime() fails to
1615 recognize English dates, which renders this function useless. I
1616 assume that other non-GNU strptime's are plagued by the same
1617 disease. We solve this by setting only LC_MESSAGES in
1618 i18n_initialize(), instead of LC_ALL.
1620 Another solution could be to temporarily set locale to C, invoke
1621 strptime(), and restore it back. This is slow and dirty,
1622 however, and locale support other than LC_MESSAGES can mess other
1623 things, so I rather chose to stick with just setting LC_MESSAGES.
1625 Also note that none of this is necessary under GNU strptime(),
1626 because it recognizes both international and local dates. */
1628 /* NOTE: We don't use `%n' for white space, as OSF's strptime uses
1629 it to eat all white space up to (and including) a newline, and
1630 the function fails if there is no newline (!).
1632 Let's hope all strptime() implementations use ` ' to skip *all*
1633 whitespace instead of just one (it works that way on all the
1634 systems I've tested it on). */
1636 /* RFC1123: Thu, 29 Jan 1998 22:12:57 */
1637 if (check_end (strptime (time_string, "%a, %d %b %Y %T", &t)))
1638 return mktime_from_utc (&t);
1639 /* RFC850: Thu, 29-Jan-98 22:12:57 */
1640 if (check_end (strptime (time_string, "%a, %d-%b-%y %T", &t)))
1641 return mktime_from_utc (&t);
1642 /* asctime: Thu Jan 29 22:12:57 1998 */
1643 if (check_end (strptime (time_string, "%a %b %d %T %Y", &t)))
1644 return mktime_from_utc (&t);
1649 /* Authorization support: We support two authorization schemes:
1651 * `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
1653 * `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
1654 consisting of answering to the server's challenge with the proper
1657 /* How many bytes it will take to store LEN bytes in base64. */
1658 #define BASE64_LENGTH(len) (4 * (((len) + 2) / 3))
1660 /* Encode the string S of length LENGTH to base64 format and place it
1661 to STORE. STORE will be 0-terminated, and must point to a writable
1662 buffer of at least 1+BASE64_LENGTH(length) bytes. */
1664 base64_encode (const char *s, char *store, int length)
1666 /* Conversion table. */
1667 static char tbl[64] = {
1668 'A','B','C','D','E','F','G','H',
1669 'I','J','K','L','M','N','O','P',
1670 'Q','R','S','T','U','V','W','X',
1671 'Y','Z','a','b','c','d','e','f',
1672 'g','h','i','j','k','l','m','n',
1673 'o','p','q','r','s','t','u','v',
1674 'w','x','y','z','0','1','2','3',
1675 '4','5','6','7','8','9','+','/'
1678 unsigned char *p = (unsigned char *)store;
1680 /* Transform the 3x8 bits to 4x6 bits, as required by base64. */
1681 for (i = 0; i < length; i += 3)
1683 *p++ = tbl[s[0] >> 2];
1684 *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
1685 *p++ = tbl[((s[1] & 0xf) << 2) + (s[2] >> 6)];
1686 *p++ = tbl[s[2] & 0x3f];
1689 /* Pad the result if necessary... */
1690 if (i == length + 1)
1692 else if (i == length + 2)
1693 *(p - 1) = *(p - 2) = '=';
1694 /* ...and zero-terminate it. */
1698 /* Create the authentication header contents for the `Basic' scheme.
1699 This is done by encoding the string `USER:PASS' in base64 and
1700 prepending `HEADER: Basic ' to it. */
1702 basic_authentication_encode (const char *user, const char *passwd,
1705 char *t1, *t2, *res;
1706 int len1 = strlen (user) + 1 + strlen (passwd);
1707 int len2 = BASE64_LENGTH (len1);
1709 t1 = (char *)alloca (len1 + 1);
1710 sprintf (t1, "%s:%s", user, passwd);
1711 t2 = (char *)alloca (1 + len2);
1712 base64_encode (t1, t2, len1);
1713 res = (char *)malloc (len2 + 11 + strlen (header));
1714 sprintf (res, "%s: Basic %s\r\n", header, t2);
1720 /* Parse HTTP `WWW-Authenticate:' header. AU points to the beginning
1721 of a field in such a header. If the field is the one specified by
1722 ATTR_NAME ("realm", "opaque", and "nonce" are used by the current
1723 digest authorization code), extract its value in the (char*)
1724 variable pointed by RET. Returns negative on a malformed header,
1725 or number of bytes that have been parsed by this call. */
1727 extract_header_attr (const char *au, const char *attr_name, char **ret)
1729 const char *cp, *ep;
1733 if (strncmp (cp, attr_name, strlen (attr_name)) == 0)
1735 cp += strlen (attr_name);
1738 cp += skip_lws (cp);
1743 cp += skip_lws (cp);
1748 for (ep = cp; *ep && *ep != '\"'; ep++)
1753 *ret = strdupdelim (cp, ep);
1760 /* Response value needs to be in lowercase, so we cannot use HEXD2ASC
1761 from url.h. See RFC 2069 2.1.2 for the syntax of response-digest. */
1762 #define HEXD2asc(x) (((x) < 10) ? ((x) + '0') : ((x) - 10 + 'a'))
1764 /* Dump the hexadecimal representation of HASH to BUF. HASH should be
1765 an array of 16 bytes containing the hash keys, and BUF should be a
1766 buffer of 33 writable characters (32 for hex digits plus one for
1767 zero termination). */
1769 dump_hash (unsigned char *buf, const unsigned char *hash)
1773 for (i = 0; i < MD5_HASHLEN; i++, hash++)
1775 *buf++ = HEXD2asc (*hash >> 4);
1776 *buf++ = HEXD2asc (*hash & 0xf);
1781 /* Take the line apart to find the challenge, and compose a digest
1782 authorization header. See RFC2069 section 2.1.2. */
1784 digest_authentication_encode (const char *au, const char *user,
1785 const char *passwd, const char *method,
1788 static char *realm, *opaque, *nonce;
1793 { "realm", &realm },
1794 { "opaque", &opaque },
1799 realm = opaque = nonce = NULL;
1801 au += 6; /* skip over `Digest' */
1806 au += skip_lws (au);
1807 for (i = 0; i < ARRAY_SIZE (options); i++)
1809 int skip = extract_header_attr (au, options[i].name,
1810 options[i].variable);
1814 FREE_MAYBE (opaque);
1824 if (i == ARRAY_SIZE (options))
1826 while (*au && *au != '=')
1830 au += skip_lws (au);
1834 while (*au && *au != '\"')
1841 while (*au && *au != ',')
1846 if (!realm || !nonce || !user || !passwd || !path || !method)
1849 FREE_MAYBE (opaque);
1854 /* Calculate the digest value. */
1857 unsigned char hash[MD5_HASHLEN];
1858 unsigned char a1buf[MD5_HASHLEN * 2 + 1], a2buf[MD5_HASHLEN * 2 + 1];
1859 unsigned char response_digest[MD5_HASHLEN * 2 + 1];
1861 /* A1BUF = H(user ":" realm ":" password) */
1862 md5_init_ctx (&ctx);
1863 md5_process_bytes (user, strlen (user), &ctx);
1864 md5_process_bytes (":", 1, &ctx);
1865 md5_process_bytes (realm, strlen (realm), &ctx);
1866 md5_process_bytes (":", 1, &ctx);
1867 md5_process_bytes (passwd, strlen (passwd), &ctx);
1868 md5_finish_ctx (&ctx, hash);
1869 dump_hash (a1buf, hash);
1871 /* A2BUF = H(method ":" path) */
1872 md5_init_ctx (&ctx);
1873 md5_process_bytes (method, strlen (method), &ctx);
1874 md5_process_bytes (":", 1, &ctx);
1875 md5_process_bytes (path, strlen (path), &ctx);
1876 md5_finish_ctx (&ctx, hash);
1877 dump_hash (a2buf, hash);
1879 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */
1880 md5_init_ctx (&ctx);
1881 md5_process_bytes (a1buf, MD5_HASHLEN * 2, &ctx);
1882 md5_process_bytes (":", 1, &ctx);
1883 md5_process_bytes (nonce, strlen (nonce), &ctx);
1884 md5_process_bytes (":", 1, &ctx);
1885 md5_process_bytes (a2buf, MD5_HASHLEN * 2, &ctx);
1886 md5_finish_ctx (&ctx, hash);
1887 dump_hash (response_digest, hash);
1889 res = (char*) xmalloc (strlen (user)
1894 + 2 * MD5_HASHLEN /*strlen (response_digest)*/
1895 + (opaque ? strlen (opaque) : 0)
1897 sprintf (res, "Authorization: Digest \
1898 username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
1899 user, realm, nonce, path, response_digest);
1902 char *p = res + strlen (res);
1903 strcat (p, ", opaque=\"");
1907 strcat (res, "\r\n");
1911 #endif /* USE_DIGEST */
1914 #define HACK_O_MATIC(line, string_constant) \
1915 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
1916 && (ISSPACE (line[sizeof (string_constant) - 1]) \
1917 || !line[sizeof (string_constant) - 1]))
1920 known_authentication_scheme_p (const char *au)
1922 return HACK_O_MATIC (au, "Basic")
1923 || HACK_O_MATIC (au, "Digest")
1924 || HACK_O_MATIC (au, "NTLM");
1929 /* Create the HTTP authorization request header. When the
1930 `WWW-Authenticate' response header is seen, according to the
1931 authorization scheme specified in that header (`Basic' and `Digest'
1932 are supported by the current implementation), produce an
1933 appropriate HTTP authorization request header. */
1935 create_authorization_line (const char *au, const char *user,
1936 const char *passwd, const char *method,
1939 char *wwwauth = NULL;
1941 if (!strncasecmp (au, "Basic", 5))
1942 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
1943 if (!strncasecmp (au, "NTLM", 4))
1944 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
1946 else if (!strncasecmp (au, "Digest", 6))
1947 wwwauth = digest_authentication_encode (au, user, passwd, method, path);
1948 #endif /* USE_DIGEST */