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);
260 long len; /* received length */
261 long contlen; /* expected length */
262 long restval; /* the restart value */
263 int res; /* the result of last read */
264 char *newloc; /* new location (redirection) */
265 char *remote_time; /* remote time-stamp string */
266 char *error; /* textual HTTP error */
267 int statcode; /* status code */
268 long dltime; /* time of the download */
271 /* Free the elements of hstat X. */
272 #define FREEHSTAT(x) do \
274 FREE_MAYBE ((x).newloc); \
275 FREE_MAYBE ((x).remote_time); \
276 FREE_MAYBE ((x).error); \
277 (x).newloc = (x).remote_time = (x).error = NULL; \
280 static char *create_authorization_line PARAMS ((const char *, const char *,
281 const char *, const char *,
283 static char *basic_authentication_encode PARAMS ((const char *, const char *,
285 static int known_authentication_scheme_p PARAMS ((const char *));
287 static time_t http_atotm PARAMS ((char *));
289 /* Retrieve a document through HTTP protocol. It recognizes status
290 code, and correctly handles redirections. It closes the network
291 socket. If it receives an error from the functions below it, it
292 will print it if there is enough information to do so (almost
293 always), returning the error to the caller (i.e. http_loop).
295 Various HTTP parameters are stored to hs. Although it parses the
296 response code correctly, it is not used in a sane way. The caller
299 If u->proxy is non-NULL, the URL u will be taken as a proxy URL,
300 and u->proxy->url will be given to the proxy server (bad naming,
303 gethttp (struct urlinfo *u, struct http_stat *hs, int *dt)
305 char *request, *type, *command, *path;
307 char *pragma_h, *referer, *useragent, *range, *wwwauth, *remhost;
308 char *authenticate_h;
313 int sock, hcount, num_written, all_length, remport, statcode;
314 long contlen, contrange;
318 int auth_tried_already;
321 if (!(*dt & HEAD_ONLY))
322 /* If we're doing a GET on the URL, as opposed to just a HEAD, we need to
323 know the local filename so we can save to it. */
324 assert (u->local != NULL);
327 auth_tried_already = 0;
330 /* We need to come back here when the initial attempt to retrieve
331 without authorization header fails. */
333 /* Initialize certain elements of struct http_stat. */
338 hs->remote_time = NULL;
341 /* Which structure to use to retrieve the original URL data. */
347 /* First: establish the connection. */
348 logprintf (LOG_VERBOSE, _("Connecting to %s:%hu... "), u->host, u->port);
349 err = make_connection (&sock, u->host, u->port);
353 logputs (LOG_VERBOSE, "\n");
354 logprintf (LOG_NOTQUIET, "%s: %s.\n", u->host, herrmsg (h_errno));
358 logputs (LOG_VERBOSE, "\n");
359 logprintf (LOG_NOTQUIET, "socket: %s\n", strerror (errno));
363 logputs (LOG_VERBOSE, "\n");
364 logprintf (LOG_NOTQUIET,
365 _("Connection to %s:%hu refused.\n"), u->host, u->port);
369 logputs (LOG_VERBOSE, "\n");
370 logprintf (LOG_NOTQUIET, "connect: %s\n", strerror (errno));
375 /* Everything is fine! */
376 logputs (LOG_VERBOSE, _("connected!\n"));
384 path = u->proxy->url;
388 command = (*dt & HEAD_ONLY) ? "HEAD" : "GET";
392 referer = (char *)alloca (9 + strlen (ou->referer) + 3);
393 sprintf (referer, "Referer: %s\r\n", ou->referer);
395 if (*dt & SEND_NOCACHE)
396 pragma_h = "Pragma: no-cache\r\n";
401 range = (char *)alloca (13 + numdigit (hs->restval) + 4);
402 /* #### Gag me! Some servers (e.g. WebSitePro) have been known
403 to misinterpret the following `Range' format, and return the
404 document as multipart/x-byte-ranges MIME type!
406 #### TODO: Interpret MIME types, recognize bullshits similar
407 the one described above, and deal with them! */
408 sprintf (range, "Range: bytes=%ld-\r\n", hs->restval);
413 STRDUP_ALLOCA (useragent, opt.useragent);
416 useragent = (char *)alloca (10 + strlen (version_string));
417 sprintf (useragent, "Wget/%s", version_string);
419 /* Construct the authentication, if userid is present. */
422 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
423 user = user ? user : opt.http_user;
424 passwd = passwd ? passwd : opt.http_passwd;
427 if (authenticate_h && user && passwd)
429 wwwauth = create_authorization_line (authenticate_h, user, passwd,
436 char *proxy_user, *proxy_passwd;
437 /* For normal username and password, URL components override
438 command-line/wgetrc parameters. With proxy authentication,
439 it's the reverse, because proxy URLs are normally the
440 "permanent" ones, so command-line args should take
442 if (opt.proxy_user && opt.proxy_passwd)
444 proxy_user = opt.proxy_user;
445 proxy_passwd = opt.proxy_passwd;
449 proxy_user = u->user;
450 proxy_passwd = u->passwd;
452 /* #### This is junky. Can't the proxy request, say, `Digest'
454 if (proxy_user && proxy_passwd)
455 proxyauth = basic_authentication_encode (proxy_user, proxy_passwd,
456 "Proxy-Authorization");
468 host_port = (char *)alloca (numdigit (remport) + 2);
469 host_port_len = sprintf (host_port, ":%d", remport);
472 /* Allocate the memory for the request. */
473 request = (char *)alloca (strlen (command) + strlen (path)
475 + strlen (remhost) + host_port_len
476 + strlen (HTTP_ACCEPT)
477 + (referer ? strlen (referer) : 0)
478 + (wwwauth ? strlen (wwwauth) : 0)
479 + (proxyauth ? strlen (proxyauth) : 0)
480 + (range ? strlen (range) : 0)
482 + (opt.user_header ? strlen (opt.user_header) : 0)
484 /* Construct the request. */
491 command, path, useragent, remhost,
492 host_port ? host_port : "",
493 HTTP_ACCEPT, referer ? referer : "",
494 wwwauth ? wwwauth : "",
495 proxyauth ? proxyauth : "",
498 opt.user_header ? opt.user_header : "");
499 DEBUGP (("---request begin---\n%s---request end---\n", request));
500 /* Free the temporary memory. */
501 FREE_MAYBE (wwwauth);
502 FREE_MAYBE (proxyauth);
504 /* Send the request to server. */
505 num_written = iwrite (sock, request, strlen (request));
508 logputs (LOG_VERBOSE, _("Failed writing HTTP request.\n"));
512 logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
513 u->proxy ? "Proxy" : "HTTP");
514 contlen = contrange = -1;
519 /* Before reading anything, initialize the rbuf. */
520 rbuf_initialize (&rbuf, sock);
524 /* Header-fetching loop. */
532 /* Get the header. */
533 status = header_get (&rbuf, &hdr,
534 /* Disallow continuations for status line. */
535 (hcount == 1 ? HG_NO_CONTINUATIONS : HG_NONE));
537 /* Check for errors. */
538 if (status == HG_EOF && *hdr)
540 /* This used to be an unconditional error, but that was
541 somewhat controversial, because of a large number of
542 broken CGI's that happily "forget" to send the second EOL
543 before closing the connection of a HEAD request.
545 So, the deal is to check whether the header is empty
546 (*hdr is zero if it is); if yes, it means that the
547 previous header was fully retrieved, and that -- most
548 probably -- the request is complete. "...be liberal in
549 what you accept." Oh boy. */
550 logputs (LOG_VERBOSE, "\n");
551 logputs (LOG_NOTQUIET, _("End of file while parsing headers.\n"));
554 FREE_MAYBE (hs->newloc);
555 FREE_MAYBE (all_headers);
559 else if (status == HG_ERROR)
561 logputs (LOG_VERBOSE, "\n");
562 logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
566 FREE_MAYBE (hs->newloc);
567 FREE_MAYBE (all_headers);
572 /* If the headers are to be saved to a file later, save them to
574 if (opt.save_headers)
576 int lh = strlen (hdr);
577 all_headers = (char *)xrealloc (all_headers, all_length + lh + 2);
578 memcpy (all_headers + all_length, hdr, lh);
580 all_headers[all_length++] = '\n';
581 all_headers[all_length] = '\0';
584 /* Print the header if requested. */
585 if (opt.server_response && hcount != 1)
586 logprintf (LOG_VERBOSE, "\n%d %s", hcount, hdr);
588 /* Check for status line. */
592 /* Parse the first line of server response. */
593 statcode = parse_http_status_line (hdr, &error);
594 hs->statcode = statcode;
595 /* Store the descriptive response. */
596 if (statcode == -1) /* malformed response */
598 /* A common reason for "malformed response" error is the
599 case when no data was actually received. Handle this
602 hs->error = xstrdup (_("No data received"));
604 hs->error = xstrdup (_("Malformed status line"));
609 hs->error = xstrdup (_("(no description)"));
611 hs->error = xstrdup (error);
618 logprintf (LOG_VERBOSE, "%d %s", statcode, error);
623 /* Exit on empty header. */
630 /* Try getting content-length. */
631 if (contlen == -1 && !opt.ignore_length)
632 if (header_process (hdr, "Content-Length", header_extract_number,
635 /* Try getting content-type. */
637 if (header_process (hdr, "Content-Type", http_process_type, &type))
639 /* Try getting location. */
641 if (header_process (hdr, "Location", header_strdup, &hs->newloc))
643 /* Try getting last-modified. */
644 if (!hs->remote_time)
645 if (header_process (hdr, "Last-Modified", header_strdup,
648 /* Try getting www-authentication. */
650 if (header_process (hdr, "WWW-Authenticate", header_strdup,
653 /* Check for accept-ranges header. If it contains the word
654 `none', disable the ranges. */
655 if (*dt & ACCEPTRANGES)
658 if (header_process (hdr, "Accept-Ranges", http_process_none, &nonep))
661 *dt &= ~ACCEPTRANGES;
665 /* Try getting content-range. */
668 struct http_process_range_closure closure;
669 if (header_process (hdr, "Content-Range", http_process_range, &closure))
671 contrange = closure.first_byte_pos;
679 logputs (LOG_VERBOSE, "\n");
681 if ((statcode == HTTP_STATUS_UNAUTHORIZED)
684 /* Authorization is required. */
689 if (auth_tried_already)
691 /* If we have tried it already, then there is not point
693 logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
694 free (authenticate_h);
697 else if (!known_authentication_scheme_p (authenticate_h))
699 free (authenticate_h);
700 logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
705 auth_tried_already = 1;
709 /* We do not need this anymore. */
712 free (authenticate_h);
713 authenticate_h = NULL;
716 /* 20x responses are counted among successful by default. */
717 if (H_20X (statcode))
720 if (type && !strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S)))
723 /* We don't assume text/html by default. */
726 if (opt.html_extension && (*dt & TEXTHTML))
727 /* -E / --html-extension / html_extension = on was specified, and this is a
728 text/html file. If some case-insensitive variation on ".htm[l]" isn't
729 already the file's suffix, tack on ".html". */
731 char* last_period_in_local_filename = strrchr(u->local, '.');
733 if (last_period_in_local_filename == NULL ||
734 !(strcasecmp(last_period_in_local_filename, ".htm") == EQ ||
735 strcasecmp(last_period_in_local_filename, ".html") == EQ))
737 size_t local_filename_len = strlen(u->local);
739 u->local = xrealloc(u->local, local_filename_len + sizeof(".html"));
740 strcpy(u->local + local_filename_len, ".html");
742 *dt |= ADDED_HTML_EXTENSION;
748 else if (contrange != hs->restval ||
749 (H_PARTIAL (statcode) && contrange == -1))
751 /* This means the whole request was somehow misunderstood by the
754 FREE_MAYBE (hs->newloc);
755 FREE_MAYBE (all_headers);
763 contlen += contrange;
765 contrange = -1; /* If conent-length was not sent,
766 content-range will be ignored. */
768 hs->contlen = contlen;
770 /* Return if redirected. */
771 if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES)
773 /* RFC2068 says that in case of the 300 (multiple choices)
774 response, the server can output a preferred URL through
775 `Location' header; otherwise, the request should be treated
776 like GET. So, if the location is set, it will be a
777 redirection; otherwise, just proceed normally. */
778 if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc)
782 logprintf (LOG_VERBOSE,
783 _("Location: %s%s\n"),
784 hs->newloc ? hs->newloc : _("unspecified"),
785 hs->newloc ? _(" [following]") : "");
788 FREE_MAYBE (all_headers);
794 if ((*dt & RETROKF) && !opt.server_response)
796 /* No need to print this output if the body won't be
797 downloaded at all, or if the original server response is
799 logputs (LOG_VERBOSE, _("Length: "));
802 logputs (LOG_VERBOSE, legible (contlen));
804 logprintf (LOG_VERBOSE, _(" (%s to go)"),
805 legible (contlen - contrange));
808 logputs (LOG_VERBOSE,
809 opt.ignore_length ? _("ignored") : _("unspecified"));
811 logprintf (LOG_VERBOSE, " [%s]\n", type);
813 logputs (LOG_VERBOSE, "\n");
817 type = NULL; /* We don't need it any more. */
819 /* Return if we have no intention of further downloading. */
820 if (!(*dt & RETROKF) || (*dt & HEAD_ONLY))
822 /* In case someone cares to look... */
826 FREE_MAYBE (all_headers);
831 /* Open the local file. */
834 mkalldirs (u->local);
836 rotate_backups (u->local);
837 fp = fopen (u->local, hs->restval ? "ab" : "wb");
840 logprintf (LOG_NOTQUIET, "%s: %s\n", u->local, strerror (errno));
842 FREE_MAYBE (all_headers);
851 /* This will silently fail for streams that don't correspond
852 to regular files, but that's OK. */
858 /* #### This confuses the code that checks for file size. There
859 should be some overhead information. */
860 if (opt.save_headers)
861 fwrite (all_headers, 1, all_length, fp);
863 /* Get the contents of the document. */
864 hs->res = get_contents (sock, fp, &hs->len, hs->restval,
865 (contlen != -1 ? contlen : 0),
867 hs->dltime = elapsed_time ();
869 /* Close or flush the file. We have to be careful to check for
870 error here. Checking the result of fwrite() is not enough --
871 errors could go unnoticed! */
874 flush_res = fclose (fp);
876 flush_res = fflush (fp);
877 if (flush_res == EOF)
880 FREE_MAYBE (all_headers);
887 /* The genuine HTTP loop! This is the part where the retrieval is
888 retried, and retried, and retried, and... */
890 http_loop (struct urlinfo *u, char **newloc, int *dt)
892 static int first_retrieval = 1;
895 int use_ts, got_head = 0; /* time-stamping info */
896 char *filename_plus_orig_suffix;
897 char *local_filename = NULL;
898 char *tms, *suf, *locf, *tmrate;
900 time_t tml = -1, tmr = -1; /* local and remote time-stamps */
901 long local_size = 0; /* the size of the local file */
903 struct http_stat hstat; /* HTTP status */
908 /* Warn on (likely bogus) wildcard usage in HTTP. Don't use
909 has_wildcards_p because it would also warn on `?', and we know that
910 shows up in CGI paths a *lot*. */
911 if (strchr (u->url, '*'))
912 logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
914 /* Determine the local filename. */
916 u->local = url_filename (u->proxy ? u->proxy : u);
918 if (!opt.output_document)
921 locf = opt.output_document;
923 /* Yuck. Multiple returns suck. We need to remember to free() the space we
924 xmalloc() here before EACH return. This is one reason it's better to set
925 flags that influence flow control and then return once at the end. */
926 filename_len = strlen(u->local);
927 filename_plus_orig_suffix = xmalloc(filename_len + sizeof(".orig"));
929 if (opt.noclobber && file_exists_p (u->local))
931 /* If opt.noclobber is turned on and file already exists, do not
933 logprintf (LOG_VERBOSE, _("\
934 File `%s' already there, will not retrieve.\n"), u->local);
935 /* If the file is there, we suppose it's retrieved OK. */
938 /* #### Bogusness alert. */
939 /* If its suffix is "html" or (yuck!) "htm", we suppose it's
940 text/html, a harmless lie. */
941 if (((suf = suffix (u->local)) != NULL)
942 && (!strcmp (suf, "html") || !strcmp (suf, "htm")))
945 free(filename_plus_orig_suffix); /* must precede every return! */
946 /* Another harmless lie: */
951 if (opt.timestamping)
953 boolean local_dot_orig_file_exists = FALSE;
955 if (opt.backup_converted)
956 /* If -K is specified, we'll act on the assumption that it was specified
957 last time these files were downloaded as well, and instead of just
958 comparing local file X against server file X, we'll compare local
959 file X.orig (if extant, else X) against server file X. If -K
960 _wasn't_ specified last time, or the server contains files called
961 *.orig, -N will be back to not operating correctly with -k. */
963 /* Would a single s[n]printf() call be faster? --dan
965 It wouldn't. sprintf() is horribly slow. At one point I
966 profiled Wget, and found that a measurable and
967 non-negligible amount of time was lost calling sprintf()
968 in url.c. Replacing sprintf with inline calls to
969 strcpy() and long_to_string() made a difference.
971 strcpy(filename_plus_orig_suffix, u->local);
972 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
974 /* Try to stat() the .orig file. */
975 if (stat(filename_plus_orig_suffix, &st) == 0)
977 local_dot_orig_file_exists = TRUE;
978 local_filename = filename_plus_orig_suffix;
982 if (!local_dot_orig_file_exists)
983 /* Couldn't stat() <file>.orig, so try to stat() <file>. */
984 if (stat (u->local, &st) == 0)
985 local_filename = u->local;
987 if (local_filename != NULL)
988 /* There was a local file, so we'll check later to see if the version
989 the server has is the same version we already have, allowing us to
994 local_size = st.st_size;
998 /* Reset the counter. */
1000 *dt = 0 | ACCEPTRANGES;
1004 /* Increment the pass counter. */
1006 /* Wait before the retrieval (unless this is the very first
1008 Check if we are retrying or not, wait accordingly - HEH */
1009 if (!first_retrieval && (opt.wait || (count && opt.waitretry)))
1013 if (count<opt.waitretry)
1016 sleep(opt.waitretry);
1021 if (first_retrieval)
1022 first_retrieval = 0;
1023 /* Get the current time string. */
1024 tms = time_str (NULL);
1025 /* Print fetch message, if opt.verbose. */
1028 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
1032 sprintf (tmp, _("(try:%2d)"), count);
1033 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
1034 tms, hurl, tmp, locf);
1036 ws_changetitle (hurl, 1);
1041 /* Default document type is empty. However, if spider mode is
1042 on or time-stamping is employed, HEAD_ONLY commands is
1043 encoded within *dt. */
1044 if (opt.spider || (use_ts && !got_head))
1048 /* Assume no restarting. */
1050 /* Decide whether or not to restart. */
1051 if (((count > 1 && (*dt & ACCEPTRANGES)) || opt.always_rest)
1052 && file_exists_p (u->local))
1053 if (stat (u->local, &st) == 0)
1054 hstat.restval = st.st_size;
1055 /* Decide whether to send the no-cache directive. */
1056 if (u->proxy && (count > 1 || (opt.proxy_cache == 0)))
1057 *dt |= SEND_NOCACHE;
1059 *dt &= ~SEND_NOCACHE;
1061 /* Try fetching the document, or at least its head. :-) */
1062 err = gethttp (u, &hstat, dt);
1064 /* It's unfortunate that wget determines the local filename before finding
1065 out the Content-Type of the file. Barring a major restructuring of the
1066 code, we need to re-set locf here, since gethttp() may have xrealloc()d
1067 u->local to tack on ".html". */
1068 if (!opt.output_document)
1071 locf = opt.output_document;
1074 tms = time_str (NULL);
1075 /* Get the new location (with or without the redirection). */
1077 *newloc = xstrdup (hstat.newloc);
1080 case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
1081 case CONERROR: case READERR: case WRITEFAILED:
1083 /* Non-fatal errors continue executing the loop, which will
1084 bring them to "while" statement at the end, to judge
1085 whether the number of tries was exceeded. */
1087 printwhat (count, opt.ntry);
1090 case HOSTERR: case CONREFUSED: case PROXERR: case AUTHFAILED:
1091 /* Fatal errors just return from the function. */
1093 free(filename_plus_orig_suffix); /* must precede every return! */
1096 case FWRITEERR: case FOPENERR:
1097 /* Another fatal error. */
1098 logputs (LOG_VERBOSE, "\n");
1099 logprintf (LOG_NOTQUIET, _("Cannot write to `%s' (%s).\n"),
1100 u->local, strerror (errno));
1102 free(filename_plus_orig_suffix); /* must precede every return! */
1106 /* Return the new location to the caller. */
1109 logprintf (LOG_NOTQUIET,
1110 _("ERROR: Redirection (%d) without location.\n"),
1112 free(filename_plus_orig_suffix); /* must precede every return! */
1116 free(filename_plus_orig_suffix); /* must precede every return! */
1120 /* Deal with you later. */
1123 /* All possibilities should have been exhausted. */
1126 if (!(*dt & RETROKF))
1130 /* #### Ugly ugly ugly! */
1131 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
1132 logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
1135 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
1136 tms, hstat.statcode, hstat.error);
1137 logputs (LOG_VERBOSE, "\n");
1139 free(filename_plus_orig_suffix); /* must precede every return! */
1143 /* Did we get the time-stamp? */
1146 if (opt.timestamping && !hstat.remote_time)
1148 logputs (LOG_NOTQUIET, _("\
1149 Last-modified header missing -- time-stamps turned off.\n"));
1151 else if (hstat.remote_time)
1153 /* Convert the date-string into struct tm. */
1154 tmr = http_atotm (hstat.remote_time);
1155 if (tmr == (time_t) (-1))
1156 logputs (LOG_VERBOSE, _("\
1157 Last-modified header invalid -- time-stamp ignored.\n"));
1161 /* The time-stamping section. */
1166 use_ts = 0; /* no more time-stamping */
1167 count = 0; /* the retrieve count for HEAD is
1169 if (hstat.remote_time && tmr != (time_t) (-1))
1171 /* Now time-stamping can be used validly. Time-stamping
1172 means that if the sizes of the local and remote file
1173 match, and local file is newer than the remote file,
1174 it will not be retrieved. Otherwise, the normal
1175 download procedure is resumed. */
1177 (hstat.contlen == -1 || local_size == hstat.contlen))
1179 logprintf (LOG_VERBOSE, _("\
1180 Server file no newer than local file `%s' -- not retrieving.\n\n"),
1183 free(filename_plus_orig_suffix);/*must precede every return!*/
1186 else if (tml >= tmr)
1187 logprintf (LOG_VERBOSE, _("\
1188 The sizes do not match (local %ld) -- retrieving.\n"), local_size);
1190 logputs (LOG_VERBOSE,
1191 _("Remote file is newer, retrieving.\n"));
1197 && (tmr != (time_t) (-1))
1199 && ((hstat.len == hstat.contlen) ||
1200 ((hstat.res == 0) &&
1201 ((hstat.contlen == -1) ||
1202 (hstat.len >= hstat.contlen && !opt.kill_longer)))))
1204 touch (u->local, tmr);
1206 /* End of time-stamping section. */
1210 logprintf (LOG_NOTQUIET, "%d %s\n\n", hstat.statcode, hstat.error);
1211 free(filename_plus_orig_suffix); /* must precede every return! */
1215 /* It is now safe to free the remainder of hstat, since the
1216 strings within it will no longer be used. */
1219 tmrate = rate (hstat.len - hstat.restval, hstat.dltime);
1221 if (hstat.len == hstat.contlen)
1225 logprintf (LOG_VERBOSE,
1226 _("%s (%s) - `%s' saved [%ld/%ld]\n\n"),
1227 tms, tmrate, locf, hstat.len, hstat.contlen);
1228 logprintf (LOG_NONVERBOSE,
1229 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1230 tms, u->url, hstat.len, hstat.contlen, locf, count);
1233 downloaded_increase (hstat.len);
1235 /* Remember that we downloaded the file for later ".orig" code. */
1236 if (*dt & ADDED_HTML_EXTENSION)
1237 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1239 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1241 free(filename_plus_orig_suffix); /* must precede every return! */
1244 else if (hstat.res == 0) /* No read error */
1246 if (hstat.contlen == -1) /* We don't know how much we were supposed
1247 to get, so assume we succeeded. */
1251 logprintf (LOG_VERBOSE,
1252 _("%s (%s) - `%s' saved [%ld]\n\n"),
1253 tms, tmrate, locf, hstat.len);
1254 logprintf (LOG_NONVERBOSE,
1255 "%s URL:%s [%ld] -> \"%s\" [%d]\n",
1256 tms, u->url, hstat.len, locf, count);
1259 downloaded_increase (hstat.len);
1261 /* Remember that we downloaded the file for later ".orig" code. */
1262 if (*dt & ADDED_HTML_EXTENSION)
1263 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1265 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1267 free(filename_plus_orig_suffix); /* must precede every return! */
1270 else if (hstat.len < hstat.contlen) /* meaning we lost the
1271 connection too soon */
1273 logprintf (LOG_VERBOSE,
1274 _("%s (%s) - Connection closed at byte %ld. "),
1275 tms, tmrate, hstat.len);
1276 printwhat (count, opt.ntry);
1279 else if (!opt.kill_longer) /* meaning we got more than expected */
1281 logprintf (LOG_VERBOSE,
1282 _("%s (%s) - `%s' saved [%ld/%ld])\n\n"),
1283 tms, tmrate, locf, hstat.len, hstat.contlen);
1284 logprintf (LOG_NONVERBOSE,
1285 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1286 tms, u->url, hstat.len, hstat.contlen, locf, count);
1288 downloaded_increase (hstat.len);
1290 /* Remember that we downloaded the file for later ".orig" code. */
1291 if (*dt & ADDED_HTML_EXTENSION)
1292 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1294 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1296 free(filename_plus_orig_suffix); /* must precede every return! */
1299 else /* the same, but not accepted */
1301 logprintf (LOG_VERBOSE,
1302 _("%s (%s) - Connection closed at byte %ld/%ld. "),
1303 tms, tmrate, hstat.len, hstat.contlen);
1304 printwhat (count, opt.ntry);
1308 else /* now hstat.res can only be -1 */
1310 if (hstat.contlen == -1)
1312 logprintf (LOG_VERBOSE,
1313 _("%s (%s) - Read error at byte %ld (%s)."),
1314 tms, tmrate, hstat.len, strerror (errno));
1315 printwhat (count, opt.ntry);
1318 else /* hstat.res == -1 and contlen is given */
1320 logprintf (LOG_VERBOSE,
1321 _("%s (%s) - Read error at byte %ld/%ld (%s). "),
1322 tms, tmrate, hstat.len, hstat.contlen,
1324 printwhat (count, opt.ntry);
1331 while (!opt.ntry || (count < opt.ntry));
1332 free(filename_plus_orig_suffix); /* must precede every return! */
1336 /* Converts struct tm to time_t, assuming the data in tm is UTC rather
1337 than local timezone (mktime assumes the latter).
1339 Contributed by Roger Beeman <beeman@cisco.com>, with the help of
1340 Mark Baushke <mdb@cisco.com> and the rest of the Gurus at CISCO. */
1342 mktime_from_utc (struct tm *t)
1349 tb = mktime (gmtime (&tl));
1350 return (tl <= tb ? (tl + (tl - tb)) : (tl - (tb - tl)));
1353 /* Check whether the result of strptime() indicates success.
1354 strptime() returns the pointer to how far it got to in the string.
1355 The processing has been successful if the string is at `GMT' or
1356 `+X', or at the end of the string.
1358 In extended regexp parlance, the function returns 1 if P matches
1359 "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (a valid result of
1360 strptime()) is considered a failure and 0 is returned. */
1366 while (ISSPACE (*p))
1369 || (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
1370 || ((p[0] == '+' || p[1] == '-') && ISDIGIT (p[1])))
1376 /* Convert TIME_STRING time to time_t. TIME_STRING can be in any of
1377 the three formats RFC2068 allows the HTTP servers to emit --
1378 RFC1123-date, RFC850-date or asctime-date. Timezones are ignored,
1381 We use strptime() to recognize various dates, which makes it a
1382 little bit slacker than the RFC1123/RFC850/asctime (e.g. it always
1383 allows shortened dates and months, one-digit days, etc.). It also
1384 allows more than one space anywhere where the specs require one SP.
1385 The routine should probably be even more forgiving (as recommended
1386 by RFC2068), but I do not have the time to write one.
1388 Return the computed time_t representation, or -1 if all the
1391 Needless to say, what we *really* need here is something like
1392 Marcus Hennecke's atotm(), which is forgiving, fast, to-the-point,
1393 and does not use strptime(). atotm() is to be found in the sources
1394 of `phttpd', a little-known HTTP server written by Peter Erikson. */
1396 http_atotm (char *time_string)
1400 /* Roger Beeman says: "This function dynamically allocates struct tm
1401 t, but does no initialization. The only field that actually
1402 needs initialization is tm_isdst, since the others will be set by
1403 strptime. Since strptime does not set tm_isdst, it will return
1404 the data structure with whatever data was in tm_isdst to begin
1405 with. For those of us in timezones where DST can occur, there
1406 can be a one hour shift depending on the previous contents of the
1407 data area where the data structure is allocated." */
1410 /* Note that under foreign locales Solaris strptime() fails to
1411 recognize English dates, which renders this function useless. I
1412 assume that other non-GNU strptime's are plagued by the same
1413 disease. We solve this by setting only LC_MESSAGES in
1414 i18n_initialize(), instead of LC_ALL.
1416 Another solution could be to temporarily set locale to C, invoke
1417 strptime(), and restore it back. This is slow and dirty,
1418 however, and locale support other than LC_MESSAGES can mess other
1419 things, so I rather chose to stick with just setting LC_MESSAGES.
1421 Also note that none of this is necessary under GNU strptime(),
1422 because it recognizes both international and local dates. */
1424 /* NOTE: We don't use `%n' for white space, as OSF's strptime uses
1425 it to eat all white space up to (and including) a newline, and
1426 the function fails if there is no newline (!).
1428 Let's hope all strptime() implementations use ` ' to skip *all*
1429 whitespace instead of just one (it works that way on all the
1430 systems I've tested it on). */
1432 /* RFC1123: Thu, 29 Jan 1998 22:12:57 */
1433 if (check_end (strptime (time_string, "%a, %d %b %Y %T", &t)))
1434 return mktime_from_utc (&t);
1435 /* RFC850: Thu, 29-Jan-98 22:12:57 */
1436 if (check_end (strptime (time_string, "%a, %d-%b-%y %T", &t)))
1437 return mktime_from_utc (&t);
1438 /* asctime: Thu Jan 29 22:12:57 1998 */
1439 if (check_end (strptime (time_string, "%a %b %d %T %Y", &t)))
1440 return mktime_from_utc (&t);
1445 /* Authorization support: We support two authorization schemes:
1447 * `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
1449 * `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
1450 consisting of answering to the server's challenge with the proper
1453 /* How many bytes it will take to store LEN bytes in base64. */
1454 #define BASE64_LENGTH(len) (4 * (((len) + 2) / 3))
1456 /* Encode the string S of length LENGTH to base64 format and place it
1457 to STORE. STORE will be 0-terminated, and must point to a writable
1458 buffer of at least 1+BASE64_LENGTH(length) bytes. */
1460 base64_encode (const char *s, char *store, int length)
1462 /* Conversion table. */
1463 static char tbl[64] = {
1464 'A','B','C','D','E','F','G','H',
1465 'I','J','K','L','M','N','O','P',
1466 'Q','R','S','T','U','V','W','X',
1467 'Y','Z','a','b','c','d','e','f',
1468 'g','h','i','j','k','l','m','n',
1469 'o','p','q','r','s','t','u','v',
1470 'w','x','y','z','0','1','2','3',
1471 '4','5','6','7','8','9','+','/'
1474 unsigned char *p = (unsigned char *)store;
1476 /* Transform the 3x8 bits to 4x6 bits, as required by base64. */
1477 for (i = 0; i < length; i += 3)
1479 *p++ = tbl[s[0] >> 2];
1480 *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
1481 *p++ = tbl[((s[1] & 0xf) << 2) + (s[2] >> 6)];
1482 *p++ = tbl[s[2] & 0x3f];
1485 /* Pad the result if necessary... */
1486 if (i == length + 1)
1488 else if (i == length + 2)
1489 *(p - 1) = *(p - 2) = '=';
1490 /* ...and zero-terminate it. */
1494 /* Create the authentication header contents for the `Basic' scheme.
1495 This is done by encoding the string `USER:PASS' in base64 and
1496 prepending `HEADER: Basic ' to it. */
1498 basic_authentication_encode (const char *user, const char *passwd,
1501 char *t1, *t2, *res;
1502 int len1 = strlen (user) + 1 + strlen (passwd);
1503 int len2 = BASE64_LENGTH (len1);
1505 t1 = (char *)alloca (len1 + 1);
1506 sprintf (t1, "%s:%s", user, passwd);
1507 t2 = (char *)alloca (1 + len2);
1508 base64_encode (t1, t2, len1);
1509 res = (char *)malloc (len2 + 11 + strlen (header));
1510 sprintf (res, "%s: Basic %s\r\n", header, t2);
1516 /* Parse HTTP `WWW-Authenticate:' header. AU points to the beginning
1517 of a field in such a header. If the field is the one specified by
1518 ATTR_NAME ("realm", "opaque", and "nonce" are used by the current
1519 digest authorization code), extract its value in the (char*)
1520 variable pointed by RET. Returns negative on a malformed header,
1521 or number of bytes that have been parsed by this call. */
1523 extract_header_attr (const char *au, const char *attr_name, char **ret)
1525 const char *cp, *ep;
1529 if (strncmp (cp, attr_name, strlen (attr_name)) == 0)
1531 cp += strlen (attr_name);
1534 cp += skip_lws (cp);
1539 cp += skip_lws (cp);
1544 for (ep = cp; *ep && *ep != '\"'; ep++)
1549 *ret = strdupdelim (cp, ep);
1556 /* Response value needs to be in lowercase, so we cannot use HEXD2ASC
1557 from url.h. See RFC 2069 2.1.2 for the syntax of response-digest. */
1558 #define HEXD2asc(x) (((x) < 10) ? ((x) + '0') : ((x) - 10 + 'a'))
1560 /* Dump the hexadecimal representation of HASH to BUF. HASH should be
1561 an array of 16 bytes containing the hash keys, and BUF should be a
1562 buffer of 33 writable characters (32 for hex digits plus one for
1563 zero termination). */
1565 dump_hash (unsigned char *buf, const unsigned char *hash)
1569 for (i = 0; i < MD5_HASHLEN; i++, hash++)
1571 *buf++ = HEXD2asc (*hash >> 4);
1572 *buf++ = HEXD2asc (*hash & 0xf);
1577 /* Take the line apart to find the challenge, and compose a digest
1578 authorization header. See RFC2069 section 2.1.2. */
1580 digest_authentication_encode (const char *au, const char *user,
1581 const char *passwd, const char *method,
1584 static char *realm, *opaque, *nonce;
1589 { "realm", &realm },
1590 { "opaque", &opaque },
1595 realm = opaque = nonce = NULL;
1597 au += 6; /* skip over `Digest' */
1602 au += skip_lws (au);
1603 for (i = 0; i < ARRAY_SIZE (options); i++)
1605 int skip = extract_header_attr (au, options[i].name,
1606 options[i].variable);
1610 FREE_MAYBE (opaque);
1620 if (i == ARRAY_SIZE (options))
1622 while (*au && *au != '=')
1626 au += skip_lws (au);
1630 while (*au && *au != '\"')
1637 while (*au && *au != ',')
1642 if (!realm || !nonce || !user || !passwd || !path || !method)
1645 FREE_MAYBE (opaque);
1650 /* Calculate the digest value. */
1653 unsigned char hash[MD5_HASHLEN];
1654 unsigned char a1buf[MD5_HASHLEN * 2 + 1], a2buf[MD5_HASHLEN * 2 + 1];
1655 unsigned char response_digest[MD5_HASHLEN * 2 + 1];
1657 /* A1BUF = H(user ":" realm ":" password) */
1658 md5_init_ctx (&ctx);
1659 md5_process_bytes (user, strlen (user), &ctx);
1660 md5_process_bytes (":", 1, &ctx);
1661 md5_process_bytes (realm, strlen (realm), &ctx);
1662 md5_process_bytes (":", 1, &ctx);
1663 md5_process_bytes (passwd, strlen (passwd), &ctx);
1664 md5_finish_ctx (&ctx, hash);
1665 dump_hash (a1buf, hash);
1667 /* A2BUF = H(method ":" path) */
1668 md5_init_ctx (&ctx);
1669 md5_process_bytes (method, strlen (method), &ctx);
1670 md5_process_bytes (":", 1, &ctx);
1671 md5_process_bytes (path, strlen (path), &ctx);
1672 md5_finish_ctx (&ctx, hash);
1673 dump_hash (a2buf, hash);
1675 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */
1676 md5_init_ctx (&ctx);
1677 md5_process_bytes (a1buf, MD5_HASHLEN * 2, &ctx);
1678 md5_process_bytes (":", 1, &ctx);
1679 md5_process_bytes (nonce, strlen (nonce), &ctx);
1680 md5_process_bytes (":", 1, &ctx);
1681 md5_process_bytes (a2buf, MD5_HASHLEN * 2, &ctx);
1682 md5_finish_ctx (&ctx, hash);
1683 dump_hash (response_digest, hash);
1685 res = (char*) xmalloc (strlen (user)
1690 + 2 * MD5_HASHLEN /*strlen (response_digest)*/
1691 + (opaque ? strlen (opaque) : 0)
1693 sprintf (res, "Authorization: Digest \
1694 username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
1695 user, realm, nonce, path, response_digest);
1698 char *p = res + strlen (res);
1699 strcat (p, ", opaque=\"");
1703 strcat (res, "\r\n");
1707 #endif /* USE_DIGEST */
1710 #define HACK_O_MATIC(line, string_constant) \
1711 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
1712 && (ISSPACE (line[sizeof (string_constant) - 1]) \
1713 || !line[sizeof (string_constant) - 1]))
1716 known_authentication_scheme_p (const char *au)
1718 return HACK_O_MATIC (au, "Basic")
1719 || HACK_O_MATIC (au, "Digest")
1720 || HACK_O_MATIC (au, "NTLM");
1725 /* Create the HTTP authorization request header. When the
1726 `WWW-Authenticate' response header is seen, according to the
1727 authorization scheme specified in that header (`Basic' and `Digest'
1728 are supported by the current implementation), produce an
1729 appropriate HTTP authorization request header. */
1731 create_authorization_line (const char *au, const char *user,
1732 const char *passwd, const char *method,
1735 char *wwwauth = NULL;
1737 if (!strncasecmp (au, "Basic", 5))
1738 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
1739 if (!strncasecmp (au, "NTLM", 4))
1740 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
1742 else if (!strncasecmp (au, "Digest", 6))
1743 wwwauth = digest_authentication_encode (au, user, passwd, method, path);
1744 #endif /* USE_DIGEST */