2 Copyright (C) 1995, 1996, 1997, 1998 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 *result = xstrdup (hdr);
245 p = strrchr (hdr, ';');
249 *result = (char *)xmalloc (len + 1);
250 memcpy (*result, hdr, len);
251 (*result)[len] = '\0';
254 *result = xstrdup (hdr);
261 long len; /* received length */
262 long contlen; /* expected length */
263 long restval; /* the restart value */
264 int res; /* the result of last read */
265 char *newloc; /* new location (redirection) */
266 char *remote_time; /* remote time-stamp string */
267 char *error; /* textual HTTP error */
268 int statcode; /* status code */
269 long dltime; /* time of the download */
272 /* Free the elements of hstat X. */
273 #define FREEHSTAT(x) do \
275 FREE_MAYBE ((x).newloc); \
276 FREE_MAYBE ((x).remote_time); \
277 FREE_MAYBE ((x).error); \
278 (x).newloc = (x).remote_time = (x).error = NULL; \
281 static char *create_authorization_line PARAMS ((const char *, const char *,
282 const char *, const char *,
284 static char *basic_authentication_encode PARAMS ((const char *, const char *,
286 static int known_authentication_scheme_p PARAMS ((const char *));
288 static time_t http_atotm PARAMS ((char *));
290 /* Retrieve a document through HTTP protocol. It recognizes status
291 code, and correctly handles redirections. It closes the network
292 socket. If it receives an error from the functions below it, it
293 will print it if there is enough information to do so (almost
294 always), returning the error to the caller (i.e. http_loop).
296 Various HTTP parameters are stored to hs. Although it parses the
297 response code correctly, it is not used in a sane way. The caller
300 If u->proxy is non-NULL, the URL u will be taken as a proxy URL,
301 and u->proxy->url will be given to the proxy server (bad naming,
304 gethttp (struct urlinfo *u, struct http_stat *hs, int *dt)
306 char *request, *type, *command, *path, *qstring;
308 char *pragma_h, *referer, *useragent, *range, *wwwauth, *remhost;
309 char *authenticate_h;
314 int sock, hcount, num_written, all_length, remport, statcode;
315 long contlen, contrange;
319 int auth_tried_already;
322 /* Let the others worry about local filename... */
323 if (!(*dt & HEAD_ONLY))
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 hstat. */
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 qstring = u->qstring;
390 command = (*dt & HEAD_ONLY) ? "HEAD" : "GET";
394 referer = (char *)alloca (9 + strlen (ou->referer) + 3);
395 sprintf (referer, "Referer: %s\r\n", ou->referer);
397 if (*dt & SEND_NOCACHE)
398 pragma_h = "Pragma: no-cache\r\n";
403 range = (char *)alloca (13 + numdigit (hs->restval) + 4);
404 /* #### Gag me! Some servers (e.g. WebSitePro) have been known
405 to misinterpret the following `Range' format, and return the
406 document as multipart/x-byte-ranges MIME type!
408 #### TODO: Interpret MIME types, recognize bullshits similar
409 the one described above, and deal with them! */
410 sprintf (range, "Range: bytes=%ld-\r\n", hs->restval);
415 STRDUP_ALLOCA (useragent, opt.useragent);
418 useragent = (char *)alloca (10 + strlen (version_string));
419 sprintf (useragent, "Wget/%s", version_string);
421 /* Construct the authentication, if userid is present. */
424 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
425 user = user ? user : opt.http_user;
426 passwd = passwd ? passwd : opt.http_passwd;
429 if (authenticate_h && user && passwd)
431 wwwauth = create_authorization_line (authenticate_h, user, passwd,
438 char *proxy_user, *proxy_passwd;
439 /* For normal username and password, URL components override
440 command-line/wgetrc parameters. With proxy authentication,
441 it's the reverse, because proxy URLs are normally the
442 "permanent" ones, so command-line args should take
444 if (opt.proxy_user && opt.proxy_passwd)
446 proxy_user = opt.proxy_user;
447 proxy_passwd = opt.proxy_passwd;
451 proxy_user = u->user;
452 proxy_passwd = u->passwd;
454 /* #### This is junky. Can't the proxy request, say, `Digest'
456 if (proxy_user && proxy_passwd)
457 proxyauth = basic_authentication_encode (proxy_user, proxy_passwd,
458 "Proxy-Authorization");
464 host_port = NULL; host_port_len = 0;
467 host_port = (char *)alloca (numdigit (remport) + 2);
468 host_port_len = sprintf (host_port, ":%d", remport);
471 /* Allocate the memory for the request. */
472 request = (char *)alloca (strlen (command) + strlen (path)
473 + (qstring ? strlen (qstring) : 0)
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. */
486 %s %s%s HTTP/1.0\r\n\
491 command, path, qstring ? qstring : "", 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"));
513 logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
514 u->proxy ? "Proxy" : "HTTP");
515 contlen = contrange = -1;
520 /* Before reading anything, initialize the rbuf. */
521 rbuf_initialize (&rbuf, sock);
525 /* Header-fetching loop. */
533 /* Get the header. */
534 status = header_get (&rbuf, &hdr,
535 /* Disallow continuations for status line. */
536 (hcount == 1 ? HG_NO_CONTINUATIONS : HG_NONE));
538 /* Check for errors. */
539 if (status == HG_EOF && *hdr)
541 /* This used to be an unconditional error, but that was
542 somewhat controversial, because of a large number of
543 broken CGI's that happily "forget" to send the second EOL
544 before closing the connection of a HEAD request.
546 So, the deal is to check whether the header is empty
547 (*hdr is zero if it is); if yes, it means that the
548 previous header was fully retrieved, and that -- most
549 probably -- the request is complete. "...be liberal in
550 what you accept." Oh boy. */
551 logputs (LOG_VERBOSE, "\n");
552 logputs (LOG_NOTQUIET, _("End of file while parsing headers.\n"));
555 FREE_MAYBE (hs->newloc);
556 FREE_MAYBE (all_headers);
560 else if (status == HG_ERROR)
562 logputs (LOG_VERBOSE, "\n");
563 logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
567 FREE_MAYBE (hs->newloc);
568 FREE_MAYBE (all_headers);
573 /* If the headers are to be saved to a file later, save them to
575 if (opt.save_headers)
577 int lh = strlen (hdr);
578 all_headers = (char *)xrealloc (all_headers, all_length + lh + 2);
579 memcpy (all_headers + all_length, hdr, lh);
581 all_headers[all_length++] = '\n';
582 all_headers[all_length] = '\0';
585 /* Print the header if requested. */
586 if (opt.server_response && hcount != 1)
587 logprintf (LOG_VERBOSE, "\n%d %s", hcount, hdr);
589 /* Check for status line. */
593 /* Parse the first line of server response. */
594 statcode = parse_http_status_line (hdr, &error);
595 hs->statcode = statcode;
596 /* Store the descriptive response. */
597 if (statcode == -1) /* malformed response */
599 /* A common reason for "malformed response" error is the
600 case when no data was actually received. Handle this
603 hs->error = xstrdup (_("No data received"));
605 hs->error = xstrdup (_("Malformed status line"));
610 hs->error = xstrdup (_("(no description)"));
612 hs->error = xstrdup (error);
619 logprintf (LOG_VERBOSE, "%d %s", statcode, error);
624 /* Exit on empty header. */
631 /* Try getting content-length. */
632 if (contlen == -1 && !opt.ignore_length)
633 if (header_process (hdr, "Content-Length", header_extract_number,
636 /* Try getting content-type. */
638 if (header_process (hdr, "Content-Type", http_process_type, &type))
640 /* Try getting location. */
642 if (header_process (hdr, "Location", header_strdup, &hs->newloc))
644 /* Try getting last-modified. */
645 if (!hs->remote_time)
646 if (header_process (hdr, "Last-Modified", header_strdup,
649 /* Try getting www-authentication. */
651 if (header_process (hdr, "WWW-Authenticate", header_strdup,
654 /* Check for accept-ranges header. If it contains the word
655 `none', disable the ranges. */
656 if (*dt & ACCEPTRANGES)
659 if (header_process (hdr, "Accept-Ranges", http_process_none, &nonep))
662 *dt &= ~ACCEPTRANGES;
666 /* Try getting content-range. */
669 struct http_process_range_closure closure;
670 if (header_process (hdr, "Content-Range", http_process_range, &closure))
672 contrange = closure.first_byte_pos;
680 logputs (LOG_VERBOSE, "\n");
682 if ((statcode == HTTP_STATUS_UNAUTHORIZED)
685 /* Authorization is required. */
690 if (auth_tried_already)
692 /* If we have tried it already, then there is not point
694 logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
695 free (authenticate_h);
698 else if (!known_authentication_scheme_p (authenticate_h))
700 free (authenticate_h);
701 logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
706 auth_tried_already = 1;
710 /* We do not need this anymore. */
713 free (authenticate_h);
714 authenticate_h = NULL;
717 /* 20x responses are counted among successful by default. */
718 if (H_20X (statcode))
721 if (type && !strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S)))
724 /* We don't assume text/html by default. */
729 else if (contrange != hs->restval ||
730 (H_PARTIAL (statcode) && contrange == -1))
732 /* This means the whole request was somehow misunderstood by the
735 FREE_MAYBE (hs->newloc);
736 FREE_MAYBE (all_headers);
744 contlen += contrange;
746 contrange = -1; /* If conent-length was not sent,
747 content-range will be ignored. */
749 hs->contlen = contlen;
751 /* Return if redirected. */
752 if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES)
754 /* RFC2068 says that in case of the 300 (multiple choices)
755 response, the server can output a preferred URL through
756 `Location' header; otherwise, the request should be treated
757 like GET. So, if the location is set, it will be a
758 redirection; otherwise, just proceed normally. */
759 if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc)
763 logprintf (LOG_VERBOSE,
764 _("Location: %s%s\n"),
765 hs->newloc ? hs->newloc : _("unspecified"),
766 hs->newloc ? _(" [following]") : "");
769 FREE_MAYBE (all_headers);
775 if ((*dt & RETROKF) && !opt.server_response)
777 /* No need to print this output if the body won't be
778 downloaded at all, or if the original server response is
780 logputs (LOG_VERBOSE, _("Length: "));
783 logputs (LOG_VERBOSE, legible (contlen));
785 logprintf (LOG_VERBOSE, _(" (%s to go)"),
786 legible (contlen - contrange));
789 logputs (LOG_VERBOSE,
790 opt.ignore_length ? _("ignored") : _("unspecified"));
792 logprintf (LOG_VERBOSE, " [%s]\n", type);
794 logputs (LOG_VERBOSE, "\n");
798 type = NULL; /* We don't need it any more. */
800 /* Return if we have no intention of further downloading. */
801 if (!(*dt & RETROKF) || (*dt & HEAD_ONLY))
803 /* In case someone cares to look... */
807 FREE_MAYBE (all_headers);
812 /* Open the local file. */
815 mkalldirs (u->local);
817 rotate_backups (u->local);
818 fp = fopen (u->local, hs->restval ? "ab" : "wb");
821 logprintf (LOG_NOTQUIET, "%s: %s\n", u->local, strerror (errno));
823 FREE_MAYBE (all_headers);
830 /* #### This confuses the code that checks for file size. There
831 should be some overhead information. */
832 if (opt.save_headers)
833 fwrite (all_headers, 1, all_length, fp);
835 /* Get the contents of the document. */
836 hs->res = get_contents (sock, fp, &hs->len, hs->restval,
837 (contlen != -1 ? contlen : 0),
839 hs->dltime = elapsed_time ();
844 FREE_MAYBE (all_headers);
851 /* The genuine HTTP loop! This is the part where the retrieval is
852 retried, and retried, and retried, and... */
854 http_loop (struct urlinfo *u, char **newloc, int *dt)
856 static int first_retrieval = 1;
859 int local_dot_orig_file_exists = FALSE;
860 int use_ts, got_head = 0; /* time-stamping info */
861 char *tms, *suf, *locf, *tmrate;
863 time_t tml = -1, tmr = -1; /* local and remote time-stamps */
864 long local_size = 0; /* the size of the local file */
865 struct http_stat hstat; /* HTTP status */
870 /* Warn on (likely bogus) wildcard usage in HTTP. Don't use
871 has_wildcards_p because it would also warn on `?', and we know that
872 shows up in CGI paths a *lot*. */
873 if (strchr (u->url, '*'))
874 logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
876 /* Determine the local filename. */
878 u->local = url_filename (u->proxy ? u->proxy : u);
880 if (!opt.output_document)
883 locf = opt.output_document;
885 if (opt.noclobber && file_exists_p (u->local))
887 /* If opt.noclobber is turned on and file already exists, do not
889 logprintf (LOG_VERBOSE, _("\
890 File `%s' already there, will not retrieve.\n"), u->local);
891 /* If the file is there, we suppose it's retrieved OK. */
894 /* #### Bogusness alert. */
895 /* If its suffix is "html" or (yuck!) "htm", we suppose it's
896 text/html, a harmless lie. */
897 if (((suf = suffix (u->local)) != NULL)
898 && (!strcmp (suf, "html") || !strcmp (suf, "htm")))
901 /* Another harmless lie: */
906 if (opt.timestamping)
908 boolean local_file_exists = FALSE;
910 if (opt.backup_converted)
911 /* If -K is specified, we'll act on the assumption that it was specified
912 last time these files were downloaded as well, and instead of just
913 comparing local file X against server file X, we'll compare local
914 file X.orig (if extant, else X) against server file X. If -K
915 _wasn't_ specified last time, or the server contains files called
916 *.orig, -N will be back to not operating correctly with -k. */
918 size_t filename_len = strlen(u->local);
919 char* filename_plus_orig_suffix = malloc(filename_len +
922 /* Would a single s[n]printf() call be faster? */
923 strcpy(filename_plus_orig_suffix, u->local);
924 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
926 /* Try to stat() the .orig file. */
927 if (stat(filename_plus_orig_suffix, &st) == 0)
929 local_file_exists = TRUE;
930 local_dot_orig_file_exists = TRUE;
933 free(filename_plus_orig_suffix);
936 if (!local_dot_orig_file_exists)
937 /* Couldn't stat() <file>.orig, so try to stat() <file>. */
938 if (stat (u->local, &st) == 0)
939 local_file_exists = TRUE;
941 if (local_file_exists)
942 /* There was a local file, so we'll check later to see if the version
943 the server has is the same version we already have, allowing us to
948 local_size = st.st_size;
952 /* Reset the counter. */
954 *dt = 0 | ACCEPTRANGES;
958 /* Increment the pass counter. */
960 /* Wait before the retrieval (unless this is the very first
962 Check if we are retrying or not, wait accordingly - HEH */
963 if (!first_retrieval && (opt.wait || (count && opt.waitretry)))
967 if (count<opt.waitretry)
970 sleep(opt.waitretry);
977 /* Get the current time string. */
978 tms = time_str (NULL);
979 /* Print fetch message, if opt.verbose. */
982 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
986 sprintf (tmp, _("(try:%2d)"), count);
987 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
988 tms, hurl, tmp, locf);
990 ws_changetitle (hurl, 1);
995 /* Default document type is empty. However, if spider mode is
996 on or time-stamping is employed, HEAD_ONLY commands is
997 encoded within *dt. */
998 if (opt.spider || (use_ts && !got_head))
1002 /* Assume no restarting. */
1004 /* Decide whether or not to restart. */
1005 if (((count > 1 && (*dt & ACCEPTRANGES)) || opt.always_rest)
1006 && file_exists_p (u->local))
1007 if (stat (u->local, &st) == 0)
1008 hstat.restval = st.st_size;
1009 /* Decide whether to send the no-cache directive. */
1010 if (u->proxy && (count > 1 || (opt.proxy_cache == 0)))
1011 *dt |= SEND_NOCACHE;
1013 *dt &= ~SEND_NOCACHE;
1015 /* Try fetching the document, or at least its head. :-) */
1016 err = gethttp (u, &hstat, dt);
1018 tms = time_str (NULL);
1019 /* Get the new location (with or without the redirection). */
1021 *newloc = xstrdup (hstat.newloc);
1024 case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
1025 case CONERROR: case READERR: case WRITEFAILED:
1027 /* Non-fatal errors continue executing the loop, which will
1028 bring them to "while" statement at the end, to judge
1029 whether the number of tries was exceeded. */
1031 printwhat (count, opt.ntry);
1034 case HOSTERR: case CONREFUSED: case PROXERR: case AUTHFAILED:
1035 /* Fatal errors just return from the function. */
1039 case FWRITEERR: case FOPENERR:
1040 /* Another fatal error. */
1041 logputs (LOG_VERBOSE, "\n");
1042 logprintf (LOG_NOTQUIET, _("Cannot write to `%s' (%s).\n"),
1043 u->local, strerror (errno));
1048 /* Return the new location to the caller. */
1051 logprintf (LOG_NOTQUIET,
1052 _("ERROR: Redirection (%d) without location.\n"),
1060 /* Deal with you later. */
1063 /* All possibilities should have been exhausted. */
1066 if (!(*dt & RETROKF))
1070 /* #### Ugly ugly ugly! */
1071 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
1072 logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
1075 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
1076 tms, hstat.statcode, hstat.error);
1077 logputs (LOG_VERBOSE, "\n");
1082 /* Did we get the time-stamp? */
1085 if (opt.timestamping && !hstat.remote_time)
1087 logputs (LOG_NOTQUIET, _("\
1088 Last-modified header missing -- time-stamps turned off.\n"));
1090 else if (hstat.remote_time)
1092 /* Convert the date-string into struct tm. */
1093 tmr = http_atotm (hstat.remote_time);
1094 if (tmr == (time_t) (-1))
1095 logputs (LOG_VERBOSE, _("\
1096 Last-modified header invalid -- time-stamp ignored.\n"));
1100 /* The time-stamping section. */
1105 use_ts = 0; /* no more time-stamping */
1106 count = 0; /* the retrieve count for HEAD is
1108 if (hstat.remote_time && tmr != (time_t) (-1))
1110 /* Now time-stamping can be used validly. Time-stamping
1111 means that if the sizes of the local and remote file
1112 match, and local file is newer than the remote file,
1113 it will not be retrieved. Otherwise, the normal
1114 download procedure is resumed. */
1116 (hstat.contlen == -1 || local_size == hstat.contlen))
1118 if (local_dot_orig_file_exists)
1119 /* We can't collapse this down into just one logprintf()
1120 call with a variable set to u->local or the .orig
1121 filename because we have to malloc() space for the
1122 latter, and because there are multiple returns above (a
1123 coding style no-no by many measures, for reasons such as
1124 this) we'd have to remember to free() the string at each
1125 one to avoid a memory leak. */
1126 logprintf (LOG_VERBOSE, _("\
1127 Server file no newer than local file `%s.orig' -- not retrieving.\n\n"),
1130 logprintf (LOG_VERBOSE, _("\
1131 Server file no newer than local file `%s' -- not retrieving.\n\n"), u->local);
1135 else if (tml >= tmr)
1136 logprintf (LOG_VERBOSE, _("\
1137 The sizes do not match (local %ld) -- retrieving.\n"), local_size);
1139 logputs (LOG_VERBOSE,
1140 _("Remote file is newer, retrieving.\n"));
1146 && (tmr != (time_t) (-1))
1148 && ((hstat.len == hstat.contlen) ||
1149 ((hstat.res == 0) &&
1150 ((hstat.contlen == -1) ||
1151 (hstat.len >= hstat.contlen && !opt.kill_longer)))))
1153 touch (u->local, tmr);
1155 /* End of time-stamping section. */
1159 logprintf (LOG_NOTQUIET, "%d %s\n\n", hstat.statcode, hstat.error);
1163 /* It is now safe to free the remainder of hstat, since the
1164 strings within it will no longer be used. */
1167 tmrate = rate (hstat.len - hstat.restval, hstat.dltime);
1169 if (hstat.len == hstat.contlen)
1173 logprintf (LOG_VERBOSE,
1174 _("%s (%s) - `%s' saved [%ld/%ld]\n\n"),
1175 tms, tmrate, locf, hstat.len, hstat.contlen);
1176 logprintf (LOG_NONVERBOSE,
1177 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1178 tms, u->url, hstat.len, hstat.contlen, locf, count);
1181 opt.downloaded += hstat.len;
1182 downloaded_file(ADD_FILE, locf);
1185 else if (hstat.res == 0) /* No read error */
1187 if (hstat.contlen == -1) /* We don't know how much we were supposed
1188 to get, so assume we succeeded. */
1192 logprintf (LOG_VERBOSE,
1193 _("%s (%s) - `%s' saved [%ld]\n\n"),
1194 tms, tmrate, locf, hstat.len);
1195 logprintf (LOG_NONVERBOSE,
1196 "%s URL:%s [%ld] -> \"%s\" [%d]\n",
1197 tms, u->url, hstat.len, locf, count);
1200 opt.downloaded += hstat.len;
1201 downloaded_file(ADD_FILE, locf);
1204 else if (hstat.len < hstat.contlen) /* meaning we lost the
1205 connection too soon */
1207 logprintf (LOG_VERBOSE,
1208 _("%s (%s) - Connection closed at byte %ld. "),
1209 tms, tmrate, hstat.len);
1210 printwhat (count, opt.ntry);
1213 else if (!opt.kill_longer) /* meaning we got more than expected */
1215 logprintf (LOG_VERBOSE,
1216 _("%s (%s) - `%s' saved [%ld/%ld])\n\n"),
1217 tms, tmrate, locf, hstat.len, hstat.contlen);
1218 logprintf (LOG_NONVERBOSE,
1219 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1220 tms, u->url, hstat.len, hstat.contlen, locf, count);
1222 opt.downloaded += hstat.len;
1223 downloaded_file(ADD_FILE, locf);
1226 else /* the same, but not accepted */
1228 logprintf (LOG_VERBOSE,
1229 _("%s (%s) - Connection closed at byte %ld/%ld. "),
1230 tms, tmrate, hstat.len, hstat.contlen);
1231 printwhat (count, opt.ntry);
1235 else /* now hstat.res can only be -1 */
1237 if (hstat.contlen == -1)
1239 logprintf (LOG_VERBOSE,
1240 _("%s (%s) - Read error at byte %ld (%s)."),
1241 tms, tmrate, hstat.len, strerror (errno));
1242 printwhat (count, opt.ntry);
1245 else /* hstat.res == -1 and contlen is given */
1247 logprintf (LOG_VERBOSE,
1248 _("%s (%s) - Read error at byte %ld/%ld (%s). "),
1249 tms, tmrate, hstat.len, hstat.contlen,
1251 printwhat (count, opt.ntry);
1258 while (!opt.ntry || (count < opt.ntry));
1262 /* Converts struct tm to time_t, assuming the data in tm is UTC rather
1263 than local timezone (mktime assumes the latter).
1265 Contributed by Roger Beeman <beeman@cisco.com>, with the help of
1266 Mark Baushke <mdb@cisco.com> and the rest of the Gurus at CISCO. */
1268 mktime_from_utc (struct tm *t)
1275 tb = mktime (gmtime (&tl));
1276 return (tl <= tb ? (tl + (tl - tb)) : (tl - (tb - tl)));
1279 /* Check whether the result of strptime() indicates success.
1280 strptime() returns the pointer to how far it got to in the string.
1281 The processing has been successful if the string is at `GMT' or
1282 `+X', or at the end of the string.
1284 In extended regexp parlance, the function returns 1 if P matches
1285 "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (a valid result of
1286 strptime()) is considered a failure and 0 is returned. */
1292 while (ISSPACE (*p))
1295 || (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
1296 || ((p[0] == '+' || p[1] == '-') && ISDIGIT (p[1])))
1302 /* Convert TIME_STRING time to time_t. TIME_STRING can be in any of
1303 the three formats RFC2068 allows the HTTP servers to emit --
1304 RFC1123-date, RFC850-date or asctime-date. Timezones are ignored,
1307 We use strptime() to recognize various dates, which makes it a
1308 little bit slacker than the RFC1123/RFC850/asctime (e.g. it always
1309 allows shortened dates and months, one-digit days, etc.). It also
1310 allows more than one space anywhere where the specs require one SP.
1311 The routine should probably be even more forgiving (as recommended
1312 by RFC2068), but I do not have the time to write one.
1314 Return the computed time_t representation, or -1 if all the
1317 Needless to say, what we *really* need here is something like
1318 Marcus Hennecke's atotm(), which is forgiving, fast, to-the-point,
1319 and does not use strptime(). atotm() is to be found in the sources
1320 of `phttpd', a little-known HTTP server written by Peter Erikson. */
1322 http_atotm (char *time_string)
1326 /* Roger Beeman says: "This function dynamically allocates struct tm
1327 t, but does no initialization. The only field that actually
1328 needs initialization is tm_isdst, since the others will be set by
1329 strptime. Since strptime does not set tm_isdst, it will return
1330 the data structure with whatever data was in tm_isdst to begin
1331 with. For those of us in timezones where DST can occur, there
1332 can be a one hour shift depending on the previous contents of the
1333 data area where the data structure is allocated." */
1336 /* Note that under foreign locales Solaris strptime() fails to
1337 recognize English dates, which renders this function useless. I
1338 assume that other non-GNU strptime's are plagued by the same
1339 disease. We solve this by setting only LC_MESSAGES in
1340 i18n_initialize(), instead of LC_ALL.
1342 Another solution could be to temporarily set locale to C, invoke
1343 strptime(), and restore it back. This is slow and dirty,
1344 however, and locale support other than LC_MESSAGES can mess other
1345 things, so I rather chose to stick with just setting LC_MESSAGES.
1347 Also note that none of this is necessary under GNU strptime(),
1348 because it recognizes both international and local dates. */
1350 /* NOTE: We don't use `%n' for white space, as OSF's strptime uses
1351 it to eat all white space up to (and including) a newline, and
1352 the function fails if there is no newline (!).
1354 Let's hope all strptime() implementations use ` ' to skip *all*
1355 whitespace instead of just one (it works that way on all the
1356 systems I've tested it on). */
1358 /* RFC1123: Thu, 29 Jan 1998 22:12:57 */
1359 if (check_end (strptime (time_string, "%a, %d %b %Y %T", &t)))
1360 return mktime_from_utc (&t);
1361 /* RFC850: Thu, 29-Jan-98 22:12:57 */
1362 if (check_end (strptime (time_string, "%a, %d-%b-%y %T", &t)))
1363 return mktime_from_utc (&t);
1364 /* asctime: Thu Jan 29 22:12:57 1998 */
1365 if (check_end (strptime (time_string, "%a %b %d %T %Y", &t)))
1366 return mktime_from_utc (&t);
1371 /* Authorization support: We support two authorization schemes:
1373 * `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
1375 * `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
1376 consisting of answering to the server's challenge with the proper
1379 /* How many bytes it will take to store LEN bytes in base64. */
1380 #define BASE64_LENGTH(len) (4 * (((len) + 2) / 3))
1382 /* Encode the string S of length LENGTH to base64 format and place it
1383 to STORE. STORE will be 0-terminated, and must point to a writable
1384 buffer of at least 1+BASE64_LENGTH(length) bytes. */
1386 base64_encode (const char *s, char *store, int length)
1388 /* Conversion table. */
1389 static char tbl[64] = {
1390 'A','B','C','D','E','F','G','H',
1391 'I','J','K','L','M','N','O','P',
1392 'Q','R','S','T','U','V','W','X',
1393 'Y','Z','a','b','c','d','e','f',
1394 'g','h','i','j','k','l','m','n',
1395 'o','p','q','r','s','t','u','v',
1396 'w','x','y','z','0','1','2','3',
1397 '4','5','6','7','8','9','+','/'
1400 unsigned char *p = (unsigned char *)store;
1402 /* Transform the 3x8 bits to 4x6 bits, as required by base64. */
1403 for (i = 0; i < length; i += 3)
1405 *p++ = tbl[s[0] >> 2];
1406 *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
1407 *p++ = tbl[((s[1] & 0xf) << 2) + (s[2] >> 6)];
1408 *p++ = tbl[s[2] & 0x3f];
1411 /* Pad the result if necessary... */
1412 if (i == length + 1)
1414 else if (i == length + 2)
1415 *(p - 1) = *(p - 2) = '=';
1416 /* ...and zero-terminate it. */
1420 /* Create the authentication header contents for the `Basic' scheme.
1421 This is done by encoding the string `USER:PASS' in base64 and
1422 prepending `HEADER: Basic ' to it. */
1424 basic_authentication_encode (const char *user, const char *passwd,
1427 char *t1, *t2, *res;
1428 int len1 = strlen (user) + 1 + strlen (passwd);
1429 int len2 = BASE64_LENGTH (len1);
1431 t1 = (char *)alloca (len1 + 1);
1432 sprintf (t1, "%s:%s", user, passwd);
1433 t2 = (char *)alloca (1 + len2);
1434 base64_encode (t1, t2, len1);
1435 res = (char *)malloc (len2 + 11 + strlen (header));
1436 sprintf (res, "%s: Basic %s\r\n", header, t2);
1442 /* Parse HTTP `WWW-Authenticate:' header. AU points to the beginning
1443 of a field in such a header. If the field is the one specified by
1444 ATTR_NAME ("realm", "opaque", and "nonce" are used by the current
1445 digest authorization code), extract its value in the (char*)
1446 variable pointed by RET. Returns negative on a malformed header,
1447 or number of bytes that have been parsed by this call. */
1449 extract_header_attr (const char *au, const char *attr_name, char **ret)
1451 const char *cp, *ep;
1455 if (strncmp (cp, attr_name, strlen (attr_name)) == 0)
1457 cp += strlen (attr_name);
1460 cp += skip_lws (cp);
1465 cp += skip_lws (cp);
1470 for (ep = cp; *ep && *ep != '\"'; ep++)
1475 *ret = strdupdelim (cp, ep);
1482 /* Response value needs to be in lowercase, so we cannot use HEXD2ASC
1483 from url.h. See RFC 2069 2.1.2 for the syntax of response-digest. */
1484 #define HEXD2asc(x) (((x) < 10) ? ((x) + '0') : ((x) - 10 + 'a'))
1486 /* Dump the hexadecimal representation of HASH to BUF. HASH should be
1487 an array of 16 bytes containing the hash keys, and BUF should be a
1488 buffer of 33 writable characters (32 for hex digits plus one for
1489 zero termination). */
1491 dump_hash (unsigned char *buf, const unsigned char *hash)
1495 for (i = 0; i < MD5_HASHLEN; i++, hash++)
1497 *buf++ = HEXD2asc (*hash >> 4);
1498 *buf++ = HEXD2asc (*hash & 0xf);
1503 /* Take the line apart to find the challenge, and compose a digest
1504 authorization header. See RFC2069 section 2.1.2. */
1506 digest_authentication_encode (const char *au, const char *user,
1507 const char *passwd, const char *method,
1510 static char *realm, *opaque, *nonce;
1515 { "realm", &realm },
1516 { "opaque", &opaque },
1521 realm = opaque = nonce = NULL;
1523 au += 6; /* skip over `Digest' */
1528 au += skip_lws (au);
1529 for (i = 0; i < ARRAY_SIZE (options); i++)
1531 int skip = extract_header_attr (au, options[i].name,
1532 options[i].variable);
1536 FREE_MAYBE (opaque);
1546 if (i == ARRAY_SIZE (options))
1548 while (*au && *au != '=')
1552 au += skip_lws (au);
1556 while (*au && *au != '\"')
1563 while (*au && *au != ',')
1568 if (!realm || !nonce || !user || !passwd || !path || !method)
1571 FREE_MAYBE (opaque);
1576 /* Calculate the digest value. */
1579 unsigned char hash[MD5_HASHLEN];
1580 unsigned char a1buf[MD5_HASHLEN * 2 + 1], a2buf[MD5_HASHLEN * 2 + 1];
1581 unsigned char response_digest[MD5_HASHLEN * 2 + 1];
1583 /* A1BUF = H(user ":" realm ":" password) */
1584 md5_init_ctx (&ctx);
1585 md5_process_bytes (user, strlen (user), &ctx);
1586 md5_process_bytes (":", 1, &ctx);
1587 md5_process_bytes (realm, strlen (realm), &ctx);
1588 md5_process_bytes (":", 1, &ctx);
1589 md5_process_bytes (passwd, strlen (passwd), &ctx);
1590 md5_finish_ctx (&ctx, hash);
1591 dump_hash (a1buf, hash);
1593 /* A2BUF = H(method ":" path) */
1594 md5_init_ctx (&ctx);
1595 md5_process_bytes (method, strlen (method), &ctx);
1596 md5_process_bytes (":", 1, &ctx);
1597 md5_process_bytes (path, strlen (path), &ctx);
1598 md5_finish_ctx (&ctx, hash);
1599 dump_hash (a2buf, hash);
1601 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */
1602 md5_init_ctx (&ctx);
1603 md5_process_bytes (a1buf, MD5_HASHLEN * 2, &ctx);
1604 md5_process_bytes (":", 1, &ctx);
1605 md5_process_bytes (nonce, strlen (nonce), &ctx);
1606 md5_process_bytes (":", 1, &ctx);
1607 md5_process_bytes (a2buf, MD5_HASHLEN * 2, &ctx);
1608 md5_finish_ctx (&ctx, hash);
1609 dump_hash (response_digest, hash);
1611 res = (char*) xmalloc (strlen (user)
1616 + 2 * MD5_HASHLEN /*strlen (response_digest)*/
1617 + (opaque ? strlen (opaque) : 0)
1619 sprintf (res, "Authorization: Digest \
1620 username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
1621 user, realm, nonce, path, response_digest);
1624 char *p = res + strlen (res);
1625 strcat (p, ", opaque=\"");
1629 strcat (res, "\r\n");
1633 #endif /* USE_DIGEST */
1636 #define HACK_O_MATIC(line, string_constant) \
1637 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
1638 && (ISSPACE (line[sizeof (string_constant) - 1]) \
1639 || !line[sizeof (string_constant) - 1]))
1642 known_authentication_scheme_p (const char *au)
1644 return HACK_O_MATIC (au, "Basic") || HACK_O_MATIC (au, "Digest");
1649 /* Create the HTTP authorization request header. When the
1650 `WWW-Authenticate' response header is seen, according to the
1651 authorization scheme specified in that header (`Basic' and `Digest'
1652 are supported by the current implementation), produce an
1653 appropriate HTTP authorization request header. */
1655 create_authorization_line (const char *au, const char *user,
1656 const char *passwd, const char *method,
1659 char *wwwauth = NULL;
1661 if (!strncasecmp (au, "Basic", 5))
1662 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
1664 else if (!strncasecmp (au, "Digest", 6))
1665 wwwauth = digest_authentication_encode (au, user, passwd, method, path);
1666 #endif /* USE_DIGEST */