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 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 /* Let the others worry about local filename... */
322 if (!(*dt & HEAD_ONLY))
323 assert (u->local != NULL);
326 auth_tried_already = 0;
329 /* We need to come back here when the initial attempt to retrieve
330 without authorization header fails. */
332 /* Initialize certain elements of struct hstat. */
337 hs->remote_time = NULL;
340 /* Which structure to use to retrieve the original URL data. */
346 /* First: establish the connection. */
347 logprintf (LOG_VERBOSE, _("Connecting to %s:%hu... "), u->host, u->port);
348 err = make_connection (&sock, u->host, u->port);
352 logputs (LOG_VERBOSE, "\n");
353 logprintf (LOG_NOTQUIET, "%s: %s.\n", u->host, herrmsg (h_errno));
357 logputs (LOG_VERBOSE, "\n");
358 logprintf (LOG_NOTQUIET, "socket: %s\n", strerror (errno));
362 logputs (LOG_VERBOSE, "\n");
363 logprintf (LOG_NOTQUIET,
364 _("Connection to %s:%hu refused.\n"), u->host, u->port);
368 logputs (LOG_VERBOSE, "\n");
369 logprintf (LOG_NOTQUIET, "connect: %s\n", strerror (errno));
374 /* Everything is fine! */
375 logputs (LOG_VERBOSE, _("connected!\n"));
383 path = u->proxy->url;
387 command = (*dt & HEAD_ONLY) ? "HEAD" : "GET";
391 referer = (char *)alloca (9 + strlen (ou->referer) + 3);
392 sprintf (referer, "Referer: %s\r\n", ou->referer);
394 if (*dt & SEND_NOCACHE)
395 pragma_h = "Pragma: no-cache\r\n";
400 range = (char *)alloca (13 + numdigit (hs->restval) + 4);
401 /* #### Gag me! Some servers (e.g. WebSitePro) have been known
402 to misinterpret the following `Range' format, and return the
403 document as multipart/x-byte-ranges MIME type!
405 #### TODO: Interpret MIME types, recognize bullshits similar
406 the one described above, and deal with them! */
407 sprintf (range, "Range: bytes=%ld-\r\n", hs->restval);
412 STRDUP_ALLOCA (useragent, opt.useragent);
415 useragent = (char *)alloca (10 + strlen (version_string));
416 sprintf (useragent, "Wget/%s", version_string);
418 /* Construct the authentication, if userid is present. */
421 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
422 user = user ? user : opt.http_user;
423 passwd = passwd ? passwd : opt.http_passwd;
426 if (authenticate_h && user && passwd)
428 wwwauth = create_authorization_line (authenticate_h, user, passwd,
435 char *proxy_user, *proxy_passwd;
436 /* For normal username and password, URL components override
437 command-line/wgetrc parameters. With proxy authentication,
438 it's the reverse, because proxy URLs are normally the
439 "permanent" ones, so command-line args should take
441 if (opt.proxy_user && opt.proxy_passwd)
443 proxy_user = opt.proxy_user;
444 proxy_passwd = opt.proxy_passwd;
448 proxy_user = u->user;
449 proxy_passwd = u->passwd;
451 /* #### This is junky. Can't the proxy request, say, `Digest'
453 if (proxy_user && proxy_passwd)
454 proxyauth = basic_authentication_encode (proxy_user, proxy_passwd,
455 "Proxy-Authorization");
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)
474 + strlen (remhost) + host_port_len
475 + strlen (HTTP_ACCEPT)
476 + (referer ? strlen (referer) : 0)
477 + (wwwauth ? strlen (wwwauth) : 0)
478 + (proxyauth ? strlen (proxyauth) : 0)
479 + (range ? strlen (range) : 0)
481 + (opt.user_header ? strlen (opt.user_header) : 0)
483 /* Construct the request. */
490 command, path, useragent, remhost,
491 host_port ? host_port : "",
492 HTTP_ACCEPT, referer ? referer : "",
493 wwwauth ? wwwauth : "",
494 proxyauth ? proxyauth : "",
497 opt.user_header ? opt.user_header : "");
498 DEBUGP (("---request begin---\n%s---request end---\n", request));
499 /* Free the temporary memory. */
500 FREE_MAYBE (wwwauth);
501 FREE_MAYBE (proxyauth);
503 /* Send the request to server. */
504 num_written = iwrite (sock, request, strlen (request));
507 logputs (LOG_VERBOSE, _("Failed writing HTTP request.\n"));
511 logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
512 u->proxy ? "Proxy" : "HTTP");
513 contlen = contrange = -1;
518 /* Before reading anything, initialize the rbuf. */
519 rbuf_initialize (&rbuf, sock);
523 /* Header-fetching loop. */
531 /* Get the header. */
532 status = header_get (&rbuf, &hdr,
533 /* Disallow continuations for status line. */
534 (hcount == 1 ? HG_NO_CONTINUATIONS : HG_NONE));
536 /* Check for errors. */
537 if (status == HG_EOF && *hdr)
539 /* This used to be an unconditional error, but that was
540 somewhat controversial, because of a large number of
541 broken CGI's that happily "forget" to send the second EOL
542 before closing the connection of a HEAD request.
544 So, the deal is to check whether the header is empty
545 (*hdr is zero if it is); if yes, it means that the
546 previous header was fully retrieved, and that -- most
547 probably -- the request is complete. "...be liberal in
548 what you accept." Oh boy. */
549 logputs (LOG_VERBOSE, "\n");
550 logputs (LOG_NOTQUIET, _("End of file while parsing headers.\n"));
553 FREE_MAYBE (hs->newloc);
554 FREE_MAYBE (all_headers);
558 else if (status == HG_ERROR)
560 logputs (LOG_VERBOSE, "\n");
561 logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
565 FREE_MAYBE (hs->newloc);
566 FREE_MAYBE (all_headers);
571 /* If the headers are to be saved to a file later, save them to
573 if (opt.save_headers)
575 int lh = strlen (hdr);
576 all_headers = (char *)xrealloc (all_headers, all_length + lh + 2);
577 memcpy (all_headers + all_length, hdr, lh);
579 all_headers[all_length++] = '\n';
580 all_headers[all_length] = '\0';
583 /* Print the header if requested. */
584 if (opt.server_response && hcount != 1)
585 logprintf (LOG_VERBOSE, "\n%d %s", hcount, hdr);
587 /* Check for status line. */
591 /* Parse the first line of server response. */
592 statcode = parse_http_status_line (hdr, &error);
593 hs->statcode = statcode;
594 /* Store the descriptive response. */
595 if (statcode == -1) /* malformed response */
597 /* A common reason for "malformed response" error is the
598 case when no data was actually received. Handle this
601 hs->error = xstrdup (_("No data received"));
603 hs->error = xstrdup (_("Malformed status line"));
608 hs->error = xstrdup (_("(no description)"));
610 hs->error = xstrdup (error);
617 logprintf (LOG_VERBOSE, "%d %s", statcode, error);
622 /* Exit on empty header. */
629 /* Try getting content-length. */
630 if (contlen == -1 && !opt.ignore_length)
631 if (header_process (hdr, "Content-Length", header_extract_number,
634 /* Try getting content-type. */
636 if (header_process (hdr, "Content-Type", http_process_type, &type))
638 /* Try getting location. */
640 if (header_process (hdr, "Location", header_strdup, &hs->newloc))
642 /* Try getting last-modified. */
643 if (!hs->remote_time)
644 if (header_process (hdr, "Last-Modified", header_strdup,
647 /* Try getting www-authentication. */
649 if (header_process (hdr, "WWW-Authenticate", header_strdup,
652 /* Check for accept-ranges header. If it contains the word
653 `none', disable the ranges. */
654 if (*dt & ACCEPTRANGES)
657 if (header_process (hdr, "Accept-Ranges", http_process_none, &nonep))
660 *dt &= ~ACCEPTRANGES;
664 /* Try getting content-range. */
667 struct http_process_range_closure closure;
668 if (header_process (hdr, "Content-Range", http_process_range, &closure))
670 contrange = closure.first_byte_pos;
678 logputs (LOG_VERBOSE, "\n");
680 if ((statcode == HTTP_STATUS_UNAUTHORIZED)
683 /* Authorization is required. */
688 if (auth_tried_already)
690 /* If we have tried it already, then there is not point
692 logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
693 free (authenticate_h);
696 else if (!known_authentication_scheme_p (authenticate_h))
698 free (authenticate_h);
699 logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
704 auth_tried_already = 1;
708 /* We do not need this anymore. */
711 free (authenticate_h);
712 authenticate_h = NULL;
715 /* 20x responses are counted among successful by default. */
716 if (H_20X (statcode))
719 if (type && !strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S)))
722 /* We don't assume text/html by default. */
727 else if (contrange != hs->restval ||
728 (H_PARTIAL (statcode) && contrange == -1))
730 /* This means the whole request was somehow misunderstood by the
733 FREE_MAYBE (hs->newloc);
734 FREE_MAYBE (all_headers);
742 contlen += contrange;
744 contrange = -1; /* If conent-length was not sent,
745 content-range will be ignored. */
747 hs->contlen = contlen;
749 /* Return if redirected. */
750 if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES)
752 /* RFC2068 says that in case of the 300 (multiple choices)
753 response, the server can output a preferred URL through
754 `Location' header; otherwise, the request should be treated
755 like GET. So, if the location is set, it will be a
756 redirection; otherwise, just proceed normally. */
757 if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc)
761 logprintf (LOG_VERBOSE,
762 _("Location: %s%s\n"),
763 hs->newloc ? hs->newloc : _("unspecified"),
764 hs->newloc ? _(" [following]") : "");
767 FREE_MAYBE (all_headers);
773 if ((*dt & RETROKF) && !opt.server_response)
775 /* No need to print this output if the body won't be
776 downloaded at all, or if the original server response is
778 logputs (LOG_VERBOSE, _("Length: "));
781 logputs (LOG_VERBOSE, legible (contlen));
783 logprintf (LOG_VERBOSE, _(" (%s to go)"),
784 legible (contlen - contrange));
787 logputs (LOG_VERBOSE,
788 opt.ignore_length ? _("ignored") : _("unspecified"));
790 logprintf (LOG_VERBOSE, " [%s]\n", type);
792 logputs (LOG_VERBOSE, "\n");
796 type = NULL; /* We don't need it any more. */
798 /* Return if we have no intention of further downloading. */
799 if (!(*dt & RETROKF) || (*dt & HEAD_ONLY))
801 /* In case someone cares to look... */
805 FREE_MAYBE (all_headers);
810 /* Open the local file. */
813 mkalldirs (u->local);
815 rotate_backups (u->local);
816 fp = fopen (u->local, hs->restval ? "ab" : "wb");
819 logprintf (LOG_NOTQUIET, "%s: %s\n", u->local, strerror (errno));
821 FREE_MAYBE (all_headers);
828 /* #### This confuses the code that checks for file size. There
829 should be some overhead information. */
830 if (opt.save_headers)
831 fwrite (all_headers, 1, all_length, fp);
833 /* Get the contents of the document. */
834 hs->res = get_contents (sock, fp, &hs->len, hs->restval,
835 (contlen != -1 ? contlen : 0),
837 hs->dltime = elapsed_time ();
839 /* Close or flush the file. We have to be careful to check for
840 error here. Checking the result of fwrite() is not enough --
841 errors could go unnoticed! */
844 flush_res = fclose (fp);
846 flush_res = fflush (fp);
847 if (flush_res == EOF)
850 FREE_MAYBE (all_headers);
857 /* The genuine HTTP loop! This is the part where the retrieval is
858 retried, and retried, and retried, and... */
860 http_loop (struct urlinfo *u, char **newloc, int *dt)
862 static int first_retrieval = 1;
865 int local_dot_orig_file_exists = FALSE;
866 int use_ts, got_head = 0; /* time-stamping info */
867 char *tms, *suf, *locf, *tmrate;
869 time_t tml = -1, tmr = -1; /* local and remote time-stamps */
870 long local_size = 0; /* the size of the local file */
871 struct http_stat hstat; /* HTTP status */
876 /* Warn on (likely bogus) wildcard usage in HTTP. Don't use
877 has_wildcards_p because it would also warn on `?', and we know that
878 shows up in CGI paths a *lot*. */
879 if (strchr (u->url, '*'))
880 logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
882 /* Determine the local filename. */
884 u->local = url_filename (u->proxy ? u->proxy : u);
886 if (!opt.output_document)
889 locf = opt.output_document;
891 if (opt.noclobber && file_exists_p (u->local))
893 /* If opt.noclobber is turned on and file already exists, do not
895 logprintf (LOG_VERBOSE, _("\
896 File `%s' already there, will not retrieve.\n"), u->local);
897 /* If the file is there, we suppose it's retrieved OK. */
900 /* #### Bogusness alert. */
901 /* If its suffix is "html" or (yuck!) "htm", we suppose it's
902 text/html, a harmless lie. */
903 if (((suf = suffix (u->local)) != NULL)
904 && (!strcmp (suf, "html") || !strcmp (suf, "htm")))
907 /* Another harmless lie: */
912 if (opt.timestamping)
914 boolean local_file_exists = FALSE;
916 if (opt.backup_converted)
917 /* If -K is specified, we'll act on the assumption that it was specified
918 last time these files were downloaded as well, and instead of just
919 comparing local file X against server file X, we'll compare local
920 file X.orig (if extant, else X) against server file X. If -K
921 _wasn't_ specified last time, or the server contains files called
922 *.orig, -N will be back to not operating correctly with -k. */
924 size_t filename_len = strlen(u->local);
925 char* filename_plus_orig_suffix = malloc(filename_len +
928 /* Would a single s[n]printf() call be faster? */
929 strcpy(filename_plus_orig_suffix, u->local);
930 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
932 /* Try to stat() the .orig file. */
933 if (stat(filename_plus_orig_suffix, &st) == 0)
935 local_file_exists = TRUE;
936 local_dot_orig_file_exists = TRUE;
939 free(filename_plus_orig_suffix);
942 if (!local_dot_orig_file_exists)
943 /* Couldn't stat() <file>.orig, so try to stat() <file>. */
944 if (stat (u->local, &st) == 0)
945 local_file_exists = TRUE;
947 if (local_file_exists)
948 /* There was a local file, so we'll check later to see if the version
949 the server has is the same version we already have, allowing us to
954 local_size = st.st_size;
958 /* Reset the counter. */
960 *dt = 0 | ACCEPTRANGES;
964 /* Increment the pass counter. */
966 /* Wait before the retrieval (unless this is the very first
968 Check if we are retrying or not, wait accordingly - HEH */
969 if (!first_retrieval && (opt.wait || (count && opt.waitretry)))
973 if (count<opt.waitretry)
976 sleep(opt.waitretry);
983 /* Get the current time string. */
984 tms = time_str (NULL);
985 /* Print fetch message, if opt.verbose. */
988 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
992 sprintf (tmp, _("(try:%2d)"), count);
993 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
994 tms, hurl, tmp, locf);
996 ws_changetitle (hurl, 1);
1001 /* Default document type is empty. However, if spider mode is
1002 on or time-stamping is employed, HEAD_ONLY commands is
1003 encoded within *dt. */
1004 if (opt.spider || (use_ts && !got_head))
1008 /* Assume no restarting. */
1010 /* Decide whether or not to restart. */
1011 if (((count > 1 && (*dt & ACCEPTRANGES)) || opt.always_rest)
1012 && file_exists_p (u->local))
1013 if (stat (u->local, &st) == 0)
1014 hstat.restval = st.st_size;
1015 /* Decide whether to send the no-cache directive. */
1016 if (u->proxy && (count > 1 || (opt.proxy_cache == 0)))
1017 *dt |= SEND_NOCACHE;
1019 *dt &= ~SEND_NOCACHE;
1021 /* Try fetching the document, or at least its head. :-) */
1022 err = gethttp (u, &hstat, dt);
1024 tms = time_str (NULL);
1025 /* Get the new location (with or without the redirection). */
1027 *newloc = xstrdup (hstat.newloc);
1030 case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
1031 case CONERROR: case READERR: case WRITEFAILED:
1033 /* Non-fatal errors continue executing the loop, which will
1034 bring them to "while" statement at the end, to judge
1035 whether the number of tries was exceeded. */
1037 printwhat (count, opt.ntry);
1040 case HOSTERR: case CONREFUSED: case PROXERR: case AUTHFAILED:
1041 /* Fatal errors just return from the function. */
1045 case FWRITEERR: case FOPENERR:
1046 /* Another fatal error. */
1047 logputs (LOG_VERBOSE, "\n");
1048 logprintf (LOG_NOTQUIET, _("Cannot write to `%s' (%s).\n"),
1049 u->local, strerror (errno));
1054 /* Return the new location to the caller. */
1057 logprintf (LOG_NOTQUIET,
1058 _("ERROR: Redirection (%d) without location.\n"),
1066 /* Deal with you later. */
1069 /* All possibilities should have been exhausted. */
1072 if (!(*dt & RETROKF))
1076 /* #### Ugly ugly ugly! */
1077 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
1078 logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
1081 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
1082 tms, hstat.statcode, hstat.error);
1083 logputs (LOG_VERBOSE, "\n");
1088 /* Did we get the time-stamp? */
1091 if (opt.timestamping && !hstat.remote_time)
1093 logputs (LOG_NOTQUIET, _("\
1094 Last-modified header missing -- time-stamps turned off.\n"));
1096 else if (hstat.remote_time)
1098 /* Convert the date-string into struct tm. */
1099 tmr = http_atotm (hstat.remote_time);
1100 if (tmr == (time_t) (-1))
1101 logputs (LOG_VERBOSE, _("\
1102 Last-modified header invalid -- time-stamp ignored.\n"));
1106 /* The time-stamping section. */
1111 use_ts = 0; /* no more time-stamping */
1112 count = 0; /* the retrieve count for HEAD is
1114 if (hstat.remote_time && tmr != (time_t) (-1))
1116 /* Now time-stamping can be used validly. Time-stamping
1117 means that if the sizes of the local and remote file
1118 match, and local file is newer than the remote file,
1119 it will not be retrieved. Otherwise, the normal
1120 download procedure is resumed. */
1122 (hstat.contlen == -1 || local_size == hstat.contlen))
1124 if (local_dot_orig_file_exists)
1125 /* We can't collapse this down into just one logprintf()
1126 call with a variable set to u->local or the .orig
1127 filename because we have to malloc() space for the
1128 latter, and because there are multiple returns above (a
1129 coding style no-no by many measures, for reasons such as
1130 this) we'd have to remember to free() the string at each
1131 one to avoid a memory leak. */
1132 logprintf (LOG_VERBOSE, _("\
1133 Server file no newer than local file `%s.orig' -- not retrieving.\n\n"),
1136 logprintf (LOG_VERBOSE, _("\
1137 Server file no newer than local file `%s' -- not retrieving.\n\n"), u->local);
1141 else if (tml >= tmr)
1142 logprintf (LOG_VERBOSE, _("\
1143 The sizes do not match (local %ld) -- retrieving.\n"), local_size);
1145 logputs (LOG_VERBOSE,
1146 _("Remote file is newer, retrieving.\n"));
1152 && (tmr != (time_t) (-1))
1154 && ((hstat.len == hstat.contlen) ||
1155 ((hstat.res == 0) &&
1156 ((hstat.contlen == -1) ||
1157 (hstat.len >= hstat.contlen && !opt.kill_longer)))))
1159 touch (u->local, tmr);
1161 /* End of time-stamping section. */
1165 logprintf (LOG_NOTQUIET, "%d %s\n\n", hstat.statcode, hstat.error);
1169 /* It is now safe to free the remainder of hstat, since the
1170 strings within it will no longer be used. */
1173 tmrate = rate (hstat.len - hstat.restval, hstat.dltime);
1175 if (hstat.len == hstat.contlen)
1179 logprintf (LOG_VERBOSE,
1180 _("%s (%s) - `%s' saved [%ld/%ld]\n\n"),
1181 tms, tmrate, locf, hstat.len, hstat.contlen);
1182 logprintf (LOG_NONVERBOSE,
1183 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1184 tms, u->url, hstat.len, hstat.contlen, locf, count);
1187 opt.downloaded += hstat.len;
1188 downloaded_file(ADD_FILE, locf);
1191 else if (hstat.res == 0) /* No read error */
1193 if (hstat.contlen == -1) /* We don't know how much we were supposed
1194 to get, so assume we succeeded. */
1198 logprintf (LOG_VERBOSE,
1199 _("%s (%s) - `%s' saved [%ld]\n\n"),
1200 tms, tmrate, locf, hstat.len);
1201 logprintf (LOG_NONVERBOSE,
1202 "%s URL:%s [%ld] -> \"%s\" [%d]\n",
1203 tms, u->url, hstat.len, locf, count);
1206 opt.downloaded += hstat.len;
1207 downloaded_file(ADD_FILE, locf);
1210 else if (hstat.len < hstat.contlen) /* meaning we lost the
1211 connection too soon */
1213 logprintf (LOG_VERBOSE,
1214 _("%s (%s) - Connection closed at byte %ld. "),
1215 tms, tmrate, hstat.len);
1216 printwhat (count, opt.ntry);
1219 else if (!opt.kill_longer) /* meaning we got more than expected */
1221 logprintf (LOG_VERBOSE,
1222 _("%s (%s) - `%s' saved [%ld/%ld])\n\n"),
1223 tms, tmrate, locf, hstat.len, hstat.contlen);
1224 logprintf (LOG_NONVERBOSE,
1225 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1226 tms, u->url, hstat.len, hstat.contlen, locf, count);
1228 opt.downloaded += hstat.len;
1229 downloaded_file(ADD_FILE, locf);
1232 else /* the same, but not accepted */
1234 logprintf (LOG_VERBOSE,
1235 _("%s (%s) - Connection closed at byte %ld/%ld. "),
1236 tms, tmrate, hstat.len, hstat.contlen);
1237 printwhat (count, opt.ntry);
1241 else /* now hstat.res can only be -1 */
1243 if (hstat.contlen == -1)
1245 logprintf (LOG_VERBOSE,
1246 _("%s (%s) - Read error at byte %ld (%s)."),
1247 tms, tmrate, hstat.len, strerror (errno));
1248 printwhat (count, opt.ntry);
1251 else /* hstat.res == -1 and contlen is given */
1253 logprintf (LOG_VERBOSE,
1254 _("%s (%s) - Read error at byte %ld/%ld (%s). "),
1255 tms, tmrate, hstat.len, hstat.contlen,
1257 printwhat (count, opt.ntry);
1264 while (!opt.ntry || (count < opt.ntry));
1268 /* Converts struct tm to time_t, assuming the data in tm is UTC rather
1269 than local timezone (mktime assumes the latter).
1271 Contributed by Roger Beeman <beeman@cisco.com>, with the help of
1272 Mark Baushke <mdb@cisco.com> and the rest of the Gurus at CISCO. */
1274 mktime_from_utc (struct tm *t)
1281 tb = mktime (gmtime (&tl));
1282 return (tl <= tb ? (tl + (tl - tb)) : (tl - (tb - tl)));
1285 /* Check whether the result of strptime() indicates success.
1286 strptime() returns the pointer to how far it got to in the string.
1287 The processing has been successful if the string is at `GMT' or
1288 `+X', or at the end of the string.
1290 In extended regexp parlance, the function returns 1 if P matches
1291 "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (a valid result of
1292 strptime()) is considered a failure and 0 is returned. */
1298 while (ISSPACE (*p))
1301 || (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
1302 || ((p[0] == '+' || p[1] == '-') && ISDIGIT (p[1])))
1308 /* Convert TIME_STRING time to time_t. TIME_STRING can be in any of
1309 the three formats RFC2068 allows the HTTP servers to emit --
1310 RFC1123-date, RFC850-date or asctime-date. Timezones are ignored,
1313 We use strptime() to recognize various dates, which makes it a
1314 little bit slacker than the RFC1123/RFC850/asctime (e.g. it always
1315 allows shortened dates and months, one-digit days, etc.). It also
1316 allows more than one space anywhere where the specs require one SP.
1317 The routine should probably be even more forgiving (as recommended
1318 by RFC2068), but I do not have the time to write one.
1320 Return the computed time_t representation, or -1 if all the
1323 Needless to say, what we *really* need here is something like
1324 Marcus Hennecke's atotm(), which is forgiving, fast, to-the-point,
1325 and does not use strptime(). atotm() is to be found in the sources
1326 of `phttpd', a little-known HTTP server written by Peter Erikson. */
1328 http_atotm (char *time_string)
1332 /* Roger Beeman says: "This function dynamically allocates struct tm
1333 t, but does no initialization. The only field that actually
1334 needs initialization is tm_isdst, since the others will be set by
1335 strptime. Since strptime does not set tm_isdst, it will return
1336 the data structure with whatever data was in tm_isdst to begin
1337 with. For those of us in timezones where DST can occur, there
1338 can be a one hour shift depending on the previous contents of the
1339 data area where the data structure is allocated." */
1342 /* Note that under foreign locales Solaris strptime() fails to
1343 recognize English dates, which renders this function useless. I
1344 assume that other non-GNU strptime's are plagued by the same
1345 disease. We solve this by setting only LC_MESSAGES in
1346 i18n_initialize(), instead of LC_ALL.
1348 Another solution could be to temporarily set locale to C, invoke
1349 strptime(), and restore it back. This is slow and dirty,
1350 however, and locale support other than LC_MESSAGES can mess other
1351 things, so I rather chose to stick with just setting LC_MESSAGES.
1353 Also note that none of this is necessary under GNU strptime(),
1354 because it recognizes both international and local dates. */
1356 /* NOTE: We don't use `%n' for white space, as OSF's strptime uses
1357 it to eat all white space up to (and including) a newline, and
1358 the function fails if there is no newline (!).
1360 Let's hope all strptime() implementations use ` ' to skip *all*
1361 whitespace instead of just one (it works that way on all the
1362 systems I've tested it on). */
1364 /* RFC1123: Thu, 29 Jan 1998 22:12:57 */
1365 if (check_end (strptime (time_string, "%a, %d %b %Y %T", &t)))
1366 return mktime_from_utc (&t);
1367 /* RFC850: Thu, 29-Jan-98 22:12:57 */
1368 if (check_end (strptime (time_string, "%a, %d-%b-%y %T", &t)))
1369 return mktime_from_utc (&t);
1370 /* asctime: Thu Jan 29 22:12:57 1998 */
1371 if (check_end (strptime (time_string, "%a %b %d %T %Y", &t)))
1372 return mktime_from_utc (&t);
1377 /* Authorization support: We support two authorization schemes:
1379 * `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
1381 * `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
1382 consisting of answering to the server's challenge with the proper
1385 /* How many bytes it will take to store LEN bytes in base64. */
1386 #define BASE64_LENGTH(len) (4 * (((len) + 2) / 3))
1388 /* Encode the string S of length LENGTH to base64 format and place it
1389 to STORE. STORE will be 0-terminated, and must point to a writable
1390 buffer of at least 1+BASE64_LENGTH(length) bytes. */
1392 base64_encode (const char *s, char *store, int length)
1394 /* Conversion table. */
1395 static char tbl[64] = {
1396 'A','B','C','D','E','F','G','H',
1397 'I','J','K','L','M','N','O','P',
1398 'Q','R','S','T','U','V','W','X',
1399 'Y','Z','a','b','c','d','e','f',
1400 'g','h','i','j','k','l','m','n',
1401 'o','p','q','r','s','t','u','v',
1402 'w','x','y','z','0','1','2','3',
1403 '4','5','6','7','8','9','+','/'
1406 unsigned char *p = (unsigned char *)store;
1408 /* Transform the 3x8 bits to 4x6 bits, as required by base64. */
1409 for (i = 0; i < length; i += 3)
1411 *p++ = tbl[s[0] >> 2];
1412 *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
1413 *p++ = tbl[((s[1] & 0xf) << 2) + (s[2] >> 6)];
1414 *p++ = tbl[s[2] & 0x3f];
1417 /* Pad the result if necessary... */
1418 if (i == length + 1)
1420 else if (i == length + 2)
1421 *(p - 1) = *(p - 2) = '=';
1422 /* ...and zero-terminate it. */
1426 /* Create the authentication header contents for the `Basic' scheme.
1427 This is done by encoding the string `USER:PASS' in base64 and
1428 prepending `HEADER: Basic ' to it. */
1430 basic_authentication_encode (const char *user, const char *passwd,
1433 char *t1, *t2, *res;
1434 int len1 = strlen (user) + 1 + strlen (passwd);
1435 int len2 = BASE64_LENGTH (len1);
1437 t1 = (char *)alloca (len1 + 1);
1438 sprintf (t1, "%s:%s", user, passwd);
1439 t2 = (char *)alloca (1 + len2);
1440 base64_encode (t1, t2, len1);
1441 res = (char *)malloc (len2 + 11 + strlen (header));
1442 sprintf (res, "%s: Basic %s\r\n", header, t2);
1448 /* Parse HTTP `WWW-Authenticate:' header. AU points to the beginning
1449 of a field in such a header. If the field is the one specified by
1450 ATTR_NAME ("realm", "opaque", and "nonce" are used by the current
1451 digest authorization code), extract its value in the (char*)
1452 variable pointed by RET. Returns negative on a malformed header,
1453 or number of bytes that have been parsed by this call. */
1455 extract_header_attr (const char *au, const char *attr_name, char **ret)
1457 const char *cp, *ep;
1461 if (strncmp (cp, attr_name, strlen (attr_name)) == 0)
1463 cp += strlen (attr_name);
1466 cp += skip_lws (cp);
1471 cp += skip_lws (cp);
1476 for (ep = cp; *ep && *ep != '\"'; ep++)
1481 *ret = strdupdelim (cp, ep);
1488 /* Response value needs to be in lowercase, so we cannot use HEXD2ASC
1489 from url.h. See RFC 2069 2.1.2 for the syntax of response-digest. */
1490 #define HEXD2asc(x) (((x) < 10) ? ((x) + '0') : ((x) - 10 + 'a'))
1492 /* Dump the hexadecimal representation of HASH to BUF. HASH should be
1493 an array of 16 bytes containing the hash keys, and BUF should be a
1494 buffer of 33 writable characters (32 for hex digits plus one for
1495 zero termination). */
1497 dump_hash (unsigned char *buf, const unsigned char *hash)
1501 for (i = 0; i < MD5_HASHLEN; i++, hash++)
1503 *buf++ = HEXD2asc (*hash >> 4);
1504 *buf++ = HEXD2asc (*hash & 0xf);
1509 /* Take the line apart to find the challenge, and compose a digest
1510 authorization header. See RFC2069 section 2.1.2. */
1512 digest_authentication_encode (const char *au, const char *user,
1513 const char *passwd, const char *method,
1516 static char *realm, *opaque, *nonce;
1521 { "realm", &realm },
1522 { "opaque", &opaque },
1527 realm = opaque = nonce = NULL;
1529 au += 6; /* skip over `Digest' */
1534 au += skip_lws (au);
1535 for (i = 0; i < ARRAY_SIZE (options); i++)
1537 int skip = extract_header_attr (au, options[i].name,
1538 options[i].variable);
1542 FREE_MAYBE (opaque);
1552 if (i == ARRAY_SIZE (options))
1554 while (*au && *au != '=')
1558 au += skip_lws (au);
1562 while (*au && *au != '\"')
1569 while (*au && *au != ',')
1574 if (!realm || !nonce || !user || !passwd || !path || !method)
1577 FREE_MAYBE (opaque);
1582 /* Calculate the digest value. */
1585 unsigned char hash[MD5_HASHLEN];
1586 unsigned char a1buf[MD5_HASHLEN * 2 + 1], a2buf[MD5_HASHLEN * 2 + 1];
1587 unsigned char response_digest[MD5_HASHLEN * 2 + 1];
1589 /* A1BUF = H(user ":" realm ":" password) */
1590 md5_init_ctx (&ctx);
1591 md5_process_bytes (user, strlen (user), &ctx);
1592 md5_process_bytes (":", 1, &ctx);
1593 md5_process_bytes (realm, strlen (realm), &ctx);
1594 md5_process_bytes (":", 1, &ctx);
1595 md5_process_bytes (passwd, strlen (passwd), &ctx);
1596 md5_finish_ctx (&ctx, hash);
1597 dump_hash (a1buf, hash);
1599 /* A2BUF = H(method ":" path) */
1600 md5_init_ctx (&ctx);
1601 md5_process_bytes (method, strlen (method), &ctx);
1602 md5_process_bytes (":", 1, &ctx);
1603 md5_process_bytes (path, strlen (path), &ctx);
1604 md5_finish_ctx (&ctx, hash);
1605 dump_hash (a2buf, hash);
1607 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */
1608 md5_init_ctx (&ctx);
1609 md5_process_bytes (a1buf, MD5_HASHLEN * 2, &ctx);
1610 md5_process_bytes (":", 1, &ctx);
1611 md5_process_bytes (nonce, strlen (nonce), &ctx);
1612 md5_process_bytes (":", 1, &ctx);
1613 md5_process_bytes (a2buf, MD5_HASHLEN * 2, &ctx);
1614 md5_finish_ctx (&ctx, hash);
1615 dump_hash (response_digest, hash);
1617 res = (char*) xmalloc (strlen (user)
1622 + 2 * MD5_HASHLEN /*strlen (response_digest)*/
1623 + (opaque ? strlen (opaque) : 0)
1625 sprintf (res, "Authorization: Digest \
1626 username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
1627 user, realm, nonce, path, response_digest);
1630 char *p = res + strlen (res);
1631 strcat (p, ", opaque=\"");
1635 strcat (res, "\r\n");
1639 #endif /* USE_DIGEST */
1642 #define HACK_O_MATIC(line, string_constant) \
1643 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
1644 && (ISSPACE (line[sizeof (string_constant) - 1]) \
1645 || !line[sizeof (string_constant) - 1]))
1648 known_authentication_scheme_p (const char *au)
1650 return HACK_O_MATIC (au, "Basic") || HACK_O_MATIC (au, "Digest");
1655 /* Create the HTTP authorization request header. When the
1656 `WWW-Authenticate' response header is seen, according to the
1657 authorization scheme specified in that header (`Basic' and `Digest'
1658 are supported by the current implementation), produce an
1659 appropriate HTTP authorization request header. */
1661 create_authorization_line (const char *au, const char *user,
1662 const char *passwd, const char *method,
1665 char *wwwauth = NULL;
1667 if (!strncasecmp (au, "Basic", 5))
1668 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
1670 else if (!strncasecmp (au, "Digest", 6))
1671 wwwauth = digest_authentication_encode (au, user, passwd, method, path);
1672 #endif /* USE_DIGEST */