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>
35 #if TIME_WITH_SYS_TIME
36 # include <sys/time.h>
40 # include <sys/time.h>
49 # include <netdb.h> /* for h_errno */
66 # include "gen_sslfunc.h"
70 extern char *version_string;
81 static int cookies_loaded_p;
83 #define TEXTHTML_S "text/html"
84 #define HTTP_ACCEPT "*/*"
86 /* Some status code validation macros: */
87 #define H_20X(x) (((x) >= 200) && ((x) < 300))
88 #define H_PARTIAL(x) ((x) == HTTP_STATUS_PARTIAL_CONTENTS)
89 #define H_REDIRECTED(x) (((x) == HTTP_STATUS_MOVED_PERMANENTLY) \
90 || ((x) == HTTP_STATUS_MOVED_TEMPORARILY))
92 /* HTTP/1.0 status codes from RFC1945, provided for reference. */
94 #define HTTP_STATUS_OK 200
95 #define HTTP_STATUS_CREATED 201
96 #define HTTP_STATUS_ACCEPTED 202
97 #define HTTP_STATUS_NO_CONTENT 204
98 #define HTTP_STATUS_PARTIAL_CONTENTS 206
100 /* Redirection 3xx. */
101 #define HTTP_STATUS_MULTIPLE_CHOICES 300
102 #define HTTP_STATUS_MOVED_PERMANENTLY 301
103 #define HTTP_STATUS_MOVED_TEMPORARILY 302
104 #define HTTP_STATUS_NOT_MODIFIED 304
106 /* Client error 4xx. */
107 #define HTTP_STATUS_BAD_REQUEST 400
108 #define HTTP_STATUS_UNAUTHORIZED 401
109 #define HTTP_STATUS_FORBIDDEN 403
110 #define HTTP_STATUS_NOT_FOUND 404
112 /* Server errors 5xx. */
113 #define HTTP_STATUS_INTERNAL 500
114 #define HTTP_STATUS_NOT_IMPLEMENTED 501
115 #define HTTP_STATUS_BAD_GATEWAY 502
116 #define HTTP_STATUS_UNAVAILABLE 503
119 /* Parse the HTTP status line, which is of format:
121 HTTP-Version SP Status-Code SP Reason-Phrase
123 The function returns the status-code, or -1 if the status line is
124 malformed. The pointer to reason-phrase is returned in RP. */
126 parse_http_status_line (const char *line, const char **reason_phrase_ptr)
128 /* (the variables must not be named `major' and `minor', because
129 that breaks compilation with SunOS4 cc.) */
130 int mjr, mnr, statcode;
133 *reason_phrase_ptr = NULL;
135 /* The standard format of HTTP-Version is: `HTTP/X.Y', where X is
136 major version, and Y is minor version. */
137 if (strncmp (line, "HTTP/", 5) != 0)
141 /* Calculate major HTTP version. */
143 for (mjr = 0; ISDIGIT (*line); line++)
144 mjr = 10 * mjr + (*line - '0');
145 if (*line != '.' || p == line)
149 /* Calculate minor HTTP version. */
151 for (mnr = 0; ISDIGIT (*line); line++)
152 mnr = 10 * mnr + (*line - '0');
153 if (*line != ' ' || p == line)
155 /* Wget will accept only 1.0 and higher HTTP-versions. The value of
156 minor version can be safely ignored. */
161 /* Calculate status code. */
162 if (!(ISDIGIT (*line) && ISDIGIT (line[1]) && ISDIGIT (line[2])))
164 statcode = 100 * (*line - '0') + 10 * (line[1] - '0') + (line[2] - '0');
166 /* Set up the reason phrase pointer. */
168 /* RFC2068 requires SPC here, but we allow the string to finish
169 here, in case no reason-phrase is present. */
173 *reason_phrase_ptr = line;
178 *reason_phrase_ptr = line + 1;
183 /* Functions to be used as arguments to header_process(): */
185 struct http_process_range_closure {
191 /* Parse the `Content-Range' header and extract the information it
192 contains. Returns 1 if successful, -1 otherwise. */
194 http_process_range (const char *hdr, void *arg)
196 struct http_process_range_closure *closure
197 = (struct http_process_range_closure *)arg;
200 /* Certain versions of Nutscape proxy server send out
201 `Content-Length' without "bytes" specifier, which is a breach of
202 RFC2068 (as well as the HTTP/1.1 draft which was current at the
203 time). But hell, I must support it... */
204 if (!strncasecmp (hdr, "bytes", 5))
207 hdr += skip_lws (hdr);
213 for (num = 0; ISDIGIT (*hdr); hdr++)
214 num = 10 * num + (*hdr - '0');
215 if (*hdr != '-' || !ISDIGIT (*(hdr + 1)))
217 closure->first_byte_pos = num;
219 for (num = 0; ISDIGIT (*hdr); hdr++)
220 num = 10 * num + (*hdr - '0');
221 if (*hdr != '/' || !ISDIGIT (*(hdr + 1)))
223 closure->last_byte_pos = num;
225 for (num = 0; ISDIGIT (*hdr); hdr++)
226 num = 10 * num + (*hdr - '0');
227 closure->entity_length = num;
231 /* Place 1 to ARG if the HDR contains the word "none", 0 otherwise.
232 Used for `Accept-Ranges'. */
234 http_process_none (const char *hdr, void *arg)
236 int *where = (int *)arg;
238 if (strstr (hdr, "none"))
245 /* Place the malloc-ed copy of HDR hdr, to the first `;' to ARG. */
247 http_process_type (const char *hdr, void *arg)
249 char **result = (char **)arg;
250 /* Locate P on `;' or the terminating zero, whichever comes first. */
251 const char *p = strchr (hdr, ';');
253 p = hdr + strlen (hdr);
254 while (p > hdr && ISSPACE (*(p - 1)))
256 *result = strdupdelim (hdr, p);
260 /* Check whether the `Connection' header is set to "keep-alive". */
262 http_process_connection (const char *hdr, void *arg)
264 int *flag = (int *)arg;
265 if (!strcasecmp (hdr, "Keep-Alive"))
270 /* Persistent connections. Currently, we cache the most recently used
271 connection as persistent, provided that the HTTP server agrees to
272 make it such. The persistence data is stored in the variables
273 below. Ideally, it would be in a structure, and it should be
274 possible to cache an arbitrary fixed number of these connections.
276 I think the code is quite easy to extend in that direction. */
278 /* Whether a persistent connection is active. */
279 static int pc_active_p;
280 /* Host and port of currently active persistent connection. */
281 static unsigned char pc_last_host[4];
282 static unsigned short pc_last_port;
284 /* File descriptor of the currently active persistent connection. */
285 static int pc_last_fd;
288 /* Whether a ssl handshake has occoured on this connection */
289 static int pc_active_ssl;
290 /* SSL connection of the currently active persistent connection. */
291 static SSL *pc_last_ssl;
292 #endif /* HAVE_SSL */
294 /* Mark the persistent connection as invalid. This is used by the
295 CLOSE_* macros after they forcefully close a registered persistent
296 connection. This does not close the file descriptor -- it is left
297 to the caller to do that. (Maybe it should, though.) */
300 invalidate_persistent (void)
305 #endif /* HAVE_SSL */
306 DEBUGP (("Invalidating fd %d from further reuse.\n", pc_last_fd));
309 /* Register FD, which should be a TCP/IP connection to HOST:PORT, as
310 persistent. This will enable someone to use the same connection
311 later. In the context of HTTP, this must be called only AFTER the
312 response has been received and the server has promised that the
313 connection will remain alive.
315 If a previous connection was persistent, it is closed. */
318 register_persistent (const char *host, unsigned short port, int fd
328 if (pc_last_fd == fd)
330 /* The connection FD is already registered. Nothing to
336 /* The old persistent connection is still active; let's
337 close it first. This situation arises whenever a
338 persistent connection exists, but we then connect to a
339 different host, and try to register a persistent
340 connection to that one. */
342 /* The ssl disconnect has to take place before the closing
345 shutdown_ssl(pc_last_ssl);
348 invalidate_persistent ();
352 /* This store_hostaddress may not fail, because it has the results
354 success = store_hostaddress (pc_last_host, host);
361 pc_active_ssl = ssl ? 1 : 0;
363 DEBUGP (("Registered fd %d for persistent reuse.\n", fd));
366 /* Return non-zero if a persistent connection is available for
367 connecting to HOST:PORT. */
370 persistent_available_p (const char *host, unsigned short port
376 unsigned char this_host[4];
377 /* First, check whether a persistent connection is active at all. */
380 /* Second, check if the active connection pertains to the correct
381 (HOST, PORT) ordered pair. */
382 if (port != pc_last_port)
385 /* Second, a): check if current connection is (not) ssl, too. This
386 test is unlikely to fail because HTTP and HTTPS typicaly use
387 different ports. Yet it is possible, or so I [Christian
388 Fraenkel] have been told, to run HTTPS and HTTP simultaneus on
390 if (ssl != pc_active_ssl)
392 #endif /* HAVE_SSL */
393 if (!store_hostaddress (this_host, host))
395 if (memcmp (pc_last_host, this_host, 4))
397 /* Third: check whether the connection is still open. This is
398 important because most server implement a liberal (short) timeout
399 on persistent connections. Wget can of course always reconnect
400 if the connection doesn't work out, but it's nicer to know in
401 advance. This test is a logical followup of the first test, but
402 is "expensive" and therefore placed at the end of the list. */
403 if (!test_socket_open (pc_last_fd))
405 /* Oops, the socket is no longer open. Now that we know that,
406 let's invalidate the persistent connection before returning
409 invalidate_persistent ();
416 # define SHUTDOWN_SSL(ssl) do { \
418 shutdown_ssl (ssl); \
421 # define SHUTDOWN_SSL(ssl)
424 /* The idea behind these two CLOSE macros is to distinguish between
425 two cases: one when the job we've been doing is finished, and we
426 want to close the connection and leave, and two when something is
427 seriously wrong and we're closing the connection as part of
430 In case of keep_alive, CLOSE_FINISH should leave the connection
431 open, while CLOSE_INVALIDATE should still close it.
433 Note that the semantics of the flag `keep_alive' is "this
434 connection *will* be reused (the server has promised not to close
435 the connection once we're done)", while the semantics of
436 `pc_active_p && (fd) == pc_last_fd' is "we're *now* using an
437 active, registered connection". */
439 #define CLOSE_FINISH(fd) do { \
442 SHUTDOWN_SSL (ssl); \
444 if (pc_active_p && (fd) == pc_last_fd) \
445 invalidate_persistent (); \
449 #define CLOSE_INVALIDATE(fd) do { \
450 SHUTDOWN_SSL (ssl); \
452 if (pc_active_p && (fd) == pc_last_fd) \
453 invalidate_persistent (); \
458 long len; /* received length */
459 long contlen; /* expected length */
460 long restval; /* the restart value */
461 int res; /* the result of last read */
462 char *newloc; /* new location (redirection) */
463 char *remote_time; /* remote time-stamp string */
464 char *error; /* textual HTTP error */
465 int statcode; /* status code */
466 long dltime; /* time of the download */
467 int no_truncate; /* whether truncating the file is
471 /* Free the elements of hstat X. */
472 #define FREEHSTAT(x) do \
474 FREE_MAYBE ((x).newloc); \
475 FREE_MAYBE ((x).remote_time); \
476 FREE_MAYBE ((x).error); \
477 (x).newloc = (x).remote_time = (x).error = NULL; \
480 static char *create_authorization_line PARAMS ((const char *, const char *,
481 const char *, const char *,
483 static char *basic_authentication_encode PARAMS ((const char *, const char *,
485 static int known_authentication_scheme_p PARAMS ((const char *));
487 time_t http_atotm PARAMS ((char *));
489 #define BEGINS_WITH(line, string_constant) \
490 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
491 && (ISSPACE (line[sizeof (string_constant) - 1]) \
492 || !line[sizeof (string_constant) - 1]))
494 /* Retrieve a document through HTTP protocol. It recognizes status
495 code, and correctly handles redirections. It closes the network
496 socket. If it receives an error from the functions below it, it
497 will print it if there is enough information to do so (almost
498 always), returning the error to the caller (i.e. http_loop).
500 Various HTTP parameters are stored to hs. Although it parses the
501 response code correctly, it is not used in a sane way. The caller
504 If u->proxy is non-NULL, the URL u will be taken as a proxy URL,
505 and u->proxy->url will be given to the proxy server (bad naming,
508 gethttp (struct urlinfo *u, struct http_stat *hs, int *dt)
510 char *request, *type, *command, *path;
512 char *pragma_h, *referer, *useragent, *range, *wwwauth, *remhost;
513 char *authenticate_h;
517 char *request_keep_alive;
518 int sock, hcount, num_written, all_length, remport, statcode;
519 long contlen, contrange;
523 int auth_tried_already;
526 static SSL_CTX *ssl_ctx = NULL;
528 #endif /* HAVE_SSL */
529 struct wget_timer *timer;
530 char *cookies = NULL;
532 /* Whether this connection will be kept alive after the HTTP request
536 /* Flags that detect the two ways of specifying HTTP keep-alive
538 int http_keep_alive_1, http_keep_alive_2;
540 /* Whether keep-alive should be inhibited. */
541 int inhibit_keep_alive;
544 /* initialize ssl_ctx on first run */
547 err=init_ssl (&ssl_ctx);
552 case SSLERRCTXCREATE:
554 logprintf (LOG_NOTQUIET, _("Failed to set up an SSL context\n"));
558 /* try without certfile */
559 logprintf (LOG_NOTQUIET,
560 _("Failed to load certificates from %s\n"),
563 logprintf (LOG_NOTQUIET,
564 _("Trying without the specified certificate\n"));
567 logprintf (LOG_NOTQUIET,
568 _("Failed to get certificate key from %s\n"),
571 logprintf (LOG_NOTQUIET,
572 _("Trying without the specified certificate\n"));
579 #endif /* HAVE_SSL */
581 if (!(*dt & HEAD_ONLY))
582 /* If we're doing a GET on the URL, as opposed to just a HEAD, we need to
583 know the local filename so we can save to it. */
584 assert (u->local != NULL);
587 auth_tried_already = 0;
589 inhibit_keep_alive = (!opt.http_keep_alive || u->proxy != NULL);
592 /* We need to come back here when the initial attempt to retrieve
593 without authorization header fails. (Expected to happen at least
594 for the Digest authorization scheme.) */
597 http_keep_alive_1 = http_keep_alive_2 = 0;
600 cookies = build_cookies_request (u->host, u->port, u->path,
601 u->proto == URLHTTPS);
603 /* Initialize certain elements of struct http_stat. */
608 hs->remote_time = NULL;
611 /* Which structure to use to retrieve the original URL data. */
617 /* First: establish the connection. */
618 if (inhibit_keep_alive
621 !persistent_available_p (u->host, u->port)
623 !persistent_available_p (u->host, u->port, (u->proto==URLHTTPS ? 1 : 0))
624 #endif /* HAVE_SSL */
627 logprintf (LOG_VERBOSE, _("Connecting to %s:%hu... "), u->host, u->port);
628 err = make_connection (&sock, u->host, u->port);
632 logputs (LOG_VERBOSE, "\n");
633 logprintf (LOG_NOTQUIET, "%s: %s.\n", u->host, herrmsg (h_errno));
637 logputs (LOG_VERBOSE, "\n");
638 logprintf (LOG_NOTQUIET, "socket: %s\n", strerror (errno));
642 logputs (LOG_VERBOSE, "\n");
643 logprintf (LOG_NOTQUIET,
644 _("Connection to %s:%hu refused.\n"), u->host, u->port);
648 logputs (LOG_VERBOSE, "\n");
649 logprintf (LOG_NOTQUIET, "connect: %s\n", strerror (errno));
654 /* Everything is fine! */
655 logputs (LOG_VERBOSE, _("connected!\n"));
662 if (u->proto == URLHTTPS)
663 if (connect_ssl (&ssl, ssl_ctx,sock) != 0)
665 logputs (LOG_VERBOSE, "\n");
666 logprintf (LOG_NOTQUIET, _("Unable to establish SSL connection.\n"));
670 #endif /* HAVE_SSL */
674 logprintf (LOG_VERBOSE, _("Reusing connection to %s:%hu.\n"), u->host, u->port);
675 /* #### pc_last_fd should be accessed through an accessor
680 #endif /* HAVE_SSL */
681 DEBUGP (("Reusing fd %d.\n", sock));
685 path = u->proxy->url;
689 command = (*dt & HEAD_ONLY) ? "HEAD" : "GET";
693 referer = (char *)alloca (9 + strlen (ou->referer) + 3);
694 sprintf (referer, "Referer: %s\r\n", ou->referer);
696 if (*dt & SEND_NOCACHE)
697 pragma_h = "Pragma: no-cache\r\n";
702 range = (char *)alloca (13 + numdigit (hs->restval) + 4);
703 /* Gag me! Some servers (e.g. WebSitePro) have been known to
704 respond to the following `Range' format by generating a
705 multipart/x-byte-ranges MIME document! This MIME type was
706 present in an old draft of the byteranges specification.
707 HTTP/1.1 specifies a multipart/byte-ranges MIME type, but
708 only if multiple non-overlapping ranges are requested --
709 which Wget never does. */
710 sprintf (range, "Range: bytes=%ld-\r\n", hs->restval);
715 STRDUP_ALLOCA (useragent, opt.useragent);
718 useragent = (char *)alloca (10 + strlen (version_string));
719 sprintf (useragent, "Wget/%s", version_string);
721 /* Construct the authentication, if userid is present. */
724 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
725 user = user ? user : opt.http_user;
726 passwd = passwd ? passwd : opt.http_passwd;
733 /* We have the username and the password, but haven't tried
734 any authorization yet. Let's see if the "Basic" method
735 works. If not, we'll come back here and construct a
736 proper authorization method with the right challenges.
738 If we didn't employ this kind of logic, every URL that
739 requires authorization would have to be processed twice,
740 which is very suboptimal and generates a bunch of false
741 "unauthorized" errors in the server log.
743 #### But this logic also has a serious problem when used
744 with stronger authentications: we *first* transmit the
745 username and the password in clear text, and *then*
746 attempt a stronger authentication scheme. That cannot be
747 right! We are only fortunate that almost everyone still
748 uses the `Basic' scheme anyway.
750 There should be an option to prevent this from happening,
751 for those who use strong authentication schemes and value
753 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
757 wwwauth = create_authorization_line (authenticate_h, user, passwd,
765 char *proxy_user, *proxy_passwd;
766 /* For normal username and password, URL components override
767 command-line/wgetrc parameters. With proxy authentication,
768 it's the reverse, because proxy URLs are normally the
769 "permanent" ones, so command-line args should take
771 if (opt.proxy_user && opt.proxy_passwd)
773 proxy_user = opt.proxy_user;
774 proxy_passwd = opt.proxy_passwd;
778 proxy_user = u->user;
779 proxy_passwd = u->passwd;
781 /* #### This is junky. Can't the proxy request, say, `Digest'
783 if (proxy_user && proxy_passwd)
784 proxyauth = basic_authentication_encode (proxy_user, proxy_passwd,
785 "Proxy-Authorization");
790 /* String of the form :PORT. Used only for non-standard ports. */
794 && remport != (u->proto == URLHTTPS
795 ? DEFAULT_HTTPS_PORT : DEFAULT_HTTP_PORT)
797 && remport != DEFAULT_HTTP_PORT
801 port_maybe = (char *)alloca (numdigit (remport) + 2);
802 sprintf (port_maybe, ":%d", remport);
805 if (!inhibit_keep_alive)
806 request_keep_alive = "Connection: Keep-Alive\r\n";
808 request_keep_alive = NULL;
810 /* Allocate the memory for the request. */
811 request = (char *)alloca (strlen (command) + strlen (path)
814 + (port_maybe ? strlen (port_maybe) : 0)
815 + strlen (HTTP_ACCEPT)
816 + (request_keep_alive
817 ? strlen (request_keep_alive) : 0)
818 + (referer ? strlen (referer) : 0)
819 + (cookies ? strlen (cookies) : 0)
820 + (wwwauth ? strlen (wwwauth) : 0)
821 + (proxyauth ? strlen (proxyauth) : 0)
822 + (range ? strlen (range) : 0)
824 + (opt.user_header ? strlen (opt.user_header) : 0)
826 /* Construct the request. */
832 %s%s%s%s%s%s%s%s\r\n",
833 command, path, useragent, remhost,
834 port_maybe ? port_maybe : "",
836 request_keep_alive ? request_keep_alive : "",
837 referer ? referer : "",
838 cookies ? cookies : "",
839 wwwauth ? wwwauth : "",
840 proxyauth ? proxyauth : "",
843 opt.user_header ? opt.user_header : "");
844 DEBUGP (("---request begin---\n%s---request end---\n", request));
845 /* Free the temporary memory. */
846 FREE_MAYBE (wwwauth);
847 FREE_MAYBE (proxyauth);
848 FREE_MAYBE (cookies);
850 /* Send the request to server. */
852 if (u->proto == URLHTTPS)
853 num_written = ssl_iwrite (ssl, request, strlen (request));
855 #endif /* HAVE_SSL */
856 num_written = iwrite (sock, request, strlen (request));
860 logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"),
862 CLOSE_INVALIDATE (sock);
865 logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
866 u->proxy ? "Proxy" : "HTTP");
867 contlen = contrange = -1;
872 /* Before reading anything, initialize the rbuf. */
873 rbuf_initialize (&rbuf, sock);
875 if (u->proto == URLHTTPS)
879 #endif /* HAVE_SSL */
882 /* Header-fetching loop. */
890 /* Get the header. */
891 status = header_get (&rbuf, &hdr,
892 /* Disallow continuations for status line. */
893 (hcount == 1 ? HG_NO_CONTINUATIONS : HG_NONE));
895 /* Check for errors. */
896 if (status == HG_EOF && *hdr)
898 /* This used to be an unconditional error, but that was
899 somewhat controversial, because of a large number of
900 broken CGI's that happily "forget" to send the second EOL
901 before closing the connection of a HEAD request.
903 So, the deal is to check whether the header is empty
904 (*hdr is zero if it is); if yes, it means that the
905 previous header was fully retrieved, and that -- most
906 probably -- the request is complete. "...be liberal in
907 what you accept." Oh boy. */
908 logputs (LOG_VERBOSE, "\n");
909 logputs (LOG_NOTQUIET, _("End of file while parsing headers.\n"));
912 FREE_MAYBE (hs->newloc);
913 FREE_MAYBE (all_headers);
914 CLOSE_INVALIDATE (sock);
917 else if (status == HG_ERROR)
919 logputs (LOG_VERBOSE, "\n");
920 logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
924 FREE_MAYBE (hs->newloc);
925 FREE_MAYBE (all_headers);
926 CLOSE_INVALIDATE (sock);
930 /* If the headers are to be saved to a file later, save them to
932 if (opt.save_headers)
934 int lh = strlen (hdr);
935 all_headers = (char *)xrealloc (all_headers, all_length + lh + 2);
936 memcpy (all_headers + all_length, hdr, lh);
938 all_headers[all_length++] = '\n';
939 all_headers[all_length] = '\0';
942 /* Print the header if requested. */
943 if (opt.server_response && hcount != 1)
944 logprintf (LOG_VERBOSE, "\n%d %s", hcount, hdr);
946 /* Check for status line. */
950 /* Parse the first line of server response. */
951 statcode = parse_http_status_line (hdr, &error);
952 hs->statcode = statcode;
953 /* Store the descriptive response. */
954 if (statcode == -1) /* malformed response */
956 /* A common reason for "malformed response" error is the
957 case when no data was actually received. Handle this
960 hs->error = xstrdup (_("No data received"));
962 hs->error = xstrdup (_("Malformed status line"));
967 hs->error = xstrdup (_("(no description)"));
969 hs->error = xstrdup (error);
976 logprintf (LOG_VERBOSE, "%d %s", statcode, error);
981 /* Exit on empty header. */
988 /* Try getting content-length. */
989 if (contlen == -1 && !opt.ignore_length)
990 if (header_process (hdr, "Content-Length", header_extract_number,
993 /* Try getting content-type. */
995 if (header_process (hdr, "Content-Type", http_process_type, &type))
997 /* Try getting location. */
999 if (header_process (hdr, "Location", header_strdup, &hs->newloc))
1001 /* Try getting last-modified. */
1002 if (!hs->remote_time)
1003 if (header_process (hdr, "Last-Modified", header_strdup,
1006 /* Try getting cookies. */
1008 if (header_process (hdr, "Set-Cookie", set_cookie_header_cb, u))
1010 /* Try getting www-authentication. */
1011 if (!authenticate_h)
1012 if (header_process (hdr, "WWW-Authenticate", header_strdup,
1015 /* Check for accept-ranges header. If it contains the word
1016 `none', disable the ranges. */
1017 if (*dt & ACCEPTRANGES)
1020 if (header_process (hdr, "Accept-Ranges", http_process_none, &nonep))
1023 *dt &= ~ACCEPTRANGES;
1027 /* Try getting content-range. */
1028 if (contrange == -1)
1030 struct http_process_range_closure closure;
1031 if (header_process (hdr, "Content-Range", http_process_range, &closure))
1033 contrange = closure.first_byte_pos;
1037 /* Check for keep-alive related responses. */
1038 if (!inhibit_keep_alive)
1040 /* Check for the `Keep-Alive' header. */
1041 if (!http_keep_alive_1)
1043 if (header_process (hdr, "Keep-Alive", header_exists,
1044 &http_keep_alive_1))
1047 /* Check for `Connection: Keep-Alive'. */
1048 if (!http_keep_alive_2)
1050 if (header_process (hdr, "Connection", http_process_connection,
1051 &http_keep_alive_2))
1059 logputs (LOG_VERBOSE, "\n");
1062 && (http_keep_alive_1 || http_keep_alive_2))
1064 assert (inhibit_keep_alive == 0);
1068 /* The server has promised that it will not close the connection
1069 when we're done. This means that we can register it. */
1071 register_persistent (u->host, u->port, sock);
1073 register_persistent (u->host, u->port, sock, ssl);
1074 #endif /* HAVE_SSL */
1076 if ((statcode == HTTP_STATUS_UNAUTHORIZED)
1079 /* Authorization is required. */
1083 CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
1084 might be more bytes in the body. */
1085 if (auth_tried_already)
1087 /* If we have tried it already, then there is not point
1090 logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
1091 xfree (authenticate_h);
1094 else if (!known_authentication_scheme_p (authenticate_h))
1096 xfree (authenticate_h);
1097 logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
1100 else if (BEGINS_WITH (authenticate_h, "Basic"))
1102 /* The authentication scheme is basic, the one we try by
1103 default, and it failed. There's no sense in trying
1109 auth_tried_already = 1;
1113 /* We do not need this anymore. */
1116 xfree (authenticate_h);
1117 authenticate_h = NULL;
1120 /* 20x responses are counted among successful by default. */
1121 if (H_20X (statcode))
1124 if (type && !strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S)))
1127 /* We don't assume text/html by default. */
1130 if (opt.html_extension && (*dt & TEXTHTML))
1131 /* -E / --html-extension / html_extension = on was specified, and this is a
1132 text/html file. If some case-insensitive variation on ".htm[l]" isn't
1133 already the file's suffix, tack on ".html". */
1135 char* last_period_in_local_filename = strrchr(u->local, '.');
1137 if (last_period_in_local_filename == NULL ||
1138 !(strcasecmp(last_period_in_local_filename, ".htm") == EQ ||
1139 strcasecmp(last_period_in_local_filename, ".html") == EQ))
1141 size_t local_filename_len = strlen(u->local);
1143 u->local = xrealloc(u->local, local_filename_len + sizeof(".html"));
1144 strcpy(u->local + local_filename_len, ".html");
1146 *dt |= ADDED_HTML_EXTENSION;
1150 if (contrange == -1)
1152 /* We did not get a content-range header. This means that the
1153 server did not honor our `Range' request. Normally, this
1154 means we should reset hs->restval and continue normally. */
1156 /* However, if `-c' is used, we need to be a bit more careful:
1158 1. If `-c' is specified and the file already existed when
1159 Wget was started, it would be a bad idea for us to start
1160 downloading it from scratch, effectively truncating it. I
1161 believe this cannot happen unless `-c' was specified.
1163 2. If `-c' is used on a file that is already fully
1164 downloaded, we're requesting bytes after the end of file,
1165 which can result in server not honoring `Range'. If this is
1166 the case, `Content-Length' will be equal to the length of the
1168 if (opt.always_rest)
1170 /* Check for condition #2. */
1171 if (hs->restval == contlen)
1173 logputs (LOG_VERBOSE, _("\
1174 \n The file is already fully retrieved; nothing to do.\n\n"));
1175 /* In case the caller inspects. */
1179 FREE_MAYBE (hs->newloc);
1180 FREE_MAYBE (all_headers);
1181 CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
1182 might be more bytes in the body. */
1183 return RETRFINISHED;
1186 /* Check for condition #1. */
1187 if (hs->no_truncate)
1189 logprintf (LOG_NOTQUIET,
1192 The server does not support continued download;\n\
1193 refusing to truncate `%s'.\n\n"), u->local);
1194 return CONTNOTSUPPORTED;
1203 else if (contrange != hs->restval ||
1204 (H_PARTIAL (statcode) && contrange == -1))
1206 /* This means the whole request was somehow misunderstood by the
1207 server. Bail out. */
1209 FREE_MAYBE (hs->newloc);
1210 FREE_MAYBE (all_headers);
1211 CLOSE_INVALIDATE (sock);
1218 contlen += contrange;
1220 contrange = -1; /* If conent-length was not sent,
1221 content-range will be ignored. */
1223 hs->contlen = contlen;
1225 /* Return if redirected. */
1226 if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES)
1228 /* RFC2068 says that in case of the 300 (multiple choices)
1229 response, the server can output a preferred URL through
1230 `Location' header; otherwise, the request should be treated
1231 like GET. So, if the location is set, it will be a
1232 redirection; otherwise, just proceed normally. */
1233 if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc)
1237 logprintf (LOG_VERBOSE,
1238 _("Location: %s%s\n"),
1239 hs->newloc ? hs->newloc : _("unspecified"),
1240 hs->newloc ? _(" [following]") : "");
1241 CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
1242 might be more bytes in the body. */
1244 FREE_MAYBE (all_headers);
1250 if ((*dt & RETROKF) && !opt.server_response)
1252 /* No need to print this output if the body won't be
1253 downloaded at all, or if the original server response is
1255 logputs (LOG_VERBOSE, _("Length: "));
1258 logputs (LOG_VERBOSE, legible (contlen));
1259 if (contrange != -1)
1260 logprintf (LOG_VERBOSE, _(" (%s to go)"),
1261 legible (contlen - contrange));
1264 logputs (LOG_VERBOSE,
1265 opt.ignore_length ? _("ignored") : _("unspecified"));
1267 logprintf (LOG_VERBOSE, " [%s]\n", type);
1269 logputs (LOG_VERBOSE, "\n");
1273 type = NULL; /* We don't need it any more. */
1275 /* Return if we have no intention of further downloading. */
1276 if (!(*dt & RETROKF) || (*dt & HEAD_ONLY))
1278 /* In case the caller cares to look... */
1282 FREE_MAYBE (all_headers);
1283 CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
1284 might be more bytes in the body. */
1285 return RETRFINISHED;
1288 /* Open the local file. */
1291 mkalldirs (u->local);
1293 rotate_backups (u->local);
1294 fp = fopen (u->local, hs->restval ? "ab" : "wb");
1297 logprintf (LOG_NOTQUIET, "%s: %s\n", u->local, strerror (errno));
1298 CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
1299 might be more bytes in the body. */
1300 FREE_MAYBE (all_headers);
1306 extern int global_download_count;
1308 /* To ensure that repeated "from scratch" downloads work for -O
1309 files, we rewind the file pointer, unless restval is
1310 non-zero. (This works only when -O is used on regular files,
1311 but it's still a valuable feature.)
1313 However, this loses when more than one URL is specified on
1314 the command line the second rewinds eradicates the contents
1315 of the first download. Thus we disable the above trick for
1316 all the downloads except the very first one.
1318 #### A possible solution to this would be to remember the
1319 file position in the output document and to seek to that
1320 position, instead of rewinding. */
1321 if (!hs->restval && global_download_count == 0)
1323 /* This will silently fail for streams that don't correspond
1324 to regular files, but that's OK. */
1326 /* ftruncate is needed because opt.dfp is opened in append
1327 mode if opt.always_rest is set. */
1328 ftruncate (fileno (fp), 0);
1333 /* #### This confuses the code that checks for file size. There
1334 should be some overhead information. */
1335 if (opt.save_headers)
1336 fwrite (all_headers, 1, all_length, fp);
1337 timer = wtimer_new ();
1338 /* Get the contents of the document. */
1339 hs->res = get_contents (sock, fp, &hs->len, hs->restval,
1340 (contlen != -1 ? contlen : 0),
1342 hs->dltime = wtimer_elapsed (timer);
1343 wtimer_delete (timer);
1345 /* Close or flush the file. We have to be careful to check for
1346 error here. Checking the result of fwrite() is not enough --
1347 errors could go unnoticed! */
1350 flush_res = fclose (fp);
1352 flush_res = fflush (fp);
1353 if (flush_res == EOF)
1356 FREE_MAYBE (all_headers);
1357 CLOSE_FINISH (sock);
1360 return RETRFINISHED;
1363 /* The genuine HTTP loop! This is the part where the retrieval is
1364 retried, and retried, and retried, and... */
1366 http_loop (struct urlinfo *u, char **newloc, int *dt)
1369 int use_ts, got_head = 0; /* time-stamping info */
1370 char *filename_plus_orig_suffix;
1371 char *local_filename = NULL;
1372 char *tms, *suf, *locf, *tmrate;
1374 time_t tml = -1, tmr = -1; /* local and remote time-stamps */
1375 long local_size = 0; /* the size of the local file */
1376 size_t filename_len;
1377 struct http_stat hstat; /* HTTP status */
1380 /* This used to be done in main(), but it's a better idea to do it
1381 here so that we don't go through the hoops if we're just using
1383 if (opt.cookies && opt.cookies_input && !cookies_loaded_p)
1384 load_cookies (opt.cookies_input);
1388 /* Warn on (likely bogus) wildcard usage in HTTP. Don't use
1389 has_wildcards_p because it would also warn on `?', and we know that
1390 shows up in CGI paths a *lot*. */
1391 if (strchr (u->url, '*'))
1392 logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
1394 /* Determine the local filename. */
1396 u->local = url_filename (u->proxy ? u->proxy : u);
1398 if (!opt.output_document)
1401 locf = opt.output_document;
1403 /* Yuck. Multiple returns suck. We need to remember to free() the space we
1404 xmalloc() here before EACH return. This is one reason it's better to set
1405 flags that influence flow control and then return once at the end. */
1406 filename_len = strlen(u->local);
1407 filename_plus_orig_suffix = xmalloc(filename_len + sizeof(".orig"));
1409 if (opt.noclobber && file_exists_p (u->local))
1411 /* If opt.noclobber is turned on and file already exists, do not
1412 retrieve the file */
1413 logprintf (LOG_VERBOSE, _("\
1414 File `%s' already there, will not retrieve.\n"), u->local);
1415 /* If the file is there, we suppose it's retrieved OK. */
1418 /* #### Bogusness alert. */
1419 /* If its suffix is "html" or (yuck!) "htm", we suppose it's
1420 text/html, a harmless lie. */
1421 if (((suf = suffix (u->local)) != NULL)
1422 && (!strcmp (suf, "html") || !strcmp (suf, "htm")))
1425 xfree (filename_plus_orig_suffix); /* must precede every return! */
1426 /* Another harmless lie: */
1431 if (opt.timestamping)
1433 boolean local_dot_orig_file_exists = FALSE;
1435 if (opt.backup_converted)
1436 /* If -K is specified, we'll act on the assumption that it was specified
1437 last time these files were downloaded as well, and instead of just
1438 comparing local file X against server file X, we'll compare local
1439 file X.orig (if extant, else X) against server file X. If -K
1440 _wasn't_ specified last time, or the server contains files called
1441 *.orig, -N will be back to not operating correctly with -k. */
1443 /* Would a single s[n]printf() call be faster? --dan
1445 Definitely not. sprintf() is horribly slow. It's a
1446 different question whether the difference between the two
1447 affects a program. Usually I'd say "no", but at one
1448 point I profiled Wget, and found that a measurable and
1449 non-negligible amount of time was lost calling sprintf()
1450 in url.c. Replacing sprintf with inline calls to
1451 strcpy() and long_to_string() made a difference.
1453 strcpy(filename_plus_orig_suffix, u->local);
1454 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
1456 /* Try to stat() the .orig file. */
1457 if (stat(filename_plus_orig_suffix, &st) == 0)
1459 local_dot_orig_file_exists = TRUE;
1460 local_filename = filename_plus_orig_suffix;
1464 if (!local_dot_orig_file_exists)
1465 /* Couldn't stat() <file>.orig, so try to stat() <file>. */
1466 if (stat (u->local, &st) == 0)
1467 local_filename = u->local;
1469 if (local_filename != NULL)
1470 /* There was a local file, so we'll check later to see if the version
1471 the server has is the same version we already have, allowing us to
1476 local_size = st.st_size;
1480 /* Reset the counter. */
1482 *dt = 0 | ACCEPTRANGES;
1486 /* Increment the pass counter. */
1488 sleep_between_retrievals (count);
1489 /* Get the current time string. */
1490 tms = time_str (NULL);
1491 /* Print fetch message, if opt.verbose. */
1494 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
1498 sprintf (tmp, _("(try:%2d)"), count);
1499 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
1500 tms, hurl, tmp, locf);
1502 ws_changetitle (hurl, 1);
1507 /* Default document type is empty. However, if spider mode is
1508 on or time-stamping is employed, HEAD_ONLY commands is
1509 encoded within *dt. */
1510 if (opt.spider || (use_ts && !got_head))
1514 /* Assume no restarting. */
1516 /* Decide whether or not to restart. */
1517 if (((count > 1 && (*dt & ACCEPTRANGES)) || opt.always_rest)
1518 && file_exists_p (locf))
1519 if (stat (locf, &st) == 0 && S_ISREG (st.st_mode))
1520 hstat.restval = st.st_size;
1521 /* Decide whether to send the no-cache directive. */
1522 if (u->proxy && (count > 1 || (opt.proxy_cache == 0)))
1523 *dt |= SEND_NOCACHE;
1525 *dt &= ~SEND_NOCACHE;
1527 /* Try fetching the document, or at least its head. :-) */
1528 err = gethttp (u, &hstat, dt);
1530 /* It's unfortunate that wget determines the local filename before finding
1531 out the Content-Type of the file. Barring a major restructuring of the
1532 code, we need to re-set locf here, since gethttp() may have xrealloc()d
1533 u->local to tack on ".html". */
1534 if (!opt.output_document)
1537 locf = opt.output_document;
1539 /* In `-c' is used, check whether the file we're writing to
1540 exists before we've done anything. If so, we'll refuse to
1541 truncate it if the server doesn't support continued
1543 if (opt.always_rest)
1544 hstat.no_truncate = file_exists_p (locf);
1547 tms = time_str (NULL);
1548 /* Get the new location (with or without the redirection). */
1550 *newloc = xstrdup (hstat.newloc);
1553 case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
1554 case CONERROR: case READERR: case WRITEFAILED:
1556 /* Non-fatal errors continue executing the loop, which will
1557 bring them to "while" statement at the end, to judge
1558 whether the number of tries was exceeded. */
1560 printwhat (count, opt.ntry);
1563 case HOSTERR: case CONREFUSED: case PROXERR: case AUTHFAILED:
1564 case SSLERRCTXCREATE: case CONTNOTSUPPORTED:
1565 /* Fatal errors just return from the function. */
1567 xfree (filename_plus_orig_suffix); /* must precede every return! */
1570 case FWRITEERR: case FOPENERR:
1571 /* Another fatal error. */
1572 logputs (LOG_VERBOSE, "\n");
1573 logprintf (LOG_NOTQUIET, _("Cannot write to `%s' (%s).\n"),
1574 u->local, strerror (errno));
1579 /* Another fatal error. */
1580 logputs (LOG_VERBOSE, "\n");
1581 logprintf (LOG_NOTQUIET, _("Unable to establish SSL connection.\n"));
1583 xfree (filename_plus_orig_suffix); /* must precede every return! */
1587 /* Return the new location to the caller. */
1590 logprintf (LOG_NOTQUIET,
1591 _("ERROR: Redirection (%d) without location.\n"),
1593 xfree (filename_plus_orig_suffix); /* must precede every return! */
1597 xfree (filename_plus_orig_suffix); /* must precede every return! */
1601 /* Deal with you later. */
1604 /* All possibilities should have been exhausted. */
1607 if (!(*dt & RETROKF))
1611 /* #### Ugly ugly ugly! */
1612 char *hurl = str_url (u->proxy ? u->proxy : u, 1);
1613 logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
1616 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
1617 tms, hstat.statcode, hstat.error);
1618 logputs (LOG_VERBOSE, "\n");
1620 xfree (filename_plus_orig_suffix); /* must precede every return! */
1624 /* Did we get the time-stamp? */
1627 if (opt.timestamping && !hstat.remote_time)
1629 logputs (LOG_NOTQUIET, _("\
1630 Last-modified header missing -- time-stamps turned off.\n"));
1632 else if (hstat.remote_time)
1634 /* Convert the date-string into struct tm. */
1635 tmr = http_atotm (hstat.remote_time);
1636 if (tmr == (time_t) (-1))
1637 logputs (LOG_VERBOSE, _("\
1638 Last-modified header invalid -- time-stamp ignored.\n"));
1642 /* The time-stamping section. */
1647 use_ts = 0; /* no more time-stamping */
1648 count = 0; /* the retrieve count for HEAD is
1650 if (hstat.remote_time && tmr != (time_t) (-1))
1652 /* Now time-stamping can be used validly. Time-stamping
1653 means that if the sizes of the local and remote file
1654 match, and local file is newer than the remote file,
1655 it will not be retrieved. Otherwise, the normal
1656 download procedure is resumed. */
1658 (hstat.contlen == -1 || local_size == hstat.contlen))
1660 logprintf (LOG_VERBOSE, _("\
1661 Server file no newer than local file `%s' -- not retrieving.\n\n"),
1664 xfree (filename_plus_orig_suffix); /*must precede every return!*/
1667 else if (tml >= tmr)
1668 logprintf (LOG_VERBOSE, _("\
1669 The sizes do not match (local %ld) -- retrieving.\n"), local_size);
1671 logputs (LOG_VERBOSE,
1672 _("Remote file is newer, retrieving.\n"));
1677 if ((tmr != (time_t) (-1))
1679 && ((hstat.len == hstat.contlen) ||
1680 ((hstat.res == 0) &&
1681 ((hstat.contlen == -1) ||
1682 (hstat.len >= hstat.contlen && !opt.kill_longer)))))
1684 /* #### This code repeats in http.c and ftp.c. Move it to a
1686 const char *fl = NULL;
1687 if (opt.output_document)
1689 if (opt.od_known_regular)
1690 fl = opt.output_document;
1697 /* End of time-stamping section. */
1701 logprintf (LOG_NOTQUIET, "%d %s\n\n", hstat.statcode, hstat.error);
1702 xfree (filename_plus_orig_suffix); /* must precede every return! */
1706 /* It is now safe to free the remainder of hstat, since the
1707 strings within it will no longer be used. */
1710 tmrate = rate (hstat.len - hstat.restval, hstat.dltime, 0);
1712 if (hstat.len == hstat.contlen)
1716 logprintf (LOG_VERBOSE,
1717 _("%s (%s) - `%s' saved [%ld/%ld]\n\n"),
1718 tms, tmrate, locf, hstat.len, hstat.contlen);
1719 logprintf (LOG_NONVERBOSE,
1720 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1721 tms, u->url, hstat.len, hstat.contlen, locf, count);
1724 downloaded_increase (hstat.len);
1726 /* Remember that we downloaded the file for later ".orig" code. */
1727 if (*dt & ADDED_HTML_EXTENSION)
1728 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1730 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1732 xfree(filename_plus_orig_suffix); /* must precede every return! */
1735 else if (hstat.res == 0) /* No read error */
1737 if (hstat.contlen == -1) /* We don't know how much we were supposed
1738 to get, so assume we succeeded. */
1742 logprintf (LOG_VERBOSE,
1743 _("%s (%s) - `%s' saved [%ld]\n\n"),
1744 tms, tmrate, locf, hstat.len);
1745 logprintf (LOG_NONVERBOSE,
1746 "%s URL:%s [%ld] -> \"%s\" [%d]\n",
1747 tms, u->url, hstat.len, locf, count);
1750 downloaded_increase (hstat.len);
1752 /* Remember that we downloaded the file for later ".orig" code. */
1753 if (*dt & ADDED_HTML_EXTENSION)
1754 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1756 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1758 xfree (filename_plus_orig_suffix); /* must precede every return! */
1761 else if (hstat.len < hstat.contlen) /* meaning we lost the
1762 connection too soon */
1764 logprintf (LOG_VERBOSE,
1765 _("%s (%s) - Connection closed at byte %ld. "),
1766 tms, tmrate, hstat.len);
1767 printwhat (count, opt.ntry);
1770 else if (!opt.kill_longer) /* meaning we got more than expected */
1772 logprintf (LOG_VERBOSE,
1773 _("%s (%s) - `%s' saved [%ld/%ld])\n\n"),
1774 tms, tmrate, locf, hstat.len, hstat.contlen);
1775 logprintf (LOG_NONVERBOSE,
1776 "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
1777 tms, u->url, hstat.len, hstat.contlen, locf, count);
1779 downloaded_increase (hstat.len);
1781 /* Remember that we downloaded the file for later ".orig" code. */
1782 if (*dt & ADDED_HTML_EXTENSION)
1783 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
1785 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
1787 xfree (filename_plus_orig_suffix); /* must precede every return! */
1790 else /* the same, but not accepted */
1792 logprintf (LOG_VERBOSE,
1793 _("%s (%s) - Connection closed at byte %ld/%ld. "),
1794 tms, tmrate, hstat.len, hstat.contlen);
1795 printwhat (count, opt.ntry);
1799 else /* now hstat.res can only be -1 */
1801 if (hstat.contlen == -1)
1803 logprintf (LOG_VERBOSE,
1804 _("%s (%s) - Read error at byte %ld (%s)."),
1805 tms, tmrate, hstat.len, strerror (errno));
1806 printwhat (count, opt.ntry);
1809 else /* hstat.res == -1 and contlen is given */
1811 logprintf (LOG_VERBOSE,
1812 _("%s (%s) - Read error at byte %ld/%ld (%s). "),
1813 tms, tmrate, hstat.len, hstat.contlen,
1815 printwhat (count, opt.ntry);
1822 while (!opt.ntry || (count < opt.ntry));
1823 xfree (filename_plus_orig_suffix); /* must precede every return! */
1827 /* Converts struct tm to time_t, assuming the data in tm is UTC rather
1828 than local timezone (mktime assumes the latter).
1830 Contributed by Roger Beeman <beeman@cisco.com>, with the help of
1831 Mark Baushke <mdb@cisco.com> and the rest of the Gurus at CISCO. */
1833 mktime_from_utc (struct tm *t)
1840 tb = mktime (gmtime (&tl));
1841 return (tl <= tb ? (tl + (tl - tb)) : (tl - (tb - tl)));
1844 /* Check whether the result of strptime() indicates success.
1845 strptime() returns the pointer to how far it got to in the string.
1846 The processing has been successful if the string is at `GMT' or
1847 `+X', or at the end of the string.
1849 In extended regexp parlance, the function returns 1 if P matches
1850 "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (a valid result of
1851 strptime()) is considered a failure and 0 is returned. */
1853 check_end (const char *p)
1857 while (ISSPACE (*p))
1860 || (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
1861 || ((p[0] == '+' || p[0] == '-') && ISDIGIT (p[1])))
1867 /* Convert TIME_STRING time to time_t. TIME_STRING can be in any of
1868 the three formats RFC2068 allows the HTTP servers to emit --
1869 RFC1123-date, RFC850-date or asctime-date. Timezones are ignored,
1872 We use strptime() to recognize various dates, which makes it a
1873 little bit slacker than the RFC1123/RFC850/asctime (e.g. it always
1874 allows shortened dates and months, one-digit days, etc.). It also
1875 allows more than one space anywhere where the specs require one SP.
1876 The routine should probably be even more forgiving (as recommended
1877 by RFC2068), but I do not have the time to write one.
1879 Return the computed time_t representation, or -1 if all the
1882 Needless to say, what we *really* need here is something like
1883 Marcus Hennecke's atotm(), which is forgiving, fast, to-the-point,
1884 and does not use strptime(). atotm() is to be found in the sources
1885 of `phttpd', a little-known HTTP server written by Peter Erikson. */
1887 http_atotm (char *time_string)
1891 /* Roger Beeman says: "This function dynamically allocates struct tm
1892 t, but does no initialization. The only field that actually
1893 needs initialization is tm_isdst, since the others will be set by
1894 strptime. Since strptime does not set tm_isdst, it will return
1895 the data structure with whatever data was in tm_isdst to begin
1896 with. For those of us in timezones where DST can occur, there
1897 can be a one hour shift depending on the previous contents of the
1898 data area where the data structure is allocated." */
1901 /* Note that under foreign locales Solaris strptime() fails to
1902 recognize English dates, which renders this function useless. I
1903 assume that other non-GNU strptime's are plagued by the same
1904 disease. We solve this by setting only LC_MESSAGES in
1905 i18n_initialize(), instead of LC_ALL.
1907 Another solution could be to temporarily set locale to C, invoke
1908 strptime(), and restore it back. This is slow and dirty,
1909 however, and locale support other than LC_MESSAGES can mess other
1910 things, so I rather chose to stick with just setting LC_MESSAGES.
1912 Also note that none of this is necessary under GNU strptime(),
1913 because it recognizes both international and local dates. */
1915 /* NOTE: We don't use `%n' for white space, as OSF's strptime uses
1916 it to eat all white space up to (and including) a newline, and
1917 the function fails if there is no newline (!).
1919 Let's hope all strptime() implementations use ` ' to skip *all*
1920 whitespace instead of just one (it works that way on all the
1921 systems I've tested it on). */
1923 /* RFC1123: Thu, 29 Jan 1998 22:12:57 */
1924 if (check_end (strptime (time_string, "%a, %d %b %Y %T", &t)))
1925 return mktime_from_utc (&t);
1926 /* RFC850: Thursday, 29-Jan-98 22:12:57 */
1927 if (check_end (strptime (time_string, "%A, %d-%b-%y %T", &t)))
1928 return mktime_from_utc (&t);
1929 /* pseudo-RFC850: Thu, 29-Jan-1998 22:12:57
1930 (google.com uses this for their cookies.)*/
1931 if (check_end (strptime (time_string, "%a, %d-%b-%Y %T", &t)))
1932 return mktime_from_utc (&t);
1933 /* asctime: Thu Jan 29 22:12:57 1998 */
1934 if (check_end (strptime (time_string, "%a %b %d %T %Y", &t)))
1935 return mktime_from_utc (&t);
1940 /* Authorization support: We support two authorization schemes:
1942 * `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
1944 * `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
1945 consisting of answering to the server's challenge with the proper
1948 /* How many bytes it will take to store LEN bytes in base64. */
1949 #define BASE64_LENGTH(len) (4 * (((len) + 2) / 3))
1951 /* Encode the string S of length LENGTH to base64 format and place it
1952 to STORE. STORE will be 0-terminated, and must point to a writable
1953 buffer of at least 1+BASE64_LENGTH(length) bytes. */
1955 base64_encode (const char *s, char *store, int length)
1957 /* Conversion table. */
1958 static char tbl[64] = {
1959 'A','B','C','D','E','F','G','H',
1960 'I','J','K','L','M','N','O','P',
1961 'Q','R','S','T','U','V','W','X',
1962 'Y','Z','a','b','c','d','e','f',
1963 'g','h','i','j','k','l','m','n',
1964 'o','p','q','r','s','t','u','v',
1965 'w','x','y','z','0','1','2','3',
1966 '4','5','6','7','8','9','+','/'
1969 unsigned char *p = (unsigned char *)store;
1971 /* Transform the 3x8 bits to 4x6 bits, as required by base64. */
1972 for (i = 0; i < length; i += 3)
1974 *p++ = tbl[s[0] >> 2];
1975 *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
1976 *p++ = tbl[((s[1] & 0xf) << 2) + (s[2] >> 6)];
1977 *p++ = tbl[s[2] & 0x3f];
1980 /* Pad the result if necessary... */
1981 if (i == length + 1)
1983 else if (i == length + 2)
1984 *(p - 1) = *(p - 2) = '=';
1985 /* ...and zero-terminate it. */
1989 /* Create the authentication header contents for the `Basic' scheme.
1990 This is done by encoding the string `USER:PASS' in base64 and
1991 prepending `HEADER: Basic ' to it. */
1993 basic_authentication_encode (const char *user, const char *passwd,
1996 char *t1, *t2, *res;
1997 int len1 = strlen (user) + 1 + strlen (passwd);
1998 int len2 = BASE64_LENGTH (len1);
2000 t1 = (char *)alloca (len1 + 1);
2001 sprintf (t1, "%s:%s", user, passwd);
2002 t2 = (char *)alloca (1 + len2);
2003 base64_encode (t1, t2, len1);
2004 res = (char *)xmalloc (len2 + 11 + strlen (header));
2005 sprintf (res, "%s: Basic %s\r\n", header, t2);
2011 /* Parse HTTP `WWW-Authenticate:' header. AU points to the beginning
2012 of a field in such a header. If the field is the one specified by
2013 ATTR_NAME ("realm", "opaque", and "nonce" are used by the current
2014 digest authorization code), extract its value in the (char*)
2015 variable pointed by RET. Returns negative on a malformed header,
2016 or number of bytes that have been parsed by this call. */
2018 extract_header_attr (const char *au, const char *attr_name, char **ret)
2020 const char *cp, *ep;
2024 if (strncmp (cp, attr_name, strlen (attr_name)) == 0)
2026 cp += strlen (attr_name);
2029 cp += skip_lws (cp);
2034 cp += skip_lws (cp);
2039 for (ep = cp; *ep && *ep != '\"'; ep++)
2044 *ret = strdupdelim (cp, ep);
2051 /* Dump the hexadecimal representation of HASH to BUF. HASH should be
2052 an array of 16 bytes containing the hash keys, and BUF should be a
2053 buffer of 33 writable characters (32 for hex digits plus one for
2054 zero termination). */
2056 dump_hash (unsigned char *buf, const unsigned char *hash)
2060 for (i = 0; i < MD5_HASHLEN; i++, hash++)
2062 *buf++ = XDIGIT_TO_xchar (*hash >> 4);
2063 *buf++ = XDIGIT_TO_xchar (*hash & 0xf);
2068 /* Take the line apart to find the challenge, and compose a digest
2069 authorization header. See RFC2069 section 2.1.2. */
2071 digest_authentication_encode (const char *au, const char *user,
2072 const char *passwd, const char *method,
2075 static char *realm, *opaque, *nonce;
2080 { "realm", &realm },
2081 { "opaque", &opaque },
2086 realm = opaque = nonce = NULL;
2088 au += 6; /* skip over `Digest' */
2093 au += skip_lws (au);
2094 for (i = 0; i < ARRAY_SIZE (options); i++)
2096 int skip = extract_header_attr (au, options[i].name,
2097 options[i].variable);
2101 FREE_MAYBE (opaque);
2111 if (i == ARRAY_SIZE (options))
2113 while (*au && *au != '=')
2117 au += skip_lws (au);
2121 while (*au && *au != '\"')
2128 while (*au && *au != ',')
2133 if (!realm || !nonce || !user || !passwd || !path || !method)
2136 FREE_MAYBE (opaque);
2141 /* Calculate the digest value. */
2144 unsigned char hash[MD5_HASHLEN];
2145 unsigned char a1buf[MD5_HASHLEN * 2 + 1], a2buf[MD5_HASHLEN * 2 + 1];
2146 unsigned char response_digest[MD5_HASHLEN * 2 + 1];
2148 /* A1BUF = H(user ":" realm ":" password) */
2149 md5_init_ctx (&ctx);
2150 md5_process_bytes (user, strlen (user), &ctx);
2151 md5_process_bytes (":", 1, &ctx);
2152 md5_process_bytes (realm, strlen (realm), &ctx);
2153 md5_process_bytes (":", 1, &ctx);
2154 md5_process_bytes (passwd, strlen (passwd), &ctx);
2155 md5_finish_ctx (&ctx, hash);
2156 dump_hash (a1buf, hash);
2158 /* A2BUF = H(method ":" path) */
2159 md5_init_ctx (&ctx);
2160 md5_process_bytes (method, strlen (method), &ctx);
2161 md5_process_bytes (":", 1, &ctx);
2162 md5_process_bytes (path, strlen (path), &ctx);
2163 md5_finish_ctx (&ctx, hash);
2164 dump_hash (a2buf, hash);
2166 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */
2167 md5_init_ctx (&ctx);
2168 md5_process_bytes (a1buf, MD5_HASHLEN * 2, &ctx);
2169 md5_process_bytes (":", 1, &ctx);
2170 md5_process_bytes (nonce, strlen (nonce), &ctx);
2171 md5_process_bytes (":", 1, &ctx);
2172 md5_process_bytes (a2buf, MD5_HASHLEN * 2, &ctx);
2173 md5_finish_ctx (&ctx, hash);
2174 dump_hash (response_digest, hash);
2176 res = (char*) xmalloc (strlen (user)
2181 + 2 * MD5_HASHLEN /*strlen (response_digest)*/
2182 + (opaque ? strlen (opaque) : 0)
2184 sprintf (res, "Authorization: Digest \
2185 username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
2186 user, realm, nonce, path, response_digest);
2189 char *p = res + strlen (res);
2190 strcat (p, ", opaque=\"");
2194 strcat (res, "\r\n");
2198 #endif /* USE_DIGEST */
2201 #define BEGINS_WITH(line, string_constant) \
2202 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
2203 && (ISSPACE (line[sizeof (string_constant) - 1]) \
2204 || !line[sizeof (string_constant) - 1]))
2207 known_authentication_scheme_p (const char *au)
2209 return BEGINS_WITH (au, "Basic")
2210 || BEGINS_WITH (au, "Digest")
2211 || BEGINS_WITH (au, "NTLM");
2216 /* Create the HTTP authorization request header. When the
2217 `WWW-Authenticate' response header is seen, according to the
2218 authorization scheme specified in that header (`Basic' and `Digest'
2219 are supported by the current implementation), produce an
2220 appropriate HTTP authorization request header. */
2222 create_authorization_line (const char *au, const char *user,
2223 const char *passwd, const char *method,
2226 char *wwwauth = NULL;
2228 if (!strncasecmp (au, "Basic", 5))
2229 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
2230 if (!strncasecmp (au, "NTLM", 4))
2231 wwwauth = basic_authentication_encode (user, passwd, "Authorization");
2233 else if (!strncasecmp (au, "Digest", 6))
2234 wwwauth = digest_authentication_encode (au, user, passwd, method, path);
2235 #endif /* USE_DIGEST */