2 Copyright (C) 2005 Free Software Foundation, Inc.
4 This file is part of GNU Wget.
6 GNU Wget 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 GNU Wget 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 Wget; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
20 In addition, as a special exception, the Free Software Foundation
21 gives permission to link the code of its release of Wget with the
22 OpenSSL project's "OpenSSL" library (or with modified versions of it
23 that use the same license as the "OpenSSL" library), and distribute
24 the linked executables. You must obey the GNU General Public License
25 in all respects for all of the code used other than "OpenSSL". If you
26 modify this file, you may extend this exception to your version of the
27 file, but you are not obligated to do so. If you do not wish to do
28 so, delete this exception statement from your version. */
34 #include <sys/types.h>
45 #if TIME_WITH_SYS_TIME
46 # include <sys/time.h>
50 # include <sys/time.h>
67 # include "gen_sslfunc.h"
70 # include "http-ntlm.h"
78 extern char *version_string;
79 extern LARGE_INT total_downloaded_bytes;
81 extern FILE *output_stream;
82 extern int output_stream_regular;
85 # define MIN(x, y) ((x) > (y) ? (y) : (x))
89 static int cookies_loaded_p;
90 static struct cookie_jar *wget_cookie_jar;
92 #define TEXTHTML_S "text/html"
93 #define TEXTXHTML_S "application/xhtml+xml"
95 /* Some status code validation macros: */
96 #define H_20X(x) (((x) >= 200) && ((x) < 300))
97 #define H_PARTIAL(x) ((x) == HTTP_STATUS_PARTIAL_CONTENTS)
98 #define H_REDIRECTED(x) ((x) == HTTP_STATUS_MOVED_PERMANENTLY \
99 || (x) == HTTP_STATUS_MOVED_TEMPORARILY \
100 || (x) == HTTP_STATUS_SEE_OTHER \
101 || (x) == HTTP_STATUS_TEMPORARY_REDIRECT)
103 /* HTTP/1.0 status codes from RFC1945, provided for reference. */
104 /* Successful 2xx. */
105 #define HTTP_STATUS_OK 200
106 #define HTTP_STATUS_CREATED 201
107 #define HTTP_STATUS_ACCEPTED 202
108 #define HTTP_STATUS_NO_CONTENT 204
109 #define HTTP_STATUS_PARTIAL_CONTENTS 206
111 /* Redirection 3xx. */
112 #define HTTP_STATUS_MULTIPLE_CHOICES 300
113 #define HTTP_STATUS_MOVED_PERMANENTLY 301
114 #define HTTP_STATUS_MOVED_TEMPORARILY 302
115 #define HTTP_STATUS_SEE_OTHER 303 /* from HTTP/1.1 */
116 #define HTTP_STATUS_NOT_MODIFIED 304
117 #define HTTP_STATUS_TEMPORARY_REDIRECT 307 /* from HTTP/1.1 */
119 /* Client error 4xx. */
120 #define HTTP_STATUS_BAD_REQUEST 400
121 #define HTTP_STATUS_UNAUTHORIZED 401
122 #define HTTP_STATUS_FORBIDDEN 403
123 #define HTTP_STATUS_NOT_FOUND 404
124 #define HTTP_STATUS_RANGE_NOT_SATISFIABLE 416
126 /* Server errors 5xx. */
127 #define HTTP_STATUS_INTERNAL 500
128 #define HTTP_STATUS_NOT_IMPLEMENTED 501
129 #define HTTP_STATUS_BAD_GATEWAY 502
130 #define HTTP_STATUS_UNAVAILABLE 503
133 rel_none, rel_name, rel_value, rel_both
140 struct request_header {
142 enum rp release_policy;
144 int hcount, hcapacity;
147 /* Create a new, empty request. At least request_set_method must be
148 called before the request can be used. */
150 static struct request *
153 struct request *req = xnew0 (struct request);
155 req->headers = xnew_array (struct request_header, req->hcapacity);
159 /* Set the request's method and its arguments. METH should be a
160 literal string (or it should outlive the request) because it will
161 not be freed. ARG will be freed by request_free. */
164 request_set_method (struct request *req, const char *meth, char *arg)
170 /* Return the method string passed with the last call to
171 request_set_method. */
174 request_method (const struct request *req)
179 /* Free one header according to the release policy specified with
180 request_set_header. */
183 release_header (struct request_header *hdr)
185 switch (hdr->release_policy)
202 /* Set the request named NAME to VALUE. Specifically, this means that
203 a "NAME: VALUE\r\n" header line will be used in the request. If a
204 header with the same name previously existed in the request, its
205 value will be replaced by this one.
207 RELEASE_POLICY determines whether NAME and VALUE should be released
208 (freed) with request_free. Allowed values are:
210 - rel_none - don't free NAME or VALUE
211 - rel_name - free NAME when done
212 - rel_value - free VALUE when done
213 - rel_both - free both NAME and VALUE when done
215 Setting release policy is useful when arguments come from different
216 sources. For example:
218 // Don't free literal strings!
219 request_set_header (req, "Pragma", "no-cache", rel_none);
221 // Don't free a global variable, we'll need it later.
222 request_set_header (req, "Referer", opt.referer, rel_none);
224 // Value freshly allocated, free it when done.
225 request_set_header (req, "Range",
226 aprintf ("bytes=%s-", number_to_static_string (hs->restval)),
231 request_set_header (struct request *req, char *name, char *value,
232 enum rp release_policy)
234 struct request_header *hdr;
238 for (i = 0; i < req->hcount; i++)
240 hdr = &req->headers[i];
241 if (0 == strcasecmp (name, hdr->name))
243 /* Replace existing header. */
244 release_header (hdr);
247 hdr->release_policy = release_policy;
252 /* Install new header. */
254 if (req->hcount >= req->hcount)
256 req->hcapacity <<= 1;
257 req->headers = xrealloc (req->headers,
258 req->hcapacity * sizeof (struct request_header));
260 hdr = &req->headers[req->hcount++];
263 hdr->release_policy = release_policy;
266 /* Like request_set_header, but sets the whole header line, as
267 provided by the user using the `--header' option. For example,
268 request_set_user_header (req, "Foo: bar") works just like
269 request_set_header (req, "Foo", "bar"). */
272 request_set_user_header (struct request *req, const char *header)
275 const char *p = strchr (header, ':');
278 BOUNDED_TO_ALLOCA (header, p, name);
282 request_set_header (req, xstrdup (name), (char *) p, rel_name);
285 #define APPEND(p, str) do { \
286 int A_len = strlen (str); \
287 memcpy (p, str, A_len); \
291 /* Construct the request and write it to FD using fd_write. */
294 request_send (const struct request *req, int fd)
296 char *request_string, *p;
297 int i, size, write_error;
299 /* Count the request size. */
302 /* METHOD " " ARG " " "HTTP/1.0" "\r\n" */
303 size += strlen (req->method) + 1 + strlen (req->arg) + 1 + 8 + 2;
305 for (i = 0; i < req->hcount; i++)
307 struct request_header *hdr = &req->headers[i];
308 /* NAME ": " VALUE "\r\n" */
309 size += strlen (hdr->name) + 2 + strlen (hdr->value) + 2;
315 p = request_string = alloca_array (char, size);
317 /* Generate the request. */
319 APPEND (p, req->method); *p++ = ' ';
320 APPEND (p, req->arg); *p++ = ' ';
321 memcpy (p, "HTTP/1.0\r\n", 10); p += 10;
323 for (i = 0; i < req->hcount; i++)
325 struct request_header *hdr = &req->headers[i];
326 APPEND (p, hdr->name);
327 *p++ = ':', *p++ = ' ';
328 APPEND (p, hdr->value);
329 *p++ = '\r', *p++ = '\n';
332 *p++ = '\r', *p++ = '\n', *p++ = '\0';
333 assert (p - request_string == size);
337 DEBUGP (("\n---request begin---\n%s---request end---\n", request_string));
339 /* Send the request to the server. */
341 write_error = fd_write (fd, request_string, size - 1, -1);
343 logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"),
348 /* Release the resources used by REQ. */
351 request_free (struct request *req)
354 xfree_null (req->arg);
355 for (i = 0; i < req->hcount; i++)
356 release_header (&req->headers[i]);
357 xfree_null (req->headers);
361 /* Send the contents of FILE_NAME to SOCK. Make sure that exactly
362 PROMISED_SIZE bytes are sent over the wire -- if the file is
363 longer, read only that much; if the file is shorter, report an error. */
366 post_file (int sock, const char *file_name, wgint promised_size)
368 static char chunk[8192];
373 DEBUGP (("[writing POST file %s ... ", file_name));
375 fp = fopen (file_name, "rb");
378 while (!feof (fp) && written < promised_size)
381 int length = fread (chunk, 1, sizeof (chunk), fp);
384 towrite = MIN (promised_size - written, length);
385 write_error = fd_write (sock, chunk, towrite, -1);
395 /* If we've written less than was promised, report a (probably
396 nonsensical) error rather than break the promise. */
397 if (written < promised_size)
403 assert (written == promised_size);
404 DEBUGP (("done]\n"));
409 response_head_terminator (const char *hunk, int oldlen, int peeklen)
411 const char *start, *end;
413 /* If at first peek, verify whether HUNK starts with "HTTP". If
414 not, this is a HTTP/0.9 request and we must bail out without
416 if (oldlen == 0 && 0 != memcmp (hunk, "HTTP", MIN (peeklen, 4)))
422 start = hunk + oldlen - 4;
423 end = hunk + oldlen + peeklen;
425 for (; start < end - 1; start++)
432 if (start[1] == '\n')
438 /* The maximum size of a single HTTP response we care to read. This
439 is not meant to impose an arbitrary limit, but to protect the user
440 from Wget slurping up available memory upon encountering malicious
441 or buggy server output. Define it to 0 to remove the limit. */
443 #define HTTP_RESPONSE_MAX_SIZE 65536
445 /* Read the HTTP request head from FD and return it. The error
446 conditions are the same as with fd_read_hunk.
448 To support HTTP/0.9 responses, this function tries to make sure
449 that the data begins with "HTTP". If this is not the case, no data
450 is read and an empty request is returned, so that the remaining
451 data can be treated as body. */
454 read_http_response_head (int fd)
456 return fd_read_hunk (fd, response_head_terminator, 512,
457 HTTP_RESPONSE_MAX_SIZE);
461 /* The response data. */
464 /* The array of pointers that indicate where each header starts.
465 For example, given this HTTP response:
472 The headers are located like this:
474 "HTTP/1.0 200 Ok\r\nDescription: some\r\n text\r\nEtag: x\r\n\r\n"
476 headers[0] headers[1] headers[2] headers[3]
478 I.e. headers[0] points to the beginning of the request,
479 headers[1] points to the end of the first header and the
480 beginning of the second one, etc. */
482 const char **headers;
485 /* Create a new response object from the text of the HTTP response,
486 available in HEAD. That text is automatically split into
487 constituent header lines for fast retrieval using
490 static struct response *
491 resp_new (const char *head)
496 struct response *resp = xnew0 (struct response);
501 /* Empty head means that we're dealing with a headerless
502 (HTTP/0.9) response. In that case, don't set HEADERS at
507 /* Split HEAD into header lines, so that resp_header_* functions
508 don't need to do this over and over again. */
514 DO_REALLOC (resp->headers, size, count + 1, const char *);
515 resp->headers[count++] = hdr;
517 /* Break upon encountering an empty line. */
518 if (!hdr[0] || (hdr[0] == '\r' && hdr[1] == '\n') || hdr[0] == '\n')
521 /* Find the end of HDR, including continuations. */
524 const char *end = strchr (hdr, '\n');
530 while (*hdr == ' ' || *hdr == '\t');
532 DO_REALLOC (resp->headers, size, count + 1, const char *);
533 resp->headers[count] = NULL;
538 /* Locate the header named NAME in the request data, starting with
539 position START. This allows the code to loop through the request
540 data, filtering for all requests of a given name. Returns the
541 found position, or -1 for failure. The code that uses this
542 function typically looks like this:
544 for (pos = 0; (pos = resp_header_locate (...)) != -1; pos++)
545 ... do something with header ...
547 If you only care about one header, use resp_header_get instead of
551 resp_header_locate (const struct response *resp, const char *name, int start,
552 const char **begptr, const char **endptr)
555 const char **headers = resp->headers;
558 if (!headers || !headers[1])
561 name_len = strlen (name);
567 for (; headers[i + 1]; i++)
569 const char *b = headers[i];
570 const char *e = headers[i + 1];
572 && b[name_len] == ':'
573 && 0 == strncasecmp (b, name, name_len))
576 while (b < e && ISSPACE (*b))
578 while (b < e && ISSPACE (e[-1]))
588 /* Find and retrieve the header named NAME in the request data. If
589 found, set *BEGPTR to its starting, and *ENDPTR to its ending
590 position, and return 1. Otherwise return 0.
592 This function is used as a building block for resp_header_copy
593 and resp_header_strdup. */
596 resp_header_get (const struct response *resp, const char *name,
597 const char **begptr, const char **endptr)
599 int pos = resp_header_locate (resp, name, 0, begptr, endptr);
603 /* Copy the response header named NAME to buffer BUF, no longer than
604 BUFSIZE (BUFSIZE includes the terminating 0). If the header
605 exists, 1 is returned, otherwise 0. If there should be no limit on
606 the size of the header, use resp_header_strdup instead.
608 If BUFSIZE is 0, no data is copied, but the boolean indication of
609 whether the header is present is still returned. */
612 resp_header_copy (const struct response *resp, const char *name,
613 char *buf, int bufsize)
616 if (!resp_header_get (resp, name, &b, &e))
620 int len = MIN (e - b, bufsize - 1);
621 memcpy (buf, b, len);
627 /* Return the value of header named NAME in RESP, allocated with
628 malloc. If such a header does not exist in RESP, return NULL. */
631 resp_header_strdup (const struct response *resp, const char *name)
634 if (!resp_header_get (resp, name, &b, &e))
636 return strdupdelim (b, e);
639 /* Parse the HTTP status line, which is of format:
641 HTTP-Version SP Status-Code SP Reason-Phrase
643 The function returns the status-code, or -1 if the status line
644 appears malformed. The pointer to "reason-phrase" message is
645 returned in *MESSAGE. */
648 resp_status (const struct response *resp, char **message)
655 /* For a HTTP/0.9 response, assume status 200. */
657 *message = xstrdup (_("No headers, assuming HTTP/0.9"));
661 p = resp->headers[0];
662 end = resp->headers[1];
668 if (end - p < 4 || 0 != strncmp (p, "HTTP", 4))
672 /* Match the HTTP version. This is optional because Gnutella
673 servers have been reported to not specify HTTP version. */
674 if (p < end && *p == '/')
677 while (p < end && ISDIGIT (*p))
679 if (p < end && *p == '.')
681 while (p < end && ISDIGIT (*p))
685 while (p < end && ISSPACE (*p))
687 if (end - p < 3 || !ISDIGIT (p[0]) || !ISDIGIT (p[1]) || !ISDIGIT (p[2]))
690 status = 100 * (p[0] - '0') + 10 * (p[1] - '0') + (p[2] - '0');
695 while (p < end && ISSPACE (*p))
697 while (p < end && ISSPACE (end[-1]))
699 *message = strdupdelim (p, end);
705 /* Release the resources used by RESP. */
708 resp_free (struct response *resp)
710 xfree_null (resp->headers);
714 /* Print the server response, line by line, omitting the trailing CRLF
715 from individual header lines, and prefixed with PREFIX. */
718 print_server_response (const struct response *resp, const char *prefix)
723 for (i = 0; resp->headers[i + 1]; i++)
725 const char *b = resp->headers[i];
726 const char *e = resp->headers[i + 1];
728 if (b < e && e[-1] == '\n')
730 if (b < e && e[-1] == '\r')
732 /* This is safe even on printfs with broken handling of "%.<n>s"
733 because resp->headers ends with \0. */
734 logprintf (LOG_VERBOSE, "%s%.*s\n", prefix, e - b, b);
738 /* Parse the `Content-Range' header and extract the information it
739 contains. Returns 1 if successful, -1 otherwise. */
741 parse_content_range (const char *hdr, wgint *first_byte_ptr,
742 wgint *last_byte_ptr, wgint *entity_length_ptr)
746 /* Ancient versions of Netscape proxy server, presumably predating
747 rfc2068, sent out `Content-Range' without the "bytes"
749 if (!strncasecmp (hdr, "bytes", 5))
752 /* "JavaWebServer/1.1.1" sends "bytes: x-y/z", contrary to the
756 while (ISSPACE (*hdr))
763 for (num = 0; ISDIGIT (*hdr); hdr++)
764 num = 10 * num + (*hdr - '0');
765 if (*hdr != '-' || !ISDIGIT (*(hdr + 1)))
767 *first_byte_ptr = num;
769 for (num = 0; ISDIGIT (*hdr); hdr++)
770 num = 10 * num + (*hdr - '0');
771 if (*hdr != '/' || !ISDIGIT (*(hdr + 1)))
773 *last_byte_ptr = num;
775 for (num = 0; ISDIGIT (*hdr); hdr++)
776 num = 10 * num + (*hdr - '0');
777 *entity_length_ptr = num;
781 /* Read the body of the request, but don't store it anywhere and don't
782 display a progress gauge. This is useful for reading the bodies of
783 administrative responses to which we will soon issue another
784 request. The response is not useful to the user, but reading it
785 allows us to continue using the same connection to the server.
787 If reading fails, 0 is returned, non-zero otherwise. In debug
788 mode, the body is displayed for debugging purposes. */
791 skip_short_body (int fd, wgint contlen)
794 SKIP_SIZE = 512, /* size of the download buffer */
795 SKIP_THRESHOLD = 4096 /* the largest size we read */
797 char dlbuf[SKIP_SIZE + 1];
798 dlbuf[SKIP_SIZE] = '\0'; /* so DEBUGP can safely print it */
800 /* We shouldn't get here with unknown contlen. (This will change
801 with HTTP/1.1, which supports "chunked" transfer.) */
802 assert (contlen != -1);
804 /* If the body is too large, it makes more sense to simply close the
805 connection than to try to read the body. */
806 if (contlen > SKIP_THRESHOLD)
809 DEBUGP (("Skipping %s bytes of body: [", number_to_static_string (contlen)));
813 int ret = fd_read (fd, dlbuf, MIN (contlen, SKIP_SIZE), -1);
816 /* Don't normally report the error since this is an
817 optimization that should be invisible to the user. */
818 DEBUGP (("] aborting (%s).\n",
819 ret < 0 ? strerror (errno) : "EOF received"));
823 /* Safe even if %.*s bogusly expects terminating \0 because
824 we've zero-terminated dlbuf above. */
825 DEBUGP (("%.*s", ret, dlbuf));
828 DEBUGP (("] done.\n"));
832 /* Persistent connections. Currently, we cache the most recently used
833 connection as persistent, provided that the HTTP server agrees to
834 make it such. The persistence data is stored in the variables
835 below. Ideally, it should be possible to cache an arbitrary fixed
836 number of these connections. */
838 /* Whether a persistent connection is active. */
839 static int pconn_active;
842 /* The socket of the connection. */
845 /* Host and port of the currently active persistent connection. */
849 /* Whether a ssl handshake has occoured on this connection. */
853 /* NTLM data of the current connection. */
854 struct ntlmdata ntlm;
858 /* Mark the persistent connection as invalid and free the resources it
859 uses. This is used by the CLOSE_* macros after they forcefully
860 close a registered persistent connection. */
863 invalidate_persistent (void)
865 DEBUGP (("Disabling further reuse of socket %d.\n", pconn.socket));
867 fd_close (pconn.socket);
872 /* Register FD, which should be a TCP/IP connection to HOST:PORT, as
873 persistent. This will enable someone to use the same connection
874 later. In the context of HTTP, this must be called only AFTER the
875 response has been received and the server has promised that the
876 connection will remain alive.
878 If a previous connection was persistent, it is closed. */
881 register_persistent (const char *host, int port, int fd, int ssl)
885 if (pconn.socket == fd)
887 /* The connection FD is already registered. */
892 /* The old persistent connection is still active; close it
893 first. This situation arises whenever a persistent
894 connection exists, but we then connect to a different
895 host, and try to register a persistent connection to that
897 invalidate_persistent ();
903 pconn.host = xstrdup (host);
907 DEBUGP (("Registered socket %d for persistent reuse.\n", fd));
910 /* Return non-zero if a persistent connection is available for
911 connecting to HOST:PORT. */
914 persistent_available_p (const char *host, int port, int ssl,
915 int *host_lookup_failed)
917 /* First, check whether a persistent connection is active at all. */
921 /* If we want SSL and the last connection wasn't or vice versa,
922 don't use it. Checking for host and port is not enough because
923 HTTP and HTTPS can apparently coexist on the same port. */
924 if (ssl != pconn.ssl)
927 /* If we're not connecting to the same port, we're not interested. */
928 if (port != pconn.port)
931 /* If the host is the same, we're in business. If not, there is
932 still hope -- read below. */
933 if (0 != strcasecmp (host, pconn.host))
935 /* If pconn.socket is already talking to HOST, we needn't
936 reconnect. This happens often when both sites are virtual
937 hosts distinguished only by name and served by the same
938 network interface, and hence the same web server (possibly
939 set up by the ISP and serving many different web sites).
940 This admittedly non-standard optimization does not contradict
941 HTTP and works well with popular server software. */
945 struct address_list *al;
948 /* Don't try to talk to two different SSL sites over the same
949 secure connection! (Besides, it's not clear if name-based
950 virtual hosting is even possible with SSL.) */
953 /* If pconn.socket's peer is one of the IP addresses HOST
954 resolves to, pconn.socket is for all intents and purposes
955 already talking to HOST. */
957 if (!socket_ip_address (pconn.socket, &ip, ENDPOINT_PEER))
959 /* Can't get the peer's address -- something must be very
960 wrong with the connection. */
961 invalidate_persistent ();
964 al = lookup_host (host, 0);
967 *host_lookup_failed = 1;
971 found = address_list_contains (al, &ip);
972 address_list_release (al);
977 /* The persistent connection's peer address was found among the
978 addresses HOST resolved to; therefore, pconn.sock is in fact
979 already talking to HOST -- no need to reconnect. */
982 /* Finally, check whether the connection is still open. This is
983 important because most server implement a liberal (short) timeout
984 on persistent connections. Wget can of course always reconnect
985 if the connection doesn't work out, but it's nicer to know in
986 advance. This test is a logical followup of the first test, but
987 is "expensive" and therefore placed at the end of the list. */
989 if (!test_socket_open (pconn.socket))
991 /* Oops, the socket is no longer open. Now that we know that,
992 let's invalidate the persistent connection before returning
994 invalidate_persistent ();
1001 /* The idea behind these two CLOSE macros is to distinguish between
1002 two cases: one when the job we've been doing is finished, and we
1003 want to close the connection and leave, and two when something is
1004 seriously wrong and we're closing the connection as part of
1007 In case of keep_alive, CLOSE_FINISH should leave the connection
1008 open, while CLOSE_INVALIDATE should still close it.
1010 Note that the semantics of the flag `keep_alive' is "this
1011 connection *will* be reused (the server has promised not to close
1012 the connection once we're done)", while the semantics of
1013 `pc_active_p && (fd) == pc_last_fd' is "we're *now* using an
1014 active, registered connection". */
1016 #define CLOSE_FINISH(fd) do { \
1019 if (pconn_active && (fd) == pconn.socket) \
1020 invalidate_persistent (); \
1029 #define CLOSE_INVALIDATE(fd) do { \
1030 if (pconn_active && (fd) == pconn.socket) \
1031 invalidate_persistent (); \
1039 wgint len; /* received length */
1040 wgint contlen; /* expected length */
1041 wgint restval; /* the restart value */
1042 int res; /* the result of last read */
1043 char *newloc; /* new location (redirection) */
1044 char *remote_time; /* remote time-stamp string */
1045 char *error; /* textual HTTP error */
1046 int statcode; /* status code */
1047 wgint rd_size; /* amount of data read from socket */
1048 double dltime; /* time it took to download the data */
1049 const char *referer; /* value of the referer header. */
1050 char **local_file; /* local file. */
1054 free_hstat (struct http_stat *hs)
1056 xfree_null (hs->newloc);
1057 xfree_null (hs->remote_time);
1058 xfree_null (hs->error);
1060 /* Guard against being called twice. */
1062 hs->remote_time = NULL;
1066 static char *create_authorization_line PARAMS ((const char *, const char *,
1067 const char *, const char *,
1069 static char *basic_authentication_encode PARAMS ((const char *, const char *));
1070 static int known_authentication_scheme_p PARAMS ((const char *));
1072 time_t http_atotm PARAMS ((const char *));
1074 #define BEGINS_WITH(line, string_constant) \
1075 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
1076 && (ISSPACE (line[sizeof (string_constant) - 1]) \
1077 || !line[sizeof (string_constant) - 1]))
1079 /* Retrieve a document through HTTP protocol. It recognizes status
1080 code, and correctly handles redirections. It closes the network
1081 socket. If it receives an error from the functions below it, it
1082 will print it if there is enough information to do so (almost
1083 always), returning the error to the caller (i.e. http_loop).
1085 Various HTTP parameters are stored to hs.
1087 If PROXY is non-NULL, the connection will be made to the proxy
1088 server, and u->url will be requested. */
1090 gethttp (struct url *u, struct http_stat *hs, int *dt, struct url *proxy)
1092 struct request *req;
1095 char *user, *passwd;
1099 wgint contlen, contrange;
1106 /* Whether authorization has been already tried. */
1107 int auth_tried_already;
1109 /* Whether our connection to the remote host is through SSL. */
1113 struct response *resp;
1117 /* Whether this connection will be kept alive after the HTTP request
1121 /* Whether keep-alive should be inhibited.
1123 RFC 2068 requests that 1.0 clients not send keep-alive requests
1124 to proxies. This is because many 1.0 proxies do not interpret
1125 the Connection header and transfer it to the remote server,
1126 causing it to not close the connection and leave both the proxy
1127 and the client hanging. */
1128 int inhibit_keep_alive =
1129 !opt.http_keep_alive || opt.ignore_length || proxy != NULL;
1131 /* Headers sent when using POST. */
1132 wgint post_data_size = 0;
1134 int host_lookup_failed = 0;
1137 if (u->scheme == SCHEME_HTTPS)
1139 /* Initialize the SSL context. After this has once been done,
1140 it becomes a no-op. */
1141 switch (ssl_init ())
1143 case SSLERRCTXCREATE:
1145 logprintf (LOG_NOTQUIET, _("Failed to set up an SSL context\n"));
1146 return SSLERRCTXCREATE;
1147 case SSLERRCERTFILE:
1148 /* try without certfile */
1149 logprintf (LOG_NOTQUIET,
1150 _("Failed to load certificates from %s\n"),
1152 logprintf (LOG_NOTQUIET,
1153 _("Trying without the specified certificate\n"));
1156 logprintf (LOG_NOTQUIET,
1157 _("Failed to get certificate key from %s\n"),
1159 logprintf (LOG_NOTQUIET,
1160 _("Trying without the specified certificate\n"));
1166 #endif /* HAVE_SSL */
1168 if (!(*dt & HEAD_ONLY))
1169 /* If we're doing a GET on the URL, as opposed to just a HEAD, we need to
1170 know the local filename so we can save to it. */
1171 assert (*hs->local_file != NULL);
1173 auth_tried_already = 0;
1175 /* Initialize certain elements of struct http_stat. */
1180 hs->remote_time = NULL;
1185 /* Prepare the request to send. */
1187 req = request_new ();
1189 const char *meth = "GET";
1190 if (*dt & HEAD_ONLY)
1192 else if (opt.post_file_name || opt.post_data)
1194 /* Use the full path, i.e. one that includes the leading slash and
1195 the query string. E.g. if u->path is "foo/bar" and u->query is
1196 "param=value", full_path will be "/foo/bar?param=value". */
1197 request_set_method (req, meth,
1198 proxy ? xstrdup (u->url) : url_full_path (u));
1201 request_set_header (req, "Referer", (char *) hs->referer, rel_none);
1202 if (*dt & SEND_NOCACHE)
1203 request_set_header (req, "Pragma", "no-cache", rel_none);
1205 request_set_header (req, "Range",
1206 aprintf ("bytes=%s-",
1207 number_to_static_string (hs->restval)),
1210 request_set_header (req, "User-Agent", opt.useragent, rel_none);
1212 request_set_header (req, "User-Agent",
1213 aprintf ("Wget/%s", version_string), rel_value);
1214 request_set_header (req, "Accept", "*/*", rel_none);
1216 /* Find the username and password for authentication. */
1219 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
1220 user = user ? user : opt.http_user;
1221 passwd = passwd ? passwd : opt.http_passwd;
1225 /* We have the username and the password, but haven't tried
1226 any authorization yet. Let's see if the "Basic" method
1227 works. If not, we'll come back here and construct a
1228 proper authorization method with the right challenges.
1230 If we didn't employ this kind of logic, every URL that
1231 requires authorization would have to be processed twice,
1232 which is very suboptimal and generates a bunch of false
1233 "unauthorized" errors in the server log.
1235 #### But this logic also has a serious problem when used
1236 with stronger authentications: we *first* transmit the
1237 username and the password in clear text, and *then* attempt a
1238 stronger authentication scheme. That cannot be right! We
1239 are only fortunate that almost everyone still uses the
1240 `Basic' scheme anyway.
1242 There should be an option to prevent this from happening, for
1243 those who use strong authentication schemes and value their
1245 request_set_header (req, "Authorization",
1246 basic_authentication_encode (user, passwd),
1253 char *proxy_user, *proxy_passwd;
1254 /* For normal username and password, URL components override
1255 command-line/wgetrc parameters. With proxy
1256 authentication, it's the reverse, because proxy URLs are
1257 normally the "permanent" ones, so command-line args
1258 should take precedence. */
1259 if (opt.proxy_user && opt.proxy_passwd)
1261 proxy_user = opt.proxy_user;
1262 proxy_passwd = opt.proxy_passwd;
1266 proxy_user = proxy->user;
1267 proxy_passwd = proxy->passwd;
1269 /* #### This does not appear right. Can't the proxy request,
1270 say, `Digest' authentication? */
1271 if (proxy_user && proxy_passwd)
1272 proxyauth = basic_authentication_encode (proxy_user, proxy_passwd);
1274 /* If we're using a proxy, we will be connecting to the proxy
1278 /* Proxy authorization over SSL is handled below. */
1280 if (u->scheme != SCHEME_HTTPS)
1282 request_set_header (req, "Proxy-Authorization", proxyauth, rel_value);
1286 /* Whether we need to print the host header with braces around
1287 host, e.g. "Host: [3ffe:8100:200:2::2]:1234" instead of the
1288 usual "Host: symbolic-name:1234". */
1289 int squares = strchr (u->host, ':') != NULL;
1290 if (u->port == scheme_default_port (u->scheme))
1291 request_set_header (req, "Host",
1292 aprintf (squares ? "[%s]" : "%s", u->host),
1295 request_set_header (req, "Host",
1296 aprintf (squares ? "[%s]:%d" : "%s:%d",
1301 if (!inhibit_keep_alive)
1302 request_set_header (req, "Connection", "Keep-Alive", rel_none);
1305 request_set_header (req, "Cookie",
1306 cookie_header (wget_cookie_jar,
1307 u->host, u->port, u->path,
1309 u->scheme == SCHEME_HTTPS
1316 if (opt.post_data || opt.post_file_name)
1318 request_set_header (req, "Content-Type",
1319 "application/x-www-form-urlencoded", rel_none);
1321 post_data_size = strlen (opt.post_data);
1324 post_data_size = file_size (opt.post_file_name);
1325 if (post_data_size == -1)
1327 logprintf (LOG_NOTQUIET, "POST data file missing: %s\n",
1328 opt.post_file_name);
1332 request_set_header (req, "Content-Length",
1333 xstrdup (number_to_static_string (post_data_size)),
1337 /* Add the user headers. */
1338 if (opt.user_headers)
1341 for (i = 0; opt.user_headers[i]; i++)
1342 request_set_user_header (req, opt.user_headers[i]);
1346 /* We need to come back here when the initial attempt to retrieve
1347 without authorization header fails. (Expected to happen at least
1348 for the Digest authorization scheme.) */
1352 /* Establish the connection. */
1354 if (!inhibit_keep_alive)
1356 /* Look for a persistent connection to target host, unless a
1357 proxy is used. The exception is when SSL is in use, in which
1358 case the proxy is nothing but a passthrough to the target
1359 host, registered as a connection to the latter. */
1360 struct url *relevant = conn;
1362 if (u->scheme == SCHEME_HTTPS)
1366 if (persistent_available_p (relevant->host, relevant->port,
1368 relevant->scheme == SCHEME_HTTPS,
1372 &host_lookup_failed))
1374 sock = pconn.socket;
1375 using_ssl = pconn.ssl;
1376 logprintf (LOG_VERBOSE, _("Reusing existing connection to %s:%d.\n"),
1377 escnonprint (pconn.host), pconn.port);
1378 DEBUGP (("Reusing fd %d.\n", sock));
1384 /* In its current implementation, persistent_available_p will
1385 look up conn->host in some cases. If that lookup failed, we
1386 don't need to bother with connect_to_host. */
1387 if (host_lookup_failed)
1393 sock = connect_to_host (conn->host, conn->port);
1402 return (retryable_socket_connect_error (errno)
1403 ? CONERROR : CONIMPOSSIBLE);
1407 if (proxy && u->scheme == SCHEME_HTTPS)
1409 /* When requesting SSL URLs through proxies, use the
1410 CONNECT method to request passthrough. */
1411 struct request *connreq = request_new ();
1412 request_set_method (connreq, "CONNECT",
1413 aprintf ("%s:%d", u->host, u->port));
1416 request_set_header (connreq, "Proxy-Authorization",
1417 proxyauth, rel_value);
1418 /* Now that PROXYAUTH is part of the CONNECT request,
1419 zero it out so we don't send proxy authorization with
1420 the regular request below. */
1424 write_error = request_send (connreq, sock);
1425 request_free (connreq);
1426 if (write_error < 0)
1428 logprintf (LOG_VERBOSE, _("Failed writing to proxy: %s.\n"),
1430 CLOSE_INVALIDATE (sock);
1434 head = read_http_response_head (sock);
1437 logprintf (LOG_VERBOSE, _("Failed reading proxy response: %s\n"),
1439 CLOSE_INVALIDATE (sock);
1448 DEBUGP (("proxy responded with: [%s]\n", head));
1450 resp = resp_new (head);
1451 statcode = resp_status (resp, &message);
1454 if (statcode != 200)
1457 logprintf (LOG_NOTQUIET, _("Proxy tunneling failed: %s"),
1458 message ? escnonprint (message) : "?");
1459 xfree_null (message);
1462 xfree_null (message);
1464 /* SOCK is now *really* connected to u->host, so update CONN
1465 to reflect this. That way register_persistent will
1466 register SOCK as being connected to u->host:u->port. */
1470 if (conn->scheme == SCHEME_HTTPS)
1472 if (!ssl_connect (sock))
1479 #endif /* HAVE_SSL */
1482 /* Send the request to server. */
1483 write_error = request_send (req, sock);
1485 if (write_error >= 0)
1489 DEBUGP (("[POST data: %s]\n", opt.post_data));
1490 write_error = fd_write (sock, opt.post_data, post_data_size, -1);
1492 else if (opt.post_file_name && post_data_size != 0)
1493 write_error = post_file (sock, opt.post_file_name, post_data_size);
1496 if (write_error < 0)
1498 logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"),
1500 CLOSE_INVALIDATE (sock);
1504 logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
1505 proxy ? "Proxy" : "HTTP");
1510 head = read_http_response_head (sock);
1515 logputs (LOG_NOTQUIET, _("No data received.\n"));
1516 CLOSE_INVALIDATE (sock);
1522 logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
1524 CLOSE_INVALIDATE (sock);
1529 DEBUGP (("\n---response begin---\n%s---response end---\n", head));
1531 resp = resp_new (head);
1533 /* Check for status line. */
1535 statcode = resp_status (resp, &message);
1536 if (!opt.server_response)
1537 logprintf (LOG_VERBOSE, "%2d %s\n", statcode,
1538 message ? escnonprint (message) : "");
1541 logprintf (LOG_VERBOSE, "\n");
1542 print_server_response (resp, " ");
1545 if (!opt.ignore_length
1546 && resp_header_copy (resp, "Content-Length", hdrval, sizeof (hdrval)))
1550 parsed = str_to_wgint (hdrval, NULL, 10);
1551 if (parsed == WGINT_MAX && errno == ERANGE)
1553 #### If Content-Length is out of range, it most likely
1554 means that the file is larger than 2G and that we're
1555 compiled without LFS. In that case we should probably
1556 refuse to even attempt to download the file. */
1562 /* Check for keep-alive related responses. */
1563 if (!inhibit_keep_alive && contlen != -1)
1565 if (resp_header_copy (resp, "Keep-Alive", NULL, 0))
1567 else if (resp_header_copy (resp, "Connection", hdrval, sizeof (hdrval)))
1569 if (0 == strcasecmp (hdrval, "Keep-Alive"))
1574 /* The server has promised that it will not close the connection
1575 when we're done. This means that we can register it. */
1576 register_persistent (conn->host, conn->port, sock, using_ssl);
1578 if (statcode == HTTP_STATUS_UNAUTHORIZED)
1580 /* Authorization is required. */
1581 if (skip_short_body (sock, contlen))
1582 CLOSE_FINISH (sock);
1584 CLOSE_INVALIDATE (sock);
1585 if (auth_tried_already || !(user && passwd))
1587 /* If we have tried it already, then there is not point
1589 logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
1593 char *www_authenticate = resp_header_strdup (resp,
1594 "WWW-Authenticate");
1595 /* If the authentication scheme is unknown or if it's the
1596 "Basic" authentication (which we try by default), there's
1597 no sense in retrying. */
1598 if (!www_authenticate
1599 || !known_authentication_scheme_p (www_authenticate)
1600 || BEGINS_WITH (www_authenticate, "Basic"))
1602 xfree_null (www_authenticate);
1603 logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
1608 auth_tried_already = 1;
1609 pth = url_full_path (u);
1610 request_set_header (req, "Authorization",
1611 create_authorization_line (www_authenticate,
1613 request_method (req),
1617 xfree (www_authenticate);
1618 goto retry_with_auth;
1626 hs->statcode = statcode;
1628 hs->error = xstrdup (_("Malformed status line"));
1630 hs->error = xstrdup (_("(no description)"));
1632 hs->error = xstrdup (message);
1635 type = resp_header_strdup (resp, "Content-Type");
1638 char *tmp = strchr (type, ';');
1641 while (tmp > type && ISSPACE (tmp[-1]))
1646 hs->newloc = resp_header_strdup (resp, "Location");
1647 hs->remote_time = resp_header_strdup (resp, "Last-Modified");
1649 /* Handle (possibly multiple instances of) the Set-Cookie header. */
1652 const char *scbeg, *scend;
1653 /* The jar should have been created by now. */
1654 assert (wget_cookie_jar != NULL);
1656 (scpos = resp_header_locate (resp, "Set-Cookie", scpos,
1657 &scbeg, &scend)) != -1;
1660 char *set_cookie = strdupdelim (scbeg, scend);
1661 cookie_handle_set_cookie (wget_cookie_jar, u->host, u->port, u->path,
1667 if (resp_header_copy (resp, "Content-Range", hdrval, sizeof (hdrval)))
1669 wgint first_byte_pos, last_byte_pos, entity_length;
1670 if (parse_content_range (hdrval, &first_byte_pos, &last_byte_pos,
1672 contrange = first_byte_pos;
1677 /* 20x responses are counted among successful by default. */
1678 if (H_20X (statcode))
1681 /* Return if redirected. */
1682 if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES)
1684 /* RFC2068 says that in case of the 300 (multiple choices)
1685 response, the server can output a preferred URL through
1686 `Location' header; otherwise, the request should be treated
1687 like GET. So, if the location is set, it will be a
1688 redirection; otherwise, just proceed normally. */
1689 if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc)
1693 logprintf (LOG_VERBOSE,
1694 _("Location: %s%s\n"),
1695 hs->newloc ? escnonprint_uri (hs->newloc) : _("unspecified"),
1696 hs->newloc ? _(" [following]") : "");
1699 if (skip_short_body (sock, contlen))
1700 CLOSE_FINISH (sock);
1702 CLOSE_INVALIDATE (sock);
1709 /* If content-type is not given, assume text/html. This is because
1710 of the multitude of broken CGI's that "forget" to generate the
1713 0 == strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S)) ||
1714 0 == strncasecmp (type, TEXTXHTML_S, strlen (TEXTXHTML_S)))
1719 if (opt.html_extension && (*dt & TEXTHTML))
1720 /* -E / --html-extension / html_extension = on was specified, and this is a
1721 text/html file. If some case-insensitive variation on ".htm[l]" isn't
1722 already the file's suffix, tack on ".html". */
1724 char* last_period_in_local_filename = strrchr(*hs->local_file, '.');
1726 if (last_period_in_local_filename == NULL
1727 || !(0 == strcasecmp (last_period_in_local_filename, ".htm")
1728 || 0 == strcasecmp (last_period_in_local_filename, ".html")))
1730 size_t local_filename_len = strlen(*hs->local_file);
1732 *hs->local_file = xrealloc(*hs->local_file,
1733 local_filename_len + sizeof(".html"));
1734 strcpy(*hs->local_file + local_filename_len, ".html");
1736 *dt |= ADDED_HTML_EXTENSION;
1740 if (statcode == HTTP_STATUS_RANGE_NOT_SATISFIABLE)
1742 /* If `-c' is in use and the file has been fully downloaded (or
1743 the remote file has shrunk), Wget effectively requests bytes
1744 after the end of file and the server response with 416. */
1745 logputs (LOG_VERBOSE, _("\
1746 \n The file is already fully retrieved; nothing to do.\n\n"));
1747 /* In case the caller inspects. */
1750 /* Mark as successfully retrieved. */
1753 CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
1754 might be more bytes in the body. */
1755 return RETRUNNEEDED;
1757 if ((contrange != 0 && contrange != hs->restval)
1758 || (H_PARTIAL (statcode) && !contrange))
1760 /* The Range request was somehow misunderstood by the server.
1763 CLOSE_INVALIDATE (sock);
1766 hs->contlen = contlen + contrange;
1772 /* No need to print this output if the body won't be
1773 downloaded at all, or if the original server response is
1775 logputs (LOG_VERBOSE, _("Length: "));
1778 logputs (LOG_VERBOSE, with_thousand_seps (contlen + contrange));
1779 if (contlen + contrange >= 1024)
1780 logprintf (LOG_VERBOSE, " (%s)",
1781 human_readable (contlen + contrange));
1784 if (contlen >= 1024)
1785 logprintf (LOG_VERBOSE, _(", %s (%s) remaining"),
1786 with_thousand_seps (contlen),
1787 human_readable (contlen));
1789 logprintf (LOG_VERBOSE, _(", %s remaining"),
1790 with_thousand_seps (contlen));
1794 logputs (LOG_VERBOSE,
1795 opt.ignore_length ? _("ignored") : _("unspecified"));
1797 logprintf (LOG_VERBOSE, " [%s]\n", escnonprint (type));
1799 logputs (LOG_VERBOSE, "\n");
1803 type = NULL; /* We don't need it any more. */
1805 /* Return if we have no intention of further downloading. */
1806 if (!(*dt & RETROKF) || (*dt & HEAD_ONLY))
1808 /* In case the caller cares to look... */
1812 /* Pre-1.10 Wget used CLOSE_INVALIDATE here. Now we trust the
1813 servers not to send body in response to a HEAD request. If
1814 you encounter such a server (more likely a broken CGI), use
1815 `--no-http-keep-alive'. */
1816 CLOSE_FINISH (sock);
1817 return RETRFINISHED;
1820 /* Open the local file. */
1823 mkalldirs (*hs->local_file);
1825 rotate_backups (*hs->local_file);
1827 fp = fopen (*hs->local_file, "ab");
1828 else if (opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct
1829 || opt.output_document)
1830 fp = fopen (*hs->local_file, "wb");
1833 fp = fopen_excl (*hs->local_file, 1);
1834 if (!fp && errno == EEXIST)
1836 /* We cannot just invent a new name and use it (which is
1837 what functions like unique_create typically do)
1838 because we told the user we'd use this name.
1839 Instead, return and retry the download. */
1840 logprintf (LOG_NOTQUIET,
1841 _("%s has sprung into existence.\n"),
1843 CLOSE_INVALIDATE (sock);
1844 return FOPEN_EXCL_ERR;
1849 logprintf (LOG_NOTQUIET, "%s: %s\n", *hs->local_file, strerror (errno));
1850 CLOSE_INVALIDATE (sock);
1857 /* #### This confuses the timestamping code that checks for file
1858 size. Maybe we should save some additional information? */
1859 if (opt.save_headers)
1860 fwrite (head, 1, strlen (head), fp);
1862 /* Download the request body. */
1865 flags |= rb_read_exactly;
1866 if (hs->restval > 0 && contrange == 0)
1867 /* If the server ignored our range request, instruct fd_read_body
1868 to skip the first RESTVAL bytes of body. */
1869 flags |= rb_skip_startpos;
1870 hs->len = hs->restval;
1872 hs->res = fd_read_body (sock, fp, contlen != -1 ? contlen : 0,
1873 hs->restval, &hs->rd_size, &hs->len, &hs->dltime,
1877 CLOSE_FINISH (sock);
1879 CLOSE_INVALIDATE (sock);
1882 /* Close or flush the file. We have to be careful to check for
1883 error here. Checking the result of fwrite() is not enough --
1884 errors could go unnoticed! */
1887 flush_res = fclose (fp);
1889 flush_res = fflush (fp);
1890 if (flush_res == EOF)
1895 return RETRFINISHED;
1898 /* The genuine HTTP loop! This is the part where the retrieval is
1899 retried, and retried, and retried, and... */
1901 http_loop (struct url *u, char **newloc, char **local_file, const char *referer,
1902 int *dt, struct url *proxy)
1905 int use_ts, got_head = 0; /* time-stamping info */
1906 char *filename_plus_orig_suffix;
1907 char *local_filename = NULL;
1908 char *tms, *locf, *tmrate;
1910 time_t tml = -1, tmr = -1; /* local and remote time-stamps */
1911 wgint local_size = 0; /* the size of the local file */
1912 size_t filename_len;
1913 struct http_stat hstat; /* HTTP status */
1917 /* This used to be done in main(), but it's a better idea to do it
1918 here so that we don't go through the hoops if we're just using
1922 if (!wget_cookie_jar)
1923 wget_cookie_jar = cookie_jar_new ();
1924 if (opt.cookies_input && !cookies_loaded_p)
1926 cookie_jar_load (wget_cookie_jar, opt.cookies_input);
1927 cookies_loaded_p = 1;
1933 /* Warn on (likely bogus) wildcard usage in HTTP. Don't use
1934 has_wildcards_p because it would also warn on `?', and we know that
1935 shows up in CGI paths a *lot*. */
1936 if (strchr (u->url, '*'))
1937 logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
1941 /* Determine the local filename. */
1942 if (local_file && *local_file)
1943 hstat.local_file = local_file;
1944 else if (local_file && !opt.output_document)
1946 *local_file = url_file_name (u);
1947 hstat.local_file = local_file;
1951 dummy = url_file_name (u);
1952 hstat.local_file = &dummy;
1953 /* be honest about where we will save the file */
1954 if (local_file && opt.output_document)
1955 *local_file = HYPHENP (opt.output_document) ? NULL : xstrdup (opt.output_document);
1958 if (!opt.output_document)
1959 locf = *hstat.local_file;
1961 locf = opt.output_document;
1963 hstat.referer = referer;
1965 filename_len = strlen (*hstat.local_file);
1966 filename_plus_orig_suffix = alloca (filename_len + sizeof (".orig"));
1968 if (opt.noclobber && file_exists_p (*hstat.local_file))
1970 /* If opt.noclobber is turned on and file already exists, do not
1971 retrieve the file */
1972 logprintf (LOG_VERBOSE, _("\
1973 File `%s' already there, will not retrieve.\n"), *hstat.local_file);
1974 /* If the file is there, we suppose it's retrieved OK. */
1977 /* #### Bogusness alert. */
1978 /* If its suffix is "html" or "htm" or similar, assume text/html. */
1979 if (has_html_suffix_p (*hstat.local_file))
1987 if (opt.timestamping)
1989 int local_dot_orig_file_exists = 0;
1991 if (opt.backup_converted)
1992 /* If -K is specified, we'll act on the assumption that it was specified
1993 last time these files were downloaded as well, and instead of just
1994 comparing local file X against server file X, we'll compare local
1995 file X.orig (if extant, else X) against server file X. If -K
1996 _wasn't_ specified last time, or the server contains files called
1997 *.orig, -N will be back to not operating correctly with -k. */
1999 /* Would a single s[n]printf() call be faster? --dan
2001 Definitely not. sprintf() is horribly slow. It's a
2002 different question whether the difference between the two
2003 affects a program. Usually I'd say "no", but at one
2004 point I profiled Wget, and found that a measurable and
2005 non-negligible amount of time was lost calling sprintf()
2006 in url.c. Replacing sprintf with inline calls to
2007 strcpy() and number_to_string() made a difference.
2009 memcpy (filename_plus_orig_suffix, *hstat.local_file, filename_len);
2010 memcpy (filename_plus_orig_suffix + filename_len,
2011 ".orig", sizeof (".orig"));
2013 /* Try to stat() the .orig file. */
2014 if (stat (filename_plus_orig_suffix, &st) == 0)
2016 local_dot_orig_file_exists = 1;
2017 local_filename = filename_plus_orig_suffix;
2021 if (!local_dot_orig_file_exists)
2022 /* Couldn't stat() <file>.orig, so try to stat() <file>. */
2023 if (stat (*hstat.local_file, &st) == 0)
2024 local_filename = *hstat.local_file;
2026 if (local_filename != NULL)
2027 /* There was a local file, so we'll check later to see if the version
2028 the server has is the same version we already have, allowing us to
2034 /* Modification time granularity is 2 seconds for Windows, so
2035 increase local time by 1 second for later comparison. */
2038 local_size = st.st_size;
2042 /* Reset the counter. */
2048 /* Increment the pass counter. */
2050 sleep_between_retrievals (count);
2051 /* Get the current time string. */
2052 tms = time_str (NULL);
2053 /* Print fetch message, if opt.verbose. */
2056 char *hurl = url_string (u, 1);
2060 sprintf (tmp, _("(try:%2d)"), count);
2061 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
2062 tms, hurl, tmp, locf);
2064 ws_changetitle (hurl);
2069 /* Default document type is empty. However, if spider mode is
2070 on or time-stamping is employed, HEAD_ONLY commands is
2071 encoded within *dt. */
2072 if (opt.spider || (use_ts && !got_head))
2077 /* Decide whether or not to restart. */
2080 hstat.restval = hstat.len; /* continue where we left off */
2081 else if (opt.always_rest
2082 && stat (locf, &st) == 0
2083 && S_ISREG (st.st_mode))
2084 hstat.restval = st.st_size;
2086 /* Decide whether to send the no-cache directive. We send it in
2088 a) we're using a proxy, and we're past our first retrieval.
2089 Some proxies are notorious for caching incomplete data, so
2090 we require a fresh get.
2091 b) caching is explicitly inhibited. */
2092 if ((proxy && count > 1) /* a */
2093 || !opt.allow_cache /* b */
2095 *dt |= SEND_NOCACHE;
2097 *dt &= ~SEND_NOCACHE;
2099 /* Try fetching the document, or at least its head. */
2100 err = gethttp (u, &hstat, dt, proxy);
2102 /* It's unfortunate that wget determines the local filename before finding
2103 out the Content-Type of the file. Barring a major restructuring of the
2104 code, we need to re-set locf here, since gethttp() may have xrealloc()d
2105 *hstat.local_file to tack on ".html". */
2106 if (!opt.output_document)
2107 locf = *hstat.local_file;
2110 tms = time_str (NULL);
2111 /* Get the new location (with or without the redirection). */
2113 *newloc = xstrdup (hstat.newloc);
2116 case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
2117 case CONERROR: case READERR: case WRITEFAILED:
2118 case RANGEERR: case FOPEN_EXCL_ERR:
2119 /* Non-fatal errors continue executing the loop, which will
2120 bring them to "while" statement at the end, to judge
2121 whether the number of tries was exceeded. */
2122 free_hstat (&hstat);
2123 printwhat (count, opt.ntry);
2124 if (err == FOPEN_EXCL_ERR)
2126 /* Re-determine the file name. */
2127 if (local_file && *local_file)
2129 xfree (*local_file);
2130 *local_file = url_file_name (u);
2131 hstat.local_file = local_file;
2136 dummy = url_file_name (u);
2137 hstat.local_file = &dummy;
2139 /* be honest about where we will save the file */
2140 if (local_file && opt.output_document)
2141 *local_file = HYPHENP (opt.output_document) ? NULL : xstrdup (opt.output_document);
2142 if (!opt.output_document)
2143 locf = *hstat.local_file;
2145 locf = opt.output_document;
2149 case HOSTERR: case CONIMPOSSIBLE: case PROXERR: case AUTHFAILED:
2150 case SSLERRCTXCREATE: case CONTNOTSUPPORTED:
2151 /* Fatal errors just return from the function. */
2152 free_hstat (&hstat);
2156 case FWRITEERR: case FOPENERR:
2157 /* Another fatal error. */
2158 logputs (LOG_VERBOSE, "\n");
2159 logprintf (LOG_NOTQUIET, _("Cannot write to `%s' (%s).\n"),
2160 *hstat.local_file, strerror (errno));
2161 free_hstat (&hstat);
2166 /* Another fatal error. */
2167 logputs (LOG_VERBOSE, "\n");
2168 logprintf (LOG_NOTQUIET, _("Unable to establish SSL connection.\n"));
2169 free_hstat (&hstat);
2174 /* Return the new location to the caller. */
2177 logprintf (LOG_NOTQUIET,
2178 _("ERROR: Redirection (%d) without location.\n"),
2180 free_hstat (&hstat);
2184 free_hstat (&hstat);
2189 /* The file was already fully retrieved. */
2190 free_hstat (&hstat);
2195 /* Deal with you later. */
2198 /* All possibilities should have been exhausted. */
2201 if (!(*dt & RETROKF))
2205 /* #### Ugly ugly ugly! */
2206 char *hurl = url_string (u, 1);
2207 logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
2210 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
2211 tms, hstat.statcode, escnonprint (hstat.error));
2212 logputs (LOG_VERBOSE, "\n");
2213 free_hstat (&hstat);
2218 /* Did we get the time-stamp? */
2221 if (opt.timestamping && !hstat.remote_time)
2223 logputs (LOG_NOTQUIET, _("\
2224 Last-modified header missing -- time-stamps turned off.\n"));
2226 else if (hstat.remote_time)
2228 /* Convert the date-string into struct tm. */
2229 tmr = http_atotm (hstat.remote_time);
2230 if (tmr == (time_t) (-1))
2231 logputs (LOG_VERBOSE, _("\
2232 Last-modified header invalid -- time-stamp ignored.\n"));
2236 /* The time-stamping section. */
2241 use_ts = 0; /* no more time-stamping */
2242 count = 0; /* the retrieve count for HEAD is
2244 if (hstat.remote_time && tmr != (time_t) (-1))
2246 /* Now time-stamping can be used validly. Time-stamping
2247 means that if the sizes of the local and remote file
2248 match, and local file is newer than the remote file,
2249 it will not be retrieved. Otherwise, the normal
2250 download procedure is resumed. */
2252 (hstat.contlen == -1 || local_size == hstat.contlen))
2254 logprintf (LOG_VERBOSE, _("\
2255 Server file no newer than local file `%s' -- not retrieving.\n\n"),
2257 free_hstat (&hstat);
2261 else if (tml >= tmr)
2262 logprintf (LOG_VERBOSE, _("\
2263 The sizes do not match (local %s) -- retrieving.\n"),
2264 number_to_static_string (local_size));
2266 logputs (LOG_VERBOSE,
2267 _("Remote file is newer, retrieving.\n"));
2269 free_hstat (&hstat);
2272 if ((tmr != (time_t) (-1))
2274 && ((hstat.len == hstat.contlen) ||
2275 ((hstat.res == 0) &&
2276 ((hstat.contlen == -1) ||
2277 (hstat.len >= hstat.contlen && !opt.kill_longer)))))
2279 /* #### This code repeats in http.c and ftp.c. Move it to a
2281 const char *fl = NULL;
2282 if (opt.output_document)
2284 if (output_stream_regular)
2285 fl = opt.output_document;
2288 fl = *hstat.local_file;
2292 /* End of time-stamping section. */
2296 logprintf (LOG_NOTQUIET, "%d %s\n\n", hstat.statcode,
2297 escnonprint (hstat.error));
2302 tmrate = retr_rate (hstat.rd_size, hstat.dltime, 0);
2304 if (hstat.len == hstat.contlen)
2308 logprintf (LOG_VERBOSE,
2309 _("%s (%s) - `%s' saved [%s/%s]\n\n"),
2311 number_to_static_string (hstat.len),
2312 number_to_static_string (hstat.contlen));
2313 logprintf (LOG_NONVERBOSE,
2314 "%s URL:%s [%s/%s] -> \"%s\" [%d]\n",
2316 number_to_static_string (hstat.len),
2317 number_to_static_string (hstat.contlen),
2321 total_downloaded_bytes += hstat.len;
2323 /* Remember that we downloaded the file for later ".orig" code. */
2324 if (*dt & ADDED_HTML_EXTENSION)
2325 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
2327 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
2329 free_hstat (&hstat);
2333 else if (hstat.res == 0) /* No read error */
2335 if (hstat.contlen == -1) /* We don't know how much we were supposed
2336 to get, so assume we succeeded. */
2340 logprintf (LOG_VERBOSE,
2341 _("%s (%s) - `%s' saved [%s]\n\n"),
2343 number_to_static_string (hstat.len));
2344 logprintf (LOG_NONVERBOSE,
2345 "%s URL:%s [%s] -> \"%s\" [%d]\n",
2346 tms, u->url, number_to_static_string (hstat.len),
2350 total_downloaded_bytes += hstat.len;
2352 /* Remember that we downloaded the file for later ".orig" code. */
2353 if (*dt & ADDED_HTML_EXTENSION)
2354 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
2356 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
2358 free_hstat (&hstat);
2362 else if (hstat.len < hstat.contlen) /* meaning we lost the
2363 connection too soon */
2365 logprintf (LOG_VERBOSE,
2366 _("%s (%s) - Connection closed at byte %s. "),
2367 tms, tmrate, number_to_static_string (hstat.len));
2368 printwhat (count, opt.ntry);
2369 free_hstat (&hstat);
2372 else if (!opt.kill_longer) /* meaning we got more than expected */
2374 logprintf (LOG_VERBOSE,
2375 _("%s (%s) - `%s' saved [%s/%s])\n\n"),
2377 number_to_static_string (hstat.len),
2378 number_to_static_string (hstat.contlen));
2379 logprintf (LOG_NONVERBOSE,
2380 "%s URL:%s [%s/%s] -> \"%s\" [%d]\n",
2382 number_to_static_string (hstat.len),
2383 number_to_static_string (hstat.contlen),
2386 total_downloaded_bytes += hstat.len;
2388 /* Remember that we downloaded the file for later ".orig" code. */
2389 if (*dt & ADDED_HTML_EXTENSION)
2390 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
2392 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
2394 free_hstat (&hstat);
2398 else /* the same, but not accepted */
2400 logprintf (LOG_VERBOSE,
2401 _("%s (%s) - Connection closed at byte %s/%s. "),
2403 number_to_static_string (hstat.len),
2404 number_to_static_string (hstat.contlen));
2405 printwhat (count, opt.ntry);
2406 free_hstat (&hstat);
2410 else /* now hstat.res can only be -1 */
2412 if (hstat.contlen == -1)
2414 logprintf (LOG_VERBOSE,
2415 _("%s (%s) - Read error at byte %s (%s)."),
2416 tms, tmrate, number_to_static_string (hstat.len),
2418 printwhat (count, opt.ntry);
2419 free_hstat (&hstat);
2422 else /* hstat.res == -1 and contlen is given */
2424 logprintf (LOG_VERBOSE,
2425 _("%s (%s) - Read error at byte %s/%s (%s). "),
2427 number_to_static_string (hstat.len),
2428 number_to_static_string (hstat.contlen),
2430 printwhat (count, opt.ntry);
2431 free_hstat (&hstat);
2438 while (!opt.ntry || (count < opt.ntry));
2442 /* Converts struct tm to time_t, assuming the data in tm is UTC rather
2443 than local timezone.
2445 mktime is similar but assumes struct tm, also known as the
2446 "broken-down" form of time, is in local time zone. mktime_from_utc
2447 uses mktime to make the conversion understanding that an offset
2448 will be introduced by the local time assumption.
2450 mktime_from_utc then measures the introduced offset by applying
2451 gmtime to the initial result and applying mktime to the resulting
2452 "broken-down" form. The difference between the two mktime results
2453 is the measured offset which is then subtracted from the initial
2454 mktime result to yield a calendar time which is the value returned.
2456 tm_isdst in struct tm is set to 0 to force mktime to introduce a
2457 consistent offset (the non DST offset) since tm and tm+o might be
2458 on opposite sides of a DST change.
2460 Some implementations of mktime return -1 for the nonexistent
2461 localtime hour at the beginning of DST. In this event, use
2462 mktime(tm - 1hr) + 3600.
2466 gmtime(t+o) --> tm+o
2467 mktime(tm+o) --> t+2o
2468 t+o - (t+2o - t+o) = t
2470 Note that glibc contains a function of the same purpose named
2471 `timegm' (reverse of gmtime). But obviously, it is not universally
2472 available, and unfortunately it is not straightforwardly
2473 extractable for use here. Perhaps configure should detect timegm
2474 and use it where available.
2476 Contributed by Roger Beeman <beeman@cisco.com>, with the help of
2477 Mark Baushke <mdb@cisco.com> and the rest of the Gurus at CISCO.
2478 Further improved by Roger with assistance from Edward J. Sabol
2479 based on input by Jamie Zawinski. */
2482 mktime_from_utc (struct tm *t)
2493 return -1; /* can't deal with output from strptime */
2504 return -1; /* can't deal with output from gmtime */
2507 return (tl - (tb - tl));
2510 /* Check whether the result of strptime() indicates success.
2511 strptime() returns the pointer to how far it got to in the string.
2512 The processing has been successful if the string is at `GMT' or
2513 `+X', or at the end of the string.
2515 In extended regexp parlance, the function returns 1 if P matches
2516 "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (which strptime
2517 can return) is considered a failure and 0 is returned. */
2519 check_end (const char *p)
2523 while (ISSPACE (*p))
2526 || (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
2527 || ((p[0] == '+' || p[0] == '-') && ISDIGIT (p[1])))
2533 /* Convert the textual specification of time in TIME_STRING to the
2534 number of seconds since the Epoch.
2536 TIME_STRING can be in any of the three formats RFC2068 allows the
2537 HTTP servers to emit -- RFC1123-date, RFC850-date or asctime-date.
2538 Timezones are ignored, and should be GMT.
2540 Return the computed time_t representation, or -1 if the conversion
2543 This function uses strptime with various string formats for parsing
2544 TIME_STRING. This results in a parser that is not as lenient in
2545 interpreting TIME_STRING as I would like it to be. Being based on
2546 strptime, it always allows shortened months, one-digit days, etc.,
2547 but due to the multitude of formats in which time can be
2548 represented, an ideal HTTP time parser would be even more
2549 forgiving. It should completely ignore things like week days and
2550 concentrate only on the various forms of representing years,
2551 months, days, hours, minutes, and seconds. For example, it would
2552 be nice if it accepted ISO 8601 out of the box.
2554 I've investigated free and PD code for this purpose, but none was
2555 usable. getdate was big and unwieldy, and had potential copyright
2556 issues, or so I was informed. Dr. Marcus Hennecke's atotm(),
2557 distributed with phttpd, is excellent, but we cannot use it because
2558 it is not assigned to the FSF. So I stuck it with strptime. */
2561 http_atotm (const char *time_string)
2563 /* NOTE: Solaris strptime man page claims that %n and %t match white
2564 space, but that's not universally available. Instead, we simply
2565 use ` ' to mean "skip all WS", which works under all strptime
2566 implementations I've tested. */
2568 static const char *time_formats[] = {
2569 "%a, %d %b %Y %T", /* RFC1123: Thu, 29 Jan 1998 22:12:57 */
2570 "%A, %d-%b-%y %T", /* RFC850: Thursday, 29-Jan-98 22:12:57 */
2571 "%a, %d-%b-%Y %T", /* pseudo-RFC850: Thu, 29-Jan-1998 22:12:57
2572 (google.com uses this for their cookies.) */
2573 "%a %b %d %T %Y" /* asctime: Thu Jan 29 22:12:57 1998 */
2579 /* According to Roger Beeman, we need to initialize tm_isdst, since
2580 strptime won't do it. */
2583 /* Note that under foreign locales Solaris strptime() fails to
2584 recognize English dates, which renders this function useless. We
2585 solve this by being careful not to affect LC_TIME when
2586 initializing locale.
2588 Another solution would be to temporarily set locale to C, invoke
2589 strptime(), and restore it back. This is slow and dirty,
2590 however, and locale support other than LC_MESSAGES can mess other
2591 things, so I rather chose to stick with just setting LC_MESSAGES.
2593 GNU strptime does not have this problem because it recognizes
2594 both international and local dates. */
2596 for (i = 0; i < countof (time_formats); i++)
2597 if (check_end (strptime (time_string, time_formats[i], &t)))
2598 return mktime_from_utc (&t);
2600 /* All formats have failed. */
2604 /* Authorization support: We support two authorization schemes:
2606 * `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
2608 * `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
2609 consisting of answering to the server's challenge with the proper
2612 /* Create the authentication header contents for the `Basic' scheme.
2613 This is done by encoding the string `USER:PASS' in base64 and
2614 prepending `HEADER: Basic ' to it. */
2616 basic_authentication_encode (const char *user, const char *passwd)
2619 int len1 = strlen (user) + 1 + strlen (passwd);
2620 int len2 = BASE64_LENGTH (len1);
2622 t1 = (char *)alloca (len1 + 1);
2623 sprintf (t1, "%s:%s", user, passwd);
2625 t2 = (char *)alloca (len2 + 1);
2626 base64_encode (t1, t2, len1);
2628 return concat_strings ("Basic ", t2, (char *) 0);
2631 #define SKIP_WS(x) do { \
2632 while (ISSPACE (*(x))) \
2636 #ifdef ENABLE_DIGEST
2637 /* Parse HTTP `WWW-Authenticate:' header. AU points to the beginning
2638 of a field in such a header. If the field is the one specified by
2639 ATTR_NAME ("realm", "opaque", and "nonce" are used by the current
2640 digest authorization code), extract its value in the (char*)
2641 variable pointed by RET. Returns negative on a malformed header,
2642 or number of bytes that have been parsed by this call. */
2644 extract_header_attr (const char *au, const char *attr_name, char **ret)
2647 const char *cp = au;
2649 if (strncmp (cp, attr_name, strlen (attr_name)) == 0)
2651 cp += strlen (attr_name);
2664 for (ep = cp; *ep && *ep != '\"'; ep++)
2669 *ret = strdupdelim (cp, ep);
2676 /* Dump the hexadecimal representation of HASH to BUF. HASH should be
2677 an array of 16 bytes containing the hash keys, and BUF should be a
2678 buffer of 33 writable characters (32 for hex digits plus one for
2679 zero termination). */
2681 dump_hash (unsigned char *buf, const unsigned char *hash)
2685 for (i = 0; i < MD5_HASHLEN; i++, hash++)
2687 *buf++ = XNUM_TO_digit (*hash >> 4);
2688 *buf++ = XNUM_TO_digit (*hash & 0xf);
2693 /* Take the line apart to find the challenge, and compose a digest
2694 authorization header. See RFC2069 section 2.1.2. */
2696 digest_authentication_encode (const char *au, const char *user,
2697 const char *passwd, const char *method,
2700 static char *realm, *opaque, *nonce;
2705 { "realm", &realm },
2706 { "opaque", &opaque },
2711 realm = opaque = nonce = NULL;
2713 au += 6; /* skip over `Digest' */
2719 for (i = 0; i < countof (options); i++)
2721 int skip = extract_header_attr (au, options[i].name,
2722 options[i].variable);
2726 xfree_null (opaque);
2736 if (i == countof (options))
2738 while (*au && *au != '=')
2746 while (*au && *au != '\"')
2753 while (*au && *au != ',')
2758 if (!realm || !nonce || !user || !passwd || !path || !method)
2761 xfree_null (opaque);
2766 /* Calculate the digest value. */
2768 ALLOCA_MD5_CONTEXT (ctx);
2769 unsigned char hash[MD5_HASHLEN];
2770 unsigned char a1buf[MD5_HASHLEN * 2 + 1], a2buf[MD5_HASHLEN * 2 + 1];
2771 unsigned char response_digest[MD5_HASHLEN * 2 + 1];
2773 /* A1BUF = H(user ":" realm ":" password) */
2775 gen_md5_update ((unsigned char *)user, strlen (user), ctx);
2776 gen_md5_update ((unsigned char *)":", 1, ctx);
2777 gen_md5_update ((unsigned char *)realm, strlen (realm), ctx);
2778 gen_md5_update ((unsigned char *)":", 1, ctx);
2779 gen_md5_update ((unsigned char *)passwd, strlen (passwd), ctx);
2780 gen_md5_finish (ctx, hash);
2781 dump_hash (a1buf, hash);
2783 /* A2BUF = H(method ":" path) */
2785 gen_md5_update ((unsigned char *)method, strlen (method), ctx);
2786 gen_md5_update ((unsigned char *)":", 1, ctx);
2787 gen_md5_update ((unsigned char *)path, strlen (path), ctx);
2788 gen_md5_finish (ctx, hash);
2789 dump_hash (a2buf, hash);
2791 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */
2793 gen_md5_update (a1buf, MD5_HASHLEN * 2, ctx);
2794 gen_md5_update ((unsigned char *)":", 1, ctx);
2795 gen_md5_update ((unsigned char *)nonce, strlen (nonce), ctx);
2796 gen_md5_update ((unsigned char *)":", 1, ctx);
2797 gen_md5_update (a2buf, MD5_HASHLEN * 2, ctx);
2798 gen_md5_finish (ctx, hash);
2799 dump_hash (response_digest, hash);
2801 res = (char*) xmalloc (strlen (user)
2806 + 2 * MD5_HASHLEN /*strlen (response_digest)*/
2807 + (opaque ? strlen (opaque) : 0)
2809 sprintf (res, "Digest \
2810 username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
2811 user, realm, nonce, path, response_digest);
2814 char *p = res + strlen (res);
2815 strcat (p, ", opaque=\"");
2822 #endif /* ENABLE_DIGEST */
2825 #define BEGINS_WITH(line, string_constant) \
2826 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
2827 && (ISSPACE (line[sizeof (string_constant) - 1]) \
2828 || !line[sizeof (string_constant) - 1]))
2831 known_authentication_scheme_p (const char *au)
2833 return BEGINS_WITH (au, "Basic")
2834 #ifdef ENABLE_DIGEST
2835 || BEGINS_WITH (au, "Digest")
2838 || BEGINS_WITH (au, "NTLM")
2845 /* Create the HTTP authorization request header. When the
2846 `WWW-Authenticate' response header is seen, according to the
2847 authorization scheme specified in that header (`Basic' and `Digest'
2848 are supported by the current implementation), produce an
2849 appropriate HTTP authorization request header. */
2851 create_authorization_line (const char *au, const char *user,
2852 const char *passwd, const char *method,
2855 if (0 == strncasecmp (au, "Basic", 5))
2856 return basic_authentication_encode (user, passwd);
2857 #ifdef ENABLE_DIGEST
2858 if (0 == strncasecmp (au, "Digest", 6))
2859 return digest_authentication_encode (au, user, passwd, method, path);
2862 if (0 == strncasecmp (au, "NTLM", 4))
2864 int ok = ntlm_input (&pconn.ntlm, au);
2867 /* #### we shouldn't ignore the OK that ntlm_output returns. */
2868 return ntlm_output (&pconn.ntlm, user, passwd, &ok);
2877 if (wget_cookie_jar)
2878 cookie_jar_save (wget_cookie_jar, opt.cookies_output);
2884 xfree_null (pconn.host);
2885 if (wget_cookie_jar)
2886 cookie_jar_delete (wget_cookie_jar);