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"
75 extern char *version_string;
76 extern LARGE_INT total_downloaded_bytes;
78 extern FILE *output_stream;
79 extern int output_stream_regular;
82 # define MIN(x, y) ((x) > (y) ? (y) : (x))
86 static int cookies_loaded_p;
87 static struct cookie_jar *wget_cookie_jar;
89 #define TEXTHTML_S "text/html"
90 #define TEXTXHTML_S "application/xhtml+xml"
92 /* Some status code validation macros: */
93 #define H_20X(x) (((x) >= 200) && ((x) < 300))
94 #define H_PARTIAL(x) ((x) == HTTP_STATUS_PARTIAL_CONTENTS)
95 #define H_REDIRECTED(x) ((x) == HTTP_STATUS_MOVED_PERMANENTLY \
96 || (x) == HTTP_STATUS_MOVED_TEMPORARILY \
97 || (x) == HTTP_STATUS_SEE_OTHER \
98 || (x) == HTTP_STATUS_TEMPORARY_REDIRECT)
100 /* HTTP/1.0 status codes from RFC1945, provided for reference. */
101 /* Successful 2xx. */
102 #define HTTP_STATUS_OK 200
103 #define HTTP_STATUS_CREATED 201
104 #define HTTP_STATUS_ACCEPTED 202
105 #define HTTP_STATUS_NO_CONTENT 204
106 #define HTTP_STATUS_PARTIAL_CONTENTS 206
108 /* Redirection 3xx. */
109 #define HTTP_STATUS_MULTIPLE_CHOICES 300
110 #define HTTP_STATUS_MOVED_PERMANENTLY 301
111 #define HTTP_STATUS_MOVED_TEMPORARILY 302
112 #define HTTP_STATUS_SEE_OTHER 303 /* from HTTP/1.1 */
113 #define HTTP_STATUS_NOT_MODIFIED 304
114 #define HTTP_STATUS_TEMPORARY_REDIRECT 307 /* from HTTP/1.1 */
116 /* Client error 4xx. */
117 #define HTTP_STATUS_BAD_REQUEST 400
118 #define HTTP_STATUS_UNAUTHORIZED 401
119 #define HTTP_STATUS_FORBIDDEN 403
120 #define HTTP_STATUS_NOT_FOUND 404
121 #define HTTP_STATUS_RANGE_NOT_SATISFIABLE 416
123 /* Server errors 5xx. */
124 #define HTTP_STATUS_INTERNAL 500
125 #define HTTP_STATUS_NOT_IMPLEMENTED 501
126 #define HTTP_STATUS_BAD_GATEWAY 502
127 #define HTTP_STATUS_UNAVAILABLE 503
130 rel_none, rel_name, rel_value, rel_both
137 struct request_header {
139 enum rp release_policy;
141 int hcount, hcapacity;
144 /* Create a new, empty request. At least request_set_method must be
145 called before the request can be used. */
147 static struct request *
150 struct request *req = xnew0 (struct request);
152 req->headers = xnew_array (struct request_header, req->hcapacity);
156 /* Set the request's method and its arguments. METH should be a
157 literal string (or it should outlive the request) because it will
158 not be freed. ARG will be freed by request_free. */
161 request_set_method (struct request *req, const char *meth, char *arg)
167 /* Return the method string passed with the last call to
168 request_set_method. */
171 request_method (const struct request *req)
176 /* Free one header according to the release policy specified with
177 request_set_header. */
180 release_header (struct request_header *hdr)
182 switch (hdr->release_policy)
199 /* Set the request named NAME to VALUE. Specifically, this means that
200 a "NAME: VALUE\r\n" header line will be used in the request. If a
201 header with the same name previously existed in the request, its
202 value will be replaced by this one.
204 RELEASE_POLICY determines whether NAME and VALUE should be released
205 (freed) with request_free. Allowed values are:
207 - rel_none - don't free NAME or VALUE
208 - rel_name - free NAME when done
209 - rel_value - free VALUE when done
210 - rel_both - free both NAME and VALUE when done
212 Setting release policy is useful when arguments come from different
213 sources. For example:
215 // Don't free literal strings!
216 request_set_header (req, "Pragma", "no-cache", rel_none);
218 // Don't free a global variable, we'll need it later.
219 request_set_header (req, "Referer", opt.referer, rel_none);
221 // Value freshly allocated, free it when done.
222 request_set_header (req, "Range",
223 aprintf ("bytes=%s-", number_to_static_string (hs->restval)),
228 request_set_header (struct request *req, char *name, char *value,
229 enum rp release_policy)
231 struct request_header *hdr;
235 for (i = 0; i < req->hcount; i++)
237 hdr = &req->headers[i];
238 if (0 == strcasecmp (name, hdr->name))
240 /* Replace existing header. */
241 release_header (hdr);
244 hdr->release_policy = release_policy;
249 /* Install new header. */
251 if (req->hcount >= req->hcount)
253 req->hcapacity <<= 1;
254 req->headers = xrealloc (req->headers,
255 req->hcapacity * sizeof (struct request_header));
257 hdr = &req->headers[req->hcount++];
260 hdr->release_policy = release_policy;
263 /* Like request_set_header, but sets the whole header line, as
264 provided by the user using the `--header' option. For example,
265 request_set_user_header (req, "Foo: bar") works just like
266 request_set_header (req, "Foo", "bar"). */
269 request_set_user_header (struct request *req, const char *header)
272 const char *p = strchr (header, ':');
275 BOUNDED_TO_ALLOCA (header, p, name);
279 request_set_header (req, xstrdup (name), (char *) p, rel_name);
282 #define APPEND(p, str) do { \
283 int A_len = strlen (str); \
284 memcpy (p, str, A_len); \
288 /* Construct the request and write it to FD using fd_write. */
291 request_send (const struct request *req, int fd)
293 char *request_string, *p;
294 int i, size, write_error;
296 /* Count the request size. */
299 /* METHOD " " ARG " " "HTTP/1.0" "\r\n" */
300 size += strlen (req->method) + 1 + strlen (req->arg) + 1 + 8 + 2;
302 for (i = 0; i < req->hcount; i++)
304 struct request_header *hdr = &req->headers[i];
305 /* NAME ": " VALUE "\r\n" */
306 size += strlen (hdr->name) + 2 + strlen (hdr->value) + 2;
312 p = request_string = alloca_array (char, size);
314 /* Generate the request. */
316 APPEND (p, req->method); *p++ = ' ';
317 APPEND (p, req->arg); *p++ = ' ';
318 memcpy (p, "HTTP/1.0\r\n", 10); p += 10;
320 for (i = 0; i < req->hcount; i++)
322 struct request_header *hdr = &req->headers[i];
323 APPEND (p, hdr->name);
324 *p++ = ':', *p++ = ' ';
325 APPEND (p, hdr->value);
326 *p++ = '\r', *p++ = '\n';
329 *p++ = '\r', *p++ = '\n', *p++ = '\0';
330 assert (p - request_string == size);
334 DEBUGP (("\n---request begin---\n%s---request end---\n", request_string));
336 /* Send the request to the server. */
338 write_error = fd_write (fd, request_string, size - 1, -1);
340 logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"),
345 /* Release the resources used by REQ. */
348 request_free (struct request *req)
351 xfree_null (req->arg);
352 for (i = 0; i < req->hcount; i++)
353 release_header (&req->headers[i]);
354 xfree_null (req->headers);
358 /* Send the contents of FILE_NAME to SOCK/SSL. Make sure that exactly
359 PROMISED_SIZE bytes are sent over the wire -- if the file is
360 longer, read only that much; if the file is shorter, report an error. */
363 post_file (int sock, const char *file_name, wgint promised_size)
365 static char chunk[8192];
370 DEBUGP (("[writing POST file %s ... ", file_name));
372 fp = fopen (file_name, "rb");
375 while (!feof (fp) && written < promised_size)
378 int length = fread (chunk, 1, sizeof (chunk), fp);
381 towrite = MIN (promised_size - written, length);
382 write_error = fd_write (sock, chunk, towrite, -1);
392 /* If we've written less than was promised, report a (probably
393 nonsensical) error rather than break the promise. */
394 if (written < promised_size)
400 assert (written == promised_size);
401 DEBUGP (("done]\n"));
406 response_head_terminator (const char *hunk, int oldlen, int peeklen)
408 const char *start, *end;
410 /* If at first peek, verify whether HUNK starts with "HTTP". If
411 not, this is a HTTP/0.9 request and we must bail out without
413 if (oldlen == 0 && 0 != memcmp (hunk, "HTTP", MIN (peeklen, 4)))
419 start = hunk + oldlen - 4;
420 end = hunk + oldlen + peeklen;
422 for (; start < end - 1; start++)
429 if (start[1] == '\n')
435 /* The maximum size of a single HTTP response we care to read. This
436 is not meant to impose an arbitrary limit, but to protect the user
437 from Wget slurping up available memory upon encountering malicious
438 or buggy server output. Define it to 0 to remove the limit. */
440 #define HTTP_RESPONSE_MAX_SIZE 65536
442 /* Read the HTTP request head from FD and return it. The error
443 conditions are the same as with fd_read_hunk.
445 To support HTTP/0.9 responses, this function tries to make sure
446 that the data begins with "HTTP". If this is not the case, no data
447 is read and an empty request is returned, so that the remaining
448 data can be treated as body. */
451 read_http_response_head (int fd)
453 return fd_read_hunk (fd, response_head_terminator, 512,
454 HTTP_RESPONSE_MAX_SIZE);
458 /* The response data. */
461 /* The array of pointers that indicate where each header starts.
462 For example, given this HTTP response:
469 The headers are located like this:
471 "HTTP/1.0 200 Ok\r\nDescription: some\r\n text\r\nEtag: x\r\n\r\n"
473 headers[0] headers[1] headers[2] headers[3]
475 I.e. headers[0] points to the beginning of the request,
476 headers[1] points to the end of the first header and the
477 beginning of the second one, etc. */
479 const char **headers;
482 /* Create a new response object from the text of the HTTP response,
483 available in HEAD. That text is automatically split into
484 constituent header lines for fast retrieval using
487 static struct response *
488 resp_new (const char *head)
493 struct response *resp = xnew0 (struct response);
498 /* Empty head means that we're dealing with a headerless
499 (HTTP/0.9) response. In that case, don't set HEADERS at
504 /* Split HEAD into header lines, so that resp_header_* functions
505 don't need to do this over and over again. */
511 DO_REALLOC (resp->headers, size, count + 1, const char *);
512 resp->headers[count++] = hdr;
514 /* Break upon encountering an empty line. */
515 if (!hdr[0] || (hdr[0] == '\r' && hdr[1] == '\n') || hdr[0] == '\n')
518 /* Find the end of HDR, including continuations. */
521 const char *end = strchr (hdr, '\n');
527 while (*hdr == ' ' || *hdr == '\t');
529 DO_REALLOC (resp->headers, size, count + 1, const char *);
530 resp->headers[count] = NULL;
535 /* Locate the header named NAME in the request data, starting with
536 position START. This allows the code to loop through the request
537 data, filtering for all requests of a given name. Returns the
538 found position, or -1 for failure. The code that uses this
539 function typically looks like this:
541 for (pos = 0; (pos = resp_header_locate (...)) != -1; pos++)
542 ... do something with header ...
544 If you only care about one header, use resp_header_get instead of
548 resp_header_locate (const struct response *resp, const char *name, int start,
549 const char **begptr, const char **endptr)
552 const char **headers = resp->headers;
555 if (!headers || !headers[1])
558 name_len = strlen (name);
564 for (; headers[i + 1]; i++)
566 const char *b = headers[i];
567 const char *e = headers[i + 1];
569 && b[name_len] == ':'
570 && 0 == strncasecmp (b, name, name_len))
573 while (b < e && ISSPACE (*b))
575 while (b < e && ISSPACE (e[-1]))
585 /* Find and retrieve the header named NAME in the request data. If
586 found, set *BEGPTR to its starting, and *ENDPTR to its ending
587 position, and return 1. Otherwise return 0.
589 This function is used as a building block for resp_header_copy
590 and resp_header_strdup. */
593 resp_header_get (const struct response *resp, const char *name,
594 const char **begptr, const char **endptr)
596 int pos = resp_header_locate (resp, name, 0, begptr, endptr);
600 /* Copy the response header named NAME to buffer BUF, no longer than
601 BUFSIZE (BUFSIZE includes the terminating 0). If the header
602 exists, 1 is returned, otherwise 0. If there should be no limit on
603 the size of the header, use resp_header_strdup instead.
605 If BUFSIZE is 0, no data is copied, but the boolean indication of
606 whether the header is present is still returned. */
609 resp_header_copy (const struct response *resp, const char *name,
610 char *buf, int bufsize)
613 if (!resp_header_get (resp, name, &b, &e))
617 int len = MIN (e - b, bufsize - 1);
618 memcpy (buf, b, len);
624 /* Return the value of header named NAME in RESP, allocated with
625 malloc. If such a header does not exist in RESP, return NULL. */
628 resp_header_strdup (const struct response *resp, const char *name)
631 if (!resp_header_get (resp, name, &b, &e))
633 return strdupdelim (b, e);
636 /* Parse the HTTP status line, which is of format:
638 HTTP-Version SP Status-Code SP Reason-Phrase
640 The function returns the status-code, or -1 if the status line
641 appears malformed. The pointer to "reason-phrase" message is
642 returned in *MESSAGE. */
645 resp_status (const struct response *resp, char **message)
652 /* For a HTTP/0.9 response, assume status 200. */
654 *message = xstrdup (_("No headers, assuming HTTP/0.9"));
658 p = resp->headers[0];
659 end = resp->headers[1];
665 if (end - p < 4 || 0 != strncmp (p, "HTTP", 4))
669 /* Match the HTTP version. This is optional because Gnutella
670 servers have been reported to not specify HTTP version. */
671 if (p < end && *p == '/')
674 while (p < end && ISDIGIT (*p))
676 if (p < end && *p == '.')
678 while (p < end && ISDIGIT (*p))
682 while (p < end && ISSPACE (*p))
684 if (end - p < 3 || !ISDIGIT (p[0]) || !ISDIGIT (p[1]) || !ISDIGIT (p[2]))
687 status = 100 * (p[0] - '0') + 10 * (p[1] - '0') + (p[2] - '0');
692 while (p < end && ISSPACE (*p))
694 while (p < end && ISSPACE (end[-1]))
696 *message = strdupdelim (p, end);
702 /* Release the resources used by RESP. */
705 resp_free (struct response *resp)
707 xfree_null (resp->headers);
711 /* Print [b, e) to the log, omitting the trailing CRLF. */
714 print_server_response_1 (const char *prefix, const char *b, const char *e)
717 if (b < e && e[-1] == '\n')
719 if (b < e && e[-1] == '\r')
721 BOUNDED_TO_ALLOCA (b, e, ln);
722 logprintf (LOG_VERBOSE, "%s%s\n", prefix, escnonprint (ln));
725 /* Print the server response, line by line, omitting the trailing CR
726 characters, prefixed with PREFIX. */
729 print_server_response (const struct response *resp, const char *prefix)
734 for (i = 0; resp->headers[i + 1]; i++)
735 print_server_response_1 (prefix, resp->headers[i], resp->headers[i + 1]);
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 error
783 responses whose bodies don't need to be displayed or logged, but
784 which need to be read anyway. */
787 skip_short_body (int fd, wgint contlen)
789 /* Skipping the body doesn't make sense if the content length is
790 unknown because, in that case, persistent connections cannot be
791 used. (#### This is not the case with HTTP/1.1 where they can
792 still be used with the magic of the "chunked" transfer!) */
795 DEBUGP (("Skipping %s bytes of body data... ", number_to_static_string (contlen)));
800 int ret = fd_read (fd, dlbuf, MIN (contlen, sizeof (dlbuf)), -1);
805 DEBUGP (("done.\n"));
808 /* Persistent connections. Currently, we cache the most recently used
809 connection as persistent, provided that the HTTP server agrees to
810 make it such. The persistence data is stored in the variables
811 below. Ideally, it should be possible to cache an arbitrary fixed
812 number of these connections. */
814 /* Whether a persistent connection is active. */
815 static int pconn_active;
818 /* The socket of the connection. */
821 /* Host and port of the currently active persistent connection. */
825 /* Whether a ssl handshake has occoured on this connection. */
829 /* Mark the persistent connection as invalid and free the resources it
830 uses. This is used by the CLOSE_* macros after they forcefully
831 close a registered persistent connection. */
834 invalidate_persistent (void)
836 DEBUGP (("Disabling further reuse of socket %d.\n", pconn.socket));
838 fd_close (pconn.socket);
843 /* Register FD, which should be a TCP/IP connection to HOST:PORT, as
844 persistent. This will enable someone to use the same connection
845 later. In the context of HTTP, this must be called only AFTER the
846 response has been received and the server has promised that the
847 connection will remain alive.
849 If a previous connection was persistent, it is closed. */
852 register_persistent (const char *host, int port, int fd, int ssl)
856 if (pconn.socket == fd)
858 /* The connection FD is already registered. */
863 /* The old persistent connection is still active; close it
864 first. This situation arises whenever a persistent
865 connection exists, but we then connect to a different
866 host, and try to register a persistent connection to that
868 invalidate_persistent ();
874 pconn.host = xstrdup (host);
878 DEBUGP (("Registered socket %d for persistent reuse.\n", fd));
881 /* Return non-zero if a persistent connection is available for
882 connecting to HOST:PORT. */
885 persistent_available_p (const char *host, int port, int ssl,
886 int *host_lookup_failed)
888 /* First, check whether a persistent connection is active at all. */
892 /* If we want SSL and the last connection wasn't or vice versa,
893 don't use it. Checking for host and port is not enough because
894 HTTP and HTTPS can apparently coexist on the same port. */
895 if (ssl != pconn.ssl)
898 /* If we're not connecting to the same port, we're not interested. */
899 if (port != pconn.port)
902 /* If the host is the same, we're in business. If not, there is
903 still hope -- read below. */
904 if (0 != strcasecmp (host, pconn.host))
906 /* If pconn.socket is already talking to HOST, we needn't
907 reconnect. This happens often when both sites are virtual
908 hosts distinguished only by name and served by the same
909 network interface, and hence the same web server (possibly
910 set up by the ISP and serving many different web sites).
911 This admittedly non-standard optimization does not contradict
912 HTTP and works well with popular server software. */
916 struct address_list *al;
919 /* Don't try to talk to two different SSL sites over the same
920 secure connection! (Besides, it's not clear if name-based
921 virtual hosting is even possible with SSL.) */
924 /* If pconn.socket's peer is one of the IP addresses HOST
925 resolves to, pconn.socket is for all intents and purposes
926 already talking to HOST. */
928 if (!socket_ip_address (pconn.socket, &ip, ENDPOINT_PEER))
930 /* Can't get the peer's address -- something must be very
931 wrong with the connection. */
932 invalidate_persistent ();
935 al = lookup_host (host, 0);
938 *host_lookup_failed = 1;
942 found = address_list_contains (al, &ip);
943 address_list_release (al);
948 /* The persistent connection's peer address was found among the
949 addresses HOST resolved to; therefore, pconn.sock is in fact
950 already talking to HOST -- no need to reconnect. */
953 /* Finally, check whether the connection is still open. This is
954 important because most server implement a liberal (short) timeout
955 on persistent connections. Wget can of course always reconnect
956 if the connection doesn't work out, but it's nicer to know in
957 advance. This test is a logical followup of the first test, but
958 is "expensive" and therefore placed at the end of the list. */
960 if (!test_socket_open (pconn.socket))
962 /* Oops, the socket is no longer open. Now that we know that,
963 let's invalidate the persistent connection before returning
965 invalidate_persistent ();
972 /* The idea behind these two CLOSE macros is to distinguish between
973 two cases: one when the job we've been doing is finished, and we
974 want to close the connection and leave, and two when something is
975 seriously wrong and we're closing the connection as part of
978 In case of keep_alive, CLOSE_FINISH should leave the connection
979 open, while CLOSE_INVALIDATE should still close it.
981 Note that the semantics of the flag `keep_alive' is "this
982 connection *will* be reused (the server has promised not to close
983 the connection once we're done)", while the semantics of
984 `pc_active_p && (fd) == pc_last_fd' is "we're *now* using an
985 active, registered connection". */
987 #define CLOSE_FINISH(fd) do { \
990 if (pconn_active && (fd) == pconn.socket) \
991 invalidate_persistent (); \
1000 #define CLOSE_INVALIDATE(fd) do { \
1001 if (pconn_active && (fd) == pconn.socket) \
1002 invalidate_persistent (); \
1010 wgint len; /* received length */
1011 wgint contlen; /* expected length */
1012 wgint restval; /* the restart value */
1013 int res; /* the result of last read */
1014 char *newloc; /* new location (redirection) */
1015 char *remote_time; /* remote time-stamp string */
1016 char *error; /* textual HTTP error */
1017 int statcode; /* status code */
1018 wgint rd_size; /* amount of data read from socket */
1019 double dltime; /* time it took to download the data */
1020 const char *referer; /* value of the referer header. */
1021 char **local_file; /* local file. */
1025 free_hstat (struct http_stat *hs)
1027 xfree_null (hs->newloc);
1028 xfree_null (hs->remote_time);
1029 xfree_null (hs->error);
1031 /* Guard against being called twice. */
1033 hs->remote_time = NULL;
1037 static char *create_authorization_line PARAMS ((const char *, const char *,
1038 const char *, const char *,
1040 static char *basic_authentication_encode PARAMS ((const char *, const char *));
1041 static int known_authentication_scheme_p PARAMS ((const char *));
1043 time_t http_atotm PARAMS ((const char *));
1045 #define BEGINS_WITH(line, string_constant) \
1046 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
1047 && (ISSPACE (line[sizeof (string_constant) - 1]) \
1048 || !line[sizeof (string_constant) - 1]))
1050 /* Retrieve a document through HTTP protocol. It recognizes status
1051 code, and correctly handles redirections. It closes the network
1052 socket. If it receives an error from the functions below it, it
1053 will print it if there is enough information to do so (almost
1054 always), returning the error to the caller (i.e. http_loop).
1056 Various HTTP parameters are stored to hs.
1058 If PROXY is non-NULL, the connection will be made to the proxy
1059 server, and u->url will be requested. */
1061 gethttp (struct url *u, struct http_stat *hs, int *dt, struct url *proxy)
1063 struct request *req;
1066 char *user, *passwd;
1070 wgint contlen, contrange;
1077 /* Whether authorization has been already tried. */
1078 int auth_tried_already;
1080 /* Whether our connection to the remote host is through SSL. */
1084 struct response *resp;
1088 /* Whether this connection will be kept alive after the HTTP request
1092 /* Whether keep-alive should be inhibited.
1094 RFC 2068 requests that 1.0 clients not send keep-alive requests
1095 to proxies. This is because many 1.0 proxies do not interpret
1096 the Connection header and transfer it to the remote server,
1097 causing it to not close the connection and leave both the proxy
1098 and the client hanging. */
1099 int inhibit_keep_alive =
1100 !opt.http_keep_alive || opt.ignore_length /*|| proxy != NULL*/;
1102 /* Headers sent when using POST. */
1103 wgint post_data_size = 0;
1105 int host_lookup_failed = 0;
1108 if (u->scheme == SCHEME_HTTPS)
1110 /* Initialize the SSL context. After this has once been done,
1111 it becomes a no-op. */
1112 switch (ssl_init ())
1114 case SSLERRCTXCREATE:
1116 logprintf (LOG_NOTQUIET, _("Failed to set up an SSL context\n"));
1117 return SSLERRCTXCREATE;
1118 case SSLERRCERTFILE:
1119 /* try without certfile */
1120 logprintf (LOG_NOTQUIET,
1121 _("Failed to load certificates from %s\n"),
1123 logprintf (LOG_NOTQUIET,
1124 _("Trying without the specified certificate\n"));
1127 logprintf (LOG_NOTQUIET,
1128 _("Failed to get certificate key from %s\n"),
1130 logprintf (LOG_NOTQUIET,
1131 _("Trying without the specified certificate\n"));
1137 #endif /* HAVE_SSL */
1139 if (!(*dt & HEAD_ONLY))
1140 /* If we're doing a GET on the URL, as opposed to just a HEAD, we need to
1141 know the local filename so we can save to it. */
1142 assert (*hs->local_file != NULL);
1144 auth_tried_already = 0;
1146 /* Initialize certain elements of struct http_stat. */
1151 hs->remote_time = NULL;
1156 /* Prepare the request to send. */
1158 req = request_new ();
1160 const char *meth = "GET";
1161 if (*dt & HEAD_ONLY)
1163 else if (opt.post_file_name || opt.post_data)
1165 /* Use the full path, i.e. one that includes the leading slash and
1166 the query string. E.g. if u->path is "foo/bar" and u->query is
1167 "param=value", full_path will be "/foo/bar?param=value". */
1168 request_set_method (req, meth,
1169 proxy ? xstrdup (u->url) : url_full_path (u));
1172 request_set_header (req, "Referer", (char *) hs->referer, rel_none);
1173 if (*dt & SEND_NOCACHE)
1174 request_set_header (req, "Pragma", "no-cache", rel_none);
1176 request_set_header (req, "Range",
1177 aprintf ("bytes=%s-",
1178 number_to_static_string (hs->restval)),
1181 request_set_header (req, "User-Agent", opt.useragent, rel_none);
1183 request_set_header (req, "User-Agent",
1184 aprintf ("Wget/%s", version_string), rel_value);
1185 request_set_header (req, "Accept", "*/*", rel_none);
1187 /* Find the username and password for authentication. */
1190 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
1191 user = user ? user : opt.http_user;
1192 passwd = passwd ? passwd : opt.http_passwd;
1196 /* We have the username and the password, but haven't tried
1197 any authorization yet. Let's see if the "Basic" method
1198 works. If not, we'll come back here and construct a
1199 proper authorization method with the right challenges.
1201 If we didn't employ this kind of logic, every URL that
1202 requires authorization would have to be processed twice,
1203 which is very suboptimal and generates a bunch of false
1204 "unauthorized" errors in the server log.
1206 #### But this logic also has a serious problem when used
1207 with stronger authentications: we *first* transmit the
1208 username and the password in clear text, and *then* attempt a
1209 stronger authentication scheme. That cannot be right! We
1210 are only fortunate that almost everyone still uses the
1211 `Basic' scheme anyway.
1213 There should be an option to prevent this from happening, for
1214 those who use strong authentication schemes and value their
1216 request_set_header (req, "Authorization",
1217 basic_authentication_encode (user, passwd),
1224 char *proxy_user, *proxy_passwd;
1225 /* For normal username and password, URL components override
1226 command-line/wgetrc parameters. With proxy
1227 authentication, it's the reverse, because proxy URLs are
1228 normally the "permanent" ones, so command-line args
1229 should take precedence. */
1230 if (opt.proxy_user && opt.proxy_passwd)
1232 proxy_user = opt.proxy_user;
1233 proxy_passwd = opt.proxy_passwd;
1237 proxy_user = proxy->user;
1238 proxy_passwd = proxy->passwd;
1240 /* #### This does not appear right. Can't the proxy request,
1241 say, `Digest' authentication? */
1242 if (proxy_user && proxy_passwd)
1243 proxyauth = basic_authentication_encode (proxy_user, proxy_passwd);
1245 /* If we're using a proxy, we will be connecting to the proxy
1249 /* Proxy authorization over SSL is handled below. */
1251 if (u->scheme != SCHEME_HTTPS)
1253 request_set_header (req, "Proxy-Authorization", proxyauth, rel_value);
1257 /* Whether we need to print the host header with braces around
1258 host, e.g. "Host: [3ffe:8100:200:2::2]:1234" instead of the
1259 usual "Host: symbolic-name:1234". */
1260 int squares = strchr (u->host, ':') != NULL;
1261 if (u->port == scheme_default_port (u->scheme))
1262 request_set_header (req, "Host",
1263 aprintf (squares ? "[%s]" : "%s", u->host),
1266 request_set_header (req, "Host",
1267 aprintf (squares ? "[%s]:%d" : "%s:%d",
1272 if (!inhibit_keep_alive)
1273 request_set_header (req, "Connection", "Keep-Alive", rel_none);
1276 request_set_header (req, "Cookie",
1277 cookie_header (wget_cookie_jar,
1278 u->host, u->port, u->path,
1280 u->scheme == SCHEME_HTTPS
1287 if (opt.post_data || opt.post_file_name)
1289 request_set_header (req, "Content-Type",
1290 "application/x-www-form-urlencoded", rel_none);
1292 post_data_size = strlen (opt.post_data);
1295 post_data_size = file_size (opt.post_file_name);
1296 if (post_data_size == -1)
1298 logprintf (LOG_NOTQUIET, "POST data file missing: %s\n",
1299 opt.post_file_name);
1303 request_set_header (req, "Content-Length",
1304 xstrdup (number_to_static_string (post_data_size)),
1308 /* Add the user headers. */
1309 if (opt.user_headers)
1312 for (i = 0; opt.user_headers[i]; i++)
1313 request_set_user_header (req, opt.user_headers[i]);
1317 /* We need to come back here when the initial attempt to retrieve
1318 without authorization header fails. (Expected to happen at least
1319 for the Digest authorization scheme.) */
1323 /* Establish the connection. */
1325 if (!inhibit_keep_alive)
1327 /* Look for a persistent connection to target host, unless a
1328 proxy is used. The exception is when SSL is in use, in which
1329 case the proxy is nothing but a passthrough to the target
1330 host, registered as a connection to the latter. */
1331 struct url *relevant = conn;
1333 if (u->scheme == SCHEME_HTTPS)
1337 if (persistent_available_p (relevant->host, relevant->port,
1339 relevant->scheme == SCHEME_HTTPS,
1343 &host_lookup_failed))
1345 sock = pconn.socket;
1346 using_ssl = pconn.ssl;
1347 logprintf (LOG_VERBOSE, _("Reusing existing connection to %s:%d.\n"),
1348 escnonprint (pconn.host), pconn.port);
1349 DEBUGP (("Reusing fd %d.\n", sock));
1355 /* In its current implementation, persistent_available_p will
1356 look up conn->host in some cases. If that lookup failed, we
1357 don't need to bother with connect_to_host. */
1358 if (host_lookup_failed)
1364 sock = connect_to_host (conn->host, conn->port);
1373 return (retryable_socket_connect_error (errno)
1374 ? CONERROR : CONIMPOSSIBLE);
1378 if (proxy && u->scheme == SCHEME_HTTPS)
1380 /* When requesting SSL URLs through proxies, use the
1381 CONNECT method to request passthrough. */
1382 struct request *connreq = request_new ();
1383 request_set_method (connreq, "CONNECT",
1384 aprintf ("%s:%d", u->host, u->port));
1387 request_set_header (connreq, "Proxy-Authorization",
1388 proxyauth, rel_value);
1389 /* Now that PROXYAUTH is part of the CONNECT request,
1390 zero it out so we don't send proxy authorization with
1391 the regular request below. */
1395 write_error = request_send (connreq, sock);
1396 request_free (connreq);
1397 if (write_error < 0)
1399 logprintf (LOG_VERBOSE, _("Failed writing to proxy: %s.\n"),
1401 CLOSE_INVALIDATE (sock);
1405 head = read_http_response_head (sock);
1408 logprintf (LOG_VERBOSE, _("Failed reading proxy response: %s\n"),
1410 CLOSE_INVALIDATE (sock);
1419 DEBUGP (("proxy responded with: [%s]\n", head));
1421 resp = resp_new (head);
1422 statcode = resp_status (resp, &message);
1425 if (statcode != 200)
1428 logprintf (LOG_NOTQUIET, _("Proxy tunneling failed: %s"),
1429 message ? escnonprint (message) : "?");
1430 xfree_null (message);
1433 xfree_null (message);
1435 /* SOCK is now *really* connected to u->host, so update CONN
1436 to reflect this. That way register_persistent will
1437 register SOCK as being connected to u->host:u->port. */
1441 if (conn->scheme == SCHEME_HTTPS)
1443 if (!ssl_connect (sock))
1450 #endif /* HAVE_SSL */
1453 /* Send the request to server. */
1454 write_error = request_send (req, sock);
1456 if (write_error >= 0)
1460 DEBUGP (("[POST data: %s]\n", opt.post_data));
1461 write_error = fd_write (sock, opt.post_data, post_data_size, -1);
1463 else if (opt.post_file_name && post_data_size != 0)
1464 write_error = post_file (sock, opt.post_file_name, post_data_size);
1467 if (write_error < 0)
1469 logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"),
1471 CLOSE_INVALIDATE (sock);
1475 logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
1476 proxy ? "Proxy" : "HTTP");
1481 head = read_http_response_head (sock);
1486 logputs (LOG_NOTQUIET, _("No data received.\n"));
1487 CLOSE_INVALIDATE (sock);
1493 logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
1495 CLOSE_INVALIDATE (sock);
1500 DEBUGP (("\n---response begin---\n%s---response end---\n", head));
1502 resp = resp_new (head);
1504 /* Check for status line. */
1506 statcode = resp_status (resp, &message);
1507 if (!opt.server_response)
1508 logprintf (LOG_VERBOSE, "%2d %s\n", statcode,
1509 message ? escnonprint (message) : "");
1512 logprintf (LOG_VERBOSE, "\n");
1513 print_server_response (resp, " ");
1516 if (!opt.ignore_length
1517 && resp_header_copy (resp, "Content-Length", hdrval, sizeof (hdrval)))
1521 parsed = str_to_wgint (hdrval, NULL, 10);
1522 if (parsed == WGINT_MAX && errno == ERANGE)
1524 #### If Content-Length is out of range, it most likely
1525 means that the file is larger than 2G and that we're
1526 compiled without LFS. In that case we should probably
1527 refuse to even attempt to download the file. */
1533 /* Check for keep-alive related responses. */
1534 if (!inhibit_keep_alive && contlen != -1)
1536 if (resp_header_copy (resp, "Keep-Alive", NULL, 0))
1538 else if (resp_header_copy (resp, "Connection", hdrval, sizeof (hdrval)))
1540 if (0 == strcasecmp (hdrval, "Keep-Alive"))
1545 /* The server has promised that it will not close the connection
1546 when we're done. This means that we can register it. */
1547 register_persistent (conn->host, conn->port, sock, using_ssl);
1549 if (statcode == HTTP_STATUS_UNAUTHORIZED)
1551 /* Authorization is required. */
1552 skip_short_body (sock, contlen);
1553 CLOSE_FINISH (sock);
1554 if (auth_tried_already || !(user && passwd))
1556 /* If we have tried it already, then there is not point
1558 logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
1562 char *www_authenticate = resp_header_strdup (resp,
1563 "WWW-Authenticate");
1564 /* If the authentication scheme is unknown or if it's the
1565 "Basic" authentication (which we try by default), there's
1566 no sense in retrying. */
1567 if (!www_authenticate
1568 || !known_authentication_scheme_p (www_authenticate)
1569 || BEGINS_WITH (www_authenticate, "Basic"))
1571 xfree_null (www_authenticate);
1572 logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
1577 auth_tried_already = 1;
1578 pth = url_full_path (u);
1579 request_set_header (req, "Authorization",
1580 create_authorization_line (www_authenticate,
1582 request_method (req),
1586 xfree (www_authenticate);
1587 goto retry_with_auth;
1595 hs->statcode = statcode;
1597 hs->error = xstrdup (_("Malformed status line"));
1599 hs->error = xstrdup (_("(no description)"));
1601 hs->error = xstrdup (message);
1604 type = resp_header_strdup (resp, "Content-Type");
1607 char *tmp = strchr (type, ';');
1610 while (tmp > type && ISSPACE (tmp[-1]))
1615 hs->newloc = resp_header_strdup (resp, "Location");
1616 hs->remote_time = resp_header_strdup (resp, "Last-Modified");
1618 /* Handle (possibly multiple instances of) the Set-Cookie header. */
1621 const char *scbeg, *scend;
1622 /* The jar should have been created by now. */
1623 assert (wget_cookie_jar != NULL);
1625 (scpos = resp_header_locate (resp, "Set-Cookie", scpos,
1626 &scbeg, &scend)) != -1;
1629 char *set_cookie = strdupdelim (scbeg, scend);
1630 cookie_handle_set_cookie (wget_cookie_jar, u->host, u->port, u->path,
1636 if (resp_header_copy (resp, "Content-Range", hdrval, sizeof (hdrval)))
1638 wgint first_byte_pos, last_byte_pos, entity_length;
1639 if (parse_content_range (hdrval, &first_byte_pos, &last_byte_pos,
1641 contrange = first_byte_pos;
1646 /* 20x responses are counted among successful by default. */
1647 if (H_20X (statcode))
1650 /* Return if redirected. */
1651 if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES)
1653 /* RFC2068 says that in case of the 300 (multiple choices)
1654 response, the server can output a preferred URL through
1655 `Location' header; otherwise, the request should be treated
1656 like GET. So, if the location is set, it will be a
1657 redirection; otherwise, just proceed normally. */
1658 if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc)
1662 logprintf (LOG_VERBOSE,
1663 _("Location: %s%s\n"),
1664 hs->newloc ? escnonprint_uri (hs->newloc) : _("unspecified"),
1665 hs->newloc ? _(" [following]") : "");
1667 skip_short_body (sock, contlen);
1668 CLOSE_FINISH (sock);
1674 /* If content-type is not given, assume text/html. This is because
1675 of the multitude of broken CGI's that "forget" to generate the
1678 0 == strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S)) ||
1679 0 == strncasecmp (type, TEXTXHTML_S, strlen (TEXTXHTML_S)))
1684 if (opt.html_extension && (*dt & TEXTHTML))
1685 /* -E / --html-extension / html_extension = on was specified, and this is a
1686 text/html file. If some case-insensitive variation on ".htm[l]" isn't
1687 already the file's suffix, tack on ".html". */
1689 char* last_period_in_local_filename = strrchr(*hs->local_file, '.');
1691 if (last_period_in_local_filename == NULL
1692 || !(0 == strcasecmp (last_period_in_local_filename, ".htm")
1693 || 0 == strcasecmp (last_period_in_local_filename, ".html")))
1695 size_t local_filename_len = strlen(*hs->local_file);
1697 *hs->local_file = xrealloc(*hs->local_file,
1698 local_filename_len + sizeof(".html"));
1699 strcpy(*hs->local_file + local_filename_len, ".html");
1701 *dt |= ADDED_HTML_EXTENSION;
1705 if (statcode == HTTP_STATUS_RANGE_NOT_SATISFIABLE)
1707 /* If `-c' is in use and the file has been fully downloaded (or
1708 the remote file has shrunk), Wget effectively requests bytes
1709 after the end of file and the server response with 416. */
1710 logputs (LOG_VERBOSE, _("\
1711 \n The file is already fully retrieved; nothing to do.\n\n"));
1712 /* In case the caller inspects. */
1715 /* Mark as successfully retrieved. */
1718 CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
1719 might be more bytes in the body. */
1720 return RETRUNNEEDED;
1722 if ((contrange != 0 && contrange != hs->restval)
1723 || (H_PARTIAL (statcode) && !contrange))
1725 /* The Range request was somehow misunderstood by the server.
1728 CLOSE_INVALIDATE (sock);
1731 hs->contlen = contlen + contrange;
1737 /* No need to print this output if the body won't be
1738 downloaded at all, or if the original server response is
1740 logputs (LOG_VERBOSE, _("Length: "));
1743 logputs (LOG_VERBOSE, with_thousand_seps (contlen + contrange));
1744 if (contlen + contrange >= 1024)
1745 logprintf (LOG_VERBOSE, " (%s)",
1746 human_readable (contlen + contrange));
1749 if (contlen >= 1024)
1750 logprintf (LOG_VERBOSE, _(", %s (%s) remaining"),
1751 with_thousand_seps (contlen),
1752 human_readable (contlen));
1754 logprintf (LOG_VERBOSE, _(", %s remaining"),
1755 with_thousand_seps (contlen));
1759 logputs (LOG_VERBOSE,
1760 opt.ignore_length ? _("ignored") : _("unspecified"));
1762 logprintf (LOG_VERBOSE, " [%s]\n", escnonprint (type));
1764 logputs (LOG_VERBOSE, "\n");
1768 type = NULL; /* We don't need it any more. */
1770 /* Return if we have no intention of further downloading. */
1771 if (!(*dt & RETROKF) || (*dt & HEAD_ONLY))
1773 /* In case the caller cares to look... */
1777 /* Pre-1.10 Wget used CLOSE_INVALIDATE here. Now we trust the
1778 servers not to send body in response to a HEAD request. If
1779 you encounter such a server (more likely a broken CGI), use
1780 `--no-http-keep-alive'. */
1781 CLOSE_FINISH (sock);
1782 return RETRFINISHED;
1785 /* Open the local file. */
1788 mkalldirs (*hs->local_file);
1790 rotate_backups (*hs->local_file);
1792 fp = fopen (*hs->local_file, "ab");
1793 else if (opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct
1794 || opt.output_document)
1795 fp = fopen (*hs->local_file, "wb");
1798 fp = fopen_excl (*hs->local_file, 0);
1799 if (!fp && errno == EEXIST)
1801 /* We cannot just invent a new name and use it (which is
1802 what functions like unique_create typically do)
1803 because we told the user we'd use this name.
1804 Instead, return and retry the download. */
1805 logprintf (LOG_NOTQUIET,
1806 _("%s has sprung into existence.\n"),
1808 CLOSE_INVALIDATE (sock);
1809 return FOPEN_EXCL_ERR;
1814 logprintf (LOG_NOTQUIET, "%s: %s\n", *hs->local_file, strerror (errno));
1815 CLOSE_INVALIDATE (sock);
1822 /* #### This confuses the timestamping code that checks for file
1823 size. Maybe we should save some additional information? */
1824 if (opt.save_headers)
1825 fwrite (head, 1, strlen (head), fp);
1827 /* Download the request body. */
1830 flags |= rb_read_exactly;
1831 if (hs->restval > 0 && contrange == 0)
1832 /* If the server ignored our range request, instruct fd_read_body
1833 to skip the first RESTVAL bytes of body. */
1834 flags |= rb_skip_startpos;
1835 hs->len = hs->restval;
1837 hs->res = fd_read_body (sock, fp, contlen != -1 ? contlen : 0,
1838 hs->restval, &hs->rd_size, &hs->len, &hs->dltime,
1842 CLOSE_FINISH (sock);
1844 CLOSE_INVALIDATE (sock);
1847 /* Close or flush the file. We have to be careful to check for
1848 error here. Checking the result of fwrite() is not enough --
1849 errors could go unnoticed! */
1852 flush_res = fclose (fp);
1854 flush_res = fflush (fp);
1855 if (flush_res == EOF)
1860 return RETRFINISHED;
1863 /* The genuine HTTP loop! This is the part where the retrieval is
1864 retried, and retried, and retried, and... */
1866 http_loop (struct url *u, char **newloc, char **local_file, const char *referer,
1867 int *dt, struct url *proxy)
1870 int use_ts, got_head = 0; /* time-stamping info */
1871 char *filename_plus_orig_suffix;
1872 char *local_filename = NULL;
1873 char *tms, *locf, *tmrate;
1875 time_t tml = -1, tmr = -1; /* local and remote time-stamps */
1876 wgint local_size = 0; /* the size of the local file */
1877 size_t filename_len;
1878 struct http_stat hstat; /* HTTP status */
1882 /* This used to be done in main(), but it's a better idea to do it
1883 here so that we don't go through the hoops if we're just using
1887 if (!wget_cookie_jar)
1888 wget_cookie_jar = cookie_jar_new ();
1889 if (opt.cookies_input && !cookies_loaded_p)
1891 cookie_jar_load (wget_cookie_jar, opt.cookies_input);
1892 cookies_loaded_p = 1;
1898 /* Warn on (likely bogus) wildcard usage in HTTP. Don't use
1899 has_wildcards_p because it would also warn on `?', and we know that
1900 shows up in CGI paths a *lot*. */
1901 if (strchr (u->url, '*'))
1902 logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
1906 /* Determine the local filename. */
1907 if (local_file && *local_file)
1908 hstat.local_file = local_file;
1909 else if (local_file && !opt.output_document)
1911 *local_file = url_file_name (u);
1912 hstat.local_file = local_file;
1916 dummy = url_file_name (u);
1917 hstat.local_file = &dummy;
1918 /* be honest about where we will save the file */
1919 if (local_file && opt.output_document)
1920 *local_file = HYPHENP (opt.output_document) ? NULL : xstrdup (opt.output_document);
1923 if (!opt.output_document)
1924 locf = *hstat.local_file;
1926 locf = opt.output_document;
1928 hstat.referer = referer;
1930 filename_len = strlen (*hstat.local_file);
1931 filename_plus_orig_suffix = alloca (filename_len + sizeof (".orig"));
1933 if (opt.noclobber && file_exists_p (*hstat.local_file))
1935 /* If opt.noclobber is turned on and file already exists, do not
1936 retrieve the file */
1937 logprintf (LOG_VERBOSE, _("\
1938 File `%s' already there, will not retrieve.\n"), *hstat.local_file);
1939 /* If the file is there, we suppose it's retrieved OK. */
1942 /* #### Bogusness alert. */
1943 /* If its suffix is "html" or "htm" or similar, assume text/html. */
1944 if (has_html_suffix_p (*hstat.local_file))
1952 if (opt.timestamping)
1954 int local_dot_orig_file_exists = 0;
1956 if (opt.backup_converted)
1957 /* If -K is specified, we'll act on the assumption that it was specified
1958 last time these files were downloaded as well, and instead of just
1959 comparing local file X against server file X, we'll compare local
1960 file X.orig (if extant, else X) against server file X. If -K
1961 _wasn't_ specified last time, or the server contains files called
1962 *.orig, -N will be back to not operating correctly with -k. */
1964 /* Would a single s[n]printf() call be faster? --dan
1966 Definitely not. sprintf() is horribly slow. It's a
1967 different question whether the difference between the two
1968 affects a program. Usually I'd say "no", but at one
1969 point I profiled Wget, and found that a measurable and
1970 non-negligible amount of time was lost calling sprintf()
1971 in url.c. Replacing sprintf with inline calls to
1972 strcpy() and number_to_string() made a difference.
1974 memcpy (filename_plus_orig_suffix, *hstat.local_file, filename_len);
1975 memcpy (filename_plus_orig_suffix + filename_len,
1976 ".orig", sizeof (".orig"));
1978 /* Try to stat() the .orig file. */
1979 if (stat (filename_plus_orig_suffix, &st) == 0)
1981 local_dot_orig_file_exists = 1;
1982 local_filename = filename_plus_orig_suffix;
1986 if (!local_dot_orig_file_exists)
1987 /* Couldn't stat() <file>.orig, so try to stat() <file>. */
1988 if (stat (*hstat.local_file, &st) == 0)
1989 local_filename = *hstat.local_file;
1991 if (local_filename != NULL)
1992 /* There was a local file, so we'll check later to see if the version
1993 the server has is the same version we already have, allowing us to
1999 /* Modification time granularity is 2 seconds for Windows, so
2000 increase local time by 1 second for later comparison. */
2003 local_size = st.st_size;
2007 /* Reset the counter. */
2013 /* Increment the pass counter. */
2015 sleep_between_retrievals (count);
2016 /* Get the current time string. */
2017 tms = time_str (NULL);
2018 /* Print fetch message, if opt.verbose. */
2021 char *hurl = url_string (u, 1);
2025 sprintf (tmp, _("(try:%2d)"), count);
2026 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
2027 tms, hurl, tmp, locf);
2029 ws_changetitle (hurl);
2034 /* Default document type is empty. However, if spider mode is
2035 on or time-stamping is employed, HEAD_ONLY commands is
2036 encoded within *dt. */
2037 if (opt.spider || (use_ts && !got_head))
2042 /* Decide whether or not to restart. */
2045 hstat.restval = hstat.len; /* continue where we left off */
2046 else if (opt.always_rest
2047 && stat (locf, &st) == 0
2048 && S_ISREG (st.st_mode))
2049 hstat.restval = st.st_size;
2051 /* Decide whether to send the no-cache directive. We send it in
2053 a) we're using a proxy, and we're past our first retrieval.
2054 Some proxies are notorious for caching incomplete data, so
2055 we require a fresh get.
2056 b) caching is explicitly inhibited. */
2057 if ((proxy && count > 1) /* a */
2058 || !opt.allow_cache /* b */
2060 *dt |= SEND_NOCACHE;
2062 *dt &= ~SEND_NOCACHE;
2064 /* Try fetching the document, or at least its head. */
2065 err = gethttp (u, &hstat, dt, proxy);
2067 /* It's unfortunate that wget determines the local filename before finding
2068 out the Content-Type of the file. Barring a major restructuring of the
2069 code, we need to re-set locf here, since gethttp() may have xrealloc()d
2070 *hstat.local_file to tack on ".html". */
2071 if (!opt.output_document)
2072 locf = *hstat.local_file;
2075 tms = time_str (NULL);
2076 /* Get the new location (with or without the redirection). */
2078 *newloc = xstrdup (hstat.newloc);
2081 case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
2082 case CONERROR: case READERR: case WRITEFAILED:
2083 case RANGEERR: case FOPEN_EXCL_ERR:
2084 /* Non-fatal errors continue executing the loop, which will
2085 bring them to "while" statement at the end, to judge
2086 whether the number of tries was exceeded. */
2087 free_hstat (&hstat);
2088 printwhat (count, opt.ntry);
2089 if (err == FOPEN_EXCL_ERR)
2091 /* Re-determine the file name. */
2092 if (local_file && *local_file)
2094 xfree (*local_file);
2095 *local_file = url_file_name (u);
2096 hstat.local_file = local_file;
2101 dummy = url_file_name (u);
2102 hstat.local_file = &dummy;
2104 /* be honest about where we will save the file */
2105 if (local_file && opt.output_document)
2106 *local_file = HYPHENP (opt.output_document) ? NULL : xstrdup (opt.output_document);
2107 if (!opt.output_document)
2108 locf = *hstat.local_file;
2110 locf = opt.output_document;
2114 case HOSTERR: case CONIMPOSSIBLE: case PROXERR: case AUTHFAILED:
2115 case SSLERRCTXCREATE: case CONTNOTSUPPORTED:
2116 /* Fatal errors just return from the function. */
2117 free_hstat (&hstat);
2121 case FWRITEERR: case FOPENERR:
2122 /* Another fatal error. */
2123 logputs (LOG_VERBOSE, "\n");
2124 logprintf (LOG_NOTQUIET, _("Cannot write to `%s' (%s).\n"),
2125 *hstat.local_file, strerror (errno));
2126 free_hstat (&hstat);
2131 /* Another fatal error. */
2132 logputs (LOG_VERBOSE, "\n");
2133 logprintf (LOG_NOTQUIET, _("Unable to establish SSL connection.\n"));
2134 free_hstat (&hstat);
2139 /* Return the new location to the caller. */
2142 logprintf (LOG_NOTQUIET,
2143 _("ERROR: Redirection (%d) without location.\n"),
2145 free_hstat (&hstat);
2149 free_hstat (&hstat);
2154 /* The file was already fully retrieved. */
2155 free_hstat (&hstat);
2160 /* Deal with you later. */
2163 /* All possibilities should have been exhausted. */
2166 if (!(*dt & RETROKF))
2170 /* #### Ugly ugly ugly! */
2171 char *hurl = url_string (u, 1);
2172 logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
2175 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
2176 tms, hstat.statcode, escnonprint (hstat.error));
2177 logputs (LOG_VERBOSE, "\n");
2178 free_hstat (&hstat);
2183 /* Did we get the time-stamp? */
2186 if (opt.timestamping && !hstat.remote_time)
2188 logputs (LOG_NOTQUIET, _("\
2189 Last-modified header missing -- time-stamps turned off.\n"));
2191 else if (hstat.remote_time)
2193 /* Convert the date-string into struct tm. */
2194 tmr = http_atotm (hstat.remote_time);
2195 if (tmr == (time_t) (-1))
2196 logputs (LOG_VERBOSE, _("\
2197 Last-modified header invalid -- time-stamp ignored.\n"));
2201 /* The time-stamping section. */
2206 use_ts = 0; /* no more time-stamping */
2207 count = 0; /* the retrieve count for HEAD is
2209 if (hstat.remote_time && tmr != (time_t) (-1))
2211 /* Now time-stamping can be used validly. Time-stamping
2212 means that if the sizes of the local and remote file
2213 match, and local file is newer than the remote file,
2214 it will not be retrieved. Otherwise, the normal
2215 download procedure is resumed. */
2217 (hstat.contlen == -1 || local_size == hstat.contlen))
2219 logprintf (LOG_VERBOSE, _("\
2220 Server file no newer than local file `%s' -- not retrieving.\n\n"),
2222 free_hstat (&hstat);
2226 else if (tml >= tmr)
2227 logprintf (LOG_VERBOSE, _("\
2228 The sizes do not match (local %s) -- retrieving.\n"),
2229 number_to_static_string (local_size));
2231 logputs (LOG_VERBOSE,
2232 _("Remote file is newer, retrieving.\n"));
2234 free_hstat (&hstat);
2237 if ((tmr != (time_t) (-1))
2239 && ((hstat.len == hstat.contlen) ||
2240 ((hstat.res == 0) &&
2241 ((hstat.contlen == -1) ||
2242 (hstat.len >= hstat.contlen && !opt.kill_longer)))))
2244 /* #### This code repeats in http.c and ftp.c. Move it to a
2246 const char *fl = NULL;
2247 if (opt.output_document)
2249 if (output_stream_regular)
2250 fl = opt.output_document;
2253 fl = *hstat.local_file;
2257 /* End of time-stamping section. */
2261 logprintf (LOG_NOTQUIET, "%d %s\n\n", hstat.statcode,
2262 escnonprint (hstat.error));
2267 tmrate = retr_rate (hstat.rd_size, hstat.dltime, 0);
2269 if (hstat.len == hstat.contlen)
2273 logprintf (LOG_VERBOSE,
2274 _("%s (%s) - `%s' saved [%s/%s]\n\n"),
2276 number_to_static_string (hstat.len),
2277 number_to_static_string (hstat.contlen));
2278 logprintf (LOG_NONVERBOSE,
2279 "%s URL:%s [%s/%s] -> \"%s\" [%d]\n",
2281 number_to_static_string (hstat.len),
2282 number_to_static_string (hstat.contlen),
2286 total_downloaded_bytes += hstat.len;
2288 /* Remember that we downloaded the file for later ".orig" code. */
2289 if (*dt & ADDED_HTML_EXTENSION)
2290 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
2292 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
2294 free_hstat (&hstat);
2298 else if (hstat.res == 0) /* No read error */
2300 if (hstat.contlen == -1) /* We don't know how much we were supposed
2301 to get, so assume we succeeded. */
2305 logprintf (LOG_VERBOSE,
2306 _("%s (%s) - `%s' saved [%s]\n\n"),
2308 number_to_static_string (hstat.len));
2309 logprintf (LOG_NONVERBOSE,
2310 "%s URL:%s [%s] -> \"%s\" [%d]\n",
2311 tms, u->url, number_to_static_string (hstat.len),
2315 total_downloaded_bytes += hstat.len;
2317 /* Remember that we downloaded the file for later ".orig" code. */
2318 if (*dt & ADDED_HTML_EXTENSION)
2319 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
2321 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
2323 free_hstat (&hstat);
2327 else if (hstat.len < hstat.contlen) /* meaning we lost the
2328 connection too soon */
2330 logprintf (LOG_VERBOSE,
2331 _("%s (%s) - Connection closed at byte %s. "),
2332 tms, tmrate, number_to_static_string (hstat.len));
2333 printwhat (count, opt.ntry);
2334 free_hstat (&hstat);
2337 else if (!opt.kill_longer) /* meaning we got more than expected */
2339 logprintf (LOG_VERBOSE,
2340 _("%s (%s) - `%s' saved [%s/%s])\n\n"),
2342 number_to_static_string (hstat.len),
2343 number_to_static_string (hstat.contlen));
2344 logprintf (LOG_NONVERBOSE,
2345 "%s URL:%s [%s/%s] -> \"%s\" [%d]\n",
2347 number_to_static_string (hstat.len),
2348 number_to_static_string (hstat.contlen),
2351 total_downloaded_bytes += hstat.len;
2353 /* Remember that we downloaded the file for later ".orig" code. */
2354 if (*dt & ADDED_HTML_EXTENSION)
2355 downloaded_file(FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, locf);
2357 downloaded_file(FILE_DOWNLOADED_NORMALLY, locf);
2359 free_hstat (&hstat);
2363 else /* the same, but not accepted */
2365 logprintf (LOG_VERBOSE,
2366 _("%s (%s) - Connection closed at byte %s/%s. "),
2368 number_to_static_string (hstat.len),
2369 number_to_static_string (hstat.contlen));
2370 printwhat (count, opt.ntry);
2371 free_hstat (&hstat);
2375 else /* now hstat.res can only be -1 */
2377 if (hstat.contlen == -1)
2379 logprintf (LOG_VERBOSE,
2380 _("%s (%s) - Read error at byte %s (%s)."),
2381 tms, tmrate, number_to_static_string (hstat.len),
2383 printwhat (count, opt.ntry);
2384 free_hstat (&hstat);
2387 else /* hstat.res == -1 and contlen is given */
2389 logprintf (LOG_VERBOSE,
2390 _("%s (%s) - Read error at byte %s/%s (%s). "),
2392 number_to_static_string (hstat.len),
2393 number_to_static_string (hstat.contlen),
2395 printwhat (count, opt.ntry);
2396 free_hstat (&hstat);
2403 while (!opt.ntry || (count < opt.ntry));
2407 /* Converts struct tm to time_t, assuming the data in tm is UTC rather
2408 than local timezone.
2410 mktime is similar but assumes struct tm, also known as the
2411 "broken-down" form of time, is in local time zone. mktime_from_utc
2412 uses mktime to make the conversion understanding that an offset
2413 will be introduced by the local time assumption.
2415 mktime_from_utc then measures the introduced offset by applying
2416 gmtime to the initial result and applying mktime to the resulting
2417 "broken-down" form. The difference between the two mktime results
2418 is the measured offset which is then subtracted from the initial
2419 mktime result to yield a calendar time which is the value returned.
2421 tm_isdst in struct tm is set to 0 to force mktime to introduce a
2422 consistent offset (the non DST offset) since tm and tm+o might be
2423 on opposite sides of a DST change.
2425 Some implementations of mktime return -1 for the nonexistent
2426 localtime hour at the beginning of DST. In this event, use
2427 mktime(tm - 1hr) + 3600.
2431 gmtime(t+o) --> tm+o
2432 mktime(tm+o) --> t+2o
2433 t+o - (t+2o - t+o) = t
2435 Note that glibc contains a function of the same purpose named
2436 `timegm' (reverse of gmtime). But obviously, it is not universally
2437 available, and unfortunately it is not straightforwardly
2438 extractable for use here. Perhaps configure should detect timegm
2439 and use it where available.
2441 Contributed by Roger Beeman <beeman@cisco.com>, with the help of
2442 Mark Baushke <mdb@cisco.com> and the rest of the Gurus at CISCO.
2443 Further improved by Roger with assistance from Edward J. Sabol
2444 based on input by Jamie Zawinski. */
2447 mktime_from_utc (struct tm *t)
2458 return -1; /* can't deal with output from strptime */
2469 return -1; /* can't deal with output from gmtime */
2472 return (tl - (tb - tl));
2475 /* Check whether the result of strptime() indicates success.
2476 strptime() returns the pointer to how far it got to in the string.
2477 The processing has been successful if the string is at `GMT' or
2478 `+X', or at the end of the string.
2480 In extended regexp parlance, the function returns 1 if P matches
2481 "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (which strptime
2482 can return) is considered a failure and 0 is returned. */
2484 check_end (const char *p)
2488 while (ISSPACE (*p))
2491 || (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
2492 || ((p[0] == '+' || p[0] == '-') && ISDIGIT (p[1])))
2498 /* Convert the textual specification of time in TIME_STRING to the
2499 number of seconds since the Epoch.
2501 TIME_STRING can be in any of the three formats RFC2068 allows the
2502 HTTP servers to emit -- RFC1123-date, RFC850-date or asctime-date.
2503 Timezones are ignored, and should be GMT.
2505 Return the computed time_t representation, or -1 if the conversion
2508 This function uses strptime with various string formats for parsing
2509 TIME_STRING. This results in a parser that is not as lenient in
2510 interpreting TIME_STRING as I would like it to be. Being based on
2511 strptime, it always allows shortened months, one-digit days, etc.,
2512 but due to the multitude of formats in which time can be
2513 represented, an ideal HTTP time parser would be even more
2514 forgiving. It should completely ignore things like week days and
2515 concentrate only on the various forms of representing years,
2516 months, days, hours, minutes, and seconds. For example, it would
2517 be nice if it accepted ISO 8601 out of the box.
2519 I've investigated free and PD code for this purpose, but none was
2520 usable. getdate was big and unwieldy, and had potential copyright
2521 issues, or so I was informed. Dr. Marcus Hennecke's atotm(),
2522 distributed with phttpd, is excellent, but we cannot use it because
2523 it is not assigned to the FSF. So I stuck it with strptime. */
2526 http_atotm (const char *time_string)
2528 /* NOTE: Solaris strptime man page claims that %n and %t match white
2529 space, but that's not universally available. Instead, we simply
2530 use ` ' to mean "skip all WS", which works under all strptime
2531 implementations I've tested. */
2533 static const char *time_formats[] = {
2534 "%a, %d %b %Y %T", /* RFC1123: Thu, 29 Jan 1998 22:12:57 */
2535 "%A, %d-%b-%y %T", /* RFC850: Thursday, 29-Jan-98 22:12:57 */
2536 "%a, %d-%b-%Y %T", /* pseudo-RFC850: Thu, 29-Jan-1998 22:12:57
2537 (google.com uses this for their cookies.) */
2538 "%a %b %d %T %Y" /* asctime: Thu Jan 29 22:12:57 1998 */
2544 /* According to Roger Beeman, we need to initialize tm_isdst, since
2545 strptime won't do it. */
2548 /* Note that under foreign locales Solaris strptime() fails to
2549 recognize English dates, which renders this function useless. We
2550 solve this by being careful not to affect LC_TIME when
2551 initializing locale.
2553 Another solution would be to temporarily set locale to C, invoke
2554 strptime(), and restore it back. This is slow and dirty,
2555 however, and locale support other than LC_MESSAGES can mess other
2556 things, so I rather chose to stick with just setting LC_MESSAGES.
2558 GNU strptime does not have this problem because it recognizes
2559 both international and local dates. */
2561 for (i = 0; i < countof (time_formats); i++)
2562 if (check_end (strptime (time_string, time_formats[i], &t)))
2563 return mktime_from_utc (&t);
2565 /* All formats have failed. */
2569 /* Authorization support: We support two authorization schemes:
2571 * `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
2573 * `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
2574 consisting of answering to the server's challenge with the proper
2577 /* How many bytes it will take to store LEN bytes in base64. */
2578 #define BASE64_LENGTH(len) (4 * (((len) + 2) / 3))
2580 /* Encode the string S of length LENGTH to base64 format and place it
2581 to STORE. STORE will be 0-terminated, and must point to a writable
2582 buffer of at least 1+BASE64_LENGTH(length) bytes. */
2584 base64_encode (const char *s, char *store, int length)
2586 /* Conversion table. */
2587 static char tbl[64] = {
2588 'A','B','C','D','E','F','G','H',
2589 'I','J','K','L','M','N','O','P',
2590 'Q','R','S','T','U','V','W','X',
2591 'Y','Z','a','b','c','d','e','f',
2592 'g','h','i','j','k','l','m','n',
2593 'o','p','q','r','s','t','u','v',
2594 'w','x','y','z','0','1','2','3',
2595 '4','5','6','7','8','9','+','/'
2598 unsigned char *p = (unsigned char *)store;
2600 /* Transform the 3x8 bits to 4x6 bits, as required by base64. */
2601 for (i = 0; i < length; i += 3)
2603 *p++ = tbl[s[0] >> 2];
2604 *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
2605 *p++ = tbl[((s[1] & 0xf) << 2) + (s[2] >> 6)];
2606 *p++ = tbl[s[2] & 0x3f];
2609 /* Pad the result if necessary... */
2610 if (i == length + 1)
2612 else if (i == length + 2)
2613 *(p - 1) = *(p - 2) = '=';
2614 /* ...and zero-terminate it. */
2618 /* Create the authentication header contents for the `Basic' scheme.
2619 This is done by encoding the string `USER:PASS' in base64 and
2620 prepending `HEADER: Basic ' to it. */
2622 basic_authentication_encode (const char *user, const char *passwd)
2625 int len1 = strlen (user) + 1 + strlen (passwd);
2626 int len2 = BASE64_LENGTH (len1);
2628 t1 = (char *)alloca (len1 + 1);
2629 sprintf (t1, "%s:%s", user, passwd);
2631 t2 = (char *)alloca (len2 + 1);
2632 base64_encode (t1, t2, len1);
2634 return concat_strings ("Basic ", t2, (char *) 0);
2637 #define SKIP_WS(x) do { \
2638 while (ISSPACE (*(x))) \
2643 /* Parse HTTP `WWW-Authenticate:' header. AU points to the beginning
2644 of a field in such a header. If the field is the one specified by
2645 ATTR_NAME ("realm", "opaque", and "nonce" are used by the current
2646 digest authorization code), extract its value in the (char*)
2647 variable pointed by RET. Returns negative on a malformed header,
2648 or number of bytes that have been parsed by this call. */
2650 extract_header_attr (const char *au, const char *attr_name, char **ret)
2653 const char *cp = au;
2655 if (strncmp (cp, attr_name, strlen (attr_name)) == 0)
2657 cp += strlen (attr_name);
2670 for (ep = cp; *ep && *ep != '\"'; ep++)
2675 *ret = strdupdelim (cp, ep);
2682 /* Dump the hexadecimal representation of HASH to BUF. HASH should be
2683 an array of 16 bytes containing the hash keys, and BUF should be a
2684 buffer of 33 writable characters (32 for hex digits plus one for
2685 zero termination). */
2687 dump_hash (unsigned char *buf, const unsigned char *hash)
2691 for (i = 0; i < MD5_HASHLEN; i++, hash++)
2693 *buf++ = XNUM_TO_digit (*hash >> 4);
2694 *buf++ = XNUM_TO_digit (*hash & 0xf);
2699 /* Take the line apart to find the challenge, and compose a digest
2700 authorization header. See RFC2069 section 2.1.2. */
2702 digest_authentication_encode (const char *au, const char *user,
2703 const char *passwd, const char *method,
2706 static char *realm, *opaque, *nonce;
2711 { "realm", &realm },
2712 { "opaque", &opaque },
2717 realm = opaque = nonce = NULL;
2719 au += 6; /* skip over `Digest' */
2725 for (i = 0; i < countof (options); i++)
2727 int skip = extract_header_attr (au, options[i].name,
2728 options[i].variable);
2732 xfree_null (opaque);
2742 if (i == countof (options))
2744 while (*au && *au != '=')
2752 while (*au && *au != '\"')
2759 while (*au && *au != ',')
2764 if (!realm || !nonce || !user || !passwd || !path || !method)
2767 xfree_null (opaque);
2772 /* Calculate the digest value. */
2774 ALLOCA_MD5_CONTEXT (ctx);
2775 unsigned char hash[MD5_HASHLEN];
2776 unsigned char a1buf[MD5_HASHLEN * 2 + 1], a2buf[MD5_HASHLEN * 2 + 1];
2777 unsigned char response_digest[MD5_HASHLEN * 2 + 1];
2779 /* A1BUF = H(user ":" realm ":" password) */
2781 gen_md5_update ((unsigned char *)user, strlen (user), ctx);
2782 gen_md5_update ((unsigned char *)":", 1, ctx);
2783 gen_md5_update ((unsigned char *)realm, strlen (realm), ctx);
2784 gen_md5_update ((unsigned char *)":", 1, ctx);
2785 gen_md5_update ((unsigned char *)passwd, strlen (passwd), ctx);
2786 gen_md5_finish (ctx, hash);
2787 dump_hash (a1buf, hash);
2789 /* A2BUF = H(method ":" path) */
2791 gen_md5_update ((unsigned char *)method, strlen (method), ctx);
2792 gen_md5_update ((unsigned char *)":", 1, ctx);
2793 gen_md5_update ((unsigned char *)path, strlen (path), ctx);
2794 gen_md5_finish (ctx, hash);
2795 dump_hash (a2buf, hash);
2797 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */
2799 gen_md5_update (a1buf, MD5_HASHLEN * 2, ctx);
2800 gen_md5_update ((unsigned char *)":", 1, ctx);
2801 gen_md5_update ((unsigned char *)nonce, strlen (nonce), ctx);
2802 gen_md5_update ((unsigned char *)":", 1, ctx);
2803 gen_md5_update (a2buf, MD5_HASHLEN * 2, ctx);
2804 gen_md5_finish (ctx, hash);
2805 dump_hash (response_digest, hash);
2807 res = (char*) xmalloc (strlen (user)
2812 + 2 * MD5_HASHLEN /*strlen (response_digest)*/
2813 + (opaque ? strlen (opaque) : 0)
2815 sprintf (res, "Digest \
2816 username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
2817 user, realm, nonce, path, response_digest);
2820 char *p = res + strlen (res);
2821 strcat (p, ", opaque=\"");
2828 #endif /* USE_DIGEST */
2831 #define BEGINS_WITH(line, string_constant) \
2832 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
2833 && (ISSPACE (line[sizeof (string_constant) - 1]) \
2834 || !line[sizeof (string_constant) - 1]))
2837 known_authentication_scheme_p (const char *au)
2839 return BEGINS_WITH (au, "Basic")
2840 || BEGINS_WITH (au, "Digest")
2841 || BEGINS_WITH (au, "NTLM");
2846 /* Create the HTTP authorization request header. When the
2847 `WWW-Authenticate' response header is seen, according to the
2848 authorization scheme specified in that header (`Basic' and `Digest'
2849 are supported by the current implementation), produce an
2850 appropriate HTTP authorization request header. */
2852 create_authorization_line (const char *au, const char *user,
2853 const char *passwd, const char *method,
2856 if (0 == strncasecmp (au, "Basic", 5))
2857 return basic_authentication_encode (user, passwd);
2859 if (0 == strncasecmp (au, "Digest", 6))
2860 return digest_authentication_encode (au, user, passwd, method, path);
2861 #endif /* USE_DIGEST */
2868 if (wget_cookie_jar)
2869 cookie_jar_save (wget_cookie_jar, opt.cookies_output);
2875 xfree_null (pconn.host);
2876 if (wget_cookie_jar)
2877 cookie_jar_delete (wget_cookie_jar);