2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
3 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012 Free Software Foundation,
6 This file is part of GNU Wget.
8 GNU Wget is free software; you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation; either version 3 of the License, or
11 (at your option) any later version.
13 GNU Wget is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with Wget. If not, see <http://www.gnu.org/licenses/>.
21 Additional permission under GNU GPL version 3 section 7
23 If you modify this program, or any covered work, by linking or
24 combining it with the OpenSSL project's OpenSSL library (or a
25 modified version of that library), containing parts covered by the
26 terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
27 grants you additional permission to convey the resulting work.
28 Corresponding Source for a non-source form of such a combination
29 shall include the source code for the parts of OpenSSL used as well
30 as that of the covered work. */
55 # include "http-ntlm.h"
69 #endif /* def __VMS */
71 extern char *version_string;
75 static char *create_authorization_line (const char *, const char *,
76 const char *, const char *,
77 const char *, bool *);
78 static char *basic_authentication_encode (const char *, const char *);
79 static bool known_authentication_scheme_p (const char *, const char *);
80 static void ensure_extension (struct http_stat *, const char *, int *);
81 static void load_cookies (void);
84 # define MIN(x, y) ((x) > (y) ? (y) : (x))
88 static bool cookies_loaded_p;
89 static struct cookie_jar *wget_cookie_jar;
91 #define TEXTHTML_S "text/html"
92 #define TEXTXHTML_S "application/xhtml+xml"
93 #define TEXTCSS_S "text/css"
95 /* Some status code validation macros: */
96 #define H_10X(x) (((x) >= 100) && ((x) < 200))
97 #define H_20X(x) (((x) >= 200) && ((x) < 300))
98 #define H_PARTIAL(x) ((x) == HTTP_STATUS_PARTIAL_CONTENTS)
99 #define H_REDIRECTED(x) ((x) == HTTP_STATUS_MOVED_PERMANENTLY \
100 || (x) == HTTP_STATUS_MOVED_TEMPORARILY \
101 || (x) == HTTP_STATUS_SEE_OTHER \
102 || (x) == HTTP_STATUS_TEMPORARY_REDIRECT)
104 /* HTTP/1.0 status codes from RFC1945, provided for reference. */
105 /* Successful 2xx. */
106 #define HTTP_STATUS_OK 200
107 #define HTTP_STATUS_CREATED 201
108 #define HTTP_STATUS_ACCEPTED 202
109 #define HTTP_STATUS_NO_CONTENT 204
110 #define HTTP_STATUS_PARTIAL_CONTENTS 206
112 /* Redirection 3xx. */
113 #define HTTP_STATUS_MULTIPLE_CHOICES 300
114 #define HTTP_STATUS_MOVED_PERMANENTLY 301
115 #define HTTP_STATUS_MOVED_TEMPORARILY 302
116 #define HTTP_STATUS_SEE_OTHER 303 /* from HTTP/1.1 */
117 #define HTTP_STATUS_NOT_MODIFIED 304
118 #define HTTP_STATUS_TEMPORARY_REDIRECT 307 /* from HTTP/1.1 */
120 /* Client error 4xx. */
121 #define HTTP_STATUS_BAD_REQUEST 400
122 #define HTTP_STATUS_UNAUTHORIZED 401
123 #define HTTP_STATUS_FORBIDDEN 403
124 #define HTTP_STATUS_NOT_FOUND 404
125 #define HTTP_STATUS_RANGE_NOT_SATISFIABLE 416
127 /* Server errors 5xx. */
128 #define HTTP_STATUS_INTERNAL 500
129 #define HTTP_STATUS_NOT_IMPLEMENTED 501
130 #define HTTP_STATUS_BAD_GATEWAY 502
131 #define HTTP_STATUS_UNAVAILABLE 503
134 rel_none, rel_name, rel_value, rel_both
141 struct request_header {
143 enum rp release_policy;
145 int hcount, hcapacity;
150 /* Create a new, empty request. At least request_set_method must be
151 called before the request can be used. */
153 static struct request *
156 struct request *req = xnew0 (struct request);
158 req->headers = xnew_array (struct request_header, req->hcapacity);
162 /* Set the request's method and its arguments. METH should be a
163 literal string (or it should outlive the request) because it will
164 not be freed. ARG will be freed by request_free. */
167 request_set_method (struct request *req, const char *meth, char *arg)
173 /* Return the method string passed with the last call to
174 request_set_method. */
177 request_method (const struct request *req)
182 /* Free one header according to the release policy specified with
183 request_set_header. */
186 release_header (struct request_header *hdr)
188 switch (hdr->release_policy)
205 /* Set the request named NAME to VALUE. Specifically, this means that
206 a "NAME: VALUE\r\n" header line will be used in the request. If a
207 header with the same name previously existed in the request, its
208 value will be replaced by this one. A NULL value means do nothing.
210 RELEASE_POLICY determines whether NAME and VALUE should be released
211 (freed) with request_free. Allowed values are:
213 - rel_none - don't free NAME or VALUE
214 - rel_name - free NAME when done
215 - rel_value - free VALUE when done
216 - rel_both - free both NAME and VALUE when done
218 Setting release policy is useful when arguments come from different
219 sources. For example:
221 // Don't free literal strings!
222 request_set_header (req, "Pragma", "no-cache", rel_none);
224 // Don't free a global variable, we'll need it later.
225 request_set_header (req, "Referer", opt.referer, rel_none);
227 // Value freshly allocated, free it when done.
228 request_set_header (req, "Range",
229 aprintf ("bytes=%s-", number_to_static_string (hs->restval)),
234 request_set_header (struct request *req, const char *name, const char *value,
235 enum rp release_policy)
237 struct request_header *hdr;
242 /* A NULL value is a no-op; if freeing the name is requested,
243 free it now to avoid leaks. */
244 if (release_policy == rel_name || release_policy == rel_both)
245 xfree ((void *)name);
249 for (i = 0; i < req->hcount; i++)
251 hdr = &req->headers[i];
252 if (0 == strcasecmp (name, hdr->name))
254 /* Replace existing header. */
255 release_header (hdr);
256 hdr->name = (void *)name;
257 hdr->value = (void *)value;
258 hdr->release_policy = release_policy;
263 /* Install new header. */
265 if (req->hcount >= req->hcapacity)
267 req->hcapacity <<= 1;
268 req->headers = xrealloc (req->headers, req->hcapacity * sizeof (*hdr));
270 hdr = &req->headers[req->hcount++];
271 hdr->name = (void *)name;
272 hdr->value = (void *)value;
273 hdr->release_policy = release_policy;
276 /* Like request_set_header, but sets the whole header line, as
277 provided by the user using the `--header' option. For example,
278 request_set_user_header (req, "Foo: bar") works just like
279 request_set_header (req, "Foo", "bar"). */
282 request_set_user_header (struct request *req, const char *header)
285 const char *p = strchr (header, ':');
288 BOUNDED_TO_ALLOCA (header, p, name);
290 while (c_isspace (*p))
292 request_set_header (req, xstrdup (name), (char *) p, rel_name);
295 /* Remove the header with specified name from REQ. Returns true if
296 the header was actually removed, false otherwise. */
299 request_remove_header (struct request *req, const char *name)
302 for (i = 0; i < req->hcount; i++)
304 struct request_header *hdr = &req->headers[i];
305 if (0 == strcasecmp (name, hdr->name))
307 release_header (hdr);
308 /* Move the remaining headers by one. */
309 if (i < req->hcount - 1)
310 memmove (hdr, hdr + 1, (req->hcount - i - 1) * sizeof (*hdr));
318 #define APPEND(p, str) do { \
319 int A_len = strlen (str); \
320 memcpy (p, str, A_len); \
324 /* Construct the request and write it to FD using fd_write.
325 If warc_tmp is set to a file pointer, the request string will
326 also be written to that file. */
329 request_send (const struct request *req, int fd, FILE *warc_tmp)
331 char *request_string, *p;
332 int i, size, write_error;
334 /* Count the request size. */
337 /* METHOD " " ARG " " "HTTP/1.0" "\r\n" */
338 size += strlen (req->method) + 1 + strlen (req->arg) + 1 + 8 + 2;
340 for (i = 0; i < req->hcount; i++)
342 struct request_header *hdr = &req->headers[i];
343 /* NAME ": " VALUE "\r\n" */
344 size += strlen (hdr->name) + 2 + strlen (hdr->value) + 2;
350 p = request_string = alloca_array (char, size);
352 /* Generate the request. */
354 APPEND (p, req->method); *p++ = ' ';
355 APPEND (p, req->arg); *p++ = ' ';
356 memcpy (p, "HTTP/1.1\r\n", 10); p += 10;
358 for (i = 0; i < req->hcount; i++)
360 struct request_header *hdr = &req->headers[i];
361 APPEND (p, hdr->name);
362 *p++ = ':', *p++ = ' ';
363 APPEND (p, hdr->value);
364 *p++ = '\r', *p++ = '\n';
367 *p++ = '\r', *p++ = '\n', *p++ = '\0';
368 assert (p - request_string == size);
372 DEBUGP (("\n---request begin---\n%s---request end---\n", request_string));
374 /* Send the request to the server. */
376 write_error = fd_write (fd, request_string, size - 1, -1);
378 logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"),
380 else if (warc_tmp != NULL)
382 /* Write a copy of the data to the WARC record. */
383 int warc_tmp_written = fwrite (request_string, 1, size - 1, warc_tmp);
384 if (warc_tmp_written != size - 1)
390 /* Release the resources used by REQ. */
393 request_free (struct request *req)
396 xfree_null (req->arg);
397 for (i = 0; i < req->hcount; i++)
398 release_header (&req->headers[i]);
399 xfree_null (req->headers);
403 static struct hash_table *basic_authed_hosts;
405 /* Find out if this host has issued a Basic challenge yet; if so, give
406 * it the username, password. A temporary measure until we can get
407 * proper authentication in place. */
410 maybe_send_basic_creds (const char *hostname, const char *user,
411 const char *passwd, struct request *req)
413 bool do_challenge = false;
415 if (opt.auth_without_challenge)
417 DEBUGP (("Auth-without-challenge set, sending Basic credentials.\n"));
420 else if (basic_authed_hosts
421 && hash_table_contains(basic_authed_hosts, hostname))
423 DEBUGP (("Found %s in basic_authed_hosts.\n", quote (hostname)));
428 DEBUGP (("Host %s has not issued a general basic challenge.\n",
433 request_set_header (req, "Authorization",
434 basic_authentication_encode (user, passwd),
441 register_basic_auth_host (const char *hostname)
443 if (!basic_authed_hosts)
445 basic_authed_hosts = make_nocase_string_hash_table (1);
447 if (!hash_table_contains(basic_authed_hosts, hostname))
449 hash_table_put (basic_authed_hosts, xstrdup(hostname), NULL);
450 DEBUGP (("Inserted %s into basic_authed_hosts\n", quote (hostname)));
455 /* Send the contents of FILE_NAME to SOCK. Make sure that exactly
456 PROMISED_SIZE bytes are sent over the wire -- if the file is
457 longer, read only that much; if the file is shorter, report an error.
458 If warc_tmp is set to a file pointer, the post data will
459 also be written to that file. */
462 body_file_send (int sock, const char *file_name, wgint promised_size, FILE *warc_tmp)
464 static char chunk[8192];
469 DEBUGP (("[writing BODY file %s ... ", file_name));
471 fp = fopen (file_name, "rb");
474 while (!feof (fp) && written < promised_size)
477 int length = fread (chunk, 1, sizeof (chunk), fp);
480 towrite = MIN (promised_size - written, length);
481 write_error = fd_write (sock, chunk, towrite, -1);
487 if (warc_tmp != NULL)
489 /* Write a copy of the data to the WARC record. */
490 int warc_tmp_written = fwrite (chunk, 1, towrite, warc_tmp);
491 if (warc_tmp_written != towrite)
501 /* If we've written less than was promised, report a (probably
502 nonsensical) error rather than break the promise. */
503 if (written < promised_size)
509 assert (written == promised_size);
510 DEBUGP (("done]\n"));
514 /* Determine whether [START, PEEKED + PEEKLEN) contains an empty line.
515 If so, return the pointer to the position after the line, otherwise
516 return NULL. This is used as callback to fd_read_hunk. The data
517 between START and PEEKED has been read and cannot be "unread"; the
518 data after PEEKED has only been peeked. */
521 response_head_terminator (const char *start, const char *peeked, int peeklen)
525 /* If at first peek, verify whether HUNK starts with "HTTP". If
526 not, this is a HTTP/0.9 request and we must bail out without
528 if (start == peeked && 0 != memcmp (start, "HTTP", MIN (peeklen, 4)))
531 /* Look for "\n[\r]\n", and return the following position if found.
532 Start two chars before the current to cover the possibility that
533 part of the terminator (e.g. "\n\r") arrived in the previous
535 p = peeked - start < 2 ? start : peeked - 2;
536 end = peeked + peeklen;
538 /* Check for \n\r\n or \n\n anywhere in [p, end-2). */
539 for (; p < end - 2; p++)
542 if (p[1] == '\r' && p[2] == '\n')
544 else if (p[1] == '\n')
547 /* p==end-2: check for \n\n directly preceding END. */
548 if (p[0] == '\n' && p[1] == '\n')
554 /* The maximum size of a single HTTP response we care to read. Rather
555 than being a limit of the reader implementation, this limit
556 prevents Wget from slurping all available memory upon encountering
557 malicious or buggy server output, thus protecting the user. Define
558 it to 0 to remove the limit. */
560 #define HTTP_RESPONSE_MAX_SIZE 65536
562 /* Read the HTTP request head from FD and return it. The error
563 conditions are the same as with fd_read_hunk.
565 To support HTTP/0.9 responses, this function tries to make sure
566 that the data begins with "HTTP". If this is not the case, no data
567 is read and an empty request is returned, so that the remaining
568 data can be treated as body. */
571 read_http_response_head (int fd)
573 return fd_read_hunk (fd, response_head_terminator, 512,
574 HTTP_RESPONSE_MAX_SIZE);
578 /* The response data. */
581 /* The array of pointers that indicate where each header starts.
582 For example, given this HTTP response:
589 The headers are located like this:
591 "HTTP/1.0 200 Ok\r\nDescription: some\r\n text\r\nEtag: x\r\n\r\n"
593 headers[0] headers[1] headers[2] headers[3]
595 I.e. headers[0] points to the beginning of the request,
596 headers[1] points to the end of the first header and the
597 beginning of the second one, etc. */
599 const char **headers;
602 /* Create a new response object from the text of the HTTP response,
603 available in HEAD. That text is automatically split into
604 constituent header lines for fast retrieval using
607 static struct response *
608 resp_new (const char *head)
613 struct response *resp = xnew0 (struct response);
618 /* Empty head means that we're dealing with a headerless
619 (HTTP/0.9) response. In that case, don't set HEADERS at
624 /* Split HEAD into header lines, so that resp_header_* functions
625 don't need to do this over and over again. */
631 DO_REALLOC (resp->headers, size, count + 1, const char *);
632 resp->headers[count++] = hdr;
634 /* Break upon encountering an empty line. */
635 if (!hdr[0] || (hdr[0] == '\r' && hdr[1] == '\n') || hdr[0] == '\n')
638 /* Find the end of HDR, including continuations. */
641 const char *end = strchr (hdr, '\n');
647 while (*hdr == ' ' || *hdr == '\t');
649 DO_REALLOC (resp->headers, size, count + 1, const char *);
650 resp->headers[count] = NULL;
655 /* Locate the header named NAME in the request data, starting with
656 position START. This allows the code to loop through the request
657 data, filtering for all requests of a given name. Returns the
658 found position, or -1 for failure. The code that uses this
659 function typically looks like this:
661 for (pos = 0; (pos = resp_header_locate (...)) != -1; pos++)
662 ... do something with header ...
664 If you only care about one header, use resp_header_get instead of
668 resp_header_locate (const struct response *resp, const char *name, int start,
669 const char **begptr, const char **endptr)
672 const char **headers = resp->headers;
675 if (!headers || !headers[1])
678 name_len = strlen (name);
684 for (; headers[i + 1]; i++)
686 const char *b = headers[i];
687 const char *e = headers[i + 1];
689 && b[name_len] == ':'
690 && 0 == strncasecmp (b, name, name_len))
693 while (b < e && c_isspace (*b))
695 while (b < e && c_isspace (e[-1]))
705 /* Find and retrieve the header named NAME in the request data. If
706 found, set *BEGPTR to its starting, and *ENDPTR to its ending
707 position, and return true. Otherwise return false.
709 This function is used as a building block for resp_header_copy
710 and resp_header_strdup. */
713 resp_header_get (const struct response *resp, const char *name,
714 const char **begptr, const char **endptr)
716 int pos = resp_header_locate (resp, name, 0, begptr, endptr);
720 /* Copy the response header named NAME to buffer BUF, no longer than
721 BUFSIZE (BUFSIZE includes the terminating 0). If the header
722 exists, true is returned, false otherwise. If there should be no
723 limit on the size of the header, use resp_header_strdup instead.
725 If BUFSIZE is 0, no data is copied, but the boolean indication of
726 whether the header is present is still returned. */
729 resp_header_copy (const struct response *resp, const char *name,
730 char *buf, int bufsize)
733 if (!resp_header_get (resp, name, &b, &e))
737 int len = MIN (e - b, bufsize - 1);
738 memcpy (buf, b, len);
744 /* Return the value of header named NAME in RESP, allocated with
745 malloc. If such a header does not exist in RESP, return NULL. */
748 resp_header_strdup (const struct response *resp, const char *name)
751 if (!resp_header_get (resp, name, &b, &e))
753 return strdupdelim (b, e);
756 /* Parse the HTTP status line, which is of format:
758 HTTP-Version SP Status-Code SP Reason-Phrase
760 The function returns the status-code, or -1 if the status line
761 appears malformed. The pointer to "reason-phrase" message is
762 returned in *MESSAGE. */
765 resp_status (const struct response *resp, char **message)
772 /* For a HTTP/0.9 response, assume status 200. */
774 *message = xstrdup (_("No headers, assuming HTTP/0.9"));
778 p = resp->headers[0];
779 end = resp->headers[1];
785 if (end - p < 4 || 0 != strncmp (p, "HTTP", 4))
789 /* Match the HTTP version. This is optional because Gnutella
790 servers have been reported to not specify HTTP version. */
791 if (p < end && *p == '/')
794 while (p < end && c_isdigit (*p))
796 if (p < end && *p == '.')
798 while (p < end && c_isdigit (*p))
802 while (p < end && c_isspace (*p))
804 if (end - p < 3 || !c_isdigit (p[0]) || !c_isdigit (p[1]) || !c_isdigit (p[2]))
807 status = 100 * (p[0] - '0') + 10 * (p[1] - '0') + (p[2] - '0');
812 while (p < end && c_isspace (*p))
814 while (p < end && c_isspace (end[-1]))
816 *message = strdupdelim (p, end);
822 /* Release the resources used by RESP. */
825 resp_free (struct response *resp)
827 xfree_null (resp->headers);
831 /* Print a single line of response, the characters [b, e). We tried
833 logprintf (LOG_VERBOSE, "%s%.*s\n", prefix, (int) (e - b), b);
834 but that failed to escape the non-printable characters and, in fact,
835 caused crashes in UTF-8 locales. */
838 print_response_line(const char *prefix, const char *b, const char *e)
841 BOUNDED_TO_ALLOCA(b, e, copy);
842 logprintf (LOG_ALWAYS, "%s%s\n", prefix,
843 quotearg_style (escape_quoting_style, copy));
846 /* Print the server response, line by line, omitting the trailing CRLF
847 from individual header lines, and prefixed with PREFIX. */
850 print_server_response (const struct response *resp, const char *prefix)
855 for (i = 0; resp->headers[i + 1]; i++)
857 const char *b = resp->headers[i];
858 const char *e = resp->headers[i + 1];
860 if (b < e && e[-1] == '\n')
862 if (b < e && e[-1] == '\r')
864 print_response_line(prefix, b, e);
868 /* Parse the `Content-Range' header and extract the information it
869 contains. Returns true if successful, false otherwise. */
871 parse_content_range (const char *hdr, wgint *first_byte_ptr,
872 wgint *last_byte_ptr, wgint *entity_length_ptr)
876 /* Ancient versions of Netscape proxy server, presumably predating
877 rfc2068, sent out `Content-Range' without the "bytes"
879 if (0 == strncasecmp (hdr, "bytes", 5))
882 /* "JavaWebServer/1.1.1" sends "bytes: x-y/z", contrary to the
886 while (c_isspace (*hdr))
891 if (!c_isdigit (*hdr))
893 for (num = 0; c_isdigit (*hdr); hdr++)
894 num = 10 * num + (*hdr - '0');
895 if (*hdr != '-' || !c_isdigit (*(hdr + 1)))
897 *first_byte_ptr = num;
899 for (num = 0; c_isdigit (*hdr); hdr++)
900 num = 10 * num + (*hdr - '0');
901 if (*hdr != '/' || !c_isdigit (*(hdr + 1)))
903 *last_byte_ptr = num;
908 for (num = 0; c_isdigit (*hdr); hdr++)
909 num = 10 * num + (*hdr - '0');
910 *entity_length_ptr = num;
914 /* Read the body of the request, but don't store it anywhere and don't
915 display a progress gauge. This is useful for reading the bodies of
916 administrative responses to which we will soon issue another
917 request. The response is not useful to the user, but reading it
918 allows us to continue using the same connection to the server.
920 If reading fails, false is returned, true otherwise. In debug
921 mode, the body is displayed for debugging purposes. */
924 skip_short_body (int fd, wgint contlen, bool chunked)
927 SKIP_SIZE = 512, /* size of the download buffer */
928 SKIP_THRESHOLD = 4096 /* the largest size we read */
930 wgint remaining_chunk_size = 0;
931 char dlbuf[SKIP_SIZE + 1];
932 dlbuf[SKIP_SIZE] = '\0'; /* so DEBUGP can safely print it */
934 assert (contlen != -1 || contlen);
936 /* If the body is too large, it makes more sense to simply close the
937 connection than to try to read the body. */
938 if (contlen > SKIP_THRESHOLD)
941 while (contlen > 0 || chunked)
946 if (remaining_chunk_size == 0)
948 char *line = fd_read_line (fd);
953 remaining_chunk_size = strtol (line, &endl, 16);
956 if (remaining_chunk_size == 0)
958 line = fd_read_line (fd);
964 contlen = MIN (remaining_chunk_size, SKIP_SIZE);
967 DEBUGP (("Skipping %s bytes of body: [", number_to_static_string (contlen)));
969 ret = fd_read (fd, dlbuf, MIN (contlen, SKIP_SIZE), -1);
972 /* Don't normally report the error since this is an
973 optimization that should be invisible to the user. */
974 DEBUGP (("] aborting (%s).\n",
975 ret < 0 ? fd_errstr (fd) : "EOF received"));
982 remaining_chunk_size -= ret;
983 if (remaining_chunk_size == 0)
985 char *line = fd_read_line (fd);
993 /* Safe even if %.*s bogusly expects terminating \0 because
994 we've zero-terminated dlbuf above. */
995 DEBUGP (("%.*s", ret, dlbuf));
998 DEBUGP (("] done.\n"));
1002 #define NOT_RFC2231 0
1003 #define RFC2231_NOENCODING 1
1004 #define RFC2231_ENCODING 2
1006 /* extract_param extracts the parameter name into NAME.
1007 However, if the parameter name is in RFC2231 format then
1008 this function adjusts NAME by stripping of the trailing
1009 characters that are not part of the name but are present to
1010 indicate the presence of encoding information in the value
1011 or a fragment of a long parameter value
1014 modify_param_name(param_token *name)
1016 const char *delim1 = memchr (name->b, '*', name->e - name->b);
1017 const char *delim2 = memrchr (name->b, '*', name->e - name->b);
1023 result = NOT_RFC2231;
1025 else if(delim1 == delim2)
1027 if ((name->e - 1) == delim1)
1029 result = RFC2231_ENCODING;
1033 result = RFC2231_NOENCODING;
1040 result = RFC2231_ENCODING;
1045 /* extract_param extract the paramater value into VALUE.
1046 Like modify_param_name this function modifies VALUE by
1047 stripping off the encoding information from the actual value
1050 modify_param_value (param_token *value, int encoding_type )
1052 if (RFC2231_ENCODING == encoding_type)
1054 const char *delim = memrchr (value->b, '\'', value->e - value->b);
1055 if ( delim != NULL )
1057 value->b = (delim+1);
1062 /* Extract a parameter from the string (typically an HTTP header) at
1063 **SOURCE and advance SOURCE to the next parameter. Return false
1064 when there are no more parameters to extract. The name of the
1065 parameter is returned in NAME, and the value in VALUE. If the
1066 parameter has no value, the token's value is zeroed out.
1068 For example, if *SOURCE points to the string "attachment;
1069 filename=\"foo bar\"", the first call to this function will return
1070 the token named "attachment" and no value, and the second call will
1071 return the token named "filename" and value "foo bar". The third
1072 call will return false, indicating no more valid tokens. */
1075 extract_param (const char **source, param_token *name, param_token *value,
1078 const char *p = *source;
1080 while (c_isspace (*p)) ++p;
1084 return false; /* no error; nothing more to extract */
1089 while (*p && !c_isspace (*p) && *p != '=' && *p != separator) ++p;
1091 if (name->b == name->e)
1092 return false; /* empty name: error */
1093 while (c_isspace (*p)) ++p;
1094 if (*p == separator || !*p) /* no value */
1097 if (*p == separator) ++p;
1102 return false; /* error */
1104 /* *p is '=', extract value */
1106 while (c_isspace (*p)) ++p;
1107 if (*p == '"') /* quoted */
1110 while (*p && *p != '"') ++p;
1114 /* Currently at closing quote; find the end of param. */
1115 while (c_isspace (*p)) ++p;
1116 while (*p && *p != separator) ++p;
1117 if (*p == separator)
1120 /* garbage after closed quote, e.g. foo="bar"baz */
1126 while (*p && *p != separator) ++p;
1128 while (value->e != value->b && c_isspace (value->e[-1]))
1130 if (*p == separator) ++p;
1134 int param_type = modify_param_name(name);
1135 if (NOT_RFC2231 != param_type)
1137 modify_param_value(value, param_type);
1143 #undef RFC2231_NOENCODING
1144 #undef RFC2231_ENCODING
1146 /* Appends the string represented by VALUE to FILENAME */
1149 append_value_to_filename (char **filename, param_token const * const value)
1151 int original_length = strlen(*filename);
1152 int new_length = strlen(*filename) + (value->e - value->b);
1153 *filename = xrealloc (*filename, new_length+1);
1154 memcpy (*filename + original_length, value->b, (value->e - value->b));
1155 (*filename)[new_length] = '\0';
1159 #define MAX(p, q) ((p) > (q) ? (p) : (q))
1161 /* Parse the contents of the `Content-Disposition' header, extracting
1162 the information useful to Wget. Content-Disposition is a header
1163 borrowed from MIME; when used in HTTP, it typically serves for
1164 specifying the desired file name of the resource. For example:
1166 Content-Disposition: attachment; filename="flora.jpg"
1168 Wget will skip the tokens it doesn't care about, such as
1169 "attachment" in the previous example; it will also skip other
1170 unrecognized params. If the header is syntactically correct and
1171 contains a file name, a copy of the file name is stored in
1172 *filename and true is returned. Otherwise, the function returns
1175 The file name is stripped of directory components and must not be
1178 Historically, this function returned filename prefixed with opt.dir_prefix,
1179 now that logic is handled by the caller, new code should pay attention,
1180 changed by crq, Sep 2010.
1184 parse_content_disposition (const char *hdr, char **filename)
1186 param_token name, value;
1188 while (extract_param (&hdr, &name, &value, ';'))
1190 int isFilename = BOUNDED_EQUAL_NO_CASE ( name.b, name.e, "filename" );
1191 if ( isFilename && value.b != NULL)
1193 /* Make the file name begin at the last slash or backslash. */
1194 const char *last_slash = memrchr (value.b, '/', value.e - value.b);
1195 const char *last_bs = memrchr (value.b, '\\', value.e - value.b);
1196 if (last_slash && last_bs)
1197 value.b = 1 + MAX (last_slash, last_bs);
1198 else if (last_slash || last_bs)
1199 value.b = 1 + (last_slash ? last_slash : last_bs);
1200 if (value.b == value.e)
1204 append_value_to_filename (filename, &value);
1206 *filename = strdupdelim (value.b, value.e);
1217 /* Persistent connections. Currently, we cache the most recently used
1218 connection as persistent, provided that the HTTP server agrees to
1219 make it such. The persistence data is stored in the variables
1220 below. Ideally, it should be possible to cache an arbitrary fixed
1221 number of these connections. */
1223 /* Whether a persistent connection is active. */
1224 static bool pconn_active;
1227 /* The socket of the connection. */
1230 /* Host and port of the currently active persistent connection. */
1234 /* Whether a ssl handshake has occoured on this connection. */
1237 /* Whether the connection was authorized. This is only done by
1238 NTLM, which authorizes *connections* rather than individual
1239 requests. (That practice is peculiar for HTTP, but it is a
1240 useful optimization.) */
1244 /* NTLM data of the current connection. */
1245 struct ntlmdata ntlm;
1249 /* Mark the persistent connection as invalid and free the resources it
1250 uses. This is used by the CLOSE_* macros after they forcefully
1251 close a registered persistent connection. */
1254 invalidate_persistent (void)
1256 DEBUGP (("Disabling further reuse of socket %d.\n", pconn.socket));
1257 pconn_active = false;
1258 fd_close (pconn.socket);
1263 /* Register FD, which should be a TCP/IP connection to HOST:PORT, as
1264 persistent. This will enable someone to use the same connection
1265 later. In the context of HTTP, this must be called only AFTER the
1266 response has been received and the server has promised that the
1267 connection will remain alive.
1269 If a previous connection was persistent, it is closed. */
1272 register_persistent (const char *host, int port, int fd, bool ssl)
1276 if (pconn.socket == fd)
1278 /* The connection FD is already registered. */
1283 /* The old persistent connection is still active; close it
1284 first. This situation arises whenever a persistent
1285 connection exists, but we then connect to a different
1286 host, and try to register a persistent connection to that
1288 invalidate_persistent ();
1292 pconn_active = true;
1294 pconn.host = xstrdup (host);
1297 pconn.authorized = false;
1299 DEBUGP (("Registered socket %d for persistent reuse.\n", fd));
1302 /* Return true if a persistent connection is available for connecting
1306 persistent_available_p (const char *host, int port, bool ssl,
1307 bool *host_lookup_failed)
1309 /* First, check whether a persistent connection is active at all. */
1313 /* If we want SSL and the last connection wasn't or vice versa,
1314 don't use it. Checking for host and port is not enough because
1315 HTTP and HTTPS can apparently coexist on the same port. */
1316 if (ssl != pconn.ssl)
1319 /* If we're not connecting to the same port, we're not interested. */
1320 if (port != pconn.port)
1323 /* If the host is the same, we're in business. If not, there is
1324 still hope -- read below. */
1325 if (0 != strcasecmp (host, pconn.host))
1327 /* Check if pconn.socket is talking to HOST under another name.
1328 This happens often when both sites are virtual hosts
1329 distinguished only by name and served by the same network
1330 interface, and hence the same web server (possibly set up by
1331 the ISP and serving many different web sites). This
1332 admittedly unconventional optimization does not contradict
1333 HTTP and works well with popular server software. */
1337 struct address_list *al;
1340 /* Don't try to talk to two different SSL sites over the same
1341 secure connection! (Besides, it's not clear that
1342 name-based virtual hosting is even possible with SSL.) */
1345 /* If pconn.socket's peer is one of the IP addresses HOST
1346 resolves to, pconn.socket is for all intents and purposes
1347 already talking to HOST. */
1349 if (!socket_ip_address (pconn.socket, &ip, ENDPOINT_PEER))
1351 /* Can't get the peer's address -- something must be very
1352 wrong with the connection. */
1353 invalidate_persistent ();
1356 al = lookup_host (host, 0);
1359 *host_lookup_failed = true;
1363 found = address_list_contains (al, &ip);
1364 address_list_release (al);
1369 /* The persistent connection's peer address was found among the
1370 addresses HOST resolved to; therefore, pconn.sock is in fact
1371 already talking to HOST -- no need to reconnect. */
1374 /* Finally, check whether the connection is still open. This is
1375 important because most servers implement liberal (short) timeout
1376 on persistent connections. Wget can of course always reconnect
1377 if the connection doesn't work out, but it's nicer to know in
1378 advance. This test is a logical followup of the first test, but
1379 is "expensive" and therefore placed at the end of the list.
1381 (Current implementation of test_socket_open has a nice side
1382 effect that it treats sockets with pending data as "closed".
1383 This is exactly what we want: if a broken server sends message
1384 body in response to HEAD, or if it sends more than conent-length
1385 data, we won't reuse the corrupted connection.) */
1387 if (!test_socket_open (pconn.socket))
1389 /* Oops, the socket is no longer open. Now that we know that,
1390 let's invalidate the persistent connection before returning
1392 invalidate_persistent ();
1399 /* The idea behind these two CLOSE macros is to distinguish between
1400 two cases: one when the job we've been doing is finished, and we
1401 want to close the connection and leave, and two when something is
1402 seriously wrong and we're closing the connection as part of
1405 In case of keep_alive, CLOSE_FINISH should leave the connection
1406 open, while CLOSE_INVALIDATE should still close it.
1408 Note that the semantics of the flag `keep_alive' is "this
1409 connection *will* be reused (the server has promised not to close
1410 the connection once we're done)", while the semantics of
1411 `pc_active_p && (fd) == pc_last_fd' is "we're *now* using an
1412 active, registered connection". */
1414 #define CLOSE_FINISH(fd) do { \
1417 if (pconn_active && (fd) == pconn.socket) \
1418 invalidate_persistent (); \
1427 #define CLOSE_INVALIDATE(fd) do { \
1428 if (pconn_active && (fd) == pconn.socket) \
1429 invalidate_persistent (); \
1437 wgint len; /* received length */
1438 wgint contlen; /* expected length */
1439 wgint restval; /* the restart value */
1440 int res; /* the result of last read */
1441 char *rderrmsg; /* error message from read error */
1442 char *newloc; /* new location (redirection) */
1443 char *remote_time; /* remote time-stamp string */
1444 char *error; /* textual HTTP error */
1445 int statcode; /* status code */
1446 char *message; /* status message */
1447 wgint rd_size; /* amount of data read from socket */
1448 double dltime; /* time it took to download the data */
1449 const char *referer; /* value of the referer header. */
1450 char *local_file; /* local file name. */
1451 bool existence_checked; /* true if we already checked for a file's
1452 existence after having begun to download
1453 (needed in gethttp for when connection is
1454 interrupted/restarted. */
1455 bool timestamp_checked; /* true if pre-download time-stamping checks
1456 * have already been performed */
1457 char *orig_file_name; /* name of file to compare for time-stamping
1458 * (might be != local_file if -K is set) */
1459 wgint orig_file_size; /* size of file to compare for time-stamping */
1460 time_t orig_file_tstamp; /* time-stamp of file to compare for
1465 free_hstat (struct http_stat *hs)
1467 xfree_null (hs->newloc);
1468 xfree_null (hs->remote_time);
1469 xfree_null (hs->error);
1470 xfree_null (hs->rderrmsg);
1471 xfree_null (hs->local_file);
1472 xfree_null (hs->orig_file_name);
1473 xfree_null (hs->message);
1475 /* Guard against being called twice. */
1477 hs->remote_time = NULL;
1482 get_file_flags (const char *filename, int *dt)
1484 logprintf (LOG_VERBOSE, _("\
1485 File %s already there; not retrieving.\n\n"), quote (filename));
1486 /* If the file is there, we suppose it's retrieved OK. */
1489 /* #### Bogusness alert. */
1490 /* If its suffix is "html" or "htm" or similar, assume text/html. */
1491 if (has_html_suffix_p (filename))
1495 /* Download the response body from the socket and writes it to
1496 an output file. The headers have already been read from the
1497 socket. If WARC is enabled, the response body will also be
1498 written to a WARC response record.
1500 hs, contlen, contrange, chunked_transfer_encoding and url are
1501 parameters from the gethttp method. fp is a pointer to the
1504 url, warc_timestamp_str, warc_request_uuid, warc_ip, type
1505 and statcode will be saved in the headers of the WARC record.
1506 The head parameter contains the HTTP headers of the response.
1508 If fp is NULL and WARC is enabled, the response body will be
1509 written only to the WARC file. If WARC is disabled and fp
1510 is a file pointer, the data will be written to the file.
1511 If fp is a file pointer and WARC is enabled, the body will
1512 be written to both destinations.
1514 Returns the error code. */
1516 read_response_body (struct http_stat *hs, int sock, FILE *fp, wgint contlen,
1517 wgint contrange, bool chunked_transfer_encoding,
1518 char *url, char *warc_timestamp_str, char *warc_request_uuid,
1519 ip_address *warc_ip, char *type, int statcode, char *head)
1521 int warc_payload_offset = 0;
1522 FILE *warc_tmp = NULL;
1525 if (opt.warc_filename != NULL)
1527 /* Open a temporary file where we can write the response before we
1528 add it to the WARC record. */
1529 warc_tmp = warc_tempfile ();
1530 if (warc_tmp == NULL)
1531 warcerr = WARC_TMP_FOPENERR;
1535 /* We should keep the response headers for the WARC record. */
1536 int head_len = strlen (head);
1537 int warc_tmp_written = fwrite (head, 1, head_len, warc_tmp);
1538 if (warc_tmp_written != head_len)
1539 warcerr = WARC_TMP_FWRITEERR;
1540 warc_payload_offset = head_len;
1545 if (warc_tmp != NULL)
1553 /* This confuses the timestamping code that checks for file size.
1554 #### The timestamping code should be smarter about file size. */
1555 if (opt.save_headers && hs->restval == 0)
1556 fwrite (head, 1, strlen (head), fp);
1559 /* Read the response body. */
1562 /* If content-length is present, read that much; otherwise, read
1563 until EOF. The HTTP spec doesn't require the server to
1564 actually close the connection when it's done sending data. */
1565 flags |= rb_read_exactly;
1566 if (fp != NULL && hs->restval > 0 && contrange == 0)
1567 /* If the server ignored our range request, instruct fd_read_body
1568 to skip the first RESTVAL bytes of body. */
1569 flags |= rb_skip_startpos;
1570 if (chunked_transfer_encoding)
1571 flags |= rb_chunked_transfer_encoding;
1573 hs->len = hs->restval;
1575 /* Download the response body and write it to fp.
1576 If we are working on a WARC file, we simultaneously write the
1577 response body to warc_tmp. */
1578 hs->res = fd_read_body (sock, fp, contlen != -1 ? contlen : 0,
1579 hs->restval, &hs->rd_size, &hs->len, &hs->dltime,
1583 if (warc_tmp != NULL)
1585 /* Create a response record and write it to the WARC file.
1586 Note: per the WARC standard, the request and response should share
1587 the same date header. We re-use the timestamp of the request.
1588 The response record should also refer to the uuid of the request. */
1589 bool r = warc_write_response_record (url, warc_timestamp_str,
1590 warc_request_uuid, warc_ip,
1591 warc_tmp, warc_payload_offset,
1592 type, statcode, hs->newloc);
1594 /* warc_write_response_record has closed warc_tmp. */
1600 return RETRFINISHED;
1603 if (warc_tmp != NULL)
1608 /* Error while writing to fd. */
1611 else if (hs->res == -3)
1613 /* Error while writing to warc_tmp. */
1614 return WARC_TMP_FWRITEERR;
1619 hs->rderrmsg = xstrdup (fd_errstr (sock));
1620 return RETRFINISHED;
1624 #define BEGINS_WITH(line, string_constant) \
1625 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
1626 && (c_isspace (line[sizeof (string_constant) - 1]) \
1627 || !line[sizeof (string_constant) - 1]))
1630 #define SET_USER_AGENT(req) do { \
1631 if (!opt.useragent) \
1632 request_set_header (req, "User-Agent", \
1633 aprintf ("Wget/%s (VMS %s %s)", \
1634 version_string, vms_arch(), vms_vers()), \
1636 else if (*opt.useragent) \
1637 request_set_header (req, "User-Agent", opt.useragent, rel_none); \
1639 #else /* def __VMS */
1640 #define SET_USER_AGENT(req) do { \
1641 if (!opt.useragent) \
1642 request_set_header (req, "User-Agent", \
1643 aprintf ("Wget/%s (%s)", \
1644 version_string, OS_TYPE), \
1646 else if (*opt.useragent) \
1647 request_set_header (req, "User-Agent", opt.useragent, rel_none); \
1649 #endif /* def __VMS [else] */
1651 /* The flags that allow clobbering the file (opening with "wb").
1652 Defined here to avoid repetition later. #### This will require
1654 #define ALLOW_CLOBBER (opt.noclobber || opt.always_rest || opt.timestamping \
1655 || opt.dirstruct || opt.output_document)
1657 /* Retrieve a document through HTTP protocol. It recognizes status
1658 code, and correctly handles redirections. It closes the network
1659 socket. If it receives an error from the functions below it, it
1660 will print it if there is enough information to do so (almost
1661 always), returning the error to the caller (i.e. http_loop).
1663 Various HTTP parameters are stored to hs.
1665 If PROXY is non-NULL, the connection will be made to the proxy
1666 server, and u->url will be requested. */
1668 gethttp (struct url *u, struct http_stat *hs, int *dt, struct url *proxy,
1669 struct iri *iri, int count)
1671 struct request *req;
1674 char *user, *passwd;
1678 wgint contlen, contrange;
1685 /* Set to 1 when the authorization has already been sent and should
1686 not be tried again. */
1687 bool auth_finished = false;
1689 /* Set to 1 when just globally-set Basic authorization has been sent;
1690 * should prevent further Basic negotiations, but not other
1692 bool basic_auth_finished = false;
1694 /* Whether NTLM authentication is used for this request. */
1695 bool ntlm_seen = false;
1697 /* Whether our connection to the remote host is through SSL. */
1698 bool using_ssl = false;
1700 /* Whether a HEAD request will be issued (as opposed to GET or
1702 bool head_only = !!(*dt & HEAD_ONLY);
1705 struct response *resp;
1709 /* Declare WARC variables. */
1710 bool warc_enabled = (opt.warc_filename != NULL);
1711 FILE *warc_tmp = NULL;
1712 char warc_timestamp_str [21];
1713 char warc_request_uuid [48];
1714 ip_address *warc_ip = NULL;
1715 off_t warc_payload_offset = -1;
1717 /* Whether this connection will be kept alive after the HTTP request
1721 /* Is the server using the chunked transfer encoding? */
1722 bool chunked_transfer_encoding = false;
1724 /* Whether keep-alive should be inhibited. */
1725 bool inhibit_keep_alive =
1726 !opt.http_keep_alive || opt.ignore_length;
1728 /* Headers sent when using POST. */
1729 wgint body_data_size = 0;
1731 bool host_lookup_failed = false;
1734 if (u->scheme == SCHEME_HTTPS)
1736 /* Initialize the SSL context. After this has once been done,
1737 it becomes a no-op. */
1740 scheme_disable (SCHEME_HTTPS);
1741 logprintf (LOG_NOTQUIET,
1742 _("Disabling SSL due to encountered errors.\n"));
1743 return SSLINITFAILED;
1746 #endif /* HAVE_SSL */
1748 /* Initialize certain elements of struct http_stat. */
1752 hs->rderrmsg = NULL;
1754 hs->remote_time = NULL;
1760 /* Prepare the request to send. */
1762 req = request_new ();
1765 const char *meth = "GET";
1768 else if (opt.method)
1771 for (q = opt.method; *q; ++q)
1772 *q = c_toupper (*q);
1775 /* Use the full path, i.e. one that includes the leading slash and
1776 the query string. E.g. if u->path is "foo/bar" and u->query is
1777 "param=value", full_path will be "/foo/bar?param=value". */
1780 /* When using SSL over proxy, CONNECT establishes a direct
1781 connection to the HTTPS server. Therefore use the same
1782 argument as when talking to the server directly. */
1783 && u->scheme != SCHEME_HTTPS
1786 meth_arg = xstrdup (u->url);
1788 meth_arg = url_full_path (u);
1789 request_set_method (req, meth, meth_arg);
1792 request_set_header (req, "Referer", (char *) hs->referer, rel_none);
1793 if (*dt & SEND_NOCACHE)
1795 /* Cache-Control MUST be obeyed by all HTTP/1.1 caching mechanisms... */
1796 request_set_header (req, "Cache-Control", "no-cache, must-revalidate", rel_none);
1798 /* ... but some HTTP/1.0 caches doesn't implement Cache-Control. */
1799 request_set_header (req, "Pragma", "no-cache", rel_none);
1802 request_set_header (req, "Range",
1803 aprintf ("bytes=%s-",
1804 number_to_static_string (hs->restval)),
1806 SET_USER_AGENT (req);
1807 request_set_header (req, "Accept", "*/*", rel_none);
1809 /* Find the username and password for authentication. */
1812 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
1813 user = user ? user : (opt.http_user ? opt.http_user : opt.user);
1814 passwd = passwd ? passwd : (opt.http_passwd ? opt.http_passwd : opt.passwd);
1816 /* We only do "site-wide" authentication with "global" user/password
1817 * values unless --auth-no-challange has been requested; URL user/password
1818 * info overrides. */
1819 if (user && passwd && (!u->user || opt.auth_without_challenge))
1821 /* If this is a host for which we've already received a Basic
1822 * challenge, we'll go ahead and send Basic authentication creds. */
1823 basic_auth_finished = maybe_send_basic_creds(u->host, user, passwd, req);
1826 /* Generate the Host header, HOST:PORT. Take into account that:
1828 - Broken server-side software often doesn't recognize the PORT
1829 argument, so we must generate "Host: www.server.com" instead of
1830 "Host: www.server.com:80" (and likewise for https port).
1832 - IPv6 addresses contain ":", so "Host: 3ffe:8100:200:2::2:1234"
1833 becomes ambiguous and needs to be rewritten as "Host:
1834 [3ffe:8100:200:2::2]:1234". */
1836 /* Formats arranged for hfmt[add_port][add_squares]. */
1837 static const char *hfmt[][2] = {
1838 { "%s", "[%s]" }, { "%s:%d", "[%s]:%d" }
1840 int add_port = u->port != scheme_default_port (u->scheme);
1841 int add_squares = strchr (u->host, ':') != NULL;
1842 request_set_header (req, "Host",
1843 aprintf (hfmt[add_port][add_squares], u->host, u->port),
1847 if (inhibit_keep_alive)
1848 request_set_header (req, "Connection", "Close", rel_none);
1852 request_set_header (req, "Connection", "Keep-Alive", rel_none);
1855 request_set_header (req, "Connection", "Close", rel_none);
1856 request_set_header (req, "Proxy-Connection", "Keep-Alive", rel_none);
1863 if (opt.body_data || opt.body_file)
1865 request_set_header (req, "Content-Type",
1866 "application/x-www-form-urlencoded", rel_none);
1869 body_data_size = strlen (opt.body_data);
1872 body_data_size = file_size (opt.body_file);
1873 if (body_data_size == -1)
1875 logprintf (LOG_NOTQUIET, _("BODY data file %s missing: %s\n"),
1876 quote (opt.body_file), strerror (errno));
1880 request_set_header (req, "Content-Length",
1881 xstrdup (number_to_static_string (body_data_size)),
1887 /* We need to come back here when the initial attempt to retrieve
1888 without authorization header fails. (Expected to happen at least
1889 for the Digest authorization scheme.) */
1892 request_set_header (req, "Cookie",
1893 cookie_header (wget_cookie_jar,
1894 u->host, u->port, u->path,
1896 u->scheme == SCHEME_HTTPS
1903 /* Add the user headers. */
1904 if (opt.user_headers)
1907 for (i = 0; opt.user_headers[i]; i++)
1908 request_set_user_header (req, opt.user_headers[i]);
1914 char *proxy_user, *proxy_passwd;
1915 /* For normal username and password, URL components override
1916 command-line/wgetrc parameters. With proxy
1917 authentication, it's the reverse, because proxy URLs are
1918 normally the "permanent" ones, so command-line args
1919 should take precedence. */
1920 if (opt.proxy_user && opt.proxy_passwd)
1922 proxy_user = opt.proxy_user;
1923 proxy_passwd = opt.proxy_passwd;
1927 proxy_user = proxy->user;
1928 proxy_passwd = proxy->passwd;
1930 /* #### This does not appear right. Can't the proxy request,
1931 say, `Digest' authentication? */
1932 if (proxy_user && proxy_passwd)
1933 proxyauth = basic_authentication_encode (proxy_user, proxy_passwd);
1935 /* If we're using a proxy, we will be connecting to the proxy
1939 /* Proxy authorization over SSL is handled below. */
1941 if (u->scheme != SCHEME_HTTPS)
1943 request_set_header (req, "Proxy-Authorization", proxyauth, rel_value);
1948 /* Establish the connection. */
1950 if (inhibit_keep_alive)
1954 /* Look for a persistent connection to target host, unless a
1955 proxy is used. The exception is when SSL is in use, in which
1956 case the proxy is nothing but a passthrough to the target
1957 host, registered as a connection to the latter. */
1958 struct url *relevant = conn;
1960 if (u->scheme == SCHEME_HTTPS)
1964 if (persistent_available_p (relevant->host, relevant->port,
1966 relevant->scheme == SCHEME_HTTPS,
1970 &host_lookup_failed))
1972 int family = socket_family (pconn.socket, ENDPOINT_PEER);
1973 sock = pconn.socket;
1974 using_ssl = pconn.ssl;
1976 if (family == AF_INET6)
1977 logprintf (LOG_VERBOSE, _("Reusing existing connection to [%s]:%d.\n"),
1978 quotearg_style (escape_quoting_style, pconn.host),
1982 logprintf (LOG_VERBOSE, _("Reusing existing connection to %s:%d.\n"),
1983 quotearg_style (escape_quoting_style, pconn.host),
1985 DEBUGP (("Reusing fd %d.\n", sock));
1986 if (pconn.authorized)
1987 /* If the connection is already authorized, the "Basic"
1988 authorization added by code above is unnecessary and
1990 request_remove_header (req, "Authorization");
1992 else if (host_lookup_failed)
1995 logprintf(LOG_NOTQUIET,
1996 _("%s: unable to resolve host address %s\n"),
1997 exec_name, quote (relevant->host));
2004 sock = connect_to_host (conn->host, conn->port);
2013 return (retryable_socket_connect_error (errno)
2014 ? CONERROR : CONIMPOSSIBLE);
2018 if (proxy && u->scheme == SCHEME_HTTPS)
2020 /* When requesting SSL URLs through proxies, use the
2021 CONNECT method to request passthrough. */
2022 struct request *connreq = request_new ();
2023 request_set_method (connreq, "CONNECT",
2024 aprintf ("%s:%d", u->host, u->port));
2025 SET_USER_AGENT (connreq);
2028 request_set_header (connreq, "Proxy-Authorization",
2029 proxyauth, rel_value);
2030 /* Now that PROXYAUTH is part of the CONNECT request,
2031 zero it out so we don't send proxy authorization with
2032 the regular request below. */
2035 /* Examples in rfc2817 use the Host header in CONNECT
2036 requests. I don't see how that gains anything, given
2037 that the contents of Host would be exactly the same as
2038 the contents of CONNECT. */
2040 write_error = request_send (connreq, sock, 0);
2041 request_free (connreq);
2042 if (write_error < 0)
2044 CLOSE_INVALIDATE (sock);
2049 head = read_http_response_head (sock);
2052 logprintf (LOG_VERBOSE, _("Failed reading proxy response: %s\n"),
2054 CLOSE_INVALIDATE (sock);
2064 DEBUGP (("proxy responded with: [%s]\n", head));
2066 resp = resp_new (head);
2067 statcode = resp_status (resp, &message);
2070 char *tms = datetime_str (time (NULL));
2071 logprintf (LOG_VERBOSE, "%d\n", statcode);
2072 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"), tms, statcode,
2073 quotearg_style (escape_quoting_style,
2074 _("Malformed status line")));
2079 hs->message = xstrdup (message);
2082 if (statcode != 200)
2085 logprintf (LOG_NOTQUIET, _("Proxy tunneling failed: %s"),
2086 message ? quotearg_style (escape_quoting_style, message) : "?");
2087 xfree_null (message);
2091 xfree_null (message);
2093 /* SOCK is now *really* connected to u->host, so update CONN
2094 to reflect this. That way register_persistent will
2095 register SOCK as being connected to u->host:u->port. */
2099 if (conn->scheme == SCHEME_HTTPS)
2101 if (!ssl_connect_wget (sock, u->host))
2107 else if (!ssl_check_certificate (sock, u->host))
2111 return VERIFCERTERR;
2115 #endif /* HAVE_SSL */
2118 /* Open the temporary file where we will write the request. */
2121 warc_tmp = warc_tempfile ();
2122 if (warc_tmp == NULL)
2124 CLOSE_INVALIDATE (sock);
2126 return WARC_TMP_FOPENERR;
2131 warc_ip = (ip_address *) alloca (sizeof (ip_address));
2132 socket_ip_address (sock, warc_ip, ENDPOINT_PEER);
2136 /* Send the request to server. */
2137 write_error = request_send (req, sock, warc_tmp);
2139 if (write_error >= 0)
2143 DEBUGP (("[BODY data: %s]\n", opt.body_data));
2144 write_error = fd_write (sock, opt.body_data, body_data_size, -1);
2145 if (write_error >= 0 && warc_tmp != NULL)
2147 /* Remember end of headers / start of payload. */
2148 warc_payload_offset = ftello (warc_tmp);
2150 /* Write a copy of the data to the WARC record. */
2151 int warc_tmp_written = fwrite (opt.body_data, 1, body_data_size, warc_tmp);
2152 if (warc_tmp_written != body_data_size)
2156 else if (opt.body_file && body_data_size != 0)
2158 if (warc_tmp != NULL)
2159 /* Remember end of headers / start of payload */
2160 warc_payload_offset = ftello (warc_tmp);
2162 write_error = body_file_send (sock, opt.body_file, body_data_size, warc_tmp);
2166 if (write_error < 0)
2168 CLOSE_INVALIDATE (sock);
2171 if (warc_tmp != NULL)
2174 if (write_error == -2)
2175 return WARC_TMP_FWRITEERR;
2179 logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
2180 proxy ? "Proxy" : "HTTP");
2189 /* Generate a timestamp and uuid for this request. */
2190 warc_timestamp (warc_timestamp_str);
2191 warc_uuid_str (warc_request_uuid);
2193 /* Create a request record and store it in the WARC file. */
2194 warc_result = warc_write_request_record (u->url, warc_timestamp_str,
2195 warc_request_uuid, warc_ip,
2196 warc_tmp, warc_payload_offset);
2199 CLOSE_INVALIDATE (sock);
2204 /* warc_write_request_record has also closed warc_tmp. */
2209 head = read_http_response_head (sock);
2214 logputs (LOG_NOTQUIET, _("No data received.\n"));
2215 CLOSE_INVALIDATE (sock);
2221 logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
2223 CLOSE_INVALIDATE (sock);
2228 DEBUGP (("\n---response begin---\n%s---response end---\n", head));
2230 resp = resp_new (head);
2232 /* Check for status line. */
2234 statcode = resp_status (resp, &message);
2237 char *tms = datetime_str (time (NULL));
2238 logprintf (LOG_VERBOSE, "%d\n", statcode);
2239 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"), tms, statcode,
2240 quotearg_style (escape_quoting_style,
2241 _("Malformed status line")));
2242 CLOSE_INVALIDATE (sock);
2249 if (H_10X (statcode))
2251 DEBUGP (("Ignoring response\n"));
2257 hs->message = xstrdup (message);
2258 if (!opt.server_response)
2259 logprintf (LOG_VERBOSE, "%2d %s\n", statcode,
2260 message ? quotearg_style (escape_quoting_style, message) : "");
2263 logprintf (LOG_VERBOSE, "\n");
2264 print_server_response (resp, " ");
2267 if (!opt.ignore_length
2268 && resp_header_copy (resp, "Content-Length", hdrval, sizeof (hdrval)))
2272 parsed = str_to_wgint (hdrval, NULL, 10);
2273 if (parsed == WGINT_MAX && errno == ERANGE)
2276 #### If Content-Length is out of range, it most likely
2277 means that the file is larger than 2G and that we're
2278 compiled without LFS. In that case we should probably
2279 refuse to even attempt to download the file. */
2282 else if (parsed < 0)
2284 /* Negative Content-Length; nonsensical, so we can't
2285 assume any information about the content to receive. */
2292 /* Check for keep-alive related responses. */
2293 if (!inhibit_keep_alive && contlen != -1)
2295 if (resp_header_copy (resp, "Connection", hdrval, sizeof (hdrval)))
2297 if (0 == strcasecmp (hdrval, "Close"))
2302 chunked_transfer_encoding = false;
2303 if (resp_header_copy (resp, "Transfer-Encoding", hdrval, sizeof (hdrval))
2304 && 0 == strcasecmp (hdrval, "chunked"))
2305 chunked_transfer_encoding = true;
2307 /* Handle (possibly multiple instances of) the Set-Cookie header. */
2311 const char *scbeg, *scend;
2312 /* The jar should have been created by now. */
2313 assert (wget_cookie_jar != NULL);
2315 (scpos = resp_header_locate (resp, "Set-Cookie", scpos,
2316 &scbeg, &scend)) != -1;
2319 char *set_cookie; BOUNDED_TO_ALLOCA (scbeg, scend, set_cookie);
2320 cookie_handle_set_cookie (wget_cookie_jar, u->host, u->port,
2321 u->path, set_cookie);
2326 /* The server has promised that it will not close the connection
2327 when we're done. This means that we can register it. */
2328 register_persistent (conn->host, conn->port, sock, using_ssl);
2330 if (statcode == HTTP_STATUS_UNAUTHORIZED)
2332 /* Authorization is required. */
2334 /* Normally we are not interested in the response body.
2335 But if we are writing a WARC file we are: we like to keep everyting. */
2339 type = resp_header_strdup (resp, "Content-Type");
2340 err = read_response_body (hs, sock, NULL, contlen, 0,
2341 chunked_transfer_encoding,
2342 u->url, warc_timestamp_str,
2343 warc_request_uuid, warc_ip, type,
2347 if (err != RETRFINISHED || hs->res < 0)
2349 CLOSE_INVALIDATE (sock);
2351 xfree_null (message);
2357 CLOSE_FINISH (sock);
2361 /* Since WARC is disabled, we are not interested in the response body. */
2362 if (keep_alive && !head_only
2363 && skip_short_body (sock, contlen, chunked_transfer_encoding))
2364 CLOSE_FINISH (sock);
2366 CLOSE_INVALIDATE (sock);
2369 pconn.authorized = false;
2370 if (!auth_finished && (user && passwd))
2372 /* IIS sends multiple copies of WWW-Authenticate, one with
2373 the value "negotiate", and other(s) with data. Loop over
2374 all the occurrences and pick the one we recognize. */
2376 const char *wabeg, *waend;
2377 char *www_authenticate = NULL;
2379 (wapos = resp_header_locate (resp, "WWW-Authenticate", wapos,
2380 &wabeg, &waend)) != -1;
2382 if (known_authentication_scheme_p (wabeg, waend))
2384 BOUNDED_TO_ALLOCA (wabeg, waend, www_authenticate);
2388 if (!www_authenticate)
2390 /* If the authentication header is missing or
2391 unrecognized, there's no sense in retrying. */
2392 logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
2394 else if (!basic_auth_finished
2395 || !BEGINS_WITH (www_authenticate, "Basic"))
2398 pth = url_full_path (u);
2399 request_set_header (req, "Authorization",
2400 create_authorization_line (www_authenticate,
2402 request_method (req),
2406 if (BEGINS_WITH (www_authenticate, "NTLM"))
2408 else if (!u->user && BEGINS_WITH (www_authenticate, "Basic"))
2410 /* Need to register this host as using basic auth,
2411 * so we automatically send creds next time. */
2412 register_basic_auth_host (u->host);
2415 xfree_null (message);
2418 goto retry_with_auth;
2422 /* We already did Basic auth, and it failed. Gotta
2426 logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
2428 xfree_null (message);
2433 else /* statcode != HTTP_STATUS_UNAUTHORIZED */
2435 /* Kludge: if NTLM is used, mark the TCP connection as authorized. */
2437 pconn.authorized = true;
2440 /* Determine the local filename if needed. Notice that if -O is used
2441 * hstat.local_file is set by http_loop to the argument of -O. */
2442 if (!hs->local_file)
2444 char *local_file = NULL;
2446 /* Honor Content-Disposition whether possible. */
2447 if (!opt.content_disposition
2448 || !resp_header_copy (resp, "Content-Disposition",
2449 hdrval, sizeof (hdrval))
2450 || !parse_content_disposition (hdrval, &local_file))
2452 /* The Content-Disposition header is missing or broken.
2453 * Choose unique file name according to given URL. */
2454 hs->local_file = url_file_name (u, NULL);
2458 DEBUGP (("Parsed filename from Content-Disposition: %s\n",
2460 hs->local_file = url_file_name (u, local_file);
2464 /* TODO: perform this check only once. */
2465 if (!hs->existence_checked && file_exists_p (hs->local_file))
2467 if (opt.noclobber && !opt.output_document)
2469 /* If opt.noclobber is turned on and file already exists, do not
2470 retrieve the file. But if the output_document was given, then this
2471 test was already done and the file didn't exist. Hence the !opt.output_document */
2472 get_file_flags (hs->local_file, dt);
2476 xfree_null (message);
2477 return RETRUNNEEDED;
2479 else if (!ALLOW_CLOBBER)
2481 char *unique = unique_name (hs->local_file, true);
2482 if (unique != hs->local_file)
2483 xfree (hs->local_file);
2484 hs->local_file = unique;
2487 hs->existence_checked = true;
2489 /* Support timestamping */
2490 /* TODO: move this code out of gethttp. */
2491 if (opt.timestamping && !hs->timestamp_checked)
2493 size_t filename_len = strlen (hs->local_file);
2494 char *filename_plus_orig_suffix = alloca (filename_len + sizeof (ORIG_SFX));
2495 bool local_dot_orig_file_exists = false;
2496 char *local_filename = NULL;
2499 if (opt.backup_converted)
2500 /* If -K is specified, we'll act on the assumption that it was specified
2501 last time these files were downloaded as well, and instead of just
2502 comparing local file X against server file X, we'll compare local
2503 file X.orig (if extant, else X) against server file X. If -K
2504 _wasn't_ specified last time, or the server contains files called
2505 *.orig, -N will be back to not operating correctly with -k. */
2507 /* Would a single s[n]printf() call be faster? --dan
2509 Definitely not. sprintf() is horribly slow. It's a
2510 different question whether the difference between the two
2511 affects a program. Usually I'd say "no", but at one
2512 point I profiled Wget, and found that a measurable and
2513 non-negligible amount of time was lost calling sprintf()
2514 in url.c. Replacing sprintf with inline calls to
2515 strcpy() and number_to_string() made a difference.
2517 memcpy (filename_plus_orig_suffix, hs->local_file, filename_len);
2518 memcpy (filename_plus_orig_suffix + filename_len,
2519 ORIG_SFX, sizeof (ORIG_SFX));
2521 /* Try to stat() the .orig file. */
2522 if (stat (filename_plus_orig_suffix, &st) == 0)
2524 local_dot_orig_file_exists = true;
2525 local_filename = filename_plus_orig_suffix;
2529 if (!local_dot_orig_file_exists)
2530 /* Couldn't stat() <file>.orig, so try to stat() <file>. */
2531 if (stat (hs->local_file, &st) == 0)
2532 local_filename = hs->local_file;
2534 if (local_filename != NULL)
2535 /* There was a local file, so we'll check later to see if the version
2536 the server has is the same version we already have, allowing us to
2539 hs->orig_file_name = xstrdup (local_filename);
2540 hs->orig_file_size = st.st_size;
2541 hs->orig_file_tstamp = st.st_mtime;
2543 /* Modification time granularity is 2 seconds for Windows, so
2544 increase local time by 1 second for later comparison. */
2545 ++hs->orig_file_tstamp;
2552 hs->statcode = statcode;
2554 hs->error = xstrdup (_("Malformed status line"));
2556 hs->error = xstrdup (_("(no description)"));
2558 hs->error = xstrdup (message);
2559 xfree_null (message);
2561 type = resp_header_strdup (resp, "Content-Type");
2564 char *tmp = strchr (type, ';');
2567 /* sXXXav: only needed if IRI support is enabled */
2568 char *tmp2 = tmp + 1;
2570 while (tmp > type && c_isspace (tmp[-1]))
2574 /* Try to get remote encoding if needed */
2575 if (opt.enable_iri && !opt.encoding_remote)
2577 tmp = parse_charset (tmp2);
2579 set_content_encoding (iri, tmp);
2583 hs->newloc = resp_header_strdup (resp, "Location");
2584 hs->remote_time = resp_header_strdup (resp, "Last-Modified");
2586 if (resp_header_copy (resp, "Content-Range", hdrval, sizeof (hdrval)))
2588 wgint first_byte_pos, last_byte_pos, entity_length;
2589 if (parse_content_range (hdrval, &first_byte_pos, &last_byte_pos,
2592 contrange = first_byte_pos;
2593 contlen = last_byte_pos - first_byte_pos + 1;
2598 /* 20x responses are counted among successful by default. */
2599 if (H_20X (statcode))
2602 /* Return if redirected. */
2603 if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES)
2605 /* RFC2068 says that in case of the 300 (multiple choices)
2606 response, the server can output a preferred URL through
2607 `Location' header; otherwise, the request should be treated
2608 like GET. So, if the location is set, it will be a
2609 redirection; otherwise, just proceed normally. */
2610 if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc)
2614 logprintf (LOG_VERBOSE,
2615 _("Location: %s%s\n"),
2616 hs->newloc ? escnonprint_uri (hs->newloc) : _("unspecified"),
2617 hs->newloc ? _(" [following]") : "");
2619 /* In case the caller cares to look... */
2624 /* Normally we are not interested in the response body of a redirect.
2625 But if we are writing a WARC file we are: we like to keep everyting. */
2628 int err = read_response_body (hs, sock, NULL, contlen, 0,
2629 chunked_transfer_encoding,
2630 u->url, warc_timestamp_str,
2631 warc_request_uuid, warc_ip, type,
2634 if (err != RETRFINISHED || hs->res < 0)
2636 CLOSE_INVALIDATE (sock);
2642 CLOSE_FINISH (sock);
2646 /* Since WARC is disabled, we are not interested in the response body. */
2647 if (keep_alive && !head_only
2648 && skip_short_body (sock, contlen, chunked_transfer_encoding))
2649 CLOSE_FINISH (sock);
2651 CLOSE_INVALIDATE (sock);
2656 /* From RFC2616: The status codes 303 and 307 have
2657 been added for servers that wish to make unambiguously
2658 clear which kind of reaction is expected of the client.
2660 A 307 should be redirected using the same method,
2661 in other words, a POST should be preserved and not
2662 converted to a GET in that case. */
2663 if (statcode == HTTP_STATUS_TEMPORARY_REDIRECT)
2664 return NEWLOCATION_KEEP_POST;
2669 /* If content-type is not given, assume text/html. This is because
2670 of the multitude of broken CGI's that "forget" to generate the
2673 0 == strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S)) ||
2674 0 == strncasecmp (type, TEXTXHTML_S, strlen (TEXTXHTML_S)))
2680 0 == strncasecmp (type, TEXTCSS_S, strlen (TEXTCSS_S)))
2685 if (opt.adjust_extension)
2688 /* -E / --adjust-extension / adjust_extension = on was specified,
2689 and this is a text/html file. If some case-insensitive
2690 variation on ".htm[l]" isn't already the file's suffix,
2693 ensure_extension (hs, ".html", dt);
2695 else if (*dt & TEXTCSS)
2697 ensure_extension (hs, ".css", dt);
2701 if (statcode == HTTP_STATUS_RANGE_NOT_SATISFIABLE
2702 || (!opt.timestamping && hs->restval > 0 && statcode == HTTP_STATUS_OK
2703 && contrange == 0 && contlen >= 0 && hs->restval >= contlen))
2705 /* If `-c' is in use and the file has been fully downloaded (or
2706 the remote file has shrunk), Wget effectively requests bytes
2707 after the end of file and the server response with 416
2708 (or 200 with a <= Content-Length. */
2709 logputs (LOG_VERBOSE, _("\
2710 \n The file is already fully retrieved; nothing to do.\n\n"));
2711 /* In case the caller inspects. */
2714 /* Mark as successfully retrieved. */
2717 CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
2718 might be more bytes in the body. */
2720 return RETRUNNEEDED;
2722 if ((contrange != 0 && contrange != hs->restval)
2723 || (H_PARTIAL (statcode) && !contrange))
2725 /* The Range request was somehow misunderstood by the server.
2728 CLOSE_INVALIDATE (sock);
2735 hs->contlen = contlen + contrange;
2741 /* No need to print this output if the body won't be
2742 downloaded at all, or if the original server response is
2744 logputs (LOG_VERBOSE, _("Length: "));
2747 logputs (LOG_VERBOSE, number_to_static_string (contlen + contrange));
2748 if (contlen + contrange >= 1024)
2749 logprintf (LOG_VERBOSE, " (%s)",
2750 human_readable (contlen + contrange));
2753 if (contlen >= 1024)
2754 logprintf (LOG_VERBOSE, _(", %s (%s) remaining"),
2755 number_to_static_string (contlen),
2756 human_readable (contlen));
2758 logprintf (LOG_VERBOSE, _(", %s remaining"),
2759 number_to_static_string (contlen));
2763 logputs (LOG_VERBOSE,
2764 opt.ignore_length ? _("ignored") : _("unspecified"));
2766 logprintf (LOG_VERBOSE, " [%s]\n", quotearg_style (escape_quoting_style, type));
2768 logputs (LOG_VERBOSE, "\n");
2772 /* Return if we have no intention of further downloading. */
2773 if ((!(*dt & RETROKF) && !opt.content_on_error) || head_only)
2775 /* In case the caller cares to look... */
2780 /* Normally we are not interested in the response body of a error responses.
2781 But if we are writing a WARC file we are: we like to keep everyting. */
2784 int err = read_response_body (hs, sock, NULL, contlen, 0,
2785 chunked_transfer_encoding,
2786 u->url, warc_timestamp_str,
2787 warc_request_uuid, warc_ip, type,
2790 if (err != RETRFINISHED || hs->res < 0)
2792 CLOSE_INVALIDATE (sock);
2798 CLOSE_FINISH (sock);
2802 /* Since WARC is disabled, we are not interested in the response body. */
2804 /* Pre-1.10 Wget used CLOSE_INVALIDATE here. Now we trust the
2805 servers not to send body in response to a HEAD request, and
2806 those that do will likely be caught by test_socket_open.
2807 If not, they can be worked around using
2808 `--no-http-keep-alive'. */
2809 CLOSE_FINISH (sock);
2811 && skip_short_body (sock, contlen, chunked_transfer_encoding))
2812 /* Successfully skipped the body; also keep using the socket. */
2813 CLOSE_FINISH (sock);
2815 CLOSE_INVALIDATE (sock);
2820 return RETRFINISHED;
2824 For VMS, define common fopen() optional arguments.
2827 # define FOPEN_OPT_ARGS "fop=sqo", "acc", acc_cb, &open_id
2828 # define FOPEN_BIN_FLAG 3
2829 #else /* def __VMS */
2830 # define FOPEN_BIN_FLAG true
2831 #endif /* def __VMS [else] */
2833 /* Open the local file. */
2836 mkalldirs (hs->local_file);
2838 rotate_backups (hs->local_file);
2845 fp = fopen (hs->local_file, "ab", FOPEN_OPT_ARGS);
2846 #else /* def __VMS */
2847 fp = fopen (hs->local_file, "ab");
2848 #endif /* def __VMS [else] */
2850 else if (ALLOW_CLOBBER || count > 0)
2852 if (opt.unlink && file_exists_p (hs->local_file))
2854 int res = unlink (hs->local_file);
2857 logprintf (LOG_NOTQUIET, "%s: %s\n", hs->local_file,
2859 CLOSE_INVALIDATE (sock);
2870 fp = fopen (hs->local_file, "wb", FOPEN_OPT_ARGS);
2871 #else /* def __VMS */
2872 fp = fopen (hs->local_file, "wb");
2873 #endif /* def __VMS [else] */
2877 fp = fopen_excl (hs->local_file, FOPEN_BIN_FLAG);
2878 if (!fp && errno == EEXIST)
2880 /* We cannot just invent a new name and use it (which is
2881 what functions like unique_create typically do)
2882 because we told the user we'd use this name.
2883 Instead, return and retry the download. */
2884 logprintf (LOG_NOTQUIET,
2885 _("%s has sprung into existence.\n"),
2887 CLOSE_INVALIDATE (sock);
2890 return FOPEN_EXCL_ERR;
2895 logprintf (LOG_NOTQUIET, "%s: %s\n", hs->local_file, strerror (errno));
2896 CLOSE_INVALIDATE (sock);
2905 /* Print fetch message, if opt.verbose. */
2908 logprintf (LOG_NOTQUIET, _("Saving to: %s\n"),
2909 HYPHENP (hs->local_file) ? quote ("STDOUT") : quote (hs->local_file));
2913 err = read_response_body (hs, sock, fp, contlen, contrange,
2914 chunked_transfer_encoding,
2915 u->url, warc_timestamp_str,
2916 warc_request_uuid, warc_ip, type,
2919 /* Now we no longer need to store the response header. */
2924 CLOSE_FINISH (sock);
2926 CLOSE_INVALIDATE (sock);
2934 /* The genuine HTTP loop! This is the part where the retrieval is
2935 retried, and retried, and retried, and... */
2937 http_loop (struct url *u, struct url *original_url, char **newloc,
2938 char **local_file, const char *referer, int *dt, struct url *proxy,
2942 bool got_head = false; /* used for time-stamping and filename detection */
2943 bool time_came_from_head = false;
2944 bool got_name = false;
2947 uerr_t err, ret = TRYLIMEXC;
2948 time_t tmr = -1; /* remote time-stamp */
2949 struct http_stat hstat; /* HTTP status */
2951 bool send_head_first = true;
2953 bool force_full_retrieve = false;
2956 /* If we are writing to a WARC file: always retrieve the whole file. */
2957 if (opt.warc_filename != NULL)
2958 force_full_retrieve = true;
2961 /* Assert that no value for *LOCAL_FILE was passed. */
2962 assert (local_file == NULL || *local_file == NULL);
2964 /* Set LOCAL_FILE parameter. */
2965 if (local_file && opt.output_document)
2966 *local_file = HYPHENP (opt.output_document) ? NULL : xstrdup (opt.output_document);
2968 /* Reset NEWLOC parameter. */
2971 /* This used to be done in main(), but it's a better idea to do it
2972 here so that we don't go through the hoops if we're just using
2977 /* Warn on (likely bogus) wildcard usage in HTTP. */
2978 if (opt.ftp_glob && has_wildcards_p (u->path))
2979 logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
2981 /* Setup hstat struct. */
2983 hstat.referer = referer;
2985 if (opt.output_document)
2987 hstat.local_file = xstrdup (opt.output_document);
2990 else if (!opt.content_disposition)
2993 url_file_name (opt.trustservernames ? u : original_url, NULL);
2997 if (got_name && file_exists_p (hstat.local_file) && opt.noclobber && !opt.output_document)
2999 /* If opt.noclobber is turned on and file already exists, do not
3000 retrieve the file. But if the output_document was given, then this
3001 test was already done and the file didn't exist. Hence the !opt.output_document */
3002 get_file_flags (hstat.local_file, dt);
3007 /* Reset the counter. */
3010 /* Reset the document type. */
3013 /* Skip preliminary HEAD request if we're not in spider mode. */
3015 send_head_first = false;
3017 /* Send preliminary HEAD request if --content-disposition and -c are used
3019 if (opt.content_disposition && opt.always_rest)
3020 send_head_first = true;
3022 /* Send preliminary HEAD request if -N is given and we have an existing
3023 * destination file. */
3024 file_name = url_file_name (opt.trustservernames ? u : original_url, NULL);
3025 if (opt.timestamping && (file_exists_p (file_name)
3026 || opt.content_disposition))
3027 send_head_first = true;
3033 /* Increment the pass counter. */
3035 sleep_between_retrievals (count);
3037 /* Get the current time string. */
3038 tms = datetime_str (time (NULL));
3040 if (opt.spider && !got_head)
3041 logprintf (LOG_VERBOSE, _("\
3042 Spider mode enabled. Check if remote file exists.\n"));
3044 /* Print fetch message, if opt.verbose. */
3047 char *hurl = url_string (u, URL_AUTH_HIDE_PASSWD);
3052 sprintf (tmp, _("(try:%2d)"), count);
3053 logprintf (LOG_NOTQUIET, "--%s-- %s %s\n",
3058 logprintf (LOG_NOTQUIET, "--%s-- %s\n",
3063 ws_changetitle (hurl);
3068 /* Default document type is empty. However, if spider mode is
3069 on or time-stamping is employed, HEAD_ONLY commands is
3070 encoded within *dt. */
3071 if (send_head_first && !got_head)
3076 /* Decide whether or not to restart. */
3077 if (force_full_retrieve)
3078 hstat.restval = hstat.len;
3079 else if (opt.always_rest
3081 && stat (hstat.local_file, &st) == 0
3082 && S_ISREG (st.st_mode))
3083 /* When -c is used, continue from on-disk size. (Can't use
3084 hstat.len even if count>1 because we don't want a failed
3085 first attempt to clobber existing data.) */
3086 hstat.restval = st.st_size;
3088 /* otherwise, continue where the previous try left off */
3089 hstat.restval = hstat.len;
3093 /* Decide whether to send the no-cache directive. We send it in
3095 a) we're using a proxy, and we're past our first retrieval.
3096 Some proxies are notorious for caching incomplete data, so
3097 we require a fresh get.
3098 b) caching is explicitly inhibited. */
3099 if ((proxy && count > 1) /* a */
3100 || !opt.allow_cache) /* b */
3101 *dt |= SEND_NOCACHE;
3103 *dt &= ~SEND_NOCACHE;
3105 /* Try fetching the document, or at least its head. */
3106 err = gethttp (u, &hstat, dt, proxy, iri, count);
3109 tms = datetime_str (time (NULL));
3111 /* Get the new location (with or without the redirection). */
3113 *newloc = xstrdup (hstat.newloc);
3117 case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
3118 case CONERROR: case READERR: case WRITEFAILED:
3119 case RANGEERR: case FOPEN_EXCL_ERR:
3120 /* Non-fatal errors continue executing the loop, which will
3121 bring them to "while" statement at the end, to judge
3122 whether the number of tries was exceeded. */
3123 printwhat (count, opt.ntry);
3125 case FWRITEERR: case FOPENERR:
3126 /* Another fatal error. */
3127 logputs (LOG_VERBOSE, "\n");
3128 logprintf (LOG_NOTQUIET, _("Cannot write to %s (%s).\n"),
3129 quote (hstat.local_file), strerror (errno));
3130 case HOSTERR: case CONIMPOSSIBLE: case PROXERR: case AUTHFAILED:
3131 case SSLINITFAILED: case CONTNOTSUPPORTED: case VERIFCERTERR:
3133 /* Fatal errors just return from the function. */
3137 /* A fatal WARC error. */
3138 logputs (LOG_VERBOSE, "\n");
3139 logprintf (LOG_NOTQUIET, _("Cannot write to WARC file.\n"));
3142 case WARC_TMP_FOPENERR: case WARC_TMP_FWRITEERR:
3143 /* A fatal WARC error. */
3144 logputs (LOG_VERBOSE, "\n");
3145 logprintf (LOG_NOTQUIET, _("Cannot write to temporary WARC file.\n"));
3149 /* Another fatal error. */
3150 logprintf (LOG_NOTQUIET, _("Unable to establish SSL connection.\n"));
3154 /* Another fatal error. */
3155 logputs (LOG_VERBOSE, "\n");
3156 logprintf (LOG_NOTQUIET, _("Cannot unlink %s (%s).\n"),
3157 quote (hstat.local_file), strerror (errno));
3161 case NEWLOCATION_KEEP_POST:
3162 /* Return the new location to the caller. */
3165 logprintf (LOG_NOTQUIET,
3166 _("ERROR: Redirection (%d) without location.\n"),
3176 /* The file was already fully retrieved. */
3180 /* Deal with you later. */
3183 /* All possibilities should have been exhausted. */
3187 if (!(*dt & RETROKF))
3192 /* #### Ugly ugly ugly! */
3193 hurl = url_string (u, URL_AUTH_HIDE_PASSWD);
3194 logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
3197 /* Fall back to GET if HEAD fails with a 500 or 501 error code. */
3199 && (hstat.statcode == 500 || hstat.statcode == 501))
3204 /* Maybe we should always keep track of broken links, not just in
3206 * Don't log error if it was UTF-8 encoded because we will try
3207 * once unencoded. */
3208 else if (opt.spider && !iri->utf8_encode)
3210 /* #### Again: ugly ugly ugly! */
3212 hurl = url_string (u, URL_AUTH_HIDE_PASSWD);
3213 nonexisting_url (hurl);
3214 logprintf (LOG_NOTQUIET, _("\
3215 Remote file does not exist -- broken link!!!\n"));
3219 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
3220 tms, hstat.statcode,
3221 quotearg_style (escape_quoting_style, hstat.error));
3223 logputs (LOG_VERBOSE, "\n");
3229 /* Did we get the time-stamp? */
3232 got_head = true; /* no more time-stamping */
3234 if (opt.timestamping && !hstat.remote_time)
3236 logputs (LOG_NOTQUIET, _("\
3237 Last-modified header missing -- time-stamps turned off.\n"));
3239 else if (hstat.remote_time)
3241 /* Convert the date-string into struct tm. */
3242 tmr = http_atotm (hstat.remote_time);
3243 if (tmr == (time_t) (-1))
3244 logputs (LOG_VERBOSE, _("\
3245 Last-modified header invalid -- time-stamp ignored.\n"));
3246 if (*dt & HEAD_ONLY)
3247 time_came_from_head = true;
3250 if (send_head_first)
3252 /* The time-stamping section. */
3253 if (opt.timestamping)
3255 if (hstat.orig_file_name) /* Perform the following
3256 checks only if the file
3258 download already exists. */
3260 if (hstat.remote_time &&
3261 tmr != (time_t) (-1))
3263 /* Now time-stamping can be used validly.
3264 Time-stamping means that if the sizes of
3265 the local and remote file match, and local
3266 file is newer than the remote file, it will
3267 not be retrieved. Otherwise, the normal
3268 download procedure is resumed. */
3269 if (hstat.orig_file_tstamp >= tmr)
3271 if (hstat.contlen == -1
3272 || hstat.orig_file_size == hstat.contlen)
3274 logprintf (LOG_VERBOSE, _("\
3275 Server file no newer than local file %s -- not retrieving.\n\n"),
3276 quote (hstat.orig_file_name));
3282 logprintf (LOG_VERBOSE, _("\
3283 The sizes do not match (local %s) -- retrieving.\n"),
3284 number_to_static_string (hstat.orig_file_size));
3289 force_full_retrieve = true;
3290 logputs (LOG_VERBOSE,
3291 _("Remote file is newer, retrieving.\n"));
3294 logputs (LOG_VERBOSE, "\n");
3298 /* free_hstat (&hstat); */
3299 hstat.timestamp_checked = true;
3304 bool finished = true;
3309 logputs (LOG_VERBOSE, _("\
3310 Remote file exists and could contain links to other resources -- retrieving.\n\n"));
3315 logprintf (LOG_VERBOSE, _("\
3316 Remote file exists but does not contain any link -- not retrieving.\n\n"));
3317 ret = RETROK; /* RETRUNNEEDED is not for caller. */
3324 logprintf (LOG_VERBOSE, _("\
3325 Remote file exists and could contain further links,\n\
3326 but recursion is disabled -- not retrieving.\n\n"));
3330 logprintf (LOG_VERBOSE, _("\
3331 Remote file exists.\n\n"));
3333 ret = RETROK; /* RETRUNNEEDED is not for caller. */
3338 logprintf (LOG_NONVERBOSE,
3339 _("%s URL: %s %2d %s\n"),
3340 tms, u->url, hstat.statcode,
3341 hstat.message ? quotearg_style (escape_quoting_style, hstat.message) : "");
3348 count = 0; /* the retrieve count for HEAD is reset */
3350 } /* send_head_first */
3353 if (opt.useservertimestamps
3354 && (tmr != (time_t) (-1))
3355 && ((hstat.len == hstat.contlen) ||
3356 ((hstat.res == 0) && (hstat.contlen == -1))))
3358 const char *fl = NULL;
3359 set_local_file (&fl, hstat.local_file);
3363 /* Reparse time header, in case it's changed. */
3364 if (time_came_from_head
3365 && hstat.remote_time && hstat.remote_time[0])
3367 newtmr = http_atotm (hstat.remote_time);
3368 if (newtmr != (time_t)-1)
3374 /* End of time-stamping section. */
3376 tmrate = retr_rate (hstat.rd_size, hstat.dltime);
3377 total_download_time += hstat.dltime;
3379 if (hstat.len == hstat.contlen)
3383 bool write_to_stdout = (opt.output_document && HYPHENP (opt.output_document));
3385 logprintf (LOG_VERBOSE,
3387 ? _("%s (%s) - written to stdout %s[%s/%s]\n\n")
3388 : _("%s (%s) - %s saved [%s/%s]\n\n"),
3390 write_to_stdout ? "" : quote (hstat.local_file),
3391 number_to_static_string (hstat.len),
3392 number_to_static_string (hstat.contlen));
3393 logprintf (LOG_NONVERBOSE,
3394 "%s URL:%s [%s/%s] -> \"%s\" [%d]\n",
3396 number_to_static_string (hstat.len),
3397 number_to_static_string (hstat.contlen),
3398 hstat.local_file, count);
3401 total_downloaded_bytes += hstat.rd_size;
3403 /* Remember that we downloaded the file for later ".orig" code. */
3404 if (*dt & ADDED_HTML_EXTENSION)
3405 downloaded_file (FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, hstat.local_file);
3407 downloaded_file (FILE_DOWNLOADED_NORMALLY, hstat.local_file);
3412 else if (hstat.res == 0) /* No read error */
3414 if (hstat.contlen == -1) /* We don't know how much we were supposed
3415 to get, so assume we succeeded. */
3419 bool write_to_stdout = (opt.output_document && HYPHENP (opt.output_document));
3421 logprintf (LOG_VERBOSE,
3423 ? _("%s (%s) - written to stdout %s[%s]\n\n")
3424 : _("%s (%s) - %s saved [%s]\n\n"),
3426 write_to_stdout ? "" : quote (hstat.local_file),
3427 number_to_static_string (hstat.len));
3428 logprintf (LOG_NONVERBOSE,
3429 "%s URL:%s [%s] -> \"%s\" [%d]\n",
3430 tms, u->url, number_to_static_string (hstat.len),
3431 hstat.local_file, count);
3434 total_downloaded_bytes += hstat.rd_size;
3436 /* Remember that we downloaded the file for later ".orig" code. */
3437 if (*dt & ADDED_HTML_EXTENSION)
3438 downloaded_file (FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, hstat.local_file);
3440 downloaded_file (FILE_DOWNLOADED_NORMALLY, hstat.local_file);
3445 else if (hstat.len < hstat.contlen) /* meaning we lost the
3446 connection too soon */
3448 logprintf (LOG_VERBOSE,
3449 _("%s (%s) - Connection closed at byte %s. "),
3450 tms, tmrate, number_to_static_string (hstat.len));
3451 printwhat (count, opt.ntry);
3454 else if (hstat.len != hstat.restval)
3455 /* Getting here would mean reading more data than
3456 requested with content-length, which we never do. */
3460 /* Getting here probably means that the content-length was
3461 * _less_ than the original, local size. We should probably
3462 * truncate or re-read, or something. FIXME */
3467 else /* from now on hstat.res can only be -1 */
3469 if (hstat.contlen == -1)
3471 logprintf (LOG_VERBOSE,
3472 _("%s (%s) - Read error at byte %s (%s)."),
3473 tms, tmrate, number_to_static_string (hstat.len),
3475 printwhat (count, opt.ntry);
3478 else /* hstat.res == -1 and contlen is given */
3480 logprintf (LOG_VERBOSE,
3481 _("%s (%s) - Read error at byte %s/%s (%s). "),
3483 number_to_static_string (hstat.len),
3484 number_to_static_string (hstat.contlen),
3486 printwhat (count, opt.ntry);
3492 while (!opt.ntry || (count < opt.ntry));
3495 if (ret == RETROK && local_file)
3496 *local_file = xstrdup (hstat.local_file);
3497 free_hstat (&hstat);
3502 /* Check whether the result of strptime() indicates success.
3503 strptime() returns the pointer to how far it got to in the string.
3504 The processing has been successful if the string is at `GMT' or
3505 `+X', or at the end of the string.
3507 In extended regexp parlance, the function returns 1 if P matches
3508 "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (which strptime
3509 can return) is considered a failure and 0 is returned. */
3511 check_end (const char *p)
3515 while (c_isspace (*p))
3518 || (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
3519 || ((p[0] == '+' || p[0] == '-') && c_isdigit (p[1])))
3525 /* Convert the textual specification of time in TIME_STRING to the
3526 number of seconds since the Epoch.
3528 TIME_STRING can be in any of the three formats RFC2616 allows the
3529 HTTP servers to emit -- RFC1123-date, RFC850-date or asctime-date,
3530 as well as the time format used in the Set-Cookie header.
3531 Timezones are ignored, and should be GMT.
3533 Return the computed time_t representation, or -1 if the conversion
3536 This function uses strptime with various string formats for parsing
3537 TIME_STRING. This results in a parser that is not as lenient in
3538 interpreting TIME_STRING as I would like it to be. Being based on
3539 strptime, it always allows shortened months, one-digit days, etc.,
3540 but due to the multitude of formats in which time can be
3541 represented, an ideal HTTP time parser would be even more
3542 forgiving. It should completely ignore things like week days and
3543 concentrate only on the various forms of representing years,
3544 months, days, hours, minutes, and seconds. For example, it would
3545 be nice if it accepted ISO 8601 out of the box.
3547 I've investigated free and PD code for this purpose, but none was
3548 usable. getdate was big and unwieldy, and had potential copyright
3549 issues, or so I was informed. Dr. Marcus Hennecke's atotm(),
3550 distributed with phttpd, is excellent, but we cannot use it because
3551 it is not assigned to the FSF. So I stuck it with strptime. */
3554 http_atotm (const char *time_string)
3556 /* NOTE: Solaris strptime man page claims that %n and %t match white
3557 space, but that's not universally available. Instead, we simply
3558 use ` ' to mean "skip all WS", which works under all strptime
3559 implementations I've tested. */
3561 static const char *time_formats[] = {
3562 "%a, %d %b %Y %T", /* rfc1123: Thu, 29 Jan 1998 22:12:57 */
3563 "%A, %d-%b-%y %T", /* rfc850: Thursday, 29-Jan-98 22:12:57 */
3564 "%a %b %d %T %Y", /* asctime: Thu Jan 29 22:12:57 1998 */
3565 "%a, %d-%b-%Y %T" /* cookies: Thu, 29-Jan-1998 22:12:57
3566 (used in Set-Cookie, defined in the
3567 Netscape cookie specification.) */
3569 const char *oldlocale;
3570 char savedlocale[256];
3572 time_t ret = (time_t) -1;
3574 /* Solaris strptime fails to recognize English month names in
3575 non-English locales, which we work around by temporarily setting
3576 locale to C before invoking strptime. */
3577 oldlocale = setlocale (LC_TIME, NULL);
3580 size_t l = strlen (oldlocale) + 1;
3581 if (l >= sizeof savedlocale)
3582 savedlocale[0] = '\0';
3584 memcpy (savedlocale, oldlocale, l);
3586 else savedlocale[0] = '\0';
3588 setlocale (LC_TIME, "C");
3590 for (i = 0; i < countof (time_formats); i++)
3594 /* Some versions of strptime use the existing contents of struct
3595 tm to recalculate the date according to format. Zero it out
3596 to prevent stack garbage from influencing strptime. */
3599 if (check_end (strptime (time_string, time_formats[i], &t)))
3606 /* Restore the previous locale. */
3608 setlocale (LC_TIME, savedlocale);
3613 /* Authorization support: We support three authorization schemes:
3615 * `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
3617 * `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
3618 consisting of answering to the server's challenge with the proper
3621 * `NTLM' ("NT Lan Manager") scheme, based on code written by Daniel
3622 Stenberg for libcurl. Like digest, NTLM is based on a
3623 challenge-response mechanism, but unlike digest, it is non-standard
3624 (authenticates TCP connections rather than requests), undocumented
3625 and Microsoft-specific. */
3627 /* Create the authentication header contents for the `Basic' scheme.
3628 This is done by encoding the string "USER:PASS" to base64 and
3629 prepending the string "Basic " in front of it. */
3632 basic_authentication_encode (const char *user, const char *passwd)
3635 int len1 = strlen (user) + 1 + strlen (passwd);
3637 t1 = (char *)alloca (len1 + 1);
3638 sprintf (t1, "%s:%s", user, passwd);
3640 t2 = (char *)alloca (BASE64_LENGTH (len1) + 1);
3641 base64_encode (t1, len1, t2);
3643 return concat_strings ("Basic ", t2, (char *) 0);
3646 #define SKIP_WS(x) do { \
3647 while (c_isspace (*(x))) \
3651 #ifdef ENABLE_DIGEST
3652 /* Dump the hexadecimal representation of HASH to BUF. HASH should be
3653 an array of 16 bytes containing the hash keys, and BUF should be a
3654 buffer of 33 writable characters (32 for hex digits plus one for
3655 zero termination). */
3657 dump_hash (char *buf, const unsigned char *hash)
3661 for (i = 0; i < MD5_DIGEST_SIZE; i++, hash++)
3663 *buf++ = XNUM_TO_digit (*hash >> 4);
3664 *buf++ = XNUM_TO_digit (*hash & 0xf);
3669 /* Take the line apart to find the challenge, and compose a digest
3670 authorization header. See RFC2069 section 2.1.2. */
3672 digest_authentication_encode (const char *au, const char *user,
3673 const char *passwd, const char *method,
3676 static char *realm, *opaque, *nonce, *qop, *algorithm;
3681 { "realm", &realm },
3682 { "opaque", &opaque },
3683 { "nonce", &nonce },
3685 { "algorithm", &algorithm }
3687 char cnonce[16] = "";
3691 param_token name, value;
3694 realm = opaque = nonce = qop = algorithm = NULL;
3696 au += 6; /* skip over `Digest' */
3697 while (extract_param (&au, &name, &value, ','))
3700 size_t namelen = name.e - name.b;
3701 for (i = 0; i < countof (options); i++)
3702 if (namelen == strlen (options[i].name)
3703 && 0 == strncmp (name.b, options[i].name,
3706 *options[i].variable = strdupdelim (value.b, value.e);
3711 if (qop != NULL && strcmp(qop,"auth"))
3713 logprintf (LOG_NOTQUIET, _("Unsupported quality of protection '%s'.\n"), qop);
3714 user = NULL; /* force freeing mem and return */
3717 if (algorithm != NULL && strcmp (algorithm,"MD5") && strcmp (algorithm,"MD5-sess"))
3719 logprintf (LOG_NOTQUIET, _("Unsupported algorithm '%s'.\n"), algorithm);
3720 user = NULL; /* force freeing mem and return */
3723 if (!realm || !nonce || !user || !passwd || !path || !method)
3726 xfree_null (opaque);
3729 xfree_null (algorithm);
3733 /* Calculate the digest value. */
3736 unsigned char hash[MD5_DIGEST_SIZE];
3737 char a1buf[MD5_DIGEST_SIZE * 2 + 1], a2buf[MD5_DIGEST_SIZE * 2 + 1];
3738 char response_digest[MD5_DIGEST_SIZE * 2 + 1];
3740 /* A1BUF = H(user ":" realm ":" password) */
3741 md5_init_ctx (&ctx);
3742 md5_process_bytes ((unsigned char *)user, strlen (user), &ctx);
3743 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3744 md5_process_bytes ((unsigned char *)realm, strlen (realm), &ctx);
3745 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3746 md5_process_bytes ((unsigned char *)passwd, strlen (passwd), &ctx);
3747 md5_finish_ctx (&ctx, hash);
3749 dump_hash (a1buf, hash);
3751 if (! strcmp (algorithm, "MD5-sess"))
3753 /* A1BUF = H( H(user ":" realm ":" password) ":" nonce ":" cnonce ) */
3754 snprintf (cnonce, sizeof (cnonce), "%08x", random_number(INT_MAX));
3756 md5_init_ctx (&ctx);
3757 // md5_process_bytes (hash, MD5_DIGEST_SIZE, &ctx);
3758 md5_process_bytes (a1buf, MD5_DIGEST_SIZE * 2, &ctx);
3759 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3760 md5_process_bytes ((unsigned char *)nonce, strlen (nonce), &ctx);
3761 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3762 md5_process_bytes ((unsigned char *)cnonce, strlen (cnonce), &ctx);
3763 md5_finish_ctx (&ctx, hash);
3765 dump_hash (a1buf, hash);
3768 /* A2BUF = H(method ":" path) */
3769 md5_init_ctx (&ctx);
3770 md5_process_bytes ((unsigned char *)method, strlen (method), &ctx);
3771 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3772 md5_process_bytes ((unsigned char *)path, strlen (path), &ctx);
3773 md5_finish_ctx (&ctx, hash);
3774 dump_hash (a2buf, hash);
3776 if (!strcmp(qop, "auth") || !strcmp (qop, "auth-int"))
3778 /* RFC 2617 Digest Access Authentication */
3779 /* generate random hex string */
3781 snprintf(cnonce, sizeof(cnonce), "%08x", random_number(INT_MAX));
3783 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" noncecount ":" clientnonce ":" qop ": " A2BUF) */
3784 md5_init_ctx (&ctx);
3785 md5_process_bytes ((unsigned char *)a1buf, MD5_DIGEST_SIZE * 2, &ctx);
3786 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3787 md5_process_bytes ((unsigned char *)nonce, strlen (nonce), &ctx);
3788 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3789 md5_process_bytes ((unsigned char *)"00000001", 8, &ctx); /* TODO: keep track of server nonce values */
3790 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3791 md5_process_bytes ((unsigned char *)cnonce, strlen(cnonce), &ctx);
3792 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3793 md5_process_bytes ((unsigned char *)qop, strlen(qop), &ctx);
3794 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3795 md5_process_bytes ((unsigned char *)a2buf, MD5_DIGEST_SIZE * 2, &ctx);
3796 md5_finish_ctx (&ctx, hash);
3800 /* RFC 2069 Digest Access Authentication */
3801 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */
3802 md5_init_ctx (&ctx);
3803 md5_process_bytes ((unsigned char *)a1buf, MD5_DIGEST_SIZE * 2, &ctx);
3804 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3805 md5_process_bytes ((unsigned char *)nonce, strlen (nonce), &ctx);
3806 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3807 md5_process_bytes ((unsigned char *)a2buf, MD5_DIGEST_SIZE * 2, &ctx);
3808 md5_finish_ctx (&ctx, hash);
3811 dump_hash (response_digest, hash);
3813 res_size = strlen (user)
3817 + 2 * MD5_DIGEST_SIZE /*strlen (response_digest)*/
3818 + (opaque ? strlen (opaque) : 0)
3819 + (algorithm ? strlen (algorithm) : 0)
3824 res = xmalloc (res_size);
3826 if (!strcmp(qop,"auth"))
3828 res_len = snprintf (res, res_size, "Digest "\
3829 "username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\""\
3830 ", qop=auth, nc=00000001, cnonce=\"%s\"",
3831 user, realm, nonce, path, response_digest, cnonce);
3836 res_len = snprintf (res, res_size, "Digest "\
3837 "username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
3838 user, realm, nonce, path, response_digest);
3843 res_len += snprintf(res + res_len, res_size - res_len, ", opaque=\"%s\"", opaque);
3848 snprintf(res + res_len, res_size - res_len, ", algorithm=\"%s\"", algorithm);
3853 #endif /* ENABLE_DIGEST */
3855 /* Computing the size of a string literal must take into account that
3856 value returned by sizeof includes the terminating \0. */
3857 #define STRSIZE(literal) (sizeof (literal) - 1)
3859 /* Whether chars in [b, e) begin with the literal string provided as
3860 first argument and are followed by whitespace or terminating \0.
3861 The comparison is case-insensitive. */
3862 #define STARTS(literal, b, e) \
3864 && ((size_t) ((e) - (b))) >= STRSIZE (literal) \
3865 && 0 == strncasecmp (b, literal, STRSIZE (literal)) \
3866 && ((size_t) ((e) - (b)) == STRSIZE (literal) \
3867 || c_isspace (b[STRSIZE (literal)])))
3870 known_authentication_scheme_p (const char *hdrbeg, const char *hdrend)
3872 return STARTS ("Basic", hdrbeg, hdrend)
3873 #ifdef ENABLE_DIGEST
3874 || STARTS ("Digest", hdrbeg, hdrend)
3877 || STARTS ("NTLM", hdrbeg, hdrend)
3884 /* Create the HTTP authorization request header. When the
3885 `WWW-Authenticate' response header is seen, according to the
3886 authorization scheme specified in that header (`Basic' and `Digest'
3887 are supported by the current implementation), produce an
3888 appropriate HTTP authorization request header. */
3890 create_authorization_line (const char *au, const char *user,
3891 const char *passwd, const char *method,
3892 const char *path, bool *finished)
3894 /* We are called only with known schemes, so we can dispatch on the
3896 switch (c_toupper (*au))
3898 case 'B': /* Basic */
3900 return basic_authentication_encode (user, passwd);
3901 #ifdef ENABLE_DIGEST
3902 case 'D': /* Digest */
3904 return digest_authentication_encode (au, user, passwd, method, path);
3907 case 'N': /* NTLM */
3908 if (!ntlm_input (&pconn.ntlm, au))
3913 return ntlm_output (&pconn.ntlm, user, passwd, finished);
3916 /* We shouldn't get here -- this function should be only called
3917 with values approved by known_authentication_scheme_p. */
3925 if (!wget_cookie_jar)
3926 wget_cookie_jar = cookie_jar_new ();
3927 if (opt.cookies_input && !cookies_loaded_p)
3929 cookie_jar_load (wget_cookie_jar, opt.cookies_input);
3930 cookies_loaded_p = true;
3937 if (wget_cookie_jar)
3938 cookie_jar_save (wget_cookie_jar, opt.cookies_output);
3944 xfree_null (pconn.host);
3945 if (wget_cookie_jar)
3946 cookie_jar_delete (wget_cookie_jar);
3950 ensure_extension (struct http_stat *hs, const char *ext, int *dt)
3952 char *last_period_in_local_filename = strrchr (hs->local_file, '.');
3954 int len = strlen (ext);
3957 strncpy (shortext, ext, len - 1);
3958 shortext[len - 1] = '\0';
3961 if (last_period_in_local_filename == NULL
3962 || !(0 == strcasecmp (last_period_in_local_filename, shortext)
3963 || 0 == strcasecmp (last_period_in_local_filename, ext)))
3965 int local_filename_len = strlen (hs->local_file);
3966 /* Resize the local file, allowing for ".html" preceded by
3967 optional ".NUMBER". */
3968 hs->local_file = xrealloc (hs->local_file,
3969 local_filename_len + 24 + len);
3970 strcpy (hs->local_file + local_filename_len, ext);
3971 /* If clobbering is not allowed and the file, as named,
3972 exists, tack on ".NUMBER.html" instead. */
3973 if (!ALLOW_CLOBBER && file_exists_p (hs->local_file))
3977 sprintf (hs->local_file + local_filename_len,
3978 ".%d%s", ext_num++, ext);
3979 while (file_exists_p (hs->local_file));
3981 *dt |= ADDED_HTML_EXTENSION;
3989 test_parse_content_disposition()
3997 { "filename=\"file.ext\"", "file.ext", true },
3998 { "attachment; filename=\"file.ext\"", "file.ext", true },
3999 { "attachment; filename=\"file.ext\"; dummy", "file.ext", true },
4000 { "attachment", NULL, false },
4001 { "attachement; filename*=UTF-8'en-US'hello.txt", "hello.txt", true },
4002 { "attachement; filename*0=\"hello\"; filename*1=\"world.txt\"", "helloworld.txt", true },
4005 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
4010 res = parse_content_disposition (test_array[i].hdrval, &filename);
4012 mu_assert ("test_parse_content_disposition: wrong result",
4013 res == test_array[i].result
4015 || 0 == strcmp (test_array[i].filename, filename)));
4021 #endif /* TESTING */
4024 * vim: et sts=2 sw=2 cino+={s