2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
3 2005, 2006, 2007, 2008, 2009, 2010, 2011 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, char *name, 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)
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);
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++];
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, 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 post_file (int sock, const char *file_name, wgint promised_size, FILE *warc_tmp)
464 static char chunk[8192];
469 DEBUGP (("[writing POST 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);
954 if (remaining_chunk_size == 0)
961 contlen = MIN (remaining_chunk_size, SKIP_SIZE);
964 DEBUGP (("Skipping %s bytes of body: [", number_to_static_string (contlen)));
966 ret = fd_read (fd, dlbuf, MIN (contlen, SKIP_SIZE), -1);
969 /* Don't normally report the error since this is an
970 optimization that should be invisible to the user. */
971 DEBUGP (("] aborting (%s).\n",
972 ret < 0 ? fd_errstr (fd) : "EOF received"));
979 remaining_chunk_size -= ret;
980 if (remaining_chunk_size == 0)
981 if (fd_read_line (fd) == NULL)
985 /* Safe even if %.*s bogusly expects terminating \0 because
986 we've zero-terminated dlbuf above. */
987 DEBUGP (("%.*s", ret, dlbuf));
990 DEBUGP (("] done.\n"));
994 #define NOT_RFC2231 0
995 #define RFC2231_NOENCODING 1
996 #define RFC2231_ENCODING 2
998 /* extract_param extracts the parameter name into NAME.
999 However, if the parameter name is in RFC2231 format then
1000 this function adjusts NAME by stripping of the trailing
1001 characters that are not part of the name but are present to
1002 indicate the presence of encoding information in the value
1003 or a fragment of a long parameter value
1006 modify_param_name(param_token *name)
1008 const char *delim1 = memchr (name->b, '*', name->e - name->b);
1009 const char *delim2 = memrchr (name->b, '*', name->e - name->b);
1015 result = NOT_RFC2231;
1017 else if(delim1 == delim2)
1019 if ((name->e - 1) == delim1)
1021 result = RFC2231_ENCODING;
1025 result = RFC2231_NOENCODING;
1032 result = RFC2231_ENCODING;
1037 /* extract_param extract the paramater value into VALUE.
1038 Like modify_param_name this function modifies VALUE by
1039 stripping off the encoding information from the actual value
1042 modify_param_value (param_token *value, int encoding_type )
1044 if (RFC2231_ENCODING == encoding_type)
1046 const char *delim = memrchr (value->b, '\'', value->e - value->b);
1047 if ( delim != NULL )
1049 value->b = (delim+1);
1054 /* Extract a parameter from the string (typically an HTTP header) at
1055 **SOURCE and advance SOURCE to the next parameter. Return false
1056 when there are no more parameters to extract. The name of the
1057 parameter is returned in NAME, and the value in VALUE. If the
1058 parameter has no value, the token's value is zeroed out.
1060 For example, if *SOURCE points to the string "attachment;
1061 filename=\"foo bar\"", the first call to this function will return
1062 the token named "attachment" and no value, and the second call will
1063 return the token named "filename" and value "foo bar". The third
1064 call will return false, indicating no more valid tokens. */
1067 extract_param (const char **source, param_token *name, param_token *value,
1070 const char *p = *source;
1072 while (c_isspace (*p)) ++p;
1076 return false; /* no error; nothing more to extract */
1081 while (*p && !c_isspace (*p) && *p != '=' && *p != separator) ++p;
1083 if (name->b == name->e)
1084 return false; /* empty name: error */
1085 while (c_isspace (*p)) ++p;
1086 if (*p == separator || !*p) /* no value */
1089 if (*p == separator) ++p;
1094 return false; /* error */
1096 /* *p is '=', extract value */
1098 while (c_isspace (*p)) ++p;
1099 if (*p == '"') /* quoted */
1102 while (*p && *p != '"') ++p;
1106 /* Currently at closing quote; find the end of param. */
1107 while (c_isspace (*p)) ++p;
1108 while (*p && *p != separator) ++p;
1109 if (*p == separator)
1112 /* garbage after closed quote, e.g. foo="bar"baz */
1118 while (*p && *p != separator) ++p;
1120 while (value->e != value->b && c_isspace (value->e[-1]))
1122 if (*p == separator) ++p;
1126 int param_type = modify_param_name(name);
1127 if (NOT_RFC2231 != param_type)
1129 modify_param_value(value, param_type);
1135 #undef RFC2231_NOENCODING
1136 #undef RFC2231_ENCODING
1138 /* Appends the string represented by VALUE to FILENAME */
1141 append_value_to_filename (char **filename, param_token const * const value)
1143 int original_length = strlen(*filename);
1144 int new_length = strlen(*filename) + (value->e - value->b);
1145 *filename = xrealloc (*filename, new_length+1);
1146 memcpy (*filename + original_length, value->b, (value->e - value->b));
1147 (*filename)[new_length] = '\0';
1151 #define MAX(p, q) ((p) > (q) ? (p) : (q))
1153 /* Parse the contents of the `Content-Disposition' header, extracting
1154 the information useful to Wget. Content-Disposition is a header
1155 borrowed from MIME; when used in HTTP, it typically serves for
1156 specifying the desired file name of the resource. For example:
1158 Content-Disposition: attachment; filename="flora.jpg"
1160 Wget will skip the tokens it doesn't care about, such as
1161 "attachment" in the previous example; it will also skip other
1162 unrecognized params. If the header is syntactically correct and
1163 contains a file name, a copy of the file name is stored in
1164 *filename and true is returned. Otherwise, the function returns
1167 The file name is stripped of directory components and must not be
1170 Historically, this function returned filename prefixed with opt.dir_prefix,
1171 now that logic is handled by the caller, new code should pay attention,
1172 changed by crq, Sep 2010.
1176 parse_content_disposition (const char *hdr, char **filename)
1178 param_token name, value;
1180 while (extract_param (&hdr, &name, &value, ';'))
1182 int isFilename = BOUNDED_EQUAL_NO_CASE ( name.b, name.e, "filename" );
1183 if ( isFilename && value.b != NULL)
1185 /* Make the file name begin at the last slash or backslash. */
1186 const char *last_slash = memrchr (value.b, '/', value.e - value.b);
1187 const char *last_bs = memrchr (value.b, '\\', value.e - value.b);
1188 if (last_slash && last_bs)
1189 value.b = 1 + MAX (last_slash, last_bs);
1190 else if (last_slash || last_bs)
1191 value.b = 1 + (last_slash ? last_slash : last_bs);
1192 if (value.b == value.e)
1196 append_value_to_filename (filename, &value);
1198 *filename = strdupdelim (value.b, value.e);
1209 /* Persistent connections. Currently, we cache the most recently used
1210 connection as persistent, provided that the HTTP server agrees to
1211 make it such. The persistence data is stored in the variables
1212 below. Ideally, it should be possible to cache an arbitrary fixed
1213 number of these connections. */
1215 /* Whether a persistent connection is active. */
1216 static bool pconn_active;
1219 /* The socket of the connection. */
1222 /* Host and port of the currently active persistent connection. */
1226 /* Whether a ssl handshake has occoured on this connection. */
1229 /* Whether the connection was authorized. This is only done by
1230 NTLM, which authorizes *connections* rather than individual
1231 requests. (That practice is peculiar for HTTP, but it is a
1232 useful optimization.) */
1236 /* NTLM data of the current connection. */
1237 struct ntlmdata ntlm;
1241 /* Mark the persistent connection as invalid and free the resources it
1242 uses. This is used by the CLOSE_* macros after they forcefully
1243 close a registered persistent connection. */
1246 invalidate_persistent (void)
1248 DEBUGP (("Disabling further reuse of socket %d.\n", pconn.socket));
1249 pconn_active = false;
1250 fd_close (pconn.socket);
1255 /* Register FD, which should be a TCP/IP connection to HOST:PORT, as
1256 persistent. This will enable someone to use the same connection
1257 later. In the context of HTTP, this must be called only AFTER the
1258 response has been received and the server has promised that the
1259 connection will remain alive.
1261 If a previous connection was persistent, it is closed. */
1264 register_persistent (const char *host, int port, int fd, bool ssl)
1268 if (pconn.socket == fd)
1270 /* The connection FD is already registered. */
1275 /* The old persistent connection is still active; close it
1276 first. This situation arises whenever a persistent
1277 connection exists, but we then connect to a different
1278 host, and try to register a persistent connection to that
1280 invalidate_persistent ();
1284 pconn_active = true;
1286 pconn.host = xstrdup (host);
1289 pconn.authorized = false;
1291 DEBUGP (("Registered socket %d for persistent reuse.\n", fd));
1294 /* Return true if a persistent connection is available for connecting
1298 persistent_available_p (const char *host, int port, bool ssl,
1299 bool *host_lookup_failed)
1301 /* First, check whether a persistent connection is active at all. */
1305 /* If we want SSL and the last connection wasn't or vice versa,
1306 don't use it. Checking for host and port is not enough because
1307 HTTP and HTTPS can apparently coexist on the same port. */
1308 if (ssl != pconn.ssl)
1311 /* If we're not connecting to the same port, we're not interested. */
1312 if (port != pconn.port)
1315 /* If the host is the same, we're in business. If not, there is
1316 still hope -- read below. */
1317 if (0 != strcasecmp (host, pconn.host))
1319 /* Check if pconn.socket is talking to HOST under another name.
1320 This happens often when both sites are virtual hosts
1321 distinguished only by name and served by the same network
1322 interface, and hence the same web server (possibly set up by
1323 the ISP and serving many different web sites). This
1324 admittedly unconventional optimization does not contradict
1325 HTTP and works well with popular server software. */
1329 struct address_list *al;
1332 /* Don't try to talk to two different SSL sites over the same
1333 secure connection! (Besides, it's not clear that
1334 name-based virtual hosting is even possible with SSL.) */
1337 /* If pconn.socket's peer is one of the IP addresses HOST
1338 resolves to, pconn.socket is for all intents and purposes
1339 already talking to HOST. */
1341 if (!socket_ip_address (pconn.socket, &ip, ENDPOINT_PEER))
1343 /* Can't get the peer's address -- something must be very
1344 wrong with the connection. */
1345 invalidate_persistent ();
1348 al = lookup_host (host, 0);
1351 *host_lookup_failed = true;
1355 found = address_list_contains (al, &ip);
1356 address_list_release (al);
1361 /* The persistent connection's peer address was found among the
1362 addresses HOST resolved to; therefore, pconn.sock is in fact
1363 already talking to HOST -- no need to reconnect. */
1366 /* Finally, check whether the connection is still open. This is
1367 important because most servers implement liberal (short) timeout
1368 on persistent connections. Wget can of course always reconnect
1369 if the connection doesn't work out, but it's nicer to know in
1370 advance. This test is a logical followup of the first test, but
1371 is "expensive" and therefore placed at the end of the list.
1373 (Current implementation of test_socket_open has a nice side
1374 effect that it treats sockets with pending data as "closed".
1375 This is exactly what we want: if a broken server sends message
1376 body in response to HEAD, or if it sends more than conent-length
1377 data, we won't reuse the corrupted connection.) */
1379 if (!test_socket_open (pconn.socket))
1381 /* Oops, the socket is no longer open. Now that we know that,
1382 let's invalidate the persistent connection before returning
1384 invalidate_persistent ();
1391 /* The idea behind these two CLOSE macros is to distinguish between
1392 two cases: one when the job we've been doing is finished, and we
1393 want to close the connection and leave, and two when something is
1394 seriously wrong and we're closing the connection as part of
1397 In case of keep_alive, CLOSE_FINISH should leave the connection
1398 open, while CLOSE_INVALIDATE should still close it.
1400 Note that the semantics of the flag `keep_alive' is "this
1401 connection *will* be reused (the server has promised not to close
1402 the connection once we're done)", while the semantics of
1403 `pc_active_p && (fd) == pc_last_fd' is "we're *now* using an
1404 active, registered connection". */
1406 #define CLOSE_FINISH(fd) do { \
1409 if (pconn_active && (fd) == pconn.socket) \
1410 invalidate_persistent (); \
1419 #define CLOSE_INVALIDATE(fd) do { \
1420 if (pconn_active && (fd) == pconn.socket) \
1421 invalidate_persistent (); \
1429 wgint len; /* received length */
1430 wgint contlen; /* expected length */
1431 wgint restval; /* the restart value */
1432 int res; /* the result of last read */
1433 char *rderrmsg; /* error message from read error */
1434 char *newloc; /* new location (redirection) */
1435 char *remote_time; /* remote time-stamp string */
1436 char *error; /* textual HTTP error */
1437 int statcode; /* status code */
1438 char *message; /* status message */
1439 wgint rd_size; /* amount of data read from socket */
1440 double dltime; /* time it took to download the data */
1441 const char *referer; /* value of the referer header. */
1442 char *local_file; /* local file name. */
1443 bool existence_checked; /* true if we already checked for a file's
1444 existence after having begun to download
1445 (needed in gethttp for when connection is
1446 interrupted/restarted. */
1447 bool timestamp_checked; /* true if pre-download time-stamping checks
1448 * have already been performed */
1449 char *orig_file_name; /* name of file to compare for time-stamping
1450 * (might be != local_file if -K is set) */
1451 wgint orig_file_size; /* size of file to compare for time-stamping */
1452 time_t orig_file_tstamp; /* time-stamp of file to compare for
1457 free_hstat (struct http_stat *hs)
1459 xfree_null (hs->newloc);
1460 xfree_null (hs->remote_time);
1461 xfree_null (hs->error);
1462 xfree_null (hs->rderrmsg);
1463 xfree_null (hs->local_file);
1464 xfree_null (hs->orig_file_name);
1465 xfree_null (hs->message);
1467 /* Guard against being called twice. */
1469 hs->remote_time = NULL;
1474 get_file_flags (const char *filename, int *dt)
1476 logprintf (LOG_VERBOSE, _("\
1477 File %s already there; not retrieving.\n\n"), quote (filename));
1478 /* If the file is there, we suppose it's retrieved OK. */
1481 /* #### Bogusness alert. */
1482 /* If its suffix is "html" or "htm" or similar, assume text/html. */
1483 if (has_html_suffix_p (filename))
1487 /* Download the response body from the socket and writes it to
1488 an output file. The headers have already been read from the
1489 socket. If WARC is enabled, the response body will also be
1490 written to a WARC response record.
1492 hs, contlen, contrange, chunked_transfer_encoding and url are
1493 parameters from the gethttp method. fp is a pointer to the
1496 url, warc_timestamp_str, warc_request_uuid, warc_ip, type
1497 and statcode will be saved in the headers of the WARC record.
1498 The head parameter contains the HTTP headers of the response.
1500 If fp is NULL and WARC is enabled, the response body will be
1501 written only to the WARC file. If WARC is disabled and fp
1502 is a file pointer, the data will be written to the file.
1503 If fp is a file pointer and WARC is enabled, the body will
1504 be written to both destinations.
1506 Returns the error code. */
1508 read_response_body (struct http_stat *hs, int sock, FILE *fp, wgint contlen,
1509 wgint contrange, bool chunked_transfer_encoding,
1510 char *url, char *warc_timestamp_str, char *warc_request_uuid,
1511 ip_address *warc_ip, char *type, int statcode, char *head)
1513 int warc_payload_offset = 0;
1514 FILE *warc_tmp = NULL;
1517 if (opt.warc_filename != NULL)
1519 /* Open a temporary file where we can write the response before we
1520 add it to the WARC record. */
1521 warc_tmp = warc_tempfile ();
1522 if (warc_tmp == NULL)
1523 warcerr = WARC_TMP_FOPENERR;
1527 /* We should keep the response headers for the WARC record. */
1528 int head_len = strlen (head);
1529 int warc_tmp_written = fwrite (head, 1, head_len, warc_tmp);
1530 if (warc_tmp_written != head_len)
1531 warcerr = WARC_TMP_FWRITEERR;
1532 warc_payload_offset = head_len;
1537 if (warc_tmp != NULL)
1545 /* This confuses the timestamping code that checks for file size.
1546 #### The timestamping code should be smarter about file size. */
1547 if (opt.save_headers && hs->restval == 0)
1548 fwrite (head, 1, strlen (head), fp);
1551 /* Read the response body. */
1554 /* If content-length is present, read that much; otherwise, read
1555 until EOF. The HTTP spec doesn't require the server to
1556 actually close the connection when it's done sending data. */
1557 flags |= rb_read_exactly;
1558 if (fp != NULL && hs->restval > 0 && contrange == 0)
1559 /* If the server ignored our range request, instruct fd_read_body
1560 to skip the first RESTVAL bytes of body. */
1561 flags |= rb_skip_startpos;
1562 if (chunked_transfer_encoding)
1563 flags |= rb_chunked_transfer_encoding;
1565 hs->len = hs->restval;
1567 /* Download the response body and write it to fp.
1568 If we are working on a WARC file, we simultaneously write the
1569 response body to warc_tmp. */
1570 hs->res = fd_read_body (sock, fp, contlen != -1 ? contlen : 0,
1571 hs->restval, &hs->rd_size, &hs->len, &hs->dltime,
1575 if (warc_tmp != NULL)
1577 /* Create a response record and write it to the WARC file.
1578 Note: per the WARC standard, the request and response should share
1579 the same date header. We re-use the timestamp of the request.
1580 The response record should also refer to the uuid of the request. */
1581 bool r = warc_write_response_record (url, warc_timestamp_str,
1582 warc_request_uuid, warc_ip,
1583 warc_tmp, warc_payload_offset,
1584 type, statcode, hs->newloc);
1586 /* warc_write_response_record has closed warc_tmp. */
1592 return RETRFINISHED;
1595 if (warc_tmp != NULL)
1600 /* Error while writing to fd. */
1603 else if (hs->res == -3)
1605 /* Error while writing to warc_tmp. */
1606 return WARC_TMP_FWRITEERR;
1611 hs->rderrmsg = xstrdup (fd_errstr (sock));
1612 return RETRFINISHED;
1616 #define BEGINS_WITH(line, string_constant) \
1617 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
1618 && (c_isspace (line[sizeof (string_constant) - 1]) \
1619 || !line[sizeof (string_constant) - 1]))
1622 #define SET_USER_AGENT(req) do { \
1623 if (!opt.useragent) \
1624 request_set_header (req, "User-Agent", \
1625 aprintf ("Wget/%s (VMS %s %s)", \
1626 version_string, vms_arch(), vms_vers()), \
1628 else if (*opt.useragent) \
1629 request_set_header (req, "User-Agent", opt.useragent, rel_none); \
1631 #else /* def __VMS */
1632 #define SET_USER_AGENT(req) do { \
1633 if (!opt.useragent) \
1634 request_set_header (req, "User-Agent", \
1635 aprintf ("Wget/%s (%s)", \
1636 version_string, OS_TYPE), \
1638 else if (*opt.useragent) \
1639 request_set_header (req, "User-Agent", opt.useragent, rel_none); \
1641 #endif /* def __VMS [else] */
1643 /* The flags that allow clobbering the file (opening with "wb").
1644 Defined here to avoid repetition later. #### This will require
1646 #define ALLOW_CLOBBER (opt.noclobber || opt.always_rest || opt.timestamping \
1647 || opt.dirstruct || opt.output_document)
1649 /* Retrieve a document through HTTP protocol. It recognizes status
1650 code, and correctly handles redirections. It closes the network
1651 socket. If it receives an error from the functions below it, it
1652 will print it if there is enough information to do so (almost
1653 always), returning the error to the caller (i.e. http_loop).
1655 Various HTTP parameters are stored to hs.
1657 If PROXY is non-NULL, the connection will be made to the proxy
1658 server, and u->url will be requested. */
1660 gethttp (struct url *u, struct http_stat *hs, int *dt, struct url *proxy,
1661 struct iri *iri, int count)
1663 struct request *req;
1666 char *user, *passwd;
1670 wgint contlen, contrange;
1677 /* Set to 1 when the authorization has already been sent and should
1678 not be tried again. */
1679 bool auth_finished = false;
1681 /* Set to 1 when just globally-set Basic authorization has been sent;
1682 * should prevent further Basic negotiations, but not other
1684 bool basic_auth_finished = false;
1686 /* Whether NTLM authentication is used for this request. */
1687 bool ntlm_seen = false;
1689 /* Whether our connection to the remote host is through SSL. */
1690 bool using_ssl = false;
1692 /* Whether a HEAD request will be issued (as opposed to GET or
1694 bool head_only = !!(*dt & HEAD_ONLY);
1697 struct response *resp;
1701 /* Declare WARC variables. */
1702 bool warc_enabled = (opt.warc_filename != NULL);
1703 FILE *warc_tmp = NULL;
1704 char warc_timestamp_str [21];
1705 char warc_request_uuid [48];
1706 ip_address *warc_ip = NULL;
1707 long int warc_payload_offset = -1;
1709 /* Whether this connection will be kept alive after the HTTP request
1713 /* Is the server using the chunked transfer encoding? */
1714 bool chunked_transfer_encoding = false;
1716 /* Whether keep-alive should be inhibited. */
1717 bool inhibit_keep_alive =
1718 !opt.http_keep_alive || opt.ignore_length;
1720 /* Headers sent when using POST. */
1721 wgint post_data_size = 0;
1723 bool host_lookup_failed = false;
1726 if (u->scheme == SCHEME_HTTPS)
1728 /* Initialize the SSL context. After this has once been done,
1729 it becomes a no-op. */
1732 scheme_disable (SCHEME_HTTPS);
1733 logprintf (LOG_NOTQUIET,
1734 _("Disabling SSL due to encountered errors.\n"));
1735 return SSLINITFAILED;
1738 #endif /* HAVE_SSL */
1740 /* Initialize certain elements of struct http_stat. */
1744 hs->rderrmsg = NULL;
1746 hs->remote_time = NULL;
1752 /* Prepare the request to send. */
1754 req = request_new ();
1757 const char *meth = "GET";
1760 else if (opt.post_file_name || opt.post_data)
1762 /* Use the full path, i.e. one that includes the leading slash and
1763 the query string. E.g. if u->path is "foo/bar" and u->query is
1764 "param=value", full_path will be "/foo/bar?param=value". */
1767 /* When using SSL over proxy, CONNECT establishes a direct
1768 connection to the HTTPS server. Therefore use the same
1769 argument as when talking to the server directly. */
1770 && u->scheme != SCHEME_HTTPS
1773 meth_arg = xstrdup (u->url);
1775 meth_arg = url_full_path (u);
1776 request_set_method (req, meth, meth_arg);
1779 request_set_header (req, "Referer", (char *) hs->referer, rel_none);
1780 if (*dt & SEND_NOCACHE)
1782 /* Cache-Control MUST be obeyed by all HTTP/1.1 caching mechanisms... */
1783 request_set_header (req, "Cache-Control", "no-cache, must-revalidate", rel_none);
1785 /* ... but some HTTP/1.0 caches doesn't implement Cache-Control. */
1786 request_set_header (req, "Pragma", "no-cache", rel_none);
1789 request_set_header (req, "Range",
1790 aprintf ("bytes=%s-",
1791 number_to_static_string (hs->restval)),
1793 SET_USER_AGENT (req);
1794 request_set_header (req, "Accept", "*/*", rel_none);
1796 /* Find the username and password for authentication. */
1799 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
1800 user = user ? user : (opt.http_user ? opt.http_user : opt.user);
1801 passwd = passwd ? passwd : (opt.http_passwd ? opt.http_passwd : opt.passwd);
1803 /* We only do "site-wide" authentication with "global" user/password
1804 * values unless --auth-no-challange has been requested; URL user/password
1805 * info overrides. */
1806 if (user && passwd && (!u->user || opt.auth_without_challenge))
1808 /* If this is a host for which we've already received a Basic
1809 * challenge, we'll go ahead and send Basic authentication creds. */
1810 basic_auth_finished = maybe_send_basic_creds(u->host, user, passwd, req);
1813 /* Generate the Host header, HOST:PORT. Take into account that:
1815 - Broken server-side software often doesn't recognize the PORT
1816 argument, so we must generate "Host: www.server.com" instead of
1817 "Host: www.server.com:80" (and likewise for https port).
1819 - IPv6 addresses contain ":", so "Host: 3ffe:8100:200:2::2:1234"
1820 becomes ambiguous and needs to be rewritten as "Host:
1821 [3ffe:8100:200:2::2]:1234". */
1823 /* Formats arranged for hfmt[add_port][add_squares]. */
1824 static const char *hfmt[][2] = {
1825 { "%s", "[%s]" }, { "%s:%d", "[%s]:%d" }
1827 int add_port = u->port != scheme_default_port (u->scheme);
1828 int add_squares = strchr (u->host, ':') != NULL;
1829 request_set_header (req, "Host",
1830 aprintf (hfmt[add_port][add_squares], u->host, u->port),
1834 if (inhibit_keep_alive)
1835 request_set_header (req, "Connection", "Close", rel_none);
1839 request_set_header (req, "Connection", "Keep-Alive", rel_none);
1842 request_set_header (req, "Connection", "Close", rel_none);
1843 request_set_header (req, "Proxy-Connection", "Keep-Alive", rel_none);
1847 if (opt.post_data || opt.post_file_name)
1849 request_set_header (req, "Content-Type",
1850 "application/x-www-form-urlencoded", rel_none);
1852 post_data_size = strlen (opt.post_data);
1855 post_data_size = file_size (opt.post_file_name);
1856 if (post_data_size == -1)
1858 logprintf (LOG_NOTQUIET, _("POST data file %s missing: %s\n"),
1859 quote (opt.post_file_name), strerror (errno));
1863 request_set_header (req, "Content-Length",
1864 xstrdup (number_to_static_string (post_data_size)),
1869 /* We need to come back here when the initial attempt to retrieve
1870 without authorization header fails. (Expected to happen at least
1871 for the Digest authorization scheme.) */
1874 request_set_header (req, "Cookie",
1875 cookie_header (wget_cookie_jar,
1876 u->host, u->port, u->path,
1878 u->scheme == SCHEME_HTTPS
1885 /* Add the user headers. */
1886 if (opt.user_headers)
1889 for (i = 0; opt.user_headers[i]; i++)
1890 request_set_user_header (req, opt.user_headers[i]);
1896 char *proxy_user, *proxy_passwd;
1897 /* For normal username and password, URL components override
1898 command-line/wgetrc parameters. With proxy
1899 authentication, it's the reverse, because proxy URLs are
1900 normally the "permanent" ones, so command-line args
1901 should take precedence. */
1902 if (opt.proxy_user && opt.proxy_passwd)
1904 proxy_user = opt.proxy_user;
1905 proxy_passwd = opt.proxy_passwd;
1909 proxy_user = proxy->user;
1910 proxy_passwd = proxy->passwd;
1912 /* #### This does not appear right. Can't the proxy request,
1913 say, `Digest' authentication? */
1914 if (proxy_user && proxy_passwd)
1915 proxyauth = basic_authentication_encode (proxy_user, proxy_passwd);
1917 /* If we're using a proxy, we will be connecting to the proxy
1921 /* Proxy authorization over SSL is handled below. */
1923 if (u->scheme != SCHEME_HTTPS)
1925 request_set_header (req, "Proxy-Authorization", proxyauth, rel_value);
1930 /* Establish the connection. */
1932 if (inhibit_keep_alive)
1936 /* Look for a persistent connection to target host, unless a
1937 proxy is used. The exception is when SSL is in use, in which
1938 case the proxy is nothing but a passthrough to the target
1939 host, registered as a connection to the latter. */
1940 struct url *relevant = conn;
1942 if (u->scheme == SCHEME_HTTPS)
1946 if (persistent_available_p (relevant->host, relevant->port,
1948 relevant->scheme == SCHEME_HTTPS,
1952 &host_lookup_failed))
1954 sock = pconn.socket;
1955 using_ssl = pconn.ssl;
1956 logprintf (LOG_VERBOSE, _("Reusing existing connection to %s:%d.\n"),
1957 quotearg_style (escape_quoting_style, pconn.host),
1959 DEBUGP (("Reusing fd %d.\n", sock));
1960 if (pconn.authorized)
1961 /* If the connection is already authorized, the "Basic"
1962 authorization added by code above is unnecessary and
1964 request_remove_header (req, "Authorization");
1966 else if (host_lookup_failed)
1969 logprintf(LOG_NOTQUIET,
1970 _("%s: unable to resolve host address %s\n"),
1971 exec_name, quote (relevant->host));
1978 sock = connect_to_host (conn->host, conn->port);
1987 return (retryable_socket_connect_error (errno)
1988 ? CONERROR : CONIMPOSSIBLE);
1992 if (proxy && u->scheme == SCHEME_HTTPS)
1994 /* When requesting SSL URLs through proxies, use the
1995 CONNECT method to request passthrough. */
1996 struct request *connreq = request_new ();
1997 request_set_method (connreq, "CONNECT",
1998 aprintf ("%s:%d", u->host, u->port));
1999 SET_USER_AGENT (connreq);
2002 request_set_header (connreq, "Proxy-Authorization",
2003 proxyauth, rel_value);
2004 /* Now that PROXYAUTH is part of the CONNECT request,
2005 zero it out so we don't send proxy authorization with
2006 the regular request below. */
2009 /* Examples in rfc2817 use the Host header in CONNECT
2010 requests. I don't see how that gains anything, given
2011 that the contents of Host would be exactly the same as
2012 the contents of CONNECT. */
2014 write_error = request_send (connreq, sock, 0);
2015 request_free (connreq);
2016 if (write_error < 0)
2018 CLOSE_INVALIDATE (sock);
2022 head = read_http_response_head (sock);
2025 logprintf (LOG_VERBOSE, _("Failed reading proxy response: %s\n"),
2027 CLOSE_INVALIDATE (sock);
2036 DEBUGP (("proxy responded with: [%s]\n", head));
2038 resp = resp_new (head);
2039 statcode = resp_status (resp, &message);
2042 char *tms = datetime_str (time (NULL));
2043 logprintf (LOG_VERBOSE, "%d\n", statcode);
2044 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"), tms, statcode,
2045 quotearg_style (escape_quoting_style,
2046 _("Malformed status line")));
2050 hs->message = xstrdup (message);
2053 if (statcode != 200)
2056 logprintf (LOG_NOTQUIET, _("Proxy tunneling failed: %s"),
2057 message ? quotearg_style (escape_quoting_style, message) : "?");
2058 xfree_null (message);
2061 xfree_null (message);
2063 /* SOCK is now *really* connected to u->host, so update CONN
2064 to reflect this. That way register_persistent will
2065 register SOCK as being connected to u->host:u->port. */
2069 if (conn->scheme == SCHEME_HTTPS)
2071 if (!ssl_connect_wget (sock))
2076 else if (!ssl_check_certificate (sock, u->host))
2079 return VERIFCERTERR;
2083 #endif /* HAVE_SSL */
2086 /* Open the temporary file where we will write the request. */
2089 warc_tmp = warc_tempfile ();
2090 if (warc_tmp == NULL)
2092 CLOSE_INVALIDATE (sock);
2094 return WARC_TMP_FOPENERR;
2099 warc_ip = (ip_address *) alloca (sizeof (ip_address));
2100 socket_ip_address (sock, warc_ip, ENDPOINT_PEER);
2104 /* Send the request to server. */
2105 write_error = request_send (req, sock, warc_tmp);
2107 if (write_error >= 0)
2111 DEBUGP (("[POST data: %s]\n", opt.post_data));
2112 write_error = fd_write (sock, opt.post_data, post_data_size, -1);
2113 if (write_error >= 0 && warc_tmp != NULL)
2115 /* Remember end of headers / start of payload. */
2116 warc_payload_offset = ftell (warc_tmp);
2118 /* Write a copy of the data to the WARC record. */
2119 int warc_tmp_written = fwrite (opt.post_data, 1, post_data_size, warc_tmp);
2120 if (warc_tmp_written != post_data_size)
2124 else if (opt.post_file_name && post_data_size != 0)
2126 if (warc_tmp != NULL)
2127 /* Remember end of headers / start of payload. */
2128 warc_payload_offset = ftell (warc_tmp);
2130 write_error = post_file (sock, opt.post_file_name, post_data_size, warc_tmp);
2134 if (write_error < 0)
2136 CLOSE_INVALIDATE (sock);
2139 if (warc_tmp != NULL)
2142 if (write_error == -2)
2143 return WARC_TMP_FWRITEERR;
2147 logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
2148 proxy ? "Proxy" : "HTTP");
2157 /* Generate a timestamp and uuid for this request. */
2158 warc_timestamp (warc_timestamp_str);
2159 warc_uuid_str (warc_request_uuid);
2161 /* Create a request record and store it in the WARC file. */
2162 warc_result = warc_write_request_record (u->url, warc_timestamp_str,
2163 warc_request_uuid, warc_ip,
2164 warc_tmp, warc_payload_offset);
2167 CLOSE_INVALIDATE (sock);
2172 /* warc_write_request_record has also closed warc_tmp. */
2177 head = read_http_response_head (sock);
2182 logputs (LOG_NOTQUIET, _("No data received.\n"));
2183 CLOSE_INVALIDATE (sock);
2189 logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
2191 CLOSE_INVALIDATE (sock);
2196 DEBUGP (("\n---response begin---\n%s---response end---\n", head));
2198 resp = resp_new (head);
2200 /* Check for status line. */
2202 statcode = resp_status (resp, &message);
2205 char *tms = datetime_str (time (NULL));
2206 logprintf (LOG_VERBOSE, "%d\n", statcode);
2207 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"), tms, statcode,
2208 quotearg_style (escape_quoting_style,
2209 _("Malformed status line")));
2210 CLOSE_INVALIDATE (sock);
2216 if (H_10X (statcode))
2218 DEBUGP (("Ignoring response\n"));
2223 hs->message = xstrdup (message);
2224 if (!opt.server_response)
2225 logprintf (LOG_VERBOSE, "%2d %s\n", statcode,
2226 message ? quotearg_style (escape_quoting_style, message) : "");
2229 logprintf (LOG_VERBOSE, "\n");
2230 print_server_response (resp, " ");
2233 if (!opt.ignore_length
2234 && resp_header_copy (resp, "Content-Length", hdrval, sizeof (hdrval)))
2238 parsed = str_to_wgint (hdrval, NULL, 10);
2239 if (parsed == WGINT_MAX && errno == ERANGE)
2242 #### If Content-Length is out of range, it most likely
2243 means that the file is larger than 2G and that we're
2244 compiled without LFS. In that case we should probably
2245 refuse to even attempt to download the file. */
2248 else if (parsed < 0)
2250 /* Negative Content-Length; nonsensical, so we can't
2251 assume any information about the content to receive. */
2258 /* Check for keep-alive related responses. */
2259 if (!inhibit_keep_alive && contlen != -1)
2261 if (resp_header_copy (resp, "Connection", hdrval, sizeof (hdrval)))
2263 if (0 == strcasecmp (hdrval, "Close"))
2268 chunked_transfer_encoding = false;
2269 if (resp_header_copy (resp, "Transfer-Encoding", hdrval, sizeof (hdrval))
2270 && 0 == strcasecmp (hdrval, "chunked"))
2271 chunked_transfer_encoding = true;
2273 /* Handle (possibly multiple instances of) the Set-Cookie header. */
2277 const char *scbeg, *scend;
2278 /* The jar should have been created by now. */
2279 assert (wget_cookie_jar != NULL);
2281 (scpos = resp_header_locate (resp, "Set-Cookie", scpos,
2282 &scbeg, &scend)) != -1;
2285 char *set_cookie; BOUNDED_TO_ALLOCA (scbeg, scend, set_cookie);
2286 cookie_handle_set_cookie (wget_cookie_jar, u->host, u->port,
2287 u->path, set_cookie);
2292 /* The server has promised that it will not close the connection
2293 when we're done. This means that we can register it. */
2294 register_persistent (conn->host, conn->port, sock, using_ssl);
2296 if (statcode == HTTP_STATUS_UNAUTHORIZED)
2298 /* Authorization is required. */
2300 /* Normally we are not interested in the response body.
2301 But if we are writing a WARC file we are: we like to keep everyting. */
2305 type = resp_header_strdup (resp, "Content-Type");
2306 err = read_response_body (hs, sock, NULL, contlen, 0,
2307 chunked_transfer_encoding,
2308 u->url, warc_timestamp_str,
2309 warc_request_uuid, warc_ip, type,
2313 if (err != RETRFINISHED || hs->res < 0)
2315 CLOSE_INVALIDATE (sock);
2317 xfree_null (message);
2323 CLOSE_FINISH (sock);
2327 /* Since WARC is disabled, we are not interested in the response body. */
2328 if (keep_alive && !head_only
2329 && skip_short_body (sock, contlen, chunked_transfer_encoding))
2330 CLOSE_FINISH (sock);
2332 CLOSE_INVALIDATE (sock);
2335 pconn.authorized = false;
2336 if (!auth_finished && (user && passwd))
2338 /* IIS sends multiple copies of WWW-Authenticate, one with
2339 the value "negotiate", and other(s) with data. Loop over
2340 all the occurrences and pick the one we recognize. */
2342 const char *wabeg, *waend;
2343 char *www_authenticate = NULL;
2345 (wapos = resp_header_locate (resp, "WWW-Authenticate", wapos,
2346 &wabeg, &waend)) != -1;
2348 if (known_authentication_scheme_p (wabeg, waend))
2350 BOUNDED_TO_ALLOCA (wabeg, waend, www_authenticate);
2354 if (!www_authenticate)
2356 /* If the authentication header is missing or
2357 unrecognized, there's no sense in retrying. */
2358 logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
2360 else if (!basic_auth_finished
2361 || !BEGINS_WITH (www_authenticate, "Basic"))
2364 pth = url_full_path (u);
2365 request_set_header (req, "Authorization",
2366 create_authorization_line (www_authenticate,
2368 request_method (req),
2372 if (BEGINS_WITH (www_authenticate, "NTLM"))
2374 else if (!u->user && BEGINS_WITH (www_authenticate, "Basic"))
2376 /* Need to register this host as using basic auth,
2377 * so we automatically send creds next time. */
2378 register_basic_auth_host (u->host);
2381 xfree_null (message);
2384 goto retry_with_auth;
2388 /* We already did Basic auth, and it failed. Gotta
2392 logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
2394 xfree_null (message);
2399 else /* statcode != HTTP_STATUS_UNAUTHORIZED */
2401 /* Kludge: if NTLM is used, mark the TCP connection as authorized. */
2403 pconn.authorized = true;
2406 /* Determine the local filename if needed. Notice that if -O is used
2407 * hstat.local_file is set by http_loop to the argument of -O. */
2408 if (!hs->local_file)
2410 char *local_file = NULL;
2412 /* Honor Content-Disposition whether possible. */
2413 if (!opt.content_disposition
2414 || !resp_header_copy (resp, "Content-Disposition",
2415 hdrval, sizeof (hdrval))
2416 || !parse_content_disposition (hdrval, &local_file))
2418 /* The Content-Disposition header is missing or broken.
2419 * Choose unique file name according to given URL. */
2420 hs->local_file = url_file_name (u, NULL);
2424 DEBUGP (("Parsed filename from Content-Disposition: %s\n",
2426 hs->local_file = url_file_name (u, local_file);
2430 /* TODO: perform this check only once. */
2431 if (!hs->existence_checked && file_exists_p (hs->local_file))
2433 if (opt.noclobber && !opt.output_document)
2435 /* If opt.noclobber is turned on and file already exists, do not
2436 retrieve the file. But if the output_document was given, then this
2437 test was already done and the file didn't exist. Hence the !opt.output_document */
2438 get_file_flags (hs->local_file, dt);
2440 xfree_null (message);
2441 return RETRUNNEEDED;
2443 else if (!ALLOW_CLOBBER)
2445 char *unique = unique_name (hs->local_file, true);
2446 if (unique != hs->local_file)
2447 xfree (hs->local_file);
2448 hs->local_file = unique;
2451 hs->existence_checked = true;
2453 /* Support timestamping */
2454 /* TODO: move this code out of gethttp. */
2455 if (opt.timestamping && !hs->timestamp_checked)
2457 size_t filename_len = strlen (hs->local_file);
2458 char *filename_plus_orig_suffix = alloca (filename_len + sizeof (ORIG_SFX));
2459 bool local_dot_orig_file_exists = false;
2460 char *local_filename = NULL;
2463 if (opt.backup_converted)
2464 /* If -K is specified, we'll act on the assumption that it was specified
2465 last time these files were downloaded as well, and instead of just
2466 comparing local file X against server file X, we'll compare local
2467 file X.orig (if extant, else X) against server file X. If -K
2468 _wasn't_ specified last time, or the server contains files called
2469 *.orig, -N will be back to not operating correctly with -k. */
2471 /* Would a single s[n]printf() call be faster? --dan
2473 Definitely not. sprintf() is horribly slow. It's a
2474 different question whether the difference between the two
2475 affects a program. Usually I'd say "no", but at one
2476 point I profiled Wget, and found that a measurable and
2477 non-negligible amount of time was lost calling sprintf()
2478 in url.c. Replacing sprintf with inline calls to
2479 strcpy() and number_to_string() made a difference.
2481 memcpy (filename_plus_orig_suffix, hs->local_file, filename_len);
2482 memcpy (filename_plus_orig_suffix + filename_len,
2483 ORIG_SFX, sizeof (ORIG_SFX));
2485 /* Try to stat() the .orig file. */
2486 if (stat (filename_plus_orig_suffix, &st) == 0)
2488 local_dot_orig_file_exists = true;
2489 local_filename = filename_plus_orig_suffix;
2493 if (!local_dot_orig_file_exists)
2494 /* Couldn't stat() <file>.orig, so try to stat() <file>. */
2495 if (stat (hs->local_file, &st) == 0)
2496 local_filename = hs->local_file;
2498 if (local_filename != NULL)
2499 /* There was a local file, so we'll check later to see if the version
2500 the server has is the same version we already have, allowing us to
2503 hs->orig_file_name = xstrdup (local_filename);
2504 hs->orig_file_size = st.st_size;
2505 hs->orig_file_tstamp = st.st_mtime;
2507 /* Modification time granularity is 2 seconds for Windows, so
2508 increase local time by 1 second for later comparison. */
2509 ++hs->orig_file_tstamp;
2516 hs->statcode = statcode;
2518 hs->error = xstrdup (_("Malformed status line"));
2520 hs->error = xstrdup (_("(no description)"));
2522 hs->error = xstrdup (message);
2523 xfree_null (message);
2525 type = resp_header_strdup (resp, "Content-Type");
2528 char *tmp = strchr (type, ';');
2531 /* sXXXav: only needed if IRI support is enabled */
2532 char *tmp2 = tmp + 1;
2534 while (tmp > type && c_isspace (tmp[-1]))
2538 /* Try to get remote encoding if needed */
2539 if (opt.enable_iri && !opt.encoding_remote)
2541 tmp = parse_charset (tmp2);
2543 set_content_encoding (iri, tmp);
2547 hs->newloc = resp_header_strdup (resp, "Location");
2548 hs->remote_time = resp_header_strdup (resp, "Last-Modified");
2550 if (resp_header_copy (resp, "Content-Range", hdrval, sizeof (hdrval)))
2552 wgint first_byte_pos, last_byte_pos, entity_length;
2553 if (parse_content_range (hdrval, &first_byte_pos, &last_byte_pos,
2556 contrange = first_byte_pos;
2557 contlen = last_byte_pos - first_byte_pos + 1;
2562 /* 20x responses are counted among successful by default. */
2563 if (H_20X (statcode))
2566 /* Return if redirected. */
2567 if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES)
2569 /* RFC2068 says that in case of the 300 (multiple choices)
2570 response, the server can output a preferred URL through
2571 `Location' header; otherwise, the request should be treated
2572 like GET. So, if the location is set, it will be a
2573 redirection; otherwise, just proceed normally. */
2574 if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc)
2578 logprintf (LOG_VERBOSE,
2579 _("Location: %s%s\n"),
2580 hs->newloc ? escnonprint_uri (hs->newloc) : _("unspecified"),
2581 hs->newloc ? _(" [following]") : "");
2583 /* In case the caller cares to look... */
2588 /* Normally we are not interested in the response body of a redirect.
2589 But if we are writing a WARC file we are: we like to keep everyting. */
2592 int err = read_response_body (hs, sock, NULL, contlen, 0,
2593 chunked_transfer_encoding,
2594 u->url, warc_timestamp_str,
2595 warc_request_uuid, warc_ip, type,
2598 if (err != RETRFINISHED || hs->res < 0)
2600 CLOSE_INVALIDATE (sock);
2606 CLOSE_FINISH (sock);
2610 /* Since WARC is disabled, we are not interested in the response body. */
2611 if (keep_alive && !head_only
2612 && skip_short_body (sock, contlen, chunked_transfer_encoding))
2613 CLOSE_FINISH (sock);
2615 CLOSE_INVALIDATE (sock);
2620 /* From RFC2616: The status codes 303 and 307 have
2621 been added for servers that wish to make unambiguously
2622 clear which kind of reaction is expected of the client.
2624 A 307 should be redirected using the same method,
2625 in other words, a POST should be preserved and not
2626 converted to a GET in that case. */
2627 if (statcode == HTTP_STATUS_TEMPORARY_REDIRECT)
2628 return NEWLOCATION_KEEP_POST;
2633 /* If content-type is not given, assume text/html. This is because
2634 of the multitude of broken CGI's that "forget" to generate the
2637 0 == strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S)) ||
2638 0 == strncasecmp (type, TEXTXHTML_S, strlen (TEXTXHTML_S)))
2644 0 == strncasecmp (type, TEXTCSS_S, strlen (TEXTCSS_S)))
2649 if (opt.adjust_extension)
2652 /* -E / --adjust-extension / adjust_extension = on was specified,
2653 and this is a text/html file. If some case-insensitive
2654 variation on ".htm[l]" isn't already the file's suffix,
2657 ensure_extension (hs, ".html", dt);
2659 else if (*dt & TEXTCSS)
2661 ensure_extension (hs, ".css", dt);
2665 if (statcode == HTTP_STATUS_RANGE_NOT_SATISFIABLE
2666 || (!opt.timestamping && hs->restval > 0 && statcode == HTTP_STATUS_OK
2667 && contrange == 0 && contlen >= 0 && hs->restval >= contlen))
2669 /* If `-c' is in use and the file has been fully downloaded (or
2670 the remote file has shrunk), Wget effectively requests bytes
2671 after the end of file and the server response with 416
2672 (or 200 with a <= Content-Length. */
2673 logputs (LOG_VERBOSE, _("\
2674 \n The file is already fully retrieved; nothing to do.\n\n"));
2675 /* In case the caller inspects. */
2678 /* Mark as successfully retrieved. */
2681 CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
2682 might be more bytes in the body. */
2684 return RETRUNNEEDED;
2686 if ((contrange != 0 && contrange != hs->restval)
2687 || (H_PARTIAL (statcode) && !contrange))
2689 /* The Range request was somehow misunderstood by the server.
2692 CLOSE_INVALIDATE (sock);
2699 hs->contlen = contlen + contrange;
2705 /* No need to print this output if the body won't be
2706 downloaded at all, or if the original server response is
2708 logputs (LOG_VERBOSE, _("Length: "));
2711 logputs (LOG_VERBOSE, number_to_static_string (contlen + contrange));
2712 if (contlen + contrange >= 1024)
2713 logprintf (LOG_VERBOSE, " (%s)",
2714 human_readable (contlen + contrange));
2717 if (contlen >= 1024)
2718 logprintf (LOG_VERBOSE, _(", %s (%s) remaining"),
2719 number_to_static_string (contlen),
2720 human_readable (contlen));
2722 logprintf (LOG_VERBOSE, _(", %s remaining"),
2723 number_to_static_string (contlen));
2727 logputs (LOG_VERBOSE,
2728 opt.ignore_length ? _("ignored") : _("unspecified"));
2730 logprintf (LOG_VERBOSE, " [%s]\n", quotearg_style (escape_quoting_style, type));
2732 logputs (LOG_VERBOSE, "\n");
2736 /* Return if we have no intention of further downloading. */
2737 if ((!(*dt & RETROKF) && !opt.content_on_error) || head_only)
2739 /* In case the caller cares to look... */
2744 /* Normally we are not interested in the response body of a error responses.
2745 But if we are writing a WARC file we are: we like to keep everyting. */
2748 int err = read_response_body (hs, sock, NULL, contlen, 0,
2749 chunked_transfer_encoding,
2750 u->url, warc_timestamp_str,
2751 warc_request_uuid, warc_ip, type,
2754 if (err != RETRFINISHED || hs->res < 0)
2756 CLOSE_INVALIDATE (sock);
2762 CLOSE_FINISH (sock);
2766 /* Since WARC is disabled, we are not interested in the response body. */
2768 /* Pre-1.10 Wget used CLOSE_INVALIDATE here. Now we trust the
2769 servers not to send body in response to a HEAD request, and
2770 those that do will likely be caught by test_socket_open.
2771 If not, they can be worked around using
2772 `--no-http-keep-alive'. */
2773 CLOSE_FINISH (sock);
2775 && skip_short_body (sock, contlen, chunked_transfer_encoding))
2776 /* Successfully skipped the body; also keep using the socket. */
2777 CLOSE_FINISH (sock);
2779 CLOSE_INVALIDATE (sock);
2784 return RETRFINISHED;
2788 For VMS, define common fopen() optional arguments.
2791 # define FOPEN_OPT_ARGS "fop=sqo", "acc", acc_cb, &open_id
2792 # define FOPEN_BIN_FLAG 3
2793 #else /* def __VMS */
2794 # define FOPEN_BIN_FLAG true
2795 #endif /* def __VMS [else] */
2797 /* Open the local file. */
2800 mkalldirs (hs->local_file);
2802 rotate_backups (hs->local_file);
2809 fp = fopen (hs->local_file, "ab", FOPEN_OPT_ARGS);
2810 #else /* def __VMS */
2811 fp = fopen (hs->local_file, "ab");
2812 #endif /* def __VMS [else] */
2814 else if (ALLOW_CLOBBER || count > 0)
2816 if (opt.unlink && file_exists_p (hs->local_file))
2818 int res = unlink (hs->local_file);
2821 logprintf (LOG_NOTQUIET, "%s: %s\n", hs->local_file,
2823 CLOSE_INVALIDATE (sock);
2834 fp = fopen (hs->local_file, "wb", FOPEN_OPT_ARGS);
2835 #else /* def __VMS */
2836 fp = fopen (hs->local_file, "wb");
2837 #endif /* def __VMS [else] */
2841 fp = fopen_excl (hs->local_file, FOPEN_BIN_FLAG);
2842 if (!fp && errno == EEXIST)
2844 /* We cannot just invent a new name and use it (which is
2845 what functions like unique_create typically do)
2846 because we told the user we'd use this name.
2847 Instead, return and retry the download. */
2848 logprintf (LOG_NOTQUIET,
2849 _("%s has sprung into existence.\n"),
2851 CLOSE_INVALIDATE (sock);
2854 return FOPEN_EXCL_ERR;
2859 logprintf (LOG_NOTQUIET, "%s: %s\n", hs->local_file, strerror (errno));
2860 CLOSE_INVALIDATE (sock);
2869 /* Print fetch message, if opt.verbose. */
2872 logprintf (LOG_NOTQUIET, _("Saving to: %s\n"),
2873 HYPHENP (hs->local_file) ? quote ("STDOUT") : quote (hs->local_file));
2877 err = read_response_body (hs, sock, fp, contlen, contrange,
2878 chunked_transfer_encoding,
2879 u->url, warc_timestamp_str,
2880 warc_request_uuid, warc_ip, type,
2883 /* Now we no longer need to store the response header. */
2888 CLOSE_FINISH (sock);
2890 CLOSE_INVALIDATE (sock);
2898 /* The genuine HTTP loop! This is the part where the retrieval is
2899 retried, and retried, and retried, and... */
2901 http_loop (struct url *u, struct url *original_url, char **newloc,
2902 char **local_file, const char *referer, int *dt, struct url *proxy,
2906 bool got_head = false; /* used for time-stamping and filename detection */
2907 bool time_came_from_head = false;
2908 bool got_name = false;
2911 uerr_t err, ret = TRYLIMEXC;
2912 time_t tmr = -1; /* remote time-stamp */
2913 struct http_stat hstat; /* HTTP status */
2915 bool send_head_first = true;
2917 bool force_full_retrieve = false;
2920 /* If we are writing to a WARC file: always retrieve the whole file. */
2921 if (opt.warc_filename != NULL)
2922 force_full_retrieve = true;
2925 /* Assert that no value for *LOCAL_FILE was passed. */
2926 assert (local_file == NULL || *local_file == NULL);
2928 /* Set LOCAL_FILE parameter. */
2929 if (local_file && opt.output_document)
2930 *local_file = HYPHENP (opt.output_document) ? NULL : xstrdup (opt.output_document);
2932 /* Reset NEWLOC parameter. */
2935 /* This used to be done in main(), but it's a better idea to do it
2936 here so that we don't go through the hoops if we're just using
2941 /* Warn on (likely bogus) wildcard usage in HTTP. */
2942 if (opt.ftp_glob && has_wildcards_p (u->path))
2943 logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
2945 /* Setup hstat struct. */
2947 hstat.referer = referer;
2949 if (opt.output_document)
2951 hstat.local_file = xstrdup (opt.output_document);
2954 else if (!opt.content_disposition)
2957 url_file_name (opt.trustservernames ? u : original_url, NULL);
2961 if (got_name && file_exists_p (hstat.local_file) && opt.noclobber && !opt.output_document)
2963 /* If opt.noclobber is turned on and file already exists, do not
2964 retrieve the file. But if the output_document was given, then this
2965 test was already done and the file didn't exist. Hence the !opt.output_document */
2966 get_file_flags (hstat.local_file, dt);
2971 /* Reset the counter. */
2974 /* Reset the document type. */
2977 /* Skip preliminary HEAD request if we're not in spider mode. */
2979 send_head_first = false;
2981 /* Send preliminary HEAD request if -N is given and we have an existing
2982 * destination file. */
2983 file_name = url_file_name (opt.trustservernames ? u : original_url, NULL);
2984 if (opt.timestamping && (file_exists_p (file_name)
2985 || opt.content_disposition))
2986 send_head_first = true;
2992 /* Increment the pass counter. */
2994 sleep_between_retrievals (count);
2996 /* Get the current time string. */
2997 tms = datetime_str (time (NULL));
2999 if (opt.spider && !got_head)
3000 logprintf (LOG_VERBOSE, _("\
3001 Spider mode enabled. Check if remote file exists.\n"));
3003 /* Print fetch message, if opt.verbose. */
3006 char *hurl = url_string (u, URL_AUTH_HIDE_PASSWD);
3011 sprintf (tmp, _("(try:%2d)"), count);
3012 logprintf (LOG_NOTQUIET, "--%s-- %s %s\n",
3017 logprintf (LOG_NOTQUIET, "--%s-- %s\n",
3022 ws_changetitle (hurl);
3027 /* Default document type is empty. However, if spider mode is
3028 on or time-stamping is employed, HEAD_ONLY commands is
3029 encoded within *dt. */
3030 if (send_head_first && !got_head)
3035 /* Decide whether or not to restart. */
3036 if (force_full_retrieve)
3037 hstat.restval = hstat.len;
3038 else if (opt.always_rest
3040 && stat (hstat.local_file, &st) == 0
3041 && S_ISREG (st.st_mode))
3042 /* When -c is used, continue from on-disk size. (Can't use
3043 hstat.len even if count>1 because we don't want a failed
3044 first attempt to clobber existing data.) */
3045 hstat.restval = st.st_size;
3047 /* otherwise, continue where the previous try left off */
3048 hstat.restval = hstat.len;
3052 /* Decide whether to send the no-cache directive. We send it in
3054 a) we're using a proxy, and we're past our first retrieval.
3055 Some proxies are notorious for caching incomplete data, so
3056 we require a fresh get.
3057 b) caching is explicitly inhibited. */
3058 if ((proxy && count > 1) /* a */
3059 || !opt.allow_cache) /* b */
3060 *dt |= SEND_NOCACHE;
3062 *dt &= ~SEND_NOCACHE;
3064 /* Try fetching the document, or at least its head. */
3065 err = gethttp (u, &hstat, dt, proxy, iri, count);
3068 tms = datetime_str (time (NULL));
3070 /* Get the new location (with or without the redirection). */
3072 *newloc = xstrdup (hstat.newloc);
3076 case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
3077 case CONERROR: case READERR: case WRITEFAILED:
3078 case RANGEERR: case FOPEN_EXCL_ERR:
3079 /* Non-fatal errors continue executing the loop, which will
3080 bring them to "while" statement at the end, to judge
3081 whether the number of tries was exceeded. */
3082 printwhat (count, opt.ntry);
3084 case FWRITEERR: case FOPENERR:
3085 /* Another fatal error. */
3086 logputs (LOG_VERBOSE, "\n");
3087 logprintf (LOG_NOTQUIET, _("Cannot write to %s (%s).\n"),
3088 quote (hstat.local_file), strerror (errno));
3089 case HOSTERR: case CONIMPOSSIBLE: case PROXERR: case AUTHFAILED:
3090 case SSLINITFAILED: case CONTNOTSUPPORTED: case VERIFCERTERR:
3091 /* Fatal errors just return from the function. */
3095 /* A fatal WARC error. */
3096 logputs (LOG_VERBOSE, "\n");
3097 logprintf (LOG_NOTQUIET, _("Cannot write to WARC file..\n"));
3100 case WARC_TMP_FOPENERR: case WARC_TMP_FWRITEERR:
3101 /* A fatal WARC error. */
3102 logputs (LOG_VERBOSE, "\n");
3103 logprintf (LOG_NOTQUIET, _("Cannot write to temporary WARC file.\n"));
3107 /* Another fatal error. */
3108 logprintf (LOG_NOTQUIET, _("Unable to establish SSL connection.\n"));
3112 /* Another fatal error. */
3113 logputs (LOG_VERBOSE, "\n");
3114 logprintf (LOG_NOTQUIET, _("Cannot unlink %s (%s).\n"),
3115 quote (hstat.local_file), strerror (errno));
3119 case NEWLOCATION_KEEP_POST:
3120 /* Return the new location to the caller. */
3123 logprintf (LOG_NOTQUIET,
3124 _("ERROR: Redirection (%d) without location.\n"),
3134 /* The file was already fully retrieved. */
3138 /* Deal with you later. */
3141 /* All possibilities should have been exhausted. */
3145 if (!(*dt & RETROKF))
3150 /* #### Ugly ugly ugly! */
3151 hurl = url_string (u, URL_AUTH_HIDE_PASSWD);
3152 logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
3155 /* Fall back to GET if HEAD fails with a 500 or 501 error code. */
3157 && (hstat.statcode == 500 || hstat.statcode == 501))
3162 /* Maybe we should always keep track of broken links, not just in
3164 * Don't log error if it was UTF-8 encoded because we will try
3165 * once unencoded. */
3166 else if (opt.spider && !iri->utf8_encode)
3168 /* #### Again: ugly ugly ugly! */
3170 hurl = url_string (u, URL_AUTH_HIDE_PASSWD);
3171 nonexisting_url (hurl);
3172 logprintf (LOG_NOTQUIET, _("\
3173 Remote file does not exist -- broken link!!!\n"));
3177 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
3178 tms, hstat.statcode,
3179 quotearg_style (escape_quoting_style, hstat.error));
3181 logputs (LOG_VERBOSE, "\n");
3187 /* Did we get the time-stamp? */
3190 got_head = true; /* no more time-stamping */
3192 if (opt.timestamping && !hstat.remote_time)
3194 logputs (LOG_NOTQUIET, _("\
3195 Last-modified header missing -- time-stamps turned off.\n"));
3197 else if (hstat.remote_time)
3199 /* Convert the date-string into struct tm. */
3200 tmr = http_atotm (hstat.remote_time);
3201 if (tmr == (time_t) (-1))
3202 logputs (LOG_VERBOSE, _("\
3203 Last-modified header invalid -- time-stamp ignored.\n"));
3204 if (*dt & HEAD_ONLY)
3205 time_came_from_head = true;
3208 if (send_head_first)
3210 /* The time-stamping section. */
3211 if (opt.timestamping)
3213 if (hstat.orig_file_name) /* Perform the following
3214 checks only if the file
3216 download already exists. */
3218 if (hstat.remote_time &&
3219 tmr != (time_t) (-1))
3221 /* Now time-stamping can be used validly.
3222 Time-stamping means that if the sizes of
3223 the local and remote file match, and local
3224 file is newer than the remote file, it will
3225 not be retrieved. Otherwise, the normal
3226 download procedure is resumed. */
3227 if (hstat.orig_file_tstamp >= tmr)
3229 if (hstat.contlen == -1
3230 || hstat.orig_file_size == hstat.contlen)
3232 logprintf (LOG_VERBOSE, _("\
3233 Server file no newer than local file %s -- not retrieving.\n\n"),
3234 quote (hstat.orig_file_name));
3240 logprintf (LOG_VERBOSE, _("\
3241 The sizes do not match (local %s) -- retrieving.\n"),
3242 number_to_static_string (hstat.orig_file_size));
3247 force_full_retrieve = true;
3248 logputs (LOG_VERBOSE,
3249 _("Remote file is newer, retrieving.\n"));
3252 logputs (LOG_VERBOSE, "\n");
3256 /* free_hstat (&hstat); */
3257 hstat.timestamp_checked = true;
3262 bool finished = true;
3267 logputs (LOG_VERBOSE, _("\
3268 Remote file exists and could contain links to other resources -- retrieving.\n\n"));
3273 logprintf (LOG_VERBOSE, _("\
3274 Remote file exists but does not contain any link -- not retrieving.\n\n"));
3275 ret = RETROK; /* RETRUNNEEDED is not for caller. */
3282 logprintf (LOG_VERBOSE, _("\
3283 Remote file exists and could contain further links,\n\
3284 but recursion is disabled -- not retrieving.\n\n"));
3288 logprintf (LOG_VERBOSE, _("\
3289 Remote file exists.\n\n"));
3291 ret = RETROK; /* RETRUNNEEDED is not for caller. */
3296 logprintf (LOG_NONVERBOSE,
3297 _("%s URL: %s %2d %s\n"),
3298 tms, u->url, hstat.statcode,
3299 hstat.message ? quotearg_style (escape_quoting_style, hstat.message) : "");
3306 count = 0; /* the retrieve count for HEAD is reset */
3308 } /* send_head_first */
3311 if (opt.useservertimestamps
3312 && (tmr != (time_t) (-1))
3313 && ((hstat.len == hstat.contlen) ||
3314 ((hstat.res == 0) && (hstat.contlen == -1))))
3316 const char *fl = NULL;
3317 set_local_file (&fl, hstat.local_file);
3321 /* Reparse time header, in case it's changed. */
3322 if (time_came_from_head
3323 && hstat.remote_time && hstat.remote_time[0])
3325 newtmr = http_atotm (hstat.remote_time);
3326 if (newtmr != (time_t)-1)
3332 /* End of time-stamping section. */
3334 tmrate = retr_rate (hstat.rd_size, hstat.dltime);
3335 total_download_time += hstat.dltime;
3337 if (hstat.len == hstat.contlen)
3341 bool write_to_stdout = (opt.output_document && HYPHENP (opt.output_document));
3343 logprintf (LOG_VERBOSE,
3345 ? _("%s (%s) - written to stdout %s[%s/%s]\n\n")
3346 : _("%s (%s) - %s saved [%s/%s]\n\n"),
3348 write_to_stdout ? "" : quote (hstat.local_file),
3349 number_to_static_string (hstat.len),
3350 number_to_static_string (hstat.contlen));
3351 logprintf (LOG_NONVERBOSE,
3352 "%s URL:%s [%s/%s] -> \"%s\" [%d]\n",
3354 number_to_static_string (hstat.len),
3355 number_to_static_string (hstat.contlen),
3356 hstat.local_file, count);
3359 total_downloaded_bytes += hstat.rd_size;
3361 /* Remember that we downloaded the file for later ".orig" code. */
3362 if (*dt & ADDED_HTML_EXTENSION)
3363 downloaded_file (FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, hstat.local_file);
3365 downloaded_file (FILE_DOWNLOADED_NORMALLY, hstat.local_file);
3370 else if (hstat.res == 0) /* No read error */
3372 if (hstat.contlen == -1) /* We don't know how much we were supposed
3373 to get, so assume we succeeded. */
3377 bool write_to_stdout = (opt.output_document && HYPHENP (opt.output_document));
3379 logprintf (LOG_VERBOSE,
3381 ? _("%s (%s) - written to stdout %s[%s]\n\n")
3382 : _("%s (%s) - %s saved [%s]\n\n"),
3384 write_to_stdout ? "" : quote (hstat.local_file),
3385 number_to_static_string (hstat.len));
3386 logprintf (LOG_NONVERBOSE,
3387 "%s URL:%s [%s] -> \"%s\" [%d]\n",
3388 tms, u->url, number_to_static_string (hstat.len),
3389 hstat.local_file, count);
3392 total_downloaded_bytes += hstat.rd_size;
3394 /* Remember that we downloaded the file for later ".orig" code. */
3395 if (*dt & ADDED_HTML_EXTENSION)
3396 downloaded_file (FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, hstat.local_file);
3398 downloaded_file (FILE_DOWNLOADED_NORMALLY, hstat.local_file);
3403 else if (hstat.len < hstat.contlen) /* meaning we lost the
3404 connection too soon */
3406 logprintf (LOG_VERBOSE,
3407 _("%s (%s) - Connection closed at byte %s. "),
3408 tms, tmrate, number_to_static_string (hstat.len));
3409 printwhat (count, opt.ntry);
3412 else if (hstat.len != hstat.restval)
3413 /* Getting here would mean reading more data than
3414 requested with content-length, which we never do. */
3418 /* Getting here probably means that the content-length was
3419 * _less_ than the original, local size. We should probably
3420 * truncate or re-read, or something. FIXME */
3425 else /* from now on hstat.res can only be -1 */
3427 if (hstat.contlen == -1)
3429 logprintf (LOG_VERBOSE,
3430 _("%s (%s) - Read error at byte %s (%s)."),
3431 tms, tmrate, number_to_static_string (hstat.len),
3433 printwhat (count, opt.ntry);
3436 else /* hstat.res == -1 and contlen is given */
3438 logprintf (LOG_VERBOSE,
3439 _("%s (%s) - Read error at byte %s/%s (%s). "),
3441 number_to_static_string (hstat.len),
3442 number_to_static_string (hstat.contlen),
3444 printwhat (count, opt.ntry);
3450 while (!opt.ntry || (count < opt.ntry));
3453 if (ret == RETROK && local_file)
3454 *local_file = xstrdup (hstat.local_file);
3455 free_hstat (&hstat);
3460 /* Check whether the result of strptime() indicates success.
3461 strptime() returns the pointer to how far it got to in the string.
3462 The processing has been successful if the string is at `GMT' or
3463 `+X', or at the end of the string.
3465 In extended regexp parlance, the function returns 1 if P matches
3466 "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (which strptime
3467 can return) is considered a failure and 0 is returned. */
3469 check_end (const char *p)
3473 while (c_isspace (*p))
3476 || (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
3477 || ((p[0] == '+' || p[0] == '-') && c_isdigit (p[1])))
3483 /* Convert the textual specification of time in TIME_STRING to the
3484 number of seconds since the Epoch.
3486 TIME_STRING can be in any of the three formats RFC2616 allows the
3487 HTTP servers to emit -- RFC1123-date, RFC850-date or asctime-date,
3488 as well as the time format used in the Set-Cookie header.
3489 Timezones are ignored, and should be GMT.
3491 Return the computed time_t representation, or -1 if the conversion
3494 This function uses strptime with various string formats for parsing
3495 TIME_STRING. This results in a parser that is not as lenient in
3496 interpreting TIME_STRING as I would like it to be. Being based on
3497 strptime, it always allows shortened months, one-digit days, etc.,
3498 but due to the multitude of formats in which time can be
3499 represented, an ideal HTTP time parser would be even more
3500 forgiving. It should completely ignore things like week days and
3501 concentrate only on the various forms of representing years,
3502 months, days, hours, minutes, and seconds. For example, it would
3503 be nice if it accepted ISO 8601 out of the box.
3505 I've investigated free and PD code for this purpose, but none was
3506 usable. getdate was big and unwieldy, and had potential copyright
3507 issues, or so I was informed. Dr. Marcus Hennecke's atotm(),
3508 distributed with phttpd, is excellent, but we cannot use it because
3509 it is not assigned to the FSF. So I stuck it with strptime. */
3512 http_atotm (const char *time_string)
3514 /* NOTE: Solaris strptime man page claims that %n and %t match white
3515 space, but that's not universally available. Instead, we simply
3516 use ` ' to mean "skip all WS", which works under all strptime
3517 implementations I've tested. */
3519 static const char *time_formats[] = {
3520 "%a, %d %b %Y %T", /* rfc1123: Thu, 29 Jan 1998 22:12:57 */
3521 "%A, %d-%b-%y %T", /* rfc850: Thursday, 29-Jan-98 22:12:57 */
3522 "%a %b %d %T %Y", /* asctime: Thu Jan 29 22:12:57 1998 */
3523 "%a, %d-%b-%Y %T" /* cookies: Thu, 29-Jan-1998 22:12:57
3524 (used in Set-Cookie, defined in the
3525 Netscape cookie specification.) */
3527 const char *oldlocale;
3528 char savedlocale[256];
3530 time_t ret = (time_t) -1;
3532 /* Solaris strptime fails to recognize English month names in
3533 non-English locales, which we work around by temporarily setting
3534 locale to C before invoking strptime. */
3535 oldlocale = setlocale (LC_TIME, NULL);
3538 size_t l = strlen (oldlocale) + 1;
3539 if (l >= sizeof savedlocale)
3540 savedlocale[0] = '\0';
3542 memcpy (savedlocale, oldlocale, l);
3544 else savedlocale[0] = '\0';
3546 setlocale (LC_TIME, "C");
3548 for (i = 0; i < countof (time_formats); i++)
3552 /* Some versions of strptime use the existing contents of struct
3553 tm to recalculate the date according to format. Zero it out
3554 to prevent stack garbage from influencing strptime. */
3557 if (check_end (strptime (time_string, time_formats[i], &t)))
3564 /* Restore the previous locale. */
3566 setlocale (LC_TIME, savedlocale);
3571 /* Authorization support: We support three authorization schemes:
3573 * `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
3575 * `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
3576 consisting of answering to the server's challenge with the proper
3579 * `NTLM' ("NT Lan Manager") scheme, based on code written by Daniel
3580 Stenberg for libcurl. Like digest, NTLM is based on a
3581 challenge-response mechanism, but unlike digest, it is non-standard
3582 (authenticates TCP connections rather than requests), undocumented
3583 and Microsoft-specific. */
3585 /* Create the authentication header contents for the `Basic' scheme.
3586 This is done by encoding the string "USER:PASS" to base64 and
3587 prepending the string "Basic " in front of it. */
3590 basic_authentication_encode (const char *user, const char *passwd)
3593 int len1 = strlen (user) + 1 + strlen (passwd);
3595 t1 = (char *)alloca (len1 + 1);
3596 sprintf (t1, "%s:%s", user, passwd);
3598 t2 = (char *)alloca (BASE64_LENGTH (len1) + 1);
3599 base64_encode (t1, len1, t2);
3601 return concat_strings ("Basic ", t2, (char *) 0);
3604 #define SKIP_WS(x) do { \
3605 while (c_isspace (*(x))) \
3609 #ifdef ENABLE_DIGEST
3610 /* Dump the hexadecimal representation of HASH to BUF. HASH should be
3611 an array of 16 bytes containing the hash keys, and BUF should be a
3612 buffer of 33 writable characters (32 for hex digits plus one for
3613 zero termination). */
3615 dump_hash (char *buf, const unsigned char *hash)
3619 for (i = 0; i < MD5_DIGEST_SIZE; i++, hash++)
3621 *buf++ = XNUM_TO_digit (*hash >> 4);
3622 *buf++ = XNUM_TO_digit (*hash & 0xf);
3627 /* Take the line apart to find the challenge, and compose a digest
3628 authorization header. See RFC2069 section 2.1.2. */
3630 digest_authentication_encode (const char *au, const char *user,
3631 const char *passwd, const char *method,
3634 static char *realm, *opaque, *nonce;
3639 { "realm", &realm },
3640 { "opaque", &opaque },
3644 param_token name, value;
3646 realm = opaque = nonce = NULL;
3648 au += 6; /* skip over `Digest' */
3649 while (extract_param (&au, &name, &value, ','))
3652 size_t namelen = name.e - name.b;
3653 for (i = 0; i < countof (options); i++)
3654 if (namelen == strlen (options[i].name)
3655 && 0 == strncmp (name.b, options[i].name,
3658 *options[i].variable = strdupdelim (value.b, value.e);
3662 if (!realm || !nonce || !user || !passwd || !path || !method)
3665 xfree_null (opaque);
3670 /* Calculate the digest value. */
3673 unsigned char hash[MD5_DIGEST_SIZE];
3674 char a1buf[MD5_DIGEST_SIZE * 2 + 1], a2buf[MD5_DIGEST_SIZE * 2 + 1];
3675 char response_digest[MD5_DIGEST_SIZE * 2 + 1];
3677 /* A1BUF = H(user ":" realm ":" password) */
3678 md5_init_ctx (&ctx);
3679 md5_process_bytes ((unsigned char *)user, strlen (user), &ctx);
3680 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3681 md5_process_bytes ((unsigned char *)realm, strlen (realm), &ctx);
3682 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3683 md5_process_bytes ((unsigned char *)passwd, strlen (passwd), &ctx);
3684 md5_finish_ctx (&ctx, hash);
3685 dump_hash (a1buf, hash);
3687 /* A2BUF = H(method ":" path) */
3688 md5_init_ctx (&ctx);
3689 md5_process_bytes ((unsigned char *)method, strlen (method), &ctx);
3690 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3691 md5_process_bytes ((unsigned char *)path, strlen (path), &ctx);
3692 md5_finish_ctx (&ctx, hash);
3693 dump_hash (a2buf, hash);
3695 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */
3696 md5_init_ctx (&ctx);
3697 md5_process_bytes ((unsigned char *)a1buf, MD5_DIGEST_SIZE * 2, &ctx);
3698 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3699 md5_process_bytes ((unsigned char *)nonce, strlen (nonce), &ctx);
3700 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3701 md5_process_bytes ((unsigned char *)a2buf, MD5_DIGEST_SIZE * 2, &ctx);
3702 md5_finish_ctx (&ctx, hash);
3703 dump_hash (response_digest, hash);
3705 res = xmalloc (strlen (user)
3710 + 2 * MD5_DIGEST_SIZE /*strlen (response_digest)*/
3711 + (opaque ? strlen (opaque) : 0)
3713 sprintf (res, "Digest \
3714 username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
3715 user, realm, nonce, path, response_digest);
3718 char *p = res + strlen (res);
3719 strcat (p, ", opaque=\"");
3726 #endif /* ENABLE_DIGEST */
3728 /* Computing the size of a string literal must take into account that
3729 value returned by sizeof includes the terminating \0. */
3730 #define STRSIZE(literal) (sizeof (literal) - 1)
3732 /* Whether chars in [b, e) begin with the literal string provided as
3733 first argument and are followed by whitespace or terminating \0.
3734 The comparison is case-insensitive. */
3735 #define STARTS(literal, b, e) \
3737 && ((size_t) ((e) - (b))) >= STRSIZE (literal) \
3738 && 0 == strncasecmp (b, literal, STRSIZE (literal)) \
3739 && ((size_t) ((e) - (b)) == STRSIZE (literal) \
3740 || c_isspace (b[STRSIZE (literal)])))
3743 known_authentication_scheme_p (const char *hdrbeg, const char *hdrend)
3745 return STARTS ("Basic", hdrbeg, hdrend)
3746 #ifdef ENABLE_DIGEST
3747 || STARTS ("Digest", hdrbeg, hdrend)
3750 || STARTS ("NTLM", hdrbeg, hdrend)
3757 /* Create the HTTP authorization request header. When the
3758 `WWW-Authenticate' response header is seen, according to the
3759 authorization scheme specified in that header (`Basic' and `Digest'
3760 are supported by the current implementation), produce an
3761 appropriate HTTP authorization request header. */
3763 create_authorization_line (const char *au, const char *user,
3764 const char *passwd, const char *method,
3765 const char *path, bool *finished)
3767 /* We are called only with known schemes, so we can dispatch on the
3769 switch (c_toupper (*au))
3771 case 'B': /* Basic */
3773 return basic_authentication_encode (user, passwd);
3774 #ifdef ENABLE_DIGEST
3775 case 'D': /* Digest */
3777 return digest_authentication_encode (au, user, passwd, method, path);
3780 case 'N': /* NTLM */
3781 if (!ntlm_input (&pconn.ntlm, au))
3786 return ntlm_output (&pconn.ntlm, user, passwd, finished);
3789 /* We shouldn't get here -- this function should be only called
3790 with values approved by known_authentication_scheme_p. */
3798 if (!wget_cookie_jar)
3799 wget_cookie_jar = cookie_jar_new ();
3800 if (opt.cookies_input && !cookies_loaded_p)
3802 cookie_jar_load (wget_cookie_jar, opt.cookies_input);
3803 cookies_loaded_p = true;
3810 if (wget_cookie_jar)
3811 cookie_jar_save (wget_cookie_jar, opt.cookies_output);
3817 xfree_null (pconn.host);
3818 if (wget_cookie_jar)
3819 cookie_jar_delete (wget_cookie_jar);
3823 ensure_extension (struct http_stat *hs, const char *ext, int *dt)
3825 char *last_period_in_local_filename = strrchr (hs->local_file, '.');
3827 int len = strlen (ext);
3830 strncpy (shortext, ext, len - 1);
3831 shortext[len - 1] = '\0';
3834 if (last_period_in_local_filename == NULL
3835 || !(0 == strcasecmp (last_period_in_local_filename, shortext)
3836 || 0 == strcasecmp (last_period_in_local_filename, ext)))
3838 int local_filename_len = strlen (hs->local_file);
3839 /* Resize the local file, allowing for ".html" preceded by
3840 optional ".NUMBER". */
3841 hs->local_file = xrealloc (hs->local_file,
3842 local_filename_len + 24 + len);
3843 strcpy (hs->local_file + local_filename_len, ext);
3844 /* If clobbering is not allowed and the file, as named,
3845 exists, tack on ".NUMBER.html" instead. */
3846 if (!ALLOW_CLOBBER && file_exists_p (hs->local_file))
3850 sprintf (hs->local_file + local_filename_len,
3851 ".%d%s", ext_num++, ext);
3852 while (file_exists_p (hs->local_file));
3854 *dt |= ADDED_HTML_EXTENSION;
3862 test_parse_content_disposition()
3870 { "filename=\"file.ext\"", "file.ext", true },
3871 { "attachment; filename=\"file.ext\"", "file.ext", true },
3872 { "attachment; filename=\"file.ext\"; dummy", "file.ext", true },
3873 { "attachment", NULL, false },
3874 { "attachement; filename*=UTF-8'en-US'hello.txt", "hello.txt", true },
3875 { "attachement; filename*0=\"hello\"; filename*1=\"world.txt\"", "helloworld.txt", true },
3878 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
3883 res = parse_content_disposition (test_array[i].hdrval, &filename);
3885 mu_assert ("test_parse_content_disposition: wrong result",
3886 res == test_array[i].result
3888 || 0 == strcmp (test_array[i].filename, filename)));
3894 #endif /* TESTING */
3897 * vim: et sts=2 sw=2 cino+={s