2 Copyright (C) 1995, 1996, 1997, 2000 Free Software Foundation, Inc.
4 This file is part of Wget.
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or (at
9 your option) any later version.
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with this program; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
30 #include <sys/types.h>
46 /* Default port definitions */
47 #define DEFAULT_HTTP_PORT 80
48 #define DEFAULT_FTP_PORT 21
49 #define DEFAULT_HTTPS_PORT 443
51 /* Table of Unsafe chars. This is intialized in
52 init_unsafe_char_table. */
54 static char unsafe_char_table[256];
56 #define UNSAFE_CHAR(c) (unsafe_char_table[(unsigned char)(c)])
58 /* If S contains unsafe characters, free it and replace it with a
59 version that doesn't. */
60 #define URL_CLEANSE(s) do \
62 if (contains_unsafe (s)) \
64 char *uc_tmp = encode_string (s); \
70 /* Is a directory "."? */
71 #define DOTP(x) ((*(x) == '.') && (!*(x + 1)))
72 /* Is a directory ".."? */
73 #define DDOTP(x) ((*(x) == '.') && (*(x + 1) == '.') && (!*(x + 2)))
76 static void path_simplify_with_kludge PARAMS ((char *));
78 static int urlpath_length PARAMS ((const char *));
80 /* NULL-terminated list of strings to be recognized as prototypes (URL
81 schemes). Note that recognized doesn't mean supported -- only HTTP,
82 HTTPS and FTP are currently supported .
84 However, a string that does not match anything in the list will be
85 considered a relative URL. Thus it's important that this list has
86 anything anyone could think of being legal.
88 There are wild things here. :-) Take a look at
89 <URL:http://www.w3.org/pub/WWW/Addressing/schemes.html> for more
91 static char *protostrings[] =
133 /* Similar to former, but for supported protocols: */
134 static struct proto sup_protos[] =
136 { "http://", URLHTTP, DEFAULT_HTTP_PORT },
138 { "https://",URLHTTPS, DEFAULT_HTTPS_PORT},
140 { "ftp://", URLFTP, DEFAULT_FTP_PORT },
141 /*{ "file://", URLFILE, DEFAULT_FTP_PORT },*/
144 static void parse_dir PARAMS ((const char *, char **, char **));
145 static uerr_t parse_uname PARAMS ((const char *, char **, char **));
146 static char *construct PARAMS ((const char *, const char *, int , int));
147 static char *construct_relative PARAMS ((const char *, const char *));
148 static char process_ftp_type PARAMS ((char *));
151 /* Returns the number of characters to be skipped if the first thing
152 in a URL is URL: (which is 0 or 4+). The optional spaces after
153 URL: are also skipped. */
155 skip_url (const char *url)
159 if (TOUPPER (url[0]) == 'U'
160 && TOUPPER (url[1]) == 'R'
161 && TOUPPER (url[2]) == 'L'
165 for (i = 4; url[i] && ISSPACE (url[i]); i++);
174 - stuff from rfc1738 ("<>\"#%{}|\\^~[]`");
175 - @ and :, for user/password encoding.
176 - everything over 127 (but we don't bother with recording those. */
178 init_unsafe_char_table (void)
181 for (i = 0; i < 256; i++)
182 if (i < 32 || i >= 127
198 unsafe_char_table[i] = 1;
201 /* Returns 1 if the string contains unsafe characters, 0 otherwise. */
203 contains_unsafe (const char *s)
206 if (UNSAFE_CHAR (*s))
211 /* Decodes the forms %xy in a URL to the character the hexadecimal
212 code of which is xy. xy are hexadecimal digits from
213 [0123456789ABCDEF] (case-insensitive). If x or y are not
214 hex-digits or `%' precedes `\0', the sequence is inserted
218 decode_string (char *s)
228 /* Do nothing if at the end of the string, or if the chars
229 are not hex-digits. */
230 if (!*(s + 1) || !*(s + 2)
231 || !(ISXDIGIT (*(s + 1)) && ISXDIGIT (*(s + 2))))
236 *p = (ASC2HEXD (*(s + 1)) << 4) + ASC2HEXD (*(s + 2));
243 /* Encode the unsafe characters (as determined by URL_UNSAFE) in a
244 given string, returning a malloc-ed %XX encoded string. */
246 encode_string (const char *s)
253 for (i = 0; *s; s++, i++)
254 if (UNSAFE_CHAR (*s))
255 i += 2; /* Two more characters (hex digits) */
256 res = (char *)xmalloc (i + 1);
258 for (p = res; *s; s++)
259 if (UNSAFE_CHAR (*s))
261 const unsigned char c = *s;
263 *p++ = HEXD2ASC (c >> 4);
264 *p++ = HEXD2ASC (c & 0xf);
272 /* Returns the proto-type if URL's protocol is supported, or
273 URLUNKNOWN if not. */
275 urlproto (const char *url)
279 url += skip_url (url);
280 for (i = 0; i < ARRAY_SIZE (sup_protos); i++)
281 if (!strncasecmp (url, sup_protos[i].name, strlen (sup_protos[i].name)))
282 return sup_protos[i].ind;
283 for (i = 0; url[i] && url[i] != ':' && url[i] != '/'; i++);
286 for (++i; url[i] && url[i] != '/'; i++)
287 if (!ISDIGIT (url[i]))
289 if (url[i - 1] == ':')
298 /* Skip the protocol part of the URL, e.g. `http://'. If no protocol
299 part is found, returns 0. */
301 skip_proto (const char *url)
306 for (s = protostrings; *s; s++)
307 if (!strncasecmp (*s, url, strlen (*s)))
312 /* HTTP and FTP protocols are expected to yield exact host names
313 (i.e. the `//' part must be skipped, too). */
314 if (!strcmp (*s, "http:") || !strcmp (*s, "ftp:"))
319 /* Returns 1 if the URL begins with a protocol (supported or
320 unsupported), 0 otherwise. */
322 has_proto (const char *url)
326 url += skip_url (url);
327 for (s = protostrings; *s; s++)
328 if (strncasecmp (url, *s, strlen (*s)) == 0)
333 /* Skip the username and password, if present here. The function
334 should be called *not* with the complete URL, but with the part
335 right after the protocol.
337 If no username and password are found, return 0. */
339 skip_uname (const char *url)
342 for (p = url; *p && *p != '/'; p++)
345 /* If a `@' was found before the first occurrence of `/', skip
353 /* Allocate a new urlinfo structure, fill it with default values and
354 return a pointer to it. */
360 u = (struct urlinfo *)xmalloc (sizeof (struct urlinfo));
361 memset (u, 0, sizeof (*u));
362 u->proto = URLUNKNOWN;
366 /* Perform a "deep" free of the urlinfo structure. The structure
367 should have been created with newurl, but need not have been used.
368 If free_pointer is non-0, free the pointer itself. */
370 freeurl (struct urlinfo *u, int complete)
374 FREE_MAYBE (u->host);
375 FREE_MAYBE (u->path);
376 FREE_MAYBE (u->file);
378 FREE_MAYBE (u->user);
379 FREE_MAYBE (u->passwd);
380 FREE_MAYBE (u->local);
381 FREE_MAYBE (u->referer);
383 freeurl (u->proxy, 1);
389 /* Extract the given URL of the form
390 (http:|ftp:)// (user (:password)?@)?hostname (:port)? (/path)?
391 1. hostname (terminated with `/' or `:')
392 2. port number (terminated with `/'), or chosen for the protocol
393 3. dirname (everything after hostname)
394 Most errors are handled. No allocation is done, you must supply
395 pointers to allocated memory.
396 ...and a host of other stuff :-)
398 - Recognizes hostname:dir/file for FTP and
399 hostname (:portnum)?/dir/file for HTTP.
400 - Parses the path to yield directory and file
401 - Parses the URL to yield the username and passwd (if present)
402 - Decodes the strings, in case they contain "forbidden" characters
403 - Writes the result to struct urlinfo
405 If the argument STRICT is set, it recognizes only the canonical
408 parseurl (const char *url, struct urlinfo *u, int strict)
411 int recognizable; /* Recognizable URL is the one where
412 the protocol name was explicitly
413 named, i.e. it wasn't deduced from
417 DEBUGP (("parseurl (\"%s\") -> ", url));
418 url += skip_url (url);
419 recognizable = has_proto (url);
420 if (strict && !recognizable)
422 for (i = 0, l = 0; i < ARRAY_SIZE (sup_protos); i++)
424 l = strlen (sup_protos[i].name);
425 if (!strncasecmp (sup_protos[i].name, url, l))
428 /* If protocol is recognizable, but unsupported, bail out, else
430 if (recognizable && i == ARRAY_SIZE (sup_protos))
432 else if (i == ARRAY_SIZE (sup_protos))
435 u->proto = type = sup_protos[i].ind;
437 if (type == URLUNKNOWN)
439 /* Allow a username and password to be specified (i.e. just skip
442 l += skip_uname (url + l);
443 for (i = l; url[i] && url[i] != ':' && url[i] != '/'; i++);
446 /* Get the hostname. */
447 u->host = strdupdelim (url + l, url + i);
448 DEBUGP (("host %s -> ", u->host));
450 /* Assume no port has been given. */
454 /* We have a colon delimiting the hostname. It could mean that
455 a port number is following it, or a directory. */
456 if (ISDIGIT (url[++i])) /* A port number */
458 if (type == URLUNKNOWN)
459 u->proto = type = URLHTTP;
460 for (; url[i] && url[i] != '/'; i++)
461 if (ISDIGIT (url[i]))
462 u->port = 10 * u->port + (url[i] - '0');
467 DEBUGP (("port %hu -> ", u->port));
469 else if (type == URLUNKNOWN) /* or a directory */
470 u->proto = type = URLFTP;
471 else /* or just a misformed port number */
474 else if (type == URLUNKNOWN)
475 u->proto = type = URLHTTP;
479 for (ind = 0; ind < ARRAY_SIZE (sup_protos); ind++)
480 if (sup_protos[ind].ind == type)
482 if (ind == ARRAY_SIZE (sup_protos))
484 u->port = sup_protos[ind].port;
486 /* Some delimiter troubles... */
487 if (url[i] == '/' && url[i - 1] != ':')
490 while (url[i] && url[i] == '/')
492 u->path = (char *)xmalloc (strlen (url + i) + 8);
493 strcpy (u->path, url + i);
496 u->ftp_type = process_ftp_type (u->path);
497 /* #### We don't handle type `d' correctly yet. */
498 if (!u->ftp_type || TOUPPER (u->ftp_type) == 'D')
500 DEBUGP (("ftp_type %c -> ", u->ftp_type));
502 DEBUGP (("opath %s -> ", u->path));
503 /* Parse the username and password (if existing). */
504 parse_uname (url, &u->user, &u->passwd);
505 /* Decode the strings, as per RFC 1738. */
506 decode_string (u->host);
507 decode_string (u->path);
509 decode_string (u->user);
511 decode_string (u->passwd);
512 /* Parse the directory. */
513 parse_dir (u->path, &u->dir, &u->file);
514 DEBUGP (("dir %s -> file %s -> ", u->dir, u->file));
515 /* Simplify the directory. */
516 path_simplify (u->dir);
517 /* Remove the leading `/' in HTTP. */
518 if (type == URLHTTP && *u->dir == '/')
519 strcpy (u->dir, u->dir + 1);
520 DEBUGP (("ndir %s\n", u->dir));
521 /* Strip trailing `/'. */
523 if (l && u->dir[l - 1] == '/')
524 u->dir[l - 1] = '\0';
525 /* Re-create the path: */
526 abs_ftp = (u->proto == URLFTP && *u->dir == '/');
527 /* sprintf (u->path, "%s%s%s%s", abs_ftp ? "%2F": "/",
528 abs_ftp ? (u->dir + 1) : u->dir, *u->dir ? "/" : "", u->file); */
529 strcpy (u->path, abs_ftp ? "%2F" : "/");
530 strcat (u->path, abs_ftp ? (u->dir + 1) : u->dir);
531 strcat (u->path, *u->dir ? "/" : "");
532 strcat (u->path, u->file);
533 URL_CLEANSE (u->path);
534 DEBUGP (("newpath: %s\n", u->path));
535 /* Create the clean URL. */
536 u->url = str_url (u, 0);
540 /* Special versions of DOTP and DDOTP for parse_dir(). */
542 #define PD_DOTP(x) ((*(x) == '.') && (!*((x) + 1) || *((x) + 1) == '?'))
543 #define PD_DDOTP(x) ((*(x) == '.') && (*(x) == '.') \
544 && (!*((x) + 2) || *((x) + 2) == '?'))
546 /* Build the directory and filename components of the path. Both
547 components are *separately* malloc-ed strings! It does not change
548 the contents of path.
550 If the path ends with "." or "..", they are (correctly) counted as
553 parse_dir (const char *path, char **dir, char **file)
557 l = urlpath_length (path);
558 for (i = l; i && path[i] != '/'; i--);
560 if (!i && *path != '/') /* Just filename */
562 if (PD_DOTP (path) || PD_DDOTP (path))
564 *dir = strdupdelim (path, path + l);
565 *file = xstrdup (path + l); /* normally empty, but could
570 *dir = xstrdup (""); /* This is required because of FTP */
571 *file = xstrdup (path);
574 else if (!i) /* /filename */
576 if (PD_DOTP (path + 1) || PD_DDOTP (path + 1))
578 *dir = strdupdelim (path, path + l);
579 *file = xstrdup (path + l); /* normally empty, but could
584 *dir = xstrdup ("/");
585 *file = xstrdup (path + 1);
588 else /* Nonempty directory with or without a filename */
590 if (PD_DOTP (path + i + 1) || PD_DDOTP (path + i + 1))
592 *dir = strdupdelim (path, path + l);
593 *file = xstrdup (path + l); /* normally empty, but could
598 *dir = strdupdelim (path, path + i);
599 *file = xstrdup (path + i + 1);
604 /* Find the optional username and password within the URL, as per
605 RFC1738. The returned user and passwd char pointers are
608 parse_uname (const char *url, char **user, char **passwd)
616 url += skip_url (url);
617 /* Look for end of protocol string. */
618 l = skip_proto (url);
621 /* Add protocol offset. */
623 /* Is there an `@' character? */
624 for (p = url; *p && *p != '/'; p++)
627 /* If not, return. */
630 /* Else find the username and password. */
631 for (p = col = url; *p != '@'; p++)
633 if (*p == ':' && !*user)
635 *user = (char *)xmalloc (p - url + 1);
636 memcpy (*user, url, p - url);
637 (*user)[p - url] = '\0';
641 /* Decide whether you have only the username or both. */
642 where = *user ? passwd : user;
643 *where = (char *)xmalloc (p - col + 1);
644 memcpy (*where, col, p - col);
645 (*where)[p - col] = '\0';
649 /* If PATH ends with `;type=X', return the character X. */
651 process_ftp_type (char *path)
653 int len = strlen (path);
656 && !memcmp (path + len - 7, ";type=", 6))
658 path[len - 7] = '\0';
659 return path[len - 1];
665 /* Return the URL as fine-formed string, with a proper protocol, optional port
666 number, directory and optional user/password. If `hide' is non-zero (as it
667 is when we're calling this on a URL we plan to print, but not when calling it
668 to canonicalize a URL for use within the program), password will be hidden.
669 The forbidden characters in the URL will be cleansed. */
671 str_url (const struct urlinfo *u, int hide)
673 char *res, *host, *user, *passwd, *proto_name, *dir, *file;
674 int i, l, ln, lu, lh, lp, lf, ld;
675 unsigned short proto_default_port;
677 /* Look for the protocol name. */
678 for (i = 0; i < ARRAY_SIZE (sup_protos); i++)
679 if (sup_protos[i].ind == u->proto)
681 if (i == ARRAY_SIZE (sup_protos))
683 proto_name = sup_protos[i].name;
684 proto_default_port = sup_protos[i].port;
685 host = CLEANDUP (u->host);
686 dir = CLEANDUP (u->dir);
687 file = CLEANDUP (u->file);
688 user = passwd = NULL;
690 user = CLEANDUP (u->user);
694 /* Don't output the password, or someone might see it over the user's
695 shoulder (or in saved wget output). Don't give away the number of
696 characters in the password, either, as we did in past versions of
697 this code, when we replaced the password characters with 'x's. */
698 passwd = xstrdup("<password>");
700 passwd = CLEANDUP (u->passwd);
702 if (u->proto == URLFTP && *dir == '/')
704 char *tmp = (char *)xmalloc (strlen (dir) + 3);
705 /*sprintf (tmp, "%%2F%s", dir + 1);*/
709 strcpy (tmp + 3, dir + 1);
714 ln = strlen (proto_name);
715 lu = user ? strlen (user) : 0;
716 lp = passwd ? strlen (passwd) : 0;
720 res = (char *)xmalloc (ln + lu + lp + lh + ld + lf + 20); /* safe sex */
721 /* sprintf (res, "%s%s%s%s%s%s:%d/%s%s%s", proto_name,
722 (user ? user : ""), (passwd ? ":" : ""),
723 (passwd ? passwd : ""), (user ? "@" : ""),
724 host, u->port, dir, *dir ? "/" : "", file); */
726 memcpy (res, proto_name, ln);
730 memcpy (res + l, user, lu);
735 memcpy (res + l, passwd, lp);
740 memcpy (res + l, host, lh);
742 if (u->port != proto_default_port)
745 long_to_string (res + l, (long)u->port);
746 l += numdigit (u->port);
749 memcpy (res + l, dir, ld);
753 strcpy (res + l, file);
762 /* Check whether two URL-s are equivalent, i.e. pointing to the same
763 location. Uses parseurl to parse them, and compares the canonical
766 Returns 1 if the URL1 is equivalent to URL2, 0 otherwise. Also
767 return 0 on error. */
769 url_equal (const char *url1, const char *url2)
771 struct urlinfo *u1, *u2;
776 err = parseurl (url1, u1, 0);
783 err = parseurl (url2, u2, 0);
789 res = !strcmp (u1->url, u2->url);
796 get_urls_file (const char *file)
798 struct file_memory *fm;
800 const char *text, *text_end;
803 fm = read_file (file);
806 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
809 DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
812 text_end = fm->content + fm->length;
813 while (text < text_end)
815 const char *line_beg = text;
816 const char *line_end = memchr (text, '\n', text_end - text);
822 while (line_beg < line_end
823 && ISSPACE (*line_beg))
825 while (line_end > line_beg + 1
826 && ISSPACE (*(line_end - 1)))
828 if (line_end > line_beg)
830 urlpos *entry = (urlpos *)xmalloc (sizeof (urlpos));
831 memset (entry, 0, sizeof (*entry));
833 entry->url = strdupdelim (line_beg, line_end);
845 /* Free the linked list of urlpos. */
847 free_urlpos (urlpos *l)
851 urlpos *next = l->next;
853 FREE_MAYBE (l->local_name);
859 /* Rotate FNAME opt.backups times */
861 rotate_backups(const char *fname)
863 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
864 char *from = (char *)alloca (maxlen);
865 char *to = (char *)alloca (maxlen);
869 if (stat (fname, &sb) == 0)
870 if (S_ISREG (sb.st_mode) == 0)
873 for (i = opt.backups; i > 1; i--)
875 sprintf (from, "%s.%d", fname, i - 1);
876 sprintf (to, "%s.%d", fname, i);
877 /* #### This will fail on machines without the rename() system
882 sprintf (to, "%s.%d", fname, 1);
886 /* Create all the necessary directories for PATH (a file). Calls
887 mkdirhier() internally. */
889 mkalldirs (const char *path)
896 p = path + strlen (path);
897 for (; *p != '/' && p != path; p--);
898 /* Don't create if it's just a file. */
899 if ((p == path) && (*p != '/'))
901 t = strdupdelim (path, p);
902 /* Check whether the directory exists. */
903 if ((stat (t, &st) == 0))
905 if (S_ISDIR (st.st_mode))
912 /* If the dir exists as a file name, remove it first. This
913 is *only* for Wget to work with buggy old CERN http
914 servers. Here is the scenario: When Wget tries to
915 retrieve a directory without a slash, e.g.
916 http://foo/bar (bar being a directory), CERN server will
917 not redirect it too http://foo/bar/ -- it will generate a
918 directory listing containing links to bar/file1,
919 bar/file2, etc. Wget will lose because it saves this
920 HTML listing to a file `bar', so it cannot create the
921 directory. To work around this, if the file of the same
922 name exists, we just remove it and create the directory
924 DEBUGP (("Removing %s because of directory danger!\n", t));
928 res = make_directory (t);
930 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
936 count_slashes (const char *s)
945 /* Return the path name of the URL-equivalent file name, with a
946 remote-like structure of directories. */
948 mkstruct (const struct urlinfo *u)
950 char *host, *dir, *file, *res, *dirpref;
953 assert (u->dir != NULL);
954 assert (u->host != NULL);
958 char *ptr = u->dir + (*u->dir == '/');
959 int slash_count = 1 + count_slashes (ptr);
960 int cut = MINVAL (opt.cut_dirs, slash_count);
961 for (; cut && *ptr; ptr++)
964 STRDUP_ALLOCA (dir, ptr);
967 dir = u->dir + (*u->dir == '/');
969 host = xstrdup (u->host);
970 /* Check for the true name (or at least a consistent name for saving
971 to directory) of HOST, reusing the hlist if possible. */
972 if (opt.add_hostdir && !opt.simple_check)
974 char *nhost = realhost (host);
978 /* Add dir_prefix and hostname (if required) to the beginning of
982 if (!DOTP (opt.dir_prefix))
984 dirpref = (char *)alloca (strlen (opt.dir_prefix) + 1
985 + strlen (host) + 1);
986 sprintf (dirpref, "%s/%s", opt.dir_prefix, host);
989 STRDUP_ALLOCA (dirpref, host);
991 else /* not add_hostdir */
993 if (!DOTP (opt.dir_prefix))
994 dirpref = opt.dir_prefix;
1000 /* If there is a prefix, prepend it. */
1003 char *newdir = (char *)alloca (strlen (dirpref) + 1 + strlen (dir) + 2);
1004 sprintf (newdir, "%s%s%s", dirpref, *dir == '/' ? "" : "/", dir);
1007 dir = xstrdup (dir);
1010 if (l && dir[l - 1] == '/')
1014 file = "index.html";
1018 /* Finally, construct the full name. */
1019 res = (char *)xmalloc (strlen (dir) + 1 + strlen (file) + 1);
1020 sprintf (res, "%s%s%s", dir, *dir ? "/" : "", file);
1025 /* Create a unique filename, corresponding to a given URL. Calls
1026 mkstruct if necessary. Does *not* actually create any directories. */
1028 url_filename (const struct urlinfo *u)
1031 int have_prefix = 0; /* whether we must prepend opt.dir_prefix */
1035 file = mkstruct (u);
1041 file = xstrdup ("index.html");
1043 file = xstrdup (u->file);
1048 /* Check whether the prefix directory is something other than "."
1049 before prepending it. */
1050 if (!DOTP (opt.dir_prefix))
1052 char *nfile = (char *)xmalloc (strlen (opt.dir_prefix)
1053 + 1 + strlen (file) + 1);
1054 sprintf (nfile, "%s/%s", opt.dir_prefix, file);
1059 /* DOS-ish file systems don't like `%' signs in them; we change it
1064 for (p = file; *p; p++)
1068 #endif /* WINDOWS */
1070 /* Check the cases in which the unique extensions are not used:
1071 1) Clobbering is turned off (-nc).
1072 2) Retrieval with regetting.
1073 3) Timestamping is used.
1074 4) Hierarchy is built.
1076 The exception is the case when file does exist and is a
1077 directory (actually support for bad httpd-s). */
1078 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1079 && !(file_exists_p (file) && !file_non_directory_p (file)))
1082 /* Find a unique name. */
1083 name = unique_name (file);
1088 /* Like strlen(), but allow the URL to be ended with '?'. */
1090 urlpath_length (const char *url)
1092 const char *q = strchr (url, '?');
1095 return strlen (url);
1098 /* Find the last occurrence of character C in the range [b, e), or
1099 NULL, if none are present. This is almost completely equivalent to
1100 { *e = '\0'; return strrchr(b); }, except that it doesn't change
1101 the contents of the string. */
1103 find_last_char (const char *b, const char *e, char c)
1111 /* Construct a URL by concatenating an absolute URL and a path, which
1112 may or may not be absolute. This tries to behave "reasonably" in
1113 all foreseeable cases. It employs little specific knowledge about
1114 protocols or URL-specific stuff -- it just works on strings. */
1116 construct (const char *url, const char *sub, int subsize, int no_proto)
1122 const char *end = url + urlpath_length (url);
1126 /* SUB is a relative URL: we need to replace everything
1127 after last slash (possibly empty) with SUB.
1129 So, if URL is "whatever/foo/bar", and SUB is "qux/xyzzy",
1130 our result should be "whatever/foo/qux/xyzzy". */
1131 int need_explicit_slash = 0;
1133 const char *start_insert;
1134 const char *last_slash = find_last_char (url, end, '/'); /* the last slash. */
1137 /* No slash found at all. Append SUB to what we have,
1138 but we'll need a slash as a separator.
1140 Example: if url == "foo" and sub == "qux/xyzzy", then
1141 we cannot just append sub to url, because we'd get
1142 "fooqux/xyzzy", whereas what we want is
1145 To make sure the / gets inserted, we set
1146 need_explicit_slash to 1. We also set start_insert
1147 to end + 1, so that the length calculations work out
1148 correctly for one more (slash) character. Accessing
1149 that character is fine, since it will be the
1150 delimiter, '\0' or '?'. */
1151 /* example: "foo?..." */
1152 /* ^ ('?' gets changed to '/') */
1153 start_insert = end + 1;
1154 need_explicit_slash = 1;
1156 else if (last_slash && last_slash != url && *(last_slash - 1) == '/')
1158 /* example: http://host" */
1160 start_insert = end + 1;
1161 need_explicit_slash = 1;
1165 /* example: "whatever/foo/bar" */
1167 start_insert = last_slash + 1;
1170 span = start_insert - url;
1171 constr = (char *)xmalloc (span + subsize + 1);
1173 memcpy (constr, url, span);
1174 if (need_explicit_slash)
1175 constr[span - 1] = '/';
1177 memcpy (constr + span, sub, subsize);
1178 constr[span + subsize] = '\0';
1180 else /* *sub == `/' */
1182 /* SUB is an absolute path: we need to replace everything
1183 after (and including) the FIRST slash with SUB.
1185 So, if URL is "http://host/whatever/foo/bar", and SUB is
1186 "/qux/xyzzy", our result should be
1187 "http://host/qux/xyzzy". */
1190 const char *start_insert = NULL; /* for gcc to shut up. */
1191 const char *pos = url;
1192 int seen_slash_slash = 0;
1193 /* We're looking for the first slash, but want to ignore
1196 slash = memchr (pos, '/', end - pos);
1197 if (slash && !seen_slash_slash)
1198 if (*(slash + 1) == '/')
1201 seen_slash_slash = 1;
1205 /* At this point, SLASH is the location of the first / after
1206 "//", or the first slash altogether. START_INSERT is the
1207 pointer to the location where SUB will be inserted. When
1208 examining the last two examples, keep in mind that SUB
1211 if (!slash && !seen_slash_slash)
1212 /* example: "foo" */
1215 else if (!slash && seen_slash_slash)
1216 /* example: "http://foo" */
1219 else if (slash && !seen_slash_slash)
1220 /* example: "foo/bar" */
1223 else if (slash && seen_slash_slash)
1224 /* example: "http://something/" */
1226 start_insert = slash;
1228 span = start_insert - url;
1229 constr = (char *)xmalloc (span + subsize + 1);
1231 memcpy (constr, url, span);
1233 memcpy (constr + span, sub, subsize);
1234 constr[span + subsize] = '\0';
1237 else /* !no_proto */
1239 constr = strdupdelim (sub, sub + subsize);
1244 /* Like the function above, but with a saner caller interface. */
1246 url_concat (const char *base_url, const char *new_url)
1248 return construct (base_url, new_url, strlen (new_url), !has_proto (new_url));
1251 /* Optimize URL by host, destructively replacing u->host with realhost
1252 (u->host). Do this regardless of opt.simple_check. */
1254 opt_url (struct urlinfo *u)
1256 /* Find the "true" host. */
1257 char *host = realhost (u->host);
1260 assert (u->dir != NULL); /* the URL must have been parsed */
1261 /* Refresh the printed representation. */
1263 u->url = str_url (u, 0);
1266 /* This beautiful kludge is fortunately not needed, as I've made
1267 parse_dir do the (almost) right thing, so that a query can never
1268 become a part of directory. */
1270 /* Call path_simplify, but make sure that the part after the
1271 question-mark, if any, is not destroyed by path_simplify's
1274 path_simplify_with_kludge (char *path)
1276 char *query = strchr (path, '?');
1278 /* path_simplify also works destructively, so we also have the
1279 license to write. */
1281 path_simplify (path);
1284 char *newend = path + strlen (path);
1286 if (newend != query)
1287 memmove (newend, query, strlen (query) + 1);
1292 /* Returns proxy host address, in accordance with PROTO. */
1294 getproxy (uerr_t proto)
1296 if (proto == URLHTTP)
1297 return opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
1298 else if (proto == URLFTP)
1299 return opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
1301 else if (proto == URLHTTPS)
1302 return opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
1303 #endif /* HAVE_SSL */
1308 /* Should a host be accessed through proxy, concerning no_proxy? */
1310 no_proxy_match (const char *host, const char **no_proxy)
1315 return !sufmatch (no_proxy, host);
1318 static void write_backup_file PARAMS ((const char *, downloaded_file_t));
1319 static void replace_attr PARAMS ((const char **, int, FILE *, const char *));
1321 /* Change the links in an HTML document. Accepts a structure that
1322 defines the positions of all the links. */
1324 convert_links (const char *file, urlpos *l)
1326 struct file_memory *fm;
1329 downloaded_file_t downloaded_file_return;
1331 logprintf (LOG_VERBOSE, _("Converting %s... "), file);
1334 /* First we do a "dry run": go through the list L and see whether
1335 any URL needs to be converted in the first place. If not, just
1336 leave the file alone. */
1339 for (dry = l; dry; dry = dry->next)
1340 if (dry->convert != CO_NOCONVERT)
1344 logputs (LOG_VERBOSE, _("nothing to do.\n"));
1349 fm = read_file (file);
1352 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1353 file, strerror (errno));
1357 downloaded_file_return = downloaded_file (CHECK_FOR_FILE, file);
1358 if (opt.backup_converted && downloaded_file_return)
1359 write_backup_file (file, downloaded_file_return);
1361 /* Before opening the file for writing, unlink the file. This is
1362 important if the data in FM is mmaped. In such case, nulling the
1363 file, which is what fopen() below does, would make us read all
1364 zeroes from the mmaped region. */
1365 if (unlink (file) < 0 && errno != ENOENT)
1367 logprintf (LOG_NOTQUIET, _("Unable to delete `%s': %s\n"),
1368 file, strerror (errno));
1369 read_file_free (fm);
1372 /* Now open the file for writing. */
1373 fp = fopen (file, "wb");
1376 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1377 file, strerror (errno));
1378 read_file_free (fm);
1381 /* Here we loop through all the URLs in file, replacing those of
1382 them that are downloaded with relative references. */
1384 for (; l; l = l->next)
1386 char *url_start = fm->content + l->pos;
1388 if (l->pos >= fm->length)
1390 DEBUGP (("Something strange is going on. Please investigate."));
1393 /* If the URL is not to be converted, skip it. */
1394 if (l->convert == CO_NOCONVERT)
1396 DEBUGP (("Skipping %s at position %d.\n", l->url, l->pos));
1400 /* Echo the file contents, up to the offending URL's opening
1401 quote, to the outfile. */
1402 fwrite (p, 1, url_start - p, fp);
1404 if (l->convert == CO_CONVERT_TO_RELATIVE)
1406 /* Convert absolute URL to relative. */
1407 char *newname = construct_relative (file, l->local_name);
1408 char *quoted_newname = html_quote_string (newname);
1409 replace_attr (&p, l->size, fp, quoted_newname);
1410 DEBUGP (("TO_RELATIVE: %s to %s at position %d in %s.\n",
1411 l->url, newname, l->pos, file));
1413 xfree (quoted_newname);
1415 else if (l->convert == CO_CONVERT_TO_COMPLETE)
1417 /* Convert the link to absolute URL. */
1418 char *newlink = l->url;
1419 char *quoted_newlink = html_quote_string (newlink);
1420 replace_attr (&p, l->size, fp, quoted_newlink);
1421 DEBUGP (("TO_COMPLETE: <something> to %s at position %d in %s.\n",
1422 newlink, l->pos, file));
1423 xfree (quoted_newlink);
1426 /* Output the rest of the file. */
1427 if (p - fm->content < fm->length)
1428 fwrite (p, 1, fm->length - (p - fm->content), fp);
1430 read_file_free (fm);
1431 logputs (LOG_VERBOSE, _("done.\n"));
1434 /* Construct and return a malloced copy of the relative link from two
1435 pieces of information: local name S1 of the referring file and
1436 local name S2 of the referred file.
1438 So, if S1 is "jagor.srce.hr/index.html" and S2 is
1439 "jagor.srce.hr/images/news.gif", the function will return
1442 Alternately, if S1 is "fly.cc.fer.hr/ioccc/index.html", and S2 is
1443 "fly.cc.fer.hr/images/fly.gif", the function will return
1444 "../images/fly.gif".
1446 Caveats: S1 should not begin with `/', unless S2 also begins with
1447 '/'. S1 should not contain things like ".." and such --
1448 construct_relative ("fly/ioccc/../index.html",
1449 "fly/images/fly.gif") will fail. (A workaround is to call
1450 something like path_simplify() on S1). */
1452 construct_relative (const char *s1, const char *s2)
1454 int i, cnt, sepdirs1;
1458 return xstrdup (s2);
1459 /* S1 should *not* be absolute, if S2 wasn't. */
1460 assert (*s1 != '/');
1462 /* Skip the directories common to both strings. */
1465 while (s1[i] && s2[i]
1470 if (s1[i] == '/' && s2[i] == '/')
1475 for (sepdirs1 = 0; s1[i]; i++)
1478 /* Now, construct the file as of:
1479 - ../ repeated sepdirs1 time
1480 - all the non-mutual directories of S2. */
1481 res = (char *)xmalloc (3 * sepdirs1 + strlen (s2 + cnt) + 1);
1482 for (i = 0; i < sepdirs1; i++)
1483 memcpy (res + 3 * i, "../", 3);
1484 strcpy (res + 3 * i, s2 + cnt);
1488 /* Add URL to the head of the list L. */
1490 add_url (urlpos *l, const char *url, const char *file)
1494 t = (urlpos *)xmalloc (sizeof (urlpos));
1495 memset (t, 0, sizeof (*t));
1496 t->url = xstrdup (url);
1497 t->local_name = xstrdup (file);
1503 write_backup_file (const char *file, downloaded_file_t downloaded_file_return)
1505 /* Rather than just writing over the original .html file with the
1506 converted version, save the former to *.orig. Note we only do
1507 this for files we've _successfully_ downloaded, so we don't
1508 clobber .orig files sitting around from previous invocations. */
1510 /* Construct the backup filename as the original name plus ".orig". */
1511 size_t filename_len = strlen(file);
1512 char* filename_plus_orig_suffix;
1513 boolean already_wrote_backup_file = FALSE;
1514 slist* converted_file_ptr;
1515 static slist* converted_files = NULL;
1517 if (downloaded_file_return == FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED)
1519 /* Just write "orig" over "html". We need to do it this way
1520 because when we're checking to see if we've downloaded the
1521 file before (to see if we can skip downloading it), we don't
1522 know if it's a text/html file. Therefore we don't know yet
1523 at that stage that -E is going to cause us to tack on
1524 ".html", so we need to compare vs. the original URL plus
1525 ".orig", not the original URL plus ".html.orig". */
1526 filename_plus_orig_suffix = alloca (filename_len + 1);
1527 strcpy(filename_plus_orig_suffix, file);
1528 strcpy((filename_plus_orig_suffix + filename_len) - 4, "orig");
1530 else /* downloaded_file_return == FILE_DOWNLOADED_NORMALLY */
1532 /* Append ".orig" to the name. */
1533 filename_plus_orig_suffix = alloca (filename_len + sizeof(".orig"));
1534 strcpy(filename_plus_orig_suffix, file);
1535 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
1538 /* We can get called twice on the same URL thanks to the
1539 convert_all_links() call in main(). If we write the .orig file
1540 each time in such a case, it'll end up containing the first-pass
1541 conversion, not the original file. So, see if we've already been
1542 called on this file. */
1543 converted_file_ptr = converted_files;
1544 while (converted_file_ptr != NULL)
1545 if (strcmp(converted_file_ptr->string, file) == 0)
1547 already_wrote_backup_file = TRUE;
1551 converted_file_ptr = converted_file_ptr->next;
1553 if (!already_wrote_backup_file)
1555 /* Rename <file> to <file>.orig before former gets written over. */
1556 if (rename(file, filename_plus_orig_suffix) != 0)
1557 logprintf (LOG_NOTQUIET, _("Cannot back up %s as %s: %s\n"),
1558 file, filename_plus_orig_suffix, strerror (errno));
1560 /* Remember that we've already written a .orig backup for this file.
1561 Note that we never free this memory since we need it till the
1562 convert_all_links() call, which is one of the last things the
1563 program does before terminating. BTW, I'm not sure if it would be
1564 safe to just set 'converted_file_ptr->string' to 'file' below,
1565 rather than making a copy of the string... Another note is that I
1566 thought I could just add a field to the urlpos structure saying
1567 that we'd written a .orig file for this URL, but that didn't work,
1568 so I had to make this separate list.
1569 -- Dan Harkless <wget@harkless.org>
1571 This [adding a field to the urlpos structure] didn't work
1572 because convert_file() is called twice: once after all its
1573 sublinks have been retrieved in recursive_retrieve(), and
1574 once at the end of the day in convert_all_links(). The
1575 original linked list collected in recursive_retrieve() is
1576 lost after the first invocation of convert_links(), and
1577 convert_all_links() makes a new one (it calls get_urls_html()
1578 for each file it covers.) That's why your first approach didn't
1579 work. The way to make it work is perhaps to make this flag a
1580 field in the `urls_html' list.
1581 -- Hrvoje Niksic <hniksic@arsdigita.com>
1583 converted_file_ptr = xmalloc(sizeof(*converted_file_ptr));
1584 converted_file_ptr->string = xstrdup(file); /* die on out-of-mem. */
1585 converted_file_ptr->next = converted_files;
1586 converted_files = converted_file_ptr;
1590 static int find_fragment PARAMS ((const char *, int, const char **,
1594 replace_attr (const char **pp, int raw_size, FILE *fp, const char *new_str)
1596 const char *p = *pp;
1598 int size = raw_size;
1599 char quote_char = '\"';
1600 const char *frag_beg, *frag_end;
1602 /* Structure of our string is:
1603 "...old-contents..."
1604 <--- l->size ---> (with quotes)
1607 <--- l->size --> (no quotes) */
1609 if (*p == '\"' || *p == '\'')
1614 size -= 2; /* disregard opening and closing quote */
1616 putc (quote_char, fp);
1617 fputs (new_str, fp);
1619 /* Look for fragment identifier, if any. */
1620 if (find_fragment (p, size, &frag_beg, &frag_end))
1621 fwrite (frag_beg, 1, frag_end - frag_beg, fp);
1625 putc (quote_char, fp);
1629 /* Find the first occurrence of '#' in [BEG, BEG+SIZE) that is not
1630 preceded by '&'. If the character is not found, return zero. If
1631 the character is found, return 1 and set BP and EP to point to the
1632 beginning and end of the region.
1634 This is used for finding the fragment indentifiers in URLs. */
1637 find_fragment (const char *beg, int size, const char **bp, const char **ep)
1639 const char *end = beg + size;
1641 for (; beg < end; beg++)
1663 typedef struct _downloaded_file_list {
1665 downloaded_file_t download_type;
1666 struct _downloaded_file_list* next;
1667 } downloaded_file_list;
1669 static downloaded_file_list *downloaded_files;
1671 /* Remembers which files have been downloaded. In the standard case, should be
1672 called with mode == FILE_DOWNLOADED_NORMALLY for each file we actually
1673 download successfully (i.e. not for ones we have failures on or that we skip
1676 When we've downloaded a file and tacked on a ".html" extension due to -E,
1677 call this function with FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED rather than
1678 FILE_DOWNLOADED_NORMALLY.
1680 If you just want to check if a file has been previously added without adding
1681 it, call with mode == CHECK_FOR_FILE. Please be sure to call this function
1682 with local filenames, not remote URLs. */
1684 downloaded_file (downloaded_file_t mode, const char* file)
1686 boolean found_file = FALSE;
1687 downloaded_file_list* rover = downloaded_files;
1689 while (rover != NULL)
1690 if (strcmp(rover->file, file) == 0)
1696 rover = rover->next;
1699 return rover->download_type; /* file had already been downloaded */
1702 if (mode != CHECK_FOR_FILE)
1704 rover = xmalloc(sizeof(*rover));
1705 rover->file = xstrdup(file); /* use xstrdup() so die on out-of-mem. */
1706 rover->download_type = mode;
1707 rover->next = downloaded_files;
1708 downloaded_files = rover;
1711 return FILE_NOT_ALREADY_DOWNLOADED;
1716 downloaded_files_free (void)
1718 downloaded_file_list* rover = downloaded_files;
1721 downloaded_file_list *next = rover->next;
1722 xfree (rover->file);
1728 /* Initialization of static stuff. */
1732 init_unsafe_char_table ();