2 Copyright (C) 1995, 1996, 1997, 2000, 2001 Free Software Foundation, Inc.
4 This file is part of GNU Wget.
6 GNU Wget is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or (at
9 your option) any later version.
11 GNU Wget is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with Wget; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
29 #include <sys/types.h>
46 #define DOTP(x) ((*(x) == '.') && (!*(x + 1)))
48 #define DDOTP(x) ((*(x) == '.') && (*(x + 1) == '.') && (!*(x + 2)))
50 static int urlpath_length PARAMS ((const char *));
52 /* A NULL-terminated list of strings to be recognized as protocol
53 types (URL schemes). Note that recognized doesn't mean supported
54 -- only HTTP, HTTPS and FTP are currently supported.
56 However, a string that does not match anything in the list will be
57 considered a relative URL. Thus it's important that this list has
58 anything anyone could think of being legal.
60 #### This is probably broken. Wget should use other means to
61 distinguish between absolute and relative URIs in HTML links.
63 Take a look at <http://www.w3.org/pub/WWW/Addressing/schemes.html>
65 static char *protostrings[] =
107 /* Similar to former, but for supported protocols: */
108 static struct proto sup_protos[] =
110 { "http://", URLHTTP, DEFAULT_HTTP_PORT },
112 { "https://",URLHTTPS, DEFAULT_HTTPS_PORT},
114 { "ftp://", URLFTP, DEFAULT_FTP_PORT }
117 static void parse_dir PARAMS ((const char *, char **, char **));
118 static uerr_t parse_uname PARAMS ((const char *, char **, char **));
119 static char *construct_relative PARAMS ((const char *, const char *));
120 static char process_ftp_type PARAMS ((char *));
123 /* Support for encoding and decoding of URL strings. We determine
124 whether a character is unsafe through static table lookup. This
125 code assumes ASCII character set and 8-bit chars. */
132 #define R urlchr_reserved
133 #define U urlchr_unsafe
136 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
138 /* rfc1738 reserved chars. We don't use this yet; preservation of
139 reserved chars will be implemented when I integrate the new
140 `reencode_string' function. */
142 #define RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
146 - stuff from rfc1738 ("<>\"#%{}|\\^~[]`");
147 - '@' and ':'; needed for encoding URL username and password.
148 - anything >= 127. */
150 #define UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
152 const static unsigned char urlchr_table[256] =
154 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
155 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
156 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
157 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
158 U, 0, U, U, 0, U, R, 0, /* SP ! " # $ % & ' */
159 0, 0, 0, R, 0, 0, 0, R, /* ( ) * + , - . / */
160 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
161 0, 0, U, R, U, R, U, R, /* 8 9 : ; < = > ? */
162 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
163 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
164 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
165 0, 0, 0, U, U, U, U, 0, /* X Y Z [ \ ] ^ _ */
166 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
167 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
168 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
169 0, 0, 0, U, U, U, U, U, /* x y z { | } ~ DEL */
171 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
172 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
173 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
174 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
176 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
177 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
178 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
179 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
182 /* Decodes the forms %xy in a URL to the character the hexadecimal
183 code of which is xy. xy are hexadecimal digits from
184 [0123456789ABCDEF] (case-insensitive). If x or y are not
185 hex-digits or `%' precedes `\0', the sequence is inserted
189 decode_string (char *s)
191 char *t = s; /* t - tortoise */
192 char *h = s; /* h - hare */
203 /* Do nothing if '%' is not followed by two hex digits. */
204 if (!*(h + 1) || !*(h + 2)
205 || !(ISXDIGIT (*(h + 1)) && ISXDIGIT (*(h + 2))))
207 *t = (XCHAR_TO_XDIGIT (*(h + 1)) << 4) + XCHAR_TO_XDIGIT (*(h + 2));
214 /* Like encode_string, but return S if there are no unsafe chars. */
217 encode_string_maybe (const char *s)
224 for (p1 = s; *p1; p1++)
225 if (UNSAFE_CHAR (*p1))
226 addition += 2; /* Two more characters (hex digits) */
231 newlen = (p1 - s) + addition;
232 newstr = (char *)xmalloc (newlen + 1);
238 if (UNSAFE_CHAR (*p1))
240 const unsigned char c = *p1++;
242 *p2++ = XDIGIT_TO_XCHAR (c >> 4);
243 *p2++ = XDIGIT_TO_XCHAR (c & 0xf);
249 assert (p2 - newstr == newlen);
254 /* Encode the unsafe characters (as determined by UNSAFE_CHAR) in a
255 given string, returning a malloc-ed %XX encoded string. */
258 encode_string (const char *s)
260 char *encoded = encode_string_maybe (s);
267 /* Encode unsafe characters in PTR to %xx. If such encoding is done,
268 the old value of PTR is freed and PTR is made to point to the newly
269 allocated storage. */
271 #define ENCODE(ptr) do { \
272 char *e_new = encode_string_maybe (ptr); \
280 /* Returns the protocol type if URL's protocol is supported, or
281 URLUNKNOWN if not. */
283 urlproto (const char *url)
287 for (i = 0; i < ARRAY_SIZE (sup_protos); i++)
288 if (!strncasecmp (url, sup_protos[i].name, strlen (sup_protos[i].name)))
289 return sup_protos[i].ind;
290 for (i = 0; url[i] && url[i] != ':' && url[i] != '/'; i++);
293 for (++i; url[i] && url[i] != '/'; i++)
294 if (!ISDIGIT (url[i]))
296 if (url[i - 1] == ':')
305 /* Skip the protocol part of the URL, e.g. `http://'. If no protocol
306 part is found, returns 0. */
308 skip_proto (const char *url)
313 for (s = protostrings; *s; s++)
314 if (!strncasecmp (*s, url, strlen (*s)))
319 /* HTTP and FTP protocols are expected to yield exact host names
320 (i.e. the `//' part must be skipped, too). */
321 if (!strcmp (*s, "http:") || !strcmp (*s, "ftp:"))
326 /* Returns 1 if the URL begins with a protocol (supported or
327 unsupported), 0 otherwise. */
329 has_proto (const char *url)
333 for (s = protostrings; *s; s++)
334 if (strncasecmp (url, *s, strlen (*s)) == 0)
339 /* Skip the username and password, if present here. The function
340 should be called *not* with the complete URL, but with the part
341 right after the protocol.
343 If no username and password are found, return 0. */
345 skip_uname (const char *url)
348 const char *q = NULL;
349 for (p = url ; *p && *p != '/'; p++)
350 if (*p == '@') q = p;
351 /* If a `@' was found before the first occurrence of `/', skip
359 /* Allocate a new urlinfo structure, fill it with default values and
360 return a pointer to it. */
366 u = (struct urlinfo *)xmalloc (sizeof (struct urlinfo));
367 memset (u, 0, sizeof (*u));
368 u->proto = URLUNKNOWN;
372 /* Perform a "deep" free of the urlinfo structure. The structure
373 should have been created with newurl, but need not have been used.
374 If free_pointer is non-0, free the pointer itself. */
376 freeurl (struct urlinfo *u, int complete)
380 FREE_MAYBE (u->host);
381 FREE_MAYBE (u->path);
382 FREE_MAYBE (u->file);
384 FREE_MAYBE (u->user);
385 FREE_MAYBE (u->passwd);
386 FREE_MAYBE (u->local);
387 FREE_MAYBE (u->referer);
389 freeurl (u->proxy, 1);
395 /* Extract the given URL of the form
396 (http:|ftp:)// (user (:password)?@)?hostname (:port)? (/path)?
397 1. hostname (terminated with `/' or `:')
398 2. port number (terminated with `/'), or chosen for the protocol
399 3. dirname (everything after hostname)
400 Most errors are handled. No allocation is done, you must supply
401 pointers to allocated memory.
402 ...and a host of other stuff :-)
404 - Recognizes hostname:dir/file for FTP and
405 hostname (:portnum)?/dir/file for HTTP.
406 - Parses the path to yield directory and file
407 - Parses the URL to yield the username and passwd (if present)
408 - Decodes the strings, in case they contain "forbidden" characters
409 - Writes the result to struct urlinfo
411 If the argument STRICT is set, it recognizes only the canonical
414 parseurl (const char *url, struct urlinfo *u, int strict)
417 int recognizable; /* Recognizable URL is the one where
418 the protocol name was explicitly
419 named, i.e. it wasn't deduced from
423 DEBUGP (("parseurl (\"%s\") -> ", url));
424 recognizable = has_proto (url);
425 if (strict && !recognizable)
427 for (i = 0, l = 0; i < ARRAY_SIZE (sup_protos); i++)
429 l = strlen (sup_protos[i].name);
430 if (!strncasecmp (sup_protos[i].name, url, l))
433 /* If protocol is recognizable, but unsupported, bail out, else
435 if (recognizable && i == ARRAY_SIZE (sup_protos))
437 else if (i == ARRAY_SIZE (sup_protos))
440 u->proto = type = sup_protos[i].ind;
442 if (type == URLUNKNOWN)
444 /* Allow a username and password to be specified (i.e. just skip
447 l += skip_uname (url + l);
448 for (i = l; url[i] && url[i] != ':' && url[i] != '/'; i++);
451 /* Get the hostname. */
452 u->host = strdupdelim (url + l, url + i);
453 DEBUGP (("host %s -> ", u->host));
455 /* Assume no port has been given. */
459 /* We have a colon delimiting the hostname. It could mean that
460 a port number is following it, or a directory. */
461 if (ISDIGIT (url[++i])) /* A port number */
463 if (type == URLUNKNOWN)
464 u->proto = type = URLHTTP;
465 for (; url[i] && url[i] != '/'; i++)
466 if (ISDIGIT (url[i]))
467 u->port = 10 * u->port + (url[i] - '0');
472 DEBUGP (("port %hu -> ", u->port));
474 else if (type == URLUNKNOWN) /* or a directory */
475 u->proto = type = URLFTP;
476 else /* or just a misformed port number */
479 else if (type == URLUNKNOWN)
480 u->proto = type = URLHTTP;
484 for (ind = 0; ind < ARRAY_SIZE (sup_protos); ind++)
485 if (sup_protos[ind].ind == type)
487 if (ind == ARRAY_SIZE (sup_protos))
489 u->port = sup_protos[ind].port;
491 /* Some delimiter troubles... */
492 if (url[i] == '/' && url[i - 1] != ':')
495 while (url[i] && url[i] == '/')
497 u->path = (char *)xmalloc (strlen (url + i) + 8);
498 strcpy (u->path, url + i);
501 u->ftp_type = process_ftp_type (u->path);
502 /* #### We don't handle type `d' correctly yet. */
503 if (!u->ftp_type || TOUPPER (u->ftp_type) == 'D')
505 DEBUGP (("ftp_type %c -> ", u->ftp_type));
507 DEBUGP (("opath %s -> ", u->path));
508 /* Parse the username and password (if existing). */
509 parse_uname (url, &u->user, &u->passwd);
510 /* Decode the strings, as per RFC 1738. */
511 decode_string (u->host);
512 decode_string (u->path);
514 decode_string (u->user);
516 decode_string (u->passwd);
517 /* Parse the directory. */
518 parse_dir (u->path, &u->dir, &u->file);
519 DEBUGP (("dir %s -> file %s -> ", u->dir, u->file));
520 /* Simplify the directory. */
521 path_simplify (u->dir);
522 /* Remove the leading `/' in HTTP. */
523 if (type == URLHTTP && *u->dir == '/')
524 strcpy (u->dir, u->dir + 1);
525 DEBUGP (("ndir %s\n", u->dir));
526 /* Strip trailing `/'. */
528 if (l > 1 && u->dir[l - 1] == '/')
529 u->dir[l - 1] = '\0';
530 /* Re-create the path: */
531 abs_ftp = (u->proto == URLFTP && *u->dir == '/');
532 /* sprintf (u->path, "%s%s%s%s", abs_ftp ? "%2F": "/",
533 abs_ftp ? (u->dir + 1) : u->dir, *u->dir ? "/" : "", u->file); */
534 strcpy (u->path, abs_ftp ? "%2F" : "/");
535 strcat (u->path, abs_ftp ? (u->dir + 1) : u->dir);
536 strcat (u->path, *u->dir ? "/" : "");
537 strcat (u->path, u->file);
539 DEBUGP (("newpath: %s\n", u->path));
540 /* Create the clean URL. */
541 u->url = str_url (u, 0);
545 /* Special versions of DOTP and DDOTP for parse_dir(). They work like
546 DOTP and DDOTP, but they also recognize `?' as end-of-string
547 delimiter. This is needed for correct handling of query
550 #define PD_DOTP(x) ((*(x) == '.') && (!*((x) + 1) || *((x) + 1) == '?'))
551 #define PD_DDOTP(x) ((*(x) == '.') && (*(x) == '.') \
552 && (!*((x) + 2) || *((x) + 2) == '?'))
554 /* Build the directory and filename components of the path. Both
555 components are *separately* malloc-ed strings! It does not change
556 the contents of path.
558 If the path ends with "." or "..", they are (correctly) counted as
561 parse_dir (const char *path, char **dir, char **file)
565 l = urlpath_length (path);
566 for (i = l; i && path[i] != '/'; i--);
568 if (!i && *path != '/') /* Just filename */
570 if (PD_DOTP (path) || PD_DDOTP (path))
572 *dir = strdupdelim (path, path + l);
573 *file = xstrdup (path + l); /* normally empty, but could
578 *dir = xstrdup (""); /* This is required because of FTP */
579 *file = xstrdup (path);
582 else if (!i) /* /filename */
584 if (PD_DOTP (path + 1) || PD_DDOTP (path + 1))
586 *dir = strdupdelim (path, path + l);
587 *file = xstrdup (path + l); /* normally empty, but could
592 *dir = xstrdup ("/");
593 *file = xstrdup (path + 1);
596 else /* Nonempty directory with or without a filename */
598 if (PD_DOTP (path + i + 1) || PD_DDOTP (path + i + 1))
600 *dir = strdupdelim (path, path + l);
601 *file = xstrdup (path + l); /* normally empty, but could
606 *dir = strdupdelim (path, path + i);
607 *file = xstrdup (path + i + 1);
612 /* Find the optional username and password within the URL, as per
613 RFC1738. The returned user and passwd char pointers are
616 parse_uname (const char *url, char **user, char **passwd)
619 const char *p, *q, *col;
625 /* Look for the end of the protocol string. */
626 l = skip_proto (url);
629 /* Add protocol offset. */
631 /* Is there an `@' character? */
632 for (p = url; *p && *p != '/'; p++)
635 /* If not, return. */
638 /* Else find the username and password. */
639 for (p = q = col = url; *p && *p != '/'; p++)
641 if (*p == ':' && !*user)
643 *user = (char *)xmalloc (p - url + 1);
644 memcpy (*user, url, p - url);
645 (*user)[p - url] = '\0';
648 if (*p == '@') q = p;
650 /* Decide whether you have only the username or both. */
651 where = *user ? passwd : user;
652 *where = (char *)xmalloc (q - col + 1);
653 memcpy (*where, col, q - col);
654 (*where)[q - col] = '\0';
658 /* If PATH ends with `;type=X', return the character X. */
660 process_ftp_type (char *path)
662 int len = strlen (path);
665 && !memcmp (path + len - 7, ";type=", 6))
667 path[len - 7] = '\0';
668 return path[len - 1];
674 /* Return the URL as fine-formed string, with a proper protocol, optional port
675 number, directory and optional user/password. If `hide' is non-zero (as it
676 is when we're calling this on a URL we plan to print, but not when calling it
677 to canonicalize a URL for use within the program), password will be hidden.
678 The forbidden characters in the URL will be cleansed. */
680 str_url (const struct urlinfo *u, int hide)
682 char *res, *host, *user, *passwd, *proto_name, *dir, *file;
683 int i, l, ln, lu, lh, lp, lf, ld;
684 unsigned short proto_default_port;
686 /* Look for the protocol name. */
687 for (i = 0; i < ARRAY_SIZE (sup_protos); i++)
688 if (sup_protos[i].ind == u->proto)
690 if (i == ARRAY_SIZE (sup_protos))
692 proto_name = sup_protos[i].name;
693 proto_default_port = sup_protos[i].port;
694 host = encode_string (u->host);
695 dir = encode_string (u->dir);
696 file = encode_string (u->file);
697 user = passwd = NULL;
699 user = encode_string (u->user);
703 /* Don't output the password, or someone might see it over the user's
704 shoulder (or in saved wget output). Don't give away the number of
705 characters in the password, either, as we did in past versions of
706 this code, when we replaced the password characters with 'x's. */
707 passwd = xstrdup("<password>");
709 passwd = encode_string (u->passwd);
711 if (u->proto == URLFTP && *dir == '/')
713 char *tmp = (char *)xmalloc (strlen (dir) + 3);
714 /*sprintf (tmp, "%%2F%s", dir + 1);*/
718 strcpy (tmp + 3, dir + 1);
723 ln = strlen (proto_name);
724 lu = user ? strlen (user) : 0;
725 lp = passwd ? strlen (passwd) : 0;
729 res = (char *)xmalloc (ln + lu + lp + lh + ld + lf + 20); /* safe sex */
730 /* sprintf (res, "%s%s%s%s%s%s:%d/%s%s%s", proto_name,
731 (user ? user : ""), (passwd ? ":" : ""),
732 (passwd ? passwd : ""), (user ? "@" : ""),
733 host, u->port, dir, *dir ? "/" : "", file); */
735 memcpy (res, proto_name, ln);
739 memcpy (res + l, user, lu);
744 memcpy (res + l, passwd, lp);
749 memcpy (res + l, host, lh);
751 if (u->port != proto_default_port)
754 long_to_string (res + l, (long)u->port);
755 l += numdigit (u->port);
758 memcpy (res + l, dir, ld);
762 strcpy (res + l, file);
771 /* Check whether two URL-s are equivalent, i.e. pointing to the same
772 location. Uses parseurl to parse them, and compares the canonical
775 Returns 1 if the URL1 is equivalent to URL2, 0 otherwise. Also
776 return 0 on error. */
778 url_equal (const char *url1, const char *url2)
780 struct urlinfo *u1, *u2;
785 err = parseurl (url1, u1, 0);
792 err = parseurl (url2, u2, 0);
798 res = !strcmp (u1->url, u2->url);
805 get_urls_file (const char *file)
807 struct file_memory *fm;
809 const char *text, *text_end;
812 fm = read_file (file);
815 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
818 DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
821 text_end = fm->content + fm->length;
822 while (text < text_end)
824 const char *line_beg = text;
825 const char *line_end = memchr (text, '\n', text_end - text);
831 while (line_beg < line_end
832 && ISSPACE (*line_beg))
834 while (line_end > line_beg + 1
835 && ISSPACE (*(line_end - 1)))
837 if (line_end > line_beg)
839 urlpos *entry = (urlpos *)xmalloc (sizeof (urlpos));
840 memset (entry, 0, sizeof (*entry));
842 entry->url = strdupdelim (line_beg, line_end);
854 /* Free the linked list of urlpos. */
856 free_urlpos (urlpos *l)
860 urlpos *next = l->next;
862 FREE_MAYBE (l->local_name);
868 /* Rotate FNAME opt.backups times */
870 rotate_backups(const char *fname)
872 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
873 char *from = (char *)alloca (maxlen);
874 char *to = (char *)alloca (maxlen);
878 if (stat (fname, &sb) == 0)
879 if (S_ISREG (sb.st_mode) == 0)
882 for (i = opt.backups; i > 1; i--)
884 sprintf (from, "%s.%d", fname, i - 1);
885 sprintf (to, "%s.%d", fname, i);
886 /* #### This will fail on machines without the rename() system
891 sprintf (to, "%s.%d", fname, 1);
895 /* Create all the necessary directories for PATH (a file). Calls
896 mkdirhier() internally. */
898 mkalldirs (const char *path)
905 p = path + strlen (path);
906 for (; *p != '/' && p != path; p--);
907 /* Don't create if it's just a file. */
908 if ((p == path) && (*p != '/'))
910 t = strdupdelim (path, p);
911 /* Check whether the directory exists. */
912 if ((stat (t, &st) == 0))
914 if (S_ISDIR (st.st_mode))
921 /* If the dir exists as a file name, remove it first. This
922 is *only* for Wget to work with buggy old CERN http
923 servers. Here is the scenario: When Wget tries to
924 retrieve a directory without a slash, e.g.
925 http://foo/bar (bar being a directory), CERN server will
926 not redirect it too http://foo/bar/ -- it will generate a
927 directory listing containing links to bar/file1,
928 bar/file2, etc. Wget will lose because it saves this
929 HTML listing to a file `bar', so it cannot create the
930 directory. To work around this, if the file of the same
931 name exists, we just remove it and create the directory
933 DEBUGP (("Removing %s because of directory danger!\n", t));
937 res = make_directory (t);
939 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
945 count_slashes (const char *s)
954 /* Return the path name of the URL-equivalent file name, with a
955 remote-like structure of directories. */
957 mkstruct (const struct urlinfo *u)
959 char *host, *dir, *file, *res, *dirpref;
962 assert (u->dir != NULL);
963 assert (u->host != NULL);
967 char *ptr = u->dir + (*u->dir == '/');
968 int slash_count = 1 + count_slashes (ptr);
969 int cut = MINVAL (opt.cut_dirs, slash_count);
970 for (; cut && *ptr; ptr++)
973 STRDUP_ALLOCA (dir, ptr);
976 dir = u->dir + (*u->dir == '/');
978 host = xstrdup (u->host);
979 /* Check for the true name (or at least a consistent name for saving
980 to directory) of HOST, reusing the hlist if possible. */
981 if (opt.add_hostdir && !opt.simple_check)
983 char *nhost = realhost (host);
987 /* Add dir_prefix and hostname (if required) to the beginning of
991 if (!DOTP (opt.dir_prefix))
993 dirpref = (char *)alloca (strlen (opt.dir_prefix) + 1
994 + strlen (host) + 1);
995 sprintf (dirpref, "%s/%s", opt.dir_prefix, host);
998 STRDUP_ALLOCA (dirpref, host);
1000 else /* not add_hostdir */
1002 if (!DOTP (opt.dir_prefix))
1003 dirpref = opt.dir_prefix;
1009 /* If there is a prefix, prepend it. */
1012 char *newdir = (char *)alloca (strlen (dirpref) + 1 + strlen (dir) + 2);
1013 sprintf (newdir, "%s%s%s", dirpref, *dir == '/' ? "" : "/", dir);
1016 dir = encode_string (dir);
1018 if (l && dir[l - 1] == '/')
1022 file = "index.html";
1026 /* Finally, construct the full name. */
1027 res = (char *)xmalloc (strlen (dir) + 1 + strlen (file) + 1);
1028 sprintf (res, "%s%s%s", dir, *dir ? "/" : "", file);
1033 /* Create a unique filename, corresponding to a given URL. Calls
1034 mkstruct if necessary. Does *not* actually create any directories. */
1036 url_filename (const struct urlinfo *u)
1039 int have_prefix = 0; /* whether we must prepend opt.dir_prefix */
1043 file = mkstruct (u);
1049 file = xstrdup ("index.html");
1051 file = xstrdup (u->file);
1056 /* Check whether the prefix directory is something other than "."
1057 before prepending it. */
1058 if (!DOTP (opt.dir_prefix))
1060 char *nfile = (char *)xmalloc (strlen (opt.dir_prefix)
1061 + 1 + strlen (file) + 1);
1062 sprintf (nfile, "%s/%s", opt.dir_prefix, file);
1067 /* DOS-ish file systems don't like `%' signs in them; we change it
1072 for (p = file; *p; p++)
1076 #endif /* WINDOWS */
1078 /* Check the cases in which the unique extensions are not used:
1079 1) Clobbering is turned off (-nc).
1080 2) Retrieval with regetting.
1081 3) Timestamping is used.
1082 4) Hierarchy is built.
1084 The exception is the case when file does exist and is a
1085 directory (actually support for bad httpd-s). */
1086 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1087 && !(file_exists_p (file) && !file_non_directory_p (file)))
1090 /* Find a unique name. */
1091 name = unique_name (file);
1096 /* Like strlen(), but allow the URL to be ended with '?'. */
1098 urlpath_length (const char *url)
1100 const char *q = strchr (url, '?');
1103 return strlen (url);
1106 /* Find the last occurrence of character C in the range [b, e), or
1107 NULL, if none are present. This is almost completely equivalent to
1108 { *e = '\0'; return strrchr(b); }, except that it doesn't change
1109 the contents of the string. */
1111 find_last_char (const char *b, const char *e, char c)
1119 /* Resolve the result of "linking" a base URI (BASE) to a
1120 link-specified URI (LINK).
1122 Either of the URIs may be absolute or relative, complete with the
1123 host name, or path only. This tries to behave "reasonably" in all
1124 foreseeable cases. It employs little specific knowledge about
1125 protocols or URL-specific stuff -- it just works on strings.
1127 The parameters LINKLENGTH is useful if LINK is not zero-terminated.
1128 See uri_merge for a gentler interface to this functionality.
1130 #### This function should handle `./' and `../' so that the evil
1131 path_simplify can go. */
1133 uri_merge_1 (const char *base, const char *link, int linklength, int no_proto)
1139 const char *end = base + urlpath_length (base);
1143 /* LINK is a relative URL: we need to replace everything
1144 after last slash (possibly empty) with LINK.
1146 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1147 our result should be "whatever/foo/qux/xyzzy". */
1148 int need_explicit_slash = 0;
1150 const char *start_insert;
1151 const char *last_slash = find_last_char (base, end, '/');
1154 /* No slash found at all. Append LINK to what we have,
1155 but we'll need a slash as a separator.
1157 Example: if base == "foo" and link == "qux/xyzzy", then
1158 we cannot just append link to base, because we'd get
1159 "fooqux/xyzzy", whereas what we want is
1162 To make sure the / gets inserted, we set
1163 need_explicit_slash to 1. We also set start_insert
1164 to end + 1, so that the length calculations work out
1165 correctly for one more (slash) character. Accessing
1166 that character is fine, since it will be the
1167 delimiter, '\0' or '?'. */
1168 /* example: "foo?..." */
1169 /* ^ ('?' gets changed to '/') */
1170 start_insert = end + 1;
1171 need_explicit_slash = 1;
1173 else if (last_slash && last_slash != base && *(last_slash - 1) == '/')
1175 /* example: http://host" */
1177 start_insert = end + 1;
1178 need_explicit_slash = 1;
1182 /* example: "whatever/foo/bar" */
1184 start_insert = last_slash + 1;
1187 span = start_insert - base;
1188 constr = (char *)xmalloc (span + linklength + 1);
1190 memcpy (constr, base, span);
1191 if (need_explicit_slash)
1192 constr[span - 1] = '/';
1194 memcpy (constr + span, link, linklength);
1195 constr[span + linklength] = '\0';
1197 else /* *link == `/' */
1199 /* LINK is an absolute path: we need to replace everything
1200 after (and including) the FIRST slash with LINK.
1202 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1203 "/qux/xyzzy", our result should be
1204 "http://host/qux/xyzzy". */
1207 const char *start_insert = NULL; /* for gcc to shut up. */
1208 const char *pos = base;
1209 int seen_slash_slash = 0;
1210 /* We're looking for the first slash, but want to ignore
1213 slash = memchr (pos, '/', end - pos);
1214 if (slash && !seen_slash_slash)
1215 if (*(slash + 1) == '/')
1218 seen_slash_slash = 1;
1222 /* At this point, SLASH is the location of the first / after
1223 "//", or the first slash altogether. START_INSERT is the
1224 pointer to the location where LINK will be inserted. When
1225 examining the last two examples, keep in mind that LINK
1228 if (!slash && !seen_slash_slash)
1229 /* example: "foo" */
1231 start_insert = base;
1232 else if (!slash && seen_slash_slash)
1233 /* example: "http://foo" */
1236 else if (slash && !seen_slash_slash)
1237 /* example: "foo/bar" */
1239 start_insert = base;
1240 else if (slash && seen_slash_slash)
1241 /* example: "http://something/" */
1243 start_insert = slash;
1245 span = start_insert - base;
1246 constr = (char *)xmalloc (span + linklength + 1);
1248 memcpy (constr, base, span);
1250 memcpy (constr + span, link, linklength);
1251 constr[span + linklength] = '\0';
1254 else /* !no_proto */
1256 constr = strdupdelim (link, link + linklength);
1261 /* Merge BASE with LINK and return the resulting URI. This is an
1262 interface to uri_merge_1 that assumes that LINK is a
1263 zero-terminated string. */
1265 uri_merge (const char *base, const char *link)
1267 return uri_merge_1 (base, link, strlen (link), !has_proto (link));
1270 /* Optimize URL by host, destructively replacing u->host with realhost
1271 (u->host). Do this regardless of opt.simple_check. */
1273 opt_url (struct urlinfo *u)
1275 /* Find the "true" host. */
1276 char *host = realhost (u->host);
1279 assert (u->dir != NULL); /* the URL must have been parsed */
1280 /* Refresh the printed representation. */
1282 u->url = str_url (u, 0);
1285 /* Returns proxy host address, in accordance with PROTO. */
1287 getproxy (uerr_t proto)
1291 if (proto == URLHTTP)
1292 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
1293 else if (proto == URLFTP)
1294 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
1296 else if (proto == URLHTTPS)
1297 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
1298 #endif /* HAVE_SSL */
1301 if (!proxy || !*proxy)
1306 /* Should a host be accessed through proxy, concerning no_proxy? */
1308 no_proxy_match (const char *host, const char **no_proxy)
1313 return !sufmatch (no_proxy, host);
1316 static void write_backup_file PARAMS ((const char *, downloaded_file_t));
1317 static void replace_attr PARAMS ((const char **, int, FILE *, const char *));
1319 /* Change the links in an HTML document. Accepts a structure that
1320 defines the positions of all the links. */
1322 convert_links (const char *file, urlpos *l)
1324 struct file_memory *fm;
1327 downloaded_file_t downloaded_file_return;
1329 logprintf (LOG_VERBOSE, _("Converting %s... "), file);
1332 /* First we do a "dry run": go through the list L and see whether
1333 any URL needs to be converted in the first place. If not, just
1334 leave the file alone. */
1337 for (dry = l; dry; dry = dry->next)
1338 if (dry->convert != CO_NOCONVERT)
1342 logputs (LOG_VERBOSE, _("nothing to do.\n"));
1347 fm = read_file (file);
1350 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1351 file, strerror (errno));
1355 downloaded_file_return = downloaded_file (CHECK_FOR_FILE, file);
1356 if (opt.backup_converted && downloaded_file_return)
1357 write_backup_file (file, downloaded_file_return);
1359 /* Before opening the file for writing, unlink the file. This is
1360 important if the data in FM is mmaped. In such case, nulling the
1361 file, which is what fopen() below does, would make us read all
1362 zeroes from the mmaped region. */
1363 if (unlink (file) < 0 && errno != ENOENT)
1365 logprintf (LOG_NOTQUIET, _("Unable to delete `%s': %s\n"),
1366 file, strerror (errno));
1367 read_file_free (fm);
1370 /* Now open the file for writing. */
1371 fp = fopen (file, "wb");
1374 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1375 file, strerror (errno));
1376 read_file_free (fm);
1379 /* Here we loop through all the URLs in file, replacing those of
1380 them that are downloaded with relative references. */
1382 for (; l; l = l->next)
1384 char *url_start = fm->content + l->pos;
1386 if (l->pos >= fm->length)
1388 DEBUGP (("Something strange is going on. Please investigate."));
1391 /* If the URL is not to be converted, skip it. */
1392 if (l->convert == CO_NOCONVERT)
1394 DEBUGP (("Skipping %s at position %d.\n", l->url, l->pos));
1398 /* Echo the file contents, up to the offending URL's opening
1399 quote, to the outfile. */
1400 fwrite (p, 1, url_start - p, fp);
1402 if (l->convert == CO_CONVERT_TO_RELATIVE)
1404 /* Convert absolute URL to relative. */
1405 char *newname = construct_relative (file, l->local_name);
1406 char *quoted_newname = html_quote_string (newname);
1407 replace_attr (&p, l->size, fp, quoted_newname);
1408 DEBUGP (("TO_RELATIVE: %s to %s at position %d in %s.\n",
1409 l->url, newname, l->pos, file));
1411 xfree (quoted_newname);
1413 else if (l->convert == CO_CONVERT_TO_COMPLETE)
1415 /* Convert the link to absolute URL. */
1416 char *newlink = l->url;
1417 char *quoted_newlink = html_quote_string (newlink);
1418 replace_attr (&p, l->size, fp, quoted_newlink);
1419 DEBUGP (("TO_COMPLETE: <something> to %s at position %d in %s.\n",
1420 newlink, l->pos, file));
1421 xfree (quoted_newlink);
1424 /* Output the rest of the file. */
1425 if (p - fm->content < fm->length)
1426 fwrite (p, 1, fm->length - (p - fm->content), fp);
1428 read_file_free (fm);
1429 logputs (LOG_VERBOSE, _("done.\n"));
1432 /* Construct and return a malloced copy of the relative link from two
1433 pieces of information: local name S1 of the referring file and
1434 local name S2 of the referred file.
1436 So, if S1 is "jagor.srce.hr/index.html" and S2 is
1437 "jagor.srce.hr/images/news.gif", the function will return
1440 Alternately, if S1 is "fly.cc.fer.hr/ioccc/index.html", and S2 is
1441 "fly.cc.fer.hr/images/fly.gif", the function will return
1442 "../images/fly.gif".
1444 Caveats: S1 should not begin with `/', unless S2 also begins with
1445 '/'. S1 should not contain things like ".." and such --
1446 construct_relative ("fly/ioccc/../index.html",
1447 "fly/images/fly.gif") will fail. (A workaround is to call
1448 something like path_simplify() on S1). */
1450 construct_relative (const char *s1, const char *s2)
1452 int i, cnt, sepdirs1;
1456 return xstrdup (s2);
1457 /* S1 should *not* be absolute, if S2 wasn't. */
1458 assert (*s1 != '/');
1460 /* Skip the directories common to both strings. */
1463 while (s1[i] && s2[i]
1468 if (s1[i] == '/' && s2[i] == '/')
1473 for (sepdirs1 = 0; s1[i]; i++)
1476 /* Now, construct the file as of:
1477 - ../ repeated sepdirs1 time
1478 - all the non-mutual directories of S2. */
1479 res = (char *)xmalloc (3 * sepdirs1 + strlen (s2 + cnt) + 1);
1480 for (i = 0; i < sepdirs1; i++)
1481 memcpy (res + 3 * i, "../", 3);
1482 strcpy (res + 3 * i, s2 + cnt);
1486 /* Add URL to the head of the list L. */
1488 add_url (urlpos *l, const char *url, const char *file)
1492 t = (urlpos *)xmalloc (sizeof (urlpos));
1493 memset (t, 0, sizeof (*t));
1494 t->url = xstrdup (url);
1495 t->local_name = xstrdup (file);
1501 write_backup_file (const char *file, downloaded_file_t downloaded_file_return)
1503 /* Rather than just writing over the original .html file with the
1504 converted version, save the former to *.orig. Note we only do
1505 this for files we've _successfully_ downloaded, so we don't
1506 clobber .orig files sitting around from previous invocations. */
1508 /* Construct the backup filename as the original name plus ".orig". */
1509 size_t filename_len = strlen(file);
1510 char* filename_plus_orig_suffix;
1511 boolean already_wrote_backup_file = FALSE;
1512 slist* converted_file_ptr;
1513 static slist* converted_files = NULL;
1515 if (downloaded_file_return == FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED)
1517 /* Just write "orig" over "html". We need to do it this way
1518 because when we're checking to see if we've downloaded the
1519 file before (to see if we can skip downloading it), we don't
1520 know if it's a text/html file. Therefore we don't know yet
1521 at that stage that -E is going to cause us to tack on
1522 ".html", so we need to compare vs. the original URL plus
1523 ".orig", not the original URL plus ".html.orig". */
1524 filename_plus_orig_suffix = alloca (filename_len + 1);
1525 strcpy(filename_plus_orig_suffix, file);
1526 strcpy((filename_plus_orig_suffix + filename_len) - 4, "orig");
1528 else /* downloaded_file_return == FILE_DOWNLOADED_NORMALLY */
1530 /* Append ".orig" to the name. */
1531 filename_plus_orig_suffix = alloca (filename_len + sizeof(".orig"));
1532 strcpy(filename_plus_orig_suffix, file);
1533 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
1536 /* We can get called twice on the same URL thanks to the
1537 convert_all_links() call in main(). If we write the .orig file
1538 each time in such a case, it'll end up containing the first-pass
1539 conversion, not the original file. So, see if we've already been
1540 called on this file. */
1541 converted_file_ptr = converted_files;
1542 while (converted_file_ptr != NULL)
1543 if (strcmp(converted_file_ptr->string, file) == 0)
1545 already_wrote_backup_file = TRUE;
1549 converted_file_ptr = converted_file_ptr->next;
1551 if (!already_wrote_backup_file)
1553 /* Rename <file> to <file>.orig before former gets written over. */
1554 if (rename(file, filename_plus_orig_suffix) != 0)
1555 logprintf (LOG_NOTQUIET, _("Cannot back up %s as %s: %s\n"),
1556 file, filename_plus_orig_suffix, strerror (errno));
1558 /* Remember that we've already written a .orig backup for this file.
1559 Note that we never free this memory since we need it till the
1560 convert_all_links() call, which is one of the last things the
1561 program does before terminating. BTW, I'm not sure if it would be
1562 safe to just set 'converted_file_ptr->string' to 'file' below,
1563 rather than making a copy of the string... Another note is that I
1564 thought I could just add a field to the urlpos structure saying
1565 that we'd written a .orig file for this URL, but that didn't work,
1566 so I had to make this separate list.
1567 -- Dan Harkless <wget@harkless.org>
1569 This [adding a field to the urlpos structure] didn't work
1570 because convert_file() is called twice: once after all its
1571 sublinks have been retrieved in recursive_retrieve(), and
1572 once at the end of the day in convert_all_links(). The
1573 original linked list collected in recursive_retrieve() is
1574 lost after the first invocation of convert_links(), and
1575 convert_all_links() makes a new one (it calls get_urls_html()
1576 for each file it covers.) That's why your first approach didn't
1577 work. The way to make it work is perhaps to make this flag a
1578 field in the `urls_html' list.
1579 -- Hrvoje Niksic <hniksic@arsdigita.com>
1581 converted_file_ptr = xmalloc(sizeof(*converted_file_ptr));
1582 converted_file_ptr->string = xstrdup(file); /* die on out-of-mem. */
1583 converted_file_ptr->next = converted_files;
1584 converted_files = converted_file_ptr;
1588 static int find_fragment PARAMS ((const char *, int, const char **,
1592 replace_attr (const char **pp, int raw_size, FILE *fp, const char *new_str)
1594 const char *p = *pp;
1596 int size = raw_size;
1597 char quote_char = '\"';
1598 const char *frag_beg, *frag_end;
1600 /* Structure of our string is:
1601 "...old-contents..."
1602 <--- l->size ---> (with quotes)
1605 <--- l->size --> (no quotes) */
1607 if (*p == '\"' || *p == '\'')
1612 size -= 2; /* disregard opening and closing quote */
1614 putc (quote_char, fp);
1615 fputs (new_str, fp);
1617 /* Look for fragment identifier, if any. */
1618 if (find_fragment (p, size, &frag_beg, &frag_end))
1619 fwrite (frag_beg, 1, frag_end - frag_beg, fp);
1623 putc (quote_char, fp);
1627 /* Find the first occurrence of '#' in [BEG, BEG+SIZE) that is not
1628 preceded by '&'. If the character is not found, return zero. If
1629 the character is found, return 1 and set BP and EP to point to the
1630 beginning and end of the region.
1632 This is used for finding the fragment indentifiers in URLs. */
1635 find_fragment (const char *beg, int size, const char **bp, const char **ep)
1637 const char *end = beg + size;
1639 for (; beg < end; beg++)
1661 typedef struct _downloaded_file_list {
1663 downloaded_file_t download_type;
1664 struct _downloaded_file_list* next;
1665 } downloaded_file_list;
1667 static downloaded_file_list *downloaded_files;
1669 /* Remembers which files have been downloaded. In the standard case, should be
1670 called with mode == FILE_DOWNLOADED_NORMALLY for each file we actually
1671 download successfully (i.e. not for ones we have failures on or that we skip
1674 When we've downloaded a file and tacked on a ".html" extension due to -E,
1675 call this function with FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED rather than
1676 FILE_DOWNLOADED_NORMALLY.
1678 If you just want to check if a file has been previously added without adding
1679 it, call with mode == CHECK_FOR_FILE. Please be sure to call this function
1680 with local filenames, not remote URLs. */
1682 downloaded_file (downloaded_file_t mode, const char* file)
1684 boolean found_file = FALSE;
1685 downloaded_file_list* rover = downloaded_files;
1687 while (rover != NULL)
1688 if (strcmp(rover->file, file) == 0)
1694 rover = rover->next;
1697 return rover->download_type; /* file had already been downloaded */
1700 if (mode != CHECK_FOR_FILE)
1702 rover = xmalloc(sizeof(*rover));
1703 rover->file = xstrdup(file); /* use xstrdup() so die on out-of-mem. */
1704 rover->download_type = mode;
1705 rover->next = downloaded_files;
1706 downloaded_files = rover;
1709 return FILE_NOT_ALREADY_DOWNLOADED;
1714 downloaded_files_free (void)
1716 downloaded_file_list* rover = downloaded_files;
1719 downloaded_file_list *next = rover->next;
1720 xfree (rover->file);