2 Copyright (C) 1995, 1996, 1997 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
9 (at 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>
47 /* Default port definitions */
48 #define DEFAULT_HTTP_PORT 80
49 #define DEFAULT_FTP_PORT 21
51 /* URL separator (for findurl) */
52 #define URL_SEPARATOR "!\"#'(),>`{}|<>"
54 /* A list of unsafe characters for encoding, as per RFC1738. '@' and
55 ':' (not listed in RFC) were added because of user/password
59 # define URL_UNSAFE_CHARS "<>\"#%{}|\\^~[]`@:"
61 # define URL_UNSAFE_CHARS "<>\"%{}|\\^[]`"
64 #define UNSAFE_CHAR(c) ( ((unsigned char)(c) <= ' ') /* ASCII 32 */ \
65 || ((unsigned char)(c) > '~') /* ASCII 127 */ \
66 || strchr (URL_UNSAFE_CHARS, c))
68 /* If S contains unsafe characters, free it and replace it with a
69 version that doesn't. */
70 #define URL_CLEANSE(s) do \
72 if (contains_unsafe (s)) \
74 char *uc_tmp = encode_string (s); \
80 /* Is a directory "."? */
81 #define DOTP(x) ((*(x) == '.') && (!*(x + 1)))
82 /* Is a directory ".."? */
83 #define DDOTP(x) ((*(x) == '.') && (*(x + 1) == '.') && (!*(x + 2)))
85 /* NULL-terminated list of strings to be recognized as prototypes (URL
86 schemes). Note that recognized doesn't mean supported -- only HTTP
87 and FTP are currently supported.
89 However, a string that does not match anything in the list will be
90 considered a relative URL. Thus it's important that this list has
91 anything anyone could think of being legal.
93 There are wild things here. :-) Take a look at
94 <URL:http://www.w3.org/pub/WWW/Addressing/schemes.html> for more
96 static char *protostrings[] =
138 /* Similar to former, but for supported protocols: */
139 static struct proto sup_protos[] =
141 { "http://", URLHTTP, DEFAULT_HTTP_PORT },
142 { "ftp://", URLFTP, DEFAULT_FTP_PORT },
143 /*{ "file://", URLFILE, DEFAULT_FTP_PORT },*/
146 static void parse_dir PARAMS ((const char *, char **, char **));
147 static uerr_t parse_uname PARAMS ((const char *, char **, char **));
148 static char *construct PARAMS ((const char *, const char *, int , int));
149 static char *construct_relative PARAMS ((const char *, const char *));
150 static char process_ftp_type PARAMS ((char *));
153 /* Returns the number of characters to be skipped if the first thing
154 in a URL is URL: (which is 0 or 4+). The optional spaces after
155 URL: are also skipped. */
157 skip_url (const char *url)
161 if (TOUPPER (url[0]) == 'U'
162 && TOUPPER (url[1]) == 'R'
163 && TOUPPER (url[2]) == 'L'
167 for (i = 4; url[i] && ISSPACE (url[i]); i++);
174 /* Returns 1 if the string contains unsafe characters, 0 otherwise. */
176 contains_unsafe (const char *s)
179 if (UNSAFE_CHAR (*s))
184 /* Decodes the forms %xy in a URL to the character the hexadecimal
185 code of which is xy. xy are hexadecimal digits from
186 [0123456789ABCDEF] (case-insensitive). If x or y are not
187 hex-digits or `%' precedes `\0', the sequence is inserted
191 decode_string (char *s)
201 /* Do nothing if at the end of the string, or if the chars
202 are not hex-digits. */
203 if (!*(s + 1) || !*(s + 2)
204 || !(ISXDIGIT (*(s + 1)) && ISXDIGIT (*(s + 2))))
209 *p = (ASC2HEXD (*(s + 1)) << 4) + ASC2HEXD (*(s + 2));
216 /* Encode the unsafe characters (as determined by URL_UNSAFE) in a
217 given string, returning a malloc-ed %XX encoded string. */
219 encode_string (const char *s)
226 for (i = 0; *s; s++, i++)
227 if (UNSAFE_CHAR (*s))
228 i += 2; /* Two more characters (hex digits) */
229 res = (char *)xmalloc (i + 1);
231 for (p = res; *s; s++)
232 if (UNSAFE_CHAR (*s))
234 const unsigned char c = *s;
236 *p++ = HEXD2ASC (c >> 4);
237 *p++ = HEXD2ASC (c & 0xf);
245 /* Returns the proto-type if URL's protocol is supported, or
246 URLUNKNOWN if not. */
248 urlproto (const char *url)
252 url += skip_url (url);
253 for (i = 0; i < ARRAY_SIZE (sup_protos); i++)
254 if (!strncasecmp (url, sup_protos[i].name, strlen (sup_protos[i].name)))
255 return sup_protos[i].ind;
256 for (i = 0; url[i] && url[i] != ':' && url[i] != '/'; i++);
259 for (++i; url[i] && url[i] != '/'; i++)
260 if (!ISDIGIT (url[i]))
262 if (url[i - 1] == ':')
271 /* Skip the protocol part of the URL, e.g. `http://'. If no protocol
272 part is found, returns 0. */
274 skip_proto (const char *url)
279 for (s = protostrings; *s; s++)
280 if (!strncasecmp (*s, url, strlen (*s)))
285 /* HTTP and FTP protocols are expected to yield exact host names
286 (i.e. the `//' part must be skipped, too). */
287 if (!strcmp (*s, "http:") || !strcmp (*s, "ftp:"))
292 /* Returns 1 if the URL begins with a protocol (supported or
293 unsupported), 0 otherwise. */
295 has_proto (const char *url)
299 url += skip_url (url);
300 for (s = protostrings; *s; s++)
301 if (strncasecmp (url, *s, strlen (*s)) == 0)
306 /* Skip the username and password, if present here. The function
307 should be called *not* with the complete URL, but with the part
308 right after the protocol.
310 If no username and password are found, return 0. */
312 skip_uname (const char *url)
315 for (p = url; *p && *p != '/'; p++)
318 /* If a `@' was found before the first occurrence of `/', skip
326 /* Allocate a new urlinfo structure, fill it with default values and
327 return a pointer to it. */
333 u = (struct urlinfo *)xmalloc (sizeof (struct urlinfo));
334 memset (u, 0, sizeof (*u));
335 u->proto = URLUNKNOWN;
339 /* Perform a "deep" free of the urlinfo structure. The structure
340 should have been created with newurl, but need not have been used.
341 If free_pointer is non-0, free the pointer itself. */
343 freeurl (struct urlinfo *u, int complete)
347 FREE_MAYBE (u->host);
348 FREE_MAYBE (u->path);
349 FREE_MAYBE (u->file);
351 FREE_MAYBE (u->user);
352 FREE_MAYBE (u->passwd);
353 FREE_MAYBE (u->local);
354 FREE_MAYBE (u->referer);
356 freeurl (u->proxy, 1);
362 /* Extract the given URL of the form
363 (http:|ftp:)// (user (:password)?@)?hostname (:port)? (/path)?
364 1. hostname (terminated with `/' or `:')
365 2. port number (terminated with `/'), or chosen for the protocol
366 3. dirname (everything after hostname)
367 Most errors are handled. No allocation is done, you must supply
368 pointers to allocated memory.
369 ...and a host of other stuff :-)
371 - Recognizes hostname:dir/file for FTP and
372 hostname (:portnum)?/dir/file for HTTP.
373 - Parses the path to yield directory and file
374 - Parses the URL to yield the username and passwd (if present)
375 - Decodes the strings, in case they contain "forbidden" characters
376 - Writes the result to struct urlinfo
378 If the argument STRICT is set, it recognizes only the canonical
381 parseurl (const char *url, struct urlinfo *u, int strict)
384 int recognizable; /* Recognizable URL is the one where
385 the protocol name was explicitly
386 named, i.e. it wasn't deduced from
390 DEBUGP (("parseurl (\"%s\") -> ", url));
391 url += skip_url (url);
392 recognizable = has_proto (url);
393 if (strict && !recognizable)
395 for (i = 0, l = 0; i < ARRAY_SIZE (sup_protos); i++)
397 l = strlen (sup_protos[i].name);
398 if (!strncasecmp (sup_protos[i].name, url, l))
401 /* If protocol is recognizable, but unsupported, bail out, else
403 if (recognizable && !sup_protos[i].name)
405 else if (i == ARRAY_SIZE (sup_protos))
408 u->proto = type = sup_protos[i].ind;
410 if (type == URLUNKNOWN)
412 /* Allow a username and password to be specified (i.e. just skip
415 l += skip_uname (url + l);
416 for (i = l; url[i] && url[i] != ':' && url[i] != '/'; i++);
419 /* Get the hostname. */
420 u->host = strdupdelim (url + l, url + i);
421 DEBUGP (("host %s -> ", u->host));
423 /* Assume no port has been given. */
427 /* We have a colon delimiting the hostname. It could mean that
428 a port number is following it, or a directory. */
429 if (ISDIGIT (url[++i])) /* A port number */
431 if (type == URLUNKNOWN)
432 u->proto = type = URLHTTP;
433 for (; url[i] && url[i] != '/'; i++)
434 if (ISDIGIT (url[i]))
435 u->port = 10 * u->port + (url[i] - '0');
440 DEBUGP (("port %hu -> ", u->port));
442 else if (type == URLUNKNOWN) /* or a directory */
443 u->proto = type = URLFTP;
444 else /* or just a misformed port number */
447 else if (type == URLUNKNOWN)
448 u->proto = type = URLHTTP;
452 for (i = 0; i < ARRAY_SIZE (sup_protos); i++)
453 if (sup_protos[i].ind == type)
455 if (i == ARRAY_SIZE (sup_protos))
457 u->port = sup_protos[i].port;
459 /* Some delimiter troubles... */
460 if (url[i] == '/' && url[i - 1] != ':')
463 while (url[i] && url[i] == '/')
465 u->path = (char *)xmalloc (strlen (url + i) + 8);
466 strcpy (u->path, url + i);
469 u->ftp_type = process_ftp_type (u->path);
470 /* #### We don't handle type `d' correctly yet. */
471 if (!u->ftp_type || TOUPPER (u->ftp_type) == 'D')
474 DEBUGP (("opath %s -> ", u->path));
475 /* Parse the username and password (if existing). */
476 parse_uname (url, &u->user, &u->passwd);
477 /* Decode the strings, as per RFC 1738. */
478 decode_string (u->host);
479 decode_string (u->path);
481 decode_string (u->user);
483 decode_string (u->passwd);
484 /* Parse the directory. */
485 parse_dir (u->path, &u->dir, &u->file);
486 DEBUGP (("dir %s -> file %s -> ", u->dir, u->file));
487 /* Simplify the directory. */
488 path_simplify (u->dir);
489 /* Remove the leading `/' in HTTP. */
490 if (type == URLHTTP && *u->dir == '/')
491 strcpy (u->dir, u->dir + 1);
492 DEBUGP (("ndir %s\n", u->dir));
493 /* Strip trailing `/'. */
495 if (l && u->dir[l - 1] == '/')
496 u->dir[l - 1] = '\0';
497 /* Re-create the path: */
498 abs_ftp = (u->proto == URLFTP && *u->dir == '/');
499 /* sprintf (u->path, "%s%s%s%s", abs_ftp ? "%2F": "/",
500 abs_ftp ? (u->dir + 1) : u->dir, *u->dir ? "/" : "", u->file); */
501 strcpy (u->path, abs_ftp ? "%2F" : "/");
502 strcat (u->path, abs_ftp ? (u->dir + 1) : u->dir);
503 strcat (u->path, *u->dir ? "/" : "");
504 strcat (u->path, u->file);
505 URL_CLEANSE (u->path);
506 /* Create the clean URL. */
507 u->url = str_url (u, 0);
511 /* Build the directory and filename components of the path. Both
512 components are *separately* malloc-ed strings! It does not change
513 the contents of path.
515 If the path ends with "." or "..", they are (correctly) counted as
518 parse_dir (const char *path, char **dir, char **file)
522 for (i = l = strlen (path); i && path[i] != '/'; i--);
523 if (!i && *path != '/') /* Just filename */
525 if (DOTP (path) || DDOTP (path))
527 *dir = xstrdup (path);
528 *file = xstrdup ("");
532 *dir = xstrdup (""); /* This is required because of FTP */
533 *file = xstrdup (path);
536 else if (!i) /* /filename */
538 if (DOTP (path + 1) || DDOTP (path + 1))
540 *dir = xstrdup (path);
541 *file = xstrdup ("");
545 *dir = xstrdup ("/");
546 *file = xstrdup (path + 1);
549 else /* Nonempty directory with or without a filename */
551 if (DOTP (path + i + 1) || DDOTP (path + i + 1))
553 *dir = xstrdup (path);
554 *file = xstrdup ("");
558 *dir = strdupdelim (path, path + i);
559 *file = strdupdelim (path + i + 1, path + l + 1);
564 /* Find the optional username and password within the URL, as per
565 RFC1738. The returned user and passwd char pointers are
568 parse_uname (const char *url, char **user, char **passwd)
576 url += skip_url (url);
577 /* Look for end of protocol string. */
578 l = skip_proto (url);
581 /* Add protocol offset. */
583 /* Is there an `@' character? */
584 for (p = url; *p && *p != '/'; p++)
587 /* If not, return. */
590 /* Else find the username and password. */
591 for (p = col = url; *p != '@'; p++)
593 if (*p == ':' && !*user)
595 *user = (char *)xmalloc (p - url + 1);
596 memcpy (*user, url, p - url);
597 (*user)[p - url] = '\0';
601 /* Decide whether you have only the username or both. */
602 where = *user ? passwd : user;
603 *where = (char *)xmalloc (p - col + 1);
604 memcpy (*where, col, p - col);
605 (*where)[p - col] = '\0';
609 /* If PATH ends with `;type=X', return the character X. */
611 process_ftp_type (char *path)
613 int len = strlen (path);
616 && !memcmp (path + len - 7, ";type=", 6))
618 path[len - 7] = '\0';
619 return path[len - 1];
625 /* Return the URL as fine-formed string, with a proper protocol, port
626 number, directory and optional user/password. If HIDE is non-zero,
627 password will be hidden. The forbidden characters in the URL will
630 str_url (const struct urlinfo *u, int hide)
632 char *res, *host, *user, *passwd, *proto_name, *dir, *file;
633 int i, l, ln, lu, lh, lp, lf, ld;
634 unsigned short proto_default_port;
636 /* Look for the protocol name. */
637 for (i = 0; i < ARRAY_SIZE (sup_protos); i++)
638 if (sup_protos[i].ind == u->proto)
640 if (i == ARRAY_SIZE (sup_protos))
642 proto_name = sup_protos[i].name;
643 proto_default_port = sup_protos[i].port;
644 host = CLEANDUP (u->host);
645 dir = CLEANDUP (u->dir);
646 file = CLEANDUP (u->file);
647 user = passwd = NULL;
649 user = CLEANDUP (u->user);
653 passwd = CLEANDUP (u->passwd);
655 for (i = 0; passwd[i]; i++)
658 if (u->proto == URLFTP && *dir == '/')
660 char *tmp = (char *)xmalloc (strlen (dir) + 3);
661 /*sprintf (tmp, "%%2F%s", dir + 1);*/
665 strcpy (tmp + 3, dir + 1);
670 ln = strlen (proto_name);
671 lu = user ? strlen (user) : 0;
672 lp = passwd ? strlen (passwd) : 0;
676 res = (char *)xmalloc (ln + lu + lp + lh + ld + lf + 20); /* safe sex */
677 /* sprintf (res, "%s%s%s%s%s%s:%d/%s%s%s", proto_name,
678 (user ? user : ""), (passwd ? ":" : ""),
679 (passwd ? passwd : ""), (user ? "@" : ""),
680 host, u->port, dir, *dir ? "/" : "", file); */
682 memcpy (res, proto_name, ln);
686 memcpy (res + l, user, lu);
691 memcpy (res + l, passwd, lp);
696 memcpy (res + l, host, lh);
698 if (u->port != proto_default_port)
701 long_to_string (res + l, (long)u->port);
702 l += numdigit (u->port);
705 memcpy (res + l, dir, ld);
709 strcpy (res + l, file);
718 /* Check whether two URL-s are equivalent, i.e. pointing to the same
719 location. Uses parseurl to parse them, and compares the canonical
722 Returns 1 if the URL1 is equivalent to URL2, 0 otherwise. Also
723 return 0 on error. */
725 url_equal (const char *url1, const char *url2)
727 struct urlinfo *u1, *u2;
732 err = parseurl (url1, u1, 0);
739 err = parseurl (url2, u2, 0);
745 res = !strcmp (u1->url, u2->url);
751 /* Find URL of format scheme:hostname[:port]/dir in a buffer. The
752 buffer may contain pretty much anything; no errors are signaled. */
754 findurl (const char *buf, int howmuch, int *count)
759 for (s1 = buf; howmuch; s1++, howmuch--)
760 for (prot = protostrings; *prot; prot++)
761 if (howmuch <= strlen (*prot))
763 else if (!strncasecmp (*prot, s1, strlen (*prot)))
765 for (s2 = s1, *count = 0;
766 howmuch && *s2 && *s2 >= 32 && *s2 < 127 && !ISSPACE (*s2) &&
767 !strchr (URL_SEPARATOR, *s2);
768 s2++, (*count)++, howmuch--);
774 /* Scans the file for signs of URL-s. Returns a vector of pointers,
775 each pointer representing a URL string. The file is *not* assumed
778 get_urls_file (const char *file)
785 urlpos *first, *current, *old;
787 if (file && !HYPHENP (file))
789 fp = fopen (file, "rb");
792 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
799 load_file (fp, &buf, &nread);
800 if (file && !HYPHENP (file))
802 DEBUGP (("Loaded %s (size %ld).\n", file, nread));
803 first = current = NULL;
804 /* Fill the linked list with URLs. */
805 for (pbuf = buf; (pbuf = findurl (pbuf, nread - (pbuf - buf), &size));
808 /* Allocate the space. */
810 current = (urlpos *)xmalloc (sizeof (urlpos));
813 memset (current, 0, sizeof (*current));
814 current->next = NULL;
815 current->url = (char *)xmalloc (size + 1);
816 memcpy (current->url, pbuf, size);
817 current->url[size] = '\0';
821 /* Free the buffer. */
827 /* Similar to get_urls_file, but for HTML files. FILE is scanned as
828 an HTML document using htmlfindurl(), which see. get_urls_html()
829 constructs the HTML-s from the relative href-s.
831 If SILENT is non-zero, do not barf on baseless relative links. */
833 get_urls_html (const char *file, const char *this_url, int silent,
834 int dash_p_leaf_HTML)
840 int step, first_time;
841 urlpos *first, *current, *old;
843 if (file && !HYPHENP (file))
845 fp = fopen (file, "rb");
848 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
855 load_file (fp, &orig_buf, &nread);
856 if (file && !HYPHENP (file))
858 DEBUGP (("Loaded HTML file %s (size %ld).\n", file, nread));
859 first = current = NULL;
861 /* Iterate over the URLs in BUF, picked by htmlfindurl(). */
863 (buf = htmlfindurl (buf, nread - (buf - orig_buf), &step, first_time,
869 const char *pbuf = buf;
875 /* A frequent phenomenon that needs to be handled are pages
876 generated by brain-damaged HTML generators, which refer to to
877 URI-s as <a href="<spaces>URI<spaces>">. We simply ignore
878 any spaces at the beginning or at the end of the string.
879 This is probably not strictly correct, but that's what the
880 browsers do, so we may follow. May the authors of "WYSIWYG"
881 HTML tools burn in hell for the damage they've inflicted! */
882 while ((pbuf < buf + step) && ISSPACE (*pbuf))
887 while (size && ISSPACE (pbuf[size - 1]))
892 for (i = 0; protostrings[i]; i++)
894 if (!strncasecmp (protostrings[i], pbuf,
895 MINVAL (strlen (protostrings[i]), size)))
898 /* Check for http:RELATIVE_URI. See below for details. */
900 && !(strncasecmp (pbuf, "http:", 5) == 0
901 && strncasecmp (pbuf, "http://", 7) != 0))
908 /* This is for extremely brain-damaged pages that refer to
909 relative URI-s as <a href="http:URL">. Just strip off the
910 silly leading "http:" (as well as any leading blanks
912 if ((size > 5) && !strncasecmp ("http:", pbuf, 5))
913 pbuf += 5, size -= 5;
917 for (i = 0; i < ARRAY_SIZE (sup_protos); i++)
919 if (!strncasecmp (sup_protos[i].name, pbuf,
920 MINVAL (strlen (sup_protos[i].name), size)))
923 /* Do *not* accept a non-supported protocol. */
924 if (i == ARRAY_SIZE (sup_protos))
929 /* First, construct the base, which can be relative itself.
931 Criteria for creating the base are:
932 1) html_base created by <base href="...">
934 3) base provided from the command line */
935 cbase = html_base ();
939 cbase = opt.base_href;
940 if (!cbase) /* Error condition -- a baseless
943 if (!opt.quiet && !silent)
945 /* Use malloc, not alloca because this is called in
947 char *temp = (char *)malloc (size + 1);
948 strncpy (temp, pbuf, size);
950 logprintf (LOG_NOTQUIET,
951 _("Error (%s): Link %s without a base provided.\n"),
958 base = construct (this_url, cbase, strlen (cbase),
962 /* Base must now be absolute, with host name and
964 if (!has_proto (cbase))
966 logprintf (LOG_NOTQUIET, _("\
967 Error (%s): Base %s relative, without referer URL.\n"),
971 base = xstrdup (cbase);
973 constr = construct (base, pbuf, size, no_proto);
978 constr = (char *)xmalloc (size + 1);
979 strncpy (constr, pbuf, size);
989 /* Use malloc, not alloca because this is called in a loop. */
990 tmp = (char *)xmalloc (size + 1);
991 strncpy (tmp, pbuf, size);
993 logprintf (LOG_ALWAYS,
994 "file %s; this_url %s; base %s\nlink: %s; constr: %s\n",
995 file, this_url ? this_url : "(null)",
996 tmp2 ? tmp2 : "(null)", tmp, constr);
1001 /* Allocate the space. */
1003 current = (urlpos *)xmalloc (sizeof (urlpos));
1005 old->next = current;
1008 /* Fill the values. */
1009 memset (current, 0, sizeof (*current));
1010 current->next = NULL;
1011 current->url = constr;
1012 current->size = size;
1013 current->pos = pbuf - orig_buf;
1014 /* A URL is relative if the host and protocol are not named,
1015 and the name does not start with `/'. */
1016 if (no_proto && *pbuf != '/')
1017 current->flags |= (URELATIVE | UNOPROTO);
1019 current->flags |= UNOPROTO;
1026 /* Free the linked list of urlpos. */
1028 free_urlpos (urlpos *l)
1032 urlpos *next = l->next;
1034 FREE_MAYBE (l->local_name);
1040 /* Rotate FNAME opt.backups times */
1042 rotate_backups(const char *fname)
1044 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
1045 char *from = (char *)alloca (maxlen);
1046 char *to = (char *)alloca (maxlen);
1050 if (stat (fname, &sb) == 0)
1051 if (S_ISREG (sb.st_mode) == 0)
1054 for (i = opt.backups; i > 1; i--)
1056 sprintf (from, "%s.%d", fname, i - 1);
1057 sprintf (to, "%s.%d", fname, i);
1058 /* #### This will fail on machines without the rename() system
1063 sprintf (to, "%s.%d", fname, 1);
1067 /* Create all the necessary directories for PATH (a file). Calls
1068 mkdirhier() internally. */
1070 mkalldirs (const char *path)
1077 p = path + strlen (path);
1078 for (; *p != '/' && p != path; p--);
1079 /* Don't create if it's just a file. */
1080 if ((p == path) && (*p != '/'))
1082 t = strdupdelim (path, p);
1083 /* Check whether the directory exists. */
1084 if ((stat (t, &st) == 0))
1086 if (S_ISDIR (st.st_mode))
1093 /* If the dir exists as a file name, remove it first. This
1094 is *only* for Wget to work with buggy old CERN http
1095 servers. Here is the scenario: When Wget tries to
1096 retrieve a directory without a slash, e.g.
1097 http://foo/bar (bar being a directory), CERN server will
1098 not redirect it too http://foo/bar/ -- it will generate a
1099 directory listing containing links to bar/file1,
1100 bar/file2, etc. Wget will lose because it saves this
1101 HTML listing to a file `bar', so it cannot create the
1102 directory. To work around this, if the file of the same
1103 name exists, we just remove it and create the directory
1105 DEBUGP (("Removing %s because of directory danger!\n", t));
1109 res = make_directory (t);
1111 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1117 count_slashes (const char *s)
1126 /* Return the path name of the URL-equivalent file name, with a
1127 remote-like structure of directories. */
1129 mkstruct (const struct urlinfo *u)
1131 char *host, *dir, *file, *res, *dirpref;
1134 assert (u->dir != NULL);
1135 assert (u->host != NULL);
1139 char *ptr = u->dir + (*u->dir == '/');
1140 int slash_count = 1 + count_slashes (ptr);
1141 int cut = MINVAL (opt.cut_dirs, slash_count);
1142 for (; cut && *ptr; ptr++)
1145 STRDUP_ALLOCA (dir, ptr);
1148 dir = u->dir + (*u->dir == '/');
1150 host = xstrdup (u->host);
1151 /* Check for the true name (or at least a consistent name for saving
1152 to directory) of HOST, reusing the hlist if possible. */
1153 if (opt.add_hostdir && !opt.simple_check)
1155 char *nhost = realhost (host);
1159 /* Add dir_prefix and hostname (if required) to the beginning of
1161 if (opt.add_hostdir)
1163 if (!DOTP (opt.dir_prefix))
1165 dirpref = (char *)alloca (strlen (opt.dir_prefix) + 1
1166 + strlen (host) + 1);
1167 sprintf (dirpref, "%s/%s", opt.dir_prefix, host);
1170 STRDUP_ALLOCA (dirpref, host);
1172 else /* not add_hostdir */
1174 if (!DOTP (opt.dir_prefix))
1175 dirpref = opt.dir_prefix;
1181 /* If there is a prefix, prepend it. */
1184 char *newdir = (char *)alloca (strlen (dirpref) + 1 + strlen (dir) + 2);
1185 sprintf (newdir, "%s%s%s", dirpref, *dir == '/' ? "" : "/", dir);
1188 dir = xstrdup (dir);
1191 if (l && dir[l - 1] == '/')
1195 file = "index.html";
1199 /* Finally, construct the full name. */
1200 res = (char *)xmalloc (strlen (dir) + 1 + strlen (file) + 1);
1201 sprintf (res, "%s%s%s", dir, *dir ? "/" : "", file);
1206 /* Create a unique filename, corresponding to a given URL. Calls
1207 mkstruct if necessary. Does *not* actually create any directories. */
1209 url_filename (const struct urlinfo *u)
1212 int have_prefix = 0; /* whether we must prepend opt.dir_prefix */
1216 file = mkstruct (u);
1222 file = xstrdup ("index.html");
1224 file = xstrdup (u->file);
1229 /* Check whether the prefix directory is something other than "."
1230 before prepending it. */
1231 if (!DOTP (opt.dir_prefix))
1233 char *nfile = (char *)xmalloc (strlen (opt.dir_prefix)
1234 + 1 + strlen (file) + 1);
1235 sprintf (nfile, "%s/%s", opt.dir_prefix, file);
1240 /* DOS-ish file systems don't like `%' signs in them; we change it
1245 for (p = file; *p; p++)
1249 #endif /* WINDOWS */
1251 /* Check the cases in which the unique extensions are not used:
1252 1) Clobbering is turned off (-nc).
1253 2) Retrieval with regetting.
1254 3) Timestamping is used.
1255 4) Hierarchy is built.
1257 The exception is the case when file does exist and is a
1258 directory (actually support for bad httpd-s). */
1259 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1260 && !(file_exists_p (file) && !file_non_directory_p (file)))
1263 /* Find a unique name. */
1264 name = unique_name (file);
1269 /* Like strlen(), except if `?' is present in the URL and its protocol
1270 is HTTP, act as if `?' is the end of the string. Needed for the
1271 correct implementation of `construct' below, at least until we code
1272 up proper parsing of URLs. */
1274 urllen_http_hack (const char *url)
1276 if ((!strncmp (url, "http://", 7)
1277 || !strncmp (url, "https://", 7)))
1279 const char *q = strchr (url, '?');
1283 return strlen (url);
1286 /* Construct an absolute URL, given a (possibly) relative one. This
1287 is more tricky than it might seem, but it works. */
1289 construct (const char *url, const char *sub, int subsize, int no_proto)
1299 for (i = urllen_http_hack (url); i && url[i] != '/'; i--);
1300 if (!i || (url[i] == url[i - 1]))
1302 int l = urllen_http_hack (url);
1303 char *t = (char *)alloca (l + 2);
1310 constr = (char *)xmalloc (i + 1 + subsize + 1);
1311 strncpy (constr, url, i + 1);
1312 constr[i + 1] = '\0';
1313 strncat (constr, sub, subsize);
1315 else /* *sub == `/' */
1322 for (; url[i] && url[i] != '/'; i++);
1325 fl = (url[i] == url[i + 1] && url[i + 1] == '/');
1332 int l = urllen_http_hack (url);
1333 char *t = (char *)alloca (l + 2);
1339 constr = (char *)xmalloc (i + 1 + subsize + 1);
1340 strncpy (constr, url, i);
1342 strncat (constr + i, sub, subsize);
1343 constr[i + subsize] = '\0';
1346 else /* !no_proto */
1348 constr = (char *)xmalloc (subsize + 1);
1349 strncpy (constr, sub, subsize);
1350 constr[subsize] = '\0';
1355 /* Like the function above, but with a saner caller interface. */
1357 url_concat (const char *base_url, const char *new_url)
1359 return construct (base_url, new_url, strlen (new_url), !has_proto (new_url));
1362 /* Optimize URL by host, destructively replacing u->host with realhost
1363 (u->host). Do this regardless of opt.simple_check. */
1365 opt_url (struct urlinfo *u)
1367 /* Find the "true" host. */
1368 char *host = realhost (u->host);
1371 assert (u->dir != NULL); /* the URL must have been parsed */
1372 /* Refresh the printed representation. */
1374 u->url = str_url (u, 0);
1377 /* Returns proxy host address, in accordance with PROTO. */
1379 getproxy (uerr_t proto)
1381 if (proto == URLHTTP)
1382 return opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
1383 else if (proto == URLFTP)
1384 return opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
1389 /* Should a host be accessed through proxy, concerning no_proxy? */
1391 no_proxy_match (const char *host, const char **no_proxy)
1396 return !sufmatch (no_proxy, host);
1399 /* Change the links in an HTML document. Accepts a structure that
1400 defines the positions of all the links. */
1402 convert_links (const char *file, urlpos *l)
1406 downloaded_file_t downloaded_file_return;
1409 logprintf (LOG_VERBOSE, _("Converting %s... "), file);
1410 /* Read from the file.... */
1411 fp = fopen (file, "rb");
1414 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1415 file, strerror (errno));
1418 /* ...to a buffer. */
1419 load_file (fp, &buf, &size);
1422 downloaded_file_return = downloaded_file(CHECK_FOR_FILE, file);
1424 if (opt.backup_converted && downloaded_file_return)
1425 /* Rather than just writing over the original .html file with the converted
1426 version, save the former to *.orig. Note we only do this for files we've
1427 _successfully_ downloaded, so we don't clobber .orig files sitting around
1428 from previous invocations. */
1430 /* Construct the backup filename as the original name plus ".orig". */
1431 size_t filename_len = strlen(file);
1432 char* filename_plus_orig_suffix;
1433 boolean already_wrote_backup_file = FALSE;
1434 slist* converted_file_ptr;
1435 static slist* converted_files = NULL;
1437 if (downloaded_file_return == FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED)
1439 /* Just write "orig" over "html". We need to do it this way because
1440 when we're checking to see if we've downloaded the file before (to
1441 see if we can skip downloading it), we don't know if it's a
1442 text/html file. Therefore we don't know yet at that stage that -E
1443 is going to cause us to tack on ".html", so we need to compare
1444 vs. the original URL plus ".orig", not the original URL plus
1446 filename_plus_orig_suffix = xmalloc(filename_len + 1);
1447 strcpy(filename_plus_orig_suffix, file);
1448 strcpy((filename_plus_orig_suffix + filename_len) - 4, "orig");
1450 else /* downloaded_file_return == FILE_DOWNLOADED_NORMALLY */
1452 /* Append ".orig" to the name. */
1453 filename_plus_orig_suffix = xmalloc(filename_len + sizeof(".orig"));
1454 strcpy(filename_plus_orig_suffix, file);
1455 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
1458 /* We can get called twice on the same URL thanks to the
1459 convert_all_links() call in main(). If we write the .orig file each
1460 time in such a case, it'll end up containing the first-pass conversion,
1461 not the original file. So, see if we've already been called on this
1463 converted_file_ptr = converted_files;
1464 while (converted_file_ptr != NULL)
1465 if (strcmp(converted_file_ptr->string, file) == 0)
1467 already_wrote_backup_file = TRUE;
1471 converted_file_ptr = converted_file_ptr->next;
1473 if (!already_wrote_backup_file)
1475 /* Rename <file> to <file>.orig before former gets written over. */
1476 if (rename(file, filename_plus_orig_suffix) != 0)
1477 logprintf (LOG_NOTQUIET, _("Cannot back up %s as %s: %s\n"),
1478 file, filename_plus_orig_suffix, strerror (errno));
1480 /* Remember that we've already written a .orig backup for this file.
1481 Note that we never free this memory since we need it till the
1482 convert_all_links() call, which is one of the last things the
1483 program does before terminating. BTW, I'm not sure if it would be
1484 safe to just set 'converted_file_ptr->string' to 'file' below,
1485 rather than making a copy of the string... Another note is that I
1486 thought I could just add a field to the urlpos structure saying
1487 that we'd written a .orig file for this URL, but that didn't work,
1488 so I had to make this separate list. */
1489 converted_file_ptr = xmalloc(sizeof(*converted_file_ptr));
1490 converted_file_ptr->string = xstrdup(file); /* die on out-of-mem. */
1491 converted_file_ptr->next = converted_files;
1492 converted_files = converted_file_ptr;
1495 free(filename_plus_orig_suffix);
1497 /* Now open the file for writing. */
1498 fp = fopen (file, "wb");
1501 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1502 file, strerror (errno));
1506 /* Presumably we have to loop through multiple URLs here (even though we're
1507 only talking about a single local file) because of the -O option. */
1508 for (p = buf; l; l = l->next)
1512 DEBUGP (("Something strange is going on. Please investigate."));
1515 /* If the URL already is relative or it is not to be converted
1516 for some other reason (e.g. because of not having been
1517 downloaded in the first place), skip it. */
1518 if ((l->flags & URELATIVE) || !(l->flags & UABS2REL))
1520 DEBUGP (("Skipping %s at position %d (flags %d).\n", l->url,
1524 /* Else, reach the position of the offending URL, echoing
1525 everything up to it to the outfile. */
1526 for (p2 = buf + l->pos; p < p2; p++)
1528 if (l->flags & UABS2REL)
1529 /* Convert absolute URL to relative. */
1531 char *newname = construct_relative (file, l->local_name);
1532 fprintf (fp, "%s", newname);
1533 DEBUGP (("ABS2REL: %s to %s at position %d in %s.\n",
1534 l->url, newname, l->pos, file));
1539 /* Output the rest of the file. */
1542 for (p2 = buf + size; p < p2; p++)
1547 logputs (LOG_VERBOSE, _("done.\n"));
1550 /* Construct and return a malloced copy of the relative link from two
1551 pieces of information: local name S1 of the referring file and
1552 local name S2 of the referred file.
1554 So, if S1 is "jagor.srce.hr/index.html" and S2 is
1555 "jagor.srce.hr/images/news.gif", the function will return
1558 Alternately, if S1 is "fly.cc.fer.hr/ioccc/index.html", and S2 is
1559 "fly.cc.fer.hr/images/fly.gif", the function will return
1560 "../images/fly.gif".
1562 Caveats: S1 should not begin with `/', unless S2 also begins with
1563 '/'. S1 should not contain things like ".." and such --
1564 construct_relative ("fly/ioccc/../index.html",
1565 "fly/images/fly.gif") will fail. (A workaround is to call
1566 something like path_simplify() on S1). */
1568 construct_relative (const char *s1, const char *s2)
1570 int i, cnt, sepdirs1;
1574 return xstrdup (s2);
1575 /* S1 should *not* be absolute, if S2 wasn't. */
1576 assert (*s1 != '/');
1578 /* Skip the directories common to both strings. */
1581 while (s1[i] && s2[i]
1586 if (s1[i] == '/' && s2[i] == '/')
1591 for (sepdirs1 = 0; s1[i]; i++)
1594 /* Now, construct the file as of:
1595 - ../ repeated sepdirs1 time
1596 - all the non-mutual directories of S2. */
1597 res = (char *)xmalloc (3 * sepdirs1 + strlen (s2 + cnt) + 1);
1598 for (i = 0; i < sepdirs1; i++)
1599 memcpy (res + 3 * i, "../", 3);
1600 strcpy (res + 3 * i, s2 + cnt);
1604 /* Add URL to the head of the list L. */
1606 add_url (urlpos *l, const char *url, const char *file)
1610 t = (urlpos *)xmalloc (sizeof (urlpos));
1611 memset (t, 0, sizeof (*t));
1612 t->url = xstrdup (url);
1613 t->local_name = xstrdup (file);
1619 /* Remembers which files have been downloaded. In the standard case, should be
1620 called with mode == FILE_DOWNLOADED_NORMALLY for each file we actually
1621 download successfully (i.e. not for ones we have failures on or that we skip
1624 When we've downloaded a file and tacked on a ".html" extension due to -E,
1625 call this function with FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED rather than
1626 FILE_DOWNLOADED_NORMALLY.
1628 If you just want to check if a file has been previously added without adding
1629 it, call with mode == CHECK_FOR_FILE. Please be sure to call this function
1630 with local filenames, not remote URLs. */
1632 downloaded_file (downloaded_file_t mode, const char* file)
1634 typedef struct _downloaded_file_list
1637 downloaded_file_t download_type;
1638 struct _downloaded_file_list* next;
1639 } downloaded_file_list;
1641 boolean found_file = FALSE;
1642 static downloaded_file_list* downloaded_files = NULL;
1643 downloaded_file_list* rover = downloaded_files;
1645 while (rover != NULL)
1646 if (strcmp(rover->file, file) == 0)
1652 rover = rover->next;
1655 return rover->download_type; /* file had already been downloaded */
1658 if (mode != CHECK_FOR_FILE)
1660 rover = xmalloc(sizeof(*rover));
1661 rover->file = xstrdup(file); /* use xstrdup() so die on out-of-mem. */
1662 rover->download_type = mode;
1663 rover->next = downloaded_files;
1664 downloaded_files = rover;
1667 return FILE_NOT_ALREADY_DOWNLOADED;