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 /* Encodes the unsafe characters (listed in URL_UNSAFE_CHARS) 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)
839 int step, first_time;
840 urlpos *first, *current, *old;
842 if (file && !HYPHENP (file))
844 fp = fopen (file, "rb");
847 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
854 load_file (fp, &orig_buf, &nread);
855 if (file && !HYPHENP (file))
857 DEBUGP (("Loaded HTML file %s (size %ld).\n", file, nread));
858 first = current = NULL;
860 /* Iterate over the URLs in BUF, picked by htmlfindurl(). */
862 (buf = htmlfindurl (buf, nread - (buf - orig_buf), &step, first_time));
867 const char *pbuf = buf;
873 /* A frequent phenomenon that needs to be handled are pages
874 generated by brain-damaged HTML generators, which refer to to
875 URI-s as <a href="<spaces>URI<spaces>">. We simply ignore
876 any spaces at the beginning or at the end of the string.
877 This is probably not strictly correct, but that's what the
878 browsers do, so we may follow. May the authors of "WYSIWYG"
879 HTML tools burn in hell for the damage they've inflicted! */
880 while ((pbuf < buf + step) && ISSPACE (*pbuf))
885 while (size && ISSPACE (pbuf[size - 1]))
890 for (i = 0; protostrings[i]; i++)
892 if (!strncasecmp (protostrings[i], pbuf,
893 MINVAL (strlen (protostrings[i]), size)))
896 /* Check for http:RELATIVE_URI. See below for details. */
898 && !(strncasecmp (pbuf, "http:", 5) == 0
899 && strncasecmp (pbuf, "http://", 7) != 0))
906 /* This is for extremely brain-damaged pages that refer to
907 relative URI-s as <a href="http:URL">. Just strip off the
908 silly leading "http:" (as well as any leading blanks
910 if ((size > 5) && !strncasecmp ("http:", pbuf, 5))
911 pbuf += 5, size -= 5;
915 for (i = 0; i < ARRAY_SIZE (sup_protos); i++)
917 if (!strncasecmp (sup_protos[i].name, pbuf,
918 MINVAL (strlen (sup_protos[i].name), size)))
921 /* Do *not* accept a non-supported protocol. */
922 if (i == ARRAY_SIZE (sup_protos))
927 /* First, construct the base, which can be relative itself.
929 Criteria for creating the base are:
930 1) html_base created by <base href="...">
932 3) base provided from the command line */
933 cbase = html_base ();
937 cbase = opt.base_href;
938 if (!cbase) /* Error condition -- a baseless
941 if (!opt.quiet && !silent)
943 /* Use malloc, not alloca because this is called in
945 char *temp = (char *)malloc (size + 1);
946 strncpy (temp, pbuf, size);
948 logprintf (LOG_NOTQUIET,
949 _("Error (%s): Link %s without a base provided.\n"),
956 base = construct (this_url, cbase, strlen (cbase),
960 /* Base must now be absolute, with host name and
962 if (!has_proto (cbase))
964 logprintf (LOG_NOTQUIET, _("\
965 Error (%s): Base %s relative, without referer URL.\n"),
969 base = xstrdup (cbase);
971 constr = construct (base, pbuf, size, no_proto);
976 constr = (char *)xmalloc (size + 1);
977 strncpy (constr, pbuf, size);
987 /* Use malloc, not alloca because this is called in a loop. */
988 tmp = (char *)xmalloc (size + 1);
989 strncpy (tmp, pbuf, size);
991 logprintf (LOG_ALWAYS,
992 "file %s; this_url %s; base %s\nlink: %s; constr: %s\n",
993 file, this_url ? this_url : "(null)",
994 tmp2 ? tmp2 : "(null)", tmp, constr);
999 /* Allocate the space. */
1001 current = (urlpos *)xmalloc (sizeof (urlpos));
1003 old->next = current;
1006 /* Fill the values. */
1007 memset (current, 0, sizeof (*current));
1008 current->next = NULL;
1009 current->url = constr;
1010 current->size = size;
1011 current->pos = pbuf - orig_buf;
1012 /* A URL is relative if the host and protocol are not named,
1013 and the name does not start with `/'. */
1014 if (no_proto && *pbuf != '/')
1015 current->flags |= (URELATIVE | UNOPROTO);
1017 current->flags |= UNOPROTO;
1024 /* Free the linked list of urlpos. */
1026 free_urlpos (urlpos *l)
1030 urlpos *next = l->next;
1032 FREE_MAYBE (l->local_name);
1038 /* Rotate FNAME opt.backups times */
1040 rotate_backups(const char *fname)
1042 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
1043 char *from = (char *)alloca (maxlen);
1044 char *to = (char *)alloca (maxlen);
1048 if (stat (fname, &sb) == 0)
1049 if (S_ISREG (sb.st_mode) == 0)
1052 for (i = opt.backups; i > 1; i--)
1054 sprintf (from, "%s.%d", fname, i - 1);
1055 sprintf (to, "%s.%d", fname, i);
1056 /* #### This will fail on machines without the rename() system
1061 sprintf (to, "%s.%d", fname, 1);
1065 /* Create all the necessary directories for PATH (a file). Calls
1066 mkdirhier() internally. */
1068 mkalldirs (const char *path)
1075 p = path + strlen (path);
1076 for (; *p != '/' && p != path; p--);
1077 /* Don't create if it's just a file. */
1078 if ((p == path) && (*p != '/'))
1080 t = strdupdelim (path, p);
1081 /* Check whether the directory exists. */
1082 if ((stat (t, &st) == 0))
1084 if (S_ISDIR (st.st_mode))
1091 /* If the dir exists as a file name, remove it first. This
1092 is *only* for Wget to work with buggy old CERN http
1093 servers. Here is the scenario: When Wget tries to
1094 retrieve a directory without a slash, e.g.
1095 http://foo/bar (bar being a directory), CERN server will
1096 not redirect it too http://foo/bar/ -- it will generate a
1097 directory listing containing links to bar/file1,
1098 bar/file2, etc. Wget will lose because it saves this
1099 HTML listing to a file `bar', so it cannot create the
1100 directory. To work around this, if the file of the same
1101 name exists, we just remove it and create the directory
1103 DEBUGP (("Removing %s because of directory danger!\n", t));
1107 res = make_directory (t);
1109 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1115 count_slashes (const char *s)
1124 /* Return the path name of the URL-equivalent file name, with a
1125 remote-like structure of directories. */
1127 mkstruct (const struct urlinfo *u)
1129 char *host, *dir, *file, *res, *dirpref;
1132 assert (u->dir != NULL);
1133 assert (u->host != NULL);
1137 char *ptr = u->dir + (*u->dir == '/');
1138 int slash_count = 1 + count_slashes (ptr);
1139 int cut = MINVAL (opt.cut_dirs, slash_count);
1140 for (; cut && *ptr; ptr++)
1143 STRDUP_ALLOCA (dir, ptr);
1146 dir = u->dir + (*u->dir == '/');
1148 host = xstrdup (u->host);
1149 /* Check for the true name (or at least a consistent name for saving
1150 to directory) of HOST, reusing the hlist if possible. */
1151 if (opt.add_hostdir && !opt.simple_check)
1153 char *nhost = realhost (host);
1157 /* Add dir_prefix and hostname (if required) to the beginning of
1159 if (opt.add_hostdir)
1161 if (!DOTP (opt.dir_prefix))
1163 dirpref = (char *)alloca (strlen (opt.dir_prefix) + 1
1164 + strlen (host) + 1);
1165 sprintf (dirpref, "%s/%s", opt.dir_prefix, host);
1168 STRDUP_ALLOCA (dirpref, host);
1170 else /* not add_hostdir */
1172 if (!DOTP (opt.dir_prefix))
1173 dirpref = opt.dir_prefix;
1179 /* If there is a prefix, prepend it. */
1182 char *newdir = (char *)alloca (strlen (dirpref) + 1 + strlen (dir) + 2);
1183 sprintf (newdir, "%s%s%s", dirpref, *dir == '/' ? "" : "/", dir);
1186 dir = xstrdup (dir);
1189 if (l && dir[l - 1] == '/')
1193 file = "index.html";
1197 /* Finally, construct the full name. */
1198 res = (char *)xmalloc (strlen (dir) + 1 + strlen (file) + 1);
1199 sprintf (res, "%s%s%s", dir, *dir ? "/" : "", file);
1204 /* Create a unique filename, corresponding to a given URL. Calls
1205 mkstruct if necessary. Does *not* actually create any directories. */
1207 url_filename (const struct urlinfo *u)
1210 int have_prefix = 0; /* whether we must prepend opt.dir_prefix */
1214 file = mkstruct (u);
1220 file = xstrdup ("index.html");
1222 file = xstrdup (u->file);
1227 /* Check whether the prefix directory is something other than "."
1228 before prepending it. */
1229 if (!DOTP (opt.dir_prefix))
1231 char *nfile = (char *)xmalloc (strlen (opt.dir_prefix)
1232 + 1 + strlen (file) + 1);
1233 sprintf (nfile, "%s/%s", opt.dir_prefix, file);
1238 /* DOS-ish file systems don't like `%' signs in them; we change it
1243 for (p = file; *p; p++)
1247 #endif /* WINDOWS */
1249 /* Check the cases in which the unique extensions are not used:
1250 1) Clobbering is turned off (-nc).
1251 2) Retrieval with regetting.
1252 3) Timestamping is used.
1253 4) Hierarchy is built.
1255 The exception is the case when file does exist and is a
1256 directory (actually support for bad httpd-s). */
1257 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1258 && !(file_exists_p (file) && !file_non_directory_p (file)))
1261 /* Find a unique name. */
1262 name = unique_name (file);
1267 /* Construct an absolute URL, given a (possibly) relative one. This
1268 is more tricky than it might seem, but it works. */
1270 construct (const char *url, const char *sub, int subsize, int no_proto)
1280 for (i = strlen (url); i && url[i] != '/'; i--);
1281 if (!i || (url[i] == url[i - 1]))
1283 int l = strlen (url);
1284 char *t = (char *)alloca (l + 2);
1291 constr = (char *)xmalloc (i + 1 + subsize + 1);
1292 strncpy (constr, url, i + 1);
1293 constr[i + 1] = '\0';
1294 strncat (constr, sub, subsize);
1296 else /* *sub == `/' */
1303 for (; url[i] && url[i] != '/'; i++);
1306 fl = (url[i] == url[i + 1] && url[i + 1] == '/');
1313 int l = strlen (url);
1314 char *t = (char *)alloca (l + 2);
1320 constr = (char *)xmalloc (i + 1 + subsize + 1);
1321 strncpy (constr, url, i);
1323 strncat (constr + i, sub, subsize);
1324 constr[i + subsize] = '\0';
1327 else /* !no_proto */
1329 constr = (char *)xmalloc (subsize + 1);
1330 strncpy (constr, sub, subsize);
1331 constr[subsize] = '\0';
1336 /* Optimize URL by host, destructively replacing u->host with realhost
1337 (u->host). Do this regardless of opt.simple_check. */
1339 opt_url (struct urlinfo *u)
1341 /* Find the "true" host. */
1342 char *host = realhost (u->host);
1345 assert (u->dir != NULL); /* the URL must have been parsed */
1346 /* Refresh the printed representation. */
1348 u->url = str_url (u, 0);
1351 /* Returns proxy host address, in accordance with PROTO. */
1353 getproxy (uerr_t proto)
1355 if (proto == URLHTTP)
1356 return opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
1357 else if (proto == URLFTP)
1358 return opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
1363 /* Should a host be accessed through proxy, concerning no_proxy? */
1365 no_proxy_match (const char *host, const char **no_proxy)
1370 return !sufmatch (no_proxy, host);
1373 /* Change the links in an HTML document. Accepts a structure that
1374 defines the positions of all the links. */
1376 convert_links (const char *file, urlpos *l)
1382 logprintf (LOG_VERBOSE, _("Converting %s... "), file);
1383 /* Read from the file.... */
1384 fp = fopen (file, "rb");
1387 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1388 file, strerror (errno));
1391 /* ...to a buffer. */
1392 load_file (fp, &buf, &size);
1394 if (opt.backup_converted && downloaded_file(CHECK_FOR_FILE, file))
1395 /* Rather than just writing over the original .html file with the converted
1396 version, save the former to *.orig. Note we only do this for files we've
1397 _successfully_ downloaded, so we don't clobber .orig files sitting around
1398 from previous invocations. */
1400 /* Construct the backup filename as the original name plus ".orig". */
1401 size_t filename_len = strlen(file);
1402 char* filename_plus_orig_suffix = malloc(filename_len +
1404 boolean already_wrote_backup_file = FALSE;
1405 slist* converted_file_ptr;
1406 static slist* converted_files = NULL;
1408 /* Would a single s[n]printf() call be faster? */
1409 strcpy(filename_plus_orig_suffix, file);
1410 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
1412 /* We can get called twice on the same URL thanks to the
1413 convert_all_links() call in main(). If we write the .orig file each
1414 time in such a case, it'll end up containing the first-pass conversion,
1415 not the original file. So, see if we've already been called on this
1417 converted_file_ptr = converted_files;
1418 while (converted_file_ptr != NULL)
1419 if (strcmp(converted_file_ptr->string, file) == 0)
1421 already_wrote_backup_file = TRUE;
1425 converted_file_ptr = converted_file_ptr->next;
1427 if (!already_wrote_backup_file)
1429 /* Rename <file> to <file>.orig before former gets written over. */
1430 if (rename(file, filename_plus_orig_suffix) != 0)
1431 logprintf (LOG_NOTQUIET, _("Cannot back up %s as %s: %s\n"),
1432 file, filename_plus_orig_suffix, strerror (errno));
1434 /* Remember that we've already written a .orig backup for this file.
1435 Note that we never free this memory since we need it till the
1436 convert_all_links() call, which is one of the last things the
1437 program does before terminating. BTW, I'm not sure if it would be
1438 safe to just set 'converted_file_ptr->string' to 'file' below,
1439 rather than making a copy of the string... Another note is that I
1440 thought I could just add a field to the urlpos structure saying
1441 that we'd written a .orig file for this URL, but that didn't work,
1442 so I had to make this separate list. */
1443 converted_file_ptr = malloc(sizeof(slist));
1444 converted_file_ptr->string = xstrdup(file); /* die on out-of-mem. */
1445 converted_file_ptr->next = converted_files;
1446 converted_files = converted_file_ptr;
1449 free(filename_plus_orig_suffix);
1451 /* Now open the file for writing. */
1452 fp = fopen (file, "wb");
1455 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1456 file, strerror (errno));
1460 /* [If someone understands why multiple URLs can correspond to one local file,
1461 can they please add a comment here...?] */
1462 for (p = buf; l; l = l->next)
1466 DEBUGP (("Something strange is going on. Please investigate."));
1469 /* If the URL already is relative or it is not to be converted
1470 for some other reason (e.g. because of not having been
1471 downloaded in the first place), skip it. */
1472 if ((l->flags & URELATIVE) || !(l->flags & UABS2REL))
1474 DEBUGP (("Skipping %s at position %d (flags %d).\n", l->url,
1478 /* Else, reach the position of the offending URL, echoing
1479 everything up to it to the outfile. */
1480 for (p2 = buf + l->pos; p < p2; p++)
1482 if (l->flags & UABS2REL)
1484 char *newname = construct_relative (file, l->local_name);
1485 fprintf (fp, "%s", newname);
1486 DEBUGP (("ABS2REL: %s to %s at position %d in %s.\n",
1487 l->url, newname, l->pos, file));
1494 for (p2 = buf + size; p < p2; p++)
1499 logputs (LOG_VERBOSE, _("done.\n"));
1502 /* Construct and return a malloced copy of the relative link from two
1503 pieces of information: local name S1 of the referring file and
1504 local name S2 of the referred file.
1506 So, if S1 is "jagor.srce.hr/index.html" and S2 is
1507 "jagor.srce.hr/images/news.gif", the function will return
1510 Alternately, if S1 is "fly.cc.fer.hr/ioccc/index.html", and S2 is
1511 "fly.cc.fer.hr/images/fly.gif", the function will return
1512 "../images/fly.gif".
1514 Caveats: S1 should not begin with `/', unless S2 also begins with
1515 '/'. S1 should not contain things like ".." and such --
1516 construct_relative ("fly/ioccc/../index.html",
1517 "fly/images/fly.gif") will fail. (A workaround is to call
1518 something like path_simplify() on S1). */
1520 construct_relative (const char *s1, const char *s2)
1522 int i, cnt, sepdirs1;
1526 return xstrdup (s2);
1527 /* S1 should *not* be absolute, if S2 wasn't. */
1528 assert (*s1 != '/');
1530 /* Skip the directories common to both strings. */
1533 while (s1[i] && s2[i]
1538 if (s1[i] == '/' && s2[i] == '/')
1543 for (sepdirs1 = 0; s1[i]; i++)
1546 /* Now, construct the file as of:
1547 - ../ repeated sepdirs1 time
1548 - all the non-mutual directories of S2. */
1549 res = (char *)xmalloc (3 * sepdirs1 + strlen (s2 + cnt) + 1);
1550 for (i = 0; i < sepdirs1; i++)
1551 memcpy (res + 3 * i, "../", 3);
1552 strcpy (res + 3 * i, s2 + cnt);
1556 /* Add URL to the head of the list L. */
1558 add_url (urlpos *l, const char *url, const char *file)
1562 t = (urlpos *)xmalloc (sizeof (urlpos));
1563 memset (t, 0, sizeof (*t));
1564 t->url = xstrdup (url);
1565 t->local_name = xstrdup (file);
1571 /* Remembers which files have been downloaded. Should be called with
1572 add_or_check == ADD_FILE for each file we actually download successfully
1573 (i.e. not for ones we have failures on or that we skip due to -N). If you
1574 just want to check if a file has been previously added without adding it,
1575 call with add_or_check == CHECK_FOR_FILE. Please be sure to call this
1576 function with local filenames, not remote URLs -- by some means that isn't
1577 commented well enough for me understand, multiple remote URLs can apparently
1578 correspond to a single local file. */
1580 downloaded_file (downloaded_file_t add_or_check, const char* file)
1582 boolean found_file = FALSE;
1583 static slist* downloaded_files = NULL;
1584 slist* rover = downloaded_files;
1586 while (rover != NULL)
1587 if (strcmp(rover->string, file) == 0)
1593 rover = rover->next;
1596 return TRUE; /* file had already been downloaded */
1599 if (add_or_check == ADD_FILE)
1601 rover = malloc(sizeof(slist));
1602 rover->string = xstrdup(file); /* die on out-of-mem. */
1603 rover->next = downloaded_files;
1604 downloaded_files = rover;
1607 return FALSE; /* file had not already been downloaded */