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>
47 #define DOTP(x) ((*(x) == '.') && (!*(x + 1)))
49 #define DDOTP(x) ((*(x) == '.') && (*(x + 1) == '.') && (!*(x + 2)))
58 /* Supported schemes: */
59 static struct scheme_data supported_schemes[] =
61 { "http://", DEFAULT_HTTP_PORT, 1 },
63 { "https://", DEFAULT_HTTPS_PORT, 1 },
65 { "ftp://", DEFAULT_FTP_PORT, 1 },
71 /* Forward declarations: */
73 static char *construct_relative PARAMS ((const char *, const char *));
74 static int path_simplify PARAMS ((char *));
78 /* Support for encoding and decoding of URL strings. We determine
79 whether a character is unsafe through static table lookup. This
80 code assumes ASCII character set and 8-bit chars. */
87 #define R urlchr_reserved
88 #define U urlchr_unsafe
91 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
93 /* rfc1738 reserved chars, preserved from encoding. */
95 #define RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
97 /* rfc1738 unsafe chars, plus some more. */
99 #define UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
101 const static unsigned char urlchr_table[256] =
103 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
104 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
105 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
106 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
107 U, 0, U, RU, 0, U, R, 0, /* SP ! " # $ % & ' */
108 0, 0, 0, R, 0, 0, 0, R, /* ( ) * + , - . / */
109 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
110 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
111 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
112 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
113 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
114 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
115 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
116 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
117 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
118 0, 0, 0, U, U, U, U, U, /* x y z { | } ~ DEL */
120 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
121 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
122 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
123 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
125 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
126 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
127 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
128 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
131 /* Decodes the forms %xy in a URL to the character the hexadecimal
132 code of which is xy. xy are hexadecimal digits from
133 [0123456789ABCDEF] (case-insensitive). If x or y are not
134 hex-digits or `%' precedes `\0', the sequence is inserted
138 decode_string (char *s)
140 char *t = s; /* t - tortoise */
141 char *h = s; /* h - hare */
152 /* Do nothing if '%' is not followed by two hex digits. */
153 if (!*(h + 1) || !*(h + 2)
154 || !(ISXDIGIT (*(h + 1)) && ISXDIGIT (*(h + 2))))
156 *t = (XCHAR_TO_XDIGIT (*(h + 1)) << 4) + XCHAR_TO_XDIGIT (*(h + 2));
163 /* Like encode_string, but return S if there are no unsafe chars. */
166 encode_string_maybe (const char *s)
173 for (p1 = s; *p1; p1++)
174 if (UNSAFE_CHAR (*p1))
175 addition += 2; /* Two more characters (hex digits) */
180 newlen = (p1 - s) + addition;
181 newstr = (char *)xmalloc (newlen + 1);
187 if (UNSAFE_CHAR (*p1))
189 unsigned char c = *p1++;
191 *p2++ = XDIGIT_TO_XCHAR (c >> 4);
192 *p2++ = XDIGIT_TO_XCHAR (c & 0xf);
198 assert (p2 - newstr == newlen);
203 /* Encode the unsafe characters (as determined by UNSAFE_CHAR) in a
204 given string, returning a malloc-ed %XX encoded string. */
207 encode_string (const char *s)
209 char *encoded = encode_string_maybe (s);
216 /* Encode unsafe characters in PTR to %xx. If such encoding is done,
217 the old value of PTR is freed and PTR is made to point to the newly
218 allocated storage. */
220 #define ENCODE(ptr) do { \
221 char *e_new = encode_string_maybe (ptr); \
229 enum copy_method { CM_DECODE, CM_ENCODE, CM_PASSTHROUGH };
231 /* Decide whether to encode, decode, or pass through the char at P.
232 This used to be a macro, but it got a little too convoluted. */
233 static inline enum copy_method
234 decide_copy_method (const char *p)
238 if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2)))
240 /* %xx sequence: decode it, unless it would decode to an
241 unsafe or a reserved char; in that case, leave it as
243 char preempt = (XCHAR_TO_XDIGIT (*(p + 1)) << 4) +
244 XCHAR_TO_XDIGIT (*(p + 2));
246 if (UNSAFE_CHAR (preempt) || RESERVED_CHAR (preempt))
247 return CM_PASSTHROUGH;
252 /* Garbled %.. sequence: encode `%'. */
255 else if (UNSAFE_CHAR (*p) && !RESERVED_CHAR (*p))
258 return CM_PASSTHROUGH;
261 /* Translate a %-quoting (but possibly non-conformant) input string S
262 into a %-quoting (and conformant) output string. If no characters
263 are encoded or decoded, return the same string S; otherwise, return
264 a freshly allocated string with the new contents.
266 After a URL has been run through this function, the protocols that
267 use `%' as the quote character can use the resulting string as-is,
268 while those that don't call decode_string() to get to the intended
269 data. This function is also stable: after an input string is
270 transformed the first time, all further transformations of the
271 result yield the same result string.
273 Let's discuss why this function is needed.
275 Imagine Wget is to retrieve `http://abc.xyz/abc def'. Since a raw
276 space character would mess up the HTTP request, it needs to be
279 GET /abc%20def HTTP/1.0
281 So it appears that the unsafe chars need to be quoted, as with
282 encode_string. But what if we're requested to download
283 `abc%20def'? Remember that %-encoding is valid URL syntax, so what
284 the user meant was a literal space, and he was kind enough to quote
285 it. In that case, Wget should obviously leave the `%20' as is, and
286 send the same request as above. So in this case we may not call
289 But what if the requested URI is `abc%20 def'? If we call
290 encode_string, we end up with `/abc%2520%20def', which is almost
291 certainly not intended. If we don't call encode_string, we are
292 left with the embedded space and cannot send the request. What the
293 user meant was for Wget to request `/abc%20%20def', and this is
294 where reencode_string kicks in.
296 Wget used to solve this by first decoding %-quotes, and then
297 encoding all the "unsafe" characters found in the resulting string.
298 This was wrong because it didn't preserve certain URL special
299 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
300 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
301 whether we considered `+' reserved (it is). One of these results
302 is inevitable because by the second step we would lose information
303 on whether the `+' was originally encoded or not. Both results
304 were wrong because in CGI parameters + means space, while %2B means
305 literal plus. reencode_string correctly translates the above to
306 "a%2B+b", i.e. returns the original string.
308 This function uses an algorithm proposed by Anon Sricharoenchai:
310 1. Encode all URL_UNSAFE and the "%" that are not followed by 2
313 2. Decode all "%XX" except URL_UNSAFE, URL_RESERVED (";/?:@=&") and
316 ...except that this code conflates the two steps, and decides
317 whether to encode, decode, or pass through each character in turn.
318 The function still uses two passes, but their logic is the same --
319 the first pass exists merely for the sake of allocation. Another
320 small difference is that we include `+' to URL_RESERVED.
324 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
326 "http://abc.xyz/%20%3F%2561%25aa%25%20a?a=a+a%2Ba&b=b%26c%3Dc"
330 "foo bar" -> "foo%20bar"
331 "foo%20bar" -> "foo%20bar"
332 "foo %20bar" -> "foo%20%20bar"
333 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
334 "foo%25%20bar" -> "foo%25%20bar"
335 "foo%2%20bar" -> "foo%252%20bar"
336 "foo+bar" -> "foo+bar" (plus is reserved!)
337 "foo%2b+bar" -> "foo%2b+bar" */
340 reencode_string (const char *s)
346 int encode_count = 0;
347 int decode_count = 0;
349 /* First, pass through the string to see if there's anything to do,
350 and to calculate the new length. */
351 for (p1 = s; *p1; p1++)
353 switch (decide_copy_method (p1))
366 if (!encode_count && !decode_count)
367 /* The string is good as it is. */
368 return (char *)s; /* C const model sucks. */
371 /* Each encoding adds two characters (hex digits), while each
372 decoding removes two characters. */
373 newlen = oldlen + 2 * (encode_count - decode_count);
374 newstr = xmalloc (newlen + 1);
381 switch (decide_copy_method (p1))
385 unsigned char c = *p1++;
387 *p2++ = XDIGIT_TO_XCHAR (c >> 4);
388 *p2++ = XDIGIT_TO_XCHAR (c & 0xf);
392 *p2++ = ((XCHAR_TO_XDIGIT (*(p1 + 1)) << 4)
393 + (XCHAR_TO_XDIGIT (*(p1 + 2))));
394 p1 += 3; /* skip %xx */
401 assert (p2 - newstr == newlen);
405 /* Run PTR_VAR through reencode_string. If a new string is consed,
406 free PTR_VAR and make it point to the new storage. Obviously,
407 PTR_VAR needs to be an lvalue. */
409 #define REENCODE(ptr_var) do { \
410 char *rf_new = reencode_string (ptr_var); \
411 if (rf_new != ptr_var) \
418 /* Returns the scheme type if the scheme is supported, or
419 SCHEME_INVALID if not. */
421 url_scheme (const char *url)
425 for (i = 0; supported_schemes[i].leading_string; i++)
426 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
427 strlen (supported_schemes[i].leading_string)))
429 if (supported_schemes[i].enabled)
430 return (enum url_scheme) i;
432 return SCHEME_INVALID;
435 return SCHEME_INVALID;
438 /* Return the number of characters needed to skip the scheme part of
439 the URL, e.g. `http://'. If no scheme is found, returns 0. */
441 url_skip_scheme (const char *url)
445 /* Skip the scheme name. We allow `-' and `+' because of `whois++',
447 while (ISALNUM (*p) || *p == '-' || *p == '+')
454 /* Skip "//" if found. */
455 if (*p == '/' && *(p + 1) == '/')
461 /* Returns 1 if the URL begins with a scheme (supported or
462 unsupported), 0 otherwise. */
464 url_has_scheme (const char *url)
467 while (ISALNUM (*p) || *p == '-' || *p == '+')
473 scheme_default_port (enum url_scheme scheme)
475 return supported_schemes[scheme].default_port;
479 scheme_disable (enum url_scheme scheme)
481 supported_schemes[scheme].enabled = 0;
484 /* Skip the username and password, if present here. The function
485 should be called *not* with the complete URL, but with the part
486 right after the scheme.
488 If no username and password are found, return 0. */
490 url_skip_uname (const char *url)
494 /* Look for '@' that comes before '/' or '?'. */
495 p = (const char *)strpbrk (url, "/?@");
503 parse_uname (const char *str, int len, char **user, char **passwd)
508 /* Empty user name not allowed. */
511 colon = memchr (str, ':', len);
513 /* Empty user name again. */
518 int pwlen = len - (colon + 1 - str);
519 *passwd = xmalloc (pwlen + 1);
520 memcpy (*passwd, colon + 1, pwlen);
521 (*passwd)[pwlen] = '\0';
527 *user = xmalloc (len + 1);
528 memcpy (*user, str, len);
532 decode_string (*user);
534 decode_string (*passwd);
539 /* Used by main.c: detect URLs written using the "shorthand" URL forms
540 popularized by Netscape and NcFTP. HTTP shorthands look like this:
542 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
543 www.foo.com[:port] -> http://www.foo.com[:port]
545 FTP shorthands look like this:
547 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
548 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
550 If the URL needs not or cannot be rewritten, return NULL. */
552 rewrite_shorthand_url (const char *url)
556 if (url_has_scheme (url))
559 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
561 for (p = url; *p && *p != ':' && *p != '/'; p++)
571 /* If the characters after the colon and before the next slash
572 or end of string are all digits, it's HTTP. */
574 for (pp = p + 1; ISDIGIT (*pp); pp++)
576 if (digits > 0 && (*pp == '/' || *pp == '\0'))
579 /* Prepend "ftp://" to the entire URL... */
580 res = xmalloc (6 + strlen (url) + 1);
581 sprintf (res, "ftp://%s", url);
582 /* ...and replace ':' with '/'. */
583 res[6 + (p - url)] = '/';
590 /* Just prepend "http://" to what we have. */
591 res = xmalloc (7 + strlen (url) + 1);
592 sprintf (res, "http://%s", url);
597 static void parse_path PARAMS ((const char *, char **, char **));
600 strpbrk_or_eos (const char *s, const char *accept)
602 char *p = strpbrk (s, accept);
604 p = (char *)s + strlen (s);
608 /* Turn STR into lowercase; return non-zero if a character was
612 lowercase_str (char *str)
619 *str = TOLOWER (*str);
624 static char *parse_errors[] = {
625 #define PE_NO_ERROR 0
627 #define PE_UNSUPPORTED_SCHEME 1
628 "Unsupported scheme",
629 #define PE_EMPTY_HOST 2
631 #define PE_BAD_PORT_NUMBER 3
633 #define PE_INVALID_USER_NAME 4
635 #define PE_UNTERMINATED_IPV6_ADDRESS 5
636 "Unterminated IPv6 numeric address",
637 #define PE_INVALID_IPV6_ADDRESS 6
638 "Invalid char in IPv6 numeric address"
641 #define SETERR(p, v) do { \
648 Return a new struct url if successful, NULL on error. In case of
649 error, and if ERROR is not NULL, also set *ERROR to the appropriate
652 url_parse (const char *url, int *error)
656 int path_modified, host_modified;
658 enum url_scheme scheme;
660 const char *uname_b, *uname_e;
661 const char *host_b, *host_e;
662 const char *path_b, *path_e;
663 const char *params_b, *params_e;
664 const char *query_b, *query_e;
665 const char *fragment_b, *fragment_e;
668 char *user = NULL, *passwd = NULL;
672 scheme = url_scheme (url);
673 if (scheme == SCHEME_INVALID)
675 SETERR (error, PE_UNSUPPORTED_SCHEME);
679 url_encoded = reencode_string (url);
682 p += strlen (supported_schemes[scheme].leading_string);
684 p += url_skip_uname (p);
687 /* scheme://user:pass@host[:port]... */
690 /* We attempt to break down the URL into the components path,
691 params, query, and fragment. They are ordered like this:
693 scheme://host[:port][/path][;params][?query][#fragment] */
695 params_b = params_e = NULL;
696 query_b = query_e = NULL;
697 fragment_b = fragment_e = NULL;
703 /* Support http://[::1]/ used by IPv6. */
712 SETERR (error, PE_UNTERMINATED_IPV6_ADDRESS);
714 case '0': case '1': case '2': case '3': case '4':
715 case '5': case '6': case '7': case '8': case '9':
724 SETERR (error, PE_INVALID_IPV6_ADDRESS);
727 /* Don't include brackets in [host_b, host_p). */
733 p = strpbrk_or_eos (p, ":/;?#");
737 if (host_b == host_e)
739 SETERR (error, PE_EMPTY_HOST);
743 port = scheme_default_port (scheme);
746 const char *port_b, *port_e, *pp;
748 /* scheme://host:port/tralala */
752 p = strpbrk_or_eos (p, "/;?#");
755 if (port_b == port_e)
757 /* http://host:/whatever */
759 SETERR (error, PE_BAD_PORT_NUMBER);
763 for (port = 0, pp = port_b; pp < port_e; pp++)
767 /* http://host:12randomgarbage/blah */
769 SETERR (error, PE_BAD_PORT_NUMBER);
772 port = 10 * port + (*pp - '0');
780 p = strpbrk_or_eos (p, ";?#");
785 /* Path is not allowed not to exist. */
793 p = strpbrk_or_eos (p, "?#");
800 p = strpbrk_or_eos (p, "#");
812 if (uname_b != uname_e)
814 /* http://user:pass@host */
816 /* uname_b uname_e */
817 if (!parse_uname (uname_b, uname_e - uname_b - 1, &user, &passwd))
819 SETERR (error, PE_INVALID_USER_NAME);
824 u = (struct url *)xmalloc (sizeof (struct url));
825 memset (u, 0, sizeof (*u));
828 u->host = strdupdelim (host_b, host_e);
833 u->path = strdupdelim (path_b, path_e);
834 path_modified = path_simplify (u->path);
835 parse_path (u->path, &u->dir, &u->file);
837 host_modified = lowercase_str (u->host);
840 u->params = strdupdelim (params_b, params_e);
842 u->query = strdupdelim (query_b, query_e);
844 u->fragment = strdupdelim (fragment_b, fragment_e);
846 if (path_modified || u->fragment || host_modified || path_b == path_e)
848 /* If we suspect that a transformation has rendered what
849 url_string might return different from URL_ENCODED, rebuild
850 u->url using url_string. */
851 u->url = url_string (u, 0);
853 if (url_encoded != url)
854 xfree ((char *) url_encoded);
858 if (url_encoded == url)
859 u->url = xstrdup (url);
861 u->url = url_encoded;
869 url_error (int error_code)
871 assert (error_code >= 0 && error_code < ARRAY_SIZE (parse_errors));
872 return parse_errors[error_code];
876 parse_path (const char *quoted_path, char **dir, char **file)
878 char *path, *last_slash;
880 STRDUP_ALLOCA (path, quoted_path);
881 decode_string (path);
883 last_slash = strrchr (path, '/');
887 *file = xstrdup (path);
891 *dir = strdupdelim (path, last_slash);
892 *file = xstrdup (last_slash + 1);
896 /* Note: URL's "full path" is the path with the query string and
897 params appended. The "fragment" (#foo) is intentionally ignored,
898 but that might be changed. For example, if the original URL was
899 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
900 the full path will be "/foo/bar/baz;bullshit?querystring". */
902 /* Return the length of the full path, without the terminating
906 full_path_length (const struct url *url)
910 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
921 /* Write out the full path. */
924 full_path_write (const struct url *url, char *where)
926 #define FROB(el, chr) do { \
927 char *f_el = url->el; \
929 int l = strlen (f_el); \
931 memcpy (where, f_el, l); \
943 /* Public function for getting the "full path". E.g. if u->path is
944 "foo/bar" and u->query is "param=value", full_path will be
945 "/foo/bar?param=value". */
948 url_full_path (const struct url *url)
950 int length = full_path_length (url);
951 char *full_path = (char *)xmalloc(length + 1);
953 full_path_write (url, full_path);
954 full_path[length] = '\0';
959 /* Sync u->path and u->url with u->dir and u->file. */
962 sync_path (struct url *url)
970 newpath = xstrdup (url->file);
975 int dirlen = strlen (url->dir);
976 int filelen = strlen (url->file);
978 newpath = xmalloc (dirlen + 1 + filelen + 1);
979 memcpy (newpath, url->dir, dirlen);
980 newpath[dirlen] = '/';
981 memcpy (newpath + dirlen + 1, url->file, filelen);
982 newpath[dirlen + 1 + filelen] = '\0';
988 /* Synchronize u->url. */
990 url->url = url_string (url, 0);
993 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
994 This way we can sync u->path and u->url when they get changed. */
997 url_set_dir (struct url *url, const char *newdir)
1000 url->dir = xstrdup (newdir);
1005 url_set_file (struct url *url, const char *newfile)
1008 url->file = xstrdup (newfile);
1013 url_free (struct url *url)
1019 FREE_MAYBE (url->params);
1020 FREE_MAYBE (url->query);
1021 FREE_MAYBE (url->fragment);
1022 FREE_MAYBE (url->user);
1023 FREE_MAYBE (url->passwd);
1032 get_urls_file (const char *file)
1034 struct file_memory *fm;
1035 struct urlpos *head, *tail;
1036 const char *text, *text_end;
1038 /* Load the file. */
1039 fm = read_file (file);
1042 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
1045 DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
1049 text_end = fm->content + fm->length;
1050 while (text < text_end)
1052 const char *line_beg = text;
1053 const char *line_end = memchr (text, '\n', text_end - text);
1055 line_end = text_end;
1060 /* Strip whitespace from the beginning and end of line. */
1061 while (line_beg < line_end && ISSPACE (*line_beg))
1063 while (line_end > line_beg && ISSPACE (*(line_end - 1)))
1066 if (line_end > line_beg)
1068 /* URL is in the [line_beg, line_end) region. */
1072 struct urlpos *entry;
1075 /* We must copy the URL to a zero-terminated string, and we
1076 can't use alloca because we're in a loop. *sigh*. */
1077 url_text = strdupdelim (line_beg, line_end);
1081 /* Merge opt.base_href with URL. */
1082 char *merged = uri_merge (opt.base_href, url_text);
1087 url = url_parse (url_text, &up_error_code);
1090 logprintf (LOG_NOTQUIET, "%s: Invalid URL %s: %s\n",
1091 file, url_text, url_error (up_error_code));
1097 entry = (struct urlpos *)xmalloc (sizeof (struct urlpos));
1098 memset (entry, 0, sizeof (*entry));
1109 read_file_free (fm);
1113 /* Free the linked list of urlpos. */
1115 free_urlpos (struct urlpos *l)
1119 struct urlpos *next = l->next;
1122 FREE_MAYBE (l->local_name);
1128 /* Rotate FNAME opt.backups times */
1130 rotate_backups(const char *fname)
1132 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
1133 char *from = (char *)alloca (maxlen);
1134 char *to = (char *)alloca (maxlen);
1138 if (stat (fname, &sb) == 0)
1139 if (S_ISREG (sb.st_mode) == 0)
1142 for (i = opt.backups; i > 1; i--)
1144 sprintf (from, "%s.%d", fname, i - 1);
1145 sprintf (to, "%s.%d", fname, i);
1146 /* #### This will fail on machines without the rename() system
1151 sprintf (to, "%s.%d", fname, 1);
1155 /* Create all the necessary directories for PATH (a file). Calls
1156 mkdirhier() internally. */
1158 mkalldirs (const char *path)
1165 p = path + strlen (path);
1166 for (; *p != '/' && p != path; p--);
1167 /* Don't create if it's just a file. */
1168 if ((p == path) && (*p != '/'))
1170 t = strdupdelim (path, p);
1171 /* Check whether the directory exists. */
1172 if ((stat (t, &st) == 0))
1174 if (S_ISDIR (st.st_mode))
1181 /* If the dir exists as a file name, remove it first. This
1182 is *only* for Wget to work with buggy old CERN http
1183 servers. Here is the scenario: When Wget tries to
1184 retrieve a directory without a slash, e.g.
1185 http://foo/bar (bar being a directory), CERN server will
1186 not redirect it too http://foo/bar/ -- it will generate a
1187 directory listing containing links to bar/file1,
1188 bar/file2, etc. Wget will lose because it saves this
1189 HTML listing to a file `bar', so it cannot create the
1190 directory. To work around this, if the file of the same
1191 name exists, we just remove it and create the directory
1193 DEBUGP (("Removing %s because of directory danger!\n", t));
1197 res = make_directory (t);
1199 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1205 count_slashes (const char *s)
1214 /* Return the path name of the URL-equivalent file name, with a
1215 remote-like structure of directories. */
1217 mkstruct (const struct url *u)
1219 char *dir, *dir_preencoding;
1220 char *file, *res, *dirpref;
1221 char *query = u->query && *u->query ? u->query : NULL;
1226 char *ptr = u->dir + (*u->dir == '/');
1227 int slash_count = 1 + count_slashes (ptr);
1228 int cut = MINVAL (opt.cut_dirs, slash_count);
1229 for (; cut && *ptr; ptr++)
1232 STRDUP_ALLOCA (dir, ptr);
1235 dir = u->dir + (*u->dir == '/');
1237 /* Check for the true name (or at least a consistent name for saving
1238 to directory) of HOST, reusing the hlist if possible. */
1239 if (opt.add_hostdir)
1241 /* Add dir_prefix and hostname (if required) to the beginning of
1243 dirpref = (char *)alloca (strlen (opt.dir_prefix) + 1
1245 + 1 + numdigit (u->port)
1247 if (!DOTP (opt.dir_prefix))
1248 sprintf (dirpref, "%s/%s", opt.dir_prefix, u->host);
1250 strcpy (dirpref, u->host);
1252 if (u->port != scheme_default_port (u->scheme))
1254 int len = strlen (dirpref);
1256 number_to_string (dirpref + len + 1, u->port);
1259 else /* not add_hostdir */
1261 if (!DOTP (opt.dir_prefix))
1262 dirpref = opt.dir_prefix;
1267 /* If there is a prefix, prepend it. */
1270 char *newdir = (char *)alloca (strlen (dirpref) + 1 + strlen (dir) + 2);
1271 sprintf (newdir, "%s%s%s", dirpref, *dir == '/' ? "" : "/", dir);
1275 dir_preencoding = dir;
1276 dir = reencode_string (dir_preencoding);
1279 if (l && dir[l - 1] == '/')
1283 file = "index.html";
1287 /* Finally, construct the full name. */
1288 res = (char *)xmalloc (strlen (dir) + 1 + strlen (file)
1289 + (query ? (1 + strlen (query)) : 0)
1291 sprintf (res, "%s%s%s", dir, *dir ? "/" : "", file);
1295 strcat (res, query);
1297 if (dir != dir_preencoding)
1302 /* Compose a file name out of BASE, an unescaped file name, and QUERY,
1303 an escaped query string. The trick is to make sure that unsafe
1304 characters in BASE are escaped, and that slashes in QUERY are also
1308 compose_file_name (char *base, char *query)
1314 /* Copy BASE to RESULT and encode all unsafe characters. */
1316 while (*from && to - result < sizeof (result))
1318 if (UNSAFE_CHAR (*from))
1320 unsigned char c = *from++;
1322 *to++ = XDIGIT_TO_XCHAR (c >> 4);
1323 *to++ = XDIGIT_TO_XCHAR (c & 0xf);
1329 if (query && to - result < sizeof (result))
1333 /* Copy QUERY to RESULT and encode all '/' characters. */
1335 while (*from && to - result < sizeof (result))
1349 if (to - result < sizeof (result))
1352 /* Truncate input which is too long, presumably due to a huge
1354 result[sizeof (result) - 1] = '\0';
1356 return xstrdup (result);
1359 /* Create a unique filename, corresponding to a given URL. Calls
1360 mkstruct if necessary. Does *not* actually create any directories. */
1362 url_filename (const struct url *u)
1365 int have_prefix = 0; /* whether we must prepend opt.dir_prefix */
1369 file = mkstruct (u);
1374 char *base = *u->file ? u->file : "index.html";
1375 char *query = u->query && *u->query ? u->query : NULL;
1376 file = compose_file_name (base, query);
1381 /* Check whether the prefix directory is something other than "."
1382 before prepending it. */
1383 if (!DOTP (opt.dir_prefix))
1385 char *nfile = (char *)xmalloc (strlen (opt.dir_prefix)
1386 + 1 + strlen (file) + 1);
1387 sprintf (nfile, "%s/%s", opt.dir_prefix, file);
1392 /* DOS-ish file systems don't like `%' signs in them; we change it
1397 for (p = file; *p; p++)
1401 #endif /* WINDOWS */
1403 /* Check the cases in which the unique extensions are not used:
1404 1) Clobbering is turned off (-nc).
1405 2) Retrieval with regetting.
1406 3) Timestamping is used.
1407 4) Hierarchy is built.
1409 The exception is the case when file does exist and is a
1410 directory (actually support for bad httpd-s). */
1411 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1412 && !(file_exists_p (file) && !file_non_directory_p (file)))
1415 /* Find a unique name. */
1416 name = unique_name (file);
1421 /* Return the langth of URL's path. Path is considered to be
1422 terminated by one of '?', ';', '#', or by the end of the
1425 path_length (const char *url)
1427 const char *q = strpbrk_or_eos (url, "?;#");
1431 /* Find the last occurrence of character C in the range [b, e), or
1432 NULL, if none are present. This is equivalent to strrchr(b, c),
1433 except that it accepts an END argument instead of requiring the
1434 string to be zero-terminated. Why is there no memrchr()? */
1436 find_last_char (const char *b, const char *e, char c)
1444 /* Resolve "." and ".." elements of PATH by destructively modifying
1445 PATH. "." is resolved by removing that path element, and ".." is
1446 resolved by removing the preceding path element. Leading and
1447 trailing slashes are preserved.
1449 Return non-zero if any changes have been made.
1451 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1452 test examples are provided below. If you change anything in this
1453 function, run test_path_simplify to make sure you haven't broken a
1456 A previous version of this function was based on path_simplify()
1457 from GNU Bash, but it has been rewritten for Wget 1.8.1. */
1460 path_simplify (char *path)
1466 ++path; /* preserve the leading '/'. */
1469 end = p + strlen (p) + 1; /* position past the terminating zero. */
1474 /* P should point to the beginning of a path element. */
1476 if (*p == '.' && (*(p + 1) == '/' || *(p + 1) == '\0'))
1478 /* Handle "./foo" by moving "foo" two characters to the
1480 if (*(p + 1) == '/')
1483 memmove (p, p + 2, end - p);
1494 else if (*p == '.' && *(p + 1) == '.'
1495 && (*(p + 2) == '/' || *(p + 2) == '\0'))
1497 /* Handle "../foo" by moving "foo" one path element to the
1499 char *b = p; /* not p-1 because P can equal PATH */
1501 /* Backtrack by one path element, but not past the beginning
1504 /* foo/bar/../baz */
1510 /* Move backwards until B hits the beginning of the
1511 previous path element or the beginning of path. */
1512 for (--b; b > path && *(b - 1) != '/'; b--)
1517 if (*(p + 2) == '/')
1519 memmove (b, p + 3, end - (p + 3));
1533 /* Remove empty path elements. Not mandated by rfc1808 et
1534 al, but empty path elements are not all that useful, and
1535 the rest of Wget might not deal with them well. */
1545 memmove (p, q, end - q);
1550 /* Skip to the next path element. */
1551 while (*p && *p != '/')
1556 /* Make sure P points to the beginning of the next path element,
1557 which is location after the slash. */
1564 /* Resolve the result of "linking" a base URI (BASE) to a
1565 link-specified URI (LINK).
1567 Either of the URIs may be absolute or relative, complete with the
1568 host name, or path only. This tries to behave "reasonably" in all
1569 foreseeable cases. It employs little specific knowledge about
1570 schemes or URL-specific stuff -- it just works on strings.
1572 The parameters LINKLENGTH is useful if LINK is not zero-terminated.
1573 See uri_merge for a gentler interface to this functionality.
1575 Perhaps this function should call path_simplify so that the callers
1576 don't have to call url_parse unconditionally. */
1578 uri_merge_1 (const char *base, const char *link, int linklength, int no_scheme)
1584 const char *end = base + path_length (base);
1588 /* Empty LINK points back to BASE, query string and all. */
1589 constr = xstrdup (base);
1591 else if (*link == '?')
1593 /* LINK points to the same location, but changes the query
1594 string. Examples: */
1595 /* uri_merge("path", "?new") -> "path?new" */
1596 /* uri_merge("path?foo", "?new") -> "path?new" */
1597 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1598 /* uri_merge("path#foo", "?new") -> "path?new" */
1599 int baselength = end - base;
1600 constr = xmalloc (baselength + linklength + 1);
1601 memcpy (constr, base, baselength);
1602 memcpy (constr + baselength, link, linklength);
1603 constr[baselength + linklength] = '\0';
1605 else if (*link == '#')
1607 /* uri_merge("path", "#new") -> "path#new" */
1608 /* uri_merge("path#foo", "#new") -> "path#new" */
1609 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1610 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1612 const char *end1 = strchr (base, '#');
1614 end1 = base + strlen (base);
1615 baselength = end1 - base;
1616 constr = xmalloc (baselength + linklength + 1);
1617 memcpy (constr, base, baselength);
1618 memcpy (constr + baselength, link, linklength);
1619 constr[baselength + linklength] = '\0';
1621 else if (linklength > 1 && *link == '/' && *(link + 1) == '/')
1623 /* LINK begins with "//" and so is a net path: we need to
1624 replace everything after (and including) the double slash
1627 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1628 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1629 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1633 const char *start_insert;
1635 /* Look for first slash. */
1636 slash = memchr (base, '/', end - base);
1637 /* If found slash and it is a double slash, then replace
1638 from this point, else default to replacing from the
1640 if (slash && *(slash + 1) == '/')
1641 start_insert = slash;
1643 start_insert = base;
1645 span = start_insert - base;
1646 constr = (char *)xmalloc (span + linklength + 1);
1648 memcpy (constr, base, span);
1649 memcpy (constr + span, link, linklength);
1650 constr[span + linklength] = '\0';
1652 else if (*link == '/')
1654 /* LINK is an absolute path: we need to replace everything
1655 after (and including) the FIRST slash with LINK.
1657 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1658 "/qux/xyzzy", our result should be
1659 "http://host/qux/xyzzy". */
1662 const char *start_insert = NULL; /* for gcc to shut up. */
1663 const char *pos = base;
1664 int seen_slash_slash = 0;
1665 /* We're looking for the first slash, but want to ignore
1668 slash = memchr (pos, '/', end - pos);
1669 if (slash && !seen_slash_slash)
1670 if (*(slash + 1) == '/')
1673 seen_slash_slash = 1;
1677 /* At this point, SLASH is the location of the first / after
1678 "//", or the first slash altogether. START_INSERT is the
1679 pointer to the location where LINK will be inserted. When
1680 examining the last two examples, keep in mind that LINK
1683 if (!slash && !seen_slash_slash)
1684 /* example: "foo" */
1686 start_insert = base;
1687 else if (!slash && seen_slash_slash)
1688 /* example: "http://foo" */
1691 else if (slash && !seen_slash_slash)
1692 /* example: "foo/bar" */
1694 start_insert = base;
1695 else if (slash && seen_slash_slash)
1696 /* example: "http://something/" */
1698 start_insert = slash;
1700 span = start_insert - base;
1701 constr = (char *)xmalloc (span + linklength + 1);
1703 memcpy (constr, base, span);
1705 memcpy (constr + span, link, linklength);
1706 constr[span + linklength] = '\0';
1710 /* LINK is a relative URL: we need to replace everything
1711 after last slash (possibly empty) with LINK.
1713 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1714 our result should be "whatever/foo/qux/xyzzy". */
1715 int need_explicit_slash = 0;
1717 const char *start_insert;
1718 const char *last_slash = find_last_char (base, end, '/');
1721 /* No slash found at all. Append LINK to what we have,
1722 but we'll need a slash as a separator.
1724 Example: if base == "foo" and link == "qux/xyzzy", then
1725 we cannot just append link to base, because we'd get
1726 "fooqux/xyzzy", whereas what we want is
1729 To make sure the / gets inserted, we set
1730 need_explicit_slash to 1. We also set start_insert
1731 to end + 1, so that the length calculations work out
1732 correctly for one more (slash) character. Accessing
1733 that character is fine, since it will be the
1734 delimiter, '\0' or '?'. */
1735 /* example: "foo?..." */
1736 /* ^ ('?' gets changed to '/') */
1737 start_insert = end + 1;
1738 need_explicit_slash = 1;
1740 else if (last_slash && last_slash != base && *(last_slash - 1) == '/')
1742 /* example: http://host" */
1744 start_insert = end + 1;
1745 need_explicit_slash = 1;
1749 /* example: "whatever/foo/bar" */
1751 start_insert = last_slash + 1;
1754 span = start_insert - base;
1755 constr = (char *)xmalloc (span + linklength + 1);
1757 memcpy (constr, base, span);
1758 if (need_explicit_slash)
1759 constr[span - 1] = '/';
1761 memcpy (constr + span, link, linklength);
1762 constr[span + linklength] = '\0';
1765 else /* !no_scheme */
1767 constr = strdupdelim (link, link + linklength);
1772 /* Merge BASE with LINK and return the resulting URI. This is an
1773 interface to uri_merge_1 that assumes that LINK is a
1774 zero-terminated string. */
1776 uri_merge (const char *base, const char *link)
1778 return uri_merge_1 (base, link, strlen (link), !url_has_scheme (link));
1781 #define APPEND(p, s) do { \
1782 int len = strlen (s); \
1783 memcpy (p, s, len); \
1787 /* Use this instead of password when the actual password is supposed
1788 to be hidden. We intentionally use a generic string without giving
1789 away the number of characters in the password, like previous
1791 #define HIDDEN_PASSWORD "*password*"
1793 /* Recreate the URL string from the data in URL.
1795 If HIDE is non-zero (as it is when we're calling this on a URL we
1796 plan to print, but not when calling it to canonicalize a URL for
1797 use within the program), password will be hidden. Unsafe
1798 characters in the URL will be quoted. */
1801 url_string (const struct url *url, int hide_password)
1805 char *quoted_user = NULL, *quoted_passwd = NULL;
1807 int scheme_port = supported_schemes[url->scheme].default_port;
1808 char *scheme_str = supported_schemes[url->scheme].leading_string;
1809 int fplen = full_path_length (url);
1811 int brackets_around_host = 0;
1813 assert (scheme_str != NULL);
1815 /* Make sure the user name and password are quoted. */
1818 quoted_user = encode_string_maybe (url->user);
1822 quoted_passwd = HIDDEN_PASSWORD;
1824 quoted_passwd = encode_string_maybe (url->passwd);
1828 if (strchr (url->host, ':'))
1829 brackets_around_host = 1;
1831 size = (strlen (scheme_str)
1832 + strlen (url->host)
1833 + (brackets_around_host ? 2 : 0)
1836 if (url->port != scheme_port)
1837 size += 1 + numdigit (url->port);
1840 size += 1 + strlen (quoted_user);
1842 size += 1 + strlen (quoted_passwd);
1845 p = result = xmalloc (size);
1847 APPEND (p, scheme_str);
1850 APPEND (p, quoted_user);
1854 APPEND (p, quoted_passwd);
1859 if (brackets_around_host)
1861 APPEND (p, url->host);
1862 if (brackets_around_host)
1864 if (url->port != scheme_port)
1867 p = number_to_string (p, url->port);
1870 full_path_write (url, p);
1874 assert (p - result == size);
1876 if (quoted_user && quoted_user != url->user)
1877 xfree (quoted_user);
1878 if (quoted_passwd && !hide_password
1879 && quoted_passwd != url->passwd)
1880 xfree (quoted_passwd);
1885 /* Returns proxy host address, in accordance with SCHEME. */
1887 getproxy (enum url_scheme scheme)
1890 char *rewritten_url;
1891 static char rewritten_storage[1024];
1896 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
1900 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
1904 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
1906 case SCHEME_INVALID:
1909 if (!proxy || !*proxy)
1912 /* Handle shorthands. */
1913 rewritten_url = rewrite_shorthand_url (proxy);
1916 strncpy (rewritten_storage, rewritten_url, sizeof(rewritten_storage));
1917 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
1918 proxy = rewritten_storage;
1924 /* Should a host be accessed through proxy, concerning no_proxy? */
1926 no_proxy_match (const char *host, const char **no_proxy)
1931 return !sufmatch (no_proxy, host);
1934 /* Support for converting links for local viewing in downloaded HTML
1935 files. This should be moved to another file, because it has
1936 nothing to do with processing URLs. */
1938 static void write_backup_file PARAMS ((const char *, downloaded_file_t));
1939 static const char *replace_attr PARAMS ((const char *, int, FILE *,
1941 static const char *replace_attr_refresh_hack PARAMS ((const char *, int, FILE *,
1942 const char *, int));
1943 static char *local_quote_string PARAMS ((const char *));
1945 /* Change the links in one HTML file. LINKS is a list of links in the
1946 document, along with their positions and the desired direction of
1949 convert_links (const char *file, struct urlpos *links)
1951 struct file_memory *fm;
1954 downloaded_file_t downloaded_file_return;
1956 struct urlpos *link;
1957 int to_url_count = 0, to_file_count = 0;
1959 logprintf (LOG_VERBOSE, _("Converting %s... "), file);
1962 /* First we do a "dry run": go through the list L and see whether
1963 any URL needs to be converted in the first place. If not, just
1964 leave the file alone. */
1966 struct urlpos *dry = links;
1967 for (dry = links; dry; dry = dry->next)
1968 if (dry->convert != CO_NOCONVERT)
1972 logputs (LOG_VERBOSE, _("nothing to do.\n"));
1977 fm = read_file (file);
1980 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1981 file, strerror (errno));
1985 downloaded_file_return = downloaded_file (CHECK_FOR_FILE, file);
1986 if (opt.backup_converted && downloaded_file_return)
1987 write_backup_file (file, downloaded_file_return);
1989 /* Before opening the file for writing, unlink the file. This is
1990 important if the data in FM is mmaped. In such case, nulling the
1991 file, which is what fopen() below does, would make us read all
1992 zeroes from the mmaped region. */
1993 if (unlink (file) < 0 && errno != ENOENT)
1995 logprintf (LOG_NOTQUIET, _("Unable to delete `%s': %s\n"),
1996 file, strerror (errno));
1997 read_file_free (fm);
2000 /* Now open the file for writing. */
2001 fp = fopen (file, "wb");
2004 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
2005 file, strerror (errno));
2006 read_file_free (fm);
2010 /* Here we loop through all the URLs in file, replacing those of
2011 them that are downloaded with relative references. */
2013 for (link = links; link; link = link->next)
2015 char *url_start = fm->content + link->pos;
2017 if (link->pos >= fm->length)
2019 DEBUGP (("Something strange is going on. Please investigate."));
2022 /* If the URL is not to be converted, skip it. */
2023 if (link->convert == CO_NOCONVERT)
2025 DEBUGP (("Skipping %s at position %d.\n", link->url->url, link->pos));
2029 /* Echo the file contents, up to the offending URL's opening
2030 quote, to the outfile. */
2031 fwrite (p, 1, url_start - p, fp);
2034 switch (link->convert)
2036 case CO_CONVERT_TO_RELATIVE:
2037 /* Convert absolute URL to relative. */
2039 char *newname = construct_relative (file, link->local_name);
2040 char *quoted_newname = local_quote_string (newname);
2042 if (!link->link_refresh_p)
2043 p = replace_attr (p, link->size, fp, quoted_newname);
2045 p = replace_attr_refresh_hack (p, link->size, fp, quoted_newname,
2046 link->refresh_timeout);
2048 DEBUGP (("TO_RELATIVE: %s to %s at position %d in %s.\n",
2049 link->url->url, newname, link->pos, file));
2051 xfree (quoted_newname);
2055 case CO_CONVERT_TO_COMPLETE:
2056 /* Convert the link to absolute URL. */
2058 char *newlink = link->url->url;
2059 char *quoted_newlink = html_quote_string (newlink);
2061 if (!link->link_refresh_p)
2062 p = replace_attr (p, link->size, fp, quoted_newlink);
2064 p = replace_attr_refresh_hack (p, link->size, fp, quoted_newlink,
2065 link->refresh_timeout);
2067 DEBUGP (("TO_COMPLETE: <something> to %s at position %d in %s.\n",
2068 newlink, link->pos, file));
2069 xfree (quoted_newlink);
2073 case CO_NULLIFY_BASE:
2074 /* Change the base href to "". */
2075 p = replace_attr (p, link->size, fp, "");
2083 /* Output the rest of the file. */
2084 if (p - fm->content < fm->length)
2085 fwrite (p, 1, fm->length - (p - fm->content), fp);
2087 read_file_free (fm);
2089 logprintf (LOG_VERBOSE, "%d-%d\n", to_file_count, to_url_count);
2092 /* Construct and return a malloced copy of the relative link from two
2093 pieces of information: local name S1 of the referring file and
2094 local name S2 of the referred file.
2096 So, if S1 is "jagor.srce.hr/index.html" and S2 is
2097 "jagor.srce.hr/images/news.gif", the function will return
2100 Alternately, if S1 is "fly.cc.fer.hr/ioccc/index.html", and S2 is
2101 "fly.cc.fer.hr/images/fly.gif", the function will return
2102 "../images/fly.gif".
2104 Caveats: S1 should not begin with `/', unless S2 also begins with
2105 '/'. S1 should not contain things like ".." and such --
2106 construct_relative ("fly/ioccc/../index.html",
2107 "fly/images/fly.gif") will fail. (A workaround is to call
2108 something like path_simplify() on S1). */
2110 construct_relative (const char *s1, const char *s2)
2112 int i, cnt, sepdirs1;
2116 return xstrdup (s2);
2117 /* S1 should *not* be absolute, if S2 wasn't. */
2118 assert (*s1 != '/');
2120 /* Skip the directories common to both strings. */
2123 while (s1[i] && s2[i]
2128 if (s1[i] == '/' && s2[i] == '/')
2133 for (sepdirs1 = 0; s1[i]; i++)
2136 /* Now, construct the file as of:
2137 - ../ repeated sepdirs1 time
2138 - all the non-mutual directories of S2. */
2139 res = (char *)xmalloc (3 * sepdirs1 + strlen (s2 + cnt) + 1);
2140 for (i = 0; i < sepdirs1; i++)
2141 memcpy (res + 3 * i, "../", 3);
2142 strcpy (res + 3 * i, s2 + cnt);
2147 write_backup_file (const char *file, downloaded_file_t downloaded_file_return)
2149 /* Rather than just writing over the original .html file with the
2150 converted version, save the former to *.orig. Note we only do
2151 this for files we've _successfully_ downloaded, so we don't
2152 clobber .orig files sitting around from previous invocations. */
2154 /* Construct the backup filename as the original name plus ".orig". */
2155 size_t filename_len = strlen(file);
2156 char* filename_plus_orig_suffix;
2157 boolean already_wrote_backup_file = FALSE;
2158 slist* converted_file_ptr;
2159 static slist* converted_files = NULL;
2161 if (downloaded_file_return == FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED)
2163 /* Just write "orig" over "html". We need to do it this way
2164 because when we're checking to see if we've downloaded the
2165 file before (to see if we can skip downloading it), we don't
2166 know if it's a text/html file. Therefore we don't know yet
2167 at that stage that -E is going to cause us to tack on
2168 ".html", so we need to compare vs. the original URL plus
2169 ".orig", not the original URL plus ".html.orig". */
2170 filename_plus_orig_suffix = alloca (filename_len + 1);
2171 strcpy(filename_plus_orig_suffix, file);
2172 strcpy((filename_plus_orig_suffix + filename_len) - 4, "orig");
2174 else /* downloaded_file_return == FILE_DOWNLOADED_NORMALLY */
2176 /* Append ".orig" to the name. */
2177 filename_plus_orig_suffix = alloca (filename_len + sizeof(".orig"));
2178 strcpy(filename_plus_orig_suffix, file);
2179 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
2182 /* We can get called twice on the same URL thanks to the
2183 convert_all_links() call in main(). If we write the .orig file
2184 each time in such a case, it'll end up containing the first-pass
2185 conversion, not the original file. So, see if we've already been
2186 called on this file. */
2187 converted_file_ptr = converted_files;
2188 while (converted_file_ptr != NULL)
2189 if (strcmp(converted_file_ptr->string, file) == 0)
2191 already_wrote_backup_file = TRUE;
2195 converted_file_ptr = converted_file_ptr->next;
2197 if (!already_wrote_backup_file)
2199 /* Rename <file> to <file>.orig before former gets written over. */
2200 if (rename(file, filename_plus_orig_suffix) != 0)
2201 logprintf (LOG_NOTQUIET, _("Cannot back up %s as %s: %s\n"),
2202 file, filename_plus_orig_suffix, strerror (errno));
2204 /* Remember that we've already written a .orig backup for this file.
2205 Note that we never free this memory since we need it till the
2206 convert_all_links() call, which is one of the last things the
2207 program does before terminating. BTW, I'm not sure if it would be
2208 safe to just set 'converted_file_ptr->string' to 'file' below,
2209 rather than making a copy of the string... Another note is that I
2210 thought I could just add a field to the urlpos structure saying
2211 that we'd written a .orig file for this URL, but that didn't work,
2212 so I had to make this separate list.
2213 -- Dan Harkless <wget@harkless.org>
2215 This [adding a field to the urlpos structure] didn't work
2216 because convert_file() is called from convert_all_links at
2217 the end of the retrieval with a freshly built new urlpos
2219 -- Hrvoje Niksic <hniksic@arsdigita.com>
2221 converted_file_ptr = xmalloc(sizeof(*converted_file_ptr));
2222 converted_file_ptr->string = xstrdup(file); /* die on out-of-mem. */
2223 converted_file_ptr->next = converted_files;
2224 converted_files = converted_file_ptr;
2228 static int find_fragment PARAMS ((const char *, int, const char **,
2231 /* Replace an attribute's original text with NEW_TEXT. */
2234 replace_attr (const char *p, int size, FILE *fp, const char *new_text)
2237 char quote_char = '\"'; /* use "..." for quoting, unless the
2238 original value is quoted, in which
2239 case reuse its quoting char. */
2240 const char *frag_beg, *frag_end;
2242 /* Structure of our string is:
2243 "...old-contents..."
2244 <--- size ---> (with quotes)
2247 <--- size --> (no quotes) */
2249 if (*p == '\"' || *p == '\'')
2254 size -= 2; /* disregard opening and closing quote */
2256 putc (quote_char, fp);
2257 fputs (new_text, fp);
2259 /* Look for fragment identifier, if any. */
2260 if (find_fragment (p, size, &frag_beg, &frag_end))
2261 fwrite (frag_beg, 1, frag_end - frag_beg, fp);
2265 putc (quote_char, fp);
2270 /* The same as REPLACE_ATTR, but used when replacing
2271 <meta http-equiv=refresh content="new_text"> because we need to
2272 append "timeout_value; URL=" before the next_text. */
2275 replace_attr_refresh_hack (const char *p, int size, FILE *fp,
2276 const char *new_text, int timeout)
2279 char *new_with_timeout = (char *)alloca (numdigit (timeout)
2283 sprintf (new_with_timeout, "%d; URL=%s", timeout, new_text);
2285 return replace_attr (p, size, fp, new_with_timeout);
2288 /* Find the first occurrence of '#' in [BEG, BEG+SIZE) that is not
2289 preceded by '&'. If the character is not found, return zero. If
2290 the character is found, return 1 and set BP and EP to point to the
2291 beginning and end of the region.
2293 This is used for finding the fragment indentifiers in URLs. */
2296 find_fragment (const char *beg, int size, const char **bp, const char **ep)
2298 const char *end = beg + size;
2300 for (; beg < end; beg++)
2322 /* Quote FILE for use as local reference to an HTML file.
2324 We quote ? as %3F to avoid passing part of the file name as the
2325 parameter when browsing the converted file through HTTP. However,
2326 it is safe to do this only when `--html-extension' is turned on.
2327 This is because converting "index.html?foo=bar" to
2328 "index.html%3Ffoo=bar" would break local browsing, as the latter
2329 isn't even recognized as an HTML file! However, converting
2330 "index.html?foo=bar.html" to "index.html%3Ffoo=bar.html" should be
2331 safe for both local and HTTP-served browsing. */
2334 local_quote_string (const char *file)
2336 const char *file_sans_qmark;
2339 if (!opt.html_extension)
2340 return html_quote_string (file);
2342 qm = count_char (file, '?');
2346 const char *from = file;
2349 /* qm * 2 because we replace each question mark with "%3F",
2350 i.e. replace one char with three, hence two more. */
2351 int fsqlen = strlen (file) + qm * 2;
2353 to = newname = (char *)alloca (fsqlen + 1);
2354 for (; *from; from++)
2365 assert (to - newname == fsqlen);
2368 file_sans_qmark = newname;
2371 file_sans_qmark = file;
2373 return html_quote_string (file_sans_qmark);
2376 /* We're storing "modes" of type downloaded_file_t in the hash table.
2377 However, our hash tables only accept pointers for keys and values.
2378 So when we need a pointer, we use the address of a
2379 downloaded_file_t variable of static storage. */
2381 static downloaded_file_t *
2382 downloaded_mode_to_ptr (downloaded_file_t mode)
2384 static downloaded_file_t
2385 v1 = FILE_NOT_ALREADY_DOWNLOADED,
2386 v2 = FILE_DOWNLOADED_NORMALLY,
2387 v3 = FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED,
2388 v4 = CHECK_FOR_FILE;
2392 case FILE_NOT_ALREADY_DOWNLOADED:
2394 case FILE_DOWNLOADED_NORMALLY:
2396 case FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED:
2398 case CHECK_FOR_FILE:
2404 /* This should really be merged with dl_file_url_map and
2405 downloaded_html_files in recur.c. This was originally a list, but
2406 I changed it to a hash table beause it was actually taking a lot of
2407 time to find things in it. */
2409 static struct hash_table *downloaded_files_hash;
2411 /* Remembers which files have been downloaded. In the standard case, should be
2412 called with mode == FILE_DOWNLOADED_NORMALLY for each file we actually
2413 download successfully (i.e. not for ones we have failures on or that we skip
2416 When we've downloaded a file and tacked on a ".html" extension due to -E,
2417 call this function with FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED rather than
2418 FILE_DOWNLOADED_NORMALLY.
2420 If you just want to check if a file has been previously added without adding
2421 it, call with mode == CHECK_FOR_FILE. Please be sure to call this function
2422 with local filenames, not remote URLs. */
2424 downloaded_file (downloaded_file_t mode, const char *file)
2426 downloaded_file_t *ptr;
2428 if (mode == CHECK_FOR_FILE)
2430 if (!downloaded_files_hash)
2431 return FILE_NOT_ALREADY_DOWNLOADED;
2432 ptr = hash_table_get (downloaded_files_hash, file);
2434 return FILE_NOT_ALREADY_DOWNLOADED;
2438 if (!downloaded_files_hash)
2439 downloaded_files_hash = make_string_hash_table (0);
2441 ptr = hash_table_get (downloaded_files_hash, file);
2445 ptr = downloaded_mode_to_ptr (mode);
2446 hash_table_put (downloaded_files_hash, xstrdup (file), &ptr);
2448 return FILE_NOT_ALREADY_DOWNLOADED;
2452 df_free_mapper (void *key, void *value, void *ignored)
2459 downloaded_files_free (void)
2461 if (downloaded_files_hash)
2463 hash_table_map (downloaded_files_hash, df_free_mapper, NULL);
2464 hash_table_destroy (downloaded_files_hash);
2465 downloaded_files_hash = NULL;
2470 /* Debugging and testing support for path_simplify. */
2472 /* Debug: run path_simplify on PATH and return the result in a new
2473 string. Useful for calling from the debugger. */
2477 char *copy = xstrdup (path);
2478 path_simplify (copy);
2483 run_test (char *test, char *expected_result, int expected_change)
2485 char *test_copy = xstrdup (test);
2486 int modified = path_simplify (test_copy);
2488 if (0 != strcmp (test_copy, expected_result))
2490 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2491 test, expected_result, test_copy);
2493 if (modified != expected_change)
2495 if (expected_change == 1)
2496 printf ("Expected no modification with path_simplify(\"%s\").\n",
2499 printf ("Expected modification with path_simplify(\"%s\").\n",
2506 test_path_simplify (void)
2509 char *test, *result;
2515 { "foo", "foo", 0 },
2516 { "foo/bar", "foo/bar", 0 },
2517 { "foo///bar", "foo/bar", 1 },
2518 { "foo/.", "foo/", 1 },
2519 { "foo/./", "foo/", 1 },
2520 { "foo./", "foo./", 0 },
2521 { "foo/../bar", "bar", 1 },
2522 { "foo/../bar/", "bar/", 1 },
2523 { "foo/bar/..", "foo/", 1 },
2524 { "foo/bar/../x", "foo/x", 1 },
2525 { "foo/bar/../x/", "foo/x/", 1 },
2526 { "foo/..", "", 1 },
2527 { "foo/../..", "", 1 },
2528 { "a/b/../../c", "c", 1 },
2529 { "./a/../b", "b", 1 }
2533 for (i = 0; i < ARRAY_SIZE (tests); i++)
2535 char *test = tests[i].test;
2536 char *expected_result = tests[i].result;
2537 int expected_change = tests[i].should_modify;
2538 run_test (test, expected_result, expected_change);
2541 /* Now run all the tests with a leading slash before the test case,
2542 to prove that the slash is being preserved. */
2543 for (i = 0; i < ARRAY_SIZE (tests); i++)
2545 char *test, *expected_result;
2546 int expected_change = tests[i].should_modify;
2548 test = xmalloc (1 + strlen (tests[i].test) + 1);
2549 sprintf (test, "/%s", tests[i].test);
2551 expected_result = xmalloc (1 + strlen (tests[i].result) + 1);
2552 sprintf (expected_result, "/%s", tests[i].result);
2554 run_test (test, expected_result, expected_change);
2557 xfree (expected_result);