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, U, U, U, 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);
534 /* Used by main.c: detect URLs written using the "shorthand" URL forms
535 popularized by Netscape and NcFTP. HTTP shorthands look like this:
537 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
538 www.foo.com[:port] -> http://www.foo.com[:port]
540 FTP shorthands look like this:
542 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
543 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
545 If the URL needs not or cannot be rewritten, return NULL. */
547 rewrite_shorthand_url (const char *url)
551 if (url_has_scheme (url))
554 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
556 for (p = url; *p && *p != ':' && *p != '/'; p++)
566 /* If the characters after the colon and before the next slash
567 or end of string are all digits, it's HTTP. */
569 for (pp = p + 1; ISDIGIT (*pp); pp++)
571 if (digits > 0 && (*pp == '/' || *pp == '\0'))
574 /* Prepend "ftp://" to the entire URL... */
575 res = xmalloc (6 + strlen (url) + 1);
576 sprintf (res, "ftp://%s", url);
577 /* ...and replace ':' with '/'. */
578 res[6 + (p - url)] = '/';
585 /* Just prepend "http://" to what we have. */
586 res = xmalloc (7 + strlen (url) + 1);
587 sprintf (res, "http://%s", url);
592 static void parse_path PARAMS ((const char *, char **, char **));
595 strpbrk_or_eos (const char *s, const char *accept)
597 char *p = strpbrk (s, accept);
599 p = (char *)s + strlen (s);
603 /* Turn STR into lowercase; return non-zero if a character was
607 lowercase_str (char *str)
614 *str = TOLOWER (*str);
619 static char *parse_errors[] = {
620 #define PE_NO_ERROR 0
622 #define PE_UNSUPPORTED_SCHEME 1
623 "Unsupported scheme",
624 #define PE_EMPTY_HOST 2
626 #define PE_BAD_PORT_NUMBER 3
628 #define PE_INVALID_USER_NAME 4
632 #define SETERR(p, v) do { \
639 Return a new struct url if successful, NULL on error. In case of
640 error, and if ERROR is not NULL, also set *ERROR to the appropriate
643 url_parse (const char *url, int *error)
647 int path_modified, host_modified;
649 enum url_scheme scheme;
651 const char *uname_b, *uname_e;
652 const char *host_b, *host_e;
653 const char *path_b, *path_e;
654 const char *params_b, *params_e;
655 const char *query_b, *query_e;
656 const char *fragment_b, *fragment_e;
659 char *user = NULL, *passwd = NULL;
663 scheme = url_scheme (url);
664 if (scheme == SCHEME_INVALID)
666 SETERR (error, PE_UNSUPPORTED_SCHEME);
670 url_encoded = reencode_string (url);
673 p += strlen (supported_schemes[scheme].leading_string);
675 p += url_skip_uname (p);
678 /* scheme://user:pass@host[:port]... */
681 /* We attempt to break down the URL into the components path,
682 params, query, and fragment. They are ordered like this:
684 scheme://host[:port][/path][;params][?query][#fragment] */
686 params_b = params_e = NULL;
687 query_b = query_e = NULL;
688 fragment_b = fragment_e = NULL;
691 p = strpbrk_or_eos (p, ":/;?#");
694 if (host_b == host_e)
696 SETERR (error, PE_EMPTY_HOST);
700 port = scheme_default_port (scheme);
703 const char *port_b, *port_e, *pp;
705 /* scheme://host:port/tralala */
709 p = strpbrk_or_eos (p, "/;?#");
712 if (port_b == port_e)
714 /* http://host:/whatever */
716 SETERR (error, PE_BAD_PORT_NUMBER);
720 for (port = 0, pp = port_b; pp < port_e; pp++)
724 /* http://host:12randomgarbage/blah */
726 SETERR (error, PE_BAD_PORT_NUMBER);
729 port = 10 * port + (*pp - '0');
737 p = strpbrk_or_eos (p, ";?#");
742 /* Path is not allowed not to exist. */
750 p = strpbrk_or_eos (p, "?#");
757 p = strpbrk_or_eos (p, "#");
769 if (uname_b != uname_e)
771 /* http://user:pass@host */
773 /* uname_b uname_e */
774 if (!parse_uname (uname_b, uname_e - uname_b - 1, &user, &passwd))
776 SETERR (error, PE_INVALID_USER_NAME);
781 u = (struct url *)xmalloc (sizeof (struct url));
782 memset (u, 0, sizeof (*u));
785 u->host = strdupdelim (host_b, host_e);
790 u->path = strdupdelim (path_b, path_e);
791 path_modified = path_simplify (u->path);
792 parse_path (u->path, &u->dir, &u->file);
794 host_modified = lowercase_str (u->host);
797 u->params = strdupdelim (params_b, params_e);
799 u->query = strdupdelim (query_b, query_e);
801 u->fragment = strdupdelim (fragment_b, fragment_e);
803 if (path_modified || u->fragment || host_modified || path_b == path_e)
805 /* If we suspect that a transformation has rendered what
806 url_string might return different from URL_ENCODED, rebuild
807 u->url using url_string. */
808 u->url = url_string (u, 0);
810 if (url_encoded != url)
811 xfree ((char *) url_encoded);
815 if (url_encoded == url)
816 u->url = xstrdup (url);
818 u->url = url_encoded;
826 url_error (int error_code)
828 assert (error_code >= 0 && error_code < ARRAY_SIZE (parse_errors));
829 return parse_errors[error_code];
833 parse_path (const char *quoted_path, char **dir, char **file)
835 char *path, *last_slash;
837 STRDUP_ALLOCA (path, quoted_path);
838 decode_string (path);
840 last_slash = strrchr (path, '/');
844 *file = xstrdup (path);
848 *dir = strdupdelim (path, last_slash);
849 *file = xstrdup (last_slash + 1);
853 /* Note: URL's "full path" is the path with the query string and
854 params appended. The "fragment" (#foo) is intentionally ignored,
855 but that might be changed. For example, if the original URL was
856 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
857 the full path will be "/foo/bar/baz;bullshit?querystring". */
859 /* Return the length of the full path, without the terminating
863 full_path_length (const struct url *url)
867 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
878 /* Write out the full path. */
881 full_path_write (const struct url *url, char *where)
883 #define FROB(el, chr) do { \
884 char *f_el = url->el; \
886 int l = strlen (f_el); \
888 memcpy (where, f_el, l); \
900 /* Public function for getting the "full path". E.g. if u->path is
901 "foo/bar" and u->query is "param=value", full_path will be
902 "/foo/bar?param=value". */
905 url_full_path (const struct url *url)
907 int length = full_path_length (url);
908 char *full_path = (char *)xmalloc(length + 1);
910 full_path_write (url, full_path);
911 full_path[length] = '\0';
916 /* Sync u->path and u->url with u->dir and u->file. */
919 sync_path (struct url *url)
927 newpath = xstrdup (url->file);
932 int dirlen = strlen (url->dir);
933 int filelen = strlen (url->file);
935 newpath = xmalloc (dirlen + 1 + filelen + 1);
936 memcpy (newpath, url->dir, dirlen);
937 newpath[dirlen] = '/';
938 memcpy (newpath + dirlen + 1, url->file, filelen);
939 newpath[dirlen + 1 + filelen] = '\0';
945 /* Synchronize u->url. */
947 url->url = url_string (url, 0);
950 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
951 This way we can sync u->path and u->url when they get changed. */
954 url_set_dir (struct url *url, const char *newdir)
957 url->dir = xstrdup (newdir);
962 url_set_file (struct url *url, const char *newfile)
965 url->file = xstrdup (newfile);
970 url_free (struct url *url)
976 FREE_MAYBE (url->params);
977 FREE_MAYBE (url->query);
978 FREE_MAYBE (url->fragment);
979 FREE_MAYBE (url->user);
980 FREE_MAYBE (url->passwd);
989 get_urls_file (const char *file)
991 struct file_memory *fm;
992 struct urlpos *head, *tail;
993 const char *text, *text_end;
996 fm = read_file (file);
999 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
1002 DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
1006 text_end = fm->content + fm->length;
1007 while (text < text_end)
1009 const char *line_beg = text;
1010 const char *line_end = memchr (text, '\n', text_end - text);
1012 line_end = text_end;
1017 /* Strip whitespace from the beginning and end of line. */
1018 while (line_beg < line_end && ISSPACE (*line_beg))
1020 while (line_end > line_beg && ISSPACE (*(line_end - 1)))
1023 if (line_end > line_beg)
1025 /* URL is in the [line_beg, line_end) region. */
1029 struct urlpos *entry;
1032 /* We must copy the URL to a zero-terminated string, and we
1033 can't use alloca because we're in a loop. *sigh*. */
1034 url_text = strdupdelim (line_beg, line_end);
1038 /* Merge opt.base_href with URL. */
1039 char *merged = uri_merge (opt.base_href, url_text);
1044 url = url_parse (url_text, &up_error_code);
1047 logprintf (LOG_NOTQUIET, "%s: Invalid URL %s: %s\n",
1048 file, url_text, url_error (up_error_code));
1054 entry = (struct urlpos *)xmalloc (sizeof (struct urlpos));
1055 memset (entry, 0, sizeof (*entry));
1066 read_file_free (fm);
1070 /* Free the linked list of urlpos. */
1072 free_urlpos (struct urlpos *l)
1076 struct urlpos *next = l->next;
1079 FREE_MAYBE (l->local_name);
1085 /* Rotate FNAME opt.backups times */
1087 rotate_backups(const char *fname)
1089 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
1090 char *from = (char *)alloca (maxlen);
1091 char *to = (char *)alloca (maxlen);
1095 if (stat (fname, &sb) == 0)
1096 if (S_ISREG (sb.st_mode) == 0)
1099 for (i = opt.backups; i > 1; i--)
1101 sprintf (from, "%s.%d", fname, i - 1);
1102 sprintf (to, "%s.%d", fname, i);
1103 /* #### This will fail on machines without the rename() system
1108 sprintf (to, "%s.%d", fname, 1);
1112 /* Create all the necessary directories for PATH (a file). Calls
1113 mkdirhier() internally. */
1115 mkalldirs (const char *path)
1122 p = path + strlen (path);
1123 for (; *p != '/' && p != path; p--);
1124 /* Don't create if it's just a file. */
1125 if ((p == path) && (*p != '/'))
1127 t = strdupdelim (path, p);
1128 /* Check whether the directory exists. */
1129 if ((stat (t, &st) == 0))
1131 if (S_ISDIR (st.st_mode))
1138 /* If the dir exists as a file name, remove it first. This
1139 is *only* for Wget to work with buggy old CERN http
1140 servers. Here is the scenario: When Wget tries to
1141 retrieve a directory without a slash, e.g.
1142 http://foo/bar (bar being a directory), CERN server will
1143 not redirect it too http://foo/bar/ -- it will generate a
1144 directory listing containing links to bar/file1,
1145 bar/file2, etc. Wget will lose because it saves this
1146 HTML listing to a file `bar', so it cannot create the
1147 directory. To work around this, if the file of the same
1148 name exists, we just remove it and create the directory
1150 DEBUGP (("Removing %s because of directory danger!\n", t));
1154 res = make_directory (t);
1156 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1162 count_slashes (const char *s)
1171 /* Return the path name of the URL-equivalent file name, with a
1172 remote-like structure of directories. */
1174 mkstruct (const struct url *u)
1176 char *dir, *dir_preencoding;
1177 char *file, *res, *dirpref;
1178 char *query = u->query && *u->query ? u->query : NULL;
1183 char *ptr = u->dir + (*u->dir == '/');
1184 int slash_count = 1 + count_slashes (ptr);
1185 int cut = MINVAL (opt.cut_dirs, slash_count);
1186 for (; cut && *ptr; ptr++)
1189 STRDUP_ALLOCA (dir, ptr);
1192 dir = u->dir + (*u->dir == '/');
1194 /* Check for the true name (or at least a consistent name for saving
1195 to directory) of HOST, reusing the hlist if possible. */
1196 if (opt.add_hostdir)
1198 /* Add dir_prefix and hostname (if required) to the beginning of
1200 dirpref = (char *)alloca (strlen (opt.dir_prefix) + 1
1202 + 1 + numdigit (u->port)
1204 if (!DOTP (opt.dir_prefix))
1205 sprintf (dirpref, "%s/%s", opt.dir_prefix, u->host);
1207 strcpy (dirpref, u->host);
1209 if (u->port != scheme_default_port (u->scheme))
1211 int len = strlen (dirpref);
1213 number_to_string (dirpref + len + 1, u->port);
1216 else /* not add_hostdir */
1218 if (!DOTP (opt.dir_prefix))
1219 dirpref = opt.dir_prefix;
1224 /* If there is a prefix, prepend it. */
1227 char *newdir = (char *)alloca (strlen (dirpref) + 1 + strlen (dir) + 2);
1228 sprintf (newdir, "%s%s%s", dirpref, *dir == '/' ? "" : "/", dir);
1232 dir_preencoding = dir;
1233 dir = reencode_string (dir_preencoding);
1236 if (l && dir[l - 1] == '/')
1240 file = "index.html";
1244 /* Finally, construct the full name. */
1245 res = (char *)xmalloc (strlen (dir) + 1 + strlen (file)
1246 + (query ? (1 + strlen (query)) : 0)
1248 sprintf (res, "%s%s%s", dir, *dir ? "/" : "", file);
1252 strcat (res, query);
1254 if (dir != dir_preencoding)
1259 /* Compose a file name out of BASE, an unescaped file name, and QUERY,
1260 an escaped query string. The trick is to make sure that unsafe
1261 characters in BASE are escaped, and that slashes in QUERY are also
1265 compose_file_name (char *base, char *query)
1271 /* Copy BASE to RESULT and encode all unsafe characters. */
1273 while (*from && to - result < sizeof (result))
1275 if (UNSAFE_CHAR (*from))
1277 unsigned char c = *from++;
1279 *to++ = XDIGIT_TO_XCHAR (c >> 4);
1280 *to++ = XDIGIT_TO_XCHAR (c & 0xf);
1286 if (query && to - result < sizeof (result))
1290 /* Copy QUERY to RESULT and encode all '/' characters. */
1292 while (*from && to - result < sizeof (result))
1306 if (to - result < sizeof (result))
1309 /* Truncate input which is too long, presumably due to a huge
1311 result[sizeof (result) - 1] = '\0';
1313 return xstrdup (result);
1316 /* Create a unique filename, corresponding to a given URL. Calls
1317 mkstruct if necessary. Does *not* actually create any directories. */
1319 url_filename (const struct url *u)
1322 int have_prefix = 0; /* whether we must prepend opt.dir_prefix */
1326 file = mkstruct (u);
1331 char *base = *u->file ? u->file : "index.html";
1332 char *query = u->query && *u->query ? u->query : NULL;
1333 file = compose_file_name (base, query);
1338 /* Check whether the prefix directory is something other than "."
1339 before prepending it. */
1340 if (!DOTP (opt.dir_prefix))
1342 char *nfile = (char *)xmalloc (strlen (opt.dir_prefix)
1343 + 1 + strlen (file) + 1);
1344 sprintf (nfile, "%s/%s", opt.dir_prefix, file);
1349 /* DOS-ish file systems don't like `%' signs in them; we change it
1354 for (p = file; *p; p++)
1358 #endif /* WINDOWS */
1360 /* Check the cases in which the unique extensions are not used:
1361 1) Clobbering is turned off (-nc).
1362 2) Retrieval with regetting.
1363 3) Timestamping is used.
1364 4) Hierarchy is built.
1366 The exception is the case when file does exist and is a
1367 directory (actually support for bad httpd-s). */
1368 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1369 && !(file_exists_p (file) && !file_non_directory_p (file)))
1372 /* Find a unique name. */
1373 name = unique_name (file);
1378 /* Return the langth of URL's path. Path is considered to be
1379 terminated by one of '?', ';', '#', or by the end of the
1382 path_length (const char *url)
1384 const char *q = strpbrk_or_eos (url, "?;#");
1388 /* Find the last occurrence of character C in the range [b, e), or
1389 NULL, if none are present. This is equivalent to strrchr(b, c),
1390 except that it accepts an END argument instead of requiring the
1391 string to be zero-terminated. Why is there no memrchr()? */
1393 find_last_char (const char *b, const char *e, char c)
1401 /* Resolve "." and ".." elements of PATH by destructively modifying
1402 PATH. "." is resolved by removing that path element, and ".." is
1403 resolved by removing the preceding path element. Leading and
1404 trailing slashes are preserved.
1406 Return non-zero if any changes have been made.
1408 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1409 test examples are provided below. If you change anything in this
1410 function, run test_path_simplify to make sure you haven't broken a
1413 A previous version of this function was based on path_simplify()
1414 from GNU Bash, but it has been rewritten for Wget 1.8.1. */
1417 path_simplify (char *path)
1423 ++path; /* preserve the leading '/'. */
1426 end = p + strlen (p) + 1; /* position past the terminating zero. */
1431 /* P should point to the beginning of a path element. */
1433 if (*p == '.' && (*(p + 1) == '/' || *(p + 1) == '\0'))
1435 /* Handle "./foo" by moving "foo" two characters to the
1437 if (*(p + 1) == '/')
1440 memmove (p, p + 2, end - p);
1451 else if (*p == '.' && *(p + 1) == '.'
1452 && (*(p + 2) == '/' || *(p + 2) == '\0'))
1454 /* Handle "../foo" by moving "foo" one path element to the
1456 char *b = p; /* not p-1 because P can equal PATH */
1458 /* Backtrack by one path element, but not past the beginning
1461 /* foo/bar/../baz */
1467 /* Move backwards until B hits the beginning of the
1468 previous path element or the beginning of path. */
1469 for (--b; b > path && *(b - 1) != '/'; b--)
1474 if (*(p + 2) == '/')
1476 memmove (b, p + 3, end - (p + 3));
1490 /* Remove empty path elements. Not mandated by rfc1808 et
1491 al, but empty path elements are not all that useful, and
1492 the rest of Wget might not deal with them well. */
1502 memmove (p, q, end - q);
1507 /* Skip to the next path element. */
1508 while (*p && *p != '/')
1513 /* Make sure P points to the beginning of the next path element,
1514 which is location after the slash. */
1521 /* Resolve the result of "linking" a base URI (BASE) to a
1522 link-specified URI (LINK).
1524 Either of the URIs may be absolute or relative, complete with the
1525 host name, or path only. This tries to behave "reasonably" in all
1526 foreseeable cases. It employs little specific knowledge about
1527 schemes or URL-specific stuff -- it just works on strings.
1529 The parameters LINKLENGTH is useful if LINK is not zero-terminated.
1530 See uri_merge for a gentler interface to this functionality.
1532 Perhaps this function should call path_simplify so that the callers
1533 don't have to call url_parse unconditionally. */
1535 uri_merge_1 (const char *base, const char *link, int linklength, int no_scheme)
1541 const char *end = base + path_length (base);
1545 /* Empty LINK points back to BASE, query string and all. */
1546 constr = xstrdup (base);
1548 else if (*link == '?')
1550 /* LINK points to the same location, but changes the query
1551 string. Examples: */
1552 /* uri_merge("path", "?new") -> "path?new" */
1553 /* uri_merge("path?foo", "?new") -> "path?new" */
1554 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1555 /* uri_merge("path#foo", "?new") -> "path?new" */
1556 int baselength = end - base;
1557 constr = xmalloc (baselength + linklength + 1);
1558 memcpy (constr, base, baselength);
1559 memcpy (constr + baselength, link, linklength);
1560 constr[baselength + linklength] = '\0';
1562 else if (*link == '#')
1564 /* uri_merge("path", "#new") -> "path#new" */
1565 /* uri_merge("path#foo", "#new") -> "path#new" */
1566 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1567 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1569 const char *end1 = strchr (base, '#');
1571 end1 = base + strlen (base);
1572 baselength = end1 - base;
1573 constr = xmalloc (baselength + linklength + 1);
1574 memcpy (constr, base, baselength);
1575 memcpy (constr + baselength, link, linklength);
1576 constr[baselength + linklength] = '\0';
1578 else if (*link == '/')
1580 /* LINK is an absolute path: we need to replace everything
1581 after (and including) the FIRST slash with LINK.
1583 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1584 "/qux/xyzzy", our result should be
1585 "http://host/qux/xyzzy". */
1588 const char *start_insert = NULL; /* for gcc to shut up. */
1589 const char *pos = base;
1590 int seen_slash_slash = 0;
1591 /* We're looking for the first slash, but want to ignore
1594 slash = memchr (pos, '/', end - pos);
1595 if (slash && !seen_slash_slash)
1596 if (*(slash + 1) == '/')
1599 seen_slash_slash = 1;
1603 /* At this point, SLASH is the location of the first / after
1604 "//", or the first slash altogether. START_INSERT is the
1605 pointer to the location where LINK will be inserted. When
1606 examining the last two examples, keep in mind that LINK
1609 if (!slash && !seen_slash_slash)
1610 /* example: "foo" */
1612 start_insert = base;
1613 else if (!slash && seen_slash_slash)
1614 /* example: "http://foo" */
1617 else if (slash && !seen_slash_slash)
1618 /* example: "foo/bar" */
1620 start_insert = base;
1621 else if (slash && seen_slash_slash)
1622 /* example: "http://something/" */
1624 start_insert = slash;
1626 span = start_insert - base;
1627 constr = (char *)xmalloc (span + linklength + 1);
1629 memcpy (constr, base, span);
1631 memcpy (constr + span, link, linklength);
1632 constr[span + linklength] = '\0';
1636 /* LINK is a relative URL: we need to replace everything
1637 after last slash (possibly empty) with LINK.
1639 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1640 our result should be "whatever/foo/qux/xyzzy". */
1641 int need_explicit_slash = 0;
1643 const char *start_insert;
1644 const char *last_slash = find_last_char (base, end, '/');
1647 /* No slash found at all. Append LINK to what we have,
1648 but we'll need a slash as a separator.
1650 Example: if base == "foo" and link == "qux/xyzzy", then
1651 we cannot just append link to base, because we'd get
1652 "fooqux/xyzzy", whereas what we want is
1655 To make sure the / gets inserted, we set
1656 need_explicit_slash to 1. We also set start_insert
1657 to end + 1, so that the length calculations work out
1658 correctly for one more (slash) character. Accessing
1659 that character is fine, since it will be the
1660 delimiter, '\0' or '?'. */
1661 /* example: "foo?..." */
1662 /* ^ ('?' gets changed to '/') */
1663 start_insert = end + 1;
1664 need_explicit_slash = 1;
1666 else if (last_slash && last_slash != base && *(last_slash - 1) == '/')
1668 /* example: http://host" */
1670 start_insert = end + 1;
1671 need_explicit_slash = 1;
1675 /* example: "whatever/foo/bar" */
1677 start_insert = last_slash + 1;
1680 span = start_insert - base;
1681 constr = (char *)xmalloc (span + linklength + 1);
1683 memcpy (constr, base, span);
1684 if (need_explicit_slash)
1685 constr[span - 1] = '/';
1687 memcpy (constr + span, link, linklength);
1688 constr[span + linklength] = '\0';
1691 else /* !no_scheme */
1693 constr = strdupdelim (link, link + linklength);
1698 /* Merge BASE with LINK and return the resulting URI. This is an
1699 interface to uri_merge_1 that assumes that LINK is a
1700 zero-terminated string. */
1702 uri_merge (const char *base, const char *link)
1704 return uri_merge_1 (base, link, strlen (link), !url_has_scheme (link));
1707 #define APPEND(p, s) do { \
1708 int len = strlen (s); \
1709 memcpy (p, s, len); \
1713 /* Use this instead of password when the actual password is supposed
1714 to be hidden. We intentionally use a generic string without giving
1715 away the number of characters in the password, like previous
1717 #define HIDDEN_PASSWORD "*password*"
1719 /* Recreate the URL string from the data in URL.
1721 If HIDE is non-zero (as it is when we're calling this on a URL we
1722 plan to print, but not when calling it to canonicalize a URL for
1723 use within the program), password will be hidden. Unsafe
1724 characters in the URL will be quoted. */
1727 url_string (const struct url *url, int hide_password)
1731 char *quoted_user = NULL, *quoted_passwd = NULL;
1733 int scheme_port = supported_schemes[url->scheme].default_port;
1734 char *scheme_str = supported_schemes[url->scheme].leading_string;
1735 int fplen = full_path_length (url);
1737 assert (scheme_str != NULL);
1739 /* Make sure the user name and password are quoted. */
1742 quoted_user = encode_string_maybe (url->user);
1746 quoted_passwd = HIDDEN_PASSWORD;
1748 quoted_passwd = encode_string_maybe (url->passwd);
1752 size = (strlen (scheme_str)
1753 + strlen (url->host)
1756 if (url->port != scheme_port)
1757 size += 1 + numdigit (url->port);
1760 size += 1 + strlen (quoted_user);
1762 size += 1 + strlen (quoted_passwd);
1765 p = result = xmalloc (size);
1767 APPEND (p, scheme_str);
1770 APPEND (p, quoted_user);
1774 APPEND (p, quoted_passwd);
1779 APPEND (p, url->host);
1780 if (url->port != scheme_port)
1783 p = number_to_string (p, url->port);
1786 full_path_write (url, p);
1790 assert (p - result == size);
1792 if (quoted_user && quoted_user != url->user)
1793 xfree (quoted_user);
1794 if (quoted_passwd && !hide_password
1795 && quoted_passwd != url->passwd)
1796 xfree (quoted_passwd);
1801 /* Returns proxy host address, in accordance with SCHEME. */
1803 getproxy (enum url_scheme scheme)
1806 char *rewritten_url;
1807 static char rewritten_storage[1024];
1812 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
1816 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
1820 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
1822 case SCHEME_INVALID:
1825 if (!proxy || !*proxy)
1828 /* Handle shorthands. */
1829 rewritten_url = rewrite_shorthand_url (proxy);
1832 strncpy (rewritten_storage, rewritten_url, sizeof(rewritten_storage));
1833 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
1834 proxy = rewritten_storage;
1840 /* Should a host be accessed through proxy, concerning no_proxy? */
1842 no_proxy_match (const char *host, const char **no_proxy)
1847 return !sufmatch (no_proxy, host);
1850 /* Support for converting links for local viewing in downloaded HTML
1851 files. This should be moved to another file, because it has
1852 nothing to do with processing URLs. */
1854 static void write_backup_file PARAMS ((const char *, downloaded_file_t));
1855 static const char *replace_attr PARAMS ((const char *, int, FILE *,
1857 static const char *replace_attr_refresh_hack PARAMS ((const char *, int, FILE *,
1858 const char *, int));
1859 static char *local_quote_string PARAMS ((const char *));
1861 /* Change the links in one HTML file. LINKS is a list of links in the
1862 document, along with their positions and the desired direction of
1865 convert_links (const char *file, struct urlpos *links)
1867 struct file_memory *fm;
1870 downloaded_file_t downloaded_file_return;
1872 struct urlpos *link;
1873 int to_url_count = 0, to_file_count = 0;
1875 logprintf (LOG_VERBOSE, _("Converting %s... "), file);
1878 /* First we do a "dry run": go through the list L and see whether
1879 any URL needs to be converted in the first place. If not, just
1880 leave the file alone. */
1882 struct urlpos *dry = links;
1883 for (dry = links; dry; dry = dry->next)
1884 if (dry->convert != CO_NOCONVERT)
1888 logputs (LOG_VERBOSE, _("nothing to do.\n"));
1893 fm = read_file (file);
1896 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1897 file, strerror (errno));
1901 downloaded_file_return = downloaded_file (CHECK_FOR_FILE, file);
1902 if (opt.backup_converted && downloaded_file_return)
1903 write_backup_file (file, downloaded_file_return);
1905 /* Before opening the file for writing, unlink the file. This is
1906 important if the data in FM is mmaped. In such case, nulling the
1907 file, which is what fopen() below does, would make us read all
1908 zeroes from the mmaped region. */
1909 if (unlink (file) < 0 && errno != ENOENT)
1911 logprintf (LOG_NOTQUIET, _("Unable to delete `%s': %s\n"),
1912 file, strerror (errno));
1913 read_file_free (fm);
1916 /* Now open the file for writing. */
1917 fp = fopen (file, "wb");
1920 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1921 file, strerror (errno));
1922 read_file_free (fm);
1926 /* Here we loop through all the URLs in file, replacing those of
1927 them that are downloaded with relative references. */
1929 for (link = links; link; link = link->next)
1931 char *url_start = fm->content + link->pos;
1933 if (link->pos >= fm->length)
1935 DEBUGP (("Something strange is going on. Please investigate."));
1938 /* If the URL is not to be converted, skip it. */
1939 if (link->convert == CO_NOCONVERT)
1941 DEBUGP (("Skipping %s at position %d.\n", link->url->url, link->pos));
1945 /* Echo the file contents, up to the offending URL's opening
1946 quote, to the outfile. */
1947 fwrite (p, 1, url_start - p, fp);
1950 switch (link->convert)
1952 case CO_CONVERT_TO_RELATIVE:
1953 /* Convert absolute URL to relative. */
1955 char *newname = construct_relative (file, link->local_name);
1956 char *quoted_newname = local_quote_string (newname);
1958 if (!link->link_refresh_p)
1959 p = replace_attr (p, link->size, fp, quoted_newname);
1961 p = replace_attr_refresh_hack (p, link->size, fp, quoted_newname,
1962 link->refresh_timeout);
1964 DEBUGP (("TO_RELATIVE: %s to %s at position %d in %s.\n",
1965 link->url->url, newname, link->pos, file));
1967 xfree (quoted_newname);
1971 case CO_CONVERT_TO_COMPLETE:
1972 /* Convert the link to absolute URL. */
1974 char *newlink = link->url->url;
1975 char *quoted_newlink = html_quote_string (newlink);
1977 if (!link->link_refresh_p)
1978 p = replace_attr (p, link->size, fp, quoted_newlink);
1980 p = replace_attr_refresh_hack (p, link->size, fp, quoted_newlink,
1981 link->refresh_timeout);
1983 DEBUGP (("TO_COMPLETE: <something> to %s at position %d in %s.\n",
1984 newlink, link->pos, file));
1985 xfree (quoted_newlink);
1989 case CO_NULLIFY_BASE:
1990 /* Change the base href to "". */
1991 p = replace_attr (p, link->size, fp, "");
1999 /* Output the rest of the file. */
2000 if (p - fm->content < fm->length)
2001 fwrite (p, 1, fm->length - (p - fm->content), fp);
2003 read_file_free (fm);
2005 logprintf (LOG_VERBOSE, "%d-%d\n", to_file_count, to_url_count);
2008 /* Construct and return a malloced copy of the relative link from two
2009 pieces of information: local name S1 of the referring file and
2010 local name S2 of the referred file.
2012 So, if S1 is "jagor.srce.hr/index.html" and S2 is
2013 "jagor.srce.hr/images/news.gif", the function will return
2016 Alternately, if S1 is "fly.cc.fer.hr/ioccc/index.html", and S2 is
2017 "fly.cc.fer.hr/images/fly.gif", the function will return
2018 "../images/fly.gif".
2020 Caveats: S1 should not begin with `/', unless S2 also begins with
2021 '/'. S1 should not contain things like ".." and such --
2022 construct_relative ("fly/ioccc/../index.html",
2023 "fly/images/fly.gif") will fail. (A workaround is to call
2024 something like path_simplify() on S1). */
2026 construct_relative (const char *s1, const char *s2)
2028 int i, cnt, sepdirs1;
2032 return xstrdup (s2);
2033 /* S1 should *not* be absolute, if S2 wasn't. */
2034 assert (*s1 != '/');
2036 /* Skip the directories common to both strings. */
2039 while (s1[i] && s2[i]
2044 if (s1[i] == '/' && s2[i] == '/')
2049 for (sepdirs1 = 0; s1[i]; i++)
2052 /* Now, construct the file as of:
2053 - ../ repeated sepdirs1 time
2054 - all the non-mutual directories of S2. */
2055 res = (char *)xmalloc (3 * sepdirs1 + strlen (s2 + cnt) + 1);
2056 for (i = 0; i < sepdirs1; i++)
2057 memcpy (res + 3 * i, "../", 3);
2058 strcpy (res + 3 * i, s2 + cnt);
2063 write_backup_file (const char *file, downloaded_file_t downloaded_file_return)
2065 /* Rather than just writing over the original .html file with the
2066 converted version, save the former to *.orig. Note we only do
2067 this for files we've _successfully_ downloaded, so we don't
2068 clobber .orig files sitting around from previous invocations. */
2070 /* Construct the backup filename as the original name plus ".orig". */
2071 size_t filename_len = strlen(file);
2072 char* filename_plus_orig_suffix;
2073 boolean already_wrote_backup_file = FALSE;
2074 slist* converted_file_ptr;
2075 static slist* converted_files = NULL;
2077 if (downloaded_file_return == FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED)
2079 /* Just write "orig" over "html". We need to do it this way
2080 because when we're checking to see if we've downloaded the
2081 file before (to see if we can skip downloading it), we don't
2082 know if it's a text/html file. Therefore we don't know yet
2083 at that stage that -E is going to cause us to tack on
2084 ".html", so we need to compare vs. the original URL plus
2085 ".orig", not the original URL plus ".html.orig". */
2086 filename_plus_orig_suffix = alloca (filename_len + 1);
2087 strcpy(filename_plus_orig_suffix, file);
2088 strcpy((filename_plus_orig_suffix + filename_len) - 4, "orig");
2090 else /* downloaded_file_return == FILE_DOWNLOADED_NORMALLY */
2092 /* Append ".orig" to the name. */
2093 filename_plus_orig_suffix = alloca (filename_len + sizeof(".orig"));
2094 strcpy(filename_plus_orig_suffix, file);
2095 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
2098 /* We can get called twice on the same URL thanks to the
2099 convert_all_links() call in main(). If we write the .orig file
2100 each time in such a case, it'll end up containing the first-pass
2101 conversion, not the original file. So, see if we've already been
2102 called on this file. */
2103 converted_file_ptr = converted_files;
2104 while (converted_file_ptr != NULL)
2105 if (strcmp(converted_file_ptr->string, file) == 0)
2107 already_wrote_backup_file = TRUE;
2111 converted_file_ptr = converted_file_ptr->next;
2113 if (!already_wrote_backup_file)
2115 /* Rename <file> to <file>.orig before former gets written over. */
2116 if (rename(file, filename_plus_orig_suffix) != 0)
2117 logprintf (LOG_NOTQUIET, _("Cannot back up %s as %s: %s\n"),
2118 file, filename_plus_orig_suffix, strerror (errno));
2120 /* Remember that we've already written a .orig backup for this file.
2121 Note that we never free this memory since we need it till the
2122 convert_all_links() call, which is one of the last things the
2123 program does before terminating. BTW, I'm not sure if it would be
2124 safe to just set 'converted_file_ptr->string' to 'file' below,
2125 rather than making a copy of the string... Another note is that I
2126 thought I could just add a field to the urlpos structure saying
2127 that we'd written a .orig file for this URL, but that didn't work,
2128 so I had to make this separate list.
2129 -- Dan Harkless <wget@harkless.org>
2131 This [adding a field to the urlpos structure] didn't work
2132 because convert_file() is called from convert_all_links at
2133 the end of the retrieval with a freshly built new urlpos
2135 -- Hrvoje Niksic <hniksic@arsdigita.com>
2137 converted_file_ptr = xmalloc(sizeof(*converted_file_ptr));
2138 converted_file_ptr->string = xstrdup(file); /* die on out-of-mem. */
2139 converted_file_ptr->next = converted_files;
2140 converted_files = converted_file_ptr;
2144 static int find_fragment PARAMS ((const char *, int, const char **,
2147 /* Replace an attribute's original text with NEW_TEXT. */
2150 replace_attr (const char *p, int size, FILE *fp, const char *new_text)
2153 char quote_char = '\"'; /* use "..." for quoting, unless the
2154 original value is quoted, in which
2155 case reuse its quoting char. */
2156 const char *frag_beg, *frag_end;
2158 /* Structure of our string is:
2159 "...old-contents..."
2160 <--- size ---> (with quotes)
2163 <--- size --> (no quotes) */
2165 if (*p == '\"' || *p == '\'')
2170 size -= 2; /* disregard opening and closing quote */
2172 putc (quote_char, fp);
2173 fputs (new_text, fp);
2175 /* Look for fragment identifier, if any. */
2176 if (find_fragment (p, size, &frag_beg, &frag_end))
2177 fwrite (frag_beg, 1, frag_end - frag_beg, fp);
2181 putc (quote_char, fp);
2186 /* The same as REPLACE_ATTR, but used when replacing
2187 <meta http-equiv=refresh content="new_text"> because we need to
2188 append "timeout_value; URL=" before the next_text. */
2191 replace_attr_refresh_hack (const char *p, int size, FILE *fp,
2192 const char *new_text, int timeout)
2195 char *new_with_timeout = (char *)alloca (numdigit (timeout)
2199 sprintf (new_with_timeout, "%d; URL=%s", timeout, new_text);
2201 return replace_attr (p, size, fp, new_with_timeout);
2204 /* Find the first occurrence of '#' in [BEG, BEG+SIZE) that is not
2205 preceded by '&'. If the character is not found, return zero. If
2206 the character is found, return 1 and set BP and EP to point to the
2207 beginning and end of the region.
2209 This is used for finding the fragment indentifiers in URLs. */
2212 find_fragment (const char *beg, int size, const char **bp, const char **ep)
2214 const char *end = beg + size;
2216 for (; beg < end; beg++)
2238 /* Quote FILE for use as local reference to an HTML file.
2240 We quote ? as %3F to avoid passing part of the file name as the
2241 parameter when browsing the converted file through HTTP. However,
2242 it is safe to do this only when `--html-extension' is turned on.
2243 This is because converting "index.html?foo=bar" to
2244 "index.html%3Ffoo=bar" would break local browsing, as the latter
2245 isn't even recognized as an HTML file! However, converting
2246 "index.html?foo=bar.html" to "index.html%3Ffoo=bar.html" should be
2247 safe for both local and HTTP-served browsing. */
2250 local_quote_string (const char *file)
2252 const char *file_sans_qmark;
2255 if (!opt.html_extension)
2256 return html_quote_string (file);
2258 qm = count_char (file, '?');
2262 const char *from = file;
2265 /* qm * 2 because we replace each question mark with "%3F",
2266 i.e. replace one char with three, hence two more. */
2267 int fsqlen = strlen (file) + qm * 2;
2269 to = newname = (char *)alloca (fsqlen + 1);
2270 for (; *from; from++)
2281 assert (to - newname == fsqlen);
2284 file_sans_qmark = newname;
2287 file_sans_qmark = file;
2289 return html_quote_string (file_sans_qmark);
2292 /* We're storing "modes" of type downloaded_file_t in the hash table.
2293 However, our hash tables only accept pointers for keys and values.
2294 So when we need a pointer, we use the address of a
2295 downloaded_file_t variable of static storage. */
2297 static downloaded_file_t *
2298 downloaded_mode_to_ptr (downloaded_file_t mode)
2300 static downloaded_file_t
2301 v1 = FILE_NOT_ALREADY_DOWNLOADED,
2302 v2 = FILE_DOWNLOADED_NORMALLY,
2303 v3 = FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED,
2304 v4 = CHECK_FOR_FILE;
2308 case FILE_NOT_ALREADY_DOWNLOADED:
2310 case FILE_DOWNLOADED_NORMALLY:
2312 case FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED:
2314 case CHECK_FOR_FILE:
2320 /* This should really be merged with dl_file_url_map and
2321 downloaded_html_files in recur.c. This was originally a list, but
2322 I changed it to a hash table beause it was actually taking a lot of
2323 time to find things in it. */
2325 static struct hash_table *downloaded_files_hash;
2327 /* Remembers which files have been downloaded. In the standard case, should be
2328 called with mode == FILE_DOWNLOADED_NORMALLY for each file we actually
2329 download successfully (i.e. not for ones we have failures on or that we skip
2332 When we've downloaded a file and tacked on a ".html" extension due to -E,
2333 call this function with FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED rather than
2334 FILE_DOWNLOADED_NORMALLY.
2336 If you just want to check if a file has been previously added without adding
2337 it, call with mode == CHECK_FOR_FILE. Please be sure to call this function
2338 with local filenames, not remote URLs. */
2340 downloaded_file (downloaded_file_t mode, const char *file)
2342 downloaded_file_t *ptr;
2344 if (mode == CHECK_FOR_FILE)
2346 if (!downloaded_files_hash)
2347 return FILE_NOT_ALREADY_DOWNLOADED;
2348 ptr = hash_table_get (downloaded_files_hash, file);
2350 return FILE_NOT_ALREADY_DOWNLOADED;
2354 if (!downloaded_files_hash)
2355 downloaded_files_hash = make_string_hash_table (0);
2357 ptr = hash_table_get (downloaded_files_hash, file);
2361 ptr = downloaded_mode_to_ptr (mode);
2362 hash_table_put (downloaded_files_hash, xstrdup (file), &ptr);
2364 return FILE_NOT_ALREADY_DOWNLOADED;
2368 df_free_mapper (void *key, void *value, void *ignored)
2375 downloaded_files_free (void)
2377 if (downloaded_files_hash)
2379 hash_table_map (downloaded_files_hash, df_free_mapper, NULL);
2380 hash_table_destroy (downloaded_files_hash);
2381 downloaded_files_hash = NULL;
2386 /* Debugging and testing support for path_simplify. */
2388 /* Debug: run path_simplify on PATH and return the result in a new
2389 string. Useful for calling from the debugger. */
2393 char *copy = xstrdup (path);
2394 path_simplify (copy);
2399 run_test (char *test, char *expected_result, int expected_change)
2401 char *test_copy = xstrdup (test);
2402 int modified = path_simplify (test_copy);
2404 if (0 != strcmp (test_copy, expected_result))
2406 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2407 test, expected_result, test_copy);
2409 if (modified != expected_change)
2411 if (expected_change == 1)
2412 printf ("Expected no modification with path_simplify(\"%s\").\n",
2415 printf ("Expected modification with path_simplify(\"%s\").\n",
2422 test_path_simplify (void)
2425 char *test, *result;
2431 { "foo", "foo", 0 },
2432 { "foo/bar", "foo/bar", 0 },
2433 { "foo///bar", "foo/bar", 1 },
2434 { "foo/.", "foo/", 1 },
2435 { "foo/./", "foo/", 1 },
2436 { "foo./", "foo./", 0 },
2437 { "foo/../bar", "bar", 1 },
2438 { "foo/../bar/", "bar/", 1 },
2439 { "foo/bar/..", "foo/", 1 },
2440 { "foo/bar/../x", "foo/x", 1 },
2441 { "foo/bar/../x/", "foo/x/", 1 },
2442 { "foo/..", "", 1 },
2443 { "foo/../..", "", 1 },
2444 { "a/b/../../c", "c", 1 },
2445 { "./a/../b", "b", 1 }
2449 for (i = 0; i < ARRAY_SIZE (tests); i++)
2451 char *test = tests[i].test;
2452 char *expected_result = tests[i].result;
2453 int expected_change = tests[i].should_modify;
2454 run_test (test, expected_result, expected_change);
2457 /* Now run all the tests with a leading slash before the test case,
2458 to prove that the slash is being preserved. */
2459 for (i = 0; i < ARRAY_SIZE (tests); i++)
2461 char *test, *expected_result;
2462 int expected_change = tests[i].should_modify;
2464 test = xmalloc (1 + strlen (tests[i].test) + 1);
2465 sprintf (test, "/%s", tests[i].test);
2467 expected_result = xmalloc (1 + strlen (tests[i].result) + 1);
2468 sprintf (expected_result, "/%s", tests[i].result);
2470 run_test (test, expected_result, expected_change);
2473 xfree (expected_result);