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)))
51 static int urlpath_length PARAMS ((const char *));
60 /* Supported schemes: */
61 static struct scheme_data supported_schemes[] =
63 { "http://", DEFAULT_HTTP_PORT, 1 },
65 { "https://", DEFAULT_HTTPS_PORT, 1 },
67 { "ftp://", DEFAULT_FTP_PORT, 1 },
73 static char *construct_relative PARAMS ((const char *, const char *));
76 /* Support for encoding and decoding of URL strings. We determine
77 whether a character is unsafe through static table lookup. This
78 code assumes ASCII character set and 8-bit chars. */
85 #define R urlchr_reserved
86 #define U urlchr_unsafe
89 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
91 /* rfc1738 reserved chars, preserved from encoding. */
93 #define RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
95 /* rfc1738 unsafe chars, plus some more. */
97 #define UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
99 const static unsigned char urlchr_table[256] =
101 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
102 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
103 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
104 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
105 U, 0, U, RU, 0, U, R, 0, /* SP ! " # $ % & ' */
106 0, 0, 0, R, 0, 0, 0, R, /* ( ) * + , - . / */
107 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
108 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
109 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
110 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
111 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
112 0, 0, 0, U, U, U, U, 0, /* X Y Z [ \ ] ^ _ */
113 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
114 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
115 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
116 0, 0, 0, U, U, U, U, U, /* x y z { | } ~ DEL */
118 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
119 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
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,
123 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
124 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,
129 /* Decodes the forms %xy in a URL to the character the hexadecimal
130 code of which is xy. xy are hexadecimal digits from
131 [0123456789ABCDEF] (case-insensitive). If x or y are not
132 hex-digits or `%' precedes `\0', the sequence is inserted
136 decode_string (char *s)
138 char *t = s; /* t - tortoise */
139 char *h = s; /* h - hare */
150 /* Do nothing if '%' is not followed by two hex digits. */
151 if (!*(h + 1) || !*(h + 2)
152 || !(ISXDIGIT (*(h + 1)) && ISXDIGIT (*(h + 2))))
154 *t = (XCHAR_TO_XDIGIT (*(h + 1)) << 4) + XCHAR_TO_XDIGIT (*(h + 2));
161 /* Like encode_string, but return S if there are no unsafe chars. */
164 encode_string_maybe (const char *s)
171 for (p1 = s; *p1; p1++)
172 if (UNSAFE_CHAR (*p1))
173 addition += 2; /* Two more characters (hex digits) */
178 newlen = (p1 - s) + addition;
179 newstr = (char *)xmalloc (newlen + 1);
185 if (UNSAFE_CHAR (*p1))
187 unsigned char c = *p1++;
189 *p2++ = XDIGIT_TO_XCHAR (c >> 4);
190 *p2++ = XDIGIT_TO_XCHAR (c & 0xf);
196 assert (p2 - newstr == newlen);
201 /* Encode the unsafe characters (as determined by UNSAFE_CHAR) in a
202 given string, returning a malloc-ed %XX encoded string. */
205 encode_string (const char *s)
207 char *encoded = encode_string_maybe (s);
214 /* Encode unsafe characters in PTR to %xx. If such encoding is done,
215 the old value of PTR is freed and PTR is made to point to the newly
216 allocated storage. */
218 #define ENCODE(ptr) do { \
219 char *e_new = encode_string_maybe (ptr); \
227 enum copy_method { CM_DECODE, CM_ENCODE, CM_PASSTHROUGH };
229 /* Decide whether to encode, decode, or pass through the char at P.
230 This used to be a macro, but it got a little too convoluted. */
231 static inline enum copy_method
232 decide_copy_method (const char *p)
236 if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2)))
238 /* %xx sequence: decode it, unless it would decode to an
239 unsafe or a reserved char; in that case, leave it as
241 char preempt = (XCHAR_TO_XDIGIT (*(p + 1)) << 4) +
242 XCHAR_TO_XDIGIT (*(p + 2));
244 if (UNSAFE_CHAR (preempt) || RESERVED_CHAR (preempt))
245 return CM_PASSTHROUGH;
250 /* Garbled %.. sequence: encode `%'. */
253 else if (UNSAFE_CHAR (*p) && !RESERVED_CHAR (*p))
256 return CM_PASSTHROUGH;
259 /* Translate a %-quoting (but possibly non-conformant) input string S
260 into a %-quoting (and conformant) output string. If no characters
261 are encoded or decoded, return the same string S; otherwise, return
262 a freshly allocated string with the new contents.
264 After a URL has been run through this function, the protocols that
265 use `%' as the quote character can use the resulting string as-is,
266 while those that don't call decode_string() to get to the intended
267 data. This function is also stable: after an input string is
268 transformed the first time, all further transformations of the
269 result yield the same result string.
271 Let's discuss why this function is needed.
273 Imagine Wget is to retrieve `http://abc.xyz/abc def'. Since a raw
274 space character would mess up the HTTP request, it needs to be
277 GET /abc%20def HTTP/1.0
279 So it appears that the unsafe chars need to be quoted, as with
280 encode_string. But what if we're requested to download
281 `abc%20def'? Remember that %-encoding is valid URL syntax, so what
282 the user meant was a literal space, and he was kind enough to quote
283 it. In that case, Wget should obviously leave the `%20' as is, and
284 send the same request as above. So in this case we may not call
287 But what if the requested URI is `abc%20 def'? If we call
288 encode_string, we end up with `/abc%2520%20def', which is almost
289 certainly not intended. If we don't call encode_string, we are
290 left with the embedded space and cannot send the request. What the
291 user meant was for Wget to request `/abc%20%20def', and this is
292 where reencode_string kicks in.
294 Wget used to solve this by first decoding %-quotes, and then
295 encoding all the "unsafe" characters found in the resulting string.
296 This was wrong because it didn't preserve certain URL special
297 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
298 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
299 whether we considered `+' reserved (it is). One of these results
300 is inevitable because by the second step we would lose information
301 on whether the `+' was originally encoded or not. Both results
302 were wrong because in CGI parameters + means space, while %2B means
303 literal plus. reencode_string correctly translates the above to
304 "a%2B+b", i.e. returns the original string.
306 This function uses an algorithm proposed by Anon Sricharoenchai:
308 1. Encode all URL_UNSAFE and the "%" that are not followed by 2
311 2. Decode all "%XX" except URL_UNSAFE, URL_RESERVED (";/?:@=&") and
314 ...except that this code conflates the two steps, and decides
315 whether to encode, decode, or pass through each character in turn.
316 The function still uses two passes, but their logic is the same --
317 the first pass exists merely for the sake of allocation. Another
318 small difference is that we include `+' to URL_RESERVED.
322 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
324 "http://abc.xyz/%20%3F%2561%25aa%25%20a?a=a+a%2Ba&b=b%26c%3Dc"
328 "foo bar" -> "foo%20bar"
329 "foo%20bar" -> "foo%20bar"
330 "foo %20bar" -> "foo%20%20bar"
331 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
332 "foo%25%20bar" -> "foo%25%20bar"
333 "foo%2%20bar" -> "foo%252%20bar"
334 "foo+bar" -> "foo+bar" (plus is reserved!)
335 "foo%2b+bar" -> "foo%2b+bar" */
338 reencode_string (const char *s)
344 int encode_count = 0;
345 int decode_count = 0;
347 /* First, pass through the string to see if there's anything to do,
348 and to calculate the new length. */
349 for (p1 = s; *p1; p1++)
351 switch (decide_copy_method (p1))
364 if (!encode_count && !decode_count)
365 /* The string is good as it is. */
366 return (char *)s; /* C const model sucks. */
369 /* Each encoding adds two characters (hex digits), while each
370 decoding removes two characters. */
371 newlen = oldlen + 2 * (encode_count - decode_count);
372 newstr = xmalloc (newlen + 1);
379 switch (decide_copy_method (p1))
383 unsigned char c = *p1++;
385 *p2++ = XDIGIT_TO_XCHAR (c >> 4);
386 *p2++ = XDIGIT_TO_XCHAR (c & 0xf);
390 *p2++ = ((XCHAR_TO_XDIGIT (*(p1 + 1)) << 4)
391 + (XCHAR_TO_XDIGIT (*(p1 + 2))));
392 p1 += 3; /* skip %xx */
399 assert (p2 - newstr == newlen);
403 /* Run PTR_VAR through reencode_string. If a new string is consed,
404 free PTR_VAR and make it point to the new storage. Obviously,
405 PTR_VAR needs to be an lvalue. */
407 #define REENCODE(ptr_var) do { \
408 char *rf_new = reencode_string (ptr_var); \
409 if (rf_new != ptr_var) \
416 /* Returns the scheme type if the scheme is supported, or
417 SCHEME_INVALID if not. */
419 url_scheme (const char *url)
423 for (i = 0; supported_schemes[i].leading_string; i++)
424 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
425 strlen (supported_schemes[i].leading_string)))
427 if (supported_schemes[i].enabled)
428 return (enum url_scheme) i;
430 return SCHEME_INVALID;
433 return SCHEME_INVALID;
436 /* Return the number of characters needed to skip the scheme part of
437 the URL, e.g. `http://'. If no scheme is found, returns 0. */
439 url_skip_scheme (const char *url)
443 /* Skip the scheme name. We allow `-' and `+' because of `whois++',
445 while (ISALNUM (*p) || *p == '-' || *p == '+')
452 /* Skip "//" if found. */
453 if (*p == '/' && *(p + 1) == '/')
459 /* Returns 1 if the URL begins with a scheme (supported or
460 unsupported), 0 otherwise. */
462 url_has_scheme (const char *url)
465 while (ISALNUM (*p) || *p == '-' || *p == '+')
471 scheme_default_port (enum url_scheme scheme)
473 return supported_schemes[scheme].default_port;
477 scheme_disable (enum url_scheme scheme)
479 supported_schemes[scheme].enabled = 0;
482 /* Skip the username and password, if present here. The function
483 should be called *not* with the complete URL, but with the part
484 right after the scheme.
486 If no username and password are found, return 0. */
488 url_skip_uname (const char *url)
492 /* Look for '@' that comes before '/' or '?'. */
493 p = (const char *)strpbrk (url, "/?@");
501 parse_uname (const char *str, int len, char **user, char **passwd)
506 /* Empty user name not allowed. */
509 colon = memchr (str, ':', len);
511 /* Empty user name again. */
516 int pwlen = len - (colon + 1 - str);
517 *passwd = xmalloc (pwlen + 1);
518 memcpy (*passwd, colon + 1, pwlen);
519 (*passwd)[pwlen] = '\0';
525 *user = xmalloc (len + 1);
526 memcpy (*user, str, len);
532 /* Used by main.c: detect URLs written using the "shorthand" URL forms
533 popularized by Netscape and NcFTP. HTTP shorthands look like this:
535 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
536 www.foo.com[:port] -> http://www.foo.com[:port]
538 FTP shorthands look like this:
540 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
541 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
543 If the URL needs not or cannot be rewritten, return NULL. */
545 rewrite_shorthand_url (const char *url)
549 if (url_has_scheme (url))
552 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
554 for (p = url; *p && *p != ':' && *p != '/'; p++)
564 /* If the characters after the colon and before the next slash
565 or end of string are all digits, it's HTTP. */
567 for (pp = p + 1; ISDIGIT (*pp); pp++)
569 if (digits > 0 && (*pp == '/' || *pp == '\0'))
572 /* Prepend "ftp://" to the entire URL... */
573 res = xmalloc (6 + strlen (url) + 1);
574 sprintf (res, "ftp://%s", url);
575 /* ...and replace ':' with '/'. */
576 res[6 + (p - url)] = '/';
583 /* Just prepend "http://" to what we have. */
584 res = xmalloc (7 + strlen (url) + 1);
585 sprintf (res, "http://%s", url);
590 static void parse_path PARAMS ((const char *, char **, char **));
593 strpbrk_or_eos (const char *s, const char *accept)
595 char *p = strpbrk (s, accept);
597 p = (char *)s + strlen (s);
601 /* Turn STR into lowercase; return non-zero if a character was
605 lowercase_str (char *str)
612 *str = TOLOWER (*str);
617 static char *parse_errors[] = {
618 #define PE_NO_ERROR 0
620 #define PE_UNSUPPORTED_SCHEME 1
621 "Unsupported scheme",
622 #define PE_EMPTY_HOST 2
624 #define PE_BAD_PORT_NUMBER 3
626 #define PE_INVALID_USER_NAME 4
630 #define SETERR(p, v) do { \
637 Return a new struct url if successful, NULL on error. In case of
638 error, and if ERROR is not NULL, also set *ERROR to the appropriate
641 url_parse (const char *url, int *error)
645 int path_modified, host_modified;
647 enum url_scheme scheme;
649 const char *uname_b, *uname_e;
650 const char *host_b, *host_e;
651 const char *path_b, *path_e;
652 const char *params_b, *params_e;
653 const char *query_b, *query_e;
654 const char *fragment_b, *fragment_e;
657 char *user = NULL, *passwd = NULL;
661 scheme = url_scheme (url);
662 if (scheme == SCHEME_INVALID)
664 SETERR (error, PE_UNSUPPORTED_SCHEME);
668 url_encoded = reencode_string (url);
671 p += strlen (supported_schemes[scheme].leading_string);
673 p += url_skip_uname (p);
676 /* scheme://user:pass@host[:port]... */
679 /* We attempt to break down the URL into the components path,
680 params, query, and fragment. They are ordered like this:
682 scheme://host[:port][/path][;params][?query][#fragment] */
684 params_b = params_e = NULL;
685 query_b = query_e = NULL;
686 fragment_b = fragment_e = NULL;
689 p = strpbrk_or_eos (p, ":/;?#");
692 if (host_b == host_e)
694 SETERR (error, PE_EMPTY_HOST);
698 port = scheme_default_port (scheme);
701 const char *port_b, *port_e, *pp;
703 /* scheme://host:port/tralala */
707 p = strpbrk_or_eos (p, "/;?#");
710 if (port_b == port_e)
712 /* http://host:/whatever */
714 SETERR (error, PE_BAD_PORT_NUMBER);
718 for (port = 0, pp = port_b; pp < port_e; pp++)
722 /* http://host:12randomgarbage/blah */
724 SETERR (error, PE_BAD_PORT_NUMBER);
727 port = 10 * port + (*pp - '0');
735 p = strpbrk_or_eos (p, ";?#");
740 /* Path is not allowed not to exist. */
748 p = strpbrk_or_eos (p, "?#");
755 p = strpbrk_or_eos (p, "#");
767 if (uname_b != uname_e)
769 /* http://user:pass@host */
771 /* uname_b uname_e */
772 if (!parse_uname (uname_b, uname_e - uname_b - 1, &user, &passwd))
774 SETERR (error, PE_INVALID_USER_NAME);
779 u = (struct url *)xmalloc (sizeof (struct url));
780 memset (u, 0, sizeof (*u));
783 u->host = strdupdelim (host_b, host_e);
788 u->path = strdupdelim (path_b, path_e);
789 path_modified = path_simplify (u->path);
790 parse_path (u->path, &u->dir, &u->file);
792 host_modified = lowercase_str (u->host);
795 u->params = strdupdelim (params_b, params_e);
797 u->query = strdupdelim (query_b, query_e);
799 u->fragment = strdupdelim (fragment_b, fragment_e);
801 if (path_modified || u->fragment || host_modified || path_b == path_e)
803 /* If we suspect that a transformation has rendered what
804 url_string might return different from URL_ENCODED, rebuild
805 u->url using url_string. */
806 u->url = url_string (u, 0);
808 if (url_encoded != url)
809 xfree ((char *) url_encoded);
813 if (url_encoded == url)
814 u->url = xstrdup (url);
816 u->url = url_encoded;
824 url_error (int error_code)
826 assert (error_code >= 0 && error_code < ARRAY_SIZE (parse_errors));
827 return parse_errors[error_code];
831 parse_path (const char *quoted_path, char **dir, char **file)
833 char *path, *last_slash;
835 STRDUP_ALLOCA (path, quoted_path);
836 decode_string (path);
838 last_slash = strrchr (path, '/');
842 *file = xstrdup (path);
846 *dir = strdupdelim (path, last_slash);
847 *file = xstrdup (last_slash + 1);
851 /* Note: URL's "full path" is the path with the query string and
852 params appended. The "fragment" (#foo) is intentionally ignored,
853 but that might be changed. For example, if the original URL was
854 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
855 the full path will be "/foo/bar/baz;bullshit?querystring". */
857 /* Return the length of the full path, without the terminating
861 full_path_length (const struct url *url)
865 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
876 /* Write out the full path. */
879 full_path_write (const struct url *url, char *where)
881 #define FROB(el, chr) do { \
882 char *f_el = url->el; \
884 int l = strlen (f_el); \
886 memcpy (where, f_el, l); \
898 /* Public function for getting the "full path". E.g. if u->path is
899 "foo/bar" and u->query is "param=value", full_path will be
900 "/foo/bar?param=value". */
903 url_full_path (const struct url *url)
905 int length = full_path_length (url);
906 char *full_path = (char *)xmalloc(length + 1);
908 full_path_write (url, full_path);
909 full_path[length] = '\0';
914 /* Sync u->path and u->url with u->dir and u->file. */
917 sync_path (struct url *url)
925 newpath = xstrdup (url->file);
930 int dirlen = strlen (url->dir);
931 int filelen = strlen (url->file);
933 newpath = xmalloc (dirlen + 1 + filelen + 1);
934 memcpy (newpath, url->dir, dirlen);
935 newpath[dirlen] = '/';
936 memcpy (newpath + dirlen + 1, url->file, filelen);
937 newpath[dirlen + 1 + filelen] = '\0';
943 /* Synchronize u->url. */
945 url->url = url_string (url, 0);
948 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
949 This way we can sync u->path and u->url when they get changed. */
952 url_set_dir (struct url *url, const char *newdir)
955 url->dir = xstrdup (newdir);
960 url_set_file (struct url *url, const char *newfile)
963 url->file = xstrdup (newfile);
968 url_free (struct url *url)
974 FREE_MAYBE (url->params);
975 FREE_MAYBE (url->query);
976 FREE_MAYBE (url->fragment);
977 FREE_MAYBE (url->user);
978 FREE_MAYBE (url->passwd);
987 get_urls_file (const char *file)
989 struct file_memory *fm;
990 struct urlpos *head, *tail;
991 const char *text, *text_end;
994 fm = read_file (file);
997 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
1000 DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
1003 text_end = fm->content + fm->length;
1004 while (text < text_end)
1006 const char *line_beg = text;
1007 const char *line_end = memchr (text, '\n', text_end - text);
1009 line_end = text_end;
1013 while (line_beg < line_end
1014 && ISSPACE (*line_beg))
1016 while (line_end > line_beg + 1
1017 && ISSPACE (*(line_end - 1)))
1019 if (line_end > line_beg)
1021 /* URL is in the [line_beg, line_end) region. */
1025 struct urlpos *entry;
1028 /* We must copy the URL to a zero-terminated string, and we
1029 can't use alloca because we're in a loop. *sigh*. */
1030 url_text = strdupdelim (line_beg, line_end);
1034 /* Merge opt.base_href with URL. */
1035 char *merged = uri_merge (opt.base_href, url_text);
1040 url = url_parse (url_text, &up_error_code);
1043 logprintf (LOG_NOTQUIET, "%s: Invalid URL %s: %s\n",
1044 file, url_text, url_error (up_error_code));
1050 entry = (struct urlpos *)xmalloc (sizeof (struct urlpos));
1051 memset (entry, 0, sizeof (*entry));
1062 read_file_free (fm);
1066 /* Free the linked list of urlpos. */
1068 free_urlpos (struct urlpos *l)
1072 struct urlpos *next = l->next;
1075 FREE_MAYBE (l->local_name);
1081 /* Rotate FNAME opt.backups times */
1083 rotate_backups(const char *fname)
1085 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
1086 char *from = (char *)alloca (maxlen);
1087 char *to = (char *)alloca (maxlen);
1091 if (stat (fname, &sb) == 0)
1092 if (S_ISREG (sb.st_mode) == 0)
1095 for (i = opt.backups; i > 1; i--)
1097 sprintf (from, "%s.%d", fname, i - 1);
1098 sprintf (to, "%s.%d", fname, i);
1099 /* #### This will fail on machines without the rename() system
1104 sprintf (to, "%s.%d", fname, 1);
1108 /* Create all the necessary directories for PATH (a file). Calls
1109 mkdirhier() internally. */
1111 mkalldirs (const char *path)
1118 p = path + strlen (path);
1119 for (; *p != '/' && p != path; p--);
1120 /* Don't create if it's just a file. */
1121 if ((p == path) && (*p != '/'))
1123 t = strdupdelim (path, p);
1124 /* Check whether the directory exists. */
1125 if ((stat (t, &st) == 0))
1127 if (S_ISDIR (st.st_mode))
1134 /* If the dir exists as a file name, remove it first. This
1135 is *only* for Wget to work with buggy old CERN http
1136 servers. Here is the scenario: When Wget tries to
1137 retrieve a directory without a slash, e.g.
1138 http://foo/bar (bar being a directory), CERN server will
1139 not redirect it too http://foo/bar/ -- it will generate a
1140 directory listing containing links to bar/file1,
1141 bar/file2, etc. Wget will lose because it saves this
1142 HTML listing to a file `bar', so it cannot create the
1143 directory. To work around this, if the file of the same
1144 name exists, we just remove it and create the directory
1146 DEBUGP (("Removing %s because of directory danger!\n", t));
1150 res = make_directory (t);
1152 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1158 count_slashes (const char *s)
1167 /* Return the path name of the URL-equivalent file name, with a
1168 remote-like structure of directories. */
1170 mkstruct (const struct url *u)
1172 char *dir, *dir_preencoding;
1173 char *file, *res, *dirpref;
1174 char *query = u->query && *u->query ? u->query : NULL;
1179 char *ptr = u->dir + (*u->dir == '/');
1180 int slash_count = 1 + count_slashes (ptr);
1181 int cut = MINVAL (opt.cut_dirs, slash_count);
1182 for (; cut && *ptr; ptr++)
1185 STRDUP_ALLOCA (dir, ptr);
1188 dir = u->dir + (*u->dir == '/');
1190 /* Check for the true name (or at least a consistent name for saving
1191 to directory) of HOST, reusing the hlist if possible. */
1192 if (opt.add_hostdir)
1194 /* Add dir_prefix and hostname (if required) to the beginning of
1196 dirpref = (char *)alloca (strlen (opt.dir_prefix) + 1
1198 + 1 + numdigit (u->port)
1200 if (!DOTP (opt.dir_prefix))
1201 sprintf (dirpref, "%s/%s", opt.dir_prefix, u->host);
1203 strcpy (dirpref, u->host);
1205 if (u->port != scheme_default_port (u->scheme))
1207 int len = strlen (dirpref);
1209 long_to_string (dirpref + len + 1, u->port);
1212 else /* not add_hostdir */
1214 if (!DOTP (opt.dir_prefix))
1215 dirpref = opt.dir_prefix;
1220 /* If there is a prefix, prepend it. */
1223 char *newdir = (char *)alloca (strlen (dirpref) + 1 + strlen (dir) + 2);
1224 sprintf (newdir, "%s%s%s", dirpref, *dir == '/' ? "" : "/", dir);
1228 dir_preencoding = dir;
1229 dir = reencode_string (dir_preencoding);
1232 if (l && dir[l - 1] == '/')
1236 file = "index.html";
1240 /* Finally, construct the full name. */
1241 res = (char *)xmalloc (strlen (dir) + 1 + strlen (file)
1242 + (query ? (1 + strlen (query)) : 0)
1244 sprintf (res, "%s%s%s", dir, *dir ? "/" : "", file);
1248 strcat (res, query);
1250 if (dir != dir_preencoding)
1255 /* Compose a file name out of BASE, an unescaped file name, and QUERY,
1256 an escaped query string. The trick is to make sure that unsafe
1257 characters in BASE are escaped, and that slashes in QUERY are also
1261 compose_file_name (char *base, char *query)
1267 /* Copy BASE to RESULT and encode all unsafe characters. */
1269 while (*from && to - result < sizeof (result))
1271 if (UNSAFE_CHAR (*from))
1273 unsigned char c = *from++;
1275 *to++ = XDIGIT_TO_XCHAR (c >> 4);
1276 *to++ = XDIGIT_TO_XCHAR (c & 0xf);
1282 if (query && to - result < sizeof (result))
1286 /* Copy QUERY to RESULT and encode all '/' characters. */
1288 while (*from && to - result < sizeof (result))
1302 if (to - result < sizeof (result))
1305 /* Truncate input which is too long, presumably due to a huge
1307 result[sizeof (result) - 1] = '\0';
1309 return xstrdup (result);
1312 /* Create a unique filename, corresponding to a given URL. Calls
1313 mkstruct if necessary. Does *not* actually create any directories. */
1315 url_filename (const struct url *u)
1318 int have_prefix = 0; /* whether we must prepend opt.dir_prefix */
1322 file = mkstruct (u);
1327 char *base = *u->file ? u->file : "index.html";
1328 char *query = u->query && *u->query ? u->query : NULL;
1329 file = compose_file_name (base, query);
1334 /* Check whether the prefix directory is something other than "."
1335 before prepending it. */
1336 if (!DOTP (opt.dir_prefix))
1338 char *nfile = (char *)xmalloc (strlen (opt.dir_prefix)
1339 + 1 + strlen (file) + 1);
1340 sprintf (nfile, "%s/%s", opt.dir_prefix, file);
1345 /* DOS-ish file systems don't like `%' signs in them; we change it
1350 for (p = file; *p; p++)
1354 #endif /* WINDOWS */
1356 /* Check the cases in which the unique extensions are not used:
1357 1) Clobbering is turned off (-nc).
1358 2) Retrieval with regetting.
1359 3) Timestamping is used.
1360 4) Hierarchy is built.
1362 The exception is the case when file does exist and is a
1363 directory (actually support for bad httpd-s). */
1364 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1365 && !(file_exists_p (file) && !file_non_directory_p (file)))
1368 /* Find a unique name. */
1369 name = unique_name (file);
1374 /* Like strlen(), but allow the URL to be ended with '?'. */
1376 urlpath_length (const char *url)
1378 const char *q = strpbrk_or_eos (url, "?;#");
1382 /* Find the last occurrence of character C in the range [b, e), or
1383 NULL, if none are present. This is almost completely equivalent to
1384 { *e = '\0'; return strrchr(b); }, except that it doesn't change
1385 the contents of the string. */
1387 find_last_char (const char *b, const char *e, char c)
1395 /* Resolve the result of "linking" a base URI (BASE) to a
1396 link-specified URI (LINK).
1398 Either of the URIs may be absolute or relative, complete with the
1399 host name, or path only. This tries to behave "reasonably" in all
1400 foreseeable cases. It employs little specific knowledge about
1401 schemes or URL-specific stuff -- it just works on strings.
1403 The parameters LINKLENGTH is useful if LINK is not zero-terminated.
1404 See uri_merge for a gentler interface to this functionality.
1406 Perhaps this function should handle `./' and `../' so that the evil
1407 path_simplify can go. */
1409 uri_merge_1 (const char *base, const char *link, int linklength, int no_scheme)
1415 const char *end = base + urlpath_length (base);
1419 /* Empty LINK points back to BASE, query string and all. */
1420 constr = xstrdup (base);
1422 else if (*link == '?')
1424 /* LINK points to the same location, but changes the query
1425 string. Examples: */
1426 /* uri_merge("path", "?new") -> "path?new" */
1427 /* uri_merge("path?foo", "?new") -> "path?new" */
1428 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1429 /* uri_merge("path#foo", "?new") -> "path?new" */
1430 int baselength = end - base;
1431 constr = xmalloc (baselength + linklength + 1);
1432 memcpy (constr, base, baselength);
1433 memcpy (constr + baselength, link, linklength);
1434 constr[baselength + linklength] = '\0';
1436 else if (*link == '#')
1438 /* uri_merge("path", "#new") -> "path#new" */
1439 /* uri_merge("path#foo", "#new") -> "path#new" */
1440 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1441 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1443 const char *end1 = strchr (base, '#');
1445 end1 = base + strlen (base);
1446 baselength = end1 - base;
1447 constr = xmalloc (baselength + linklength + 1);
1448 memcpy (constr, base, baselength);
1449 memcpy (constr + baselength, link, linklength);
1450 constr[baselength + linklength] = '\0';
1452 else if (*link == '/')
1454 /* LINK is an absolute path: we need to replace everything
1455 after (and including) the FIRST slash with LINK.
1457 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1458 "/qux/xyzzy", our result should be
1459 "http://host/qux/xyzzy". */
1462 const char *start_insert = NULL; /* for gcc to shut up. */
1463 const char *pos = base;
1464 int seen_slash_slash = 0;
1465 /* We're looking for the first slash, but want to ignore
1468 slash = memchr (pos, '/', end - pos);
1469 if (slash && !seen_slash_slash)
1470 if (*(slash + 1) == '/')
1473 seen_slash_slash = 1;
1477 /* At this point, SLASH is the location of the first / after
1478 "//", or the first slash altogether. START_INSERT is the
1479 pointer to the location where LINK will be inserted. When
1480 examining the last two examples, keep in mind that LINK
1483 if (!slash && !seen_slash_slash)
1484 /* example: "foo" */
1486 start_insert = base;
1487 else if (!slash && seen_slash_slash)
1488 /* example: "http://foo" */
1491 else if (slash && !seen_slash_slash)
1492 /* example: "foo/bar" */
1494 start_insert = base;
1495 else if (slash && seen_slash_slash)
1496 /* example: "http://something/" */
1498 start_insert = slash;
1500 span = start_insert - base;
1501 constr = (char *)xmalloc (span + linklength + 1);
1503 memcpy (constr, base, span);
1505 memcpy (constr + span, link, linklength);
1506 constr[span + linklength] = '\0';
1510 /* LINK is a relative URL: we need to replace everything
1511 after last slash (possibly empty) with LINK.
1513 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1514 our result should be "whatever/foo/qux/xyzzy". */
1515 int need_explicit_slash = 0;
1517 const char *start_insert;
1518 const char *last_slash = find_last_char (base, end, '/');
1521 /* No slash found at all. Append LINK to what we have,
1522 but we'll need a slash as a separator.
1524 Example: if base == "foo" and link == "qux/xyzzy", then
1525 we cannot just append link to base, because we'd get
1526 "fooqux/xyzzy", whereas what we want is
1529 To make sure the / gets inserted, we set
1530 need_explicit_slash to 1. We also set start_insert
1531 to end + 1, so that the length calculations work out
1532 correctly for one more (slash) character. Accessing
1533 that character is fine, since it will be the
1534 delimiter, '\0' or '?'. */
1535 /* example: "foo?..." */
1536 /* ^ ('?' gets changed to '/') */
1537 start_insert = end + 1;
1538 need_explicit_slash = 1;
1540 else if (last_slash && last_slash != base && *(last_slash - 1) == '/')
1542 /* example: http://host" */
1544 start_insert = end + 1;
1545 need_explicit_slash = 1;
1549 /* example: "whatever/foo/bar" */
1551 start_insert = last_slash + 1;
1554 span = start_insert - base;
1555 constr = (char *)xmalloc (span + linklength + 1);
1557 memcpy (constr, base, span);
1558 if (need_explicit_slash)
1559 constr[span - 1] = '/';
1561 memcpy (constr + span, link, linklength);
1562 constr[span + linklength] = '\0';
1565 else /* !no_scheme */
1567 constr = strdupdelim (link, link + linklength);
1572 /* Merge BASE with LINK and return the resulting URI. This is an
1573 interface to uri_merge_1 that assumes that LINK is a
1574 zero-terminated string. */
1576 uri_merge (const char *base, const char *link)
1578 return uri_merge_1 (base, link, strlen (link), !url_has_scheme (link));
1581 #define APPEND(p, s) do { \
1582 int len = strlen (s); \
1583 memcpy (p, s, len); \
1587 /* Use this instead of password when the actual password is supposed
1588 to be hidden. We intentionally use a generic string without giving
1589 away the number of characters in the password, like previous
1591 #define HIDDEN_PASSWORD "*password*"
1593 /* Recreate the URL string from the data in URL.
1595 If HIDE is non-zero (as it is when we're calling this on a URL we
1596 plan to print, but not when calling it to canonicalize a URL for
1597 use within the program), password will be hidden. Unsafe
1598 characters in the URL will be quoted. */
1601 url_string (const struct url *url, int hide_password)
1605 char *quoted_user = NULL, *quoted_passwd = NULL;
1607 int scheme_port = supported_schemes[url->scheme].default_port;
1608 char *scheme_str = supported_schemes[url->scheme].leading_string;
1609 int fplen = full_path_length (url);
1611 assert (scheme_str != NULL);
1613 /* Make sure the user name and password are quoted. */
1616 quoted_user = encode_string_maybe (url->user);
1620 quoted_passwd = HIDDEN_PASSWORD;
1622 quoted_passwd = encode_string_maybe (url->passwd);
1626 size = (strlen (scheme_str)
1627 + strlen (url->host)
1630 if (url->port != scheme_port)
1631 size += 1 + numdigit (url->port);
1634 size += 1 + strlen (quoted_user);
1636 size += 1 + strlen (quoted_passwd);
1639 p = result = xmalloc (size);
1641 APPEND (p, scheme_str);
1644 APPEND (p, quoted_user);
1648 APPEND (p, quoted_passwd);
1653 APPEND (p, url->host);
1654 if (url->port != scheme_port)
1657 long_to_string (p, url->port);
1661 full_path_write (url, p);
1665 assert (p - result == size);
1667 if (quoted_user && quoted_user != url->user)
1668 xfree (quoted_user);
1669 if (quoted_passwd && !hide_password
1670 && quoted_passwd != url->passwd)
1671 xfree (quoted_passwd);
1676 /* Returns proxy host address, in accordance with SCHEME. */
1678 getproxy (enum url_scheme scheme)
1681 char *rewritten_url;
1682 static char rewritten_storage[1024];
1687 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
1691 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
1695 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
1697 case SCHEME_INVALID:
1700 if (!proxy || !*proxy)
1703 /* Handle shorthands. */
1704 rewritten_url = rewrite_shorthand_url (proxy);
1707 strncpy (rewritten_storage, rewritten_url, sizeof(rewritten_storage));
1708 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
1709 proxy = rewritten_storage;
1715 /* Should a host be accessed through proxy, concerning no_proxy? */
1717 no_proxy_match (const char *host, const char **no_proxy)
1722 return !sufmatch (no_proxy, host);
1725 static void write_backup_file PARAMS ((const char *, downloaded_file_t));
1726 static const char *replace_attr PARAMS ((const char *, int, FILE *,
1728 static const char *replace_attr_refresh_hack PARAMS ((const char *, int, FILE *,
1729 const char *, int));
1730 static char *local_quote_string PARAMS ((const char *));
1732 /* Change the links in one HTML file. LINKS is a list of links in the
1733 document, along with their positions and the desired direction of
1736 convert_links (const char *file, struct urlpos *links)
1738 struct file_memory *fm;
1741 downloaded_file_t downloaded_file_return;
1743 struct urlpos *link;
1744 int to_url_count = 0, to_file_count = 0;
1746 logprintf (LOG_VERBOSE, _("Converting %s... "), file);
1749 /* First we do a "dry run": go through the list L and see whether
1750 any URL needs to be converted in the first place. If not, just
1751 leave the file alone. */
1753 struct urlpos *dry = links;
1754 for (dry = links; dry; dry = dry->next)
1755 if (dry->convert != CO_NOCONVERT)
1759 logputs (LOG_VERBOSE, _("nothing to do.\n"));
1764 fm = read_file (file);
1767 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1768 file, strerror (errno));
1772 downloaded_file_return = downloaded_file (CHECK_FOR_FILE, file);
1773 if (opt.backup_converted && downloaded_file_return)
1774 write_backup_file (file, downloaded_file_return);
1776 /* Before opening the file for writing, unlink the file. This is
1777 important if the data in FM is mmaped. In such case, nulling the
1778 file, which is what fopen() below does, would make us read all
1779 zeroes from the mmaped region. */
1780 if (unlink (file) < 0 && errno != ENOENT)
1782 logprintf (LOG_NOTQUIET, _("Unable to delete `%s': %s\n"),
1783 file, strerror (errno));
1784 read_file_free (fm);
1787 /* Now open the file for writing. */
1788 fp = fopen (file, "wb");
1791 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1792 file, strerror (errno));
1793 read_file_free (fm);
1797 /* Here we loop through all the URLs in file, replacing those of
1798 them that are downloaded with relative references. */
1800 for (link = links; link; link = link->next)
1802 char *url_start = fm->content + link->pos;
1804 if (link->pos >= fm->length)
1806 DEBUGP (("Something strange is going on. Please investigate."));
1809 /* If the URL is not to be converted, skip it. */
1810 if (link->convert == CO_NOCONVERT)
1812 DEBUGP (("Skipping %s at position %d.\n", link->url->url, link->pos));
1816 /* Echo the file contents, up to the offending URL's opening
1817 quote, to the outfile. */
1818 fwrite (p, 1, url_start - p, fp);
1821 switch (link->convert)
1823 case CO_CONVERT_TO_RELATIVE:
1824 /* Convert absolute URL to relative. */
1826 char *newname = construct_relative (file, link->local_name);
1827 char *quoted_newname = local_quote_string (newname);
1829 if (!link->link_refresh_p)
1830 p = replace_attr (p, link->size, fp, quoted_newname);
1832 p = replace_attr_refresh_hack (p, link->size, fp, quoted_newname,
1833 link->refresh_timeout);
1835 DEBUGP (("TO_RELATIVE: %s to %s at position %d in %s.\n",
1836 link->url->url, newname, link->pos, file));
1838 xfree (quoted_newname);
1842 case CO_CONVERT_TO_COMPLETE:
1843 /* Convert the link to absolute URL. */
1845 char *newlink = link->url->url;
1846 char *quoted_newlink = html_quote_string (newlink);
1848 if (!link->link_refresh_p)
1849 p = replace_attr (p, link->size, fp, quoted_newlink);
1851 p = replace_attr_refresh_hack (p, link->size, fp, quoted_newlink,
1852 link->refresh_timeout);
1854 DEBUGP (("TO_COMPLETE: <something> to %s at position %d in %s.\n",
1855 newlink, link->pos, file));
1856 xfree (quoted_newlink);
1860 case CO_NULLIFY_BASE:
1861 /* Change the base href to "". */
1862 p = replace_attr (p, link->size, fp, "");
1870 /* Output the rest of the file. */
1871 if (p - fm->content < fm->length)
1872 fwrite (p, 1, fm->length - (p - fm->content), fp);
1874 read_file_free (fm);
1876 logprintf (LOG_VERBOSE, "%d-%d\n", to_file_count, to_url_count);
1879 /* Construct and return a malloced copy of the relative link from two
1880 pieces of information: local name S1 of the referring file and
1881 local name S2 of the referred file.
1883 So, if S1 is "jagor.srce.hr/index.html" and S2 is
1884 "jagor.srce.hr/images/news.gif", the function will return
1887 Alternately, if S1 is "fly.cc.fer.hr/ioccc/index.html", and S2 is
1888 "fly.cc.fer.hr/images/fly.gif", the function will return
1889 "../images/fly.gif".
1891 Caveats: S1 should not begin with `/', unless S2 also begins with
1892 '/'. S1 should not contain things like ".." and such --
1893 construct_relative ("fly/ioccc/../index.html",
1894 "fly/images/fly.gif") will fail. (A workaround is to call
1895 something like path_simplify() on S1). */
1897 construct_relative (const char *s1, const char *s2)
1899 int i, cnt, sepdirs1;
1903 return xstrdup (s2);
1904 /* S1 should *not* be absolute, if S2 wasn't. */
1905 assert (*s1 != '/');
1907 /* Skip the directories common to both strings. */
1910 while (s1[i] && s2[i]
1915 if (s1[i] == '/' && s2[i] == '/')
1920 for (sepdirs1 = 0; s1[i]; i++)
1923 /* Now, construct the file as of:
1924 - ../ repeated sepdirs1 time
1925 - all the non-mutual directories of S2. */
1926 res = (char *)xmalloc (3 * sepdirs1 + strlen (s2 + cnt) + 1);
1927 for (i = 0; i < sepdirs1; i++)
1928 memcpy (res + 3 * i, "../", 3);
1929 strcpy (res + 3 * i, s2 + cnt);
1934 write_backup_file (const char *file, downloaded_file_t downloaded_file_return)
1936 /* Rather than just writing over the original .html file with the
1937 converted version, save the former to *.orig. Note we only do
1938 this for files we've _successfully_ downloaded, so we don't
1939 clobber .orig files sitting around from previous invocations. */
1941 /* Construct the backup filename as the original name plus ".orig". */
1942 size_t filename_len = strlen(file);
1943 char* filename_plus_orig_suffix;
1944 boolean already_wrote_backup_file = FALSE;
1945 slist* converted_file_ptr;
1946 static slist* converted_files = NULL;
1948 if (downloaded_file_return == FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED)
1950 /* Just write "orig" over "html". We need to do it this way
1951 because when we're checking to see if we've downloaded the
1952 file before (to see if we can skip downloading it), we don't
1953 know if it's a text/html file. Therefore we don't know yet
1954 at that stage that -E is going to cause us to tack on
1955 ".html", so we need to compare vs. the original URL plus
1956 ".orig", not the original URL plus ".html.orig". */
1957 filename_plus_orig_suffix = alloca (filename_len + 1);
1958 strcpy(filename_plus_orig_suffix, file);
1959 strcpy((filename_plus_orig_suffix + filename_len) - 4, "orig");
1961 else /* downloaded_file_return == FILE_DOWNLOADED_NORMALLY */
1963 /* Append ".orig" to the name. */
1964 filename_plus_orig_suffix = alloca (filename_len + sizeof(".orig"));
1965 strcpy(filename_plus_orig_suffix, file);
1966 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
1969 /* We can get called twice on the same URL thanks to the
1970 convert_all_links() call in main(). If we write the .orig file
1971 each time in such a case, it'll end up containing the first-pass
1972 conversion, not the original file. So, see if we've already been
1973 called on this file. */
1974 converted_file_ptr = converted_files;
1975 while (converted_file_ptr != NULL)
1976 if (strcmp(converted_file_ptr->string, file) == 0)
1978 already_wrote_backup_file = TRUE;
1982 converted_file_ptr = converted_file_ptr->next;
1984 if (!already_wrote_backup_file)
1986 /* Rename <file> to <file>.orig before former gets written over. */
1987 if (rename(file, filename_plus_orig_suffix) != 0)
1988 logprintf (LOG_NOTQUIET, _("Cannot back up %s as %s: %s\n"),
1989 file, filename_plus_orig_suffix, strerror (errno));
1991 /* Remember that we've already written a .orig backup for this file.
1992 Note that we never free this memory since we need it till the
1993 convert_all_links() call, which is one of the last things the
1994 program does before terminating. BTW, I'm not sure if it would be
1995 safe to just set 'converted_file_ptr->string' to 'file' below,
1996 rather than making a copy of the string... Another note is that I
1997 thought I could just add a field to the urlpos structure saying
1998 that we'd written a .orig file for this URL, but that didn't work,
1999 so I had to make this separate list.
2000 -- Dan Harkless <wget@harkless.org>
2002 This [adding a field to the urlpos structure] didn't work
2003 because convert_file() is called from convert_all_links at
2004 the end of the retrieval with a freshly built new urlpos
2006 -- Hrvoje Niksic <hniksic@arsdigita.com>
2008 converted_file_ptr = xmalloc(sizeof(*converted_file_ptr));
2009 converted_file_ptr->string = xstrdup(file); /* die on out-of-mem. */
2010 converted_file_ptr->next = converted_files;
2011 converted_files = converted_file_ptr;
2015 static int find_fragment PARAMS ((const char *, int, const char **,
2018 /* Replace an attribute's original text with NEW_TEXT. */
2021 replace_attr (const char *p, int size, FILE *fp, const char *new_text)
2024 char quote_char = '\"'; /* use "..." for quoting, unless the
2025 original value is quoted, in which
2026 case reuse its quoting char. */
2027 const char *frag_beg, *frag_end;
2029 /* Structure of our string is:
2030 "...old-contents..."
2031 <--- size ---> (with quotes)
2034 <--- size --> (no quotes) */
2036 if (*p == '\"' || *p == '\'')
2041 size -= 2; /* disregard opening and closing quote */
2043 putc (quote_char, fp);
2044 fputs (new_text, fp);
2046 /* Look for fragment identifier, if any. */
2047 if (find_fragment (p, size, &frag_beg, &frag_end))
2048 fwrite (frag_beg, 1, frag_end - frag_beg, fp);
2052 putc (quote_char, fp);
2057 /* The same as REPLACE_ATTR, but used when replacing
2058 <meta http-equiv=refresh content="new_text"> because we need to
2059 append "timeout_value; URL=" before the next_text. */
2062 replace_attr_refresh_hack (const char *p, int size, FILE *fp,
2063 const char *new_text, int timeout)
2066 char *new_with_timeout = (char *)alloca (numdigit (timeout)
2070 sprintf (new_with_timeout, "%d; URL=%s", timeout, new_text);
2072 return replace_attr (p, size, fp, new_with_timeout);
2075 /* Find the first occurrence of '#' in [BEG, BEG+SIZE) that is not
2076 preceded by '&'. If the character is not found, return zero. If
2077 the character is found, return 1 and set BP and EP to point to the
2078 beginning and end of the region.
2080 This is used for finding the fragment indentifiers in URLs. */
2083 find_fragment (const char *beg, int size, const char **bp, const char **ep)
2085 const char *end = beg + size;
2087 for (; beg < end; beg++)
2109 /* Quote FILE for use as local reference to an HTML file.
2111 We quote ? as %3F to avoid passing part of the file name as the
2112 parameter when browsing the converted file through HTTP. However,
2113 it is safe to do this only when `--html-extension' is turned on.
2114 This is because converting "index.html?foo=bar" to
2115 "index.html%3Ffoo=bar" would break local browsing, as the latter
2116 isn't even recognized as an HTML file! However, converting
2117 "index.html?foo=bar.html" to "index.html%3Ffoo=bar.html" should be
2118 safe for both local and HTTP-served browsing. */
2121 local_quote_string (const char *file)
2123 const char *file_sans_qmark;
2126 if (!opt.html_extension)
2127 return html_quote_string (file);
2129 qm = count_char (file, '?');
2133 const char *from = file;
2136 /* qm * 2 because we replace each question mark with "%3F",
2137 i.e. replace one char with three, hence two more. */
2138 int fsqlen = strlen (file) + qm * 2;
2140 to = newname = (char *)alloca (fsqlen + 1);
2141 for (; *from; from++)
2152 assert (to - newname == fsqlen);
2155 file_sans_qmark = newname;
2158 file_sans_qmark = file;
2160 return html_quote_string (file_sans_qmark);
2163 /* We're storing "modes" of type downloaded_file_t in the hash table.
2164 However, our hash tables only accept pointers for keys and values.
2165 So when we need a pointer, we use the address of a
2166 downloaded_file_t variable of static storage. */
2168 static downloaded_file_t *
2169 downloaded_mode_to_ptr (downloaded_file_t mode)
2171 static downloaded_file_t
2172 v1 = FILE_NOT_ALREADY_DOWNLOADED,
2173 v2 = FILE_DOWNLOADED_NORMALLY,
2174 v3 = FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED,
2175 v4 = CHECK_FOR_FILE;
2179 case FILE_NOT_ALREADY_DOWNLOADED:
2181 case FILE_DOWNLOADED_NORMALLY:
2183 case FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED:
2185 case CHECK_FOR_FILE:
2191 /* This should really be merged with dl_file_url_map and
2192 downloaded_html_files in recur.c. This was originally a list, but
2193 I changed it to a hash table beause it was actually taking a lot of
2194 time to find things in it. */
2196 static struct hash_table *downloaded_files_hash;
2198 /* Remembers which files have been downloaded. In the standard case, should be
2199 called with mode == FILE_DOWNLOADED_NORMALLY for each file we actually
2200 download successfully (i.e. not for ones we have failures on or that we skip
2203 When we've downloaded a file and tacked on a ".html" extension due to -E,
2204 call this function with FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED rather than
2205 FILE_DOWNLOADED_NORMALLY.
2207 If you just want to check if a file has been previously added without adding
2208 it, call with mode == CHECK_FOR_FILE. Please be sure to call this function
2209 with local filenames, not remote URLs. */
2211 downloaded_file (downloaded_file_t mode, const char *file)
2213 downloaded_file_t *ptr;
2215 if (mode == CHECK_FOR_FILE)
2217 if (!downloaded_files_hash)
2218 return FILE_NOT_ALREADY_DOWNLOADED;
2219 ptr = hash_table_get (downloaded_files_hash, file);
2221 return FILE_NOT_ALREADY_DOWNLOADED;
2225 if (!downloaded_files_hash)
2226 downloaded_files_hash = make_string_hash_table (0);
2228 ptr = hash_table_get (downloaded_files_hash, file);
2232 ptr = downloaded_mode_to_ptr (mode);
2233 hash_table_put (downloaded_files_hash, xstrdup (file), &ptr);
2235 return FILE_NOT_ALREADY_DOWNLOADED;
2239 df_free_mapper (void *key, void *value, void *ignored)
2246 downloaded_files_free (void)
2248 if (downloaded_files_hash)
2250 hash_table_map (downloaded_files_hash, df_free_mapper, NULL);
2251 hash_table_destroy (downloaded_files_hash);
2252 downloaded_files_hash = NULL;