2 Copyright (C) 1995, 1996, 1997, 2000, 2001, 2003, 2003
3 Free Software Foundation, Inc.
5 This file is part of GNU Wget.
7 GNU Wget is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 2 of the License, or (at
10 your option) any later version.
12 GNU Wget is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with Wget; if not, write to the Free Software
19 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
21 In addition, as a special exception, the Free Software Foundation
22 gives permission to link the code of its release of Wget with the
23 OpenSSL project's "OpenSSL" library (or with modified versions of it
24 that use the same license as the "OpenSSL" library), and distribute
25 the linked executables. You must obey the GNU General Public License
26 in all respects for all of the code used other than "OpenSSL". If you
27 modify this file, you may extend this exception to your version of the
28 file, but you are not obligated to do so. If you do not wish to do
29 so, delete this exception statement from your version. */
40 #include <sys/types.h>
50 #include "host.h" /* for is_valid_ipv6_address */
59 const char *leading_string;
64 /* Supported schemes: */
65 static struct scheme_data supported_schemes[] =
67 { "http", "http://", DEFAULT_HTTP_PORT, 1 },
69 { "https", "https://", DEFAULT_HTTPS_PORT, 1 },
71 { "ftp", "ftp://", DEFAULT_FTP_PORT, 1 },
77 /* Forward declarations: */
79 static int path_simplify PARAMS ((char *));
81 /* Support for escaping and unescaping of URL strings. */
83 /* Table of "reserved" and "unsafe" characters. Those terms are
84 rfc1738-speak, as such largely obsoleted by rfc2396 and later
85 specs, but the general idea remains.
87 A reserved character is the one that you can't decode without
88 changing the meaning of the URL. For example, you can't decode
89 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
90 path components is different. Non-reserved characters can be
91 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". Wget
92 uses the rfc1738 set of reserved characters, plus "$" and ",", as
93 recommended by rfc2396.
95 An unsafe characters is the one that should be encoded when URLs
96 are placed in foreign environments. E.g. space and newline are
97 unsafe in HTTP contexts because HTTP uses them as separator and
98 terminator, so they must be encoded to %20 and %0A respectively.
99 "*" is unsafe in shell context, etc.
101 We determine whether a character is unsafe through static table
102 lookup. This code assumes ASCII character set and 8-bit chars. */
105 /* rfc1738 reserved chars + "$" and ",". */
108 /* rfc1738 unsafe chars, plus non-printables. */
112 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
113 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
114 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
116 /* Shorthands for the table: */
117 #define R urlchr_reserved
118 #define U urlchr_unsafe
121 const static unsigned char urlchr_table[256] =
123 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
124 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
125 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
126 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
127 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
128 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
129 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
130 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
131 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
132 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
133 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
134 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
135 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
136 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
137 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
138 0, 0, 0, U, U, U, U, U, /* x y z { | } ~ DEL */
140 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
141 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
142 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
143 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
145 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
146 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
147 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
148 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
154 /* URL-unescape the string S.
156 This is done by transforming the sequences "%HH" to the character
157 represented by the hexadecimal digits HH. If % is not followed by
158 two hexadecimal digits, it is inserted literally.
160 The transformation is done in place. If you need the original
161 string intact, make a copy before calling this function. */
164 url_unescape (char *s)
166 char *t = s; /* t - tortoise */
167 char *h = s; /* h - hare */
178 /* Do nothing if '%' is not followed by two hex digits. */
179 if (!h[1] || !h[2] || !(ISXDIGIT (h[1]) && ISXDIGIT (h[2])))
181 *t = X2DIGITS_TO_NUM (h[1], h[2]);
188 /* The core of url_escape_* functions. Escapes the characters that
189 match the provided mask in urlchr_table.
191 If ALLOW_PASSTHROUGH is non-zero, a string with no unsafe chars
192 will be returned unchanged. If ALLOW_PASSTHROUGH is zero, a
193 freshly allocated string will be returned in all cases. */
196 url_escape_1 (const char *s, unsigned char mask, int allow_passthrough)
203 for (p1 = s; *p1; p1++)
204 if (urlchr_test (*p1, mask))
205 addition += 2; /* Two more characters (hex digits) */
208 return allow_passthrough ? (char *)s : xstrdup (s);
210 newlen = (p1 - s) + addition;
211 newstr = (char *)xmalloc (newlen + 1);
217 /* Quote the characters that match the test mask. */
218 if (urlchr_test (*p1, mask))
220 unsigned char c = *p1++;
222 *p2++ = XNUM_TO_DIGIT (c >> 4);
223 *p2++ = XNUM_TO_DIGIT (c & 0xf);
228 assert (p2 - newstr == newlen);
234 /* URL-escape the unsafe characters (see urlchr_table) in a given
235 string, returning a freshly allocated string. */
238 url_escape (const char *s)
240 return url_escape_1 (s, urlchr_unsafe, 0);
243 /* URL-escape the unsafe characters (see urlchr_table) in a given
244 string. If no characters are unsafe, S is returned. */
247 url_escape_allow_passthrough (const char *s)
249 return url_escape_1 (s, urlchr_unsafe, 1);
252 enum copy_method { CM_DECODE, CM_ENCODE, CM_PASSTHROUGH };
254 /* Decide whether to encode, decode, or pass through the char at P.
255 This used to be a macro, but it got a little too convoluted. */
256 static inline enum copy_method
257 decide_copy_method (const char *p)
261 if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2)))
263 /* %xx sequence: decode it, unless it would decode to an
264 unsafe or a reserved char; in that case, leave it as
266 char preempt = X2DIGITS_TO_NUM (*(p + 1), *(p + 2));
267 if (URL_UNSAFE_CHAR (preempt) || URL_RESERVED_CHAR (preempt))
268 return CM_PASSTHROUGH;
273 /* Garbled %.. sequence: encode `%'. */
276 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
279 return CM_PASSTHROUGH;
282 /* Translate a %-escaped (but possibly non-conformant) input string S
283 into a %-escaped (and conformant) output string. If no characters
284 are encoded or decoded, return the same string S; otherwise, return
285 a freshly allocated string with the new contents.
287 After a URL has been run through this function, the protocols that
288 use `%' as the quote character can use the resulting string as-is,
289 while those that don't call url_unescape() to get to the intended
290 data. This function is also stable: after an input string is
291 transformed the first time, all further transformations of the
292 result yield the same result string.
294 Let's discuss why this function is needed.
296 Imagine Wget is to retrieve `http://abc.xyz/abc def'. Since a raw
297 space character would mess up the HTTP request, it needs to be
300 GET /abc%20def HTTP/1.0
302 It appears that the unsafe chars need to be quoted, for example
303 with url_escape. But what if we're requested to download
304 `abc%20def'? url_escape transforms "%" to "%25", which would leave
305 us with `abc%2520def'. This is incorrect -- since %-escapes are
306 part of URL syntax, "%20" is the correct way to denote a literal
307 space on the Wget command line. This leaves us in the conclusion
308 that in that case Wget should not call url_escape, but leave the
311 And what if the requested URI is `abc%20 def'? If we call
312 url_escape, we end up with `/abc%2520%20def', which is almost
313 certainly not intended. If we don't call url_escape, we are left
314 with the embedded space and cannot complete the request. What the
315 user meant was for Wget to request `/abc%20%20def', and this is
316 where reencode_escapes kicks in.
318 Wget used to solve this by first decoding %-quotes, and then
319 encoding all the "unsafe" characters found in the resulting string.
320 This was wrong because it didn't preserve certain URL special
321 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
322 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
323 whether we considered `+' reserved (it is). One of these results
324 is inevitable because by the second step we would lose information
325 on whether the `+' was originally encoded or not. Both results
326 were wrong because in CGI parameters + means space, while %2B means
327 literal plus. reencode_escapes correctly translates the above to
328 "a%2B+b", i.e. returns the original string.
330 This function uses an algorithm proposed by Anon Sricharoenchai:
332 1. Encode all URL_UNSAFE and the "%" that are not followed by 2
335 2. Decode all "%XX" except URL_UNSAFE, URL_RESERVED (";/?:@=&") and
338 ...except that this code conflates the two steps, and decides
339 whether to encode, decode, or pass through each character in turn.
340 The function still uses two passes, but their logic is the same --
341 the first pass exists merely for the sake of allocation. Another
342 small difference is that we include `+' to URL_RESERVED.
346 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
348 "http://abc.xyz/%20%3F%2561%25aa%25%20a?a=a+a%2Ba&b=b%26c%3Dc"
352 "foo bar" -> "foo%20bar"
353 "foo%20bar" -> "foo%20bar"
354 "foo %20bar" -> "foo%20%20bar"
355 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
356 "foo%25%20bar" -> "foo%25%20bar"
357 "foo%2%20bar" -> "foo%252%20bar"
358 "foo+bar" -> "foo+bar" (plus is reserved!)
359 "foo%2b+bar" -> "foo%2b+bar" */
362 reencode_escapes (const char *s)
368 int encode_count = 0;
369 int decode_count = 0;
371 /* First, pass through the string to see if there's anything to do,
372 and to calculate the new length. */
373 for (p1 = s; *p1; p1++)
375 switch (decide_copy_method (p1))
388 if (!encode_count && !decode_count)
389 /* The string is good as it is. */
390 return (char *)s; /* C const model sucks. */
393 /* Each encoding adds two characters (hex digits), while each
394 decoding removes two characters. */
395 newlen = oldlen + 2 * (encode_count - decode_count);
396 newstr = xmalloc (newlen + 1);
403 switch (decide_copy_method (p1))
407 unsigned char c = *p1++;
409 *p2++ = XNUM_TO_DIGIT (c >> 4);
410 *p2++ = XNUM_TO_DIGIT (c & 0xf);
414 *p2++ = X2DIGITS_TO_NUM (p1[1], p1[2]);
415 p1 += 3; /* skip %xx */
422 assert (p2 - newstr == newlen);
426 /* Returns the scheme type if the scheme is supported, or
427 SCHEME_INVALID if not. */
430 url_scheme (const char *url)
434 for (i = 0; supported_schemes[i].leading_string; i++)
435 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
436 strlen (supported_schemes[i].leading_string)))
438 if (supported_schemes[i].enabled)
439 return (enum url_scheme) i;
441 return SCHEME_INVALID;
444 return SCHEME_INVALID;
447 #define SCHEME_CHAR(ch) (ISALNUM (ch) || (ch) == '-' || (ch) == '+')
449 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
450 currently implemented, it returns true if URL begins with
454 url_has_scheme (const char *url)
458 /* The first char must be a scheme char. */
459 if (!*p || !SCHEME_CHAR (*p))
462 /* Followed by 0 or more scheme chars. */
463 while (*p && SCHEME_CHAR (*p))
465 /* Terminated by ':'. */
470 scheme_default_port (enum url_scheme scheme)
472 return supported_schemes[scheme].default_port;
476 scheme_disable (enum url_scheme scheme)
478 supported_schemes[scheme].enabled = 0;
481 /* Skip the username and password, if present in the URL. The
482 function should *not* be called with the complete URL, but with the
483 portion after the scheme.
485 If no username and password are found, return URL. */
488 url_skip_credentials (const char *url)
490 /* Look for '@' that comes before terminators, such as '/', '?',
492 const char *p = (const char *)strpbrk (url, "@/?#;");
498 /* Parse credentials contained in [BEG, END). The region is expected
499 to have come from a URL and is unescaped. */
502 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
508 return 0; /* empty user name */
510 colon = memchr (beg, ':', end - beg);
512 return 0; /* again empty user name */
516 *passwd = strdupdelim (colon + 1, end);
518 url_unescape (*passwd);
525 *user = strdupdelim (beg, userend);
526 url_unescape (*user);
530 /* Used by main.c: detect URLs written using the "shorthand" URL forms
531 popularized by Netscape and NcFTP. HTTP shorthands look like this:
533 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
534 www.foo.com[:port] -> http://www.foo.com[:port]
536 FTP shorthands look like this:
538 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
539 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
541 If the URL needs not or cannot be rewritten, return NULL. */
544 rewrite_shorthand_url (const char *url)
548 if (url_has_scheme (url))
551 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
553 for (p = url; *p && *p != ':' && *p != '/'; p++)
563 /* If the characters after the colon and before the next slash
564 or end of string are all digits, it's HTTP. */
566 for (pp = p + 1; ISDIGIT (*pp); pp++)
568 if (digits > 0 && (*pp == '/' || *pp == '\0'))
571 /* Prepend "ftp://" to the entire URL... */
572 res = xmalloc (6 + strlen (url) + 1);
573 sprintf (res, "ftp://%s", url);
574 /* ...and replace ':' with '/'. */
575 res[6 + (p - url)] = '/';
582 /* Just prepend "http://" to what we have. */
583 res = xmalloc (7 + strlen (url) + 1);
584 sprintf (res, "http://%s", url);
589 static void split_path PARAMS ((const char *, char **, char **));
591 /* Like strpbrk, with the exception that it returns the pointer to the
592 terminating zero (end-of-string aka "eos") if no matching character
595 Although I normally balk at Gcc-specific optimizations, it probably
596 makes sense here: glibc has optimizations that detect strpbrk being
597 called with literal string as ACCEPT and inline the search. That
598 optimization is defeated if strpbrk is hidden within the call to
599 another function. (And no, making strpbrk_or_eos inline doesn't
600 help because the check for literal accept is in the
605 #define strpbrk_or_eos(s, accept) ({ \
606 char *SOE_p = strpbrk (s, accept); \
608 SOE_p = (char *)s + strlen (s); \
612 #else /* not __GNUC__ */
615 strpbrk_or_eos (const char *s, const char *accept)
617 char *p = strpbrk (s, accept);
619 p = (char *)s + strlen (s);
624 /* Turn STR into lowercase; return non-zero if a character was
628 lowercase_str (char *str)
635 *str = TOLOWER (*str);
640 static const char *parse_errors[] = {
641 #define PE_NO_ERROR 0
643 #define PE_UNSUPPORTED_SCHEME 1
644 N_("Unsupported scheme"),
645 #define PE_EMPTY_HOST 2
647 #define PE_BAD_PORT_NUMBER 3
648 N_("Bad port number"),
649 #define PE_INVALID_USER_NAME 4
650 N_("Invalid user name"),
651 #define PE_UNTERMINATED_IPV6_ADDRESS 5
652 N_("Unterminated IPv6 numeric address"),
653 #define PE_IPV6_NOT_SUPPORTED 6
654 N_("IPv6 addresses not supported"),
655 #define PE_INVALID_IPV6_ADDRESS 7
656 N_("Invalid IPv6 numeric address")
661 Return a new struct url if successful, NULL on error. In case of
662 error, and if ERROR is not NULL, also set *ERROR to the appropriate
665 url_parse (const char *url, int *error)
669 int path_modified, host_modified;
671 enum url_scheme scheme;
673 const char *uname_b, *uname_e;
674 const char *host_b, *host_e;
675 const char *path_b, *path_e;
676 const char *params_b, *params_e;
677 const char *query_b, *query_e;
678 const char *fragment_b, *fragment_e;
681 char *user = NULL, *passwd = NULL;
683 char *url_encoded = NULL;
687 scheme = url_scheme (url);
688 if (scheme == SCHEME_INVALID)
690 error_code = PE_UNSUPPORTED_SCHEME;
694 url_encoded = reencode_escapes (url);
697 p += strlen (supported_schemes[scheme].leading_string);
699 p = url_skip_credentials (p);
702 /* scheme://user:pass@host[:port]... */
705 /* We attempt to break down the URL into the components path,
706 params, query, and fragment. They are ordered like this:
708 scheme://host[:port][/path][;params][?query][#fragment] */
710 params_b = params_e = NULL;
711 query_b = query_e = NULL;
712 fragment_b = fragment_e = NULL;
718 /* Handle IPv6 address inside square brackets. Ideally we'd
719 just look for the terminating ']', but rfc2732 mandates
720 rejecting invalid IPv6 addresses. */
722 /* The address begins after '['. */
724 host_e = strchr (host_b, ']');
728 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
733 /* Check if the IPv6 address is valid. */
734 if (!is_valid_ipv6_address(host_b, host_e))
736 error_code = PE_INVALID_IPV6_ADDRESS;
740 /* Continue parsing after the closing ']'. */
743 error_code = PE_IPV6_NOT_SUPPORTED;
749 p = strpbrk_or_eos (p, ":/;?#");
753 if (host_b == host_e)
755 error_code = PE_EMPTY_HOST;
759 port = scheme_default_port (scheme);
762 const char *port_b, *port_e, *pp;
764 /* scheme://host:port/tralala */
768 p = strpbrk_or_eos (p, "/;?#");
771 /* Allow empty port, as per rfc2396. */
772 if (port_b != port_e)
774 for (port = 0, pp = port_b; pp < port_e; pp++)
778 /* http://host:12randomgarbage/blah */
780 error_code = PE_BAD_PORT_NUMBER;
783 port = 10 * port + (*pp - '0');
792 p = strpbrk_or_eos (p, ";?#");
797 /* Path is not allowed not to exist. */
805 p = strpbrk_or_eos (p, "?#");
812 p = strpbrk_or_eos (p, "#");
815 /* Hack that allows users to use '?' (a wildcard character) in
816 FTP URLs without it being interpreted as a query string
818 if (scheme == SCHEME_FTP)
820 query_b = query_e = NULL;
833 if (uname_b != uname_e)
835 /* http://user:pass@host */
837 /* uname_b uname_e */
838 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
840 error_code = PE_INVALID_USER_NAME;
845 u = xnew0 (struct url);
847 u->host = strdupdelim (host_b, host_e);
852 u->path = strdupdelim (path_b, path_e);
853 path_modified = path_simplify (u->path);
854 split_path (u->path, &u->dir, &u->file);
856 host_modified = lowercase_str (u->host);
858 /* Decode %HH sequences in host name. This is important not so much
859 to support %HH sequences, but to support binary characters (which
860 will have been converted to %HH by reencode_escapes). */
861 if (strchr (u->host, '%'))
863 url_unescape (u->host);
868 u->params = strdupdelim (params_b, params_e);
870 u->query = strdupdelim (query_b, query_e);
872 u->fragment = strdupdelim (fragment_b, fragment_e);
874 if (path_modified || u->fragment || host_modified || path_b == path_e)
876 /* If we suspect that a transformation has rendered what
877 url_string might return different from URL_ENCODED, rebuild
878 u->url using url_string. */
879 u->url = url_string (u, 0);
881 if (url_encoded != url)
882 xfree ((char *) url_encoded);
886 if (url_encoded == url)
887 u->url = xstrdup (url);
889 u->url = url_encoded;
896 /* Cleanup in case of error: */
897 if (url_encoded && url_encoded != url)
900 /* Transmit the error code to the caller, if the caller wants to
907 /* Return the error message string from ERROR_CODE, which should have
908 been retrieved from url_parse. The error message is translated. */
911 url_error (int error_code)
913 assert (error_code >= 0 && error_code < countof (parse_errors));
914 return _(parse_errors[error_code]);
917 /* Split PATH into DIR and FILE. PATH comes from the URL and is
918 expected to be URL-escaped.
920 The path is split into directory (the part up to the last slash)
921 and file (the part after the last slash), which are subsequently
925 "foo/bar/baz" "foo/bar" "baz"
926 "foo/bar/" "foo/bar" ""
928 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
930 DIR and FILE are freshly allocated. */
933 split_path (const char *path, char **dir, char **file)
935 char *last_slash = strrchr (path, '/');
939 *file = xstrdup (path);
943 *dir = strdupdelim (path, last_slash);
944 *file = xstrdup (last_slash + 1);
947 url_unescape (*file);
950 /* Note: URL's "full path" is the path with the query string and
951 params appended. The "fragment" (#foo) is intentionally ignored,
952 but that might be changed. For example, if the original URL was
953 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
954 the full path will be "/foo/bar/baz;bullshit?querystring". */
956 /* Return the length of the full path, without the terminating
960 full_path_length (const struct url *url)
964 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
975 /* Write out the full path. */
978 full_path_write (const struct url *url, char *where)
980 #define FROB(el, chr) do { \
981 char *f_el = url->el; \
983 int l = strlen (f_el); \
985 memcpy (where, f_el, l); \
997 /* Public function for getting the "full path". E.g. if u->path is
998 "foo/bar" and u->query is "param=value", full_path will be
999 "/foo/bar?param=value". */
1002 url_full_path (const struct url *url)
1004 int length = full_path_length (url);
1005 char *full_path = (char *)xmalloc(length + 1);
1007 full_path_write (url, full_path);
1008 full_path[length] = '\0';
1013 /* Escape unsafe and reserved characters, except for the slash
1017 url_escape_dir (const char *dir)
1019 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1024 /* Unescape slashes in NEWDIR. */
1026 h = newdir; /* hare */
1027 t = newdir; /* tortoise */
1029 for (; *h; h++, t++)
1031 /* url_escape_1 having converted '/' to "%2F" exactly. */
1032 if (*h == '%' && h[1] == '2' && h[2] == 'F')
1045 /* Sync u->path and u->url with u->dir and u->file. Called after
1046 u->file or u->dir have been changed, typically by the FTP code. */
1049 sync_path (struct url *u)
1051 char *newpath, *efile, *edir;
1055 /* u->dir and u->file are not escaped. URL-escape them before
1056 reassembling them into u->path. That way, if they contain
1057 separators like '?' or even if u->file contains slashes, the
1058 path will be correctly assembled. (u->file can contain slashes
1059 if the URL specifies it with %2f, or if an FTP server returns
1061 edir = url_escape_dir (u->dir);
1062 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1065 newpath = xstrdup (efile);
1068 int dirlen = strlen (edir);
1069 int filelen = strlen (efile);
1071 /* Copy "DIR/FILE" to newpath. */
1072 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1073 memcpy (p, edir, dirlen);
1076 memcpy (p, efile, filelen);
1085 if (efile != u->file)
1088 /* Regenerate u->url as well. */
1090 u->url = url_string (u, 0);
1093 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1094 This way we can sync u->path and u->url when they get changed. */
1097 url_set_dir (struct url *url, const char *newdir)
1100 url->dir = xstrdup (newdir);
1105 url_set_file (struct url *url, const char *newfile)
1108 url->file = xstrdup (newfile);
1113 url_free (struct url *url)
1119 xfree_null (url->params);
1120 xfree_null (url->query);
1121 xfree_null (url->fragment);
1122 xfree_null (url->user);
1123 xfree_null (url->passwd);
1131 /* Create all the necessary directories for PATH (a file). Calls
1132 mkdirhier() internally. */
1134 mkalldirs (const char *path)
1141 p = path + strlen (path);
1142 for (; *p != '/' && p != path; p--)
1145 /* Don't create if it's just a file. */
1146 if ((p == path) && (*p != '/'))
1148 t = strdupdelim (path, p);
1150 /* Check whether the directory exists. */
1151 if ((stat (t, &st) == 0))
1153 if (S_ISDIR (st.st_mode))
1160 /* If the dir exists as a file name, remove it first. This
1161 is *only* for Wget to work with buggy old CERN http
1162 servers. Here is the scenario: When Wget tries to
1163 retrieve a directory without a slash, e.g.
1164 http://foo/bar (bar being a directory), CERN server will
1165 not redirect it too http://foo/bar/ -- it will generate a
1166 directory listing containing links to bar/file1,
1167 bar/file2, etc. Wget will lose because it saves this
1168 HTML listing to a file `bar', so it cannot create the
1169 directory. To work around this, if the file of the same
1170 name exists, we just remove it and create the directory
1172 DEBUGP (("Removing %s because of directory danger!\n", t));
1176 res = make_directory (t);
1178 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1183 /* Functions for constructing the file name out of URL components. */
1185 /* A growable string structure, used by url_file_name and friends.
1186 This should perhaps be moved to utils.c.
1188 The idea is to have a convenient and efficient way to construct a
1189 string by having various functions append data to it. Instead of
1190 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1191 functions in questions, we pass the pointer to this struct. */
1199 /* Ensure that the string can accept APPEND_COUNT more characters past
1200 the current TAIL position. If necessary, this will grow the string
1201 and update its allocated size. If the string is already large
1202 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1203 #define GROW(g, append_size) do { \
1204 struct growable *G_ = g; \
1205 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1208 /* Return the tail position of the string. */
1209 #define TAIL(r) ((r)->base + (r)->tail)
1211 /* Move the tail position by APPEND_COUNT characters. */
1212 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1214 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1218 append_string (const char *str, struct growable *dest)
1220 int l = strlen (str);
1222 memcpy (TAIL (dest), str, l);
1223 TAIL_INCR (dest, l);
1226 /* Append CH to DEST. For example, append_char (0, DEST)
1227 zero-terminates DEST. */
1230 append_char (char ch, struct growable *dest)
1234 TAIL_INCR (dest, 1);
1238 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1239 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1240 filechr_control = 4 /* a control character, e.g. 0-31 */
1243 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1245 /* Shorthands for the table: */
1246 #define U filechr_not_unix
1247 #define W filechr_not_windows
1248 #define C filechr_control
1253 /* Table of characters unsafe under various conditions (see above).
1255 Arguably we could also claim `%' to be unsafe, since we use it as
1256 the escape character. If we ever want to be able to reliably
1257 translate file name back to URL, this would become important
1258 crucial. Right now, it's better to be minimal in escaping. */
1260 const static unsigned char filechr_table[256] =
1262 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1263 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1264 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1265 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1266 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1267 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1268 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1269 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1270 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1271 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1272 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1273 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1274 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1275 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1276 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1277 0, 0, 0, 0, 0, 0, 0, 0, /* x y z { | } ~ DEL */
1279 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1280 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1281 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1282 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1284 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1285 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1286 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1287 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1295 /* FN_PORT_SEP is the separator between host and port in file names
1296 for non-standard port numbers. On Unix this is normally ':', as in
1297 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1298 because Windows can't handle ':' in file names. */
1299 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1301 /* FN_QUERY_SEP is the separator between the file name and the URL
1302 query, normally '?'. Since Windows cannot handle '?' as part of
1303 file name, we use '@' instead there. */
1304 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1306 /* Quote path element, characters in [b, e), as file name, and append
1307 the quoted string to DEST. Each character is quoted as per
1308 file_unsafe_char and the corresponding table.
1310 If ESCAPED_P is non-zero, the path element is considered to be
1311 URL-escaped and will be unescaped prior to inspection. */
1314 append_uri_pathel (const char *b, const char *e, int escaped_p,
1315 struct growable *dest)
1321 if (opt.restrict_files_os == restrict_unix)
1322 mask = filechr_not_unix;
1324 mask = filechr_not_windows;
1325 if (opt.restrict_files_ctrl)
1326 mask |= filechr_control;
1328 /* Copy [b, e) to PATHEL and URL-unescape it. */
1332 BOUNDED_TO_ALLOCA (b, e, unescaped);
1333 url_unescape (unescaped);
1335 e = unescaped + strlen (unescaped);
1338 /* Defang ".." when found as component of path. Remember that path
1339 comes from the URL and might contain malicious input. */
1340 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1346 /* Walk the PATHEL string and check how many characters we'll need
1349 for (p = b; p < e; p++)
1350 if (FILE_CHAR_TEST (*p, mask))
1353 /* Calculate the length of the output string. e-b is the input
1354 string length. Each quoted char introduces two additional
1355 characters in the string, hence 2*quoted. */
1356 outlen = (e - b) + (2 * quoted);
1357 GROW (dest, outlen);
1361 /* If there's nothing to quote, we can simply append the string
1362 without processing it again. */
1363 memcpy (TAIL (dest), b, outlen);
1367 char *q = TAIL (dest);
1368 for (p = b; p < e; p++)
1370 if (!FILE_CHAR_TEST (*p, mask))
1374 unsigned char ch = *p;
1376 *q++ = XNUM_TO_DIGIT (ch >> 4);
1377 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1380 assert (q - TAIL (dest) == outlen);
1382 TAIL_INCR (dest, outlen);
1385 /* Append to DEST the directory structure that corresponds the
1386 directory part of URL's path. For example, if the URL is
1387 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1389 Each path element ("dir1" and "dir2" in the above example) is
1390 examined, url-unescaped, and re-escaped as file name element.
1392 Additionally, it cuts as many directories from the path as
1393 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1394 will produce "bar" for the above example. For 2 or more, it will
1397 Each component of the path is quoted for use as file name. */
1400 append_dir_structure (const struct url *u, struct growable *dest)
1402 char *pathel, *next;
1403 int cut = opt.cut_dirs;
1405 /* Go through the path components, de-URL-quote them, and quote them
1406 (if necessary) as file names. */
1409 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1414 /* Ignore empty pathels. */
1418 append_char ('/', dest);
1419 append_uri_pathel (pathel, next, 1, dest);
1423 /* Return a unique file name that matches the given URL as good as
1424 possible. Does not create directories on the file system. */
1427 url_file_name (const struct url *u)
1429 struct growable fnres; /* stands for "file name result" */
1431 const char *u_file, *u_query;
1432 char *fname, *unique;
1438 /* Start with the directory prefix, if specified. */
1440 append_string (opt.dir_prefix, &fnres);
1442 /* If "dirstruct" is turned on (typically the case with -r), add
1443 the host and port (unless those have been turned off) and
1444 directory structure. */
1447 if (opt.protocol_directories)
1450 append_char ('/', &fnres);
1451 append_string (supported_schemes[u->scheme].name, &fnres);
1453 if (opt.add_hostdir)
1456 append_char ('/', &fnres);
1457 if (0 != strcmp (u->host, ".."))
1458 append_string (u->host, &fnres);
1460 /* Host name can come from the network; malicious DNS may
1461 allow ".." to be resolved, causing us to write to
1462 "../<file>". Defang such host names. */
1463 append_string ("%2E%2E", &fnres);
1464 if (u->port != scheme_default_port (u->scheme))
1467 number_to_string (portstr, u->port);
1468 append_char (FN_PORT_SEP, &fnres);
1469 append_string (portstr, &fnres);
1473 append_dir_structure (u, &fnres);
1476 /* Add the file name. */
1478 append_char ('/', &fnres);
1479 u_file = *u->file ? u->file : "index.html";
1480 append_uri_pathel (u_file, u_file + strlen (u_file), 0, &fnres);
1482 /* Append "?query" to the file name. */
1483 u_query = u->query && *u->query ? u->query : NULL;
1486 append_char (FN_QUERY_SEP, &fnres);
1487 append_uri_pathel (u_query, u_query + strlen (u_query), 1, &fnres);
1490 /* Zero-terminate the file name. */
1491 append_char ('\0', &fnres);
1495 /* Check the cases in which the unique extensions are not used:
1496 1) Clobbering is turned off (-nc).
1497 2) Retrieval with regetting.
1498 3) Timestamping is used.
1499 4) Hierarchy is built.
1501 The exception is the case when file does exist and is a
1502 directory (see `mkalldirs' for explanation). */
1504 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1505 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1508 unique = unique_name (fname, 1);
1509 if (unique != fname)
1514 /* Resolve "." and ".." elements of PATH by destructively modifying
1515 PATH and return non-zero if PATH has been modified, zero otherwise.
1517 The algorithm is in spirit similar to the one described in rfc1808,
1518 although implemented differently, in one pass. To recap, path
1519 elements containing only "." are removed, and ".." is taken to mean
1520 "back up one element". Single leading and trailing slashes are
1523 This function does not handle URL escapes explicitly. If you're
1524 passing paths from URLs, make sure to unquote "%2e" and "%2E" to
1525 ".", so that this function can find the dots. (Wget's URL parser
1526 calls reencode_escapes, which see.)
1528 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1529 test examples are provided below. If you change anything in this
1530 function, run test_path_simplify to make sure you haven't broken a
1534 path_simplify (char *path)
1536 char *h = path; /* hare */
1537 char *t = path; /* tortoise */
1538 char *beg = path; /* boundary for backing the tortoise */
1539 char *end = path + strlen (path);
1543 /* Hare should be at the beginning of a path element. */
1545 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1550 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1552 /* Handle "../" by retreating the tortoise by one path
1553 element -- but not past beggining. */
1556 /* Move backwards until T hits the beginning of the
1557 previous path element or the beginning of path. */
1558 for (--t; t > beg && t[-1] != '/'; t--)
1563 /* If we're at the beginning, copy the "../" literally
1564 move the beginning so a later ".." doesn't remove
1574 /* A regular path element. If H hasn't advanced past T,
1575 simply skip to the next path element. Otherwise, copy
1576 the path element until the next slash. */
1579 /* Skip the path element, including the slash. */
1580 while (h < end && *h != '/')
1587 /* Copy the path element, including the final slash. */
1588 while (h < end && *h != '/')
1602 /* Return the length of URL's path. Path is considered to be
1603 terminated by one of '?', ';', '#', or by the end of the
1607 path_length (const char *url)
1609 const char *q = strpbrk_or_eos (url, "?;#");
1613 /* Find the last occurrence of character C in the range [b, e), or
1614 NULL, if none are present. We might want to use memrchr (a GNU
1615 extension) under GNU libc. */
1618 find_last_char (const char *b, const char *e, char c)
1626 /* Merge BASE with LINK and return the resulting URI.
1628 Either of the URIs may be absolute or relative, complete with the
1629 host name, or path only. This tries to reasonably handle all
1630 foreseeable cases. It only employs minimal URL parsing, without
1631 knowledge of the specifics of schemes.
1633 I briefly considered making this function call path_simplify after
1634 the merging process, as rfc1738 seems to suggest. This is a bad
1635 idea for several reasons: 1) it complexifies the code, and 2)
1636 url_parse has to simplify path anyway, so it's wasteful to boot. */
1639 uri_merge (const char *base, const char *link)
1645 if (url_has_scheme (link))
1646 return xstrdup (link);
1648 /* We may not examine BASE past END. */
1649 end = base + path_length (base);
1650 linklength = strlen (link);
1654 /* Empty LINK points back to BASE, query string and all. */
1655 return xstrdup (base);
1657 else if (*link == '?')
1659 /* LINK points to the same location, but changes the query
1660 string. Examples: */
1661 /* uri_merge("path", "?new") -> "path?new" */
1662 /* uri_merge("path?foo", "?new") -> "path?new" */
1663 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1664 /* uri_merge("path#foo", "?new") -> "path?new" */
1665 int baselength = end - base;
1666 merge = xmalloc (baselength + linklength + 1);
1667 memcpy (merge, base, baselength);
1668 memcpy (merge + baselength, link, linklength);
1669 merge[baselength + linklength] = '\0';
1671 else if (*link == '#')
1673 /* uri_merge("path", "#new") -> "path#new" */
1674 /* uri_merge("path#foo", "#new") -> "path#new" */
1675 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1676 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1678 const char *end1 = strchr (base, '#');
1680 end1 = base + strlen (base);
1681 baselength = end1 - base;
1682 merge = xmalloc (baselength + linklength + 1);
1683 memcpy (merge, base, baselength);
1684 memcpy (merge + baselength, link, linklength);
1685 merge[baselength + linklength] = '\0';
1687 else if (*link == '/' && *(link + 1) == '/')
1689 /* LINK begins with "//" and so is a net path: we need to
1690 replace everything after (and including) the double slash
1693 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1694 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1695 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1699 const char *start_insert;
1701 /* Look for first slash. */
1702 slash = memchr (base, '/', end - base);
1703 /* If found slash and it is a double slash, then replace
1704 from this point, else default to replacing from the
1706 if (slash && *(slash + 1) == '/')
1707 start_insert = slash;
1709 start_insert = base;
1711 span = start_insert - base;
1712 merge = (char *)xmalloc (span + linklength + 1);
1714 memcpy (merge, base, span);
1715 memcpy (merge + span, link, linklength);
1716 merge[span + linklength] = '\0';
1718 else if (*link == '/')
1720 /* LINK is an absolute path: we need to replace everything
1721 after (and including) the FIRST slash with LINK.
1723 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1724 "/qux/xyzzy", our result should be
1725 "http://host/qux/xyzzy". */
1728 const char *start_insert = NULL; /* for gcc to shut up. */
1729 const char *pos = base;
1730 int seen_slash_slash = 0;
1731 /* We're looking for the first slash, but want to ignore
1734 slash = memchr (pos, '/', end - pos);
1735 if (slash && !seen_slash_slash)
1736 if (*(slash + 1) == '/')
1739 seen_slash_slash = 1;
1743 /* At this point, SLASH is the location of the first / after
1744 "//", or the first slash altogether. START_INSERT is the
1745 pointer to the location where LINK will be inserted. When
1746 examining the last two examples, keep in mind that LINK
1749 if (!slash && !seen_slash_slash)
1750 /* example: "foo" */
1752 start_insert = base;
1753 else if (!slash && seen_slash_slash)
1754 /* example: "http://foo" */
1757 else if (slash && !seen_slash_slash)
1758 /* example: "foo/bar" */
1760 start_insert = base;
1761 else if (slash && seen_slash_slash)
1762 /* example: "http://something/" */
1764 start_insert = slash;
1766 span = start_insert - base;
1767 merge = (char *)xmalloc (span + linklength + 1);
1769 memcpy (merge, base, span);
1770 memcpy (merge + span, link, linklength);
1771 merge[span + linklength] = '\0';
1775 /* LINK is a relative URL: we need to replace everything
1776 after last slash (possibly empty) with LINK.
1778 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1779 our result should be "whatever/foo/qux/xyzzy". */
1780 int need_explicit_slash = 0;
1782 const char *start_insert;
1783 const char *last_slash = find_last_char (base, end, '/');
1786 /* No slash found at all. Replace what we have with LINK. */
1787 start_insert = base;
1789 else if (last_slash && last_slash >= base + 2
1790 && last_slash[-2] == ':' && last_slash[-1] == '/')
1792 /* example: http://host" */
1794 start_insert = end + 1;
1795 need_explicit_slash = 1;
1799 /* example: "whatever/foo/bar" */
1801 start_insert = last_slash + 1;
1804 span = start_insert - base;
1805 merge = (char *)xmalloc (span + linklength + 1);
1807 memcpy (merge, base, span);
1808 if (need_explicit_slash)
1809 merge[span - 1] = '/';
1810 memcpy (merge + span, link, linklength);
1811 merge[span + linklength] = '\0';
1817 #define APPEND(p, s) do { \
1818 int len = strlen (s); \
1819 memcpy (p, s, len); \
1823 /* Use this instead of password when the actual password is supposed
1824 to be hidden. We intentionally use a generic string without giving
1825 away the number of characters in the password, like previous
1827 #define HIDDEN_PASSWORD "*password*"
1829 /* Recreate the URL string from the data in URL.
1831 If HIDE is non-zero (as it is when we're calling this on a URL we
1832 plan to print, but not when calling it to canonicalize a URL for
1833 use within the program), password will be hidden. Unsafe
1834 characters in the URL will be quoted. */
1837 url_string (const struct url *url, int hide_password)
1841 char *quoted_user = NULL, *quoted_passwd = NULL;
1843 int scheme_port = supported_schemes[url->scheme].default_port;
1844 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1845 int fplen = full_path_length (url);
1847 int brackets_around_host;
1849 assert (scheme_str != NULL);
1851 /* Make sure the user name and password are quoted. */
1854 quoted_user = url_escape_allow_passthrough (url->user);
1858 quoted_passwd = HIDDEN_PASSWORD;
1860 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1864 /* Numeric IPv6 addresses can contain ':' and need to be quoted with
1866 brackets_around_host = strchr (url->host, ':') != NULL;
1868 size = (strlen (scheme_str)
1869 + strlen (url->host)
1870 + (brackets_around_host ? 2 : 0)
1873 if (url->port != scheme_port)
1874 size += 1 + numdigit (url->port);
1877 size += 1 + strlen (quoted_user);
1879 size += 1 + strlen (quoted_passwd);
1882 p = result = xmalloc (size);
1884 APPEND (p, scheme_str);
1887 APPEND (p, quoted_user);
1891 APPEND (p, quoted_passwd);
1896 if (brackets_around_host)
1898 APPEND (p, url->host);
1899 if (brackets_around_host)
1901 if (url->port != scheme_port)
1904 p = number_to_string (p, url->port);
1907 full_path_write (url, p);
1911 assert (p - result == size);
1913 if (quoted_user && quoted_user != url->user)
1914 xfree (quoted_user);
1915 if (quoted_passwd && !hide_password
1916 && quoted_passwd != url->passwd)
1917 xfree (quoted_passwd);
1922 /* Return non-zero if scheme a is similar to scheme b.
1924 Schemes are similar if they are equal. If SSL is supported, schemes
1925 are also similar if one is http (SCHEME_HTTP) and the other is https
1928 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1933 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1934 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1941 /* Debugging and testing support for path_simplify. */
1943 /* Debug: run path_simplify on PATH and return the result in a new
1944 string. Useful for calling from the debugger. */
1948 char *copy = xstrdup (path);
1949 path_simplify (copy);
1954 run_test (char *test, char *expected_result, int expected_change)
1956 char *test_copy = xstrdup (test);
1957 int modified = path_simplify (test_copy);
1959 if (0 != strcmp (test_copy, expected_result))
1961 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
1962 test, expected_result, test_copy);
1964 if (modified != expected_change)
1966 if (expected_change == 1)
1967 printf ("Expected modification with path_simplify(\"%s\").\n",
1970 printf ("Expected no modification with path_simplify(\"%s\").\n",
1977 test_path_simplify (void)
1980 char *test, *result;
1987 { "../", "../", 0 },
1988 { "foo", "foo", 0 },
1989 { "foo/bar", "foo/bar", 0 },
1990 { "foo///bar", "foo///bar", 0 },
1991 { "foo/.", "foo/", 1 },
1992 { "foo/./", "foo/", 1 },
1993 { "foo./", "foo./", 0 },
1994 { "foo/../bar", "bar", 1 },
1995 { "foo/../bar/", "bar/", 1 },
1996 { "foo/bar/..", "foo/", 1 },
1997 { "foo/bar/../x", "foo/x", 1 },
1998 { "foo/bar/../x/", "foo/x/", 1 },
1999 { "foo/..", "", 1 },
2000 { "foo/../..", "..", 1 },
2001 { "foo/../../..", "../..", 1 },
2002 { "foo/../../bar/../../baz", "../../baz", 1 },
2003 { "a/b/../../c", "c", 1 },
2004 { "./a/../b", "b", 1 }
2008 for (i = 0; i < countof (tests); i++)
2010 char *test = tests[i].test;
2011 char *expected_result = tests[i].result;
2012 int expected_change = tests[i].should_modify;
2013 run_test (test, expected_result, expected_change);