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>
62 /* Supported schemes: */
63 static struct scheme_data supported_schemes[] =
65 { "http://", DEFAULT_HTTP_PORT, 1 },
67 { "https://", DEFAULT_HTTPS_PORT, 1 },
69 { "ftp://", DEFAULT_FTP_PORT, 1 },
75 /* Forward declarations: */
77 static int path_simplify PARAMS ((char *));
79 /* Support for encoding and decoding of URL strings. We determine
80 whether a character is unsafe through static table lookup. This
81 code assumes ASCII character set and 8-bit chars. */
84 /* rfc1738 reserved chars, preserved from encoding. */
87 /* rfc1738 unsafe chars, plus some more. */
91 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
92 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
93 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
95 /* Shorthands for the table: */
96 #define R urlchr_reserved
97 #define U urlchr_unsafe
100 const static unsigned char urlchr_table[256] =
102 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
103 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
104 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
105 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
106 U, 0, U, RU, 0, U, R, 0, /* SP ! " # $ % & ' */
107 0, 0, 0, R, 0, 0, 0, R, /* ( ) * + , - . / */
108 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
109 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
110 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
111 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
112 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
113 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
114 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
115 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
116 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
117 0, 0, 0, U, U, U, U, U, /* x y z { | } ~ DEL */
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,
122 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,
127 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
133 /* URL-unescape the string S.
135 This is done by transforming the sequences "%HH" to the character
136 represented by the hexadecimal digits HH. If % is not followed by
137 two hexadecimal digits, it is inserted literally.
139 The transformation is done in place. If you need the original
140 string intact, make a copy before calling this function. */
143 url_unescape (char *s)
145 char *t = s; /* t - tortoise */
146 char *h = s; /* h - hare */
157 /* Do nothing if '%' is not followed by two hex digits. */
158 if (!h[1] || !h[2] || !(ISXDIGIT (h[1]) && ISXDIGIT (h[2])))
160 *t = X2DIGITS_TO_NUM (h[1], h[2]);
167 /* The core of url_escape_* functions. Escapes the characters that
168 match the provided mask in urlchr_table.
170 If ALLOW_PASSTHROUGH is non-zero, a string with no unsafe chars
171 will be returned unchanged. If ALLOW_PASSTHROUGH is zero, a
172 freshly allocated string will be returned in all cases. */
175 url_escape_1 (const char *s, unsigned char mask, int allow_passthrough)
182 for (p1 = s; *p1; p1++)
183 if (urlchr_test (*p1, mask))
184 addition += 2; /* Two more characters (hex digits) */
187 return allow_passthrough ? (char *)s : xstrdup (s);
189 newlen = (p1 - s) + addition;
190 newstr = (char *)xmalloc (newlen + 1);
196 /* Quote the characters that match the test mask. */
197 if (urlchr_test (*p1, mask))
199 unsigned char c = *p1++;
201 *p2++ = XNUM_TO_digit (c >> 4);
202 *p2++ = XNUM_TO_digit (c & 0xf);
207 assert (p2 - newstr == newlen);
213 /* URL-escape the unsafe characters (see urlchr_table) in a given
214 string, returning a freshly allocated string. */
217 url_escape (const char *s)
219 return url_escape_1 (s, urlchr_unsafe, 0);
222 /* URL-escape the unsafe characters (see urlchr_table) in a given
223 string. If no characters are unsafe, S is returned. */
226 url_escape_allow_passthrough (const char *s)
228 return url_escape_1 (s, urlchr_unsafe, 1);
231 enum copy_method { CM_DECODE, CM_ENCODE, CM_PASSTHROUGH };
233 /* Decide whether to encode, decode, or pass through the char at P.
234 This used to be a macro, but it got a little too convoluted. */
235 static inline enum copy_method
236 decide_copy_method (const char *p)
240 if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2)))
242 /* %xx sequence: decode it, unless it would decode to an
243 unsafe or a reserved char; in that case, leave it as
245 char preempt = X2DIGITS_TO_NUM (*(p + 1), *(p + 2));
246 if (URL_UNSAFE_CHAR (preempt) || URL_RESERVED_CHAR (preempt))
247 return CM_PASSTHROUGH;
252 /* Garbled %.. sequence: encode `%'. */
255 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
258 return CM_PASSTHROUGH;
261 /* Translate a %-escaped (but possibly non-conformant) input string S
262 into a %-escaped (and conformant) output string. If no characters
263 are encoded or decoded, return the same string S; otherwise, return
264 a freshly allocated string with the new contents.
266 After a URL has been run through this function, the protocols that
267 use `%' as the quote character can use the resulting string as-is,
268 while those that don't call url_unescape() to get to the intended
269 data. This function is also stable: after an input string is
270 transformed the first time, all further transformations of the
271 result yield the same result string.
273 Let's discuss why this function is needed.
275 Imagine Wget is to retrieve `http://abc.xyz/abc def'. Since a raw
276 space character would mess up the HTTP request, it needs to be
279 GET /abc%20def HTTP/1.0
281 It appears that the unsafe chars need to be quoted, for example
282 with url_escape. But what if we're requested to download
283 `abc%20def'? url_escape transforms "%" to "%25", which would leave
284 us with `abc%2520def'. This is incorrect -- since %-escapes are
285 part of URL syntax, "%20" is the correct way to denote a literal
286 space on the Wget command line. This leaves us in the conclusion
287 that in that case Wget should not call url_escape, but leave the
290 And what if the requested URI is `abc%20 def'? If we call
291 url_escape, we end up with `/abc%2520%20def', which is almost
292 certainly not intended. If we don't call url_escape, we are left
293 with the embedded space and cannot complete the request. What the
294 user meant was for Wget to request `/abc%20%20def', and this is
295 where reencode_escapes kicks in.
297 Wget used to solve this by first decoding %-quotes, and then
298 encoding all the "unsafe" characters found in the resulting string.
299 This was wrong because it didn't preserve certain URL special
300 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
301 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
302 whether we considered `+' reserved (it is). One of these results
303 is inevitable because by the second step we would lose information
304 on whether the `+' was originally encoded or not. Both results
305 were wrong because in CGI parameters + means space, while %2B means
306 literal plus. reencode_escapes correctly translates the above to
307 "a%2B+b", i.e. returns the original string.
309 This function uses an algorithm proposed by Anon Sricharoenchai:
311 1. Encode all URL_UNSAFE and the "%" that are not followed by 2
314 2. Decode all "%XX" except URL_UNSAFE, URL_RESERVED (";/?:@=&") and
317 ...except that this code conflates the two steps, and decides
318 whether to encode, decode, or pass through each character in turn.
319 The function still uses two passes, but their logic is the same --
320 the first pass exists merely for the sake of allocation. Another
321 small difference is that we include `+' to URL_RESERVED.
325 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
327 "http://abc.xyz/%20%3F%2561%25aa%25%20a?a=a+a%2Ba&b=b%26c%3Dc"
331 "foo bar" -> "foo%20bar"
332 "foo%20bar" -> "foo%20bar"
333 "foo %20bar" -> "foo%20%20bar"
334 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
335 "foo%25%20bar" -> "foo%25%20bar"
336 "foo%2%20bar" -> "foo%252%20bar"
337 "foo+bar" -> "foo+bar" (plus is reserved!)
338 "foo%2b+bar" -> "foo%2b+bar" */
341 reencode_escapes (const char *s)
347 int encode_count = 0;
348 int decode_count = 0;
350 /* First, pass through the string to see if there's anything to do,
351 and to calculate the new length. */
352 for (p1 = s; *p1; p1++)
354 switch (decide_copy_method (p1))
367 if (!encode_count && !decode_count)
368 /* The string is good as it is. */
369 return (char *)s; /* C const model sucks. */
372 /* Each encoding adds two characters (hex digits), while each
373 decoding removes two characters. */
374 newlen = oldlen + 2 * (encode_count - decode_count);
375 newstr = xmalloc (newlen + 1);
382 switch (decide_copy_method (p1))
386 unsigned char c = *p1++;
388 *p2++ = XNUM_TO_DIGIT (c >> 4);
389 *p2++ = XNUM_TO_DIGIT (c & 0xf);
393 *p2++ = X2DIGITS_TO_NUM (p1[1], p1[2]);
394 p1 += 3; /* skip %xx */
401 assert (p2 - newstr == newlen);
405 /* Returns the scheme type if the scheme is supported, or
406 SCHEME_INVALID if not. */
409 url_scheme (const char *url)
413 for (i = 0; supported_schemes[i].leading_string; i++)
414 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
415 strlen (supported_schemes[i].leading_string)))
417 if (supported_schemes[i].enabled)
418 return (enum url_scheme) i;
420 return SCHEME_INVALID;
423 return SCHEME_INVALID;
426 #define SCHEME_CHAR(ch) (ISALNUM (ch) || (ch) == '-' || (ch) == '+')
428 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
429 currently implemented, it returns true if URL begins with
433 url_has_scheme (const char *url)
437 /* The first char must be a scheme char. */
438 if (!*p || !SCHEME_CHAR (*p))
441 /* Followed by 0 or more scheme chars. */
442 while (*p && SCHEME_CHAR (*p))
444 /* Terminated by ':'. */
449 scheme_default_port (enum url_scheme scheme)
451 return supported_schemes[scheme].default_port;
455 scheme_disable (enum url_scheme scheme)
457 supported_schemes[scheme].enabled = 0;
460 /* Skip the username and password, if present here. The function
461 should *not* be called with the complete URL, but with the part
462 right after the scheme.
464 If no username and password are found, return 0. */
467 url_skip_credentials (const char *url)
469 /* Look for '@' that comes before terminators, such as '/', '?',
471 const char *p = (const char *)strpbrk (url, "@/?#;");
477 /* Parse credentials contained in [BEG, END). The region is expected
478 to have come from a URL and is unescaped. */
481 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
487 return 0; /* empty user name */
489 colon = memchr (beg, ':', end - beg);
491 return 0; /* again empty user name */
495 *passwd = strdupdelim (colon + 1, end);
497 url_unescape (*passwd);
504 *user = strdupdelim (beg, userend);
505 url_unescape (*user);
509 /* Used by main.c: detect URLs written using the "shorthand" URL forms
510 popularized by Netscape and NcFTP. HTTP shorthands look like this:
512 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
513 www.foo.com[:port] -> http://www.foo.com[:port]
515 FTP shorthands look like this:
517 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
518 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
520 If the URL needs not or cannot be rewritten, return NULL. */
523 rewrite_shorthand_url (const char *url)
527 if (url_has_scheme (url))
530 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
532 for (p = url; *p && *p != ':' && *p != '/'; p++)
542 /* If the characters after the colon and before the next slash
543 or end of string are all digits, it's HTTP. */
545 for (pp = p + 1; ISDIGIT (*pp); pp++)
547 if (digits > 0 && (*pp == '/' || *pp == '\0'))
550 /* Prepend "ftp://" to the entire URL... */
551 res = xmalloc (6 + strlen (url) + 1);
552 sprintf (res, "ftp://%s", url);
553 /* ...and replace ':' with '/'. */
554 res[6 + (p - url)] = '/';
561 /* Just prepend "http://" to what we have. */
562 res = xmalloc (7 + strlen (url) + 1);
563 sprintf (res, "http://%s", url);
568 static void split_path PARAMS ((const char *, char **, char **));
570 /* Like strpbrk, with the exception that it returns the pointer to the
571 terminating zero (end-of-string aka "eos") if no matching character
574 Although I normally balk at Gcc-specific optimizations, it probably
575 makes sense here: glibc has optimizations that detect strpbrk being
576 called with literal string as ACCEPT and inline the search. That
577 optimization is defeated if strpbrk is hidden within the call to
578 another function. (And no, making strpbrk_or_eos inline doesn't
579 help because the check for literal accept is in the
584 #define strpbrk_or_eos(s, accept) ({ \
585 char *SOE_p = strpbrk (s, accept); \
587 SOE_p = (char *)s + strlen (s); \
591 #else /* not __GNUC__ */
594 strpbrk_or_eos (const char *s, const char *accept)
596 char *p = strpbrk (s, accept);
598 p = (char *)s + strlen (s);
603 /* Turn STR into lowercase; return non-zero if a character was
607 lowercase_str (char *str)
614 *str = TOLOWER (*str);
619 static char *parse_errors[] = {
620 #define PE_NO_ERROR 0
622 #define PE_UNSUPPORTED_SCHEME 1
623 "Unsupported scheme",
624 #define PE_EMPTY_HOST 2
626 #define PE_BAD_PORT_NUMBER 3
628 #define PE_INVALID_USER_NAME 4
630 #define PE_UNTERMINATED_IPV6_ADDRESS 5
631 "Unterminated IPv6 numeric address",
632 #define PE_IPV6_NOT_SUPPORTED 6
633 "IPv6 addresses not supported",
634 #define PE_INVALID_IPV6_ADDRESS 7
635 "Invalid IPv6 numeric address"
638 #define SETERR(p, v) do { \
644 /* The following two functions were adapted from glibc. */
647 is_valid_ipv4_address (const char *str, const char *end)
649 int saw_digit, octets;
659 if (ch >= '0' && ch <= '9') {
660 val = val * 10 + (ch - '0');
664 if (saw_digit == 0) {
669 } else if (ch == '.' && saw_digit == 1) {
683 static const int NS_INADDRSZ = 4;
684 static const int NS_IN6ADDRSZ = 16;
685 static const int NS_INT16SZ = 2;
688 is_valid_ipv6_address (const char *str, const char *end)
690 static const char xdigits[] = "0123456789abcdef";
703 /* Leading :: requires some special handling. */
707 if (str == end || *str != ':')
719 /* if ch is a number, add it to val. */
720 pch = strchr(xdigits, ch);
723 val |= (pch - xdigits);
730 /* if ch is a colon ... */
733 if (saw_xdigit == 0) {
738 } else if (str == end) {
741 if (tp > NS_IN6ADDRSZ - NS_INT16SZ)
749 /* if ch is a dot ... */
750 if (ch == '.' && (tp <= NS_IN6ADDRSZ - NS_INADDRSZ) &&
751 is_valid_ipv4_address(curtok, end) == 1) {
760 if (saw_xdigit == 1) {
761 if (tp > NS_IN6ADDRSZ - NS_INT16SZ)
766 if (colonp != NULL) {
767 if (tp == NS_IN6ADDRSZ)
772 if (tp != NS_IN6ADDRSZ)
781 Return a new struct url if successful, NULL on error. In case of
782 error, and if ERROR is not NULL, also set *ERROR to the appropriate
785 url_parse (const char *url, int *error)
789 int path_modified, host_modified;
791 enum url_scheme scheme;
793 const char *uname_b, *uname_e;
794 const char *host_b, *host_e;
795 const char *path_b, *path_e;
796 const char *params_b, *params_e;
797 const char *query_b, *query_e;
798 const char *fragment_b, *fragment_e;
801 char *user = NULL, *passwd = NULL;
805 scheme = url_scheme (url);
806 if (scheme == SCHEME_INVALID)
808 SETERR (error, PE_UNSUPPORTED_SCHEME);
812 url_encoded = reencode_escapes (url);
815 p += strlen (supported_schemes[scheme].leading_string);
817 p += url_skip_credentials (p);
820 /* scheme://user:pass@host[:port]... */
823 /* We attempt to break down the URL into the components path,
824 params, query, and fragment. They are ordered like this:
826 scheme://host[:port][/path][;params][?query][#fragment] */
828 params_b = params_e = NULL;
829 query_b = query_e = NULL;
830 fragment_b = fragment_e = NULL;
836 /* Handle IPv6 address inside square brackets. Ideally we'd
837 just look for the terminating ']', but rfc2732 mandates
838 rejecting invalid IPv6 addresses. */
840 /* The address begins after '['. */
842 host_e = strchr (host_b, ']');
846 SETERR (error, PE_UNTERMINATED_IPV6_ADDRESS);
851 /* Check if the IPv6 address is valid. */
852 if (!is_valid_ipv6_address(host_b, host_e))
854 SETERR (error, PE_INVALID_IPV6_ADDRESS);
858 /* Continue parsing after the closing ']'. */
861 SETERR (error, PE_IPV6_NOT_SUPPORTED);
867 p = strpbrk_or_eos (p, ":/;?#");
871 if (host_b == host_e)
873 SETERR (error, PE_EMPTY_HOST);
877 port = scheme_default_port (scheme);
880 const char *port_b, *port_e, *pp;
882 /* scheme://host:port/tralala */
886 p = strpbrk_or_eos (p, "/;?#");
889 if (port_b == port_e)
891 /* http://host:/whatever */
893 SETERR (error, PE_BAD_PORT_NUMBER);
897 for (port = 0, pp = port_b; pp < port_e; pp++)
901 /* http://host:12randomgarbage/blah */
903 SETERR (error, PE_BAD_PORT_NUMBER);
907 port = 10 * port + (*pp - '0');
915 p = strpbrk_or_eos (p, ";?#");
920 /* Path is not allowed not to exist. */
928 p = strpbrk_or_eos (p, "?#");
935 p = strpbrk_or_eos (p, "#");
938 /* Hack that allows users to use '?' (a wildcard character) in
939 FTP URLs without it being interpreted as a query string
941 if (scheme == SCHEME_FTP)
943 query_b = query_e = NULL;
956 if (uname_b != uname_e)
958 /* http://user:pass@host */
960 /* uname_b uname_e */
961 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
963 SETERR (error, PE_INVALID_USER_NAME);
968 u = (struct url *)xmalloc (sizeof (struct url));
969 memset (u, 0, sizeof (*u));
972 u->host = strdupdelim (host_b, host_e);
977 u->path = strdupdelim (path_b, path_e);
978 path_modified = path_simplify (u->path);
979 split_path (u->path, &u->dir, &u->file);
981 host_modified = lowercase_str (u->host);
984 u->params = strdupdelim (params_b, params_e);
986 u->query = strdupdelim (query_b, query_e);
988 u->fragment = strdupdelim (fragment_b, fragment_e);
990 if (path_modified || u->fragment || host_modified || path_b == path_e)
992 /* If we suspect that a transformation has rendered what
993 url_string might return different from URL_ENCODED, rebuild
994 u->url using url_string. */
995 u->url = url_string (u, 0);
997 if (url_encoded != url)
998 xfree ((char *) url_encoded);
1002 if (url_encoded == url)
1003 u->url = xstrdup (url);
1005 u->url = url_encoded;
1013 url_error (int error_code)
1015 assert (error_code >= 0 && error_code < countof (parse_errors));
1016 return parse_errors[error_code];
1019 /* Split PATH into DIR and FILE. PATH comes from the URL and is
1020 expected to be URL-escaped.
1022 The path is split into directory (the part up to the last slash)
1023 and file (the part after the last slash), which are subsequently
1024 unescaped. Examples:
1027 "foo/bar/baz" "foo/bar" "baz"
1028 "foo/bar/" "foo/bar" ""
1030 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
1032 DIR and FILE are freshly allocated. */
1035 split_path (const char *path, char **dir, char **file)
1037 char *last_slash = strrchr (path, '/');
1040 *dir = xstrdup ("");
1041 *file = xstrdup (path);
1045 *dir = strdupdelim (path, last_slash);
1046 *file = xstrdup (last_slash + 1);
1048 url_unescape (*dir);
1049 url_unescape (*file);
1052 /* Note: URL's "full path" is the path with the query string and
1053 params appended. The "fragment" (#foo) is intentionally ignored,
1054 but that might be changed. For example, if the original URL was
1055 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
1056 the full path will be "/foo/bar/baz;bullshit?querystring". */
1058 /* Return the length of the full path, without the terminating
1062 full_path_length (const struct url *url)
1066 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
1077 /* Write out the full path. */
1080 full_path_write (const struct url *url, char *where)
1082 #define FROB(el, chr) do { \
1083 char *f_el = url->el; \
1085 int l = strlen (f_el); \
1087 memcpy (where, f_el, l); \
1099 /* Public function for getting the "full path". E.g. if u->path is
1100 "foo/bar" and u->query is "param=value", full_path will be
1101 "/foo/bar?param=value". */
1104 url_full_path (const struct url *url)
1106 int length = full_path_length (url);
1107 char *full_path = (char *)xmalloc(length + 1);
1109 full_path_write (url, full_path);
1110 full_path[length] = '\0';
1115 /* Escape unsafe and reserved characters, except for the slash
1119 url_escape_dir (const char *dir)
1121 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1126 /* Unescape slashes in NEWDIR. */
1128 h = newdir; /* hare */
1129 t = newdir; /* tortoise */
1131 for (; *h; h++, t++)
1133 if (*h == '%' && h[1] == '2' && h[2] == 'F')
1146 /* Sync u->path and u->url with u->dir and u->file. Called after
1147 u->file or u->dir have been changed, typically by the FTP code. */
1150 sync_path (struct url *u)
1152 char *newpath, *efile, *edir;
1156 /* u->dir and u->file are not escaped. URL-escape them before
1157 reassembling them into u->path. That way, if they contain
1158 separators like '?' or even if u->file contains slashes, the
1159 path will be correctly assembled. (u->file can contain slashes
1160 if the URL specifies it with %2f, or if an FTP server returns
1162 edir = url_escape_dir (u->dir);
1163 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1166 newpath = xstrdup (efile);
1169 int dirlen = strlen (edir);
1170 int filelen = strlen (efile);
1172 /* Copy "DIR/FILE" to newpath. */
1173 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1174 memcpy (p, edir, dirlen);
1177 memcpy (p, efile, filelen);
1186 if (efile != u->file)
1189 /* Regenerate u->url as well. */
1191 u->url = url_string (u, 0);
1194 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1195 This way we can sync u->path and u->url when they get changed. */
1198 url_set_dir (struct url *url, const char *newdir)
1201 url->dir = xstrdup (newdir);
1206 url_set_file (struct url *url, const char *newfile)
1209 url->file = xstrdup (newfile);
1214 url_free (struct url *url)
1220 FREE_MAYBE (url->params);
1221 FREE_MAYBE (url->query);
1222 FREE_MAYBE (url->fragment);
1223 FREE_MAYBE (url->user);
1224 FREE_MAYBE (url->passwd);
1232 /* Create all the necessary directories for PATH (a file). Calls
1233 mkdirhier() internally. */
1235 mkalldirs (const char *path)
1242 p = path + strlen (path);
1243 for (; *p != '/' && p != path; p--)
1246 /* Don't create if it's just a file. */
1247 if ((p == path) && (*p != '/'))
1249 t = strdupdelim (path, p);
1251 /* Check whether the directory exists. */
1252 if ((stat (t, &st) == 0))
1254 if (S_ISDIR (st.st_mode))
1261 /* If the dir exists as a file name, remove it first. This
1262 is *only* for Wget to work with buggy old CERN http
1263 servers. Here is the scenario: When Wget tries to
1264 retrieve a directory without a slash, e.g.
1265 http://foo/bar (bar being a directory), CERN server will
1266 not redirect it too http://foo/bar/ -- it will generate a
1267 directory listing containing links to bar/file1,
1268 bar/file2, etc. Wget will lose because it saves this
1269 HTML listing to a file `bar', so it cannot create the
1270 directory. To work around this, if the file of the same
1271 name exists, we just remove it and create the directory
1273 DEBUGP (("Removing %s because of directory danger!\n", t));
1277 res = make_directory (t);
1279 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1284 /* Functions for constructing the file name out of URL components. */
1286 /* A growable string structure, used by url_file_name and friends.
1287 This should perhaps be moved to utils.c.
1289 The idea is to have a convenient and efficient way to construct a
1290 string by having various functions append data to it. Instead of
1291 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1292 functions in questions, we pass the pointer to this struct. */
1300 /* Ensure that the string can accept APPEND_COUNT more characters past
1301 the current TAIL position. If necessary, this will grow the string
1302 and update its allocated size. If the string is already large
1303 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1304 #define GROW(g, append_size) do { \
1305 struct growable *G_ = g; \
1306 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1309 /* Return the tail position of the string. */
1310 #define TAIL(r) ((r)->base + (r)->tail)
1312 /* Move the tail position by APPEND_COUNT characters. */
1313 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1315 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1319 append_string (const char *str, struct growable *dest)
1321 int l = strlen (str);
1323 memcpy (TAIL (dest), str, l);
1324 TAIL_INCR (dest, l);
1327 /* Append CH to DEST. For example, append_char (0, DEST)
1328 zero-terminates DEST. */
1331 append_char (char ch, struct growable *dest)
1335 TAIL_INCR (dest, 1);
1339 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1340 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1341 filechr_control = 4, /* a control character, e.g. 0-31 */
1344 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1346 /* Shorthands for the table: */
1347 #define U filechr_not_unix
1348 #define W filechr_not_windows
1349 #define C filechr_control
1354 /* Table of characters unsafe under various conditions (see above).
1356 Arguably we could also claim `%' to be unsafe, since we use it as
1357 the escape character. If we ever want to be able to reliably
1358 translate file name back to URL, this would become important
1359 crucial. Right now, it's better to be minimal in escaping. */
1361 const static unsigned char filechr_table[256] =
1363 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1364 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1365 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1366 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1367 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1368 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1369 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1370 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1371 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1372 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1373 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1374 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1375 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1376 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1377 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1378 0, 0, 0, 0, 0, 0, 0, 0, /* x y z { | } ~ DEL */
1380 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1381 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1382 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1383 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1385 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1386 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1387 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1388 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1396 /* FN_PORT_SEP is the separator between host and port in file names
1397 for non-standard port numbers. On Unix this is normally ':', as in
1398 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1399 because Windows can't handle ':' in file names. */
1400 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1402 /* FN_QUERY_SEP is the separator between the file name and the URL
1403 query, normally '?'. Since Windows cannot handle '?' as part of
1404 file name, we use '@' instead there. */
1405 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1407 /* Quote path element, characters in [b, e), as file name, and append
1408 the quoted string to DEST. Each character is quoted as per
1409 file_unsafe_char and the corresponding table. */
1412 append_uri_pathel (const char *b, const char *e, struct growable *dest)
1421 if (opt.restrict_files_os == restrict_unix)
1422 mask = filechr_not_unix;
1424 mask = filechr_not_windows;
1425 if (opt.restrict_files_ctrl)
1426 mask |= filechr_control;
1428 /* Copy [b, e) to PATHEL and URL-unescape it. */
1429 BOUNDED_TO_ALLOCA (b, e, pathel);
1430 url_unescape (pathel);
1431 pathlen = strlen (pathel);
1433 /* Go through PATHEL and check how many characters we'll need to
1434 add for file quoting. */
1436 for (p = pathel; *p; p++)
1437 if (FILE_CHAR_TEST (*p, mask))
1440 /* p - pathel is the string length. Each quoted char means two
1441 additional characters in the string, hence 2*quoted. */
1442 outlen = (p - pathel) + (2 * quoted);
1443 GROW (dest, outlen);
1447 /* If there's nothing to quote, we don't need to go through the
1448 string the second time. */
1449 memcpy (TAIL (dest), pathel, outlen);
1453 char *q = TAIL (dest);
1454 for (p = pathel; *p; p++)
1456 if (!FILE_CHAR_TEST (*p, mask))
1460 unsigned char ch = *p;
1462 *q++ = XNUM_TO_DIGIT (ch >> 4);
1463 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1466 assert (q - TAIL (dest) == outlen);
1468 TAIL_INCR (dest, outlen);
1471 /* Append to DEST the directory structure that corresponds the
1472 directory part of URL's path. For example, if the URL is
1473 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1475 Each path element ("dir1" and "dir2" in the above example) is
1476 examined, url-unescaped, and re-escaped as file name element.
1478 Additionally, it cuts as many directories from the path as
1479 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1480 will produce "bar" for the above example. For 2 or more, it will
1483 Each component of the path is quoted for use as file name. */
1486 append_dir_structure (const struct url *u, struct growable *dest)
1488 char *pathel, *next;
1489 int cut = opt.cut_dirs;
1491 /* Go through the path components, de-URL-quote them, and quote them
1492 (if necessary) as file names. */
1495 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1500 /* Ignore empty pathels. path_simplify should remove
1501 occurrences of "//" from the path, but it has special cases
1502 for starting / which generates an empty pathel here. */
1506 append_char ('/', dest);
1507 append_uri_pathel (pathel, next, dest);
1511 /* Return a unique file name that matches the given URL as good as
1512 possible. Does not create directories on the file system. */
1515 url_file_name (const struct url *u)
1517 struct growable fnres;
1519 char *u_file, *u_query;
1520 char *fname, *unique;
1526 /* Start with the directory prefix, if specified. */
1528 append_string (opt.dir_prefix, &fnres);
1530 /* If "dirstruct" is turned on (typically the case with -r), add
1531 the host and port (unless those have been turned off) and
1532 directory structure. */
1535 if (opt.add_hostdir)
1538 append_char ('/', &fnres);
1539 append_string (u->host, &fnres);
1540 if (u->port != scheme_default_port (u->scheme))
1543 number_to_string (portstr, u->port);
1544 append_char (FN_PORT_SEP, &fnres);
1545 append_string (portstr, &fnres);
1549 append_dir_structure (u, &fnres);
1552 /* Add the file name. */
1554 append_char ('/', &fnres);
1555 u_file = *u->file ? u->file : "index.html";
1556 append_uri_pathel (u_file, u_file + strlen (u_file), &fnres);
1558 /* Append "?query" to the file name. */
1559 u_query = u->query && *u->query ? u->query : NULL;
1562 append_char (FN_QUERY_SEP, &fnres);
1563 append_uri_pathel (u_query, u_query + strlen (u_query), &fnres);
1566 /* Zero-terminate the file name. */
1567 append_char ('\0', &fnres);
1571 /* Check the cases in which the unique extensions are not used:
1572 1) Clobbering is turned off (-nc).
1573 2) Retrieval with regetting.
1574 3) Timestamping is used.
1575 4) Hierarchy is built.
1577 The exception is the case when file does exist and is a
1578 directory (see `mkalldirs' for explanation). */
1580 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1581 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1584 unique = unique_name (fname, 1);
1585 if (unique != fname)
1590 /* Return the length of URL's path. Path is considered to be
1591 terminated by one of '?', ';', '#', or by the end of the
1594 path_length (const char *url)
1596 const char *q = strpbrk_or_eos (url, "?;#");
1600 /* Find the last occurrence of character C in the range [b, e), or
1601 NULL, if none are present. This is equivalent to strrchr(b, c),
1602 except that it accepts an END argument instead of requiring the
1603 string to be zero-terminated. Why is there no memrchr()? */
1605 find_last_char (const char *b, const char *e, char c)
1613 /* Resolve "." and ".." elements of PATH by destructively modifying
1614 PATH. "." is resolved by removing that path element, and ".." is
1615 resolved by removing the preceding path element. Single leading
1616 and trailing slashes are preserved.
1618 Return non-zero if any changes have been made.
1620 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1621 test examples are provided below. If you change anything in this
1622 function, run test_path_simplify to make sure you haven't broken a
1625 A previous version of this function was based on path_simplify()
1626 from GNU Bash, but it has been rewritten for Wget 1.8.1. */
1629 path_simplify (char *path)
1633 /* Preserve the leading '/'. */
1637 h = path; /* hare */
1638 t = path; /* tortoise */
1639 end = path + strlen (path);
1643 /* Hare should be at the beginning of a path element. */
1645 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1650 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1652 /* Handle "../" by retreating the tortoise by one path
1653 element -- but not past beggining of PATH. */
1657 /* Move backwards until B hits the beginning of the
1658 previous path element or the beginning of path. */
1659 for (--t; t > path && t[-1] != '/'; t--)
1666 /* Ignore empty path elements. Supporting them is hard (in
1667 which directory do you save http://x.com///y.html?), and
1668 they don't bring any practical gain. Plus, they break
1669 our filesystem-influenced assumptions: allowing empty
1670 path elements means that "x/y/../z" simplifies to
1671 "x/y/z", whereas most people would expect "x/z". */
1676 /* A regular path element. If H hasn't advanced past T,
1677 simply skip to the next path element. Otherwise, copy
1678 the path element until the next slash. */
1681 /* Skip the path element, including the slash. */
1682 while (h < end && *h != '/')
1689 /* Copy the path element, including the final slash. */
1690 while (h < end && *h != '/')
1704 /* Merge BASE with LINK and return the resulting URI.
1706 Either of the URIs may be absolute or relative, complete with the
1707 host name, or path only. This tries to reasonably handle all
1708 foreseeable cases. It only employs minimal URL parsing, without
1709 knowledge of the specifics of schemes.
1711 Perhaps this function should call path_simplify so that the callers
1712 don't have to call url_parse unconditionally. */
1715 uri_merge (const char *base, const char *link)
1721 if (url_has_scheme (link))
1722 return xstrdup (link);
1724 /* We may not examine BASE past END. */
1725 end = base + path_length (base);
1726 linklength = strlen (link);
1730 /* Empty LINK points back to BASE, query string and all. */
1731 return xstrdup (base);
1733 else if (*link == '?')
1735 /* LINK points to the same location, but changes the query
1736 string. Examples: */
1737 /* uri_merge("path", "?new") -> "path?new" */
1738 /* uri_merge("path?foo", "?new") -> "path?new" */
1739 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1740 /* uri_merge("path#foo", "?new") -> "path?new" */
1741 int baselength = end - base;
1742 merge = xmalloc (baselength + linklength + 1);
1743 memcpy (merge, base, baselength);
1744 memcpy (merge + baselength, link, linklength);
1745 merge[baselength + linklength] = '\0';
1747 else if (*link == '#')
1749 /* uri_merge("path", "#new") -> "path#new" */
1750 /* uri_merge("path#foo", "#new") -> "path#new" */
1751 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1752 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1754 const char *end1 = strchr (base, '#');
1756 end1 = base + strlen (base);
1757 baselength = end1 - base;
1758 merge = xmalloc (baselength + linklength + 1);
1759 memcpy (merge, base, baselength);
1760 memcpy (merge + baselength, link, linklength);
1761 merge[baselength + linklength] = '\0';
1763 else if (*link == '/' && *(link + 1) == '/')
1765 /* LINK begins with "//" and so is a net path: we need to
1766 replace everything after (and including) the double slash
1769 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1770 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1771 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1775 const char *start_insert;
1777 /* Look for first slash. */
1778 slash = memchr (base, '/', end - base);
1779 /* If found slash and it is a double slash, then replace
1780 from this point, else default to replacing from the
1782 if (slash && *(slash + 1) == '/')
1783 start_insert = slash;
1785 start_insert = base;
1787 span = start_insert - base;
1788 merge = (char *)xmalloc (span + linklength + 1);
1790 memcpy (merge, base, span);
1791 memcpy (merge + span, link, linklength);
1792 merge[span + linklength] = '\0';
1794 else if (*link == '/')
1796 /* LINK is an absolute path: we need to replace everything
1797 after (and including) the FIRST slash with LINK.
1799 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1800 "/qux/xyzzy", our result should be
1801 "http://host/qux/xyzzy". */
1804 const char *start_insert = NULL; /* for gcc to shut up. */
1805 const char *pos = base;
1806 int seen_slash_slash = 0;
1807 /* We're looking for the first slash, but want to ignore
1810 slash = memchr (pos, '/', end - pos);
1811 if (slash && !seen_slash_slash)
1812 if (*(slash + 1) == '/')
1815 seen_slash_slash = 1;
1819 /* At this point, SLASH is the location of the first / after
1820 "//", or the first slash altogether. START_INSERT is the
1821 pointer to the location where LINK will be inserted. When
1822 examining the last two examples, keep in mind that LINK
1825 if (!slash && !seen_slash_slash)
1826 /* example: "foo" */
1828 start_insert = base;
1829 else if (!slash && seen_slash_slash)
1830 /* example: "http://foo" */
1833 else if (slash && !seen_slash_slash)
1834 /* example: "foo/bar" */
1836 start_insert = base;
1837 else if (slash && seen_slash_slash)
1838 /* example: "http://something/" */
1840 start_insert = slash;
1842 span = start_insert - base;
1843 merge = (char *)xmalloc (span + linklength + 1);
1845 memcpy (merge, base, span);
1846 memcpy (merge + span, link, linklength);
1847 merge[span + linklength] = '\0';
1851 /* LINK is a relative URL: we need to replace everything
1852 after last slash (possibly empty) with LINK.
1854 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1855 our result should be "whatever/foo/qux/xyzzy". */
1856 int need_explicit_slash = 0;
1858 const char *start_insert;
1859 const char *last_slash = find_last_char (base, end, '/');
1862 /* No slash found at all. Append LINK to what we have,
1863 but we'll need a slash as a separator.
1865 Example: if base == "foo" and link == "qux/xyzzy", then
1866 we cannot just append link to base, because we'd get
1867 "fooqux/xyzzy", whereas what we want is
1870 To make sure the / gets inserted, we set
1871 need_explicit_slash to 1. We also set start_insert
1872 to end + 1, so that the length calculations work out
1873 correctly for one more (slash) character. Accessing
1874 that character is fine, since it will be the
1875 delimiter, '\0' or '?'. */
1876 /* example: "foo?..." */
1877 /* ^ ('?' gets changed to '/') */
1878 start_insert = end + 1;
1879 need_explicit_slash = 1;
1881 else if (last_slash && last_slash >= base + 2
1882 && last_slash[-2] == ':' && last_slash[-1] == '/')
1884 /* example: http://host" */
1886 start_insert = end + 1;
1887 need_explicit_slash = 1;
1891 /* example: "whatever/foo/bar" */
1893 start_insert = last_slash + 1;
1896 span = start_insert - base;
1897 merge = (char *)xmalloc (span + linklength + 1);
1899 memcpy (merge, base, span);
1900 if (need_explicit_slash)
1901 merge[span - 1] = '/';
1902 memcpy (merge + span, link, linklength);
1903 merge[span + linklength] = '\0';
1909 #define APPEND(p, s) do { \
1910 int len = strlen (s); \
1911 memcpy (p, s, len); \
1915 /* Use this instead of password when the actual password is supposed
1916 to be hidden. We intentionally use a generic string without giving
1917 away the number of characters in the password, like previous
1919 #define HIDDEN_PASSWORD "*password*"
1921 /* Recreate the URL string from the data in URL.
1923 If HIDE is non-zero (as it is when we're calling this on a URL we
1924 plan to print, but not when calling it to canonicalize a URL for
1925 use within the program), password will be hidden. Unsafe
1926 characters in the URL will be quoted. */
1929 url_string (const struct url *url, int hide_password)
1933 char *quoted_user = NULL, *quoted_passwd = NULL;
1935 int scheme_port = supported_schemes[url->scheme].default_port;
1936 char *scheme_str = supported_schemes[url->scheme].leading_string;
1937 int fplen = full_path_length (url);
1939 int brackets_around_host = 0;
1941 assert (scheme_str != NULL);
1943 /* Make sure the user name and password are quoted. */
1946 quoted_user = url_escape_allow_passthrough (url->user);
1950 quoted_passwd = HIDDEN_PASSWORD;
1952 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1956 if (strchr (url->host, ':'))
1957 brackets_around_host = 1;
1959 size = (strlen (scheme_str)
1960 + strlen (url->host)
1961 + (brackets_around_host ? 2 : 0)
1964 if (url->port != scheme_port)
1965 size += 1 + numdigit (url->port);
1968 size += 1 + strlen (quoted_user);
1970 size += 1 + strlen (quoted_passwd);
1973 p = result = xmalloc (size);
1975 APPEND (p, scheme_str);
1978 APPEND (p, quoted_user);
1982 APPEND (p, quoted_passwd);
1987 if (brackets_around_host)
1989 APPEND (p, url->host);
1990 if (brackets_around_host)
1992 if (url->port != scheme_port)
1995 p = number_to_string (p, url->port);
1998 full_path_write (url, p);
2002 assert (p - result == size);
2004 if (quoted_user && quoted_user != url->user)
2005 xfree (quoted_user);
2006 if (quoted_passwd && !hide_password
2007 && quoted_passwd != url->passwd)
2008 xfree (quoted_passwd);
2013 /* Return non-zero if scheme a is similar to scheme b.
2015 Schemes are similar if they are equal. If SSL is supported, schemes
2016 are also similar if one is http (SCHEME_HTTP) and the other is https
2019 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
2024 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
2025 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
2032 /* Debugging and testing support for path_simplify. */
2034 /* Debug: run path_simplify on PATH and return the result in a new
2035 string. Useful for calling from the debugger. */
2039 char *copy = xstrdup (path);
2040 path_simplify (copy);
2045 run_test (char *test, char *expected_result, int expected_change)
2047 char *test_copy = xstrdup (test);
2048 int modified = path_simplify (test_copy);
2050 if (0 != strcmp (test_copy, expected_result))
2052 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2053 test, expected_result, test_copy);
2055 if (modified != expected_change)
2057 if (expected_change == 1)
2058 printf ("Expected no modification with path_simplify(\"%s\").\n",
2061 printf ("Expected modification with path_simplify(\"%s\").\n",
2068 test_path_simplify (void)
2071 char *test, *result;
2077 { "foo", "foo", 0 },
2078 { "foo/bar", "foo/bar", 0 },
2079 { "foo///bar", "foo/bar", 1 },
2080 { "foo/.", "foo/", 1 },
2081 { "foo/./", "foo/", 1 },
2082 { "foo./", "foo./", 0 },
2083 { "foo/../bar", "bar", 1 },
2084 { "foo/../bar/", "bar/", 1 },
2085 { "foo/bar/..", "foo/", 1 },
2086 { "foo/bar/../x", "foo/x", 1 },
2087 { "foo/bar/../x/", "foo/x/", 1 },
2088 { "foo/..", "", 1 },
2089 { "foo/../..", "", 1 },
2090 { "a/b/../../c", "c", 1 },
2091 { "./a/../b", "b", 1 }
2095 for (i = 0; i < countof (tests); i++)
2097 char *test = tests[i].test;
2098 char *expected_result = tests[i].result;
2099 int expected_change = tests[i].should_modify;
2100 run_test (test, expected_result, expected_change);
2103 /* Now run all the tests with a leading slash before the test case,
2104 to prove that the slash is being preserved. */
2105 for (i = 0; i < countof (tests); i++)
2107 char *test, *expected_result;
2108 int expected_change = tests[i].should_modify;
2110 test = xmalloc (1 + strlen (tests[i].test) + 1);
2111 sprintf (test, "/%s", tests[i].test);
2113 expected_result = xmalloc (1 + strlen (tests[i].result) + 1);
2114 sprintf (expected_result, "/%s", tests[i].result);
2116 run_test (test, expected_result, expected_change);
2119 xfree (expected_result);