2 Copyright (C) 2005 Free Software Foundation, Inc.
4 This file is part of GNU Wget.
6 GNU Wget is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or (at
9 your option) any later version.
11 GNU Wget is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with Wget; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
20 In addition, as a special exception, the Free Software Foundation
21 gives permission to link the code of its release of Wget with the
22 OpenSSL project's "OpenSSL" library (or with modified versions of it
23 that use the same license as the "OpenSSL" library), and distribute
24 the linked executables. You must obey the GNU General Public License
25 in all respects for all of the code used other than "OpenSSL". If you
26 modify this file, you may extend this exception to your version of the
27 file, but you are not obligated to do so. If you do not wish to do
28 so, delete this exception statement from your version. */
39 #include <sys/types.h>
49 #include "host.h" /* for is_valid_ipv6_address */
58 const char *leading_string;
63 /* Supported schemes: */
64 static struct scheme_data supported_schemes[] =
66 { "http", "http://", DEFAULT_HTTP_PORT, 1 },
68 { "https", "https://", DEFAULT_HTTPS_PORT, 1 },
70 { "ftp", "ftp://", DEFAULT_FTP_PORT, 1 },
76 /* Forward declarations: */
78 static int path_simplify PARAMS ((char *));
80 /* Support for escaping and unescaping of URL strings. */
82 /* Table of "reserved" and "unsafe" characters. Those terms are
83 rfc1738-speak, as such largely obsoleted by rfc2396 and later
84 specs, but the general idea remains.
86 A reserved character is the one that you can't decode without
87 changing the meaning of the URL. For example, you can't decode
88 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
89 path components is different. Non-reserved characters can be
90 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
91 unsafe characters are loosely based on rfc1738, plus "$" and ",",
92 as recommended by rfc2396, and minus "~", which is very frequently
93 used (and sometimes unrecognized as %7E by broken servers).
95 An unsafe character is the one that should be encoded when URLs are
96 placed in foreign environments. E.g. space and newline are unsafe
97 in HTTP contexts because HTTP uses them as separator and line
98 terminator, so they must be encoded to %20 and %0A respectively.
99 "*" is unsafe in shell context, etc.
101 We determine whether a character is unsafe through static table
102 lookup. This code assumes ASCII character set and 8-bit chars. */
105 /* rfc1738 reserved chars + "$" 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 static const 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, 0, 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_scheme (url) != SCHEME_INVALID)
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');
784 /* Check for too large port numbers here, before we have
785 a chance to overflow on bogus port values. */
788 error_code = PE_BAD_PORT_NUMBER;
799 p = strpbrk_or_eos (p, ";?#");
804 /* Path is not allowed not to exist. */
812 p = strpbrk_or_eos (p, "?#");
819 p = strpbrk_or_eos (p, "#");
822 /* Hack that allows users to use '?' (a wildcard character) in
823 FTP URLs without it being interpreted as a query string
825 if (scheme == SCHEME_FTP)
827 query_b = query_e = NULL;
840 if (uname_b != uname_e)
842 /* http://user:pass@host */
844 /* uname_b uname_e */
845 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
847 error_code = PE_INVALID_USER_NAME;
852 u = xnew0 (struct url);
854 u->host = strdupdelim (host_b, host_e);
859 u->path = strdupdelim (path_b, path_e);
860 path_modified = path_simplify (u->path);
861 split_path (u->path, &u->dir, &u->file);
863 host_modified = lowercase_str (u->host);
865 /* Decode %HH sequences in host name. This is important not so much
866 to support %HH sequences, but to support binary characters (which
867 will have been converted to %HH by reencode_escapes). */
868 if (strchr (u->host, '%'))
870 url_unescape (u->host);
875 u->params = strdupdelim (params_b, params_e);
877 u->query = strdupdelim (query_b, query_e);
879 u->fragment = strdupdelim (fragment_b, fragment_e);
881 if (path_modified || u->fragment || host_modified || path_b == path_e)
883 /* If we suspect that a transformation has rendered what
884 url_string might return different from URL_ENCODED, rebuild
885 u->url using url_string. */
886 u->url = url_string (u, 0);
888 if (url_encoded != url)
889 xfree ((char *) url_encoded);
893 if (url_encoded == url)
894 u->url = xstrdup (url);
896 u->url = url_encoded;
903 /* Cleanup in case of error: */
904 if (url_encoded && url_encoded != url)
907 /* Transmit the error code to the caller, if the caller wants to
914 /* Return the error message string from ERROR_CODE, which should have
915 been retrieved from url_parse. The error message is translated. */
918 url_error (int error_code)
920 assert (error_code >= 0 && error_code < countof (parse_errors));
921 return _(parse_errors[error_code]);
924 /* Split PATH into DIR and FILE. PATH comes from the URL and is
925 expected to be URL-escaped.
927 The path is split into directory (the part up to the last slash)
928 and file (the part after the last slash), which are subsequently
932 "foo/bar/baz" "foo/bar" "baz"
933 "foo/bar/" "foo/bar" ""
935 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
937 DIR and FILE are freshly allocated. */
940 split_path (const char *path, char **dir, char **file)
942 char *last_slash = strrchr (path, '/');
946 *file = xstrdup (path);
950 *dir = strdupdelim (path, last_slash);
951 *file = xstrdup (last_slash + 1);
954 url_unescape (*file);
957 /* Note: URL's "full path" is the path with the query string and
958 params appended. The "fragment" (#foo) is intentionally ignored,
959 but that might be changed. For example, if the original URL was
960 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
961 the full path will be "/foo/bar/baz;bullshit?querystring". */
963 /* Return the length of the full path, without the terminating
967 full_path_length (const struct url *url)
971 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
982 /* Write out the full path. */
985 full_path_write (const struct url *url, char *where)
987 #define FROB(el, chr) do { \
988 char *f_el = url->el; \
990 int l = strlen (f_el); \
992 memcpy (where, f_el, l); \
1004 /* Public function for getting the "full path". E.g. if u->path is
1005 "foo/bar" and u->query is "param=value", full_path will be
1006 "/foo/bar?param=value". */
1009 url_full_path (const struct url *url)
1011 int length = full_path_length (url);
1012 char *full_path = (char *) xmalloc (length + 1);
1014 full_path_write (url, full_path);
1015 full_path[length] = '\0';
1020 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1021 escaping of certain characters, such as "/" and ":". Returns a
1022 count of unescaped chars. */
1025 unescape_single_char (char *str, char chr)
1027 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1028 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1029 char *h = str; /* hare */
1030 char *t = str; /* tortoise */
1031 for (; *h; h++, t++)
1033 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1044 /* Escape unsafe and reserved characters, except for the slash
1048 url_escape_dir (const char *dir)
1050 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1054 unescape_single_char (newdir, '/');
1058 /* Sync u->path and u->url with u->dir and u->file. Called after
1059 u->file or u->dir have been changed, typically by the FTP code. */
1062 sync_path (struct url *u)
1064 char *newpath, *efile, *edir;
1068 /* u->dir and u->file are not escaped. URL-escape them before
1069 reassembling them into u->path. That way, if they contain
1070 separators like '?' or even if u->file contains slashes, the
1071 path will be correctly assembled. (u->file can contain slashes
1072 if the URL specifies it with %2f, or if an FTP server returns
1074 edir = url_escape_dir (u->dir);
1075 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1078 newpath = xstrdup (efile);
1081 int dirlen = strlen (edir);
1082 int filelen = strlen (efile);
1084 /* Copy "DIR/FILE" to newpath. */
1085 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1086 memcpy (p, edir, dirlen);
1089 memcpy (p, efile, filelen);
1098 if (efile != u->file)
1101 /* Regenerate u->url as well. */
1103 u->url = url_string (u, 0);
1106 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1107 This way we can sync u->path and u->url when they get changed. */
1110 url_set_dir (struct url *url, const char *newdir)
1113 url->dir = xstrdup (newdir);
1118 url_set_file (struct url *url, const char *newfile)
1121 url->file = xstrdup (newfile);
1126 url_free (struct url *url)
1132 xfree_null (url->params);
1133 xfree_null (url->query);
1134 xfree_null (url->fragment);
1135 xfree_null (url->user);
1136 xfree_null (url->passwd);
1144 /* Create all the necessary directories for PATH (a file). Calls
1145 mkdirhier() internally. */
1147 mkalldirs (const char *path)
1154 p = path + strlen (path);
1155 for (; *p != '/' && p != path; p--)
1158 /* Don't create if it's just a file. */
1159 if ((p == path) && (*p != '/'))
1161 t = strdupdelim (path, p);
1163 /* Check whether the directory exists. */
1164 if ((stat (t, &st) == 0))
1166 if (S_ISDIR (st.st_mode))
1173 /* If the dir exists as a file name, remove it first. This
1174 is *only* for Wget to work with buggy old CERN http
1175 servers. Here is the scenario: When Wget tries to
1176 retrieve a directory without a slash, e.g.
1177 http://foo/bar (bar being a directory), CERN server will
1178 not redirect it too http://foo/bar/ -- it will generate a
1179 directory listing containing links to bar/file1,
1180 bar/file2, etc. Wget will lose because it saves this
1181 HTML listing to a file `bar', so it cannot create the
1182 directory. To work around this, if the file of the same
1183 name exists, we just remove it and create the directory
1185 DEBUGP (("Removing %s because of directory danger!\n", t));
1189 res = make_directory (t);
1191 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1196 /* Functions for constructing the file name out of URL components. */
1198 /* A growable string structure, used by url_file_name and friends.
1199 This should perhaps be moved to utils.c.
1201 The idea is to have a convenient and efficient way to construct a
1202 string by having various functions append data to it. Instead of
1203 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1204 functions in questions, we pass the pointer to this struct. */
1212 /* Ensure that the string can accept APPEND_COUNT more characters past
1213 the current TAIL position. If necessary, this will grow the string
1214 and update its allocated size. If the string is already large
1215 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1216 #define GROW(g, append_size) do { \
1217 struct growable *G_ = g; \
1218 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1221 /* Return the tail position of the string. */
1222 #define TAIL(r) ((r)->base + (r)->tail)
1224 /* Move the tail position by APPEND_COUNT characters. */
1225 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1227 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1231 append_string (const char *str, struct growable *dest)
1233 int l = strlen (str);
1235 memcpy (TAIL (dest), str, l);
1236 TAIL_INCR (dest, l);
1239 /* Append CH to DEST. For example, append_char (0, DEST)
1240 zero-terminates DEST. */
1243 append_char (char ch, struct growable *dest)
1247 TAIL_INCR (dest, 1);
1251 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1252 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1253 filechr_control = 4 /* a control character, e.g. 0-31 */
1256 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1258 /* Shorthands for the table: */
1259 #define U filechr_not_unix
1260 #define W filechr_not_windows
1261 #define C filechr_control
1266 /* Table of characters unsafe under various conditions (see above).
1268 Arguably we could also claim `%' to be unsafe, since we use it as
1269 the escape character. If we ever want to be able to reliably
1270 translate file name back to URL, this would become important
1271 crucial. Right now, it's better to be minimal in escaping. */
1273 static const unsigned char filechr_table[256] =
1275 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1276 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1277 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1278 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1279 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1280 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1281 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1282 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1283 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1284 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1285 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1286 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1287 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1288 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1289 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1290 0, 0, 0, 0, 0, 0, 0, 0, /* x y z { | } ~ DEL */
1292 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1293 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1294 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1295 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1297 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1298 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1299 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1300 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1308 /* FN_PORT_SEP is the separator between host and port in file names
1309 for non-standard port numbers. On Unix this is normally ':', as in
1310 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1311 because Windows can't handle ':' in file names. */
1312 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1314 /* FN_QUERY_SEP is the separator between the file name and the URL
1315 query, normally '?'. Since Windows cannot handle '?' as part of
1316 file name, we use '@' instead there. */
1317 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1319 /* Quote path element, characters in [b, e), as file name, and append
1320 the quoted string to DEST. Each character is quoted as per
1321 file_unsafe_char and the corresponding table.
1323 If ESCAPED_P is non-zero, the path element is considered to be
1324 URL-escaped and will be unescaped prior to inspection. */
1327 append_uri_pathel (const char *b, const char *e, int escaped_p,
1328 struct growable *dest)
1334 if (opt.restrict_files_os == restrict_unix)
1335 mask = filechr_not_unix;
1337 mask = filechr_not_windows;
1338 if (opt.restrict_files_ctrl)
1339 mask |= filechr_control;
1341 /* Copy [b, e) to PATHEL and URL-unescape it. */
1345 BOUNDED_TO_ALLOCA (b, e, unescaped);
1346 url_unescape (unescaped);
1348 e = unescaped + strlen (unescaped);
1351 /* Defang ".." when found as component of path. Remember that path
1352 comes from the URL and might contain malicious input. */
1353 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1359 /* Walk the PATHEL string and check how many characters we'll need
1362 for (p = b; p < e; p++)
1363 if (FILE_CHAR_TEST (*p, mask))
1366 /* Calculate the length of the output string. e-b is the input
1367 string length. Each quoted char introduces two additional
1368 characters in the string, hence 2*quoted. */
1369 outlen = (e - b) + (2 * quoted);
1370 GROW (dest, outlen);
1374 /* If there's nothing to quote, we can simply append the string
1375 without processing it again. */
1376 memcpy (TAIL (dest), b, outlen);
1380 char *q = TAIL (dest);
1381 for (p = b; p < e; p++)
1383 if (!FILE_CHAR_TEST (*p, mask))
1387 unsigned char ch = *p;
1389 *q++ = XNUM_TO_DIGIT (ch >> 4);
1390 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1393 assert (q - TAIL (dest) == outlen);
1395 TAIL_INCR (dest, outlen);
1398 /* Append to DEST the directory structure that corresponds the
1399 directory part of URL's path. For example, if the URL is
1400 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1402 Each path element ("dir1" and "dir2" in the above example) is
1403 examined, url-unescaped, and re-escaped as file name element.
1405 Additionally, it cuts as many directories from the path as
1406 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1407 will produce "bar" for the above example. For 2 or more, it will
1410 Each component of the path is quoted for use as file name. */
1413 append_dir_structure (const struct url *u, struct growable *dest)
1415 char *pathel, *next;
1416 int cut = opt.cut_dirs;
1418 /* Go through the path components, de-URL-quote them, and quote them
1419 (if necessary) as file names. */
1422 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1427 /* Ignore empty pathels. */
1431 append_char ('/', dest);
1432 append_uri_pathel (pathel, next, 1, dest);
1436 /* Return a unique file name that matches the given URL as good as
1437 possible. Does not create directories on the file system. */
1440 url_file_name (const struct url *u)
1442 struct growable fnres; /* stands for "file name result" */
1444 const char *u_file, *u_query;
1445 char *fname, *unique;
1451 /* Start with the directory prefix, if specified. */
1453 append_string (opt.dir_prefix, &fnres);
1455 /* If "dirstruct" is turned on (typically the case with -r), add
1456 the host and port (unless those have been turned off) and
1457 directory structure. */
1460 if (opt.protocol_directories)
1463 append_char ('/', &fnres);
1464 append_string (supported_schemes[u->scheme].name, &fnres);
1466 if (opt.add_hostdir)
1469 append_char ('/', &fnres);
1470 if (0 != strcmp (u->host, ".."))
1471 append_string (u->host, &fnres);
1473 /* Host name can come from the network; malicious DNS may
1474 allow ".." to be resolved, causing us to write to
1475 "../<file>". Defang such host names. */
1476 append_string ("%2E%2E", &fnres);
1477 if (u->port != scheme_default_port (u->scheme))
1480 number_to_string (portstr, u->port);
1481 append_char (FN_PORT_SEP, &fnres);
1482 append_string (portstr, &fnres);
1486 append_dir_structure (u, &fnres);
1489 /* Add the file name. */
1491 append_char ('/', &fnres);
1492 u_file = *u->file ? u->file : "index.html";
1493 append_uri_pathel (u_file, u_file + strlen (u_file), 0, &fnres);
1495 /* Append "?query" to the file name. */
1496 u_query = u->query && *u->query ? u->query : NULL;
1499 append_char (FN_QUERY_SEP, &fnres);
1500 append_uri_pathel (u_query, u_query + strlen (u_query), 1, &fnres);
1503 /* Zero-terminate the file name. */
1504 append_char ('\0', &fnres);
1508 /* Check the cases in which the unique extensions are not used:
1509 1) Clobbering is turned off (-nc).
1510 2) Retrieval with regetting.
1511 3) Timestamping is used.
1512 4) Hierarchy is built.
1514 The exception is the case when file does exist and is a
1515 directory (see `mkalldirs' for explanation). */
1517 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1518 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1521 unique = unique_name (fname, 1);
1522 if (unique != fname)
1527 /* Resolve "." and ".." elements of PATH by destructively modifying
1528 PATH and return non-zero if PATH has been modified, zero otherwise.
1530 The algorithm is in spirit similar to the one described in rfc1808,
1531 although implemented differently, in one pass. To recap, path
1532 elements containing only "." are removed, and ".." is taken to mean
1533 "back up one element". Single leading and trailing slashes are
1536 This function does not handle URL escapes explicitly. If you're
1537 passing paths from URLs, make sure to unquote "%2e" and "%2E" to
1538 ".", so that this function can find the dots. (Wget's URL parser
1539 calls reencode_escapes, which see.)
1541 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1542 test examples are provided below. If you change anything in this
1543 function, run test_path_simplify to make sure you haven't broken a
1547 path_simplify (char *path)
1549 char *h = path; /* hare */
1550 char *t = path; /* tortoise */
1551 char *beg = path; /* boundary for backing the tortoise */
1552 char *end = path + strlen (path);
1556 /* Hare should be at the beginning of a path element. */
1558 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1563 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1565 /* Handle "../" by retreating the tortoise by one path
1566 element -- but not past beggining. */
1569 /* Move backwards until T hits the beginning of the
1570 previous path element or the beginning of path. */
1571 for (--t; t > beg && t[-1] != '/'; t--)
1576 /* If we're at the beginning, copy the "../" literally
1577 move the beginning so a later ".." doesn't remove
1587 /* A regular path element. If H hasn't advanced past T,
1588 simply skip to the next path element. Otherwise, copy
1589 the path element until the next slash. */
1592 /* Skip the path element, including the slash. */
1593 while (h < end && *h != '/')
1600 /* Copy the path element, including the final slash. */
1601 while (h < end && *h != '/')
1615 /* Return the length of URL's path. Path is considered to be
1616 terminated by one of '?', ';', '#', or by the end of the
1620 path_length (const char *url)
1622 const char *q = strpbrk_or_eos (url, "?;#");
1626 /* Find the last occurrence of character C in the range [b, e), or
1627 NULL, if none are present. We might want to use memrchr (a GNU
1628 extension) under GNU libc. */
1631 find_last_char (const char *b, const char *e, char c)
1639 /* Merge BASE with LINK and return the resulting URI.
1641 Either of the URIs may be absolute or relative, complete with the
1642 host name, or path only. This tries to reasonably handle all
1643 foreseeable cases. It only employs minimal URL parsing, without
1644 knowledge of the specifics of schemes.
1646 I briefly considered making this function call path_simplify after
1647 the merging process, as rfc1738 seems to suggest. This is a bad
1648 idea for several reasons: 1) it complexifies the code, and 2)
1649 url_parse has to simplify path anyway, so it's wasteful to boot. */
1652 uri_merge (const char *base, const char *link)
1658 if (url_has_scheme (link))
1659 return xstrdup (link);
1661 /* We may not examine BASE past END. */
1662 end = base + path_length (base);
1663 linklength = strlen (link);
1667 /* Empty LINK points back to BASE, query string and all. */
1668 return xstrdup (base);
1670 else if (*link == '?')
1672 /* LINK points to the same location, but changes the query
1673 string. Examples: */
1674 /* uri_merge("path", "?new") -> "path?new" */
1675 /* uri_merge("path?foo", "?new") -> "path?new" */
1676 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1677 /* uri_merge("path#foo", "?new") -> "path?new" */
1678 int baselength = end - base;
1679 merge = xmalloc (baselength + linklength + 1);
1680 memcpy (merge, base, baselength);
1681 memcpy (merge + baselength, link, linklength);
1682 merge[baselength + linklength] = '\0';
1684 else if (*link == '#')
1686 /* uri_merge("path", "#new") -> "path#new" */
1687 /* uri_merge("path#foo", "#new") -> "path#new" */
1688 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1689 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1691 const char *end1 = strchr (base, '#');
1693 end1 = base + strlen (base);
1694 baselength = end1 - base;
1695 merge = xmalloc (baselength + linklength + 1);
1696 memcpy (merge, base, baselength);
1697 memcpy (merge + baselength, link, linklength);
1698 merge[baselength + linklength] = '\0';
1700 else if (*link == '/' && *(link + 1) == '/')
1702 /* LINK begins with "//" and so is a net path: we need to
1703 replace everything after (and including) the double slash
1706 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1707 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1708 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1712 const char *start_insert;
1714 /* Look for first slash. */
1715 slash = memchr (base, '/', end - base);
1716 /* If found slash and it is a double slash, then replace
1717 from this point, else default to replacing from the
1719 if (slash && *(slash + 1) == '/')
1720 start_insert = slash;
1722 start_insert = base;
1724 span = start_insert - base;
1725 merge = (char *)xmalloc (span + linklength + 1);
1727 memcpy (merge, base, span);
1728 memcpy (merge + span, link, linklength);
1729 merge[span + linklength] = '\0';
1731 else if (*link == '/')
1733 /* LINK is an absolute path: we need to replace everything
1734 after (and including) the FIRST slash with LINK.
1736 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1737 "/qux/xyzzy", our result should be
1738 "http://host/qux/xyzzy". */
1741 const char *start_insert = NULL; /* for gcc to shut up. */
1742 const char *pos = base;
1743 int seen_slash_slash = 0;
1744 /* We're looking for the first slash, but want to ignore
1747 slash = memchr (pos, '/', end - pos);
1748 if (slash && !seen_slash_slash)
1749 if (*(slash + 1) == '/')
1752 seen_slash_slash = 1;
1756 /* At this point, SLASH is the location of the first / after
1757 "//", or the first slash altogether. START_INSERT is the
1758 pointer to the location where LINK will be inserted. When
1759 examining the last two examples, keep in mind that LINK
1762 if (!slash && !seen_slash_slash)
1763 /* example: "foo" */
1765 start_insert = base;
1766 else if (!slash && seen_slash_slash)
1767 /* example: "http://foo" */
1770 else if (slash && !seen_slash_slash)
1771 /* example: "foo/bar" */
1773 start_insert = base;
1774 else if (slash && seen_slash_slash)
1775 /* example: "http://something/" */
1777 start_insert = slash;
1779 span = start_insert - base;
1780 merge = (char *)xmalloc (span + linklength + 1);
1782 memcpy (merge, base, span);
1783 memcpy (merge + span, link, linklength);
1784 merge[span + linklength] = '\0';
1788 /* LINK is a relative URL: we need to replace everything
1789 after last slash (possibly empty) with LINK.
1791 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1792 our result should be "whatever/foo/qux/xyzzy". */
1793 int need_explicit_slash = 0;
1795 const char *start_insert;
1796 const char *last_slash = find_last_char (base, end, '/');
1799 /* No slash found at all. Replace what we have with LINK. */
1800 start_insert = base;
1802 else if (last_slash && last_slash >= base + 2
1803 && last_slash[-2] == ':' && last_slash[-1] == '/')
1805 /* example: http://host" */
1807 start_insert = end + 1;
1808 need_explicit_slash = 1;
1812 /* example: "whatever/foo/bar" */
1814 start_insert = last_slash + 1;
1817 span = start_insert - base;
1818 merge = (char *)xmalloc (span + linklength + 1);
1820 memcpy (merge, base, span);
1821 if (need_explicit_slash)
1822 merge[span - 1] = '/';
1823 memcpy (merge + span, link, linklength);
1824 merge[span + linklength] = '\0';
1830 #define APPEND(p, s) do { \
1831 int len = strlen (s); \
1832 memcpy (p, s, len); \
1836 /* Use this instead of password when the actual password is supposed
1837 to be hidden. We intentionally use a generic string without giving
1838 away the number of characters in the password, like previous
1840 #define HIDDEN_PASSWORD "*password*"
1842 /* Recreate the URL string from the data in URL.
1844 If HIDE is non-zero (as it is when we're calling this on a URL we
1845 plan to print, but not when calling it to canonicalize a URL for
1846 use within the program), password will be hidden. Unsafe
1847 characters in the URL will be quoted. */
1850 url_string (const struct url *url, int hide_password)
1854 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1856 int scheme_port = supported_schemes[url->scheme].default_port;
1857 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1858 int fplen = full_path_length (url);
1860 int brackets_around_host;
1862 assert (scheme_str != NULL);
1864 /* Make sure the user name and password are quoted. */
1867 quoted_user = url_escape_allow_passthrough (url->user);
1871 quoted_passwd = HIDDEN_PASSWORD;
1873 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1877 /* In the unlikely event that the host name contains non-printable
1878 characters, quote it for displaying to the user. */
1879 quoted_host = url_escape_allow_passthrough (url->host);
1881 /* Undo the quoting of colons that URL escaping performs. IPv6
1882 addresses may legally contain colons, and in that case must be
1883 placed in square brackets. */
1884 if (quoted_host != url->host)
1885 unescape_single_char (quoted_host, ':');
1886 brackets_around_host = strchr (quoted_host, ':') != NULL;
1888 size = (strlen (scheme_str)
1889 + strlen (quoted_host)
1890 + (brackets_around_host ? 2 : 0)
1893 if (url->port != scheme_port)
1894 size += 1 + numdigit (url->port);
1897 size += 1 + strlen (quoted_user);
1899 size += 1 + strlen (quoted_passwd);
1902 p = result = xmalloc (size);
1904 APPEND (p, scheme_str);
1907 APPEND (p, quoted_user);
1911 APPEND (p, quoted_passwd);
1916 if (brackets_around_host)
1918 APPEND (p, quoted_host);
1919 if (brackets_around_host)
1921 if (url->port != scheme_port)
1924 p = number_to_string (p, url->port);
1927 full_path_write (url, p);
1931 assert (p - result == size);
1933 if (quoted_user && quoted_user != url->user)
1934 xfree (quoted_user);
1935 if (quoted_passwd && !hide_password && quoted_passwd != url->passwd)
1936 xfree (quoted_passwd);
1937 if (quoted_host != url->host)
1938 xfree (quoted_host);
1943 /* Return non-zero if scheme a is similar to scheme b.
1945 Schemes are similar if they are equal. If SSL is supported, schemes
1946 are also similar if one is http (SCHEME_HTTP) and the other is https
1949 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1954 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1955 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1962 /* Debugging and testing support for path_simplify. */
1964 /* Debug: run path_simplify on PATH and return the result in a new
1965 string. Useful for calling from the debugger. */
1969 char *copy = xstrdup (path);
1970 path_simplify (copy);
1975 run_test (char *test, char *expected_result, int expected_change)
1977 char *test_copy = xstrdup (test);
1978 int modified = path_simplify (test_copy);
1980 if (0 != strcmp (test_copy, expected_result))
1982 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
1983 test, expected_result, test_copy);
1985 if (modified != expected_change)
1987 if (expected_change == 1)
1988 printf ("Expected modification with path_simplify(\"%s\").\n",
1991 printf ("Expected no modification with path_simplify(\"%s\").\n",
1998 test_path_simplify (void)
2001 char *test, *result;
2008 { "../", "../", 0 },
2009 { "foo", "foo", 0 },
2010 { "foo/bar", "foo/bar", 0 },
2011 { "foo///bar", "foo///bar", 0 },
2012 { "foo/.", "foo/", 1 },
2013 { "foo/./", "foo/", 1 },
2014 { "foo./", "foo./", 0 },
2015 { "foo/../bar", "bar", 1 },
2016 { "foo/../bar/", "bar/", 1 },
2017 { "foo/bar/..", "foo/", 1 },
2018 { "foo/bar/../x", "foo/x", 1 },
2019 { "foo/bar/../x/", "foo/x/", 1 },
2020 { "foo/..", "", 1 },
2021 { "foo/../..", "..", 1 },
2022 { "foo/../../..", "../..", 1 },
2023 { "foo/../../bar/../../baz", "../../baz", 1 },
2024 { "a/b/../../c", "c", 1 },
2025 { "./a/../b", "b", 1 }
2029 for (i = 0; i < countof (tests); i++)
2031 char *test = tests[i].test;
2032 char *expected_result = tests[i].result;
2033 int expected_change = tests[i].should_modify;
2034 run_test (test, expected_result, expected_change);