2 Copyright (C) 1996-2005 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 Foundation, Inc.,
18 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 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. */
44 #include "host.h" /* for is_valid_ipv6_address */
47 scm_disabled = 1, /* for https when OpenSSL fails to init. */
48 scm_has_params = 2, /* whether scheme has ;params */
49 scm_has_query = 4, /* whether scheme has ?query */
50 scm_has_fragment = 8 /* whether scheme has #fragment */
55 /* Short name of the scheme, such as "http" or "ftp". */
57 /* Leading string that identifies the scheme, such as "https://". */
58 const char *leading_string;
59 /* Default port of the scheme when none is specified. */
65 /* Supported schemes: */
66 static struct scheme_data supported_schemes[] =
68 { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment },
70 { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment },
72 { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment },
78 /* Forward declarations: */
80 static bool path_simplify (char *);
82 /* Support for escaping and unescaping of URL strings. */
84 /* Table of "reserved" and "unsafe" characters. Those terms are
85 rfc1738-speak, as such largely obsoleted by rfc2396 and later
86 specs, but the general idea remains.
88 A reserved character is the one that you can't decode without
89 changing the meaning of the URL. For example, you can't decode
90 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
91 path components is different. Non-reserved characters can be
92 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
93 unsafe characters are loosely based on rfc1738, plus "$" and ",",
94 as recommended by rfc2396, and minus "~", which is very frequently
95 used (and sometimes unrecognized as %7E by broken servers).
97 An unsafe character is the one that should be encoded when URLs are
98 placed in foreign environments. E.g. space and newline are unsafe
99 in HTTP contexts because HTTP uses them as separator and line
100 terminator, so they must be encoded to %20 and %0A respectively.
101 "*" is unsafe in shell context, etc.
103 We determine whether a character is unsafe through static table
104 lookup. This code assumes ASCII character set and 8-bit chars. */
107 /* rfc1738 reserved chars + "$" and ",". */
110 /* rfc1738 unsafe chars, plus non-printables. */
114 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
115 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
116 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
118 /* Shorthands for the table: */
119 #define R urlchr_reserved
120 #define U urlchr_unsafe
123 static const unsigned char urlchr_table[256] =
125 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
126 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
127 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
128 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
129 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
130 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
131 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
132 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
133 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
134 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
135 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
136 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
137 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
138 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
139 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
140 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
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,
144 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,
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,
149 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
150 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
156 /* URL-unescape the string S.
158 This is done by transforming the sequences "%HH" to the character
159 represented by the hexadecimal digits HH. If % is not followed by
160 two hexadecimal digits, it is inserted literally.
162 The transformation is done in place. If you need the original
163 string intact, make a copy before calling this function. */
166 url_unescape (char *s)
168 char *t = s; /* t - tortoise */
169 char *h = s; /* h - hare */
181 /* Do nothing if '%' is not followed by two hex digits. */
182 if (!h[1] || !h[2] || !(ISXDIGIT (h[1]) && ISXDIGIT (h[2])))
184 c = X2DIGITS_TO_NUM (h[1], h[2]);
185 /* Don't unescape %00 because there is no way to insert it
186 into a C string without effectively truncating it. */
196 /* The core of url_escape_* functions. Escapes the characters that
197 match the provided mask in urlchr_table.
199 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
200 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
201 allocated string will be returned in all cases. */
204 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
211 for (p1 = s; *p1; p1++)
212 if (urlchr_test (*p1, mask))
213 addition += 2; /* Two more characters (hex digits) */
216 return allow_passthrough ? (char *)s : xstrdup (s);
218 newlen = (p1 - s) + addition;
219 newstr = xmalloc (newlen + 1);
225 /* Quote the characters that match the test mask. */
226 if (urlchr_test (*p1, mask))
228 unsigned char c = *p1++;
230 *p2++ = XNUM_TO_DIGIT (c >> 4);
231 *p2++ = XNUM_TO_DIGIT (c & 0xf);
236 assert (p2 - newstr == newlen);
242 /* URL-escape the unsafe characters (see urlchr_table) in a given
243 string, returning a freshly allocated string. */
246 url_escape (const char *s)
248 return url_escape_1 (s, urlchr_unsafe, false);
251 /* URL-escape the unsafe characters (see urlchr_table) in a given
252 string. If no characters are unsafe, S is returned. */
255 url_escape_allow_passthrough (const char *s)
257 return url_escape_1 (s, urlchr_unsafe, true);
260 /* Decide whether the char at position P needs to be encoded. (It is
261 not enough to pass a single char *P because the function may need
262 to inspect the surrounding context.)
264 Return true if the char should be escaped as %XX, false otherwise. */
267 char_needs_escaping (const char *p)
271 if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2)))
274 /* Garbled %.. sequence: encode `%'. */
277 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
283 /* Translate a %-escaped (but possibly non-conformant) input string S
284 into a %-escaped (and conformant) output string. If no characters
285 are encoded or decoded, return the same string S; otherwise, return
286 a freshly allocated string with the new contents.
288 After a URL has been run through this function, the protocols that
289 use `%' as the quote character can use the resulting string as-is,
290 while those that don't can use url_unescape to get to the intended
291 data. This function is stable: once the input is transformed,
292 further transformations of the result yield the same output.
294 Let's discuss why this function is needed.
296 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
297 a raw space character would mess up the HTTP request, it needs to
298 be quoted, like this:
300 GET /abc%20def HTTP/1.0
302 It would appear that the unsafe chars need to be quoted, for
303 example 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 leads to the conclusion that
308 in that case Wget should not call url_escape, but leave the `%20'
309 as is. This is clearly contradictory, but it only gets worse.
311 What if the requested URI is `abc%20 def'? If we call url_escape,
312 we end up with `/abc%2520%20def', which is almost certainly not
313 intended. If we don't call url_escape, we are left with the
314 embedded space and cannot complete the request. What the user
315 meant was for Wget to request `/abc%20%20def', and this is where
316 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 a modified version of the algorithm originally
331 proposed by Anon Sricharoenchai:
333 * Encode all "unsafe" characters, except those that are also
334 "reserved", to %XX. See urlchr_table for which characters are
337 * Encode the "%" characters not followed by two hex digits to
340 * Pass through all other characters and %XX escapes as-is. (Up to
341 Wget 1.10 this decoded %XX escapes corresponding to "safe"
342 characters, but that was obtrusive and broke some servers.)
346 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
348 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+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;
370 /* First pass: inspect the string to see if there's anything to do,
371 and to calculate the new length. */
372 for (p1 = s; *p1; p1++)
373 if (char_needs_escaping (p1))
377 /* The string is good as it is. */
378 return (char *) s; /* C const model sucks. */
381 /* Each encoding adds two characters (hex digits). */
382 newlen = oldlen + 2 * encode_count;
383 newstr = xmalloc (newlen + 1);
385 /* Second pass: copy the string to the destination address, encoding
386 chars when needed. */
391 if (char_needs_escaping (p1))
393 unsigned char c = *p1++;
395 *p2++ = XNUM_TO_DIGIT (c >> 4);
396 *p2++ = XNUM_TO_DIGIT (c & 0xf);
402 assert (p2 - newstr == newlen);
406 /* Returns the scheme type if the scheme is supported, or
407 SCHEME_INVALID if not. */
410 url_scheme (const char *url)
414 for (i = 0; supported_schemes[i].leading_string; i++)
415 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
416 strlen (supported_schemes[i].leading_string)))
418 if (!(supported_schemes[i].flags & scm_disabled))
419 return (enum url_scheme) i;
421 return SCHEME_INVALID;
424 return SCHEME_INVALID;
427 #define SCHEME_CHAR(ch) (ISALNUM (ch) || (ch) == '-' || (ch) == '+')
429 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
430 currently implemented, it returns true if URL begins with
434 url_has_scheme (const char *url)
438 /* The first char must be a scheme char. */
439 if (!*p || !SCHEME_CHAR (*p))
442 /* Followed by 0 or more scheme chars. */
443 while (*p && SCHEME_CHAR (*p))
445 /* Terminated by ':'. */
450 scheme_default_port (enum url_scheme scheme)
452 return supported_schemes[scheme].default_port;
456 scheme_disable (enum url_scheme scheme)
458 supported_schemes[scheme].flags |= scm_disabled;
461 /* Skip the username and password, if present in the URL. The
462 function should *not* be called with the complete URL, but with the
463 portion after the scheme.
465 If no username and password are found, return URL. */
468 url_skip_credentials (const char *url)
470 /* Look for '@' that comes before terminators, such as '/', '?',
472 const char *p = (const char *)strpbrk (url, "@/?#;");
478 /* Parse credentials contained in [BEG, END). The region is expected
479 to have come from a URL and is unescaped. */
482 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
488 return false; /* empty user name */
490 colon = memchr (beg, ':', end - beg);
492 return false; /* again empty user name */
496 *passwd = strdupdelim (colon + 1, end);
498 url_unescape (*passwd);
505 *user = strdupdelim (beg, userend);
506 url_unescape (*user);
510 /* Used by main.c: detect URLs written using the "shorthand" URL forms
511 popularized by Netscape and NcFTP. HTTP shorthands look like this:
513 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
514 www.foo.com[:port] -> http://www.foo.com[:port]
516 FTP shorthands look like this:
518 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
519 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
521 If the URL needs not or cannot be rewritten, return NULL. */
524 rewrite_shorthand_url (const char *url)
528 if (url_scheme (url) != SCHEME_INVALID)
531 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
533 for (p = url; *p && *p != ':' && *p != '/'; p++)
539 /* If we're looking at "://", it means the URL uses a scheme we
540 don't support, which may include "https" when compiled without
541 SSL support. Don't bogusly rewrite such URLs. */
542 if (p[0] == ':' && p[1] == '/' && p[2] == '/')
549 /* If the characters after the colon and before the next slash
550 or end of string are all digits, it's HTTP. */
552 for (pp = p + 1; ISDIGIT (*pp); pp++)
554 if (digits > 0 && (*pp == '/' || *pp == '\0'))
557 /* Prepend "ftp://" to the entire URL... */
558 res = xmalloc (6 + strlen (url) + 1);
559 sprintf (res, "ftp://%s", url);
560 /* ...and replace ':' with '/'. */
561 res[6 + (p - url)] = '/';
568 /* Just prepend "http://" to what we have. */
569 res = xmalloc (7 + strlen (url) + 1);
570 sprintf (res, "http://%s", url);
575 static void split_path (const char *, char **, char **);
577 /* Like strpbrk, with the exception that it returns the pointer to the
578 terminating zero (end-of-string aka "eos") if no matching character
581 Although I normally balk at Gcc-specific optimizations, it probably
582 makes sense here: glibc has optimizations that detect strpbrk being
583 called with literal string as ACCEPT and inline the search. That
584 optimization is defeated if strpbrk is hidden within the call to
585 another function. (And no, making strpbrk_or_eos inline doesn't
586 help because the check for literal accept is in the
589 #if defined(__GNUC__) && __GNUC__ >= 3
591 #define strpbrk_or_eos(s, accept) ({ \
592 char *SOE_p = strpbrk (s, accept); \
594 SOE_p = strchr (s, '\0'); \
598 #else /* not __GNUC__ or old gcc */
601 strpbrk_or_eos (const char *s, const char *accept)
603 char *p = strpbrk (s, accept);
605 p = strchr (s, '\0');
608 #endif /* not __GNUC__ or old gcc */
610 /* Turn STR into lowercase; return true if a character was actually
614 lowercase_str (char *str)
616 bool changed = false;
621 *str = TOLOWER (*str);
627 init_seps (enum url_scheme scheme)
629 static char seps[8] = ":/";
631 int flags = supported_schemes[scheme].flags;
633 if (flags & scm_has_params)
635 if (flags & scm_has_query)
637 if (flags & scm_has_fragment)
643 static const char *parse_errors[] = {
644 #define PE_NO_ERROR 0
646 #define PE_UNSUPPORTED_SCHEME 1
647 N_("Unsupported scheme"),
648 #define PE_INVALID_HOST_NAME 2
649 N_("Invalid host name"),
650 #define PE_BAD_PORT_NUMBER 3
651 N_("Bad port number"),
652 #define PE_INVALID_USER_NAME 4
653 N_("Invalid user name"),
654 #define PE_UNTERMINATED_IPV6_ADDRESS 5
655 N_("Unterminated IPv6 numeric address"),
656 #define PE_IPV6_NOT_SUPPORTED 6
657 N_("IPv6 addresses not supported"),
658 #define PE_INVALID_IPV6_ADDRESS 7
659 N_("Invalid IPv6 numeric address")
664 Return a new struct url if successful, NULL on error. In case of
665 error, and if ERROR is not NULL, also set *ERROR to the appropriate
668 url_parse (const char *url, int *error)
672 bool path_modified, host_modified;
674 enum url_scheme scheme;
677 const char *uname_b, *uname_e;
678 const char *host_b, *host_e;
679 const char *path_b, *path_e;
680 const char *params_b, *params_e;
681 const char *query_b, *query_e;
682 const char *fragment_b, *fragment_e;
685 char *user = NULL, *passwd = NULL;
687 char *url_encoded = NULL;
691 scheme = url_scheme (url);
692 if (scheme == SCHEME_INVALID)
694 error_code = PE_UNSUPPORTED_SCHEME;
698 url_encoded = reencode_escapes (url);
701 p += strlen (supported_schemes[scheme].leading_string);
703 p = url_skip_credentials (p);
706 /* scheme://user:pass@host[:port]... */
709 /* We attempt to break down the URL into the components path,
710 params, query, and fragment. They are ordered like this:
712 scheme://host[:port][/path][;params][?query][#fragment] */
714 path_b = path_e = NULL;
715 params_b = params_e = NULL;
716 query_b = query_e = NULL;
717 fragment_b = fragment_e = NULL;
719 /* Initialize separators for optional parts of URL, depending on the
720 scheme. For example, FTP has params, and HTTP and HTTPS have
721 query string and fragment. */
722 seps = init_seps (scheme);
728 /* Handle IPv6 address inside square brackets. Ideally we'd
729 just look for the terminating ']', but rfc2732 mandates
730 rejecting invalid IPv6 addresses. */
732 /* The address begins after '['. */
734 host_e = strchr (host_b, ']');
738 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
743 /* Check if the IPv6 address is valid. */
744 if (!is_valid_ipv6_address(host_b, host_e))
746 error_code = PE_INVALID_IPV6_ADDRESS;
750 /* Continue parsing after the closing ']'. */
753 error_code = PE_IPV6_NOT_SUPPORTED;
757 /* The closing bracket must be followed by a separator or by the
759 /* http://[::1]... */
761 if (!strchr (seps, *p))
763 /* Trailing garbage after []-delimited IPv6 address. */
764 error_code = PE_INVALID_HOST_NAME;
770 p = strpbrk_or_eos (p, seps);
773 ++seps; /* advance to '/' */
775 if (host_b == host_e)
777 error_code = PE_INVALID_HOST_NAME;
781 port = scheme_default_port (scheme);
784 const char *port_b, *port_e, *pp;
786 /* scheme://host:port/tralala */
790 p = strpbrk_or_eos (p, seps);
793 /* Allow empty port, as per rfc2396. */
794 if (port_b != port_e)
795 for (port = 0, pp = port_b; pp < port_e; pp++)
799 /* http://host:12randomgarbage/blah */
801 error_code = PE_BAD_PORT_NUMBER;
804 port = 10 * port + (*pp - '0');
805 /* Check for too large port numbers here, before we have
806 a chance to overflow on bogus port values. */
809 error_code = PE_BAD_PORT_NUMBER;
814 /* Advance to the first separator *after* '/' (either ';' or '?',
815 depending on the scheme). */
818 /* Get the optional parts of URL, each part being delimited by
819 current location and the position of the next separator. */
820 #define GET_URL_PART(sepchar, var) do { \
822 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
826 GET_URL_PART ('/', path);
827 if (supported_schemes[scheme].flags & scm_has_params)
828 GET_URL_PART (';', params);
829 if (supported_schemes[scheme].flags & scm_has_query)
830 GET_URL_PART ('?', query);
831 if (supported_schemes[scheme].flags & scm_has_fragment)
832 GET_URL_PART ('#', fragment);
837 if (uname_b != uname_e)
839 /* http://user:pass@host */
841 /* uname_b uname_e */
842 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
844 error_code = PE_INVALID_USER_NAME;
849 u = xnew0 (struct url);
851 u->host = strdupdelim (host_b, host_e);
856 u->path = strdupdelim (path_b, path_e);
857 path_modified = path_simplify (u->path);
858 split_path (u->path, &u->dir, &u->file);
860 host_modified = lowercase_str (u->host);
862 /* Decode %HH sequences in host name. This is important not so much
863 to support %HH sequences in host names (which other browser
864 don't), but to support binary characters (which will have been
865 converted to %HH by reencode_escapes). */
866 if (strchr (u->host, '%'))
868 url_unescape (u->host);
869 host_modified = true;
873 u->params = strdupdelim (params_b, params_e);
875 u->query = strdupdelim (query_b, query_e);
877 u->fragment = strdupdelim (fragment_b, fragment_e);
879 if (path_modified || u->fragment || host_modified || path_b == path_e)
881 /* If we suspect that a transformation has rendered what
882 url_string might return different from URL_ENCODED, rebuild
883 u->url using url_string. */
884 u->url = url_string (u, false);
886 if (url_encoded != url)
887 xfree ((char *) url_encoded);
891 if (url_encoded == url)
892 u->url = xstrdup (url);
894 u->url = url_encoded;
900 /* Cleanup in case of error: */
901 if (url_encoded && url_encoded != url)
904 /* Transmit the error code to the caller, if the caller wants to
911 /* Return the error message string from ERROR_CODE, which should have
912 been retrieved from url_parse. The error message is translated. */
915 url_error (int error_code)
917 assert (error_code >= 0 && error_code < countof (parse_errors));
918 return _(parse_errors[error_code]);
921 /* Split PATH into DIR and FILE. PATH comes from the URL and is
922 expected to be URL-escaped.
924 The path is split into directory (the part up to the last slash)
925 and file (the part after the last slash), which are subsequently
929 "foo/bar/baz" "foo/bar" "baz"
930 "foo/bar/" "foo/bar" ""
932 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
934 DIR and FILE are freshly allocated. */
937 split_path (const char *path, char **dir, char **file)
939 char *last_slash = strrchr (path, '/');
943 *file = xstrdup (path);
947 *dir = strdupdelim (path, last_slash);
948 *file = xstrdup (last_slash + 1);
951 url_unescape (*file);
954 /* Note: URL's "full path" is the path with the query string and
955 params appended. The "fragment" (#foo) is intentionally ignored,
956 but that might be changed. For example, if the original URL was
957 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
958 the full path will be "/foo/bar/baz;bullshit?querystring". */
960 /* Return the length of the full path, without the terminating
964 full_path_length (const struct url *url)
968 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
979 /* Write out the full path. */
982 full_path_write (const struct url *url, char *where)
984 #define FROB(el, chr) do { \
985 char *f_el = url->el; \
987 int l = strlen (f_el); \
989 memcpy (where, f_el, l); \
1001 /* Public function for getting the "full path". E.g. if u->path is
1002 "foo/bar" and u->query is "param=value", full_path will be
1003 "/foo/bar?param=value". */
1006 url_full_path (const struct url *url)
1008 int length = full_path_length (url);
1009 char *full_path = xmalloc (length + 1);
1011 full_path_write (url, full_path);
1012 full_path[length] = '\0';
1017 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1018 escaping of certain characters, such as "/" and ":". Returns a
1019 count of unescaped chars. */
1022 unescape_single_char (char *str, char chr)
1024 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1025 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1026 char *h = str; /* hare */
1027 char *t = str; /* tortoise */
1028 for (; *h; h++, t++)
1030 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1041 /* Escape unsafe and reserved characters, except for the slash
1045 url_escape_dir (const char *dir)
1047 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1051 unescape_single_char (newdir, '/');
1055 /* Sync u->path and u->url with u->dir and u->file. Called after
1056 u->file or u->dir have been changed, typically by the FTP code. */
1059 sync_path (struct url *u)
1061 char *newpath, *efile, *edir;
1065 /* u->dir and u->file are not escaped. URL-escape them before
1066 reassembling them into u->path. That way, if they contain
1067 separators like '?' or even if u->file contains slashes, the
1068 path will be correctly assembled. (u->file can contain slashes
1069 if the URL specifies it with %2f, or if an FTP server returns
1071 edir = url_escape_dir (u->dir);
1072 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1075 newpath = xstrdup (efile);
1078 int dirlen = strlen (edir);
1079 int filelen = strlen (efile);
1081 /* Copy "DIR/FILE" to newpath. */
1082 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1083 memcpy (p, edir, dirlen);
1086 memcpy (p, efile, filelen);
1095 if (efile != u->file)
1098 /* Regenerate u->url as well. */
1100 u->url = url_string (u, false);
1103 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1104 This way we can sync u->path and u->url when they get changed. */
1107 url_set_dir (struct url *url, const char *newdir)
1110 url->dir = xstrdup (newdir);
1115 url_set_file (struct url *url, const char *newfile)
1118 url->file = xstrdup (newfile);
1123 url_free (struct url *url)
1129 xfree_null (url->params);
1130 xfree_null (url->query);
1131 xfree_null (url->fragment);
1132 xfree_null (url->user);
1133 xfree_null (url->passwd);
1141 /* Create all the necessary directories for PATH (a file). Calls
1142 make_directory internally. */
1144 mkalldirs (const char *path)
1151 p = path + strlen (path);
1152 for (; *p != '/' && p != path; p--)
1155 /* Don't create if it's just a file. */
1156 if ((p == path) && (*p != '/'))
1158 t = strdupdelim (path, p);
1160 /* Check whether the directory exists. */
1161 if ((stat (t, &st) == 0))
1163 if (S_ISDIR (st.st_mode))
1170 /* If the dir exists as a file name, remove it first. This
1171 is *only* for Wget to work with buggy old CERN http
1172 servers. Here is the scenario: When Wget tries to
1173 retrieve a directory without a slash, e.g.
1174 http://foo/bar (bar being a directory), CERN server will
1175 not redirect it too http://foo/bar/ -- it will generate a
1176 directory listing containing links to bar/file1,
1177 bar/file2, etc. Wget will lose because it saves this
1178 HTML listing to a file `bar', so it cannot create the
1179 directory. To work around this, if the file of the same
1180 name exists, we just remove it and create the directory
1182 DEBUGP (("Removing %s because of directory danger!\n", t));
1186 res = make_directory (t);
1188 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1193 /* Functions for constructing the file name out of URL components. */
1195 /* A growable string structure, used by url_file_name and friends.
1196 This should perhaps be moved to utils.c.
1198 The idea is to have a convenient and efficient way to construct a
1199 string by having various functions append data to it. Instead of
1200 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1201 functions in questions, we pass the pointer to this struct. */
1209 /* Ensure that the string can accept APPEND_COUNT more characters past
1210 the current TAIL position. If necessary, this will grow the string
1211 and update its allocated size. If the string is already large
1212 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1213 #define GROW(g, append_size) do { \
1214 struct growable *G_ = g; \
1215 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1218 /* Return the tail position of the string. */
1219 #define TAIL(r) ((r)->base + (r)->tail)
1221 /* Move the tail position by APPEND_COUNT characters. */
1222 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1224 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1228 append_string (const char *str, struct growable *dest)
1230 int l = strlen (str);
1232 memcpy (TAIL (dest), str, l);
1233 TAIL_INCR (dest, l);
1236 /* Append CH to DEST. For example, append_char (0, DEST)
1237 zero-terminates DEST. */
1240 append_char (char ch, struct growable *dest)
1244 TAIL_INCR (dest, 1);
1248 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1249 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1250 filechr_control = 4 /* a control character, e.g. 0-31 */
1253 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1255 /* Shorthands for the table: */
1256 #define U filechr_not_unix
1257 #define W filechr_not_windows
1258 #define C filechr_control
1263 /* Table of characters unsafe under various conditions (see above).
1265 Arguably we could also claim `%' to be unsafe, since we use it as
1266 the escape character. If we ever want to be able to reliably
1267 translate file name back to URL, this would become important
1268 crucial. Right now, it's better to be minimal in escaping. */
1270 static const unsigned char filechr_table[256] =
1272 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1273 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1274 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1275 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1276 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1277 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1278 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1279 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1280 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1281 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1282 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1283 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1284 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1285 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1286 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1287 0, 0, 0, 0, 0, 0, 0, 0, /* x y z { | } ~ DEL */
1289 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1290 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1291 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1292 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
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,
1296 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,
1305 /* FN_PORT_SEP is the separator between host and port in file names
1306 for non-standard port numbers. On Unix this is normally ':', as in
1307 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1308 because Windows can't handle ':' in file names. */
1309 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1311 /* FN_QUERY_SEP is the separator between the file name and the URL
1312 query, normally '?'. Since Windows cannot handle '?' as part of
1313 file name, we use '@' instead there. */
1314 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1316 /* Quote path element, characters in [b, e), as file name, and append
1317 the quoted string to DEST. Each character is quoted as per
1318 file_unsafe_char and the corresponding table.
1320 If ESCAPED is true, the path element is considered to be
1321 URL-escaped and will be unescaped prior to inspection. */
1324 append_uri_pathel (const char *b, const char *e, bool escaped,
1325 struct growable *dest)
1331 if (opt.restrict_files_os == restrict_unix)
1332 mask = filechr_not_unix;
1334 mask = filechr_not_windows;
1335 if (opt.restrict_files_ctrl)
1336 mask |= filechr_control;
1338 /* Copy [b, e) to PATHEL and URL-unescape it. */
1342 BOUNDED_TO_ALLOCA (b, e, unescaped);
1343 url_unescape (unescaped);
1345 e = unescaped + strlen (unescaped);
1348 /* Defang ".." when found as component of path. Remember that path
1349 comes from the URL and might contain malicious input. */
1350 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1356 /* Walk the PATHEL string and check how many characters we'll need
1359 for (p = b; p < e; p++)
1360 if (FILE_CHAR_TEST (*p, mask))
1363 /* Calculate the length of the output string. e-b is the input
1364 string length. Each quoted char introduces two additional
1365 characters in the string, hence 2*quoted. */
1366 outlen = (e - b) + (2 * quoted);
1367 GROW (dest, outlen);
1371 /* If there's nothing to quote, we can simply append the string
1372 without processing it again. */
1373 memcpy (TAIL (dest), b, outlen);
1377 char *q = TAIL (dest);
1378 for (p = b; p < e; p++)
1380 if (!FILE_CHAR_TEST (*p, mask))
1384 unsigned char ch = *p;
1386 *q++ = XNUM_TO_DIGIT (ch >> 4);
1387 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1390 assert (q - TAIL (dest) == outlen);
1392 TAIL_INCR (dest, outlen);
1395 /* Append to DEST the directory structure that corresponds the
1396 directory part of URL's path. For example, if the URL is
1397 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1399 Each path element ("dir1" and "dir2" in the above example) is
1400 examined, url-unescaped, and re-escaped as file name element.
1402 Additionally, it cuts as many directories from the path as
1403 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1404 will produce "bar" for the above example. For 2 or more, it will
1407 Each component of the path is quoted for use as file name. */
1410 append_dir_structure (const struct url *u, struct growable *dest)
1412 char *pathel, *next;
1413 int cut = opt.cut_dirs;
1415 /* Go through the path components, de-URL-quote them, and quote them
1416 (if necessary) as file names. */
1419 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1424 /* Ignore empty pathels. */
1428 append_char ('/', dest);
1429 append_uri_pathel (pathel, next, true, dest);
1433 /* Return a unique file name that matches the given URL as good as
1434 possible. Does not create directories on the file system. */
1437 url_file_name (const struct url *u)
1439 struct growable fnres; /* stands for "file name result" */
1441 const char *u_file, *u_query;
1442 char *fname, *unique;
1448 /* Start with the directory prefix, if specified. */
1450 append_string (opt.dir_prefix, &fnres);
1452 /* If "dirstruct" is turned on (typically the case with -r), add
1453 the host and port (unless those have been turned off) and
1454 directory structure. */
1457 if (opt.protocol_directories)
1460 append_char ('/', &fnres);
1461 append_string (supported_schemes[u->scheme].name, &fnres);
1463 if (opt.add_hostdir)
1466 append_char ('/', &fnres);
1467 if (0 != strcmp (u->host, ".."))
1468 append_string (u->host, &fnres);
1470 /* Host name can come from the network; malicious DNS may
1471 allow ".." to be resolved, causing us to write to
1472 "../<file>". Defang such host names. */
1473 append_string ("%2E%2E", &fnres);
1474 if (u->port != scheme_default_port (u->scheme))
1477 number_to_string (portstr, u->port);
1478 append_char (FN_PORT_SEP, &fnres);
1479 append_string (portstr, &fnres);
1483 append_dir_structure (u, &fnres);
1486 /* Add the file name. */
1488 append_char ('/', &fnres);
1489 u_file = *u->file ? u->file : "index.html";
1490 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1492 /* Append "?query" to the file name. */
1493 u_query = u->query && *u->query ? u->query : NULL;
1496 append_char (FN_QUERY_SEP, &fnres);
1497 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1500 /* Zero-terminate the file name. */
1501 append_char ('\0', &fnres);
1505 /* Check the cases in which the unique extensions are not used:
1506 1) Clobbering is turned off (-nc).
1507 2) Retrieval with regetting.
1508 3) Timestamping is used.
1509 4) Hierarchy is built.
1511 The exception is the case when file does exist and is a
1512 directory (see `mkalldirs' for explanation). */
1514 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1515 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1518 unique = unique_name (fname, true);
1519 if (unique != fname)
1524 /* Resolve "." and ".." elements of PATH by destructively modifying
1525 PATH and return true if PATH has been modified, false otherwise.
1527 The algorithm is in spirit similar to the one described in rfc1808,
1528 although implemented differently, in one pass. To recap, path
1529 elements containing only "." are removed, and ".." is taken to mean
1530 "back up one element". Single leading and trailing slashes are
1533 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1534 test examples are provided below. If you change anything in this
1535 function, run test_path_simplify to make sure you haven't broken a
1539 path_simplify (char *path)
1541 char *h = path; /* hare */
1542 char *t = path; /* tortoise */
1543 char *beg = path; /* boundary for backing the tortoise */
1544 char *end = path + strlen (path);
1548 /* Hare should be at the beginning of a path element. */
1550 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1555 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1557 /* Handle "../" by retreating the tortoise by one path
1558 element -- but not past beggining. */
1561 /* Move backwards until T hits the beginning of the
1562 previous path element or the beginning of path. */
1563 for (--t; t > beg && t[-1] != '/'; t--)
1568 /* If we're at the beginning, copy the "../" literally
1569 move the beginning so a later ".." doesn't remove
1579 /* A regular path element. If H hasn't advanced past T,
1580 simply skip to the next path element. Otherwise, copy
1581 the path element until the next slash. */
1584 /* Skip the path element, including the slash. */
1585 while (h < end && *h != '/')
1592 /* Copy the path element, including the final slash. */
1593 while (h < end && *h != '/')
1607 /* Return the length of URL's path. Path is considered to be
1608 terminated by one of '?', ';', '#', or by the end of the
1612 path_length (const char *url)
1614 const char *q = strpbrk_or_eos (url, "?;#");
1618 /* Find the last occurrence of character C in the range [b, e), or
1619 NULL, if none are present. We might want to use memrchr (a GNU
1620 extension) under GNU libc. */
1623 find_last_char (const char *b, const char *e, char c)
1631 /* Merge BASE with LINK and return the resulting URI.
1633 Either of the URIs may be absolute or relative, complete with the
1634 host name, or path only. This tries to reasonably handle all
1635 foreseeable cases. It only employs minimal URL parsing, without
1636 knowledge of the specifics of schemes.
1638 I briefly considered making this function call path_simplify after
1639 the merging process, as rfc1738 seems to suggest. This is a bad
1640 idea for several reasons: 1) it complexifies the code, and 2)
1641 url_parse has to simplify path anyway, so it's wasteful to boot. */
1644 uri_merge (const char *base, const char *link)
1650 if (url_has_scheme (link))
1651 return xstrdup (link);
1653 /* We may not examine BASE past END. */
1654 end = base + path_length (base);
1655 linklength = strlen (link);
1659 /* Empty LINK points back to BASE, query string and all. */
1660 return xstrdup (base);
1662 else if (*link == '?')
1664 /* LINK points to the same location, but changes the query
1665 string. Examples: */
1666 /* uri_merge("path", "?new") -> "path?new" */
1667 /* uri_merge("path?foo", "?new") -> "path?new" */
1668 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1669 /* uri_merge("path#foo", "?new") -> "path?new" */
1670 int baselength = end - base;
1671 merge = xmalloc (baselength + linklength + 1);
1672 memcpy (merge, base, baselength);
1673 memcpy (merge + baselength, link, linklength);
1674 merge[baselength + linklength] = '\0';
1676 else if (*link == '#')
1678 /* uri_merge("path", "#new") -> "path#new" */
1679 /* uri_merge("path#foo", "#new") -> "path#new" */
1680 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1681 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1683 const char *end1 = strchr (base, '#');
1685 end1 = base + strlen (base);
1686 baselength = end1 - base;
1687 merge = xmalloc (baselength + linklength + 1);
1688 memcpy (merge, base, baselength);
1689 memcpy (merge + baselength, link, linklength);
1690 merge[baselength + linklength] = '\0';
1692 else if (*link == '/' && *(link + 1) == '/')
1694 /* LINK begins with "//" and so is a net path: we need to
1695 replace everything after (and including) the double slash
1698 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1699 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1700 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1704 const char *start_insert;
1706 /* Look for first slash. */
1707 slash = memchr (base, '/', end - base);
1708 /* If found slash and it is a double slash, then replace
1709 from this point, else default to replacing from the
1711 if (slash && *(slash + 1) == '/')
1712 start_insert = slash;
1714 start_insert = base;
1716 span = start_insert - base;
1717 merge = xmalloc (span + linklength + 1);
1719 memcpy (merge, base, span);
1720 memcpy (merge + span, link, linklength);
1721 merge[span + linklength] = '\0';
1723 else if (*link == '/')
1725 /* LINK is an absolute path: we need to replace everything
1726 after (and including) the FIRST slash with LINK.
1728 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1729 "/qux/xyzzy", our result should be
1730 "http://host/qux/xyzzy". */
1733 const char *start_insert = NULL; /* for gcc to shut up. */
1734 const char *pos = base;
1735 bool seen_slash_slash = false;
1736 /* We're looking for the first slash, but want to ignore
1739 slash = memchr (pos, '/', end - pos);
1740 if (slash && !seen_slash_slash)
1741 if (*(slash + 1) == '/')
1744 seen_slash_slash = true;
1748 /* At this point, SLASH is the location of the first / after
1749 "//", or the first slash altogether. START_INSERT is the
1750 pointer to the location where LINK will be inserted. When
1751 examining the last two examples, keep in mind that LINK
1754 if (!slash && !seen_slash_slash)
1755 /* example: "foo" */
1757 start_insert = base;
1758 else if (!slash && seen_slash_slash)
1759 /* example: "http://foo" */
1762 else if (slash && !seen_slash_slash)
1763 /* example: "foo/bar" */
1765 start_insert = base;
1766 else if (slash && seen_slash_slash)
1767 /* example: "http://something/" */
1769 start_insert = slash;
1771 span = start_insert - base;
1772 merge = xmalloc (span + linklength + 1);
1774 memcpy (merge, base, span);
1775 memcpy (merge + span, link, linklength);
1776 merge[span + linklength] = '\0';
1780 /* LINK is a relative URL: we need to replace everything
1781 after last slash (possibly empty) with LINK.
1783 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1784 our result should be "whatever/foo/qux/xyzzy". */
1785 bool need_explicit_slash = false;
1787 const char *start_insert;
1788 const char *last_slash = find_last_char (base, end, '/');
1791 /* No slash found at all. Replace what we have with LINK. */
1792 start_insert = base;
1794 else if (last_slash && last_slash >= base + 2
1795 && last_slash[-2] == ':' && last_slash[-1] == '/')
1797 /* example: http://host" */
1799 start_insert = end + 1;
1800 need_explicit_slash = true;
1804 /* example: "whatever/foo/bar" */
1806 start_insert = last_slash + 1;
1809 span = start_insert - base;
1810 merge = xmalloc (span + linklength + 1);
1812 memcpy (merge, base, span);
1813 if (need_explicit_slash)
1814 merge[span - 1] = '/';
1815 memcpy (merge + span, link, linklength);
1816 merge[span + linklength] = '\0';
1822 #define APPEND(p, s) do { \
1823 int len = strlen (s); \
1824 memcpy (p, s, len); \
1828 /* Use this instead of password when the actual password is supposed
1829 to be hidden. We intentionally use a generic string without giving
1830 away the number of characters in the password, like previous
1832 #define HIDDEN_PASSWORD "*password*"
1834 /* Recreate the URL string from the data in URL.
1836 If HIDE is true (as it is when we're calling this on a URL we plan
1837 to print, but not when calling it to canonicalize a URL for use
1838 within the program), password will be hidden. Unsafe characters in
1839 the URL will be quoted. */
1842 url_string (const struct url *url, bool hide_password)
1846 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1848 int scheme_port = supported_schemes[url->scheme].default_port;
1849 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1850 int fplen = full_path_length (url);
1852 bool brackets_around_host;
1854 assert (scheme_str != NULL);
1856 /* Make sure the user name and password are quoted. */
1859 quoted_user = url_escape_allow_passthrough (url->user);
1863 quoted_passwd = HIDDEN_PASSWORD;
1865 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1869 /* In the unlikely event that the host name contains non-printable
1870 characters, quote it for displaying to the user. */
1871 quoted_host = url_escape_allow_passthrough (url->host);
1873 /* Undo the quoting of colons that URL escaping performs. IPv6
1874 addresses may legally contain colons, and in that case must be
1875 placed in square brackets. */
1876 if (quoted_host != url->host)
1877 unescape_single_char (quoted_host, ':');
1878 brackets_around_host = strchr (quoted_host, ':') != NULL;
1880 size = (strlen (scheme_str)
1881 + strlen (quoted_host)
1882 + (brackets_around_host ? 2 : 0)
1885 if (url->port != scheme_port)
1886 size += 1 + numdigit (url->port);
1889 size += 1 + strlen (quoted_user);
1891 size += 1 + strlen (quoted_passwd);
1894 p = result = xmalloc (size);
1896 APPEND (p, scheme_str);
1899 APPEND (p, quoted_user);
1903 APPEND (p, quoted_passwd);
1908 if (brackets_around_host)
1910 APPEND (p, quoted_host);
1911 if (brackets_around_host)
1913 if (url->port != scheme_port)
1916 p = number_to_string (p, url->port);
1919 full_path_write (url, p);
1923 assert (p - result == size);
1925 if (quoted_user && quoted_user != url->user)
1926 xfree (quoted_user);
1927 if (quoted_passwd && !hide_password && quoted_passwd != url->passwd)
1928 xfree (quoted_passwd);
1929 if (quoted_host != url->host)
1930 xfree (quoted_host);
1935 /* Return true if scheme a is similar to scheme b.
1937 Schemes are similar if they are equal. If SSL is supported, schemes
1938 are also similar if one is http (SCHEME_HTTP) and the other is https
1941 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1946 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1947 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1954 /* Debugging and testing support for path_simplify. */
1956 /* Debug: run path_simplify on PATH and return the result in a new
1957 string. Useful for calling from the debugger. */
1961 char *copy = xstrdup (path);
1962 path_simplify (copy);
1967 run_test (char *test, char *expected_result, bool expected_change)
1969 char *test_copy = xstrdup (test);
1970 bool modified = path_simplify (test_copy);
1972 if (0 != strcmp (test_copy, expected_result))
1974 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
1975 test, expected_result, test_copy);
1977 if (modified != expected_change)
1979 if (expected_change)
1980 printf ("Expected modification with path_simplify(\"%s\").\n",
1983 printf ("Expected no modification with path_simplify(\"%s\").\n",
1990 test_path_simplify (void)
1993 char *test, *result;
1999 { "..", "..", false },
2000 { "../", "../", false },
2001 { "foo", "foo", false },
2002 { "foo/bar", "foo/bar", false },
2003 { "foo///bar", "foo///bar", false },
2004 { "foo/.", "foo/", true },
2005 { "foo/./", "foo/", true },
2006 { "foo./", "foo./", false },
2007 { "foo/../bar", "bar", true },
2008 { "foo/../bar/", "bar/", true },
2009 { "foo/bar/..", "foo/", true },
2010 { "foo/bar/../x", "foo/x", true },
2011 { "foo/bar/../x/", "foo/x/", true },
2012 { "foo/..", "", true },
2013 { "foo/../..", "..", true },
2014 { "foo/../../..", "../..", true },
2015 { "foo/../../bar/../../baz", "../../baz", true },
2016 { "a/b/../../c", "c", true },
2017 { "./a/../b", "b", true }
2021 for (i = 0; i < countof (tests); i++)
2023 char *test = tests[i].test;
2024 char *expected_result = tests[i].result;
2025 bool expected_change = tests[i].should_modify;
2026 run_test (test, expected_result, expected_change);