2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
3 2005, 2006, 2007, 2008, 2009, 2010 Free Software Foundation, Inc.
5 This file is part of GNU Wget.
7 GNU Wget is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or (at
10 your option) any later version.
12 GNU Wget is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with Wget. If not, see <http://www.gnu.org/licenses/>.
20 Additional permission under GNU GPL version 3 section 7
22 If you modify this program, or any covered work, by linking or
23 combining it with the OpenSSL project's OpenSSL library (or a
24 modified version of that library), containing parts covered by the
25 terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
26 grants you additional permission to convey the resulting work.
27 Corresponding Source for a non-source form of such a combination
28 shall include the source code for the parts of OpenSSL used as well
29 as that of the covered work. */
44 #include "host.h" /* for is_valid_ipv6_address */
48 #endif /* def __VMS */
55 scm_disabled = 1, /* for https when OpenSSL fails to init. */
56 scm_has_params = 2, /* whether scheme has ;params */
57 scm_has_query = 4, /* whether scheme has ?query */
58 scm_has_fragment = 8 /* whether scheme has #fragment */
63 /* Short name of the scheme, such as "http" or "ftp". */
65 /* Leading string that identifies the scheme, such as "https://". */
66 const char *leading_string;
67 /* Default port of the scheme when none is specified. */
73 /* Supported schemes: */
74 static struct scheme_data supported_schemes[] =
76 { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment },
78 { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment },
80 { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment },
86 /* Forward declarations: */
88 static bool path_simplify (enum url_scheme, char *);
90 /* Support for escaping and unescaping of URL strings. */
92 /* Table of "reserved" and "unsafe" characters. Those terms are
93 rfc1738-speak, as such largely obsoleted by rfc2396 and later
94 specs, but the general idea remains.
96 A reserved character is the one that you can't decode without
97 changing the meaning of the URL. For example, you can't decode
98 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
99 path components is different. Non-reserved characters can be
100 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
101 unsafe characters are loosely based on rfc1738, plus "$" and ",",
102 as recommended by rfc2396, and minus "~", which is very frequently
103 used (and sometimes unrecognized as %7E by broken servers).
105 An unsafe character is the one that should be encoded when URLs are
106 placed in foreign environments. E.g. space and newline are unsafe
107 in HTTP contexts because HTTP uses them as separator and line
108 terminator, so they must be encoded to %20 and %0A respectively.
109 "*" is unsafe in shell context, etc.
111 We determine whether a character is unsafe through static table
112 lookup. This code assumes ASCII character set and 8-bit chars. */
115 /* rfc1738 reserved chars + "$" and ",". */
118 /* rfc1738 unsafe chars, plus non-printables. */
122 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
123 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
124 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
126 /* Shorthands for the table: */
127 #define R urlchr_reserved
128 #define U urlchr_unsafe
131 static const unsigned char urlchr_table[256] =
133 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
134 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
135 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
136 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
137 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
138 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
139 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
140 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
141 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
142 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
143 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
144 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
145 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
146 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
147 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
148 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
150 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
151 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
152 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
153 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
155 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
156 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
157 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
158 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
164 /* URL-unescape the string S.
166 This is done by transforming the sequences "%HH" to the character
167 represented by the hexadecimal digits HH. If % is not followed by
168 two hexadecimal digits, it is inserted literally.
170 The transformation is done in place. If you need the original
171 string intact, make a copy before calling this function. */
174 url_unescape (char *s)
176 char *t = s; /* t - tortoise */
177 char *h = s; /* h - hare */
189 /* Do nothing if '%' is not followed by two hex digits. */
190 if (!h[1] || !h[2] || !(c_isxdigit (h[1]) && c_isxdigit (h[2])))
192 c = X2DIGITS_TO_NUM (h[1], h[2]);
193 /* Don't unescape %00 because there is no way to insert it
194 into a C string without effectively truncating it. */
204 /* The core of url_escape_* functions. Escapes the characters that
205 match the provided mask in urlchr_table.
207 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
208 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
209 allocated string will be returned in all cases. */
212 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
219 for (p1 = s; *p1; p1++)
220 if (urlchr_test (*p1, mask))
221 addition += 2; /* Two more characters (hex digits) */
224 return allow_passthrough ? (char *)s : xstrdup (s);
226 newlen = (p1 - s) + addition;
227 newstr = xmalloc (newlen + 1);
233 /* Quote the characters that match the test mask. */
234 if (urlchr_test (*p1, mask))
236 unsigned char c = *p1++;
238 *p2++ = XNUM_TO_DIGIT (c >> 4);
239 *p2++ = XNUM_TO_DIGIT (c & 0xf);
244 assert (p2 - newstr == newlen);
250 /* URL-escape the unsafe characters (see urlchr_table) in a given
251 string, returning a freshly allocated string. */
254 url_escape (const char *s)
256 return url_escape_1 (s, urlchr_unsafe, false);
259 /* URL-escape the unsafe and reserved characters (see urlchr_table) in
260 a given string, returning a freshly allocated string. */
263 url_escape_unsafe_and_reserved (const char *s)
265 return url_escape_1 (s, urlchr_unsafe|urlchr_reserved, false);
268 /* URL-escape the unsafe characters (see urlchr_table) in a given
269 string. If no characters are unsafe, S is returned. */
272 url_escape_allow_passthrough (const char *s)
274 return url_escape_1 (s, urlchr_unsafe, true);
277 /* Decide whether the char at position P needs to be encoded. (It is
278 not enough to pass a single char *P because the function may need
279 to inspect the surrounding context.)
281 Return true if the char should be escaped as %XX, false otherwise. */
284 char_needs_escaping (const char *p)
288 if (c_isxdigit (*(p + 1)) && c_isxdigit (*(p + 2)))
291 /* Garbled %.. sequence: encode `%'. */
294 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
300 /* Translate a %-escaped (but possibly non-conformant) input string S
301 into a %-escaped (and conformant) output string. If no characters
302 are encoded or decoded, return the same string S; otherwise, return
303 a freshly allocated string with the new contents.
305 After a URL has been run through this function, the protocols that
306 use `%' as the quote character can use the resulting string as-is,
307 while those that don't can use url_unescape to get to the intended
308 data. This function is stable: once the input is transformed,
309 further transformations of the result yield the same output.
311 Let's discuss why this function is needed.
313 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
314 a raw space character would mess up the HTTP request, it needs to
315 be quoted, like this:
317 GET /abc%20def HTTP/1.0
319 It would appear that the unsafe chars need to be quoted, for
320 example with url_escape. But what if we're requested to download
321 `abc%20def'? url_escape transforms "%" to "%25", which would leave
322 us with `abc%2520def'. This is incorrect -- since %-escapes are
323 part of URL syntax, "%20" is the correct way to denote a literal
324 space on the Wget command line. This leads to the conclusion that
325 in that case Wget should not call url_escape, but leave the `%20'
326 as is. This is clearly contradictory, but it only gets worse.
328 What if the requested URI is `abc%20 def'? If we call url_escape,
329 we end up with `/abc%2520%20def', which is almost certainly not
330 intended. If we don't call url_escape, we are left with the
331 embedded space and cannot complete the request. What the user
332 meant was for Wget to request `/abc%20%20def', and this is where
333 reencode_escapes kicks in.
335 Wget used to solve this by first decoding %-quotes, and then
336 encoding all the "unsafe" characters found in the resulting string.
337 This was wrong because it didn't preserve certain URL special
338 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
339 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
340 whether we considered `+' reserved (it is). One of these results
341 is inevitable because by the second step we would lose information
342 on whether the `+' was originally encoded or not. Both results
343 were wrong because in CGI parameters + means space, while %2B means
344 literal plus. reencode_escapes correctly translates the above to
345 "a%2B+b", i.e. returns the original string.
347 This function uses a modified version of the algorithm originally
348 proposed by Anon Sricharoenchai:
350 * Encode all "unsafe" characters, except those that are also
351 "reserved", to %XX. See urlchr_table for which characters are
354 * Encode the "%" characters not followed by two hex digits to
357 * Pass through all other characters and %XX escapes as-is. (Up to
358 Wget 1.10 this decoded %XX escapes corresponding to "safe"
359 characters, but that was obtrusive and broke some servers.)
363 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
365 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
369 "foo bar" -> "foo%20bar"
370 "foo%20bar" -> "foo%20bar"
371 "foo %20bar" -> "foo%20%20bar"
372 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
373 "foo%25%20bar" -> "foo%25%20bar"
374 "foo%2%20bar" -> "foo%252%20bar"
375 "foo+bar" -> "foo+bar" (plus is reserved!)
376 "foo%2b+bar" -> "foo%2b+bar" */
379 reencode_escapes (const char *s)
385 int encode_count = 0;
387 /* First pass: inspect the string to see if there's anything to do,
388 and to calculate the new length. */
389 for (p1 = s; *p1; p1++)
390 if (char_needs_escaping (p1))
394 /* The string is good as it is. */
395 return (char *) s; /* C const model sucks. */
398 /* Each encoding adds two characters (hex digits). */
399 newlen = oldlen + 2 * encode_count;
400 newstr = xmalloc (newlen + 1);
402 /* Second pass: copy the string to the destination address, encoding
403 chars when needed. */
408 if (char_needs_escaping (p1))
410 unsigned char c = *p1++;
412 *p2++ = XNUM_TO_DIGIT (c >> 4);
413 *p2++ = XNUM_TO_DIGIT (c & 0xf);
419 assert (p2 - newstr == newlen);
423 /* Returns the scheme type if the scheme is supported, or
424 SCHEME_INVALID if not. */
427 url_scheme (const char *url)
431 for (i = 0; supported_schemes[i].leading_string; i++)
432 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
433 strlen (supported_schemes[i].leading_string)))
435 if (!(supported_schemes[i].flags & scm_disabled))
436 return (enum url_scheme) i;
438 return SCHEME_INVALID;
441 return SCHEME_INVALID;
444 #define SCHEME_CHAR(ch) (c_isalnum (ch) || (ch) == '-' || (ch) == '+')
446 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
447 currently implemented, it returns true if URL begins with
451 url_has_scheme (const char *url)
455 /* The first char must be a scheme char. */
456 if (!*p || !SCHEME_CHAR (*p))
459 /* Followed by 0 or more scheme chars. */
460 while (*p && SCHEME_CHAR (*p))
462 /* Terminated by ':'. */
467 url_valid_scheme (const char *url)
469 enum url_scheme scheme = url_scheme (url);
470 return scheme != SCHEME_INVALID;
474 scheme_default_port (enum url_scheme scheme)
476 return supported_schemes[scheme].default_port;
480 scheme_disable (enum url_scheme scheme)
482 supported_schemes[scheme].flags |= scm_disabled;
485 /* Skip the username and password, if present in the URL. The
486 function should *not* be called with the complete URL, but with the
487 portion after the scheme.
489 If no username and password are found, return URL. */
492 url_skip_credentials (const char *url)
494 /* Look for '@' that comes before terminators, such as '/', '?',
496 const char *p = (const char *)strpbrk (url, "@/?#;");
502 /* Parse credentials contained in [BEG, END). The region is expected
503 to have come from a URL and is unescaped. */
506 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
512 return false; /* empty user name */
514 colon = memchr (beg, ':', end - beg);
516 return false; /* again empty user name */
520 *passwd = strdupdelim (colon + 1, end);
522 url_unescape (*passwd);
529 *user = strdupdelim (beg, userend);
530 url_unescape (*user);
534 /* Used by main.c: detect URLs written using the "shorthand" URL forms
535 originally popularized by Netscape and NcFTP. HTTP shorthands look
538 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
539 www.foo.com[:port] -> http://www.foo.com[:port]
541 FTP shorthands look like this:
543 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
544 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
546 If the URL needs not or cannot be rewritten, return NULL. */
549 rewrite_shorthand_url (const char *url)
554 if (url_scheme (url) != SCHEME_INVALID)
557 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
559 p = strpbrk (url, ":/");
563 /* If we're looking at "://", it means the URL uses a scheme we
564 don't support, which may include "https" when compiled without
565 SSL support. Don't bogusly rewrite such URLs. */
566 if (p && p[0] == ':' && p[1] == '/' && p[2] == '/')
571 /* Colon indicates ftp, as in foo.bar.com:path. Check for
572 special case of http port number ("localhost:10000"). */
573 int digits = strspn (p + 1, "0123456789");
574 if (digits && (p[1 + digits] == '/' || p[1 + digits] == '\0'))
577 /* Turn "foo.bar.com:path" to "ftp://foo.bar.com/path". */
578 ret = aprintf ("ftp://%s", url);
579 ret[6 + (p - url)] = '/';
584 /* Just prepend "http://" to URL. */
585 ret = aprintf ("http://%s", url);
590 static void split_path (const char *, char **, char **);
592 /* Like strpbrk, with the exception that it returns the pointer to the
593 terminating zero (end-of-string aka "eos") if no matching character
597 strpbrk_or_eos (const char *s, const char *accept)
599 char *p = strpbrk (s, accept);
601 p = strchr (s, '\0');
605 /* Turn STR into lowercase; return true if a character was actually
609 lowercase_str (char *str)
611 bool changed = false;
613 if (c_isupper (*str))
616 *str = c_tolower (*str);
622 init_seps (enum url_scheme scheme)
624 static char seps[8] = ":/";
626 int flags = supported_schemes[scheme].flags;
628 if (flags & scm_has_params)
630 if (flags & scm_has_query)
632 if (flags & scm_has_fragment)
638 static const char *parse_errors[] = {
639 #define PE_NO_ERROR 0
641 #define PE_UNSUPPORTED_SCHEME 1
642 N_("Unsupported scheme %s"), /* support for format token only here */
643 #define PE_MISSING_SCHEME 2
644 N_("Scheme missing"),
645 #define PE_INVALID_HOST_NAME 3
646 N_("Invalid host name"),
647 #define PE_BAD_PORT_NUMBER 4
648 N_("Bad port number"),
649 #define PE_INVALID_USER_NAME 5
650 N_("Invalid user name"),
651 #define PE_UNTERMINATED_IPV6_ADDRESS 6
652 N_("Unterminated IPv6 numeric address"),
653 #define PE_IPV6_NOT_SUPPORTED 7
654 N_("IPv6 addresses not supported"),
655 #define PE_INVALID_IPV6_ADDRESS 8
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, struct iri *iri, bool percent_encode)
669 bool path_modified, host_modified;
671 enum url_scheme scheme;
674 const char *uname_b, *uname_e;
675 const char *host_b, *host_e;
676 const char *path_b, *path_e;
677 const char *params_b, *params_e;
678 const char *query_b, *query_e;
679 const char *fragment_b, *fragment_e;
682 char *user = NULL, *passwd = NULL;
684 const char *url_encoded = NULL;
685 char *new_url = NULL;
689 scheme = url_scheme (url);
690 if (scheme == SCHEME_INVALID)
692 if (url_has_scheme (url))
693 error_code = PE_UNSUPPORTED_SCHEME;
695 error_code = PE_MISSING_SCHEME;
699 if (iri && iri->utf8_encode)
701 iri->utf8_encode = remote_to_utf8 (iri, iri->orig_url ? iri->orig_url : url, (const char **) &new_url);
702 if (!iri->utf8_encode)
705 iri->orig_url = xstrdup (url);
708 /* XXX XXX Could that change introduce (security) bugs ??? XXX XXX*/
710 url_encoded = reencode_escapes (new_url ? new_url : url);
712 url_encoded = new_url ? new_url : url;
716 if (new_url && url_encoded != new_url)
719 p += strlen (supported_schemes[scheme].leading_string);
721 p = url_skip_credentials (p);
724 /* scheme://user:pass@host[:port]... */
727 /* We attempt to break down the URL into the components path,
728 params, query, and fragment. They are ordered like this:
730 scheme://host[:port][/path][;params][?query][#fragment] */
732 path_b = path_e = NULL;
733 params_b = params_e = NULL;
734 query_b = query_e = NULL;
735 fragment_b = fragment_e = NULL;
737 /* Initialize separators for optional parts of URL, depending on the
738 scheme. For example, FTP has params, and HTTP and HTTPS have
739 query string and fragment. */
740 seps = init_seps (scheme);
746 /* Handle IPv6 address inside square brackets. Ideally we'd
747 just look for the terminating ']', but rfc2732 mandates
748 rejecting invalid IPv6 addresses. */
750 /* The address begins after '['. */
752 host_e = strchr (host_b, ']');
756 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
761 /* Check if the IPv6 address is valid. */
762 if (!is_valid_ipv6_address(host_b, host_e))
764 error_code = PE_INVALID_IPV6_ADDRESS;
768 /* Continue parsing after the closing ']'. */
771 error_code = PE_IPV6_NOT_SUPPORTED;
775 /* The closing bracket must be followed by a separator or by the
777 /* http://[::1]... */
779 if (!strchr (seps, *p))
781 /* Trailing garbage after []-delimited IPv6 address. */
782 error_code = PE_INVALID_HOST_NAME;
788 p = strpbrk_or_eos (p, seps);
791 ++seps; /* advance to '/' */
793 if (host_b == host_e)
795 error_code = PE_INVALID_HOST_NAME;
799 port = scheme_default_port (scheme);
802 const char *port_b, *port_e, *pp;
804 /* scheme://host:port/tralala */
808 p = strpbrk_or_eos (p, seps);
811 /* Allow empty port, as per rfc2396. */
812 if (port_b != port_e)
813 for (port = 0, pp = port_b; pp < port_e; pp++)
815 if (!c_isdigit (*pp))
817 /* http://host:12randomgarbage/blah */
819 error_code = PE_BAD_PORT_NUMBER;
822 port = 10 * port + (*pp - '0');
823 /* Check for too large port numbers here, before we have
824 a chance to overflow on bogus port values. */
827 error_code = PE_BAD_PORT_NUMBER;
832 /* Advance to the first separator *after* '/' (either ';' or '?',
833 depending on the scheme). */
836 /* Get the optional parts of URL, each part being delimited by
837 current location and the position of the next separator. */
838 #define GET_URL_PART(sepchar, var) do { \
840 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
844 GET_URL_PART ('/', path);
845 if (supported_schemes[scheme].flags & scm_has_params)
846 GET_URL_PART (';', params);
847 if (supported_schemes[scheme].flags & scm_has_query)
848 GET_URL_PART ('?', query);
849 if (supported_schemes[scheme].flags & scm_has_fragment)
850 GET_URL_PART ('#', fragment);
855 if (uname_b != uname_e)
857 /* http://user:pass@host */
859 /* uname_b uname_e */
860 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
862 error_code = PE_INVALID_USER_NAME;
867 u = xnew0 (struct url);
869 u->host = strdupdelim (host_b, host_e);
874 u->path = strdupdelim (path_b, path_e);
875 path_modified = path_simplify (scheme, u->path);
876 split_path (u->path, &u->dir, &u->file);
878 host_modified = lowercase_str (u->host);
880 /* Decode %HH sequences in host name. This is important not so much
881 to support %HH sequences in host names (which other browser
882 don't), but to support binary characters (which will have been
883 converted to %HH by reencode_escapes). */
884 if (strchr (u->host, '%'))
886 url_unescape (u->host);
887 host_modified = true;
889 /* Apply IDNA regardless of iri->utf8_encode status */
890 if (opt.enable_iri && iri)
892 char *new = idn_encode (iri, u->host);
897 host_modified = true;
903 u->params = strdupdelim (params_b, params_e);
905 u->query = strdupdelim (query_b, query_e);
907 u->fragment = strdupdelim (fragment_b, fragment_e);
909 if (opt.enable_iri || path_modified || u->fragment || host_modified || path_b == path_e)
911 /* If we suspect that a transformation has rendered what
912 url_string might return different from URL_ENCODED, rebuild
913 u->url using url_string. */
914 u->url = url_string (u, URL_AUTH_SHOW);
916 if (url_encoded != url)
917 xfree ((char *) url_encoded);
921 if (url_encoded == url)
922 u->url = xstrdup (url);
924 u->url = (char *) url_encoded;
930 /* Cleanup in case of error: */
931 if (url_encoded && url_encoded != url)
932 xfree ((char *) url_encoded);
934 /* Transmit the error code to the caller, if the caller wants to
941 /* Return the error message string from ERROR_CODE, which should have
942 been retrieved from url_parse. The error message is translated. */
945 url_error (const char *url, int error_code)
947 assert (error_code >= 0 && ((size_t) error_code) < countof (parse_errors));
949 if (error_code == PE_UNSUPPORTED_SCHEME)
952 char *scheme = xstrdup (url);
953 assert (url_has_scheme (url));
955 if ((p = strchr (scheme, ':')))
957 if (!strcasecmp (scheme, "https"))
958 error = aprintf (_("HTTPS support not compiled in"));
960 error = aprintf (_(parse_errors[error_code]), quote (scheme));
966 return xstrdup (_(parse_errors[error_code]));
969 /* Split PATH into DIR and FILE. PATH comes from the URL and is
970 expected to be URL-escaped.
972 The path is split into directory (the part up to the last slash)
973 and file (the part after the last slash), which are subsequently
977 "foo/bar/baz" "foo/bar" "baz"
978 "foo/bar/" "foo/bar" ""
980 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
982 DIR and FILE are freshly allocated. */
985 split_path (const char *path, char **dir, char **file)
987 char *last_slash = strrchr (path, '/');
991 *file = xstrdup (path);
995 *dir = strdupdelim (path, last_slash);
996 *file = xstrdup (last_slash + 1);
999 url_unescape (*file);
1002 /* Note: URL's "full path" is the path with the query string and
1003 params appended. The "fragment" (#foo) is intentionally ignored,
1004 but that might be changed. For example, if the original URL was
1005 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
1006 the full path will be "/foo/bar/baz;bullshit?querystring". */
1008 /* Return the length of the full path, without the terminating
1012 full_path_length (const struct url *url)
1016 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
1027 /* Write out the full path. */
1030 full_path_write (const struct url *url, char *where)
1032 #define FROB(el, chr) do { \
1033 char *f_el = url->el; \
1035 int l = strlen (f_el); \
1037 memcpy (where, f_el, l); \
1049 /* Public function for getting the "full path". E.g. if u->path is
1050 "foo/bar" and u->query is "param=value", full_path will be
1051 "/foo/bar?param=value". */
1054 url_full_path (const struct url *url)
1056 int length = full_path_length (url);
1057 char *full_path = xmalloc (length + 1);
1059 full_path_write (url, full_path);
1060 full_path[length] = '\0';
1065 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1066 escaping of certain characters, such as "/" and ":". Returns a
1067 count of unescaped chars. */
1070 unescape_single_char (char *str, char chr)
1072 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1073 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1074 char *h = str; /* hare */
1075 char *t = str; /* tortoise */
1076 for (; *h; h++, t++)
1078 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1089 /* Escape unsafe and reserved characters, except for the slash
1093 url_escape_dir (const char *dir)
1095 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1099 unescape_single_char (newdir, '/');
1103 /* Sync u->path and u->url with u->dir and u->file. Called after
1104 u->file or u->dir have been changed, typically by the FTP code. */
1107 sync_path (struct url *u)
1109 char *newpath, *efile, *edir;
1113 /* u->dir and u->file are not escaped. URL-escape them before
1114 reassembling them into u->path. That way, if they contain
1115 separators like '?' or even if u->file contains slashes, the
1116 path will be correctly assembled. (u->file can contain slashes
1117 if the URL specifies it with %2f, or if an FTP server returns
1119 edir = url_escape_dir (u->dir);
1120 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1123 newpath = xstrdup (efile);
1126 int dirlen = strlen (edir);
1127 int filelen = strlen (efile);
1129 /* Copy "DIR/FILE" to newpath. */
1130 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1131 memcpy (p, edir, dirlen);
1134 memcpy (p, efile, filelen);
1143 if (efile != u->file)
1146 /* Regenerate u->url as well. */
1148 u->url = url_string (u, URL_AUTH_SHOW);
1151 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1152 This way we can sync u->path and u->url when they get changed. */
1155 url_set_dir (struct url *url, const char *newdir)
1158 url->dir = xstrdup (newdir);
1163 url_set_file (struct url *url, const char *newfile)
1166 url->file = xstrdup (newfile);
1171 url_free (struct url *url)
1177 xfree_null (url->params);
1178 xfree_null (url->query);
1179 xfree_null (url->fragment);
1180 xfree_null (url->user);
1181 xfree_null (url->passwd);
1189 /* Create all the necessary directories for PATH (a file). Calls
1190 make_directory internally. */
1192 mkalldirs (const char *path)
1199 p = path + strlen (path);
1200 for (; *p != '/' && p != path; p--)
1203 /* Don't create if it's just a file. */
1204 if ((p == path) && (*p != '/'))
1206 t = strdupdelim (path, p);
1208 /* Check whether the directory exists. */
1209 if ((stat (t, &st) == 0))
1211 if (S_ISDIR (st.st_mode))
1218 /* If the dir exists as a file name, remove it first. This
1219 is *only* for Wget to work with buggy old CERN http
1220 servers. Here is the scenario: When Wget tries to
1221 retrieve a directory without a slash, e.g.
1222 http://foo/bar (bar being a directory), CERN server will
1223 not redirect it too http://foo/bar/ -- it will generate a
1224 directory listing containing links to bar/file1,
1225 bar/file2, etc. Wget will lose because it saves this
1226 HTML listing to a file `bar', so it cannot create the
1227 directory. To work around this, if the file of the same
1228 name exists, we just remove it and create the directory
1230 DEBUGP (("Removing %s because of directory danger!\n", t));
1234 res = make_directory (t);
1236 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1241 /* Functions for constructing the file name out of URL components. */
1243 /* A growable string structure, used by url_file_name and friends.
1244 This should perhaps be moved to utils.c.
1246 The idea is to have a convenient and efficient way to construct a
1247 string by having various functions append data to it. Instead of
1248 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1249 functions in questions, we pass the pointer to this struct. */
1257 /* Ensure that the string can accept APPEND_COUNT more characters past
1258 the current TAIL position. If necessary, this will grow the string
1259 and update its allocated size. If the string is already large
1260 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1261 #define GROW(g, append_size) do { \
1262 struct growable *G_ = g; \
1263 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1266 /* Return the tail position of the string. */
1267 #define TAIL(r) ((r)->base + (r)->tail)
1269 /* Move the tail position by APPEND_COUNT characters. */
1270 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1272 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1276 append_string (const char *str, struct growable *dest)
1278 int l = strlen (str);
1280 memcpy (TAIL (dest), str, l);
1281 TAIL_INCR (dest, l);
1284 /* Append CH to DEST. For example, append_char (0, DEST)
1285 zero-terminates DEST. */
1288 append_char (char ch, struct growable *dest)
1292 TAIL_INCR (dest, 1);
1296 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1297 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1298 filechr_control = 4 /* a control character, e.g. 0-31 */
1301 #define FILE_CHAR_TEST(c, mask) \
1302 ((opt.restrict_files_nonascii && !c_isascii ((unsigned char)(c))) || \
1303 (filechr_table[(unsigned char)(c)] & (mask)))
1305 /* Shorthands for the table: */
1306 #define U filechr_not_unix
1307 #define W filechr_not_windows
1308 #define C filechr_control
1313 /* Table of characters unsafe under various conditions (see above).
1315 Arguably we could also claim `%' to be unsafe, since we use it as
1316 the escape character. If we ever want to be able to reliably
1317 translate file name back to URL, this would become important
1318 crucial. Right now, it's better to be minimal in escaping. */
1320 static const unsigned char filechr_table[256] =
1322 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1323 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1324 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1325 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1326 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1327 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1328 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1329 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1330 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1331 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1332 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1333 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1334 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1335 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1336 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1337 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1339 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1340 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1341 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1342 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1344 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1345 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1346 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1347 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1355 /* FN_PORT_SEP is the separator between host and port in file names
1356 for non-standard port numbers. On Unix this is normally ':', as in
1357 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1358 because Windows can't handle ':' in file names. */
1359 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1361 /* FN_QUERY_SEP is the separator between the file name and the URL
1362 query, normally '?'. Since Windows cannot handle '?' as part of
1363 file name, we use '@' instead there. */
1364 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1366 /* Quote path element, characters in [b, e), as file name, and append
1367 the quoted string to DEST. Each character is quoted as per
1368 file_unsafe_char and the corresponding table.
1370 If ESCAPED is true, the path element is considered to be
1371 URL-escaped and will be unescaped prior to inspection. */
1374 append_uri_pathel (const char *b, const char *e, bool escaped,
1375 struct growable *dest)
1381 if (opt.restrict_files_os == restrict_unix)
1382 mask = filechr_not_unix;
1384 mask = filechr_not_windows;
1385 if (opt.restrict_files_ctrl)
1386 mask |= filechr_control;
1388 /* Copy [b, e) to PATHEL and URL-unescape it. */
1392 BOUNDED_TO_ALLOCA (b, e, unescaped);
1393 url_unescape (unescaped);
1395 e = unescaped + strlen (unescaped);
1398 /* Defang ".." when found as component of path. Remember that path
1399 comes from the URL and might contain malicious input. */
1400 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1406 /* Walk the PATHEL string and check how many characters we'll need
1409 for (p = b; p < e; p++)
1410 if (FILE_CHAR_TEST (*p, mask))
1413 /* Calculate the length of the output string. e-b is the input
1414 string length. Each quoted char introduces two additional
1415 characters in the string, hence 2*quoted. */
1416 outlen = (e - b) + (2 * quoted);
1417 GROW (dest, outlen);
1421 /* If there's nothing to quote, we can simply append the string
1422 without processing it again. */
1423 memcpy (TAIL (dest), b, outlen);
1427 char *q = TAIL (dest);
1428 for (p = b; p < e; p++)
1430 if (!FILE_CHAR_TEST (*p, mask))
1434 unsigned char ch = *p;
1436 *q++ = XNUM_TO_DIGIT (ch >> 4);
1437 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1440 assert (q - TAIL (dest) == outlen);
1443 /* Perform inline case transformation if required. */
1444 if (opt.restrict_files_case == restrict_lowercase
1445 || opt.restrict_files_case == restrict_uppercase)
1448 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1450 if (opt.restrict_files_case == restrict_lowercase)
1451 *q = c_tolower (*q);
1453 *q = c_toupper (*q);
1457 TAIL_INCR (dest, outlen);
1460 /* Append to DEST the directory structure that corresponds the
1461 directory part of URL's path. For example, if the URL is
1462 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1464 Each path element ("dir1" and "dir2" in the above example) is
1465 examined, url-unescaped, and re-escaped as file name element.
1467 Additionally, it cuts as many directories from the path as
1468 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1469 will produce "bar" for the above example. For 2 or more, it will
1472 Each component of the path is quoted for use as file name. */
1475 append_dir_structure (const struct url *u, struct growable *dest)
1477 char *pathel, *next;
1478 int cut = opt.cut_dirs;
1480 /* Go through the path components, de-URL-quote them, and quote them
1481 (if necessary) as file names. */
1484 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1489 /* Ignore empty pathels. */
1493 append_char ('/', dest);
1494 append_uri_pathel (pathel, next, true, dest);
1498 /* Return a unique file name that matches the given URL as good as
1499 possible. Does not create directories on the file system. */
1502 url_file_name (const struct url *u)
1504 struct growable fnres; /* stands for "file name result" */
1506 const char *u_file, *u_query;
1507 char *fname, *unique;
1508 char *index_filename = "index.html"; /* The default index file is index.html */
1514 /* If an alternative index file was defined, change index_filename */
1515 if (opt.default_page)
1516 index_filename = opt.default_page;
1519 /* Start with the directory prefix, if specified. */
1521 append_string (opt.dir_prefix, &fnres);
1523 /* If "dirstruct" is turned on (typically the case with -r), add
1524 the host and port (unless those have been turned off) and
1525 directory structure. */
1528 if (opt.protocol_directories)
1531 append_char ('/', &fnres);
1532 append_string (supported_schemes[u->scheme].name, &fnres);
1534 if (opt.add_hostdir)
1537 append_char ('/', &fnres);
1538 if (0 != strcmp (u->host, ".."))
1539 append_string (u->host, &fnres);
1541 /* Host name can come from the network; malicious DNS may
1542 allow ".." to be resolved, causing us to write to
1543 "../<file>". Defang such host names. */
1544 append_string ("%2E%2E", &fnres);
1545 if (u->port != scheme_default_port (u->scheme))
1548 number_to_string (portstr, u->port);
1549 append_char (FN_PORT_SEP, &fnres);
1550 append_string (portstr, &fnres);
1554 append_dir_structure (u, &fnres);
1557 /* Add the file name. */
1559 append_char ('/', &fnres);
1560 u_file = *u->file ? u->file : index_filename;
1561 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1563 /* Append "?query" to the file name. */
1564 u_query = u->query && *u->query ? u->query : NULL;
1567 append_char (FN_QUERY_SEP, &fnres);
1568 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1571 /* Zero-terminate the file name. */
1572 append_char ('\0', &fnres);
1576 /* Check the cases in which the unique extensions are not used:
1577 1) Clobbering is turned off (-nc).
1578 2) Retrieval with regetting.
1579 3) Timestamping is used.
1580 4) Hierarchy is built.
1582 The exception is the case when file does exist and is a
1583 directory (see `mkalldirs' for explanation). */
1585 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1586 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1592 unique = unique_name (fname, true);
1593 if (unique != fname)
1597 /* On VMS, alter the name as required. */
1602 unique2 = ods_conform( unique);
1603 if (unique2 != unique)
1609 #endif /* def __VMS */
1614 /* Resolve "." and ".." elements of PATH by destructively modifying
1615 PATH and return true if PATH has been modified, false otherwise.
1617 The algorithm is in spirit similar to the one described in rfc1808,
1618 although implemented differently, in one pass. To recap, path
1619 elements containing only "." are removed, and ".." is taken to mean
1620 "back up one element". Single leading and trailing slashes are
1623 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1624 test examples are provided below. If you change anything in this
1625 function, run test_path_simplify to make sure you haven't broken a
1629 path_simplify (enum url_scheme scheme, char *path)
1631 char *h = path; /* hare */
1632 char *t = path; /* tortoise */
1634 char *end = strchr (path, '\0');
1638 /* Hare should be at the beginning of a path element. */
1640 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1645 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1647 /* Handle "../" by retreating the tortoise by one path
1648 element -- but not past beggining. */
1651 /* Move backwards until T hits the beginning of the
1652 previous path element or the beginning of path. */
1653 for (--t; t > beg && t[-1] != '/'; t--)
1656 else if (scheme == SCHEME_FTP)
1658 /* If we're at the beginning, copy the "../" literally
1659 and move the beginning so a later ".." doesn't remove
1660 it. This violates RFC 3986; but we do it for FTP
1661 anyway because there is otherwise no way to get at a
1662 parent directory, when the FTP server drops us in a
1663 non-root directory (which is not uncommon). */
1672 /* A regular path element. If H hasn't advanced past T,
1673 simply skip to the next path element. Otherwise, copy
1674 the path element until the next slash. */
1677 /* Skip the path element, including the slash. */
1678 while (h < end && *h != '/')
1685 /* Copy the path element, including the final slash. */
1686 while (h < end && *h != '/')
1700 /* Return the length of URL's path. Path is considered to be
1701 terminated by one or more of the ?query or ;params or #fragment,
1702 depending on the scheme. */
1705 path_end (const char *url)
1707 enum url_scheme scheme = url_scheme (url);
1709 if (scheme == SCHEME_INVALID)
1710 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1711 /* +2 to ignore the first two separators ':' and '/' */
1712 seps = init_seps (scheme) + 2;
1713 return strpbrk_or_eos (url, seps);
1716 /* Find the last occurrence of character C in the range [b, e), or
1717 NULL, if none are present. */
1718 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1720 /* Merge BASE with LINK and return the resulting URI.
1722 Either of the URIs may be absolute or relative, complete with the
1723 host name, or path only. This tries to reasonably handle all
1724 foreseeable cases. It only employs minimal URL parsing, without
1725 knowledge of the specifics of schemes.
1727 I briefly considered making this function call path_simplify after
1728 the merging process, as rfc1738 seems to suggest. This is a bad
1729 idea for several reasons: 1) it complexifies the code, and 2)
1730 url_parse has to simplify path anyway, so it's wasteful to boot. */
1733 uri_merge (const char *base, const char *link)
1739 if (url_has_scheme (link))
1740 return xstrdup (link);
1742 /* We may not examine BASE past END. */
1743 end = path_end (base);
1744 linklength = strlen (link);
1748 /* Empty LINK points back to BASE, query string and all. */
1749 return xstrdup (base);
1751 else if (*link == '?')
1753 /* LINK points to the same location, but changes the query
1754 string. Examples: */
1755 /* uri_merge("path", "?new") -> "path?new" */
1756 /* uri_merge("path?foo", "?new") -> "path?new" */
1757 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1758 /* uri_merge("path#foo", "?new") -> "path?new" */
1759 int baselength = end - base;
1760 merge = xmalloc (baselength + linklength + 1);
1761 memcpy (merge, base, baselength);
1762 memcpy (merge + baselength, link, linklength);
1763 merge[baselength + linklength] = '\0';
1765 else if (*link == '#')
1767 /* uri_merge("path", "#new") -> "path#new" */
1768 /* uri_merge("path#foo", "#new") -> "path#new" */
1769 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1770 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1772 const char *end1 = strchr (base, '#');
1774 end1 = base + strlen (base);
1775 baselength = end1 - base;
1776 merge = xmalloc (baselength + linklength + 1);
1777 memcpy (merge, base, baselength);
1778 memcpy (merge + baselength, link, linklength);
1779 merge[baselength + linklength] = '\0';
1781 else if (*link == '/' && *(link + 1) == '/')
1783 /* LINK begins with "//" and so is a net path: we need to
1784 replace everything after (and including) the double slash
1787 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1788 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1789 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1793 const char *start_insert;
1795 /* Look for first slash. */
1796 slash = memchr (base, '/', end - base);
1797 /* If found slash and it is a double slash, then replace
1798 from this point, else default to replacing from the
1800 if (slash && *(slash + 1) == '/')
1801 start_insert = slash;
1803 start_insert = base;
1805 span = start_insert - base;
1806 merge = xmalloc (span + linklength + 1);
1808 memcpy (merge, base, span);
1809 memcpy (merge + span, link, linklength);
1810 merge[span + linklength] = '\0';
1812 else if (*link == '/')
1814 /* LINK is an absolute path: we need to replace everything
1815 after (and including) the FIRST slash with LINK.
1817 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1818 "/qux/xyzzy", our result should be
1819 "http://host/qux/xyzzy". */
1822 const char *start_insert = NULL; /* for gcc to shut up. */
1823 const char *pos = base;
1824 bool seen_slash_slash = false;
1825 /* We're looking for the first slash, but want to ignore
1828 slash = memchr (pos, '/', end - pos);
1829 if (slash && !seen_slash_slash)
1830 if (*(slash + 1) == '/')
1833 seen_slash_slash = true;
1837 /* At this point, SLASH is the location of the first / after
1838 "//", or the first slash altogether. START_INSERT is the
1839 pointer to the location where LINK will be inserted. When
1840 examining the last two examples, keep in mind that LINK
1843 if (!slash && !seen_slash_slash)
1844 /* example: "foo" */
1846 start_insert = base;
1847 else if (!slash && seen_slash_slash)
1848 /* example: "http://foo" */
1851 else if (slash && !seen_slash_slash)
1852 /* example: "foo/bar" */
1854 start_insert = base;
1855 else if (slash && seen_slash_slash)
1856 /* example: "http://something/" */
1858 start_insert = slash;
1860 span = start_insert - base;
1861 merge = xmalloc (span + linklength + 1);
1863 memcpy (merge, base, span);
1864 memcpy (merge + span, link, linklength);
1865 merge[span + linklength] = '\0';
1869 /* LINK is a relative URL: we need to replace everything
1870 after last slash (possibly empty) with LINK.
1872 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1873 our result should be "whatever/foo/qux/xyzzy". */
1874 bool need_explicit_slash = false;
1876 const char *start_insert;
1877 const char *last_slash = find_last_char (base, end, '/');
1880 /* No slash found at all. Replace what we have with LINK. */
1881 start_insert = base;
1883 else if (last_slash && last_slash >= base + 2
1884 && last_slash[-2] == ':' && last_slash[-1] == '/')
1886 /* example: http://host" */
1888 start_insert = end + 1;
1889 need_explicit_slash = true;
1893 /* example: "whatever/foo/bar" */
1895 start_insert = last_slash + 1;
1898 span = start_insert - base;
1899 merge = xmalloc (span + linklength + 1);
1901 memcpy (merge, base, span);
1902 if (need_explicit_slash)
1903 merge[span - 1] = '/';
1904 memcpy (merge + span, link, linklength);
1905 merge[span + linklength] = '\0';
1911 #define APPEND(p, s) do { \
1912 int len = strlen (s); \
1913 memcpy (p, s, len); \
1917 /* Use this instead of password when the actual password is supposed
1918 to be hidden. We intentionally use a generic string without giving
1919 away the number of characters in the password, like previous
1921 #define HIDDEN_PASSWORD "*password*"
1923 /* Recreate the URL string from the data in URL.
1925 If HIDE is true (as it is when we're calling this on a URL we plan
1926 to print, but not when calling it to canonicalize a URL for use
1927 within the program), password will be hidden. Unsafe characters in
1928 the URL will be quoted. */
1931 url_string (const struct url *url, enum url_auth_mode auth_mode)
1935 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1937 int scheme_port = supported_schemes[url->scheme].default_port;
1938 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1939 int fplen = full_path_length (url);
1941 bool brackets_around_host;
1943 assert (scheme_str != NULL);
1945 /* Make sure the user name and password are quoted. */
1948 if (auth_mode != URL_AUTH_HIDE)
1950 quoted_user = url_escape_allow_passthrough (url->user);
1953 if (auth_mode == URL_AUTH_HIDE_PASSWD)
1954 quoted_passwd = HIDDEN_PASSWORD;
1956 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1961 /* In the unlikely event that the host name contains non-printable
1962 characters, quote it for displaying to the user. */
1963 quoted_host = url_escape_allow_passthrough (url->host);
1965 /* Undo the quoting of colons that URL escaping performs. IPv6
1966 addresses may legally contain colons, and in that case must be
1967 placed in square brackets. */
1968 if (quoted_host != url->host)
1969 unescape_single_char (quoted_host, ':');
1970 brackets_around_host = strchr (quoted_host, ':') != NULL;
1972 size = (strlen (scheme_str)
1973 + strlen (quoted_host)
1974 + (brackets_around_host ? 2 : 0)
1977 if (url->port != scheme_port)
1978 size += 1 + numdigit (url->port);
1981 size += 1 + strlen (quoted_user);
1983 size += 1 + strlen (quoted_passwd);
1986 p = result = xmalloc (size);
1988 APPEND (p, scheme_str);
1991 APPEND (p, quoted_user);
1995 APPEND (p, quoted_passwd);
2000 if (brackets_around_host)
2002 APPEND (p, quoted_host);
2003 if (brackets_around_host)
2005 if (url->port != scheme_port)
2008 p = number_to_string (p, url->port);
2011 full_path_write (url, p);
2015 assert (p - result == size);
2017 if (quoted_user && quoted_user != url->user)
2018 xfree (quoted_user);
2019 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
2020 && quoted_passwd != url->passwd)
2021 xfree (quoted_passwd);
2022 if (quoted_host != url->host)
2023 xfree (quoted_host);
2028 /* Return true if scheme a is similar to scheme b.
2030 Schemes are similar if they are equal. If SSL is supported, schemes
2031 are also similar if one is http (SCHEME_HTTP) and the other is https
2034 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
2039 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
2040 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
2047 getchar_from_escaped_string (const char *str, char *c)
2049 const char *p = str;
2051 assert (str && *str);
2056 if (!c_isxdigit(p[1]) || !c_isxdigit(p[2]))
2064 return 0; /* error: invalid string */
2066 *c = X2DIGITS_TO_NUM (p[1], p[2]);
2067 if (URL_RESERVED_CHAR(*c))
2085 are_urls_equal (const char *u1, const char *u2)
2096 && (pp = getchar_from_escaped_string (p, &ch1))
2097 && (qq = getchar_from_escaped_string (q, &ch2))
2098 && (c_tolower(ch1) == c_tolower(ch2)))
2104 return (*p == 0 && *q == 0 ? true : false);
2108 /* Debugging and testing support for path_simplify. */
2111 /* Debug: run path_simplify on PATH and return the result in a new
2112 string. Useful for calling from the debugger. */
2116 char *copy = xstrdup (path);
2117 path_simplify (copy);
2123 run_test (char *test, char *expected_result, enum url_scheme scheme,
2124 bool expected_change)
2126 char *test_copy = xstrdup (test);
2127 bool modified = path_simplify (scheme, test_copy);
2129 if (0 != strcmp (test_copy, expected_result))
2131 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2132 test, expected_result, test_copy);
2135 if (modified != expected_change)
2137 if (expected_change)
2138 printf ("Expected modification with path_simplify(\"%s\").\n",
2141 printf ("Expected no modification with path_simplify(\"%s\").\n",
2145 mu_assert ("", modified == expected_change);
2150 test_path_simplify (void)
2153 char *test, *result;
2154 enum url_scheme scheme;
2157 { "", "", SCHEME_HTTP, false },
2158 { ".", "", SCHEME_HTTP, true },
2159 { "./", "", SCHEME_HTTP, true },
2160 { "..", "", SCHEME_HTTP, true },
2161 { "../", "", SCHEME_HTTP, true },
2162 { "..", "..", SCHEME_FTP, false },
2163 { "../", "../", SCHEME_FTP, false },
2164 { "foo", "foo", SCHEME_HTTP, false },
2165 { "foo/bar", "foo/bar", SCHEME_HTTP, false },
2166 { "foo///bar", "foo///bar", SCHEME_HTTP, false },
2167 { "foo/.", "foo/", SCHEME_HTTP, true },
2168 { "foo/./", "foo/", SCHEME_HTTP, true },
2169 { "foo./", "foo./", SCHEME_HTTP, false },
2170 { "foo/../bar", "bar", SCHEME_HTTP, true },
2171 { "foo/../bar/", "bar/", SCHEME_HTTP, true },
2172 { "foo/bar/..", "foo/", SCHEME_HTTP, true },
2173 { "foo/bar/../x", "foo/x", SCHEME_HTTP, true },
2174 { "foo/bar/../x/", "foo/x/", SCHEME_HTTP, true },
2175 { "foo/..", "", SCHEME_HTTP, true },
2176 { "foo/../..", "", SCHEME_HTTP, true },
2177 { "foo/../../..", "", SCHEME_HTTP, true },
2178 { "foo/../../bar/../../baz", "baz", SCHEME_HTTP, true },
2179 { "foo/../..", "..", SCHEME_FTP, true },
2180 { "foo/../../..", "../..", SCHEME_FTP, true },
2181 { "foo/../../bar/../../baz", "../../baz", SCHEME_FTP, true },
2182 { "a/b/../../c", "c", SCHEME_HTTP, true },
2183 { "./a/../b", "b", SCHEME_HTTP, true }
2187 for (i = 0; i < countof (tests); i++)
2189 const char *message;
2190 char *test = tests[i].test;
2191 char *expected_result = tests[i].result;
2192 enum url_scheme scheme = tests[i].scheme;
2193 bool expected_change = tests[i].should_modify;
2194 message = run_test (test, expected_result, scheme, expected_change);
2195 if (message) return message;
2201 test_append_uri_pathel()
2208 char *expected_result;
2210 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2213 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2215 struct growable dest;
2216 const char *p = test_array[i].input;
2218 memset (&dest, 0, sizeof (dest));
2220 append_string (test_array[i].original_url, &dest);
2221 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2222 append_char ('\0', &dest);
2224 mu_assert ("test_append_uri_pathel: wrong result",
2225 strcmp (dest.base, test_array[i].expected_result) == 0);
2232 test_are_urls_equal()
2238 bool expected_result;
2240 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2241 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2242 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2243 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2244 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2245 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2248 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2250 mu_assert ("test_are_urls_equal: wrong result",
2251 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2257 #endif /* TESTING */