2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
3 2004, 2005, 2006, 2007 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. */
45 #include "host.h" /* for is_valid_ipv6_address */
52 scm_disabled = 1, /* for https when OpenSSL fails to init. */
53 scm_has_params = 2, /* whether scheme has ;params */
54 scm_has_query = 4, /* whether scheme has ?query */
55 scm_has_fragment = 8 /* whether scheme has #fragment */
60 /* Short name of the scheme, such as "http" or "ftp". */
62 /* Leading string that identifies the scheme, such as "https://". */
63 const char *leading_string;
64 /* Default port of the scheme when none is specified. */
70 /* Supported schemes: */
71 static struct scheme_data supported_schemes[] =
73 { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment },
75 { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment },
77 { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment },
83 /* Forward declarations: */
85 static bool path_simplify (char *);
87 /* Support for escaping and unescaping of URL strings. */
89 /* Table of "reserved" and "unsafe" characters. Those terms are
90 rfc1738-speak, as such largely obsoleted by rfc2396 and later
91 specs, but the general idea remains.
93 A reserved character is the one that you can't decode without
94 changing the meaning of the URL. For example, you can't decode
95 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
96 path components is different. Non-reserved characters can be
97 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
98 unsafe characters are loosely based on rfc1738, plus "$" and ",",
99 as recommended by rfc2396, and minus "~", which is very frequently
100 used (and sometimes unrecognized as %7E by broken servers).
102 An unsafe character is the one that should be encoded when URLs are
103 placed in foreign environments. E.g. space and newline are unsafe
104 in HTTP contexts because HTTP uses them as separator and line
105 terminator, so they must be encoded to %20 and %0A respectively.
106 "*" is unsafe in shell context, etc.
108 We determine whether a character is unsafe through static table
109 lookup. This code assumes ASCII character set and 8-bit chars. */
112 /* rfc1738 reserved chars + "$" and ",". */
115 /* rfc1738 unsafe chars, plus non-printables. */
119 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
120 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
121 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
123 /* Shorthands for the table: */
124 #define R urlchr_reserved
125 #define U urlchr_unsafe
128 static const unsigned char urlchr_table[256] =
130 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
131 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
132 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
133 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
134 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
135 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
136 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
137 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
138 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
139 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
140 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
141 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
142 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
143 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
144 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
145 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
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,
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,
154 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,
161 /* URL-unescape the string S.
163 This is done by transforming the sequences "%HH" to the character
164 represented by the hexadecimal digits HH. If % is not followed by
165 two hexadecimal digits, it is inserted literally.
167 The transformation is done in place. If you need the original
168 string intact, make a copy before calling this function. */
171 url_unescape (char *s)
173 char *t = s; /* t - tortoise */
174 char *h = s; /* h - hare */
186 /* Do nothing if '%' is not followed by two hex digits. */
187 if (!h[1] || !h[2] || !(ISXDIGIT (h[1]) && ISXDIGIT (h[2])))
189 c = X2DIGITS_TO_NUM (h[1], h[2]);
190 /* Don't unescape %00 because there is no way to insert it
191 into a C string without effectively truncating it. */
201 /* The core of url_escape_* functions. Escapes the characters that
202 match the provided mask in urlchr_table.
204 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
205 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
206 allocated string will be returned in all cases. */
209 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
216 for (p1 = s; *p1; p1++)
217 if (urlchr_test (*p1, mask))
218 addition += 2; /* Two more characters (hex digits) */
221 return allow_passthrough ? (char *)s : xstrdup (s);
223 newlen = (p1 - s) + addition;
224 newstr = xmalloc (newlen + 1);
230 /* Quote the characters that match the test mask. */
231 if (urlchr_test (*p1, mask))
233 unsigned char c = *p1++;
235 *p2++ = XNUM_TO_DIGIT (c >> 4);
236 *p2++ = XNUM_TO_DIGIT (c & 0xf);
241 assert (p2 - newstr == newlen);
247 /* URL-escape the unsafe characters (see urlchr_table) in a given
248 string, returning a freshly allocated string. */
251 url_escape (const char *s)
253 return url_escape_1 (s, urlchr_unsafe, false);
256 /* URL-escape the unsafe characters (see urlchr_table) in a given
257 string. If no characters are unsafe, S is returned. */
260 url_escape_allow_passthrough (const char *s)
262 return url_escape_1 (s, urlchr_unsafe, true);
265 /* Decide whether the char at position P needs to be encoded. (It is
266 not enough to pass a single char *P because the function may need
267 to inspect the surrounding context.)
269 Return true if the char should be escaped as %XX, false otherwise. */
272 char_needs_escaping (const char *p)
276 if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2)))
279 /* Garbled %.. sequence: encode `%'. */
282 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
288 /* Translate a %-escaped (but possibly non-conformant) input string S
289 into a %-escaped (and conformant) output string. If no characters
290 are encoded or decoded, return the same string S; otherwise, return
291 a freshly allocated string with the new contents.
293 After a URL has been run through this function, the protocols that
294 use `%' as the quote character can use the resulting string as-is,
295 while those that don't can use url_unescape to get to the intended
296 data. This function is stable: once the input is transformed,
297 further transformations of the result yield the same output.
299 Let's discuss why this function is needed.
301 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
302 a raw space character would mess up the HTTP request, it needs to
303 be quoted, like this:
305 GET /abc%20def HTTP/1.0
307 It would appear that the unsafe chars need to be quoted, for
308 example with url_escape. But what if we're requested to download
309 `abc%20def'? url_escape transforms "%" to "%25", which would leave
310 us with `abc%2520def'. This is incorrect -- since %-escapes are
311 part of URL syntax, "%20" is the correct way to denote a literal
312 space on the Wget command line. This leads to the conclusion that
313 in that case Wget should not call url_escape, but leave the `%20'
314 as is. This is clearly contradictory, but it only gets worse.
316 What if the requested URI is `abc%20 def'? If we call url_escape,
317 we end up with `/abc%2520%20def', which is almost certainly not
318 intended. If we don't call url_escape, we are left with the
319 embedded space and cannot complete the request. What the user
320 meant was for Wget to request `/abc%20%20def', and this is where
321 reencode_escapes kicks in.
323 Wget used to solve this by first decoding %-quotes, and then
324 encoding all the "unsafe" characters found in the resulting string.
325 This was wrong because it didn't preserve certain URL special
326 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
327 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
328 whether we considered `+' reserved (it is). One of these results
329 is inevitable because by the second step we would lose information
330 on whether the `+' was originally encoded or not. Both results
331 were wrong because in CGI parameters + means space, while %2B means
332 literal plus. reencode_escapes correctly translates the above to
333 "a%2B+b", i.e. returns the original string.
335 This function uses a modified version of the algorithm originally
336 proposed by Anon Sricharoenchai:
338 * Encode all "unsafe" characters, except those that are also
339 "reserved", to %XX. See urlchr_table for which characters are
342 * Encode the "%" characters not followed by two hex digits to
345 * Pass through all other characters and %XX escapes as-is. (Up to
346 Wget 1.10 this decoded %XX escapes corresponding to "safe"
347 characters, but that was obtrusive and broke some servers.)
351 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
353 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
357 "foo bar" -> "foo%20bar"
358 "foo%20bar" -> "foo%20bar"
359 "foo %20bar" -> "foo%20%20bar"
360 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
361 "foo%25%20bar" -> "foo%25%20bar"
362 "foo%2%20bar" -> "foo%252%20bar"
363 "foo+bar" -> "foo+bar" (plus is reserved!)
364 "foo%2b+bar" -> "foo%2b+bar" */
367 reencode_escapes (const char *s)
373 int encode_count = 0;
375 /* First pass: inspect the string to see if there's anything to do,
376 and to calculate the new length. */
377 for (p1 = s; *p1; p1++)
378 if (char_needs_escaping (p1))
382 /* The string is good as it is. */
383 return (char *) s; /* C const model sucks. */
386 /* Each encoding adds two characters (hex digits). */
387 newlen = oldlen + 2 * encode_count;
388 newstr = xmalloc (newlen + 1);
390 /* Second pass: copy the string to the destination address, encoding
391 chars when needed. */
396 if (char_needs_escaping (p1))
398 unsigned char c = *p1++;
400 *p2++ = XNUM_TO_DIGIT (c >> 4);
401 *p2++ = XNUM_TO_DIGIT (c & 0xf);
407 assert (p2 - newstr == newlen);
411 /* Returns the scheme type if the scheme is supported, or
412 SCHEME_INVALID if not. */
415 url_scheme (const char *url)
419 for (i = 0; supported_schemes[i].leading_string; i++)
420 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
421 strlen (supported_schemes[i].leading_string)))
423 if (!(supported_schemes[i].flags & scm_disabled))
424 return (enum url_scheme) i;
426 return SCHEME_INVALID;
429 return SCHEME_INVALID;
432 #define SCHEME_CHAR(ch) (ISALNUM (ch) || (ch) == '-' || (ch) == '+')
434 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
435 currently implemented, it returns true if URL begins with
439 url_has_scheme (const char *url)
443 /* The first char must be a scheme char. */
444 if (!*p || !SCHEME_CHAR (*p))
447 /* Followed by 0 or more scheme chars. */
448 while (*p && SCHEME_CHAR (*p))
450 /* Terminated by ':'. */
455 scheme_default_port (enum url_scheme scheme)
457 return supported_schemes[scheme].default_port;
461 scheme_disable (enum url_scheme scheme)
463 supported_schemes[scheme].flags |= scm_disabled;
466 /* Skip the username and password, if present in the URL. The
467 function should *not* be called with the complete URL, but with the
468 portion after the scheme.
470 If no username and password are found, return URL. */
473 url_skip_credentials (const char *url)
475 /* Look for '@' that comes before terminators, such as '/', '?',
477 const char *p = (const char *)strpbrk (url, "@/?#;");
483 /* Parse credentials contained in [BEG, END). The region is expected
484 to have come from a URL and is unescaped. */
487 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
493 return false; /* empty user name */
495 colon = memchr (beg, ':', end - beg);
497 return false; /* again empty user name */
501 *passwd = strdupdelim (colon + 1, end);
503 url_unescape (*passwd);
510 *user = strdupdelim (beg, userend);
511 url_unescape (*user);
515 /* Used by main.c: detect URLs written using the "shorthand" URL forms
516 originally popularized by Netscape and NcFTP. HTTP shorthands look
519 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
520 www.foo.com[:port] -> http://www.foo.com[:port]
522 FTP shorthands look like this:
524 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
525 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
527 If the URL needs not or cannot be rewritten, return NULL. */
530 rewrite_shorthand_url (const char *url)
535 if (url_scheme (url) != SCHEME_INVALID)
538 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
540 p = strpbrk (url, ":/");
544 /* If we're looking at "://", it means the URL uses a scheme we
545 don't support, which may include "https" when compiled without
546 SSL support. Don't bogusly rewrite such URLs. */
547 if (p && p[0] == ':' && p[1] == '/' && p[2] == '/')
552 /* Colon indicates ftp, as in foo.bar.com:path. Check for
553 special case of http port number ("localhost:10000"). */
554 int digits = strspn (p + 1, "0123456789");
555 if (digits && (p[1 + digits] == '/' || p[1 + digits] == '\0'))
558 /* Turn "foo.bar.com:path" to "ftp://foo.bar.com/path". */
559 ret = aprintf ("ftp://%s", url);
560 ret[6 + (p - url)] = '/';
565 /* Just prepend "http://" to URL. */
566 ret = aprintf ("http://%s", url);
571 static void split_path (const char *, char **, char **);
573 /* Like strpbrk, with the exception that it returns the pointer to the
574 terminating zero (end-of-string aka "eos") if no matching character
578 strpbrk_or_eos (const char *s, const char *accept)
580 char *p = strpbrk (s, accept);
582 p = strchr (s, '\0');
586 /* Turn STR into lowercase; return true if a character was actually
590 lowercase_str (char *str)
592 bool changed = false;
597 *str = TOLOWER (*str);
603 init_seps (enum url_scheme scheme)
605 static char seps[8] = ":/";
607 int flags = supported_schemes[scheme].flags;
609 if (flags & scm_has_params)
611 if (flags & scm_has_query)
613 if (flags & scm_has_fragment)
619 static const char *parse_errors[] = {
620 #define PE_NO_ERROR 0
622 #define PE_UNSUPPORTED_SCHEME 1
623 N_("Unsupported scheme"),
624 #define PE_INVALID_HOST_NAME 2
625 N_("Invalid host name"),
626 #define PE_BAD_PORT_NUMBER 3
627 N_("Bad port number"),
628 #define PE_INVALID_USER_NAME 4
629 N_("Invalid user name"),
630 #define PE_UNTERMINATED_IPV6_ADDRESS 5
631 N_("Unterminated IPv6 numeric address"),
632 #define PE_IPV6_NOT_SUPPORTED 6
633 N_("IPv6 addresses not supported"),
634 #define PE_INVALID_IPV6_ADDRESS 7
635 N_("Invalid IPv6 numeric address")
640 Return a new struct url if successful, NULL on error. In case of
641 error, and if ERROR is not NULL, also set *ERROR to the appropriate
644 url_parse (const char *url, int *error)
648 bool path_modified, host_modified;
650 enum url_scheme scheme;
653 const char *uname_b, *uname_e;
654 const char *host_b, *host_e;
655 const char *path_b, *path_e;
656 const char *params_b, *params_e;
657 const char *query_b, *query_e;
658 const char *fragment_b, *fragment_e;
661 char *user = NULL, *passwd = NULL;
663 char *url_encoded = NULL;
667 scheme = url_scheme (url);
668 if (scheme == SCHEME_INVALID)
670 error_code = PE_UNSUPPORTED_SCHEME;
674 url_encoded = reencode_escapes (url);
677 p += strlen (supported_schemes[scheme].leading_string);
679 p = url_skip_credentials (p);
682 /* scheme://user:pass@host[:port]... */
685 /* We attempt to break down the URL into the components path,
686 params, query, and fragment. They are ordered like this:
688 scheme://host[:port][/path][;params][?query][#fragment] */
690 path_b = path_e = NULL;
691 params_b = params_e = NULL;
692 query_b = query_e = NULL;
693 fragment_b = fragment_e = NULL;
695 /* Initialize separators for optional parts of URL, depending on the
696 scheme. For example, FTP has params, and HTTP and HTTPS have
697 query string and fragment. */
698 seps = init_seps (scheme);
704 /* Handle IPv6 address inside square brackets. Ideally we'd
705 just look for the terminating ']', but rfc2732 mandates
706 rejecting invalid IPv6 addresses. */
708 /* The address begins after '['. */
710 host_e = strchr (host_b, ']');
714 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
719 /* Check if the IPv6 address is valid. */
720 if (!is_valid_ipv6_address(host_b, host_e))
722 error_code = PE_INVALID_IPV6_ADDRESS;
726 /* Continue parsing after the closing ']'. */
729 error_code = PE_IPV6_NOT_SUPPORTED;
733 /* The closing bracket must be followed by a separator or by the
735 /* http://[::1]... */
737 if (!strchr (seps, *p))
739 /* Trailing garbage after []-delimited IPv6 address. */
740 error_code = PE_INVALID_HOST_NAME;
746 p = strpbrk_or_eos (p, seps);
749 ++seps; /* advance to '/' */
751 if (host_b == host_e)
753 error_code = PE_INVALID_HOST_NAME;
757 port = scheme_default_port (scheme);
760 const char *port_b, *port_e, *pp;
762 /* scheme://host:port/tralala */
766 p = strpbrk_or_eos (p, seps);
769 /* Allow empty port, as per rfc2396. */
770 if (port_b != port_e)
771 for (port = 0, pp = port_b; pp < port_e; pp++)
775 /* http://host:12randomgarbage/blah */
777 error_code = PE_BAD_PORT_NUMBER;
780 port = 10 * port + (*pp - '0');
781 /* Check for too large port numbers here, before we have
782 a chance to overflow on bogus port values. */
785 error_code = PE_BAD_PORT_NUMBER;
790 /* Advance to the first separator *after* '/' (either ';' or '?',
791 depending on the scheme). */
794 /* Get the optional parts of URL, each part being delimited by
795 current location and the position of the next separator. */
796 #define GET_URL_PART(sepchar, var) do { \
798 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
802 GET_URL_PART ('/', path);
803 if (supported_schemes[scheme].flags & scm_has_params)
804 GET_URL_PART (';', params);
805 if (supported_schemes[scheme].flags & scm_has_query)
806 GET_URL_PART ('?', query);
807 if (supported_schemes[scheme].flags & scm_has_fragment)
808 GET_URL_PART ('#', fragment);
813 if (uname_b != uname_e)
815 /* http://user:pass@host */
817 /* uname_b uname_e */
818 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
820 error_code = PE_INVALID_USER_NAME;
825 u = xnew0 (struct url);
827 u->host = strdupdelim (host_b, host_e);
832 u->path = strdupdelim (path_b, path_e);
833 path_modified = path_simplify (u->path);
834 split_path (u->path, &u->dir, &u->file);
836 host_modified = lowercase_str (u->host);
838 /* Decode %HH sequences in host name. This is important not so much
839 to support %HH sequences in host names (which other browser
840 don't), but to support binary characters (which will have been
841 converted to %HH by reencode_escapes). */
842 if (strchr (u->host, '%'))
844 url_unescape (u->host);
845 host_modified = true;
849 u->params = strdupdelim (params_b, params_e);
851 u->query = strdupdelim (query_b, query_e);
853 u->fragment = strdupdelim (fragment_b, fragment_e);
855 if (path_modified || u->fragment || host_modified || path_b == path_e)
857 /* If we suspect that a transformation has rendered what
858 url_string might return different from URL_ENCODED, rebuild
859 u->url using url_string. */
860 u->url = url_string (u, URL_AUTH_SHOW);
862 if (url_encoded != url)
863 xfree ((char *) url_encoded);
867 if (url_encoded == url)
868 u->url = xstrdup (url);
870 u->url = url_encoded;
876 /* Cleanup in case of error: */
877 if (url_encoded && url_encoded != url)
880 /* Transmit the error code to the caller, if the caller wants to
887 /* Return the error message string from ERROR_CODE, which should have
888 been retrieved from url_parse. The error message is translated. */
891 url_error (int error_code)
893 assert (error_code >= 0 && error_code < countof (parse_errors));
894 return _(parse_errors[error_code]);
897 /* Split PATH into DIR and FILE. PATH comes from the URL and is
898 expected to be URL-escaped.
900 The path is split into directory (the part up to the last slash)
901 and file (the part after the last slash), which are subsequently
905 "foo/bar/baz" "foo/bar" "baz"
906 "foo/bar/" "foo/bar" ""
908 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
910 DIR and FILE are freshly allocated. */
913 split_path (const char *path, char **dir, char **file)
915 char *last_slash = strrchr (path, '/');
919 *file = xstrdup (path);
923 *dir = strdupdelim (path, last_slash);
924 *file = xstrdup (last_slash + 1);
927 url_unescape (*file);
930 /* Note: URL's "full path" is the path with the query string and
931 params appended. The "fragment" (#foo) is intentionally ignored,
932 but that might be changed. For example, if the original URL was
933 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
934 the full path will be "/foo/bar/baz;bullshit?querystring". */
936 /* Return the length of the full path, without the terminating
940 full_path_length (const struct url *url)
944 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
955 /* Write out the full path. */
958 full_path_write (const struct url *url, char *where)
960 #define FROB(el, chr) do { \
961 char *f_el = url->el; \
963 int l = strlen (f_el); \
965 memcpy (where, f_el, l); \
977 /* Public function for getting the "full path". E.g. if u->path is
978 "foo/bar" and u->query is "param=value", full_path will be
979 "/foo/bar?param=value". */
982 url_full_path (const struct url *url)
984 int length = full_path_length (url);
985 char *full_path = xmalloc (length + 1);
987 full_path_write (url, full_path);
988 full_path[length] = '\0';
993 /* Unescape CHR in an otherwise escaped STR. Used to selectively
994 escaping of certain characters, such as "/" and ":". Returns a
995 count of unescaped chars. */
998 unescape_single_char (char *str, char chr)
1000 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1001 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1002 char *h = str; /* hare */
1003 char *t = str; /* tortoise */
1004 for (; *h; h++, t++)
1006 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1017 /* Escape unsafe and reserved characters, except for the slash
1021 url_escape_dir (const char *dir)
1023 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1027 unescape_single_char (newdir, '/');
1031 /* Sync u->path and u->url with u->dir and u->file. Called after
1032 u->file or u->dir have been changed, typically by the FTP code. */
1035 sync_path (struct url *u)
1037 char *newpath, *efile, *edir;
1041 /* u->dir and u->file are not escaped. URL-escape them before
1042 reassembling them into u->path. That way, if they contain
1043 separators like '?' or even if u->file contains slashes, the
1044 path will be correctly assembled. (u->file can contain slashes
1045 if the URL specifies it with %2f, or if an FTP server returns
1047 edir = url_escape_dir (u->dir);
1048 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1051 newpath = xstrdup (efile);
1054 int dirlen = strlen (edir);
1055 int filelen = strlen (efile);
1057 /* Copy "DIR/FILE" to newpath. */
1058 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1059 memcpy (p, edir, dirlen);
1062 memcpy (p, efile, filelen);
1071 if (efile != u->file)
1074 /* Regenerate u->url as well. */
1076 u->url = url_string (u, URL_AUTH_SHOW);
1079 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1080 This way we can sync u->path and u->url when they get changed. */
1083 url_set_dir (struct url *url, const char *newdir)
1086 url->dir = xstrdup (newdir);
1091 url_set_file (struct url *url, const char *newfile)
1094 url->file = xstrdup (newfile);
1099 url_free (struct url *url)
1105 xfree_null (url->params);
1106 xfree_null (url->query);
1107 xfree_null (url->fragment);
1108 xfree_null (url->user);
1109 xfree_null (url->passwd);
1117 /* Create all the necessary directories for PATH (a file). Calls
1118 make_directory internally. */
1120 mkalldirs (const char *path)
1127 p = path + strlen (path);
1128 for (; *p != '/' && p != path; p--)
1131 /* Don't create if it's just a file. */
1132 if ((p == path) && (*p != '/'))
1134 t = strdupdelim (path, p);
1136 /* Check whether the directory exists. */
1137 if ((stat (t, &st) == 0))
1139 if (S_ISDIR (st.st_mode))
1146 /* If the dir exists as a file name, remove it first. This
1147 is *only* for Wget to work with buggy old CERN http
1148 servers. Here is the scenario: When Wget tries to
1149 retrieve a directory without a slash, e.g.
1150 http://foo/bar (bar being a directory), CERN server will
1151 not redirect it too http://foo/bar/ -- it will generate a
1152 directory listing containing links to bar/file1,
1153 bar/file2, etc. Wget will lose because it saves this
1154 HTML listing to a file `bar', so it cannot create the
1155 directory. To work around this, if the file of the same
1156 name exists, we just remove it and create the directory
1158 DEBUGP (("Removing %s because of directory danger!\n", t));
1162 res = make_directory (t);
1164 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1169 /* Functions for constructing the file name out of URL components. */
1171 /* A growable string structure, used by url_file_name and friends.
1172 This should perhaps be moved to utils.c.
1174 The idea is to have a convenient and efficient way to construct a
1175 string by having various functions append data to it. Instead of
1176 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1177 functions in questions, we pass the pointer to this struct. */
1185 /* Ensure that the string can accept APPEND_COUNT more characters past
1186 the current TAIL position. If necessary, this will grow the string
1187 and update its allocated size. If the string is already large
1188 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1189 #define GROW(g, append_size) do { \
1190 struct growable *G_ = g; \
1191 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1194 /* Return the tail position of the string. */
1195 #define TAIL(r) ((r)->base + (r)->tail)
1197 /* Move the tail position by APPEND_COUNT characters. */
1198 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1200 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1204 append_string (const char *str, struct growable *dest)
1206 int l = strlen (str);
1208 memcpy (TAIL (dest), str, l);
1209 TAIL_INCR (dest, l);
1212 /* Append CH to DEST. For example, append_char (0, DEST)
1213 zero-terminates DEST. */
1216 append_char (char ch, struct growable *dest)
1220 TAIL_INCR (dest, 1);
1224 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1225 filechr_not_windows = 2, /* unusable on MSDOS/Windows, one of \|/<>?:*" */
1226 filechr_control = 4 /* a control character, e.g. 0-31 */
1229 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1231 /* Shorthands for the table: */
1232 #define U filechr_not_unix
1233 #define W filechr_not_windows
1234 #define C filechr_control
1239 /* Table of characters unsafe under various conditions (see above).
1241 Arguably we could also claim `%' to be unsafe, since we use it as
1242 the escape character. If we ever want to be able to reliably
1243 translate file name back to URL, this would become important
1244 crucial. Right now, it's better to be minimal in escaping. */
1246 static const unsigned char filechr_table[256] =
1248 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1249 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1250 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1251 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1252 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1253 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1254 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1255 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1256 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1257 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1258 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1259 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1260 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1261 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1262 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1263 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1265 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1266 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1267 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1268 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1270 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1271 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1272 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1273 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1281 /* FN_PORT_SEP is the separator between host and port in file names
1282 for non-standard port numbers. On Unix this is normally ':', as in
1283 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1284 because Windows can't handle ':' in file names. */
1285 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1287 /* FN_QUERY_SEP is the separator between the file name and the URL
1288 query, normally '?'. Since Windows cannot handle '?' as part of
1289 file name, we use '@' instead there. */
1290 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1292 /* Quote path element, characters in [b, e), as file name, and append
1293 the quoted string to DEST. Each character is quoted as per
1294 file_unsafe_char and the corresponding table.
1296 If ESCAPED is true, the path element is considered to be
1297 URL-escaped and will be unescaped prior to inspection. */
1300 append_uri_pathel (const char *b, const char *e, bool escaped,
1301 struct growable *dest)
1307 if (opt.restrict_files_os == restrict_unix)
1308 mask = filechr_not_unix;
1310 mask = filechr_not_windows;
1311 if (opt.restrict_files_ctrl)
1312 mask |= filechr_control;
1314 /* Copy [b, e) to PATHEL and URL-unescape it. */
1318 BOUNDED_TO_ALLOCA (b, e, unescaped);
1319 url_unescape (unescaped);
1321 e = unescaped + strlen (unescaped);
1324 /* Defang ".." when found as component of path. Remember that path
1325 comes from the URL and might contain malicious input. */
1326 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1332 /* Walk the PATHEL string and check how many characters we'll need
1335 for (p = b; p < e; p++)
1336 if (FILE_CHAR_TEST (*p, mask))
1339 /* Calculate the length of the output string. e-b is the input
1340 string length. Each quoted char introduces two additional
1341 characters in the string, hence 2*quoted. */
1342 outlen = (e - b) + (2 * quoted);
1343 GROW (dest, outlen);
1347 /* If there's nothing to quote, we can simply append the string
1348 without processing it again. */
1349 memcpy (TAIL (dest), b, outlen);
1353 char *q = TAIL (dest);
1354 for (p = b; p < e; p++)
1356 if (!FILE_CHAR_TEST (*p, mask))
1360 unsigned char ch = *p;
1362 *q++ = XNUM_TO_DIGIT (ch >> 4);
1363 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1366 assert (q - TAIL (dest) == outlen);
1369 /* Perform inline case transformation if required. */
1370 if (opt.restrict_files_case == restrict_lowercase
1371 || opt.restrict_files_case == restrict_uppercase)
1374 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1376 if (opt.restrict_files_case == restrict_lowercase)
1383 TAIL_INCR (dest, outlen);
1386 /* Append to DEST the directory structure that corresponds the
1387 directory part of URL's path. For example, if the URL is
1388 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1390 Each path element ("dir1" and "dir2" in the above example) is
1391 examined, url-unescaped, and re-escaped as file name element.
1393 Additionally, it cuts as many directories from the path as
1394 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1395 will produce "bar" for the above example. For 2 or more, it will
1398 Each component of the path is quoted for use as file name. */
1401 append_dir_structure (const struct url *u, struct growable *dest)
1403 char *pathel, *next;
1404 int cut = opt.cut_dirs;
1406 /* Go through the path components, de-URL-quote them, and quote them
1407 (if necessary) as file names. */
1410 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1415 /* Ignore empty pathels. */
1419 append_char ('/', dest);
1420 append_uri_pathel (pathel, next, true, dest);
1424 /* Return a unique file name that matches the given URL as good as
1425 possible. Does not create directories on the file system. */
1428 url_file_name (const struct url *u)
1430 struct growable fnres; /* stands for "file name result" */
1432 const char *u_file, *u_query;
1433 char *fname, *unique;
1439 /* Start with the directory prefix, if specified. */
1441 append_string (opt.dir_prefix, &fnres);
1443 /* If "dirstruct" is turned on (typically the case with -r), add
1444 the host and port (unless those have been turned off) and
1445 directory structure. */
1448 if (opt.protocol_directories)
1451 append_char ('/', &fnres);
1452 append_string (supported_schemes[u->scheme].name, &fnres);
1454 if (opt.add_hostdir)
1457 append_char ('/', &fnres);
1458 if (0 != strcmp (u->host, ".."))
1459 append_string (u->host, &fnres);
1461 /* Host name can come from the network; malicious DNS may
1462 allow ".." to be resolved, causing us to write to
1463 "../<file>". Defang such host names. */
1464 append_string ("%2E%2E", &fnres);
1465 if (u->port != scheme_default_port (u->scheme))
1468 number_to_string (portstr, u->port);
1469 append_char (FN_PORT_SEP, &fnres);
1470 append_string (portstr, &fnres);
1474 append_dir_structure (u, &fnres);
1477 /* Add the file name. */
1479 append_char ('/', &fnres);
1480 u_file = *u->file ? u->file : "index.html";
1481 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1483 /* Append "?query" to the file name. */
1484 u_query = u->query && *u->query ? u->query : NULL;
1487 append_char (FN_QUERY_SEP, &fnres);
1488 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1491 /* Zero-terminate the file name. */
1492 append_char ('\0', &fnres);
1496 /* Check the cases in which the unique extensions are not used:
1497 1) Clobbering is turned off (-nc).
1498 2) Retrieval with regetting.
1499 3) Timestamping is used.
1500 4) Hierarchy is built.
1502 The exception is the case when file does exist and is a
1503 directory (see `mkalldirs' for explanation). */
1505 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1506 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1509 unique = unique_name (fname, true);
1510 if (unique != fname)
1515 /* Resolve "." and ".." elements of PATH by destructively modifying
1516 PATH and return true if PATH has been modified, false otherwise.
1518 The algorithm is in spirit similar to the one described in rfc1808,
1519 although implemented differently, in one pass. To recap, path
1520 elements containing only "." are removed, and ".." is taken to mean
1521 "back up one element". Single leading and trailing slashes are
1524 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1525 test examples are provided below. If you change anything in this
1526 function, run test_path_simplify to make sure you haven't broken a
1530 path_simplify (char *path)
1532 char *h = path; /* hare */
1533 char *t = path; /* tortoise */
1534 char *end = strchr (path, '\0');
1538 /* Hare should be at the beginning of a path element. */
1540 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1545 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1547 /* Handle "../" by retreating the tortoise by one path
1548 element -- but not past beggining. */
1551 /* Move backwards until T hits the beginning of the
1552 previous path element or the beginning of path. */
1553 for (--t; t > path && t[-1] != '/'; t--)
1560 /* A regular path element. If H hasn't advanced past T,
1561 simply skip to the next path element. Otherwise, copy
1562 the path element until the next slash. */
1565 /* Skip the path element, including the slash. */
1566 while (h < end && *h != '/')
1573 /* Copy the path element, including the final slash. */
1574 while (h < end && *h != '/')
1588 /* Return the length of URL's path. Path is considered to be
1589 terminated by one or more of the ?query or ;params or #fragment,
1590 depending on the scheme. */
1593 path_end (const char *url)
1595 enum url_scheme scheme = url_scheme (url);
1597 if (scheme == SCHEME_INVALID)
1598 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1599 /* +2 to ignore the first two separators ':' and '/' */
1600 seps = init_seps (scheme) + 2;
1601 return strpbrk_or_eos (url, seps);
1604 /* Find the last occurrence of character C in the range [b, e), or
1605 NULL, if none are present. */
1606 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1608 /* Merge BASE with LINK and return the resulting URI.
1610 Either of the URIs may be absolute or relative, complete with the
1611 host name, or path only. This tries to reasonably handle all
1612 foreseeable cases. It only employs minimal URL parsing, without
1613 knowledge of the specifics of schemes.
1615 I briefly considered making this function call path_simplify after
1616 the merging process, as rfc1738 seems to suggest. This is a bad
1617 idea for several reasons: 1) it complexifies the code, and 2)
1618 url_parse has to simplify path anyway, so it's wasteful to boot. */
1621 uri_merge (const char *base, const char *link)
1627 if (url_has_scheme (link))
1628 return xstrdup (link);
1630 /* We may not examine BASE past END. */
1631 end = path_end (base);
1632 linklength = strlen (link);
1636 /* Empty LINK points back to BASE, query string and all. */
1637 return xstrdup (base);
1639 else if (*link == '?')
1641 /* LINK points to the same location, but changes the query
1642 string. Examples: */
1643 /* uri_merge("path", "?new") -> "path?new" */
1644 /* uri_merge("path?foo", "?new") -> "path?new" */
1645 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1646 /* uri_merge("path#foo", "?new") -> "path?new" */
1647 int baselength = end - base;
1648 merge = xmalloc (baselength + linklength + 1);
1649 memcpy (merge, base, baselength);
1650 memcpy (merge + baselength, link, linklength);
1651 merge[baselength + linklength] = '\0';
1653 else if (*link == '#')
1655 /* uri_merge("path", "#new") -> "path#new" */
1656 /* uri_merge("path#foo", "#new") -> "path#new" */
1657 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1658 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1660 const char *end1 = strchr (base, '#');
1662 end1 = base + strlen (base);
1663 baselength = end1 - base;
1664 merge = xmalloc (baselength + linklength + 1);
1665 memcpy (merge, base, baselength);
1666 memcpy (merge + baselength, link, linklength);
1667 merge[baselength + linklength] = '\0';
1669 else if (*link == '/' && *(link + 1) == '/')
1671 /* LINK begins with "//" and so is a net path: we need to
1672 replace everything after (and including) the double slash
1675 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1676 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1677 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1681 const char *start_insert;
1683 /* Look for first slash. */
1684 slash = memchr (base, '/', end - base);
1685 /* If found slash and it is a double slash, then replace
1686 from this point, else default to replacing from the
1688 if (slash && *(slash + 1) == '/')
1689 start_insert = slash;
1691 start_insert = base;
1693 span = start_insert - base;
1694 merge = xmalloc (span + linklength + 1);
1696 memcpy (merge, base, span);
1697 memcpy (merge + span, link, linklength);
1698 merge[span + linklength] = '\0';
1700 else if (*link == '/')
1702 /* LINK is an absolute path: we need to replace everything
1703 after (and including) the FIRST slash with LINK.
1705 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1706 "/qux/xyzzy", our result should be
1707 "http://host/qux/xyzzy". */
1710 const char *start_insert = NULL; /* for gcc to shut up. */
1711 const char *pos = base;
1712 bool seen_slash_slash = false;
1713 /* We're looking for the first slash, but want to ignore
1716 slash = memchr (pos, '/', end - pos);
1717 if (slash && !seen_slash_slash)
1718 if (*(slash + 1) == '/')
1721 seen_slash_slash = true;
1725 /* At this point, SLASH is the location of the first / after
1726 "//", or the first slash altogether. START_INSERT is the
1727 pointer to the location where LINK will be inserted. When
1728 examining the last two examples, keep in mind that LINK
1731 if (!slash && !seen_slash_slash)
1732 /* example: "foo" */
1734 start_insert = base;
1735 else if (!slash && seen_slash_slash)
1736 /* example: "http://foo" */
1739 else if (slash && !seen_slash_slash)
1740 /* example: "foo/bar" */
1742 start_insert = base;
1743 else if (slash && seen_slash_slash)
1744 /* example: "http://something/" */
1746 start_insert = slash;
1748 span = start_insert - base;
1749 merge = xmalloc (span + linklength + 1);
1751 memcpy (merge, base, span);
1752 memcpy (merge + span, link, linklength);
1753 merge[span + linklength] = '\0';
1757 /* LINK is a relative URL: we need to replace everything
1758 after last slash (possibly empty) with LINK.
1760 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1761 our result should be "whatever/foo/qux/xyzzy". */
1762 bool need_explicit_slash = false;
1764 const char *start_insert;
1765 const char *last_slash = find_last_char (base, end, '/');
1768 /* No slash found at all. Replace what we have with LINK. */
1769 start_insert = base;
1771 else if (last_slash && last_slash >= base + 2
1772 && last_slash[-2] == ':' && last_slash[-1] == '/')
1774 /* example: http://host" */
1776 start_insert = end + 1;
1777 need_explicit_slash = true;
1781 /* example: "whatever/foo/bar" */
1783 start_insert = last_slash + 1;
1786 span = start_insert - base;
1787 merge = xmalloc (span + linklength + 1);
1789 memcpy (merge, base, span);
1790 if (need_explicit_slash)
1791 merge[span - 1] = '/';
1792 memcpy (merge + span, link, linklength);
1793 merge[span + linklength] = '\0';
1799 #define APPEND(p, s) do { \
1800 int len = strlen (s); \
1801 memcpy (p, s, len); \
1805 /* Use this instead of password when the actual password is supposed
1806 to be hidden. We intentionally use a generic string without giving
1807 away the number of characters in the password, like previous
1809 #define HIDDEN_PASSWORD "*password*"
1811 /* Recreate the URL string from the data in URL.
1813 If HIDE is true (as it is when we're calling this on a URL we plan
1814 to print, but not when calling it to canonicalize a URL for use
1815 within the program), password will be hidden. Unsafe characters in
1816 the URL will be quoted. */
1819 url_string (const struct url *url, enum url_auth_mode auth_mode)
1823 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1825 int scheme_port = supported_schemes[url->scheme].default_port;
1826 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1827 int fplen = full_path_length (url);
1829 bool brackets_around_host;
1831 assert (scheme_str != NULL);
1833 /* Make sure the user name and password are quoted. */
1836 if (auth_mode != URL_AUTH_HIDE)
1838 quoted_user = url_escape_allow_passthrough (url->user);
1841 if (auth_mode == URL_AUTH_HIDE_PASSWD)
1842 quoted_passwd = HIDDEN_PASSWORD;
1844 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1849 /* In the unlikely event that the host name contains non-printable
1850 characters, quote it for displaying to the user. */
1851 quoted_host = url_escape_allow_passthrough (url->host);
1853 /* Undo the quoting of colons that URL escaping performs. IPv6
1854 addresses may legally contain colons, and in that case must be
1855 placed in square brackets. */
1856 if (quoted_host != url->host)
1857 unescape_single_char (quoted_host, ':');
1858 brackets_around_host = strchr (quoted_host, ':') != NULL;
1860 size = (strlen (scheme_str)
1861 + strlen (quoted_host)
1862 + (brackets_around_host ? 2 : 0)
1865 if (url->port != scheme_port)
1866 size += 1 + numdigit (url->port);
1869 size += 1 + strlen (quoted_user);
1871 size += 1 + strlen (quoted_passwd);
1874 p = result = xmalloc (size);
1876 APPEND (p, scheme_str);
1879 APPEND (p, quoted_user);
1883 APPEND (p, quoted_passwd);
1888 if (brackets_around_host)
1890 APPEND (p, quoted_host);
1891 if (brackets_around_host)
1893 if (url->port != scheme_port)
1896 p = number_to_string (p, url->port);
1899 full_path_write (url, p);
1903 assert (p - result == size);
1905 if (quoted_user && quoted_user != url->user)
1906 xfree (quoted_user);
1907 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
1908 && quoted_passwd != url->passwd)
1909 xfree (quoted_passwd);
1910 if (quoted_host != url->host)
1911 xfree (quoted_host);
1916 /* Return true if scheme a is similar to scheme b.
1918 Schemes are similar if they are equal. If SSL is supported, schemes
1919 are also similar if one is http (SCHEME_HTTP) and the other is https
1922 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1927 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1928 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1935 getchar_from_escaped_string (const char *str, char *c)
1937 const char *p = str;
1939 assert (str && *str);
1944 if (!ISXDIGIT(p[1]) || !ISXDIGIT(p[2]))
1952 return 0; /* error: invalid string */
1954 *c = X2DIGITS_TO_NUM (p[1], p[2]);
1955 if (URL_RESERVED_CHAR(*c))
1973 are_urls_equal (const char *u1, const char *u2)
1984 && (pp = getchar_from_escaped_string (p, &ch1))
1985 && (qq = getchar_from_escaped_string (q, &ch2))
1986 && (TOLOWER(ch1) == TOLOWER(ch2)))
1992 return (*p == 0 && *q == 0 ? true : false);
1996 /* Debugging and testing support for path_simplify. */
1998 /* Debug: run path_simplify on PATH and return the result in a new
1999 string. Useful for calling from the debugger. */
2003 char *copy = xstrdup (path);
2004 path_simplify (copy);
2009 run_test (char *test, char *expected_result, bool expected_change)
2011 char *test_copy = xstrdup (test);
2012 bool modified = path_simplify (test_copy);
2014 if (0 != strcmp (test_copy, expected_result))
2016 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2017 test, expected_result, test_copy);
2019 if (modified != expected_change)
2021 if (expected_change)
2022 printf ("Expected modification with path_simplify(\"%s\").\n",
2025 printf ("Expected no modification with path_simplify(\"%s\").\n",
2032 test_path_simplify (void)
2035 char *test, *result;
2042 { "../", "", true },
2043 { "foo", "foo", false },
2044 { "foo/bar", "foo/bar", false },
2045 { "foo///bar", "foo///bar", false },
2046 { "foo/.", "foo/", true },
2047 { "foo/./", "foo/", true },
2048 { "foo./", "foo./", false },
2049 { "foo/../bar", "bar", true },
2050 { "foo/../bar/", "bar/", true },
2051 { "foo/bar/..", "foo/", true },
2052 { "foo/bar/../x", "foo/x", true },
2053 { "foo/bar/../x/", "foo/x/", true },
2054 { "foo/..", "", true },
2055 { "foo/../..", "", true },
2056 { "foo/../../..", "", true },
2057 { "foo/../../bar/../../baz", "baz", true },
2058 { "a/b/../../c", "c", true },
2059 { "./a/../b", "b", true }
2063 for (i = 0; i < countof (tests); i++)
2065 char *test = tests[i].test;
2066 char *expected_result = tests[i].result;
2067 bool expected_change = tests[i].should_modify;
2068 run_test (test, expected_result, expected_change);
2076 test_append_uri_pathel()
2083 char *expected_result;
2085 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2088 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2090 struct growable dest;
2091 const char *p = test_array[i].input;
2093 memset (&dest, 0, sizeof (dest));
2095 append_string (test_array[i].original_url, &dest);
2096 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2097 append_char ('\0', &dest);
2099 mu_assert ("test_append_uri_pathel: wrong result",
2100 strcmp (dest.base, test_array[i].expected_result) == 0);
2107 test_are_urls_equal()
2113 bool expected_result;
2115 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2116 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2117 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2118 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2119 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2120 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2123 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2125 mu_assert ("test_are_urls_equal: wrong result",
2126 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2132 #endif /* TESTING */