2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
3 2004, 2005, 2006, 2007, 2008 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 */
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 (enum url_scheme, 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] || !(c_isxdigit (h[1]) && c_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 (c_isxdigit (*(p + 1)) && c_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) (c_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;
594 if (c_isupper (*str))
597 *str = c_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;
676 url_unescape ((char *) url);
677 url = locale_to_utf8(url);
680 url_encoded = reencode_escapes (url);
683 p += strlen (supported_schemes[scheme].leading_string);
685 p = url_skip_credentials (p);
688 /* scheme://user:pass@host[:port]... */
691 /* We attempt to break down the URL into the components path,
692 params, query, and fragment. They are ordered like this:
694 scheme://host[:port][/path][;params][?query][#fragment] */
696 path_b = path_e = NULL;
697 params_b = params_e = NULL;
698 query_b = query_e = NULL;
699 fragment_b = fragment_e = NULL;
701 /* Initialize separators for optional parts of URL, depending on the
702 scheme. For example, FTP has params, and HTTP and HTTPS have
703 query string and fragment. */
704 seps = init_seps (scheme);
710 /* Handle IPv6 address inside square brackets. Ideally we'd
711 just look for the terminating ']', but rfc2732 mandates
712 rejecting invalid IPv6 addresses. */
714 /* The address begins after '['. */
716 host_e = strchr (host_b, ']');
720 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
725 /* Check if the IPv6 address is valid. */
726 if (!is_valid_ipv6_address(host_b, host_e))
728 error_code = PE_INVALID_IPV6_ADDRESS;
732 /* Continue parsing after the closing ']'. */
735 error_code = PE_IPV6_NOT_SUPPORTED;
739 /* The closing bracket must be followed by a separator or by the
741 /* http://[::1]... */
743 if (!strchr (seps, *p))
745 /* Trailing garbage after []-delimited IPv6 address. */
746 error_code = PE_INVALID_HOST_NAME;
752 p = strpbrk_or_eos (p, seps);
755 ++seps; /* advance to '/' */
757 if (host_b == host_e)
759 error_code = PE_INVALID_HOST_NAME;
763 port = scheme_default_port (scheme);
766 const char *port_b, *port_e, *pp;
768 /* scheme://host:port/tralala */
772 p = strpbrk_or_eos (p, seps);
775 /* Allow empty port, as per rfc2396. */
776 if (port_b != port_e)
777 for (port = 0, pp = port_b; pp < port_e; pp++)
779 if (!c_isdigit (*pp))
781 /* http://host:12randomgarbage/blah */
783 error_code = PE_BAD_PORT_NUMBER;
786 port = 10 * port + (*pp - '0');
787 /* Check for too large port numbers here, before we have
788 a chance to overflow on bogus port values. */
791 error_code = PE_BAD_PORT_NUMBER;
796 /* Advance to the first separator *after* '/' (either ';' or '?',
797 depending on the scheme). */
800 /* Get the optional parts of URL, each part being delimited by
801 current location and the position of the next separator. */
802 #define GET_URL_PART(sepchar, var) do { \
804 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
808 GET_URL_PART ('/', path);
809 if (supported_schemes[scheme].flags & scm_has_params)
810 GET_URL_PART (';', params);
811 if (supported_schemes[scheme].flags & scm_has_query)
812 GET_URL_PART ('?', query);
813 if (supported_schemes[scheme].flags & scm_has_fragment)
814 GET_URL_PART ('#', fragment);
819 if (uname_b != uname_e)
821 /* http://user:pass@host */
823 /* uname_b uname_e */
824 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
826 error_code = PE_INVALID_USER_NAME;
831 u = xnew0 (struct url);
833 u->host = strdupdelim (host_b, host_e);
838 u->path = strdupdelim (path_b, path_e);
839 path_modified = path_simplify (scheme, u->path);
840 split_path (u->path, &u->dir, &u->file);
842 host_modified = lowercase_str (u->host);
844 /* Decode %HH sequences in host name. This is important not so much
845 to support %HH sequences in host names (which other browser
846 don't), but to support binary characters (which will have been
847 converted to %HH by reencode_escapes). */
848 if (strchr (u->host, '%'))
850 url_unescape (u->host);
851 host_modified = true;
856 char *new = idn_encode (u->host);
861 host_modified = true;
866 u->params = strdupdelim (params_b, params_e);
868 u->query = strdupdelim (query_b, query_e);
870 u->fragment = strdupdelim (fragment_b, fragment_e);
872 if (opt.enable_iri || path_modified || u->fragment || host_modified || path_b == path_e)
874 /* If we suspect that a transformation has rendered what
875 url_string might return different from URL_ENCODED, rebuild
876 u->url using url_string. */
877 u->url = url_string (u, URL_AUTH_SHOW);
879 if (url_encoded != url)
880 xfree ((char *) url_encoded);
884 if (url_encoded == url)
885 u->url = xstrdup (url);
887 u->url = url_encoded;
893 /* Cleanup in case of error: */
894 if (url_encoded && url_encoded != url)
897 /* Transmit the error code to the caller, if the caller wants to
904 /* Return the error message string from ERROR_CODE, which should have
905 been retrieved from url_parse. The error message is translated. */
908 url_error (int error_code)
910 assert (error_code >= 0 && ((size_t) error_code) < countof (parse_errors));
911 return _(parse_errors[error_code]);
914 /* Split PATH into DIR and FILE. PATH comes from the URL and is
915 expected to be URL-escaped.
917 The path is split into directory (the part up to the last slash)
918 and file (the part after the last slash), which are subsequently
922 "foo/bar/baz" "foo/bar" "baz"
923 "foo/bar/" "foo/bar" ""
925 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
927 DIR and FILE are freshly allocated. */
930 split_path (const char *path, char **dir, char **file)
932 char *last_slash = strrchr (path, '/');
936 *file = xstrdup (path);
940 *dir = strdupdelim (path, last_slash);
941 *file = xstrdup (last_slash + 1);
944 url_unescape (*file);
947 /* Note: URL's "full path" is the path with the query string and
948 params appended. The "fragment" (#foo) is intentionally ignored,
949 but that might be changed. For example, if the original URL was
950 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
951 the full path will be "/foo/bar/baz;bullshit?querystring". */
953 /* Return the length of the full path, without the terminating
957 full_path_length (const struct url *url)
961 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
972 /* Write out the full path. */
975 full_path_write (const struct url *url, char *where)
977 #define FROB(el, chr) do { \
978 char *f_el = url->el; \
980 int l = strlen (f_el); \
982 memcpy (where, f_el, l); \
994 /* Public function for getting the "full path". E.g. if u->path is
995 "foo/bar" and u->query is "param=value", full_path will be
996 "/foo/bar?param=value". */
999 url_full_path (const struct url *url)
1001 int length = full_path_length (url);
1002 char *full_path = xmalloc (length + 1);
1004 full_path_write (url, full_path);
1005 full_path[length] = '\0';
1010 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1011 escaping of certain characters, such as "/" and ":". Returns a
1012 count of unescaped chars. */
1015 unescape_single_char (char *str, char chr)
1017 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1018 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1019 char *h = str; /* hare */
1020 char *t = str; /* tortoise */
1021 for (; *h; h++, t++)
1023 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1034 /* Escape unsafe and reserved characters, except for the slash
1038 url_escape_dir (const char *dir)
1040 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1044 unescape_single_char (newdir, '/');
1048 /* Sync u->path and u->url with u->dir and u->file. Called after
1049 u->file or u->dir have been changed, typically by the FTP code. */
1052 sync_path (struct url *u)
1054 char *newpath, *efile, *edir;
1058 /* u->dir and u->file are not escaped. URL-escape them before
1059 reassembling them into u->path. That way, if they contain
1060 separators like '?' or even if u->file contains slashes, the
1061 path will be correctly assembled. (u->file can contain slashes
1062 if the URL specifies it with %2f, or if an FTP server returns
1064 edir = url_escape_dir (u->dir);
1065 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1068 newpath = xstrdup (efile);
1071 int dirlen = strlen (edir);
1072 int filelen = strlen (efile);
1074 /* Copy "DIR/FILE" to newpath. */
1075 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1076 memcpy (p, edir, dirlen);
1079 memcpy (p, efile, filelen);
1088 if (efile != u->file)
1091 /* Regenerate u->url as well. */
1093 u->url = url_string (u, URL_AUTH_SHOW);
1096 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1097 This way we can sync u->path and u->url when they get changed. */
1100 url_set_dir (struct url *url, const char *newdir)
1103 url->dir = xstrdup (newdir);
1108 url_set_file (struct url *url, const char *newfile)
1111 url->file = xstrdup (newfile);
1116 url_free (struct url *url)
1122 xfree_null (url->params);
1123 xfree_null (url->query);
1124 xfree_null (url->fragment);
1125 xfree_null (url->user);
1126 xfree_null (url->passwd);
1134 /* Create all the necessary directories for PATH (a file). Calls
1135 make_directory internally. */
1137 mkalldirs (const char *path)
1144 p = path + strlen (path);
1145 for (; *p != '/' && p != path; p--)
1148 /* Don't create if it's just a file. */
1149 if ((p == path) && (*p != '/'))
1151 t = strdupdelim (path, p);
1153 /* Check whether the directory exists. */
1154 if ((stat (t, &st) == 0))
1156 if (S_ISDIR (st.st_mode))
1163 /* If the dir exists as a file name, remove it first. This
1164 is *only* for Wget to work with buggy old CERN http
1165 servers. Here is the scenario: When Wget tries to
1166 retrieve a directory without a slash, e.g.
1167 http://foo/bar (bar being a directory), CERN server will
1168 not redirect it too http://foo/bar/ -- it will generate a
1169 directory listing containing links to bar/file1,
1170 bar/file2, etc. Wget will lose because it saves this
1171 HTML listing to a file `bar', so it cannot create the
1172 directory. To work around this, if the file of the same
1173 name exists, we just remove it and create the directory
1175 DEBUGP (("Removing %s because of directory danger!\n", t));
1179 res = make_directory (t);
1181 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1186 /* Functions for constructing the file name out of URL components. */
1188 /* A growable string structure, used by url_file_name and friends.
1189 This should perhaps be moved to utils.c.
1191 The idea is to have a convenient and efficient way to construct a
1192 string by having various functions append data to it. Instead of
1193 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1194 functions in questions, we pass the pointer to this struct. */
1202 /* Ensure that the string can accept APPEND_COUNT more characters past
1203 the current TAIL position. If necessary, this will grow the string
1204 and update its allocated size. If the string is already large
1205 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1206 #define GROW(g, append_size) do { \
1207 struct growable *G_ = g; \
1208 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1211 /* Return the tail position of the string. */
1212 #define TAIL(r) ((r)->base + (r)->tail)
1214 /* Move the tail position by APPEND_COUNT characters. */
1215 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1217 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1221 append_string (const char *str, struct growable *dest)
1223 int l = strlen (str);
1225 memcpy (TAIL (dest), str, l);
1226 TAIL_INCR (dest, l);
1229 /* Append CH to DEST. For example, append_char (0, DEST)
1230 zero-terminates DEST. */
1233 append_char (char ch, struct growable *dest)
1237 TAIL_INCR (dest, 1);
1241 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1242 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1243 filechr_control = 4 /* a control character, e.g. 0-31 */
1246 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1248 /* Shorthands for the table: */
1249 #define U filechr_not_unix
1250 #define W filechr_not_windows
1251 #define C filechr_control
1256 /* Table of characters unsafe under various conditions (see above).
1258 Arguably we could also claim `%' to be unsafe, since we use it as
1259 the escape character. If we ever want to be able to reliably
1260 translate file name back to URL, this would become important
1261 crucial. Right now, it's better to be minimal in escaping. */
1263 static const unsigned char filechr_table[256] =
1265 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1266 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1267 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1268 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1269 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1270 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1271 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1272 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1273 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1274 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1275 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1276 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1277 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1278 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1279 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1280 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1282 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1283 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1284 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1285 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1287 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1288 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1289 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1290 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1298 /* FN_PORT_SEP is the separator between host and port in file names
1299 for non-standard port numbers. On Unix this is normally ':', as in
1300 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1301 because Windows can't handle ':' in file names. */
1302 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1304 /* FN_QUERY_SEP is the separator between the file name and the URL
1305 query, normally '?'. Since Windows cannot handle '?' as part of
1306 file name, we use '@' instead there. */
1307 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1309 /* Quote path element, characters in [b, e), as file name, and append
1310 the quoted string to DEST. Each character is quoted as per
1311 file_unsafe_char and the corresponding table.
1313 If ESCAPED is true, the path element is considered to be
1314 URL-escaped and will be unescaped prior to inspection. */
1317 append_uri_pathel (const char *b, const char *e, bool escaped,
1318 struct growable *dest)
1324 if (opt.restrict_files_os == restrict_unix)
1325 mask = filechr_not_unix;
1327 mask = filechr_not_windows;
1328 if (opt.restrict_files_ctrl)
1329 mask |= filechr_control;
1331 /* Copy [b, e) to PATHEL and URL-unescape it. */
1335 BOUNDED_TO_ALLOCA (b, e, unescaped);
1336 url_unescape (unescaped);
1338 e = unescaped + strlen (unescaped);
1341 /* Defang ".." when found as component of path. Remember that path
1342 comes from the URL and might contain malicious input. */
1343 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1349 /* Walk the PATHEL string and check how many characters we'll need
1352 for (p = b; p < e; p++)
1353 if (FILE_CHAR_TEST (*p, mask))
1356 /* Calculate the length of the output string. e-b is the input
1357 string length. Each quoted char introduces two additional
1358 characters in the string, hence 2*quoted. */
1359 outlen = (e - b) + (2 * quoted);
1360 GROW (dest, outlen);
1364 /* If there's nothing to quote, we can simply append the string
1365 without processing it again. */
1366 memcpy (TAIL (dest), b, outlen);
1370 char *q = TAIL (dest);
1371 for (p = b; p < e; p++)
1373 if (!FILE_CHAR_TEST (*p, mask))
1377 unsigned char ch = *p;
1379 *q++ = XNUM_TO_DIGIT (ch >> 4);
1380 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1383 assert (q - TAIL (dest) == outlen);
1386 /* Perform inline case transformation if required. */
1387 if (opt.restrict_files_case == restrict_lowercase
1388 || opt.restrict_files_case == restrict_uppercase)
1391 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1393 if (opt.restrict_files_case == restrict_lowercase)
1394 *q = c_tolower (*q);
1396 *q = c_toupper (*q);
1400 TAIL_INCR (dest, outlen);
1403 /* Append to DEST the directory structure that corresponds the
1404 directory part of URL's path. For example, if the URL is
1405 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1407 Each path element ("dir1" and "dir2" in the above example) is
1408 examined, url-unescaped, and re-escaped as file name element.
1410 Additionally, it cuts as many directories from the path as
1411 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1412 will produce "bar" for the above example. For 2 or more, it will
1415 Each component of the path is quoted for use as file name. */
1418 append_dir_structure (const struct url *u, struct growable *dest)
1420 char *pathel, *next;
1421 int cut = opt.cut_dirs;
1423 /* Go through the path components, de-URL-quote them, and quote them
1424 (if necessary) as file names. */
1427 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1432 /* Ignore empty pathels. */
1436 append_char ('/', dest);
1437 append_uri_pathel (pathel, next, true, dest);
1441 /* Return a unique file name that matches the given URL as good as
1442 possible. Does not create directories on the file system. */
1445 url_file_name (const struct url *u)
1447 struct growable fnres; /* stands for "file name result" */
1449 const char *u_file, *u_query;
1450 char *fname, *unique;
1456 /* Start with the directory prefix, if specified. */
1458 append_string (opt.dir_prefix, &fnres);
1460 /* If "dirstruct" is turned on (typically the case with -r), add
1461 the host and port (unless those have been turned off) and
1462 directory structure. */
1465 if (opt.protocol_directories)
1468 append_char ('/', &fnres);
1469 append_string (supported_schemes[u->scheme].name, &fnres);
1471 if (opt.add_hostdir)
1474 append_char ('/', &fnres);
1475 if (0 != strcmp (u->host, ".."))
1476 append_string (u->host, &fnres);
1478 /* Host name can come from the network; malicious DNS may
1479 allow ".." to be resolved, causing us to write to
1480 "../<file>". Defang such host names. */
1481 append_string ("%2E%2E", &fnres);
1482 if (u->port != scheme_default_port (u->scheme))
1485 number_to_string (portstr, u->port);
1486 append_char (FN_PORT_SEP, &fnres);
1487 append_string (portstr, &fnres);
1491 append_dir_structure (u, &fnres);
1494 /* Add the file name. */
1496 append_char ('/', &fnres);
1497 u_file = *u->file ? u->file : "index.html";
1498 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1500 /* Append "?query" to the file name. */
1501 u_query = u->query && *u->query ? u->query : NULL;
1504 append_char (FN_QUERY_SEP, &fnres);
1505 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1508 /* Zero-terminate the file name. */
1509 append_char ('\0', &fnres);
1513 /* Check the cases in which the unique extensions are not used:
1514 1) Clobbering is turned off (-nc).
1515 2) Retrieval with regetting.
1516 3) Timestamping is used.
1517 4) Hierarchy is built.
1519 The exception is the case when file does exist and is a
1520 directory (see `mkalldirs' for explanation). */
1522 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1523 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1526 unique = unique_name (fname, true);
1527 if (unique != fname)
1532 /* Resolve "." and ".." elements of PATH by destructively modifying
1533 PATH and return true if PATH has been modified, false otherwise.
1535 The algorithm is in spirit similar to the one described in rfc1808,
1536 although implemented differently, in one pass. To recap, path
1537 elements containing only "." are removed, and ".." is taken to mean
1538 "back up one element". Single leading and trailing slashes are
1541 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1542 test examples are provided below. If you change anything in this
1543 function, run test_path_simplify to make sure you haven't broken a
1547 path_simplify (enum url_scheme scheme, char *path)
1549 char *h = path; /* hare */
1550 char *t = path; /* tortoise */
1552 char *end = strchr (path, '\0');
1556 /* Hare should be at the beginning of a path element. */
1558 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1563 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1565 /* Handle "../" by retreating the tortoise by one path
1566 element -- but not past beggining. */
1569 /* Move backwards until T hits the beginning of the
1570 previous path element or the beginning of path. */
1571 for (--t; t > beg && t[-1] != '/'; t--)
1574 else if (scheme == SCHEME_FTP)
1576 /* If we're at the beginning, copy the "../" literally
1577 and move the beginning so a later ".." doesn't remove
1578 it. This violates RFC 3986; but we do it for FTP
1579 anyway because there is otherwise no way to get at a
1580 parent directory, when the FTP server drops us in a
1581 non-root directory (which is not uncommon). */
1590 /* A regular path element. If H hasn't advanced past T,
1591 simply skip to the next path element. Otherwise, copy
1592 the path element until the next slash. */
1595 /* Skip the path element, including the slash. */
1596 while (h < end && *h != '/')
1603 /* Copy the path element, including the final slash. */
1604 while (h < end && *h != '/')
1618 /* Return the length of URL's path. Path is considered to be
1619 terminated by one or more of the ?query or ;params or #fragment,
1620 depending on the scheme. */
1623 path_end (const char *url)
1625 enum url_scheme scheme = url_scheme (url);
1627 if (scheme == SCHEME_INVALID)
1628 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1629 /* +2 to ignore the first two separators ':' and '/' */
1630 seps = init_seps (scheme) + 2;
1631 return strpbrk_or_eos (url, seps);
1634 /* Find the last occurrence of character C in the range [b, e), or
1635 NULL, if none are present. */
1636 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1638 /* Merge BASE with LINK and return the resulting URI.
1640 Either of the URIs may be absolute or relative, complete with the
1641 host name, or path only. This tries to reasonably handle all
1642 foreseeable cases. It only employs minimal URL parsing, without
1643 knowledge of the specifics of schemes.
1645 I briefly considered making this function call path_simplify after
1646 the merging process, as rfc1738 seems to suggest. This is a bad
1647 idea for several reasons: 1) it complexifies the code, and 2)
1648 url_parse has to simplify path anyway, so it's wasteful to boot. */
1651 uri_merge (const char *base, const char *link)
1657 if (url_has_scheme (link))
1658 return xstrdup (link);
1660 /* We may not examine BASE past END. */
1661 end = path_end (base);
1662 linklength = strlen (link);
1666 /* Empty LINK points back to BASE, query string and all. */
1667 return xstrdup (base);
1669 else if (*link == '?')
1671 /* LINK points to the same location, but changes the query
1672 string. Examples: */
1673 /* uri_merge("path", "?new") -> "path?new" */
1674 /* uri_merge("path?foo", "?new") -> "path?new" */
1675 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1676 /* uri_merge("path#foo", "?new") -> "path?new" */
1677 int baselength = end - base;
1678 merge = xmalloc (baselength + linklength + 1);
1679 memcpy (merge, base, baselength);
1680 memcpy (merge + baselength, link, linklength);
1681 merge[baselength + linklength] = '\0';
1683 else if (*link == '#')
1685 /* uri_merge("path", "#new") -> "path#new" */
1686 /* uri_merge("path#foo", "#new") -> "path#new" */
1687 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1688 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1690 const char *end1 = strchr (base, '#');
1692 end1 = base + strlen (base);
1693 baselength = end1 - base;
1694 merge = xmalloc (baselength + linklength + 1);
1695 memcpy (merge, base, baselength);
1696 memcpy (merge + baselength, link, linklength);
1697 merge[baselength + linklength] = '\0';
1699 else if (*link == '/' && *(link + 1) == '/')
1701 /* LINK begins with "//" and so is a net path: we need to
1702 replace everything after (and including) the double slash
1705 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1706 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1707 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1711 const char *start_insert;
1713 /* Look for first slash. */
1714 slash = memchr (base, '/', end - base);
1715 /* If found slash and it is a double slash, then replace
1716 from this point, else default to replacing from the
1718 if (slash && *(slash + 1) == '/')
1719 start_insert = slash;
1721 start_insert = base;
1723 span = start_insert - base;
1724 merge = xmalloc (span + linklength + 1);
1726 memcpy (merge, base, span);
1727 memcpy (merge + span, link, linklength);
1728 merge[span + linklength] = '\0';
1730 else if (*link == '/')
1732 /* LINK is an absolute path: we need to replace everything
1733 after (and including) the FIRST slash with LINK.
1735 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1736 "/qux/xyzzy", our result should be
1737 "http://host/qux/xyzzy". */
1740 const char *start_insert = NULL; /* for gcc to shut up. */
1741 const char *pos = base;
1742 bool seen_slash_slash = false;
1743 /* We're looking for the first slash, but want to ignore
1746 slash = memchr (pos, '/', end - pos);
1747 if (slash && !seen_slash_slash)
1748 if (*(slash + 1) == '/')
1751 seen_slash_slash = true;
1755 /* At this point, SLASH is the location of the first / after
1756 "//", or the first slash altogether. START_INSERT is the
1757 pointer to the location where LINK will be inserted. When
1758 examining the last two examples, keep in mind that LINK
1761 if (!slash && !seen_slash_slash)
1762 /* example: "foo" */
1764 start_insert = base;
1765 else if (!slash && seen_slash_slash)
1766 /* example: "http://foo" */
1769 else if (slash && !seen_slash_slash)
1770 /* example: "foo/bar" */
1772 start_insert = base;
1773 else if (slash && seen_slash_slash)
1774 /* example: "http://something/" */
1776 start_insert = slash;
1778 span = start_insert - base;
1779 merge = xmalloc (span + linklength + 1);
1781 memcpy (merge, base, span);
1782 memcpy (merge + span, link, linklength);
1783 merge[span + linklength] = '\0';
1787 /* LINK is a relative URL: we need to replace everything
1788 after last slash (possibly empty) with LINK.
1790 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1791 our result should be "whatever/foo/qux/xyzzy". */
1792 bool need_explicit_slash = false;
1794 const char *start_insert;
1795 const char *last_slash = find_last_char (base, end, '/');
1798 /* No slash found at all. Replace what we have with LINK. */
1799 start_insert = base;
1801 else if (last_slash && last_slash >= base + 2
1802 && last_slash[-2] == ':' && last_slash[-1] == '/')
1804 /* example: http://host" */
1806 start_insert = end + 1;
1807 need_explicit_slash = true;
1811 /* example: "whatever/foo/bar" */
1813 start_insert = last_slash + 1;
1816 span = start_insert - base;
1817 merge = xmalloc (span + linklength + 1);
1819 memcpy (merge, base, span);
1820 if (need_explicit_slash)
1821 merge[span - 1] = '/';
1822 memcpy (merge + span, link, linklength);
1823 merge[span + linklength] = '\0';
1829 #define APPEND(p, s) do { \
1830 int len = strlen (s); \
1831 memcpy (p, s, len); \
1835 /* Use this instead of password when the actual password is supposed
1836 to be hidden. We intentionally use a generic string without giving
1837 away the number of characters in the password, like previous
1839 #define HIDDEN_PASSWORD "*password*"
1841 /* Recreate the URL string from the data in URL.
1843 If HIDE is true (as it is when we're calling this on a URL we plan
1844 to print, but not when calling it to canonicalize a URL for use
1845 within the program), password will be hidden. Unsafe characters in
1846 the URL will be quoted. */
1849 url_string (const struct url *url, enum url_auth_mode auth_mode)
1853 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1855 int scheme_port = supported_schemes[url->scheme].default_port;
1856 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1857 int fplen = full_path_length (url);
1859 bool brackets_around_host;
1861 assert (scheme_str != NULL);
1863 /* Make sure the user name and password are quoted. */
1866 if (auth_mode != URL_AUTH_HIDE)
1868 quoted_user = url_escape_allow_passthrough (url->user);
1871 if (auth_mode == URL_AUTH_HIDE_PASSWD)
1872 quoted_passwd = HIDDEN_PASSWORD;
1874 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1879 /* In the unlikely event that the host name contains non-printable
1880 characters, quote it for displaying to the user. */
1881 quoted_host = url_escape_allow_passthrough (url->host);
1883 /* Undo the quoting of colons that URL escaping performs. IPv6
1884 addresses may legally contain colons, and in that case must be
1885 placed in square brackets. */
1886 if (quoted_host != url->host)
1887 unescape_single_char (quoted_host, ':');
1888 brackets_around_host = strchr (quoted_host, ':') != NULL;
1890 size = (strlen (scheme_str)
1891 + strlen (quoted_host)
1892 + (brackets_around_host ? 2 : 0)
1895 if (url->port != scheme_port)
1896 size += 1 + numdigit (url->port);
1899 size += 1 + strlen (quoted_user);
1901 size += 1 + strlen (quoted_passwd);
1904 p = result = xmalloc (size);
1906 APPEND (p, scheme_str);
1909 APPEND (p, quoted_user);
1913 APPEND (p, quoted_passwd);
1918 if (brackets_around_host)
1920 APPEND (p, quoted_host);
1921 if (brackets_around_host)
1923 if (url->port != scheme_port)
1926 p = number_to_string (p, url->port);
1929 full_path_write (url, p);
1933 assert (p - result == size);
1935 if (quoted_user && quoted_user != url->user)
1936 xfree (quoted_user);
1937 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
1938 && quoted_passwd != url->passwd)
1939 xfree (quoted_passwd);
1940 if (quoted_host != url->host)
1941 xfree (quoted_host);
1946 /* Return true if scheme a is similar to scheme b.
1948 Schemes are similar if they are equal. If SSL is supported, schemes
1949 are also similar if one is http (SCHEME_HTTP) and the other is https
1952 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1957 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1958 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1965 getchar_from_escaped_string (const char *str, char *c)
1967 const char *p = str;
1969 assert (str && *str);
1974 if (!c_isxdigit(p[1]) || !c_isxdigit(p[2]))
1982 return 0; /* error: invalid string */
1984 *c = X2DIGITS_TO_NUM (p[1], p[2]);
1985 if (URL_RESERVED_CHAR(*c))
2003 are_urls_equal (const char *u1, const char *u2)
2014 && (pp = getchar_from_escaped_string (p, &ch1))
2015 && (qq = getchar_from_escaped_string (q, &ch2))
2016 && (c_tolower(ch1) == c_tolower(ch2)))
2022 return (*p == 0 && *q == 0 ? true : false);
2026 /* Debugging and testing support for path_simplify. */
2029 /* Debug: run path_simplify on PATH and return the result in a new
2030 string. Useful for calling from the debugger. */
2034 char *copy = xstrdup (path);
2035 path_simplify (copy);
2041 run_test (char *test, char *expected_result, enum url_scheme scheme,
2042 bool expected_change)
2044 char *test_copy = xstrdup (test);
2045 bool modified = path_simplify (scheme, test_copy);
2047 if (0 != strcmp (test_copy, expected_result))
2049 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2050 test, expected_result, test_copy);
2053 if (modified != expected_change)
2055 if (expected_change)
2056 printf ("Expected modification with path_simplify(\"%s\").\n",
2059 printf ("Expected no modification with path_simplify(\"%s\").\n",
2063 mu_assert ("", modified == expected_change);
2068 test_path_simplify (void)
2071 char *test, *result;
2072 enum url_scheme scheme;
2075 { "", "", SCHEME_HTTP, false },
2076 { ".", "", SCHEME_HTTP, true },
2077 { "./", "", SCHEME_HTTP, true },
2078 { "..", "", SCHEME_HTTP, true },
2079 { "../", "", SCHEME_HTTP, true },
2080 { "..", "..", SCHEME_FTP, false },
2081 { "../", "../", SCHEME_FTP, false },
2082 { "foo", "foo", SCHEME_HTTP, false },
2083 { "foo/bar", "foo/bar", SCHEME_HTTP, false },
2084 { "foo///bar", "foo///bar", SCHEME_HTTP, false },
2085 { "foo/.", "foo/", SCHEME_HTTP, true },
2086 { "foo/./", "foo/", SCHEME_HTTP, true },
2087 { "foo./", "foo./", SCHEME_HTTP, false },
2088 { "foo/../bar", "bar", SCHEME_HTTP, true },
2089 { "foo/../bar/", "bar/", SCHEME_HTTP, true },
2090 { "foo/bar/..", "foo/", SCHEME_HTTP, true },
2091 { "foo/bar/../x", "foo/x", SCHEME_HTTP, true },
2092 { "foo/bar/../x/", "foo/x/", SCHEME_HTTP, true },
2093 { "foo/..", "", SCHEME_HTTP, true },
2094 { "foo/../..", "", SCHEME_HTTP, true },
2095 { "foo/../../..", "", SCHEME_HTTP, true },
2096 { "foo/../../bar/../../baz", "baz", SCHEME_HTTP, true },
2097 { "foo/../..", "..", SCHEME_FTP, true },
2098 { "foo/../../..", "../..", SCHEME_FTP, true },
2099 { "foo/../../bar/../../baz", "../../baz", SCHEME_FTP, true },
2100 { "a/b/../../c", "c", SCHEME_HTTP, true },
2101 { "./a/../b", "b", SCHEME_HTTP, true }
2105 for (i = 0; i < countof (tests); i++)
2107 const char *message;
2108 char *test = tests[i].test;
2109 char *expected_result = tests[i].result;
2110 enum url_scheme scheme = tests[i].scheme;
2111 bool expected_change = tests[i].should_modify;
2112 message = run_test (test, expected_result, scheme, expected_change);
2113 if (message) return message;
2119 test_append_uri_pathel()
2126 char *expected_result;
2128 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2131 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2133 struct growable dest;
2134 const char *p = test_array[i].input;
2136 memset (&dest, 0, sizeof (dest));
2138 append_string (test_array[i].original_url, &dest);
2139 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2140 append_char ('\0', &dest);
2142 mu_assert ("test_append_uri_pathel: wrong result",
2143 strcmp (dest.base, test_array[i].expected_result) == 0);
2150 test_are_urls_equal()
2156 bool expected_result;
2158 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2159 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2160 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2161 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2162 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2163 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2166 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2168 mu_assert ("test_are_urls_equal: wrong result",
2169 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2175 #endif /* TESTING */