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 */
48 #endif /* def __VMS */
55 scm_disabled = 1, /* for https when OpenSSL fails to init. */
56 scm_has_params = 2, /* whether scheme has ;params */
57 scm_has_query = 4, /* whether scheme has ?query */
58 scm_has_fragment = 8 /* whether scheme has #fragment */
63 /* Short name of the scheme, such as "http" or "ftp". */
65 /* Leading string that identifies the scheme, such as "https://". */
66 const char *leading_string;
67 /* Default port of the scheme when none is specified. */
73 /* Supported schemes: */
74 static struct scheme_data supported_schemes[] =
76 { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment },
78 { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment },
80 { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment },
86 /* Forward declarations: */
88 static bool path_simplify (char *);
90 /* Support for escaping and unescaping of URL strings. */
92 /* Table of "reserved" and "unsafe" characters. Those terms are
93 rfc1738-speak, as such largely obsoleted by rfc2396 and later
94 specs, but the general idea remains.
96 A reserved character is the one that you can't decode without
97 changing the meaning of the URL. For example, you can't decode
98 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
99 path components is different. Non-reserved characters can be
100 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
101 unsafe characters are loosely based on rfc1738, plus "$" and ",",
102 as recommended by rfc2396, and minus "~", which is very frequently
103 used (and sometimes unrecognized as %7E by broken servers).
105 An unsafe character is the one that should be encoded when URLs are
106 placed in foreign environments. E.g. space and newline are unsafe
107 in HTTP contexts because HTTP uses them as separator and line
108 terminator, so they must be encoded to %20 and %0A respectively.
109 "*" is unsafe in shell context, etc.
111 We determine whether a character is unsafe through static table
112 lookup. This code assumes ASCII character set and 8-bit chars. */
115 /* rfc1738 reserved chars + "$" and ",". */
118 /* rfc1738 unsafe chars, plus non-printables. */
122 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
123 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
124 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
126 /* Shorthands for the table: */
127 #define R urlchr_reserved
128 #define U urlchr_unsafe
131 static const unsigned char urlchr_table[256] =
133 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
134 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
135 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
136 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
137 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
138 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
139 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
140 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
141 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
142 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
143 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
144 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
145 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
146 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
147 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
148 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
150 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
151 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
152 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
153 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
155 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
156 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
157 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
158 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
164 /* URL-unescape the string S.
166 This is done by transforming the sequences "%HH" to the character
167 represented by the hexadecimal digits HH. If % is not followed by
168 two hexadecimal digits, it is inserted literally.
170 The transformation is done in place. If you need the original
171 string intact, make a copy before calling this function. */
174 url_unescape (char *s)
176 char *t = s; /* t - tortoise */
177 char *h = s; /* h - hare */
189 /* Do nothing if '%' is not followed by two hex digits. */
190 if (!h[1] || !h[2] || !(c_isxdigit (h[1]) && c_isxdigit (h[2])))
192 c = X2DIGITS_TO_NUM (h[1], h[2]);
193 /* Don't unescape %00 because there is no way to insert it
194 into a C string without effectively truncating it. */
204 /* The core of url_escape_* functions. Escapes the characters that
205 match the provided mask in urlchr_table.
207 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
208 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
209 allocated string will be returned in all cases. */
212 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
219 for (p1 = s; *p1; p1++)
220 if (urlchr_test (*p1, mask))
221 addition += 2; /* Two more characters (hex digits) */
224 return allow_passthrough ? (char *)s : xstrdup (s);
226 newlen = (p1 - s) + addition;
227 newstr = xmalloc (newlen + 1);
233 /* Quote the characters that match the test mask. */
234 if (urlchr_test (*p1, mask))
236 unsigned char c = *p1++;
238 *p2++ = XNUM_TO_DIGIT (c >> 4);
239 *p2++ = XNUM_TO_DIGIT (c & 0xf);
244 assert (p2 - newstr == newlen);
250 /* URL-escape the unsafe characters (see urlchr_table) in a given
251 string, returning a freshly allocated string. */
254 url_escape (const char *s)
256 return url_escape_1 (s, urlchr_unsafe, false);
259 /* URL-escape the unsafe characters (see urlchr_table) in a given
260 string. If no characters are unsafe, S is returned. */
263 url_escape_allow_passthrough (const char *s)
265 return url_escape_1 (s, urlchr_unsafe, true);
268 /* Decide whether the char at position P needs to be encoded. (It is
269 not enough to pass a single char *P because the function may need
270 to inspect the surrounding context.)
272 Return true if the char should be escaped as %XX, false otherwise. */
275 char_needs_escaping (const char *p)
279 if (c_isxdigit (*(p + 1)) && c_isxdigit (*(p + 2)))
282 /* Garbled %.. sequence: encode `%'. */
285 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
291 /* Translate a %-escaped (but possibly non-conformant) input string S
292 into a %-escaped (and conformant) output string. If no characters
293 are encoded or decoded, return the same string S; otherwise, return
294 a freshly allocated string with the new contents.
296 After a URL has been run through this function, the protocols that
297 use `%' as the quote character can use the resulting string as-is,
298 while those that don't can use url_unescape to get to the intended
299 data. This function is stable: once the input is transformed,
300 further transformations of the result yield the same output.
302 Let's discuss why this function is needed.
304 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
305 a raw space character would mess up the HTTP request, it needs to
306 be quoted, like this:
308 GET /abc%20def HTTP/1.0
310 It would appear that the unsafe chars need to be quoted, for
311 example with url_escape. But what if we're requested to download
312 `abc%20def'? url_escape transforms "%" to "%25", which would leave
313 us with `abc%2520def'. This is incorrect -- since %-escapes are
314 part of URL syntax, "%20" is the correct way to denote a literal
315 space on the Wget command line. This leads to the conclusion that
316 in that case Wget should not call url_escape, but leave the `%20'
317 as is. This is clearly contradictory, but it only gets worse.
319 What if the requested URI is `abc%20 def'? If we call url_escape,
320 we end up with `/abc%2520%20def', which is almost certainly not
321 intended. If we don't call url_escape, we are left with the
322 embedded space and cannot complete the request. What the user
323 meant was for Wget to request `/abc%20%20def', and this is where
324 reencode_escapes kicks in.
326 Wget used to solve this by first decoding %-quotes, and then
327 encoding all the "unsafe" characters found in the resulting string.
328 This was wrong because it didn't preserve certain URL special
329 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
330 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
331 whether we considered `+' reserved (it is). One of these results
332 is inevitable because by the second step we would lose information
333 on whether the `+' was originally encoded or not. Both results
334 were wrong because in CGI parameters + means space, while %2B means
335 literal plus. reencode_escapes correctly translates the above to
336 "a%2B+b", i.e. returns the original string.
338 This function uses a modified version of the algorithm originally
339 proposed by Anon Sricharoenchai:
341 * Encode all "unsafe" characters, except those that are also
342 "reserved", to %XX. See urlchr_table for which characters are
345 * Encode the "%" characters not followed by two hex digits to
348 * Pass through all other characters and %XX escapes as-is. (Up to
349 Wget 1.10 this decoded %XX escapes corresponding to "safe"
350 characters, but that was obtrusive and broke some servers.)
354 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
356 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
360 "foo bar" -> "foo%20bar"
361 "foo%20bar" -> "foo%20bar"
362 "foo %20bar" -> "foo%20%20bar"
363 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
364 "foo%25%20bar" -> "foo%25%20bar"
365 "foo%2%20bar" -> "foo%252%20bar"
366 "foo+bar" -> "foo+bar" (plus is reserved!)
367 "foo%2b+bar" -> "foo%2b+bar" */
370 reencode_escapes (const char *s)
376 int encode_count = 0;
378 /* First pass: inspect the string to see if there's anything to do,
379 and to calculate the new length. */
380 for (p1 = s; *p1; p1++)
381 if (char_needs_escaping (p1))
385 /* The string is good as it is. */
386 return (char *) s; /* C const model sucks. */
389 /* Each encoding adds two characters (hex digits). */
390 newlen = oldlen + 2 * encode_count;
391 newstr = xmalloc (newlen + 1);
393 /* Second pass: copy the string to the destination address, encoding
394 chars when needed. */
399 if (char_needs_escaping (p1))
401 unsigned char c = *p1++;
403 *p2++ = XNUM_TO_DIGIT (c >> 4);
404 *p2++ = XNUM_TO_DIGIT (c & 0xf);
410 assert (p2 - newstr == newlen);
414 /* Returns the scheme type if the scheme is supported, or
415 SCHEME_INVALID if not. */
418 url_scheme (const char *url)
422 for (i = 0; supported_schemes[i].leading_string; i++)
423 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
424 strlen (supported_schemes[i].leading_string)))
426 if (!(supported_schemes[i].flags & scm_disabled))
427 return (enum url_scheme) i;
429 return SCHEME_INVALID;
432 return SCHEME_INVALID;
435 #define SCHEME_CHAR(ch) (c_isalnum (ch) || (ch) == '-' || (ch) == '+')
437 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
438 currently implemented, it returns true if URL begins with
442 url_has_scheme (const char *url)
446 /* The first char must be a scheme char. */
447 if (!*p || !SCHEME_CHAR (*p))
450 /* Followed by 0 or more scheme chars. */
451 while (*p && SCHEME_CHAR (*p))
453 /* Terminated by ':'. */
458 scheme_default_port (enum url_scheme scheme)
460 return supported_schemes[scheme].default_port;
464 scheme_disable (enum url_scheme scheme)
466 supported_schemes[scheme].flags |= scm_disabled;
469 /* Skip the username and password, if present in the URL. The
470 function should *not* be called with the complete URL, but with the
471 portion after the scheme.
473 If no username and password are found, return URL. */
476 url_skip_credentials (const char *url)
478 /* Look for '@' that comes before terminators, such as '/', '?',
480 const char *p = (const char *)strpbrk (url, "@/?#;");
486 /* Parse credentials contained in [BEG, END). The region is expected
487 to have come from a URL and is unescaped. */
490 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
496 return false; /* empty user name */
498 colon = memchr (beg, ':', end - beg);
500 return false; /* again empty user name */
504 *passwd = strdupdelim (colon + 1, end);
506 url_unescape (*passwd);
513 *user = strdupdelim (beg, userend);
514 url_unescape (*user);
518 /* Used by main.c: detect URLs written using the "shorthand" URL forms
519 originally popularized by Netscape and NcFTP. HTTP shorthands look
522 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
523 www.foo.com[:port] -> http://www.foo.com[:port]
525 FTP shorthands look like this:
527 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
528 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
530 If the URL needs not or cannot be rewritten, return NULL. */
533 rewrite_shorthand_url (const char *url)
538 if (url_scheme (url) != SCHEME_INVALID)
541 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
543 p = strpbrk (url, ":/");
547 /* If we're looking at "://", it means the URL uses a scheme we
548 don't support, which may include "https" when compiled without
549 SSL support. Don't bogusly rewrite such URLs. */
550 if (p && p[0] == ':' && p[1] == '/' && p[2] == '/')
555 /* Colon indicates ftp, as in foo.bar.com:path. Check for
556 special case of http port number ("localhost:10000"). */
557 int digits = strspn (p + 1, "0123456789");
558 if (digits && (p[1 + digits] == '/' || p[1 + digits] == '\0'))
561 /* Turn "foo.bar.com:path" to "ftp://foo.bar.com/path". */
562 ret = aprintf ("ftp://%s", url);
563 ret[6 + (p - url)] = '/';
568 /* Just prepend "http://" to URL. */
569 ret = aprintf ("http://%s", url);
574 static void split_path (const char *, char **, char **);
576 /* Like strpbrk, with the exception that it returns the pointer to the
577 terminating zero (end-of-string aka "eos") if no matching character
581 strpbrk_or_eos (const char *s, const char *accept)
583 char *p = strpbrk (s, accept);
585 p = strchr (s, '\0');
589 /* Turn STR into lowercase; return true if a character was actually
593 lowercase_str (char *str)
595 bool changed = false;
597 if (c_isupper (*str))
600 *str = c_tolower (*str);
606 init_seps (enum url_scheme scheme)
608 static char seps[8] = ":/";
610 int flags = supported_schemes[scheme].flags;
612 if (flags & scm_has_params)
614 if (flags & scm_has_query)
616 if (flags & scm_has_fragment)
622 static const char *parse_errors[] = {
623 #define PE_NO_ERROR 0
625 #define PE_UNSUPPORTED_SCHEME 1
626 N_("Unsupported scheme"),
627 #define PE_INVALID_HOST_NAME 2
628 N_("Invalid host name"),
629 #define PE_BAD_PORT_NUMBER 3
630 N_("Bad port number"),
631 #define PE_INVALID_USER_NAME 4
632 N_("Invalid user name"),
633 #define PE_UNTERMINATED_IPV6_ADDRESS 5
634 N_("Unterminated IPv6 numeric address"),
635 #define PE_IPV6_NOT_SUPPORTED 6
636 N_("IPv6 addresses not supported"),
637 #define PE_INVALID_IPV6_ADDRESS 7
638 N_("Invalid IPv6 numeric address")
643 Return a new struct url if successful, NULL on error. In case of
644 error, and if ERROR is not NULL, also set *ERROR to the appropriate
647 url_parse (const char *url, int *error)
651 bool path_modified, host_modified;
653 enum url_scheme scheme;
656 const char *uname_b, *uname_e;
657 const char *host_b, *host_e;
658 const char *path_b, *path_e;
659 const char *params_b, *params_e;
660 const char *query_b, *query_e;
661 const char *fragment_b, *fragment_e;
664 char *user = NULL, *passwd = NULL;
666 char *url_encoded = NULL;
670 scheme = url_scheme (url);
671 if (scheme == SCHEME_INVALID)
673 error_code = PE_UNSUPPORTED_SCHEME;
677 url_encoded = reencode_escapes (url);
680 p += strlen (supported_schemes[scheme].leading_string);
682 p = url_skip_credentials (p);
685 /* scheme://user:pass@host[:port]... */
688 /* We attempt to break down the URL into the components path,
689 params, query, and fragment. They are ordered like this:
691 scheme://host[:port][/path][;params][?query][#fragment] */
693 path_b = path_e = NULL;
694 params_b = params_e = NULL;
695 query_b = query_e = NULL;
696 fragment_b = fragment_e = NULL;
698 /* Initialize separators for optional parts of URL, depending on the
699 scheme. For example, FTP has params, and HTTP and HTTPS have
700 query string and fragment. */
701 seps = init_seps (scheme);
707 /* Handle IPv6 address inside square brackets. Ideally we'd
708 just look for the terminating ']', but rfc2732 mandates
709 rejecting invalid IPv6 addresses. */
711 /* The address begins after '['. */
713 host_e = strchr (host_b, ']');
717 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
722 /* Check if the IPv6 address is valid. */
723 if (!is_valid_ipv6_address(host_b, host_e))
725 error_code = PE_INVALID_IPV6_ADDRESS;
729 /* Continue parsing after the closing ']'. */
732 error_code = PE_IPV6_NOT_SUPPORTED;
736 /* The closing bracket must be followed by a separator or by the
738 /* http://[::1]... */
740 if (!strchr (seps, *p))
742 /* Trailing garbage after []-delimited IPv6 address. */
743 error_code = PE_INVALID_HOST_NAME;
749 p = strpbrk_or_eos (p, seps);
752 ++seps; /* advance to '/' */
754 if (host_b == host_e)
756 error_code = PE_INVALID_HOST_NAME;
760 port = scheme_default_port (scheme);
763 const char *port_b, *port_e, *pp;
765 /* scheme://host:port/tralala */
769 p = strpbrk_or_eos (p, seps);
772 /* Allow empty port, as per rfc2396. */
773 if (port_b != port_e)
774 for (port = 0, pp = port_b; pp < port_e; pp++)
776 if (!c_isdigit (*pp))
778 /* http://host:12randomgarbage/blah */
780 error_code = PE_BAD_PORT_NUMBER;
783 port = 10 * port + (*pp - '0');
784 /* Check for too large port numbers here, before we have
785 a chance to overflow on bogus port values. */
788 error_code = PE_BAD_PORT_NUMBER;
793 /* Advance to the first separator *after* '/' (either ';' or '?',
794 depending on the scheme). */
797 /* Get the optional parts of URL, each part being delimited by
798 current location and the position of the next separator. */
799 #define GET_URL_PART(sepchar, var) do { \
801 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
805 GET_URL_PART ('/', path);
806 if (supported_schemes[scheme].flags & scm_has_params)
807 GET_URL_PART (';', params);
808 if (supported_schemes[scheme].flags & scm_has_query)
809 GET_URL_PART ('?', query);
810 if (supported_schemes[scheme].flags & scm_has_fragment)
811 GET_URL_PART ('#', fragment);
816 if (uname_b != uname_e)
818 /* http://user:pass@host */
820 /* uname_b uname_e */
821 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
823 error_code = PE_INVALID_USER_NAME;
828 u = xnew0 (struct url);
830 u->host = strdupdelim (host_b, host_e);
835 u->path = strdupdelim (path_b, path_e);
836 path_modified = path_simplify (u->path);
837 split_path (u->path, &u->dir, &u->file);
839 host_modified = lowercase_str (u->host);
841 /* Decode %HH sequences in host name. This is important not so much
842 to support %HH sequences in host names (which other browser
843 don't), but to support binary characters (which will have been
844 converted to %HH by reencode_escapes). */
845 if (strchr (u->host, '%'))
847 url_unescape (u->host);
848 host_modified = true;
852 u->params = strdupdelim (params_b, params_e);
854 u->query = strdupdelim (query_b, query_e);
856 u->fragment = strdupdelim (fragment_b, fragment_e);
858 if (path_modified || u->fragment || host_modified || path_b == path_e)
860 /* If we suspect that a transformation has rendered what
861 url_string might return different from URL_ENCODED, rebuild
862 u->url using url_string. */
863 u->url = url_string (u, URL_AUTH_SHOW);
865 if (url_encoded != url)
866 xfree ((char *) url_encoded);
870 if (url_encoded == url)
871 u->url = xstrdup (url);
873 u->url = url_encoded;
879 /* Cleanup in case of error: */
880 if (url_encoded && url_encoded != url)
883 /* Transmit the error code to the caller, if the caller wants to
890 /* Return the error message string from ERROR_CODE, which should have
891 been retrieved from url_parse. The error message is translated. */
894 url_error (int error_code)
896 assert (error_code >= 0 && error_code < countof (parse_errors));
897 return _(parse_errors[error_code]);
900 /* Split PATH into DIR and FILE. PATH comes from the URL and is
901 expected to be URL-escaped.
903 The path is split into directory (the part up to the last slash)
904 and file (the part after the last slash), which are subsequently
908 "foo/bar/baz" "foo/bar" "baz"
909 "foo/bar/" "foo/bar" ""
911 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
913 DIR and FILE are freshly allocated. */
916 split_path (const char *path, char **dir, char **file)
918 char *last_slash = strrchr (path, '/');
922 *file = xstrdup (path);
926 *dir = strdupdelim (path, last_slash);
927 *file = xstrdup (last_slash + 1);
930 url_unescape (*file);
933 /* Note: URL's "full path" is the path with the query string and
934 params appended. The "fragment" (#foo) is intentionally ignored,
935 but that might be changed. For example, if the original URL was
936 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
937 the full path will be "/foo/bar/baz;bullshit?querystring". */
939 /* Return the length of the full path, without the terminating
943 full_path_length (const struct url *url)
947 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
958 /* Write out the full path. */
961 full_path_write (const struct url *url, char *where)
963 #define FROB(el, chr) do { \
964 char *f_el = url->el; \
966 int l = strlen (f_el); \
968 memcpy (where, f_el, l); \
980 /* Public function for getting the "full path". E.g. if u->path is
981 "foo/bar" and u->query is "param=value", full_path will be
982 "/foo/bar?param=value". */
985 url_full_path (const struct url *url)
987 int length = full_path_length (url);
988 char *full_path = xmalloc (length + 1);
990 full_path_write (url, full_path);
991 full_path[length] = '\0';
996 /* Unescape CHR in an otherwise escaped STR. Used to selectively
997 escaping of certain characters, such as "/" and ":". Returns a
998 count of unescaped chars. */
1001 unescape_single_char (char *str, char chr)
1003 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1004 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1005 char *h = str; /* hare */
1006 char *t = str; /* tortoise */
1007 for (; *h; h++, t++)
1009 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1020 /* Escape unsafe and reserved characters, except for the slash
1024 url_escape_dir (const char *dir)
1026 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1030 unescape_single_char (newdir, '/');
1034 /* Sync u->path and u->url with u->dir and u->file. Called after
1035 u->file or u->dir have been changed, typically by the FTP code. */
1038 sync_path (struct url *u)
1040 char *newpath, *efile, *edir;
1044 /* u->dir and u->file are not escaped. URL-escape them before
1045 reassembling them into u->path. That way, if they contain
1046 separators like '?' or even if u->file contains slashes, the
1047 path will be correctly assembled. (u->file can contain slashes
1048 if the URL specifies it with %2f, or if an FTP server returns
1050 edir = url_escape_dir (u->dir);
1051 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1054 newpath = xstrdup (efile);
1057 int dirlen = strlen (edir);
1058 int filelen = strlen (efile);
1060 /* Copy "DIR/FILE" to newpath. */
1061 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1062 memcpy (p, edir, dirlen);
1065 memcpy (p, efile, filelen);
1074 if (efile != u->file)
1077 /* Regenerate u->url as well. */
1079 u->url = url_string (u, URL_AUTH_SHOW);
1082 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1083 This way we can sync u->path and u->url when they get changed. */
1086 url_set_dir (struct url *url, const char *newdir)
1089 url->dir = xstrdup (newdir);
1094 url_set_file (struct url *url, const char *newfile)
1097 url->file = xstrdup (newfile);
1102 url_free (struct url *url)
1108 xfree_null (url->params);
1109 xfree_null (url->query);
1110 xfree_null (url->fragment);
1111 xfree_null (url->user);
1112 xfree_null (url->passwd);
1120 /* Create all the necessary directories for PATH (a file). Calls
1121 make_directory internally. */
1123 mkalldirs (const char *path)
1130 p = path + strlen (path);
1131 for (; *p != '/' && p != path; p--)
1134 /* Don't create if it's just a file. */
1135 if ((p == path) && (*p != '/'))
1137 t = strdupdelim (path, p);
1139 /* Check whether the directory exists. */
1140 if ((stat (t, &st) == 0))
1142 if (S_ISDIR (st.st_mode))
1149 /* If the dir exists as a file name, remove it first. This
1150 is *only* for Wget to work with buggy old CERN http
1151 servers. Here is the scenario: When Wget tries to
1152 retrieve a directory without a slash, e.g.
1153 http://foo/bar (bar being a directory), CERN server will
1154 not redirect it too http://foo/bar/ -- it will generate a
1155 directory listing containing links to bar/file1,
1156 bar/file2, etc. Wget will lose because it saves this
1157 HTML listing to a file `bar', so it cannot create the
1158 directory. To work around this, if the file of the same
1159 name exists, we just remove it and create the directory
1161 DEBUGP (("Removing %s because of directory danger!\n", t));
1165 res = make_directory (t);
1167 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1172 /* Functions for constructing the file name out of URL components. */
1174 /* A growable string structure, used by url_file_name and friends.
1175 This should perhaps be moved to utils.c.
1177 The idea is to have a convenient and efficient way to construct a
1178 string by having various functions append data to it. Instead of
1179 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1180 functions in questions, we pass the pointer to this struct. */
1188 /* Ensure that the string can accept APPEND_COUNT more characters past
1189 the current TAIL position. If necessary, this will grow the string
1190 and update its allocated size. If the string is already large
1191 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1192 #define GROW(g, append_size) do { \
1193 struct growable *G_ = g; \
1194 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1197 /* Return the tail position of the string. */
1198 #define TAIL(r) ((r)->base + (r)->tail)
1200 /* Move the tail position by APPEND_COUNT characters. */
1201 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1203 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1207 append_string (const char *str, struct growable *dest)
1209 int l = strlen (str);
1211 memcpy (TAIL (dest), str, l);
1212 TAIL_INCR (dest, l);
1215 /* Append CH to DEST. For example, append_char (0, DEST)
1216 zero-terminates DEST. */
1219 append_char (char ch, struct growable *dest)
1223 TAIL_INCR (dest, 1);
1227 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1228 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1229 filechr_control = 4 /* a control character, e.g. 0-31 */
1232 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1234 /* Shorthands for the table: */
1235 #define U filechr_not_unix
1236 #define W filechr_not_windows
1237 #define C filechr_control
1242 /* Table of characters unsafe under various conditions (see above).
1244 Arguably we could also claim `%' to be unsafe, since we use it as
1245 the escape character. If we ever want to be able to reliably
1246 translate file name back to URL, this would become important
1247 crucial. Right now, it's better to be minimal in escaping. */
1249 static const unsigned char filechr_table[256] =
1251 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1252 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1253 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1254 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1255 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1256 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1257 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1258 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1259 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1260 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1261 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1262 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1263 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1264 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1265 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1266 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1268 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1269 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
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,
1273 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1274 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1275 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1276 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1284 /* FN_PORT_SEP is the separator between host and port in file names
1285 for non-standard port numbers. On Unix this is normally ':', as in
1286 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1287 because Windows can't handle ':' in file names. */
1288 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1290 /* FN_QUERY_SEP is the separator between the file name and the URL
1291 query, normally '?'. Since Windows cannot handle '?' as part of
1292 file name, we use '@' instead there. */
1293 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1295 /* Quote path element, characters in [b, e), as file name, and append
1296 the quoted string to DEST. Each character is quoted as per
1297 file_unsafe_char and the corresponding table.
1299 If ESCAPED is true, the path element is considered to be
1300 URL-escaped and will be unescaped prior to inspection. */
1303 append_uri_pathel (const char *b, const char *e, bool escaped,
1304 struct growable *dest)
1310 if (opt.restrict_files_os == restrict_unix)
1311 mask = filechr_not_unix;
1313 mask = filechr_not_windows;
1314 if (opt.restrict_files_ctrl)
1315 mask |= filechr_control;
1317 /* Copy [b, e) to PATHEL and URL-unescape it. */
1321 BOUNDED_TO_ALLOCA (b, e, unescaped);
1322 url_unescape (unescaped);
1324 e = unescaped + strlen (unescaped);
1327 /* Defang ".." when found as component of path. Remember that path
1328 comes from the URL and might contain malicious input. */
1329 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1335 /* Walk the PATHEL string and check how many characters we'll need
1338 for (p = b; p < e; p++)
1339 if (FILE_CHAR_TEST (*p, mask))
1342 /* Calculate the length of the output string. e-b is the input
1343 string length. Each quoted char introduces two additional
1344 characters in the string, hence 2*quoted. */
1345 outlen = (e - b) + (2 * quoted);
1346 GROW (dest, outlen);
1350 /* If there's nothing to quote, we can simply append the string
1351 without processing it again. */
1352 memcpy (TAIL (dest), b, outlen);
1356 char *q = TAIL (dest);
1357 for (p = b; p < e; p++)
1359 if (!FILE_CHAR_TEST (*p, mask))
1363 unsigned char ch = *p;
1365 *q++ = XNUM_TO_DIGIT (ch >> 4);
1366 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1369 assert (q - TAIL (dest) == outlen);
1372 /* Perform inline case transformation if required. */
1373 if (opt.restrict_files_case == restrict_lowercase
1374 || opt.restrict_files_case == restrict_uppercase)
1377 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1379 if (opt.restrict_files_case == restrict_lowercase)
1380 *q = c_tolower (*q);
1382 *q = c_toupper (*q);
1386 TAIL_INCR (dest, outlen);
1389 /* Append to DEST the directory structure that corresponds the
1390 directory part of URL's path. For example, if the URL is
1391 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1393 Each path element ("dir1" and "dir2" in the above example) is
1394 examined, url-unescaped, and re-escaped as file name element.
1396 Additionally, it cuts as many directories from the path as
1397 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1398 will produce "bar" for the above example. For 2 or more, it will
1401 Each component of the path is quoted for use as file name. */
1404 append_dir_structure (const struct url *u, struct growable *dest)
1406 char *pathel, *next;
1407 int cut = opt.cut_dirs;
1409 /* Go through the path components, de-URL-quote them, and quote them
1410 (if necessary) as file names. */
1413 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1418 /* Ignore empty pathels. */
1422 append_char ('/', dest);
1423 append_uri_pathel (pathel, next, true, dest);
1427 /* Return a unique file name that matches the given URL as good as
1428 possible. Does not create directories on the file system. */
1431 url_file_name (const struct url *u)
1433 struct growable fnres; /* stands for "file name result" */
1435 const char *u_file, *u_query;
1436 char *fname, *unique;
1442 /* Start with the directory prefix, if specified. */
1444 append_string (opt.dir_prefix, &fnres);
1446 /* If "dirstruct" is turned on (typically the case with -r), add
1447 the host and port (unless those have been turned off) and
1448 directory structure. */
1451 if (opt.protocol_directories)
1454 append_char ('/', &fnres);
1455 append_string (supported_schemes[u->scheme].name, &fnres);
1457 if (opt.add_hostdir)
1460 append_char ('/', &fnres);
1461 if (0 != strcmp (u->host, ".."))
1462 append_string (u->host, &fnres);
1464 /* Host name can come from the network; malicious DNS may
1465 allow ".." to be resolved, causing us to write to
1466 "../<file>". Defang such host names. */
1467 append_string ("%2E%2E", &fnres);
1468 if (u->port != scheme_default_port (u->scheme))
1471 number_to_string (portstr, u->port);
1472 append_char (FN_PORT_SEP, &fnres);
1473 append_string (portstr, &fnres);
1477 append_dir_structure (u, &fnres);
1480 /* Add the file name. */
1482 append_char ('/', &fnres);
1483 u_file = *u->file ? u->file : "index.html";
1484 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1486 /* Append "?query" to the file name. */
1487 u_query = u->query && *u->query ? u->query : NULL;
1490 append_char (FN_QUERY_SEP, &fnres);
1491 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1494 /* Zero-terminate the file name. */
1495 append_char ('\0', &fnres);
1499 /* Check the cases in which the unique extensions are not used:
1500 1) Clobbering is turned off (-nc).
1501 2) Retrieval with regetting.
1502 3) Timestamping is used.
1503 4) Hierarchy is built.
1505 The exception is the case when file does exist and is a
1506 directory (see `mkalldirs' for explanation). */
1508 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1509 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1515 unique = unique_name (fname, true);
1516 if (unique != fname)
1520 /* On VMS, alter the name as required. */
1525 unique2 = ods_conform( unique);
1526 if (unique2 != unique)
1532 #endif /* def __VMS */
1537 /* Resolve "." and ".." elements of PATH by destructively modifying
1538 PATH and return true if PATH has been modified, false otherwise.
1540 The algorithm is in spirit similar to the one described in rfc1808,
1541 although implemented differently, in one pass. To recap, path
1542 elements containing only "." are removed, and ".." is taken to mean
1543 "back up one element". Single leading and trailing slashes are
1546 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1547 test examples are provided below. If you change anything in this
1548 function, run test_path_simplify to make sure you haven't broken a
1552 path_simplify (char *path)
1554 char *h = path; /* hare */
1555 char *t = path; /* tortoise */
1556 char *end = strchr (path, '\0');
1560 /* Hare should be at the beginning of a path element. */
1562 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1567 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1569 /* Handle "../" by retreating the tortoise by one path
1570 element -- but not past beggining. */
1573 /* Move backwards until T hits the beginning of the
1574 previous path element or the beginning of path. */
1575 for (--t; t > path && t[-1] != '/'; t--)
1582 /* A regular path element. If H hasn't advanced past T,
1583 simply skip to the next path element. Otherwise, copy
1584 the path element until the next slash. */
1587 /* Skip the path element, including the slash. */
1588 while (h < end && *h != '/')
1595 /* Copy the path element, including the final slash. */
1596 while (h < end && *h != '/')
1610 /* Return the length of URL's path. Path is considered to be
1611 terminated by one or more of the ?query or ;params or #fragment,
1612 depending on the scheme. */
1615 path_end (const char *url)
1617 enum url_scheme scheme = url_scheme (url);
1619 if (scheme == SCHEME_INVALID)
1620 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1621 /* +2 to ignore the first two separators ':' and '/' */
1622 seps = init_seps (scheme) + 2;
1623 return strpbrk_or_eos (url, seps);
1626 /* Find the last occurrence of character C in the range [b, e), or
1627 NULL, if none are present. */
1628 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1630 /* Merge BASE with LINK and return the resulting URI.
1632 Either of the URIs may be absolute or relative, complete with the
1633 host name, or path only. This tries to reasonably handle all
1634 foreseeable cases. It only employs minimal URL parsing, without
1635 knowledge of the specifics of schemes.
1637 I briefly considered making this function call path_simplify after
1638 the merging process, as rfc1738 seems to suggest. This is a bad
1639 idea for several reasons: 1) it complexifies the code, and 2)
1640 url_parse has to simplify path anyway, so it's wasteful to boot. */
1643 uri_merge (const char *base, const char *link)
1649 if (url_has_scheme (link))
1650 return xstrdup (link);
1652 /* We may not examine BASE past END. */
1653 end = path_end (base);
1654 linklength = strlen (link);
1658 /* Empty LINK points back to BASE, query string and all. */
1659 return xstrdup (base);
1661 else if (*link == '?')
1663 /* LINK points to the same location, but changes the query
1664 string. Examples: */
1665 /* uri_merge("path", "?new") -> "path?new" */
1666 /* uri_merge("path?foo", "?new") -> "path?new" */
1667 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1668 /* uri_merge("path#foo", "?new") -> "path?new" */
1669 int baselength = end - base;
1670 merge = xmalloc (baselength + linklength + 1);
1671 memcpy (merge, base, baselength);
1672 memcpy (merge + baselength, link, linklength);
1673 merge[baselength + linklength] = '\0';
1675 else if (*link == '#')
1677 /* uri_merge("path", "#new") -> "path#new" */
1678 /* uri_merge("path#foo", "#new") -> "path#new" */
1679 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1680 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1682 const char *end1 = strchr (base, '#');
1684 end1 = base + strlen (base);
1685 baselength = end1 - base;
1686 merge = xmalloc (baselength + linklength + 1);
1687 memcpy (merge, base, baselength);
1688 memcpy (merge + baselength, link, linklength);
1689 merge[baselength + linklength] = '\0';
1691 else if (*link == '/' && *(link + 1) == '/')
1693 /* LINK begins with "//" and so is a net path: we need to
1694 replace everything after (and including) the double slash
1697 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1698 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1699 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1703 const char *start_insert;
1705 /* Look for first slash. */
1706 slash = memchr (base, '/', end - base);
1707 /* If found slash and it is a double slash, then replace
1708 from this point, else default to replacing from the
1710 if (slash && *(slash + 1) == '/')
1711 start_insert = slash;
1713 start_insert = base;
1715 span = start_insert - base;
1716 merge = xmalloc (span + linklength + 1);
1718 memcpy (merge, base, span);
1719 memcpy (merge + span, link, linklength);
1720 merge[span + linklength] = '\0';
1722 else if (*link == '/')
1724 /* LINK is an absolute path: we need to replace everything
1725 after (and including) the FIRST slash with LINK.
1727 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1728 "/qux/xyzzy", our result should be
1729 "http://host/qux/xyzzy". */
1732 const char *start_insert = NULL; /* for gcc to shut up. */
1733 const char *pos = base;
1734 bool seen_slash_slash = false;
1735 /* We're looking for the first slash, but want to ignore
1738 slash = memchr (pos, '/', end - pos);
1739 if (slash && !seen_slash_slash)
1740 if (*(slash + 1) == '/')
1743 seen_slash_slash = true;
1747 /* At this point, SLASH is the location of the first / after
1748 "//", or the first slash altogether. START_INSERT is the
1749 pointer to the location where LINK will be inserted. When
1750 examining the last two examples, keep in mind that LINK
1753 if (!slash && !seen_slash_slash)
1754 /* example: "foo" */
1756 start_insert = base;
1757 else if (!slash && seen_slash_slash)
1758 /* example: "http://foo" */
1761 else if (slash && !seen_slash_slash)
1762 /* example: "foo/bar" */
1764 start_insert = base;
1765 else if (slash && seen_slash_slash)
1766 /* example: "http://something/" */
1768 start_insert = slash;
1770 span = start_insert - base;
1771 merge = xmalloc (span + linklength + 1);
1773 memcpy (merge, base, span);
1774 memcpy (merge + span, link, linklength);
1775 merge[span + linklength] = '\0';
1779 /* LINK is a relative URL: we need to replace everything
1780 after last slash (possibly empty) with LINK.
1782 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1783 our result should be "whatever/foo/qux/xyzzy". */
1784 bool need_explicit_slash = false;
1786 const char *start_insert;
1787 const char *last_slash = find_last_char (base, end, '/');
1790 /* No slash found at all. Replace what we have with LINK. */
1791 start_insert = base;
1793 else if (last_slash && last_slash >= base + 2
1794 && last_slash[-2] == ':' && last_slash[-1] == '/')
1796 /* example: http://host" */
1798 start_insert = end + 1;
1799 need_explicit_slash = true;
1803 /* example: "whatever/foo/bar" */
1805 start_insert = last_slash + 1;
1808 span = start_insert - base;
1809 merge = xmalloc (span + linklength + 1);
1811 memcpy (merge, base, span);
1812 if (need_explicit_slash)
1813 merge[span - 1] = '/';
1814 memcpy (merge + span, link, linklength);
1815 merge[span + linklength] = '\0';
1821 #define APPEND(p, s) do { \
1822 int len = strlen (s); \
1823 memcpy (p, s, len); \
1827 /* Use this instead of password when the actual password is supposed
1828 to be hidden. We intentionally use a generic string without giving
1829 away the number of characters in the password, like previous
1831 #define HIDDEN_PASSWORD "*password*"
1833 /* Recreate the URL string from the data in URL.
1835 If HIDE is true (as it is when we're calling this on a URL we plan
1836 to print, but not when calling it to canonicalize a URL for use
1837 within the program), password will be hidden. Unsafe characters in
1838 the URL will be quoted. */
1841 url_string (const struct url *url, enum url_auth_mode auth_mode)
1845 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1847 int scheme_port = supported_schemes[url->scheme].default_port;
1848 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1849 int fplen = full_path_length (url);
1851 bool brackets_around_host;
1853 assert (scheme_str != NULL);
1855 /* Make sure the user name and password are quoted. */
1858 if (auth_mode != URL_AUTH_HIDE)
1860 quoted_user = url_escape_allow_passthrough (url->user);
1863 if (auth_mode == URL_AUTH_HIDE_PASSWD)
1864 quoted_passwd = HIDDEN_PASSWORD;
1866 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1871 /* In the unlikely event that the host name contains non-printable
1872 characters, quote it for displaying to the user. */
1873 quoted_host = url_escape_allow_passthrough (url->host);
1875 /* Undo the quoting of colons that URL escaping performs. IPv6
1876 addresses may legally contain colons, and in that case must be
1877 placed in square brackets. */
1878 if (quoted_host != url->host)
1879 unescape_single_char (quoted_host, ':');
1880 brackets_around_host = strchr (quoted_host, ':') != NULL;
1882 size = (strlen (scheme_str)
1883 + strlen (quoted_host)
1884 + (brackets_around_host ? 2 : 0)
1887 if (url->port != scheme_port)
1888 size += 1 + numdigit (url->port);
1891 size += 1 + strlen (quoted_user);
1893 size += 1 + strlen (quoted_passwd);
1896 p = result = xmalloc (size);
1898 APPEND (p, scheme_str);
1901 APPEND (p, quoted_user);
1905 APPEND (p, quoted_passwd);
1910 if (brackets_around_host)
1912 APPEND (p, quoted_host);
1913 if (brackets_around_host)
1915 if (url->port != scheme_port)
1918 p = number_to_string (p, url->port);
1921 full_path_write (url, p);
1925 assert (p - result == size);
1927 if (quoted_user && quoted_user != url->user)
1928 xfree (quoted_user);
1929 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
1930 && quoted_passwd != url->passwd)
1931 xfree (quoted_passwd);
1932 if (quoted_host != url->host)
1933 xfree (quoted_host);
1938 /* Return true if scheme a is similar to scheme b.
1940 Schemes are similar if they are equal. If SSL is supported, schemes
1941 are also similar if one is http (SCHEME_HTTP) and the other is https
1944 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1949 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1950 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1957 getchar_from_escaped_string (const char *str, char *c)
1959 const char *p = str;
1961 assert (str && *str);
1966 if (!c_isxdigit(p[1]) || !c_isxdigit(p[2]))
1974 return 0; /* error: invalid string */
1976 *c = X2DIGITS_TO_NUM (p[1], p[2]);
1977 if (URL_RESERVED_CHAR(*c))
1995 are_urls_equal (const char *u1, const char *u2)
2006 && (pp = getchar_from_escaped_string (p, &ch1))
2007 && (qq = getchar_from_escaped_string (q, &ch2))
2008 && (c_tolower(ch1) == c_tolower(ch2)))
2014 return (*p == 0 && *q == 0 ? true : false);
2018 /* Debugging and testing support for path_simplify. */
2020 /* Debug: run path_simplify on PATH and return the result in a new
2021 string. Useful for calling from the debugger. */
2025 char *copy = xstrdup (path);
2026 path_simplify (copy);
2031 run_test (char *test, char *expected_result, bool expected_change)
2033 char *test_copy = xstrdup (test);
2034 bool modified = path_simplify (test_copy);
2036 if (0 != strcmp (test_copy, expected_result))
2038 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2039 test, expected_result, test_copy);
2041 if (modified != expected_change)
2043 if (expected_change)
2044 printf ("Expected modification with path_simplify(\"%s\").\n",
2047 printf ("Expected no modification with path_simplify(\"%s\").\n",
2054 test_path_simplify (void)
2057 char *test, *result;
2064 { "../", "", true },
2065 { "foo", "foo", false },
2066 { "foo/bar", "foo/bar", false },
2067 { "foo///bar", "foo///bar", false },
2068 { "foo/.", "foo/", true },
2069 { "foo/./", "foo/", true },
2070 { "foo./", "foo./", false },
2071 { "foo/../bar", "bar", true },
2072 { "foo/../bar/", "bar/", true },
2073 { "foo/bar/..", "foo/", true },
2074 { "foo/bar/../x", "foo/x", true },
2075 { "foo/bar/../x/", "foo/x/", true },
2076 { "foo/..", "", true },
2077 { "foo/../..", "", true },
2078 { "foo/../../..", "", true },
2079 { "foo/../../bar/../../baz", "baz", true },
2080 { "a/b/../../c", "c", true },
2081 { "./a/../b", "b", true }
2085 for (i = 0; i < countof (tests); i++)
2087 char *test = tests[i].test;
2088 char *expected_result = tests[i].result;
2089 bool expected_change = tests[i].should_modify;
2090 run_test (test, expected_result, expected_change);
2098 test_append_uri_pathel()
2105 char *expected_result;
2107 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2110 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2112 struct growable dest;
2113 const char *p = test_array[i].input;
2115 memset (&dest, 0, sizeof (dest));
2117 append_string (test_array[i].original_url, &dest);
2118 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2119 append_char ('\0', &dest);
2121 mu_assert ("test_append_uri_pathel: wrong result",
2122 strcmp (dest.base, test_array[i].expected_result) == 0);
2129 test_are_urls_equal()
2135 bool expected_result;
2137 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2138 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2139 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2140 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2141 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2142 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2145 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2147 mu_assert ("test_are_urls_equal: wrong result",
2148 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2154 #endif /* TESTING */