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. */
31 #define USE_GNULIB_ALLOC
46 #include "host.h" /* for is_valid_ipv6_address */
53 scm_disabled = 1, /* for https when OpenSSL fails to init. */
54 scm_has_params = 2, /* whether scheme has ;params */
55 scm_has_query = 4, /* whether scheme has ?query */
56 scm_has_fragment = 8 /* whether scheme has #fragment */
61 /* Short name of the scheme, such as "http" or "ftp". */
63 /* Leading string that identifies the scheme, such as "https://". */
64 const char *leading_string;
65 /* Default port of the scheme when none is specified. */
71 /* Supported schemes: */
72 static struct scheme_data supported_schemes[] =
74 { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment },
76 { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment },
78 { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment },
84 /* Forward declarations: */
86 static bool path_simplify (char *);
88 /* Support for escaping and unescaping of URL strings. */
90 /* Table of "reserved" and "unsafe" characters. Those terms are
91 rfc1738-speak, as such largely obsoleted by rfc2396 and later
92 specs, but the general idea remains.
94 A reserved character is the one that you can't decode without
95 changing the meaning of the URL. For example, you can't decode
96 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
97 path components is different. Non-reserved characters can be
98 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
99 unsafe characters are loosely based on rfc1738, plus "$" and ",",
100 as recommended by rfc2396, and minus "~", which is very frequently
101 used (and sometimes unrecognized as %7E by broken servers).
103 An unsafe character is the one that should be encoded when URLs are
104 placed in foreign environments. E.g. space and newline are unsafe
105 in HTTP contexts because HTTP uses them as separator and line
106 terminator, so they must be encoded to %20 and %0A respectively.
107 "*" is unsafe in shell context, etc.
109 We determine whether a character is unsafe through static table
110 lookup. This code assumes ASCII character set and 8-bit chars. */
113 /* rfc1738 reserved chars + "$" and ",". */
116 /* rfc1738 unsafe chars, plus non-printables. */
120 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
121 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
122 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
124 /* Shorthands for the table: */
125 #define R urlchr_reserved
126 #define U urlchr_unsafe
129 static const unsigned char urlchr_table[256] =
131 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
132 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
133 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
134 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
135 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
136 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
137 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
138 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
139 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
140 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
141 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
142 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
143 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
144 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
145 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
146 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
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,
151 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,
156 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
162 /* URL-unescape the string S.
164 This is done by transforming the sequences "%HH" to the character
165 represented by the hexadecimal digits HH. If % is not followed by
166 two hexadecimal digits, it is inserted literally.
168 The transformation is done in place. If you need the original
169 string intact, make a copy before calling this function. */
172 url_unescape (char *s)
174 char *t = s; /* t - tortoise */
175 char *h = s; /* h - hare */
187 /* Do nothing if '%' is not followed by two hex digits. */
188 if (!h[1] || !h[2] || !(c_isxdigit (h[1]) && c_isxdigit (h[2])))
190 c = X2DIGITS_TO_NUM (h[1], h[2]);
191 /* Don't unescape %00 because there is no way to insert it
192 into a C string without effectively truncating it. */
202 /* The core of url_escape_* functions. Escapes the characters that
203 match the provided mask in urlchr_table.
205 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
206 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
207 allocated string will be returned in all cases. */
210 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
217 for (p1 = s; *p1; p1++)
218 if (urlchr_test (*p1, mask))
219 addition += 2; /* Two more characters (hex digits) */
222 return allow_passthrough ? (char *)s : xstrdup (s);
224 newlen = (p1 - s) + addition;
225 newstr = xmalloc (newlen + 1);
231 /* Quote the characters that match the test mask. */
232 if (urlchr_test (*p1, mask))
234 unsigned char c = *p1++;
236 *p2++ = XNUM_TO_DIGIT (c >> 4);
237 *p2++ = XNUM_TO_DIGIT (c & 0xf);
242 assert (p2 - newstr == newlen);
248 /* URL-escape the unsafe characters (see urlchr_table) in a given
249 string, returning a freshly allocated string. */
252 url_escape (const char *s)
254 return url_escape_1 (s, urlchr_unsafe, false);
257 /* URL-escape the unsafe characters (see urlchr_table) in a given
258 string. If no characters are unsafe, S is returned. */
261 url_escape_allow_passthrough (const char *s)
263 return url_escape_1 (s, urlchr_unsafe, true);
266 /* Decide whether the char at position P needs to be encoded. (It is
267 not enough to pass a single char *P because the function may need
268 to inspect the surrounding context.)
270 Return true if the char should be escaped as %XX, false otherwise. */
273 char_needs_escaping (const char *p)
277 if (c_isxdigit (*(p + 1)) && c_isxdigit (*(p + 2)))
280 /* Garbled %.. sequence: encode `%'. */
283 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
289 /* Translate a %-escaped (but possibly non-conformant) input string S
290 into a %-escaped (and conformant) output string. If no characters
291 are encoded or decoded, return the same string S; otherwise, return
292 a freshly allocated string with the new contents.
294 After a URL has been run through this function, the protocols that
295 use `%' as the quote character can use the resulting string as-is,
296 while those that don't can use url_unescape to get to the intended
297 data. This function is stable: once the input is transformed,
298 further transformations of the result yield the same output.
300 Let's discuss why this function is needed.
302 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
303 a raw space character would mess up the HTTP request, it needs to
304 be quoted, like this:
306 GET /abc%20def HTTP/1.0
308 It would appear that the unsafe chars need to be quoted, for
309 example with url_escape. But what if we're requested to download
310 `abc%20def'? url_escape transforms "%" to "%25", which would leave
311 us with `abc%2520def'. This is incorrect -- since %-escapes are
312 part of URL syntax, "%20" is the correct way to denote a literal
313 space on the Wget command line. This leads to the conclusion that
314 in that case Wget should not call url_escape, but leave the `%20'
315 as is. This is clearly contradictory, but it only gets worse.
317 What if the requested URI is `abc%20 def'? If we call url_escape,
318 we end up with `/abc%2520%20def', which is almost certainly not
319 intended. If we don't call url_escape, we are left with the
320 embedded space and cannot complete the request. What the user
321 meant was for Wget to request `/abc%20%20def', and this is where
322 reencode_escapes kicks in.
324 Wget used to solve this by first decoding %-quotes, and then
325 encoding all the "unsafe" characters found in the resulting string.
326 This was wrong because it didn't preserve certain URL special
327 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
328 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
329 whether we considered `+' reserved (it is). One of these results
330 is inevitable because by the second step we would lose information
331 on whether the `+' was originally encoded or not. Both results
332 were wrong because in CGI parameters + means space, while %2B means
333 literal plus. reencode_escapes correctly translates the above to
334 "a%2B+b", i.e. returns the original string.
336 This function uses a modified version of the algorithm originally
337 proposed by Anon Sricharoenchai:
339 * Encode all "unsafe" characters, except those that are also
340 "reserved", to %XX. See urlchr_table for which characters are
343 * Encode the "%" characters not followed by two hex digits to
346 * Pass through all other characters and %XX escapes as-is. (Up to
347 Wget 1.10 this decoded %XX escapes corresponding to "safe"
348 characters, but that was obtrusive and broke some servers.)
352 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
354 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
358 "foo bar" -> "foo%20bar"
359 "foo%20bar" -> "foo%20bar"
360 "foo %20bar" -> "foo%20%20bar"
361 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
362 "foo%25%20bar" -> "foo%25%20bar"
363 "foo%2%20bar" -> "foo%252%20bar"
364 "foo+bar" -> "foo+bar" (plus is reserved!)
365 "foo%2b+bar" -> "foo%2b+bar" */
368 reencode_escapes (const char *s)
374 int encode_count = 0;
376 /* First pass: inspect the string to see if there's anything to do,
377 and to calculate the new length. */
378 for (p1 = s; *p1; p1++)
379 if (char_needs_escaping (p1))
383 /* The string is good as it is. */
384 return (char *) s; /* C const model sucks. */
387 /* Each encoding adds two characters (hex digits). */
388 newlen = oldlen + 2 * encode_count;
389 newstr = xmalloc (newlen + 1);
391 /* Second pass: copy the string to the destination address, encoding
392 chars when needed. */
397 if (char_needs_escaping (p1))
399 unsigned char c = *p1++;
401 *p2++ = XNUM_TO_DIGIT (c >> 4);
402 *p2++ = XNUM_TO_DIGIT (c & 0xf);
408 assert (p2 - newstr == newlen);
412 /* Returns the scheme type if the scheme is supported, or
413 SCHEME_INVALID if not. */
416 url_scheme (const char *url)
420 for (i = 0; supported_schemes[i].leading_string; i++)
421 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
422 strlen (supported_schemes[i].leading_string)))
424 if (!(supported_schemes[i].flags & scm_disabled))
425 return (enum url_scheme) i;
427 return SCHEME_INVALID;
430 return SCHEME_INVALID;
433 #define SCHEME_CHAR(ch) (c_isalnum (ch) || (ch) == '-' || (ch) == '+')
435 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
436 currently implemented, it returns true if URL begins with
440 url_has_scheme (const char *url)
444 /* The first char must be a scheme char. */
445 if (!*p || !SCHEME_CHAR (*p))
448 /* Followed by 0 or more scheme chars. */
449 while (*p && SCHEME_CHAR (*p))
451 /* Terminated by ':'. */
456 scheme_default_port (enum url_scheme scheme)
458 return supported_schemes[scheme].default_port;
462 scheme_disable (enum url_scheme scheme)
464 supported_schemes[scheme].flags |= scm_disabled;
467 /* Skip the username and password, if present in the URL. The
468 function should *not* be called with the complete URL, but with the
469 portion after the scheme.
471 If no username and password are found, return URL. */
474 url_skip_credentials (const char *url)
476 /* Look for '@' that comes before terminators, such as '/', '?',
478 const char *p = (const char *)strpbrk (url, "@/?#;");
484 /* Parse credentials contained in [BEG, END). The region is expected
485 to have come from a URL and is unescaped. */
488 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
494 return false; /* empty user name */
496 colon = memchr (beg, ':', end - beg);
498 return false; /* again empty user name */
502 *passwd = strdupdelim (colon + 1, end);
504 url_unescape (*passwd);
511 *user = strdupdelim (beg, userend);
512 url_unescape (*user);
516 /* Used by main.c: detect URLs written using the "shorthand" URL forms
517 originally popularized by Netscape and NcFTP. HTTP shorthands look
520 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
521 www.foo.com[:port] -> http://www.foo.com[:port]
523 FTP shorthands look like this:
525 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
526 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
528 If the URL needs not or cannot be rewritten, return NULL. */
531 rewrite_shorthand_url (const char *url)
536 if (url_scheme (url) != SCHEME_INVALID)
539 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
541 p = strpbrk (url, ":/");
545 /* If we're looking at "://", it means the URL uses a scheme we
546 don't support, which may include "https" when compiled without
547 SSL support. Don't bogusly rewrite such URLs. */
548 if (p && p[0] == ':' && p[1] == '/' && p[2] == '/')
553 /* Colon indicates ftp, as in foo.bar.com:path. Check for
554 special case of http port number ("localhost:10000"). */
555 int digits = strspn (p + 1, "0123456789");
556 if (digits && (p[1 + digits] == '/' || p[1 + digits] == '\0'))
559 /* Turn "foo.bar.com:path" to "ftp://foo.bar.com/path". */
560 ret = aprintf ("ftp://%s", url);
561 ret[6 + (p - url)] = '/';
566 /* Just prepend "http://" to URL. */
567 ret = aprintf ("http://%s", url);
572 static void split_path (const char *, char **, char **);
574 /* Like strpbrk, with the exception that it returns the pointer to the
575 terminating zero (end-of-string aka "eos") if no matching character
579 strpbrk_or_eos (const char *s, const char *accept)
581 char *p = strpbrk (s, accept);
583 p = strchr (s, '\0');
587 /* Turn STR into lowercase; return true if a character was actually
591 lowercase_str (char *str)
593 bool changed = false;
595 if (c_isupper (*str))
598 *str = c_tolower (*str);
604 init_seps (enum url_scheme scheme)
606 static char seps[8] = ":/";
608 int flags = supported_schemes[scheme].flags;
610 if (flags & scm_has_params)
612 if (flags & scm_has_query)
614 if (flags & scm_has_fragment)
620 static const char *parse_errors[] = {
621 #define PE_NO_ERROR 0
623 #define PE_UNSUPPORTED_SCHEME 1
624 N_("Unsupported scheme"),
625 #define PE_INVALID_HOST_NAME 2
626 N_("Invalid host name"),
627 #define PE_BAD_PORT_NUMBER 3
628 N_("Bad port number"),
629 #define PE_INVALID_USER_NAME 4
630 N_("Invalid user name"),
631 #define PE_UNTERMINATED_IPV6_ADDRESS 5
632 N_("Unterminated IPv6 numeric address"),
633 #define PE_IPV6_NOT_SUPPORTED 6
634 N_("IPv6 addresses not supported"),
635 #define PE_INVALID_IPV6_ADDRESS 7
636 N_("Invalid IPv6 numeric address")
641 Return a new struct url if successful, NULL on error. In case of
642 error, and if ERROR is not NULL, also set *ERROR to the appropriate
645 url_parse (const char *url, int *error)
649 bool path_modified, host_modified;
651 enum url_scheme scheme;
654 const char *uname_b, *uname_e;
655 const char *host_b, *host_e;
656 const char *path_b, *path_e;
657 const char *params_b, *params_e;
658 const char *query_b, *query_e;
659 const char *fragment_b, *fragment_e;
662 char *user = NULL, *passwd = NULL;
664 char *url_encoded = NULL;
668 scheme = url_scheme (url);
669 if (scheme == SCHEME_INVALID)
671 error_code = PE_UNSUPPORTED_SCHEME;
675 url_encoded = reencode_escapes (url);
678 p += strlen (supported_schemes[scheme].leading_string);
680 p = url_skip_credentials (p);
683 /* scheme://user:pass@host[:port]... */
686 /* We attempt to break down the URL into the components path,
687 params, query, and fragment. They are ordered like this:
689 scheme://host[:port][/path][;params][?query][#fragment] */
691 path_b = path_e = NULL;
692 params_b = params_e = NULL;
693 query_b = query_e = NULL;
694 fragment_b = fragment_e = NULL;
696 /* Initialize separators for optional parts of URL, depending on the
697 scheme. For example, FTP has params, and HTTP and HTTPS have
698 query string and fragment. */
699 seps = init_seps (scheme);
705 /* Handle IPv6 address inside square brackets. Ideally we'd
706 just look for the terminating ']', but rfc2732 mandates
707 rejecting invalid IPv6 addresses. */
709 /* The address begins after '['. */
711 host_e = strchr (host_b, ']');
715 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
720 /* Check if the IPv6 address is valid. */
721 if (!is_valid_ipv6_address(host_b, host_e))
723 error_code = PE_INVALID_IPV6_ADDRESS;
727 /* Continue parsing after the closing ']'. */
730 error_code = PE_IPV6_NOT_SUPPORTED;
734 /* The closing bracket must be followed by a separator or by the
736 /* http://[::1]... */
738 if (!strchr (seps, *p))
740 /* Trailing garbage after []-delimited IPv6 address. */
741 error_code = PE_INVALID_HOST_NAME;
747 p = strpbrk_or_eos (p, seps);
750 ++seps; /* advance to '/' */
752 if (host_b == host_e)
754 error_code = PE_INVALID_HOST_NAME;
758 port = scheme_default_port (scheme);
761 const char *port_b, *port_e, *pp;
763 /* scheme://host:port/tralala */
767 p = strpbrk_or_eos (p, seps);
770 /* Allow empty port, as per rfc2396. */
771 if (port_b != port_e)
772 for (port = 0, pp = port_b; pp < port_e; pp++)
774 if (!c_isdigit (*pp))
776 /* http://host:12randomgarbage/blah */
778 error_code = PE_BAD_PORT_NUMBER;
781 port = 10 * port + (*pp - '0');
782 /* Check for too large port numbers here, before we have
783 a chance to overflow on bogus port values. */
786 error_code = PE_BAD_PORT_NUMBER;
791 /* Advance to the first separator *after* '/' (either ';' or '?',
792 depending on the scheme). */
795 /* Get the optional parts of URL, each part being delimited by
796 current location and the position of the next separator. */
797 #define GET_URL_PART(sepchar, var) do { \
799 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
803 GET_URL_PART ('/', path);
804 if (supported_schemes[scheme].flags & scm_has_params)
805 GET_URL_PART (';', params);
806 if (supported_schemes[scheme].flags & scm_has_query)
807 GET_URL_PART ('?', query);
808 if (supported_schemes[scheme].flags & scm_has_fragment)
809 GET_URL_PART ('#', fragment);
814 if (uname_b != uname_e)
816 /* http://user:pass@host */
818 /* uname_b uname_e */
819 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
821 error_code = PE_INVALID_USER_NAME;
826 u = xnew0 (struct url);
828 u->host = strdupdelim (host_b, host_e);
833 u->path = strdupdelim (path_b, path_e);
834 path_modified = path_simplify (u->path);
835 split_path (u->path, &u->dir, &u->file);
837 host_modified = lowercase_str (u->host);
839 /* Decode %HH sequences in host name. This is important not so much
840 to support %HH sequences in host names (which other browser
841 don't), but to support binary characters (which will have been
842 converted to %HH by reencode_escapes). */
843 if (strchr (u->host, '%'))
845 url_unescape (u->host);
846 host_modified = true;
850 u->params = strdupdelim (params_b, params_e);
852 u->query = strdupdelim (query_b, query_e);
854 u->fragment = strdupdelim (fragment_b, fragment_e);
856 if (path_modified || u->fragment || host_modified || path_b == path_e)
858 /* If we suspect that a transformation has rendered what
859 url_string might return different from URL_ENCODED, rebuild
860 u->url using url_string. */
861 u->url = url_string (u, URL_AUTH_SHOW);
863 if (url_encoded != url)
864 xfree ((char *) url_encoded);
868 if (url_encoded == url)
869 u->url = xstrdup (url);
871 u->url = url_encoded;
877 /* Cleanup in case of error: */
878 if (url_encoded && url_encoded != url)
881 /* Transmit the error code to the caller, if the caller wants to
888 /* Return the error message string from ERROR_CODE, which should have
889 been retrieved from url_parse. The error message is translated. */
892 url_error (int error_code)
894 assert (error_code >= 0 && error_code < countof (parse_errors));
895 return _(parse_errors[error_code]);
898 /* Split PATH into DIR and FILE. PATH comes from the URL and is
899 expected to be URL-escaped.
901 The path is split into directory (the part up to the last slash)
902 and file (the part after the last slash), which are subsequently
906 "foo/bar/baz" "foo/bar" "baz"
907 "foo/bar/" "foo/bar" ""
909 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
911 DIR and FILE are freshly allocated. */
914 split_path (const char *path, char **dir, char **file)
916 char *last_slash = strrchr (path, '/');
920 *file = xstrdup (path);
924 *dir = strdupdelim (path, last_slash);
925 *file = xstrdup (last_slash + 1);
928 url_unescape (*file);
931 /* Note: URL's "full path" is the path with the query string and
932 params appended. The "fragment" (#foo) is intentionally ignored,
933 but that might be changed. For example, if the original URL was
934 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
935 the full path will be "/foo/bar/baz;bullshit?querystring". */
937 /* Return the length of the full path, without the terminating
941 full_path_length (const struct url *url)
945 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
956 /* Write out the full path. */
959 full_path_write (const struct url *url, char *where)
961 #define FROB(el, chr) do { \
962 char *f_el = url->el; \
964 int l = strlen (f_el); \
966 memcpy (where, f_el, l); \
978 /* Public function for getting the "full path". E.g. if u->path is
979 "foo/bar" and u->query is "param=value", full_path will be
980 "/foo/bar?param=value". */
983 url_full_path (const struct url *url)
985 int length = full_path_length (url);
986 char *full_path = xmalloc (length + 1);
988 full_path_write (url, full_path);
989 full_path[length] = '\0';
994 /* Unescape CHR in an otherwise escaped STR. Used to selectively
995 escaping of certain characters, such as "/" and ":". Returns a
996 count of unescaped chars. */
999 unescape_single_char (char *str, char chr)
1001 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1002 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1003 char *h = str; /* hare */
1004 char *t = str; /* tortoise */
1005 for (; *h; h++, t++)
1007 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1018 /* Escape unsafe and reserved characters, except for the slash
1022 url_escape_dir (const char *dir)
1024 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1028 unescape_single_char (newdir, '/');
1032 /* Sync u->path and u->url with u->dir and u->file. Called after
1033 u->file or u->dir have been changed, typically by the FTP code. */
1036 sync_path (struct url *u)
1038 char *newpath, *efile, *edir;
1042 /* u->dir and u->file are not escaped. URL-escape them before
1043 reassembling them into u->path. That way, if they contain
1044 separators like '?' or even if u->file contains slashes, the
1045 path will be correctly assembled. (u->file can contain slashes
1046 if the URL specifies it with %2f, or if an FTP server returns
1048 edir = url_escape_dir (u->dir);
1049 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1052 newpath = xstrdup (efile);
1055 int dirlen = strlen (edir);
1056 int filelen = strlen (efile);
1058 /* Copy "DIR/FILE" to newpath. */
1059 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1060 memcpy (p, edir, dirlen);
1063 memcpy (p, efile, filelen);
1072 if (efile != u->file)
1075 /* Regenerate u->url as well. */
1077 u->url = url_string (u, URL_AUTH_SHOW);
1080 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1081 This way we can sync u->path and u->url when they get changed. */
1084 url_set_dir (struct url *url, const char *newdir)
1087 url->dir = xstrdup (newdir);
1092 url_set_file (struct url *url, const char *newfile)
1095 url->file = xstrdup (newfile);
1100 url_free (struct url *url)
1106 xfree_null (url->params);
1107 xfree_null (url->query);
1108 xfree_null (url->fragment);
1109 xfree_null (url->user);
1110 xfree_null (url->passwd);
1118 /* Create all the necessary directories for PATH (a file). Calls
1119 make_directory internally. */
1121 mkalldirs (const char *path)
1128 p = path + strlen (path);
1129 for (; *p != '/' && p != path; p--)
1132 /* Don't create if it's just a file. */
1133 if ((p == path) && (*p != '/'))
1135 t = strdupdelim (path, p);
1137 /* Check whether the directory exists. */
1138 if ((stat (t, &st) == 0))
1140 if (S_ISDIR (st.st_mode))
1147 /* If the dir exists as a file name, remove it first. This
1148 is *only* for Wget to work with buggy old CERN http
1149 servers. Here is the scenario: When Wget tries to
1150 retrieve a directory without a slash, e.g.
1151 http://foo/bar (bar being a directory), CERN server will
1152 not redirect it too http://foo/bar/ -- it will generate a
1153 directory listing containing links to bar/file1,
1154 bar/file2, etc. Wget will lose because it saves this
1155 HTML listing to a file `bar', so it cannot create the
1156 directory. To work around this, if the file of the same
1157 name exists, we just remove it and create the directory
1159 DEBUGP (("Removing %s because of directory danger!\n", t));
1163 res = make_directory (t);
1165 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1170 /* Functions for constructing the file name out of URL components. */
1172 /* A growable string structure, used by url_file_name and friends.
1173 This should perhaps be moved to utils.c.
1175 The idea is to have a convenient and efficient way to construct a
1176 string by having various functions append data to it. Instead of
1177 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1178 functions in questions, we pass the pointer to this struct. */
1186 /* Ensure that the string can accept APPEND_COUNT more characters past
1187 the current TAIL position. If necessary, this will grow the string
1188 and update its allocated size. If the string is already large
1189 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1190 #define GROW(g, append_size) do { \
1191 struct growable *G_ = g; \
1192 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1195 /* Return the tail position of the string. */
1196 #define TAIL(r) ((r)->base + (r)->tail)
1198 /* Move the tail position by APPEND_COUNT characters. */
1199 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1201 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1205 append_string (const char *str, struct growable *dest)
1207 int l = strlen (str);
1209 memcpy (TAIL (dest), str, l);
1210 TAIL_INCR (dest, l);
1213 /* Append CH to DEST. For example, append_char (0, DEST)
1214 zero-terminates DEST. */
1217 append_char (char ch, struct growable *dest)
1221 TAIL_INCR (dest, 1);
1225 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1226 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1227 filechr_control = 4 /* a control character, e.g. 0-31 */
1230 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1232 /* Shorthands for the table: */
1233 #define U filechr_not_unix
1234 #define W filechr_not_windows
1235 #define C filechr_control
1240 /* Table of characters unsafe under various conditions (see above).
1242 Arguably we could also claim `%' to be unsafe, since we use it as
1243 the escape character. If we ever want to be able to reliably
1244 translate file name back to URL, this would become important
1245 crucial. Right now, it's better to be minimal in escaping. */
1247 static const unsigned char filechr_table[256] =
1249 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1250 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1251 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1252 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1253 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1254 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1255 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1256 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1257 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1258 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1259 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1260 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1261 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1262 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1263 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1264 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1266 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1267 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1268 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1269 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1271 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1272 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1273 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1274 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1282 /* FN_PORT_SEP is the separator between host and port in file names
1283 for non-standard port numbers. On Unix this is normally ':', as in
1284 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1285 because Windows can't handle ':' in file names. */
1286 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1288 /* FN_QUERY_SEP is the separator between the file name and the URL
1289 query, normally '?'. Since Windows cannot handle '?' as part of
1290 file name, we use '@' instead there. */
1291 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1293 /* Quote path element, characters in [b, e), as file name, and append
1294 the quoted string to DEST. Each character is quoted as per
1295 file_unsafe_char and the corresponding table.
1297 If ESCAPED is true, the path element is considered to be
1298 URL-escaped and will be unescaped prior to inspection. */
1301 append_uri_pathel (const char *b, const char *e, bool escaped,
1302 struct growable *dest)
1308 if (opt.restrict_files_os == restrict_unix)
1309 mask = filechr_not_unix;
1311 mask = filechr_not_windows;
1312 if (opt.restrict_files_ctrl)
1313 mask |= filechr_control;
1315 /* Copy [b, e) to PATHEL and URL-unescape it. */
1319 BOUNDED_TO_ALLOCA (b, e, unescaped);
1320 url_unescape (unescaped);
1322 e = unescaped + strlen (unescaped);
1325 /* Defang ".." when found as component of path. Remember that path
1326 comes from the URL and might contain malicious input. */
1327 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1333 /* Walk the PATHEL string and check how many characters we'll need
1336 for (p = b; p < e; p++)
1337 if (FILE_CHAR_TEST (*p, mask))
1340 /* Calculate the length of the output string. e-b is the input
1341 string length. Each quoted char introduces two additional
1342 characters in the string, hence 2*quoted. */
1343 outlen = (e - b) + (2 * quoted);
1344 GROW (dest, outlen);
1348 /* If there's nothing to quote, we can simply append the string
1349 without processing it again. */
1350 memcpy (TAIL (dest), b, outlen);
1354 char *q = TAIL (dest);
1355 for (p = b; p < e; p++)
1357 if (!FILE_CHAR_TEST (*p, mask))
1361 unsigned char ch = *p;
1363 *q++ = XNUM_TO_DIGIT (ch >> 4);
1364 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1367 assert (q - TAIL (dest) == outlen);
1370 /* Perform inline case transformation if required. */
1371 if (opt.restrict_files_case == restrict_lowercase
1372 || opt.restrict_files_case == restrict_uppercase)
1375 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1377 if (opt.restrict_files_case == restrict_lowercase)
1378 *q = c_tolower (*q);
1380 *q = c_toupper (*q);
1384 TAIL_INCR (dest, outlen);
1387 /* Append to DEST the directory structure that corresponds the
1388 directory part of URL's path. For example, if the URL is
1389 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1391 Each path element ("dir1" and "dir2" in the above example) is
1392 examined, url-unescaped, and re-escaped as file name element.
1394 Additionally, it cuts as many directories from the path as
1395 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1396 will produce "bar" for the above example. For 2 or more, it will
1399 Each component of the path is quoted for use as file name. */
1402 append_dir_structure (const struct url *u, struct growable *dest)
1404 char *pathel, *next;
1405 int cut = opt.cut_dirs;
1407 /* Go through the path components, de-URL-quote them, and quote them
1408 (if necessary) as file names. */
1411 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1416 /* Ignore empty pathels. */
1420 append_char ('/', dest);
1421 append_uri_pathel (pathel, next, true, dest);
1425 /* Return a unique file name that matches the given URL as good as
1426 possible. Does not create directories on the file system. */
1429 url_file_name (const struct url *u)
1431 struct growable fnres; /* stands for "file name result" */
1433 const char *u_file, *u_query;
1434 char *fname, *unique;
1440 /* Start with the directory prefix, if specified. */
1442 append_string (opt.dir_prefix, &fnres);
1444 /* If "dirstruct" is turned on (typically the case with -r), add
1445 the host and port (unless those have been turned off) and
1446 directory structure. */
1449 if (opt.protocol_directories)
1452 append_char ('/', &fnres);
1453 append_string (supported_schemes[u->scheme].name, &fnres);
1455 if (opt.add_hostdir)
1458 append_char ('/', &fnres);
1459 if (0 != strcmp (u->host, ".."))
1460 append_string (u->host, &fnres);
1462 /* Host name can come from the network; malicious DNS may
1463 allow ".." to be resolved, causing us to write to
1464 "../<file>". Defang such host names. */
1465 append_string ("%2E%2E", &fnres);
1466 if (u->port != scheme_default_port (u->scheme))
1469 number_to_string (portstr, u->port);
1470 append_char (FN_PORT_SEP, &fnres);
1471 append_string (portstr, &fnres);
1475 append_dir_structure (u, &fnres);
1478 /* Add the file name. */
1480 append_char ('/', &fnres);
1481 u_file = *u->file ? u->file : "index.html";
1482 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1484 /* Append "?query" to the file name. */
1485 u_query = u->query && *u->query ? u->query : NULL;
1488 append_char (FN_QUERY_SEP, &fnres);
1489 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1492 /* Zero-terminate the file name. */
1493 append_char ('\0', &fnres);
1497 /* Check the cases in which the unique extensions are not used:
1498 1) Clobbering is turned off (-nc).
1499 2) Retrieval with regetting.
1500 3) Timestamping is used.
1501 4) Hierarchy is built.
1503 The exception is the case when file does exist and is a
1504 directory (see `mkalldirs' for explanation). */
1506 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1507 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1510 unique = unique_name (fname, true);
1511 if (unique != fname)
1516 /* Resolve "." and ".." elements of PATH by destructively modifying
1517 PATH and return true if PATH has been modified, false otherwise.
1519 The algorithm is in spirit similar to the one described in rfc1808,
1520 although implemented differently, in one pass. To recap, path
1521 elements containing only "." are removed, and ".." is taken to mean
1522 "back up one element". Single leading and trailing slashes are
1525 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1526 test examples are provided below. If you change anything in this
1527 function, run test_path_simplify to make sure you haven't broken a
1531 path_simplify (char *path)
1533 char *h = path; /* hare */
1534 char *t = path; /* tortoise */
1535 char *end = strchr (path, '\0');
1539 /* Hare should be at the beginning of a path element. */
1541 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1546 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1548 /* Handle "../" by retreating the tortoise by one path
1549 element -- but not past beggining. */
1552 /* Move backwards until T hits the beginning of the
1553 previous path element or the beginning of path. */
1554 for (--t; t > path && t[-1] != '/'; t--)
1561 /* A regular path element. If H hasn't advanced past T,
1562 simply skip to the next path element. Otherwise, copy
1563 the path element until the next slash. */
1566 /* Skip the path element, including the slash. */
1567 while (h < end && *h != '/')
1574 /* Copy the path element, including the final slash. */
1575 while (h < end && *h != '/')
1589 /* Return the length of URL's path. Path is considered to be
1590 terminated by one or more of the ?query or ;params or #fragment,
1591 depending on the scheme. */
1594 path_end (const char *url)
1596 enum url_scheme scheme = url_scheme (url);
1598 if (scheme == SCHEME_INVALID)
1599 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1600 /* +2 to ignore the first two separators ':' and '/' */
1601 seps = init_seps (scheme) + 2;
1602 return strpbrk_or_eos (url, seps);
1605 /* Find the last occurrence of character C in the range [b, e), or
1606 NULL, if none are present. */
1607 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1609 /* Merge BASE with LINK and return the resulting URI.
1611 Either of the URIs may be absolute or relative, complete with the
1612 host name, or path only. This tries to reasonably handle all
1613 foreseeable cases. It only employs minimal URL parsing, without
1614 knowledge of the specifics of schemes.
1616 I briefly considered making this function call path_simplify after
1617 the merging process, as rfc1738 seems to suggest. This is a bad
1618 idea for several reasons: 1) it complexifies the code, and 2)
1619 url_parse has to simplify path anyway, so it's wasteful to boot. */
1622 uri_merge (const char *base, const char *link)
1628 if (url_has_scheme (link))
1629 return xstrdup (link);
1631 /* We may not examine BASE past END. */
1632 end = path_end (base);
1633 linklength = strlen (link);
1637 /* Empty LINK points back to BASE, query string and all. */
1638 return xstrdup (base);
1640 else if (*link == '?')
1642 /* LINK points to the same location, but changes the query
1643 string. Examples: */
1644 /* uri_merge("path", "?new") -> "path?new" */
1645 /* uri_merge("path?foo", "?new") -> "path?new" */
1646 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1647 /* uri_merge("path#foo", "?new") -> "path?new" */
1648 int baselength = end - base;
1649 merge = xmalloc (baselength + linklength + 1);
1650 memcpy (merge, base, baselength);
1651 memcpy (merge + baselength, link, linklength);
1652 merge[baselength + linklength] = '\0';
1654 else if (*link == '#')
1656 /* uri_merge("path", "#new") -> "path#new" */
1657 /* uri_merge("path#foo", "#new") -> "path#new" */
1658 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1659 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1661 const char *end1 = strchr (base, '#');
1663 end1 = base + strlen (base);
1664 baselength = end1 - base;
1665 merge = xmalloc (baselength + linklength + 1);
1666 memcpy (merge, base, baselength);
1667 memcpy (merge + baselength, link, linklength);
1668 merge[baselength + linklength] = '\0';
1670 else if (*link == '/' && *(link + 1) == '/')
1672 /* LINK begins with "//" and so is a net path: we need to
1673 replace everything after (and including) the double slash
1676 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1677 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1678 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1682 const char *start_insert;
1684 /* Look for first slash. */
1685 slash = memchr (base, '/', end - base);
1686 /* If found slash and it is a double slash, then replace
1687 from this point, else default to replacing from the
1689 if (slash && *(slash + 1) == '/')
1690 start_insert = slash;
1692 start_insert = base;
1694 span = start_insert - base;
1695 merge = xmalloc (span + linklength + 1);
1697 memcpy (merge, base, span);
1698 memcpy (merge + span, link, linklength);
1699 merge[span + linklength] = '\0';
1701 else if (*link == '/')
1703 /* LINK is an absolute path: we need to replace everything
1704 after (and including) the FIRST slash with LINK.
1706 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1707 "/qux/xyzzy", our result should be
1708 "http://host/qux/xyzzy". */
1711 const char *start_insert = NULL; /* for gcc to shut up. */
1712 const char *pos = base;
1713 bool seen_slash_slash = false;
1714 /* We're looking for the first slash, but want to ignore
1717 slash = memchr (pos, '/', end - pos);
1718 if (slash && !seen_slash_slash)
1719 if (*(slash + 1) == '/')
1722 seen_slash_slash = true;
1726 /* At this point, SLASH is the location of the first / after
1727 "//", or the first slash altogether. START_INSERT is the
1728 pointer to the location where LINK will be inserted. When
1729 examining the last two examples, keep in mind that LINK
1732 if (!slash && !seen_slash_slash)
1733 /* example: "foo" */
1735 start_insert = base;
1736 else if (!slash && seen_slash_slash)
1737 /* example: "http://foo" */
1740 else if (slash && !seen_slash_slash)
1741 /* example: "foo/bar" */
1743 start_insert = base;
1744 else if (slash && seen_slash_slash)
1745 /* example: "http://something/" */
1747 start_insert = slash;
1749 span = start_insert - base;
1750 merge = xmalloc (span + linklength + 1);
1752 memcpy (merge, base, span);
1753 memcpy (merge + span, link, linklength);
1754 merge[span + linklength] = '\0';
1758 /* LINK is a relative URL: we need to replace everything
1759 after last slash (possibly empty) with LINK.
1761 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1762 our result should be "whatever/foo/qux/xyzzy". */
1763 bool need_explicit_slash = false;
1765 const char *start_insert;
1766 const char *last_slash = find_last_char (base, end, '/');
1769 /* No slash found at all. Replace what we have with LINK. */
1770 start_insert = base;
1772 else if (last_slash && last_slash >= base + 2
1773 && last_slash[-2] == ':' && last_slash[-1] == '/')
1775 /* example: http://host" */
1777 start_insert = end + 1;
1778 need_explicit_slash = true;
1782 /* example: "whatever/foo/bar" */
1784 start_insert = last_slash + 1;
1787 span = start_insert - base;
1788 merge = xmalloc (span + linklength + 1);
1790 memcpy (merge, base, span);
1791 if (need_explicit_slash)
1792 merge[span - 1] = '/';
1793 memcpy (merge + span, link, linklength);
1794 merge[span + linklength] = '\0';
1800 #define APPEND(p, s) do { \
1801 int len = strlen (s); \
1802 memcpy (p, s, len); \
1806 /* Use this instead of password when the actual password is supposed
1807 to be hidden. We intentionally use a generic string without giving
1808 away the number of characters in the password, like previous
1810 #define HIDDEN_PASSWORD "*password*"
1812 /* Recreate the URL string from the data in URL.
1814 If HIDE is true (as it is when we're calling this on a URL we plan
1815 to print, but not when calling it to canonicalize a URL for use
1816 within the program), password will be hidden. Unsafe characters in
1817 the URL will be quoted. */
1820 url_string (const struct url *url, enum url_auth_mode auth_mode)
1824 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1826 int scheme_port = supported_schemes[url->scheme].default_port;
1827 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1828 int fplen = full_path_length (url);
1830 bool brackets_around_host;
1832 assert (scheme_str != NULL);
1834 /* Make sure the user name and password are quoted. */
1837 if (auth_mode != URL_AUTH_HIDE)
1839 quoted_user = url_escape_allow_passthrough (url->user);
1842 if (auth_mode == URL_AUTH_HIDE_PASSWD)
1843 quoted_passwd = HIDDEN_PASSWORD;
1845 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1850 /* In the unlikely event that the host name contains non-printable
1851 characters, quote it for displaying to the user. */
1852 quoted_host = url_escape_allow_passthrough (url->host);
1854 /* Undo the quoting of colons that URL escaping performs. IPv6
1855 addresses may legally contain colons, and in that case must be
1856 placed in square brackets. */
1857 if (quoted_host != url->host)
1858 unescape_single_char (quoted_host, ':');
1859 brackets_around_host = strchr (quoted_host, ':') != NULL;
1861 size = (strlen (scheme_str)
1862 + strlen (quoted_host)
1863 + (brackets_around_host ? 2 : 0)
1866 if (url->port != scheme_port)
1867 size += 1 + numdigit (url->port);
1870 size += 1 + strlen (quoted_user);
1872 size += 1 + strlen (quoted_passwd);
1875 p = result = xmalloc (size);
1877 APPEND (p, scheme_str);
1880 APPEND (p, quoted_user);
1884 APPEND (p, quoted_passwd);
1889 if (brackets_around_host)
1891 APPEND (p, quoted_host);
1892 if (brackets_around_host)
1894 if (url->port != scheme_port)
1897 p = number_to_string (p, url->port);
1900 full_path_write (url, p);
1904 assert (p - result == size);
1906 if (quoted_user && quoted_user != url->user)
1907 xfree (quoted_user);
1908 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
1909 && quoted_passwd != url->passwd)
1910 xfree (quoted_passwd);
1911 if (quoted_host != url->host)
1912 xfree (quoted_host);
1917 /* Return true if scheme a is similar to scheme b.
1919 Schemes are similar if they are equal. If SSL is supported, schemes
1920 are also similar if one is http (SCHEME_HTTP) and the other is https
1923 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1928 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1929 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1936 getchar_from_escaped_string (const char *str, char *c)
1938 const char *p = str;
1940 assert (str && *str);
1945 if (!c_isxdigit(p[1]) || !c_isxdigit(p[2]))
1953 return 0; /* error: invalid string */
1955 *c = X2DIGITS_TO_NUM (p[1], p[2]);
1956 if (URL_RESERVED_CHAR(*c))
1974 are_urls_equal (const char *u1, const char *u2)
1985 && (pp = getchar_from_escaped_string (p, &ch1))
1986 && (qq = getchar_from_escaped_string (q, &ch2))
1987 && (c_tolower(ch1) == c_tolower(ch2)))
1993 return (*p == 0 && *q == 0 ? true : false);
1997 /* Debugging and testing support for path_simplify. */
1999 /* Debug: run path_simplify on PATH and return the result in a new
2000 string. Useful for calling from the debugger. */
2004 char *copy = xstrdup (path);
2005 path_simplify (copy);
2010 run_test (char *test, char *expected_result, bool expected_change)
2012 char *test_copy = xstrdup (test);
2013 bool modified = path_simplify (test_copy);
2015 if (0 != strcmp (test_copy, expected_result))
2017 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2018 test, expected_result, test_copy);
2020 if (modified != expected_change)
2022 if (expected_change)
2023 printf ("Expected modification with path_simplify(\"%s\").\n",
2026 printf ("Expected no modification with path_simplify(\"%s\").\n",
2033 test_path_simplify (void)
2036 char *test, *result;
2043 { "../", "", true },
2044 { "foo", "foo", false },
2045 { "foo/bar", "foo/bar", false },
2046 { "foo///bar", "foo///bar", false },
2047 { "foo/.", "foo/", true },
2048 { "foo/./", "foo/", true },
2049 { "foo./", "foo./", false },
2050 { "foo/../bar", "bar", true },
2051 { "foo/../bar/", "bar/", true },
2052 { "foo/bar/..", "foo/", true },
2053 { "foo/bar/../x", "foo/x", true },
2054 { "foo/bar/../x/", "foo/x/", true },
2055 { "foo/..", "", true },
2056 { "foo/../..", "", true },
2057 { "foo/../../..", "", true },
2058 { "foo/../../bar/../../baz", "baz", true },
2059 { "a/b/../../c", "c", true },
2060 { "./a/../b", "b", true }
2064 for (i = 0; i < countof (tests); i++)
2066 char *test = tests[i].test;
2067 char *expected_result = tests[i].result;
2068 bool expected_change = tests[i].should_modify;
2069 run_test (test, expected_result, expected_change);
2077 test_append_uri_pathel()
2084 char *expected_result;
2086 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2089 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2091 struct growable dest;
2092 const char *p = test_array[i].input;
2094 memset (&dest, 0, sizeof (dest));
2096 append_string (test_array[i].original_url, &dest);
2097 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2098 append_char ('\0', &dest);
2100 mu_assert ("test_append_uri_pathel: wrong result",
2101 strcmp (dest.base, test_array[i].expected_result) == 0);
2108 test_are_urls_equal()
2114 bool expected_result;
2116 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2117 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2118 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2119 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2120 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2121 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2124 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2126 mu_assert ("test_are_urls_equal: wrong result",
2127 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2133 #endif /* TESTING */