2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
3 2005, 2006, 2007, 2008, 2009, 2010, 2011 Free Software Foundation,
6 This file is part of GNU Wget.
8 GNU Wget is free software; you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation; either version 3 of the License, or (at
11 your option) any later version.
13 GNU Wget is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with Wget. If not, see <http://www.gnu.org/licenses/>.
21 Additional permission under GNU GPL version 3 section 7
23 If you modify this program, or any covered work, by linking or
24 combining it with the OpenSSL project's OpenSSL library (or a
25 modified version of that library), containing parts covered by the
26 terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
27 grants you additional permission to convey the resulting work.
28 Corresponding Source for a non-source form of such a combination
29 shall include the source code for the parts of OpenSSL used as well
30 as that of the covered work. */
43 #include "host.h" /* for is_valid_ipv6_address */
47 #endif /* def __VMS */
54 scm_disabled = 1, /* for https when OpenSSL fails to init. */
55 scm_has_params = 2, /* whether scheme has ;params */
56 scm_has_query = 4, /* whether scheme has ?query */
57 scm_has_fragment = 8 /* whether scheme has #fragment */
62 /* Short name of the scheme, such as "http" or "ftp". */
64 /* Leading string that identifies the scheme, such as "https://". */
65 const char *leading_string;
66 /* Default port of the scheme when none is specified. */
72 /* Supported schemes: */
73 static struct scheme_data supported_schemes[] =
75 { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment },
77 { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment },
79 { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment },
85 /* Forward declarations: */
87 static bool path_simplify (enum url_scheme, char *);
89 /* Support for escaping and unescaping of URL strings. */
91 /* Table of "reserved" and "unsafe" characters. Those terms are
92 rfc1738-speak, as such largely obsoleted by rfc2396 and later
93 specs, but the general idea remains.
95 A reserved character is the one that you can't decode without
96 changing the meaning of the URL. For example, you can't decode
97 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
98 path components is different. Non-reserved characters can be
99 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
100 unsafe characters are loosely based on rfc1738, plus "$" and ",",
101 as recommended by rfc2396, and minus "~", which is very frequently
102 used (and sometimes unrecognized as %7E by broken servers).
104 An unsafe character is the one that should be encoded when URLs are
105 placed in foreign environments. E.g. space and newline are unsafe
106 in HTTP contexts because HTTP uses them as separator and line
107 terminator, so they must be encoded to %20 and %0A respectively.
108 "*" is unsafe in shell context, etc.
110 We determine whether a character is unsafe through static table
111 lookup. This code assumes ASCII character set and 8-bit chars. */
114 /* rfc1738 reserved chars + "$" and ",". */
117 /* rfc1738 unsafe chars, plus non-printables. */
121 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
122 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
123 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
125 /* Shorthands for the table: */
126 #define R urlchr_reserved
127 #define U urlchr_unsafe
130 static const unsigned char urlchr_table[256] =
132 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
133 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
134 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
135 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
136 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
137 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
138 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
139 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
140 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
141 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
142 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
143 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
144 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
145 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
146 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
147 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
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,
152 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,
157 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
163 /* URL-unescape the string S.
165 This is done by transforming the sequences "%HH" to the character
166 represented by the hexadecimal digits HH. If % is not followed by
167 two hexadecimal digits, it is inserted literally.
169 The transformation is done in place. If you need the original
170 string intact, make a copy before calling this function. */
173 url_unescape (char *s)
175 char *t = s; /* t - tortoise */
176 char *h = s; /* h - hare */
188 /* Do nothing if '%' is not followed by two hex digits. */
189 if (!h[1] || !h[2] || !(c_isxdigit (h[1]) && c_isxdigit (h[2])))
191 c = X2DIGITS_TO_NUM (h[1], h[2]);
192 /* Don't unescape %00 because there is no way to insert it
193 into a C string without effectively truncating it. */
203 /* The core of url_escape_* functions. Escapes the characters that
204 match the provided mask in urlchr_table.
206 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
207 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
208 allocated string will be returned in all cases. */
211 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
218 for (p1 = s; *p1; p1++)
219 if (urlchr_test (*p1, mask))
220 addition += 2; /* Two more characters (hex digits) */
223 return allow_passthrough ? (char *)s : xstrdup (s);
225 newlen = (p1 - s) + addition;
226 newstr = xmalloc (newlen + 1);
232 /* Quote the characters that match the test mask. */
233 if (urlchr_test (*p1, mask))
235 unsigned char c = *p1++;
237 *p2++ = XNUM_TO_DIGIT (c >> 4);
238 *p2++ = XNUM_TO_DIGIT (c & 0xf);
243 assert (p2 - newstr == newlen);
249 /* URL-escape the unsafe characters (see urlchr_table) in a given
250 string, returning a freshly allocated string. */
253 url_escape (const char *s)
255 return url_escape_1 (s, urlchr_unsafe, false);
258 /* URL-escape the unsafe and reserved characters (see urlchr_table) in
259 a given string, returning a freshly allocated string. */
262 url_escape_unsafe_and_reserved (const char *s)
264 return url_escape_1 (s, urlchr_unsafe|urlchr_reserved, false);
267 /* URL-escape the unsafe characters (see urlchr_table) in a given
268 string. If no characters are unsafe, S is returned. */
271 url_escape_allow_passthrough (const char *s)
273 return url_escape_1 (s, urlchr_unsafe, true);
276 /* Decide whether the char at position P needs to be encoded. (It is
277 not enough to pass a single char *P because the function may need
278 to inspect the surrounding context.)
280 Return true if the char should be escaped as %XX, false otherwise. */
283 char_needs_escaping (const char *p)
287 if (c_isxdigit (*(p + 1)) && c_isxdigit (*(p + 2)))
290 /* Garbled %.. sequence: encode `%'. */
293 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
299 /* Translate a %-escaped (but possibly non-conformant) input string S
300 into a %-escaped (and conformant) output string. If no characters
301 are encoded or decoded, return the same string S; otherwise, return
302 a freshly allocated string with the new contents.
304 After a URL has been run through this function, the protocols that
305 use `%' as the quote character can use the resulting string as-is,
306 while those that don't can use url_unescape to get to the intended
307 data. This function is stable: once the input is transformed,
308 further transformations of the result yield the same output.
310 Let's discuss why this function is needed.
312 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
313 a raw space character would mess up the HTTP request, it needs to
314 be quoted, like this:
316 GET /abc%20def HTTP/1.0
318 It would appear that the unsafe chars need to be quoted, for
319 example with url_escape. But what if we're requested to download
320 `abc%20def'? url_escape transforms "%" to "%25", which would leave
321 us with `abc%2520def'. This is incorrect -- since %-escapes are
322 part of URL syntax, "%20" is the correct way to denote a literal
323 space on the Wget command line. This leads to the conclusion that
324 in that case Wget should not call url_escape, but leave the `%20'
325 as is. This is clearly contradictory, but it only gets worse.
327 What if the requested URI is `abc%20 def'? If we call url_escape,
328 we end up with `/abc%2520%20def', which is almost certainly not
329 intended. If we don't call url_escape, we are left with the
330 embedded space and cannot complete the request. What the user
331 meant was for Wget to request `/abc%20%20def', and this is where
332 reencode_escapes kicks in.
334 Wget used to solve this by first decoding %-quotes, and then
335 encoding all the "unsafe" characters found in the resulting string.
336 This was wrong because it didn't preserve certain URL special
337 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
338 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
339 whether we considered `+' reserved (it is). One of these results
340 is inevitable because by the second step we would lose information
341 on whether the `+' was originally encoded or not. Both results
342 were wrong because in CGI parameters + means space, while %2B means
343 literal plus. reencode_escapes correctly translates the above to
344 "a%2B+b", i.e. returns the original string.
346 This function uses a modified version of the algorithm originally
347 proposed by Anon Sricharoenchai:
349 * Encode all "unsafe" characters, except those that are also
350 "reserved", to %XX. See urlchr_table for which characters are
353 * Encode the "%" characters not followed by two hex digits to
356 * Pass through all other characters and %XX escapes as-is. (Up to
357 Wget 1.10 this decoded %XX escapes corresponding to "safe"
358 characters, but that was obtrusive and broke some servers.)
362 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
364 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
368 "foo bar" -> "foo%20bar"
369 "foo%20bar" -> "foo%20bar"
370 "foo %20bar" -> "foo%20%20bar"
371 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
372 "foo%25%20bar" -> "foo%25%20bar"
373 "foo%2%20bar" -> "foo%252%20bar"
374 "foo+bar" -> "foo+bar" (plus is reserved!)
375 "foo%2b+bar" -> "foo%2b+bar" */
378 reencode_escapes (const char *s)
384 int encode_count = 0;
386 /* First pass: inspect the string to see if there's anything to do,
387 and to calculate the new length. */
388 for (p1 = s; *p1; p1++)
389 if (char_needs_escaping (p1))
393 /* The string is good as it is. */
394 return (char *) s; /* C const model sucks. */
397 /* Each encoding adds two characters (hex digits). */
398 newlen = oldlen + 2 * encode_count;
399 newstr = xmalloc (newlen + 1);
401 /* Second pass: copy the string to the destination address, encoding
402 chars when needed. */
407 if (char_needs_escaping (p1))
409 unsigned char c = *p1++;
411 *p2++ = XNUM_TO_DIGIT (c >> 4);
412 *p2++ = XNUM_TO_DIGIT (c & 0xf);
418 assert (p2 - newstr == newlen);
422 /* Returns the scheme type if the scheme is supported, or
423 SCHEME_INVALID if not. */
426 url_scheme (const char *url)
430 for (i = 0; supported_schemes[i].leading_string; i++)
431 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
432 strlen (supported_schemes[i].leading_string)))
434 if (!(supported_schemes[i].flags & scm_disabled))
435 return (enum url_scheme) i;
437 return SCHEME_INVALID;
440 return SCHEME_INVALID;
443 #define SCHEME_CHAR(ch) (c_isalnum (ch) || (ch) == '-' || (ch) == '+')
445 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
446 currently implemented, it returns true if URL begins with
450 url_has_scheme (const char *url)
454 /* The first char must be a scheme char. */
455 if (!*p || !SCHEME_CHAR (*p))
458 /* Followed by 0 or more scheme chars. */
459 while (*p && SCHEME_CHAR (*p))
461 /* Terminated by ':'. */
466 url_valid_scheme (const char *url)
468 enum url_scheme scheme = url_scheme (url);
469 return scheme != SCHEME_INVALID;
473 scheme_default_port (enum url_scheme scheme)
475 return supported_schemes[scheme].default_port;
479 scheme_disable (enum url_scheme scheme)
481 supported_schemes[scheme].flags |= scm_disabled;
484 /* Skip the username and password, if present in the URL. The
485 function should *not* be called with the complete URL, but with the
486 portion after the scheme.
488 If no username and password are found, return URL. */
491 url_skip_credentials (const char *url)
493 /* Look for '@' that comes before terminators, such as '/', '?',
495 const char *p = (const char *)strpbrk (url, "@/?#;");
501 /* Parse credentials contained in [BEG, END). The region is expected
502 to have come from a URL and is unescaped. */
505 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
511 return false; /* empty user name */
513 colon = memchr (beg, ':', end - beg);
515 return false; /* again empty user name */
519 *passwd = strdupdelim (colon + 1, end);
521 url_unescape (*passwd);
528 *user = strdupdelim (beg, userend);
529 url_unescape (*user);
533 /* Used by main.c: detect URLs written using the "shorthand" URL forms
534 originally popularized by Netscape and NcFTP. HTTP shorthands look
537 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
538 www.foo.com[:port] -> http://www.foo.com[:port]
540 FTP shorthands look like this:
542 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
543 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
545 If the URL needs not or cannot be rewritten, return NULL. */
548 rewrite_shorthand_url (const char *url)
553 if (url_scheme (url) != SCHEME_INVALID)
556 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
558 p = strpbrk (url, ":/");
562 /* If we're looking at "://", it means the URL uses a scheme we
563 don't support, which may include "https" when compiled without
564 SSL support. Don't bogusly rewrite such URLs. */
565 if (p && p[0] == ':' && p[1] == '/' && p[2] == '/')
570 /* Colon indicates ftp, as in foo.bar.com:path. Check for
571 special case of http port number ("localhost:10000"). */
572 int digits = strspn (p + 1, "0123456789");
573 if (digits && (p[1 + digits] == '/' || p[1 + digits] == '\0'))
576 /* Turn "foo.bar.com:path" to "ftp://foo.bar.com/path". */
577 ret = aprintf ("ftp://%s", url);
578 ret[6 + (p - url)] = '/';
583 /* Just prepend "http://" to URL. */
584 ret = aprintf ("http://%s", url);
589 static void split_path (const char *, char **, char **);
591 /* Like strpbrk, with the exception that it returns the pointer to the
592 terminating zero (end-of-string aka "eos") if no matching character
596 strpbrk_or_eos (const char *s, const char *accept)
598 char *p = strpbrk (s, accept);
600 p = strchr (s, '\0');
604 /* Turn STR into lowercase; return true if a character was actually
608 lowercase_str (char *str)
610 bool changed = false;
612 if (c_isupper (*str))
615 *str = c_tolower (*str);
621 init_seps (enum url_scheme scheme)
623 static char seps[8] = ":/";
625 int flags = supported_schemes[scheme].flags;
627 if (flags & scm_has_params)
629 if (flags & scm_has_query)
631 if (flags & scm_has_fragment)
637 static const char *parse_errors[] = {
638 #define PE_NO_ERROR 0
640 #define PE_UNSUPPORTED_SCHEME 1
641 N_("Unsupported scheme %s"), /* support for format token only here */
642 #define PE_MISSING_SCHEME 2
643 N_("Scheme missing"),
644 #define PE_INVALID_HOST_NAME 3
645 N_("Invalid host name"),
646 #define PE_BAD_PORT_NUMBER 4
647 N_("Bad port number"),
648 #define PE_INVALID_USER_NAME 5
649 N_("Invalid user name"),
650 #define PE_UNTERMINATED_IPV6_ADDRESS 6
651 N_("Unterminated IPv6 numeric address"),
652 #define PE_IPV6_NOT_SUPPORTED 7
653 N_("IPv6 addresses not supported"),
654 #define PE_INVALID_IPV6_ADDRESS 8
655 N_("Invalid IPv6 numeric address")
660 Return a new struct url if successful, NULL on error. In case of
661 error, and if ERROR is not NULL, also set *ERROR to the appropriate
664 url_parse (const char *url, int *error, struct iri *iri, bool percent_encode)
668 bool path_modified, host_modified;
670 enum url_scheme scheme;
673 const char *uname_b, *uname_e;
674 const char *host_b, *host_e;
675 const char *path_b, *path_e;
676 const char *params_b, *params_e;
677 const char *query_b, *query_e;
678 const char *fragment_b, *fragment_e;
681 char *user = NULL, *passwd = NULL;
683 const char *url_encoded = NULL;
684 char *new_url = NULL;
688 scheme = url_scheme (url);
689 if (scheme == SCHEME_INVALID)
691 if (url_has_scheme (url))
692 error_code = PE_UNSUPPORTED_SCHEME;
694 error_code = PE_MISSING_SCHEME;
698 if (iri && iri->utf8_encode)
700 iri->utf8_encode = remote_to_utf8 (iri, iri->orig_url ? iri->orig_url : url, (const char **) &new_url);
701 if (!iri->utf8_encode)
704 iri->orig_url = xstrdup (url);
707 /* XXX XXX Could that change introduce (security) bugs ??? XXX XXX*/
709 url_encoded = reencode_escapes (new_url ? new_url : url);
711 url_encoded = new_url ? new_url : url;
715 if (new_url && url_encoded != new_url)
718 p += strlen (supported_schemes[scheme].leading_string);
720 p = url_skip_credentials (p);
723 /* scheme://user:pass@host[:port]... */
726 /* We attempt to break down the URL into the components path,
727 params, query, and fragment. They are ordered like this:
729 scheme://host[:port][/path][;params][?query][#fragment] */
731 path_b = path_e = NULL;
732 params_b = params_e = NULL;
733 query_b = query_e = NULL;
734 fragment_b = fragment_e = NULL;
736 /* Initialize separators for optional parts of URL, depending on the
737 scheme. For example, FTP has params, and HTTP and HTTPS have
738 query string and fragment. */
739 seps = init_seps (scheme);
745 /* Handle IPv6 address inside square brackets. Ideally we'd
746 just look for the terminating ']', but rfc2732 mandates
747 rejecting invalid IPv6 addresses. */
749 /* The address begins after '['. */
751 host_e = strchr (host_b, ']');
755 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
760 /* Check if the IPv6 address is valid. */
761 if (!is_valid_ipv6_address(host_b, host_e))
763 error_code = PE_INVALID_IPV6_ADDRESS;
767 /* Continue parsing after the closing ']'. */
770 error_code = PE_IPV6_NOT_SUPPORTED;
774 /* The closing bracket must be followed by a separator or by the
776 /* http://[::1]... */
778 if (!strchr (seps, *p))
780 /* Trailing garbage after []-delimited IPv6 address. */
781 error_code = PE_INVALID_HOST_NAME;
787 p = strpbrk_or_eos (p, seps);
790 ++seps; /* advance to '/' */
792 if (host_b == host_e)
794 error_code = PE_INVALID_HOST_NAME;
798 port = scheme_default_port (scheme);
801 const char *port_b, *port_e, *pp;
803 /* scheme://host:port/tralala */
807 p = strpbrk_or_eos (p, seps);
810 /* Allow empty port, as per rfc2396. */
811 if (port_b != port_e)
812 for (port = 0, pp = port_b; pp < port_e; pp++)
814 if (!c_isdigit (*pp))
816 /* http://host:12randomgarbage/blah */
818 error_code = PE_BAD_PORT_NUMBER;
821 port = 10 * port + (*pp - '0');
822 /* Check for too large port numbers here, before we have
823 a chance to overflow on bogus port values. */
826 error_code = PE_BAD_PORT_NUMBER;
831 /* Advance to the first separator *after* '/' (either ';' or '?',
832 depending on the scheme). */
835 /* Get the optional parts of URL, each part being delimited by
836 current location and the position of the next separator. */
837 #define GET_URL_PART(sepchar, var) do { \
839 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
843 GET_URL_PART ('/', path);
844 if (supported_schemes[scheme].flags & scm_has_params)
845 GET_URL_PART (';', params);
846 if (supported_schemes[scheme].flags & scm_has_query)
847 GET_URL_PART ('?', query);
848 if (supported_schemes[scheme].flags & scm_has_fragment)
849 GET_URL_PART ('#', fragment);
854 if (uname_b != uname_e)
856 /* http://user:pass@host */
858 /* uname_b uname_e */
859 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
861 error_code = PE_INVALID_USER_NAME;
866 u = xnew0 (struct url);
868 u->host = strdupdelim (host_b, host_e);
873 u->path = strdupdelim (path_b, path_e);
874 path_modified = path_simplify (scheme, u->path);
875 split_path (u->path, &u->dir, &u->file);
877 host_modified = lowercase_str (u->host);
879 /* Decode %HH sequences in host name. This is important not so much
880 to support %HH sequences in host names (which other browser
881 don't), but to support binary characters (which will have been
882 converted to %HH by reencode_escapes). */
883 if (strchr (u->host, '%'))
885 url_unescape (u->host);
886 host_modified = true;
888 /* Apply IDNA regardless of iri->utf8_encode status */
889 if (opt.enable_iri && iri)
891 char *new = idn_encode (iri, u->host);
896 host_modified = true;
902 u->params = strdupdelim (params_b, params_e);
904 u->query = strdupdelim (query_b, query_e);
906 u->fragment = strdupdelim (fragment_b, fragment_e);
908 if (opt.enable_iri || path_modified || u->fragment || host_modified || path_b == path_e)
910 /* If we suspect that a transformation has rendered what
911 url_string might return different from URL_ENCODED, rebuild
912 u->url using url_string. */
913 u->url = url_string (u, URL_AUTH_SHOW);
915 if (url_encoded != url)
916 xfree ((char *) url_encoded);
920 if (url_encoded == url)
921 u->url = xstrdup (url);
923 u->url = (char *) url_encoded;
929 /* Cleanup in case of error: */
930 if (url_encoded && url_encoded != url)
931 xfree ((char *) url_encoded);
933 /* Transmit the error code to the caller, if the caller wants to
940 /* Return the error message string from ERROR_CODE, which should have
941 been retrieved from url_parse. The error message is translated. */
944 url_error (const char *url, int error_code)
946 assert (error_code >= 0 && ((size_t) error_code) < countof (parse_errors));
948 if (error_code == PE_UNSUPPORTED_SCHEME)
951 char *scheme = xstrdup (url);
952 assert (url_has_scheme (url));
954 if ((p = strchr (scheme, ':')))
956 if (!strcasecmp (scheme, "https"))
957 error = aprintf (_("HTTPS support not compiled in"));
959 error = aprintf (_(parse_errors[error_code]), quote (scheme));
965 return xstrdup (_(parse_errors[error_code]));
968 /* Split PATH into DIR and FILE. PATH comes from the URL and is
969 expected to be URL-escaped.
971 The path is split into directory (the part up to the last slash)
972 and file (the part after the last slash), which are subsequently
976 "foo/bar/baz" "foo/bar" "baz"
977 "foo/bar/" "foo/bar" ""
979 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
981 DIR and FILE are freshly allocated. */
984 split_path (const char *path, char **dir, char **file)
986 char *last_slash = strrchr (path, '/');
990 *file = xstrdup (path);
994 *dir = strdupdelim (path, last_slash);
995 *file = xstrdup (last_slash + 1);
998 url_unescape (*file);
1001 /* Note: URL's "full path" is the path with the query string and
1002 params appended. The "fragment" (#foo) is intentionally ignored,
1003 but that might be changed. For example, if the original URL was
1004 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
1005 the full path will be "/foo/bar/baz;bullshit?querystring". */
1007 /* Return the length of the full path, without the terminating
1011 full_path_length (const struct url *url)
1015 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
1026 /* Write out the full path. */
1029 full_path_write (const struct url *url, char *where)
1031 #define FROB(el, chr) do { \
1032 char *f_el = url->el; \
1034 int l = strlen (f_el); \
1036 memcpy (where, f_el, l); \
1048 /* Public function for getting the "full path". E.g. if u->path is
1049 "foo/bar" and u->query is "param=value", full_path will be
1050 "/foo/bar?param=value". */
1053 url_full_path (const struct url *url)
1055 int length = full_path_length (url);
1056 char *full_path = xmalloc (length + 1);
1058 full_path_write (url, full_path);
1059 full_path[length] = '\0';
1064 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1065 escaping of certain characters, such as "/" and ":". Returns a
1066 count of unescaped chars. */
1069 unescape_single_char (char *str, char chr)
1071 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1072 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1073 char *h = str; /* hare */
1074 char *t = str; /* tortoise */
1075 for (; *h; h++, t++)
1077 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1088 /* Escape unsafe and reserved characters, except for the slash
1092 url_escape_dir (const char *dir)
1094 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1098 unescape_single_char (newdir, '/');
1102 /* Sync u->path and u->url with u->dir and u->file. Called after
1103 u->file or u->dir have been changed, typically by the FTP code. */
1106 sync_path (struct url *u)
1108 char *newpath, *efile, *edir;
1112 /* u->dir and u->file are not escaped. URL-escape them before
1113 reassembling them into u->path. That way, if they contain
1114 separators like '?' or even if u->file contains slashes, the
1115 path will be correctly assembled. (u->file can contain slashes
1116 if the URL specifies it with %2f, or if an FTP server returns
1118 edir = url_escape_dir (u->dir);
1119 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1122 newpath = xstrdup (efile);
1125 int dirlen = strlen (edir);
1126 int filelen = strlen (efile);
1128 /* Copy "DIR/FILE" to newpath. */
1129 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1130 memcpy (p, edir, dirlen);
1133 memcpy (p, efile, filelen);
1142 if (efile != u->file)
1145 /* Regenerate u->url as well. */
1147 u->url = url_string (u, URL_AUTH_SHOW);
1150 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1151 This way we can sync u->path and u->url when they get changed. */
1154 url_set_dir (struct url *url, const char *newdir)
1157 url->dir = xstrdup (newdir);
1162 url_set_file (struct url *url, const char *newfile)
1165 url->file = xstrdup (newfile);
1170 url_free (struct url *url)
1176 xfree_null (url->params);
1177 xfree_null (url->query);
1178 xfree_null (url->fragment);
1179 xfree_null (url->user);
1180 xfree_null (url->passwd);
1188 /* Create all the necessary directories for PATH (a file). Calls
1189 make_directory internally. */
1191 mkalldirs (const char *path)
1198 p = path + strlen (path);
1199 for (; *p != '/' && p != path; p--)
1202 /* Don't create if it's just a file. */
1203 if ((p == path) && (*p != '/'))
1205 t = strdupdelim (path, p);
1207 /* Check whether the directory exists. */
1208 if ((stat (t, &st) == 0))
1210 if (S_ISDIR (st.st_mode))
1217 /* If the dir exists as a file name, remove it first. This
1218 is *only* for Wget to work with buggy old CERN http
1219 servers. Here is the scenario: When Wget tries to
1220 retrieve a directory without a slash, e.g.
1221 http://foo/bar (bar being a directory), CERN server will
1222 not redirect it too http://foo/bar/ -- it will generate a
1223 directory listing containing links to bar/file1,
1224 bar/file2, etc. Wget will lose because it saves this
1225 HTML listing to a file `bar', so it cannot create the
1226 directory. To work around this, if the file of the same
1227 name exists, we just remove it and create the directory
1229 DEBUGP (("Removing %s because of directory danger!\n", t));
1233 res = make_directory (t);
1235 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1240 /* Functions for constructing the file name out of URL components. */
1242 /* A growable string structure, used by url_file_name and friends.
1243 This should perhaps be moved to utils.c.
1245 The idea is to have a convenient and efficient way to construct a
1246 string by having various functions append data to it. Instead of
1247 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1248 functions in questions, we pass the pointer to this struct.
1250 Functions that write to the members in this struct must make sure
1251 that base remains null terminated by calling append_null().
1256 int size; /* memory allocated */
1257 int tail; /* string length */
1260 /* Ensure that the string can accept APPEND_COUNT more characters past
1261 the current TAIL position. If necessary, this will grow the string
1262 and update its allocated size. If the string is already large
1263 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1264 #define GROW(g, append_size) do { \
1265 struct growable *G_ = g; \
1266 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1269 /* Return the tail position of the string. */
1270 #define TAIL(r) ((r)->base + (r)->tail)
1272 /* Move the tail position by APPEND_COUNT characters. */
1273 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1276 /* Append NULL to DEST. */
1278 append_null (struct growable *dest)
1284 /* Shorten DEST to LENGTH. */
1286 shorten_length (size_t length, struct growable *dest)
1288 if (length < dest->tail)
1289 dest->tail = length;
1294 /* Append CH to DEST. */
1296 append_char (char ch, struct growable *dest)
1302 TAIL_INCR (dest, 1);
1308 /* Append the string STR to DEST. */
1310 append_string (const char *str, struct growable *dest)
1312 int l = strlen (str);
1317 memcpy (TAIL (dest), str, l);
1318 TAIL_INCR (dest, l);
1326 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1327 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1328 filechr_control = 4 /* a control character, e.g. 0-31 */
1331 #define FILE_CHAR_TEST(c, mask) \
1332 ((opt.restrict_files_nonascii && !c_isascii ((unsigned char)(c))) || \
1333 (filechr_table[(unsigned char)(c)] & (mask)))
1335 /* Shorthands for the table: */
1336 #define U filechr_not_unix
1337 #define W filechr_not_windows
1338 #define C filechr_control
1343 /* Table of characters unsafe under various conditions (see above).
1345 Arguably we could also claim `%' to be unsafe, since we use it as
1346 the escape character. If we ever want to be able to reliably
1347 translate file name back to URL, this would become important
1348 crucial. Right now, it's better to be minimal in escaping. */
1350 static const unsigned char filechr_table[256] =
1352 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1353 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1354 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1355 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1356 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1357 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1358 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1359 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1360 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1361 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1362 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1363 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1364 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1365 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1366 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1367 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1369 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1370 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1371 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1372 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1374 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1375 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1376 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1377 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1385 /* FN_PORT_SEP is the separator between host and port in file names
1386 for non-standard port numbers. On Unix this is normally ':', as in
1387 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1388 because Windows can't handle ':' in file names. */
1389 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1391 /* FN_QUERY_SEP is the separator between the file name and the URL
1392 query, normally '?'. Since Windows cannot handle '?' as part of
1393 file name, we use '@' instead there. */
1394 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1395 #define FN_QUERY_SEP_STR (opt.restrict_files_os != restrict_windows ? "?" : "@")
1397 /* Quote path element, characters in [b, e), as file name, and append
1398 the quoted string to DEST. Each character is quoted as per
1399 file_unsafe_char and the corresponding table.
1401 If ESCAPED is true, the path element is considered to be
1402 URL-escaped and will be unescaped prior to inspection. */
1405 append_uri_pathel (const char *b, const char *e, bool escaped,
1406 struct growable *dest)
1412 if (opt.restrict_files_os == restrict_unix)
1413 mask = filechr_not_unix;
1415 mask = filechr_not_windows;
1416 if (opt.restrict_files_ctrl)
1417 mask |= filechr_control;
1419 /* Copy [b, e) to PATHEL and URL-unescape it. */
1423 BOUNDED_TO_ALLOCA (b, e, unescaped);
1424 url_unescape (unescaped);
1426 e = unescaped + strlen (unescaped);
1429 /* Defang ".." when found as component of path. Remember that path
1430 comes from the URL and might contain malicious input. */
1431 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1437 /* Walk the PATHEL string and check how many characters we'll need
1440 for (p = b; p < e; p++)
1441 if (FILE_CHAR_TEST (*p, mask))
1444 /* Calculate the length of the output string. e-b is the input
1445 string length. Each quoted char introduces two additional
1446 characters in the string, hence 2*quoted. */
1447 outlen = (e - b) + (2 * quoted);
1448 GROW (dest, outlen);
1452 /* If there's nothing to quote, we can simply append the string
1453 without processing it again. */
1454 memcpy (TAIL (dest), b, outlen);
1458 char *q = TAIL (dest);
1459 for (p = b; p < e; p++)
1461 if (!FILE_CHAR_TEST (*p, mask))
1465 unsigned char ch = *p;
1467 *q++ = XNUM_TO_DIGIT (ch >> 4);
1468 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1471 assert (q - TAIL (dest) == outlen);
1474 /* Perform inline case transformation if required. */
1475 if (opt.restrict_files_case == restrict_lowercase
1476 || opt.restrict_files_case == restrict_uppercase)
1479 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1481 if (opt.restrict_files_case == restrict_lowercase)
1482 *q = c_tolower (*q);
1484 *q = c_toupper (*q);
1488 TAIL_INCR (dest, outlen);
1492 /* Append to DEST the directory structure that corresponds the
1493 directory part of URL's path. For example, if the URL is
1494 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1496 Each path element ("dir1" and "dir2" in the above example) is
1497 examined, url-unescaped, and re-escaped as file name element.
1499 Additionally, it cuts as many directories from the path as
1500 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1501 will produce "bar" for the above example. For 2 or more, it will
1504 Each component of the path is quoted for use as file name. */
1507 append_dir_structure (const struct url *u, struct growable *dest)
1509 char *pathel, *next;
1510 int cut = opt.cut_dirs;
1512 /* Go through the path components, de-URL-quote them, and quote them
1513 (if necessary) as file names. */
1516 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1521 /* Ignore empty pathels. */
1525 append_char ('/', dest);
1526 append_uri_pathel (pathel, next, true, dest);
1530 /* Return a unique file name that matches the given URL as well as
1531 possible. Does not create directories on the file system. */
1534 url_file_name (const struct url *u, char *replaced_filename)
1536 struct growable fnres; /* stands for "file name result" */
1537 struct growable temp_fnres;
1540 char *fname, *unique, *fname_len_check;
1541 const char *index_filename = "index.html"; /* The default index file is index.html */
1548 temp_fnres.base = NULL;
1549 temp_fnres.size = 0;
1550 temp_fnres.tail = 0;
1552 /* If an alternative index file was defined, change index_filename */
1553 if (opt.default_page)
1554 index_filename = opt.default_page;
1557 /* Start with the directory prefix, if specified. */
1559 append_string (opt.dir_prefix, &fnres);
1561 /* If "dirstruct" is turned on (typically the case with -r), add
1562 the host and port (unless those have been turned off) and
1563 directory structure. */
1566 if (opt.protocol_directories)
1569 append_char ('/', &fnres);
1570 append_string (supported_schemes[u->scheme].name, &fnres);
1572 if (opt.add_hostdir)
1575 append_char ('/', &fnres);
1576 if (0 != strcmp (u->host, ".."))
1577 append_string (u->host, &fnres);
1579 /* Host name can come from the network; malicious DNS may
1580 allow ".." to be resolved, causing us to write to
1581 "../<file>". Defang such host names. */
1582 append_string ("%2E%2E", &fnres);
1583 if (u->port != scheme_default_port (u->scheme))
1586 number_to_string (portstr, u->port);
1587 append_char (FN_PORT_SEP, &fnres);
1588 append_string (portstr, &fnres);
1592 append_dir_structure (u, &fnres);
1595 if (!replaced_filename)
1597 /* Create the filename. */
1598 u_file = *u->file ? u->file : index_filename;
1600 /* Append "?query" to the file name, even if empty,
1601 * and create fname_len_check. */
1603 fname_len_check = concat_strings (u_file, FN_QUERY_SEP_STR, u->query, NULL);
1605 fname_len_check = strdupdelim (u_file, u_file + strlen (u_file));
1609 u_file = replaced_filename;
1610 fname_len_check = strdupdelim (u_file, u_file + strlen (u_file));
1613 append_uri_pathel (fname_len_check,
1614 fname_len_check + strlen (fname_len_check), false, &temp_fnres);
1616 /* Zero-terminate the temporary file name. */
1617 append_char ('\0', &temp_fnres);
1619 /* Check that the length of the file name is acceptable. */
1621 if (MAX_PATH > (fnres.tail + CHOMP_BUFFER + 2))
1623 max_length = MAX_PATH - (fnres.tail + CHOMP_BUFFER + 2);
1624 /* FIXME: In Windows a filename is usually limited to 255 characters.
1625 To really be accurate you could call GetVolumeInformation() to get
1626 lpMaximumComponentLength
1628 if (max_length > 255)
1638 max_length = get_max_length (fnres.base, fnres.tail, _PC_NAME_MAX) - CHOMP_BUFFER;
1640 if (max_length > 0 && strlen (temp_fnres.base) > max_length)
1642 logprintf (LOG_NOTQUIET, "The name is too long, %lu chars total.\n",
1643 (unsigned long) strlen (temp_fnres.base));
1644 logprintf (LOG_NOTQUIET, "Trying to shorten...\n");
1646 /* Shorten the file name. */
1647 temp_fnres.base[max_length] = '\0';
1649 logprintf (LOG_NOTQUIET, "New name is %s.\n", temp_fnres.base);
1652 free (fname_len_check);
1654 /* The filename has already been 'cleaned' by append_uri_pathel() above. So,
1655 * just append it. */
1657 append_char ('/', &fnres);
1658 append_string (temp_fnres.base, &fnres);
1662 /* Make a final check that the path length is acceptable? */
1663 /* TODO: check fnres.base for path length problem */
1665 free (temp_fnres.base);
1667 /* Check the cases in which the unique extensions are not used:
1668 1) Clobbering is turned off (-nc).
1669 2) Retrieval with regetting.
1670 3) Timestamping is used.
1671 4) Hierarchy is built.
1672 5) Backups are specified.
1674 The exception is the case when file does exist and is a
1675 directory (see `mkalldirs' for explanation). */
1678 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1684 unique = unique_name (fname, true);
1685 if (unique != fname)
1689 /* On VMS, alter the name as required. */
1694 unique2 = ods_conform( unique);
1695 if (unique2 != unique)
1701 #endif /* def __VMS */
1706 /* Resolve "." and ".." elements of PATH by destructively modifying
1707 PATH and return true if PATH has been modified, false otherwise.
1709 The algorithm is in spirit similar to the one described in rfc1808,
1710 although implemented differently, in one pass. To recap, path
1711 elements containing only "." are removed, and ".." is taken to mean
1712 "back up one element". Single leading and trailing slashes are
1715 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1716 test examples are provided below. If you change anything in this
1717 function, run test_path_simplify to make sure you haven't broken a
1721 path_simplify (enum url_scheme scheme, char *path)
1723 char *h = path; /* hare */
1724 char *t = path; /* tortoise */
1726 char *end = strchr (path, '\0');
1730 /* Hare should be at the beginning of a path element. */
1732 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1737 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1739 /* Handle "../" by retreating the tortoise by one path
1740 element -- but not past beggining. */
1743 /* Move backwards until T hits the beginning of the
1744 previous path element or the beginning of path. */
1745 for (--t; t > beg && t[-1] != '/'; t--)
1748 else if (scheme == SCHEME_FTP)
1750 /* If we're at the beginning, copy the "../" literally
1751 and move the beginning so a later ".." doesn't remove
1752 it. This violates RFC 3986; but we do it for FTP
1753 anyway because there is otherwise no way to get at a
1754 parent directory, when the FTP server drops us in a
1755 non-root directory (which is not uncommon). */
1764 /* A regular path element. If H hasn't advanced past T,
1765 simply skip to the next path element. Otherwise, copy
1766 the path element until the next slash. */
1769 /* Skip the path element, including the slash. */
1770 while (h < end && *h != '/')
1777 /* Copy the path element, including the final slash. */
1778 while (h < end && *h != '/')
1792 /* Return the length of URL's path. Path is considered to be
1793 terminated by one or more of the ?query or ;params or #fragment,
1794 depending on the scheme. */
1797 path_end (const char *url)
1799 enum url_scheme scheme = url_scheme (url);
1801 if (scheme == SCHEME_INVALID)
1802 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1803 /* +2 to ignore the first two separators ':' and '/' */
1804 seps = init_seps (scheme) + 2;
1805 return strpbrk_or_eos (url, seps);
1808 /* Find the last occurrence of character C in the range [b, e), or
1809 NULL, if none are present. */
1810 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1812 /* Merge BASE with LINK and return the resulting URI.
1814 Either of the URIs may be absolute or relative, complete with the
1815 host name, or path only. This tries to reasonably handle all
1816 foreseeable cases. It only employs minimal URL parsing, without
1817 knowledge of the specifics of schemes.
1819 I briefly considered making this function call path_simplify after
1820 the merging process, as rfc1738 seems to suggest. This is a bad
1821 idea for several reasons: 1) it complexifies the code, and 2)
1822 url_parse has to simplify path anyway, so it's wasteful to boot. */
1825 uri_merge (const char *base, const char *link)
1831 if (url_has_scheme (link))
1832 return xstrdup (link);
1834 /* We may not examine BASE past END. */
1835 end = path_end (base);
1836 linklength = strlen (link);
1840 /* Empty LINK points back to BASE, query string and all. */
1841 return xstrdup (base);
1843 else if (*link == '?')
1845 /* LINK points to the same location, but changes the query
1846 string. Examples: */
1847 /* uri_merge("path", "?new") -> "path?new" */
1848 /* uri_merge("path?foo", "?new") -> "path?new" */
1849 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1850 /* uri_merge("path#foo", "?new") -> "path?new" */
1851 int baselength = end - base;
1852 merge = xmalloc (baselength + linklength + 1);
1853 memcpy (merge, base, baselength);
1854 memcpy (merge + baselength, link, linklength);
1855 merge[baselength + linklength] = '\0';
1857 else if (*link == '#')
1859 /* uri_merge("path", "#new") -> "path#new" */
1860 /* uri_merge("path#foo", "#new") -> "path#new" */
1861 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1862 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1864 const char *end1 = strchr (base, '#');
1866 end1 = base + strlen (base);
1867 baselength = end1 - base;
1868 merge = xmalloc (baselength + linklength + 1);
1869 memcpy (merge, base, baselength);
1870 memcpy (merge + baselength, link, linklength);
1871 merge[baselength + linklength] = '\0';
1873 else if (*link == '/' && *(link + 1) == '/')
1875 /* LINK begins with "//" and so is a net path: we need to
1876 replace everything after (and including) the double slash
1879 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1880 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1881 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1885 const char *start_insert;
1887 /* Look for first slash. */
1888 slash = memchr (base, '/', end - base);
1889 /* If found slash and it is a double slash, then replace
1890 from this point, else default to replacing from the
1892 if (slash && *(slash + 1) == '/')
1893 start_insert = slash;
1895 start_insert = base;
1897 span = start_insert - base;
1898 merge = xmalloc (span + linklength + 1);
1900 memcpy (merge, base, span);
1901 memcpy (merge + span, link, linklength);
1902 merge[span + linklength] = '\0';
1904 else if (*link == '/')
1906 /* LINK is an absolute path: we need to replace everything
1907 after (and including) the FIRST slash with LINK.
1909 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1910 "/qux/xyzzy", our result should be
1911 "http://host/qux/xyzzy". */
1914 const char *start_insert = NULL; /* for gcc to shut up. */
1915 const char *pos = base;
1916 bool seen_slash_slash = false;
1917 /* We're looking for the first slash, but want to ignore
1920 slash = memchr (pos, '/', end - pos);
1921 if (slash && !seen_slash_slash)
1922 if (*(slash + 1) == '/')
1925 seen_slash_slash = true;
1929 /* At this point, SLASH is the location of the first / after
1930 "//", or the first slash altogether. START_INSERT is the
1931 pointer to the location where LINK will be inserted. When
1932 examining the last two examples, keep in mind that LINK
1935 if (!slash && !seen_slash_slash)
1936 /* example: "foo" */
1938 start_insert = base;
1939 else if (!slash && seen_slash_slash)
1940 /* example: "http://foo" */
1943 else if (slash && !seen_slash_slash)
1944 /* example: "foo/bar" */
1946 start_insert = base;
1947 else if (slash && seen_slash_slash)
1948 /* example: "http://something/" */
1950 start_insert = slash;
1952 span = start_insert - base;
1953 merge = xmalloc (span + linklength + 1);
1955 memcpy (merge, base, span);
1956 memcpy (merge + span, link, linklength);
1957 merge[span + linklength] = '\0';
1961 /* LINK is a relative URL: we need to replace everything
1962 after last slash (possibly empty) with LINK.
1964 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1965 our result should be "whatever/foo/qux/xyzzy". */
1966 bool need_explicit_slash = false;
1968 const char *start_insert;
1969 const char *last_slash = find_last_char (base, end, '/');
1972 /* No slash found at all. Replace what we have with LINK. */
1973 start_insert = base;
1975 else if (last_slash && last_slash >= base + 2
1976 && last_slash[-2] == ':' && last_slash[-1] == '/')
1978 /* example: http://host" */
1980 start_insert = end + 1;
1981 need_explicit_slash = true;
1985 /* example: "whatever/foo/bar" */
1987 start_insert = last_slash + 1;
1990 span = start_insert - base;
1991 merge = xmalloc (span + linklength + 1);
1993 memcpy (merge, base, span);
1994 if (need_explicit_slash)
1995 merge[span - 1] = '/';
1996 memcpy (merge + span, link, linklength);
1997 merge[span + linklength] = '\0';
2003 #define APPEND(p, s) do { \
2004 int len = strlen (s); \
2005 memcpy (p, s, len); \
2009 /* Use this instead of password when the actual password is supposed
2010 to be hidden. We intentionally use a generic string without giving
2011 away the number of characters in the password, like previous
2013 #define HIDDEN_PASSWORD "*password*"
2015 /* Recreate the URL string from the data in URL.
2017 If HIDE is true (as it is when we're calling this on a URL we plan
2018 to print, but not when calling it to canonicalize a URL for use
2019 within the program), password will be hidden. Unsafe characters in
2020 the URL will be quoted. */
2023 url_string (const struct url *url, enum url_auth_mode auth_mode)
2027 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
2029 int scheme_port = supported_schemes[url->scheme].default_port;
2030 const char *scheme_str = supported_schemes[url->scheme].leading_string;
2031 int fplen = full_path_length (url);
2033 bool brackets_around_host;
2035 assert (scheme_str != NULL);
2037 /* Make sure the user name and password are quoted. */
2040 if (auth_mode != URL_AUTH_HIDE)
2042 quoted_user = url_escape_allow_passthrough (url->user);
2045 if (auth_mode == URL_AUTH_HIDE_PASSWD)
2046 quoted_passwd = HIDDEN_PASSWORD;
2048 quoted_passwd = url_escape_allow_passthrough (url->passwd);
2053 /* In the unlikely event that the host name contains non-printable
2054 characters, quote it for displaying to the user. */
2055 quoted_host = url_escape_allow_passthrough (url->host);
2057 /* Undo the quoting of colons that URL escaping performs. IPv6
2058 addresses may legally contain colons, and in that case must be
2059 placed in square brackets. */
2060 if (quoted_host != url->host)
2061 unescape_single_char (quoted_host, ':');
2062 brackets_around_host = strchr (quoted_host, ':') != NULL;
2064 size = (strlen (scheme_str)
2065 + strlen (quoted_host)
2066 + (brackets_around_host ? 2 : 0)
2069 if (url->port != scheme_port)
2070 size += 1 + numdigit (url->port);
2073 size += 1 + strlen (quoted_user);
2075 size += 1 + strlen (quoted_passwd);
2078 p = result = xmalloc (size);
2080 APPEND (p, scheme_str);
2083 APPEND (p, quoted_user);
2087 APPEND (p, quoted_passwd);
2092 if (brackets_around_host)
2094 APPEND (p, quoted_host);
2095 if (brackets_around_host)
2097 if (url->port != scheme_port)
2100 p = number_to_string (p, url->port);
2103 full_path_write (url, p);
2107 assert (p - result == size);
2109 if (quoted_user && quoted_user != url->user)
2110 xfree (quoted_user);
2111 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
2112 && quoted_passwd != url->passwd)
2113 xfree (quoted_passwd);
2114 if (quoted_host != url->host)
2115 xfree (quoted_host);
2120 /* Return true if scheme a is similar to scheme b.
2122 Schemes are similar if they are equal. If SSL is supported, schemes
2123 are also similar if one is http (SCHEME_HTTP) and the other is https
2126 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
2131 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
2132 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
2139 getchar_from_escaped_string (const char *str, char *c)
2141 const char *p = str;
2143 assert (str && *str);
2148 if (!c_isxdigit(p[1]) || !c_isxdigit(p[2]))
2156 return 0; /* error: invalid string */
2158 *c = X2DIGITS_TO_NUM (p[1], p[2]);
2159 if (URL_RESERVED_CHAR(*c))
2177 are_urls_equal (const char *u1, const char *u2)
2188 && (pp = getchar_from_escaped_string (p, &ch1))
2189 && (qq = getchar_from_escaped_string (q, &ch2))
2190 && (c_tolower(ch1) == c_tolower(ch2)))
2196 return (*p == 0 && *q == 0 ? true : false);
2200 /* Debugging and testing support for path_simplify. */
2203 /* Debug: run path_simplify on PATH and return the result in a new
2204 string. Useful for calling from the debugger. */
2208 char *copy = xstrdup (path);
2209 path_simplify (copy);
2215 run_test (char *test, char *expected_result, enum url_scheme scheme,
2216 bool expected_change)
2218 char *test_copy = xstrdup (test);
2219 bool modified = path_simplify (scheme, test_copy);
2221 if (0 != strcmp (test_copy, expected_result))
2223 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2224 test, expected_result, test_copy);
2227 if (modified != expected_change)
2229 if (expected_change)
2230 printf ("Expected modification with path_simplify(\"%s\").\n",
2233 printf ("Expected no modification with path_simplify(\"%s\").\n",
2237 mu_assert ("", modified == expected_change);
2242 test_path_simplify (void)
2245 char *test, *result;
2246 enum url_scheme scheme;
2249 { "", "", SCHEME_HTTP, false },
2250 { ".", "", SCHEME_HTTP, true },
2251 { "./", "", SCHEME_HTTP, true },
2252 { "..", "", SCHEME_HTTP, true },
2253 { "../", "", SCHEME_HTTP, true },
2254 { "..", "..", SCHEME_FTP, false },
2255 { "../", "../", SCHEME_FTP, false },
2256 { "foo", "foo", SCHEME_HTTP, false },
2257 { "foo/bar", "foo/bar", SCHEME_HTTP, false },
2258 { "foo///bar", "foo///bar", SCHEME_HTTP, false },
2259 { "foo/.", "foo/", SCHEME_HTTP, true },
2260 { "foo/./", "foo/", SCHEME_HTTP, true },
2261 { "foo./", "foo./", SCHEME_HTTP, false },
2262 { "foo/../bar", "bar", SCHEME_HTTP, true },
2263 { "foo/../bar/", "bar/", SCHEME_HTTP, true },
2264 { "foo/bar/..", "foo/", SCHEME_HTTP, true },
2265 { "foo/bar/../x", "foo/x", SCHEME_HTTP, true },
2266 { "foo/bar/../x/", "foo/x/", SCHEME_HTTP, true },
2267 { "foo/..", "", SCHEME_HTTP, true },
2268 { "foo/../..", "", SCHEME_HTTP, true },
2269 { "foo/../../..", "", SCHEME_HTTP, true },
2270 { "foo/../../bar/../../baz", "baz", SCHEME_HTTP, true },
2271 { "foo/../..", "..", SCHEME_FTP, true },
2272 { "foo/../../..", "../..", SCHEME_FTP, true },
2273 { "foo/../../bar/../../baz", "../../baz", SCHEME_FTP, true },
2274 { "a/b/../../c", "c", SCHEME_HTTP, true },
2275 { "./a/../b", "b", SCHEME_HTTP, true }
2279 for (i = 0; i < countof (tests); i++)
2281 const char *message;
2282 char *test = tests[i].test;
2283 char *expected_result = tests[i].result;
2284 enum url_scheme scheme = tests[i].scheme;
2285 bool expected_change = tests[i].should_modify;
2286 message = run_test (test, expected_result, scheme, expected_change);
2287 if (message) return message;
2293 test_append_uri_pathel()
2300 char *expected_result;
2302 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2305 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2307 struct growable dest;
2308 const char *p = test_array[i].input;
2310 memset (&dest, 0, sizeof (dest));
2312 append_string (test_array[i].original_url, &dest);
2313 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2315 mu_assert ("test_append_uri_pathel: wrong result",
2316 strcmp (dest.base, test_array[i].expected_result) == 0);
2323 test_are_urls_equal()
2329 bool expected_result;
2331 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2332 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2333 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2334 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2335 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2336 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2339 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2341 mu_assert ("test_are_urls_equal: wrong result",
2342 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2348 #endif /* TESTING */