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)
705 iri->orig_url = xstrdup (url);
706 percent_encode = true;
710 /* XXX XXX Could that change introduce (security) bugs ??? XXX XXX*/
712 url_encoded = reencode_escapes (new_url ? new_url : url);
714 url_encoded = new_url ? new_url : url;
718 if (new_url && url_encoded != new_url)
721 p += strlen (supported_schemes[scheme].leading_string);
723 p = url_skip_credentials (p);
726 /* scheme://user:pass@host[:port]... */
729 /* We attempt to break down the URL into the components path,
730 params, query, and fragment. They are ordered like this:
732 scheme://host[:port][/path][;params][?query][#fragment] */
734 path_b = path_e = NULL;
735 params_b = params_e = NULL;
736 query_b = query_e = NULL;
737 fragment_b = fragment_e = NULL;
739 /* Initialize separators for optional parts of URL, depending on the
740 scheme. For example, FTP has params, and HTTP and HTTPS have
741 query string and fragment. */
742 seps = init_seps (scheme);
748 /* Handle IPv6 address inside square brackets. Ideally we'd
749 just look for the terminating ']', but rfc2732 mandates
750 rejecting invalid IPv6 addresses. */
752 /* The address begins after '['. */
754 host_e = strchr (host_b, ']');
758 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
763 /* Check if the IPv6 address is valid. */
764 if (!is_valid_ipv6_address(host_b, host_e))
766 error_code = PE_INVALID_IPV6_ADDRESS;
770 /* Continue parsing after the closing ']'. */
773 error_code = PE_IPV6_NOT_SUPPORTED;
777 /* The closing bracket must be followed by a separator or by the
779 /* http://[::1]... */
781 if (!strchr (seps, *p))
783 /* Trailing garbage after []-delimited IPv6 address. */
784 error_code = PE_INVALID_HOST_NAME;
790 p = strpbrk_or_eos (p, seps);
793 ++seps; /* advance to '/' */
795 if (host_b == host_e)
797 error_code = PE_INVALID_HOST_NAME;
801 port = scheme_default_port (scheme);
804 const char *port_b, *port_e, *pp;
806 /* scheme://host:port/tralala */
810 p = strpbrk_or_eos (p, seps);
813 /* Allow empty port, as per rfc2396. */
814 if (port_b != port_e)
815 for (port = 0, pp = port_b; pp < port_e; pp++)
817 if (!c_isdigit (*pp))
819 /* http://host:12randomgarbage/blah */
821 error_code = PE_BAD_PORT_NUMBER;
824 port = 10 * port + (*pp - '0');
825 /* Check for too large port numbers here, before we have
826 a chance to overflow on bogus port values. */
829 error_code = PE_BAD_PORT_NUMBER;
834 /* Advance to the first separator *after* '/' (either ';' or '?',
835 depending on the scheme). */
838 /* Get the optional parts of URL, each part being delimited by
839 current location and the position of the next separator. */
840 #define GET_URL_PART(sepchar, var) do { \
842 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
846 GET_URL_PART ('/', path);
847 if (supported_schemes[scheme].flags & scm_has_params)
848 GET_URL_PART (';', params);
849 if (supported_schemes[scheme].flags & scm_has_query)
850 GET_URL_PART ('?', query);
851 if (supported_schemes[scheme].flags & scm_has_fragment)
852 GET_URL_PART ('#', fragment);
857 if (uname_b != uname_e)
859 /* http://user:pass@host */
861 /* uname_b uname_e */
862 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
864 error_code = PE_INVALID_USER_NAME;
869 u = xnew0 (struct url);
871 u->host = strdupdelim (host_b, host_e);
876 u->path = strdupdelim (path_b, path_e);
877 path_modified = path_simplify (scheme, u->path);
878 split_path (u->path, &u->dir, &u->file);
880 host_modified = lowercase_str (u->host);
882 /* Decode %HH sequences in host name. This is important not so much
883 to support %HH sequences in host names (which other browser
884 don't), but to support binary characters (which will have been
885 converted to %HH by reencode_escapes). */
886 if (strchr (u->host, '%'))
888 url_unescape (u->host);
889 host_modified = true;
891 /* Apply IDNA regardless of iri->utf8_encode status */
892 if (opt.enable_iri && iri)
894 char *new = idn_encode (iri, u->host);
899 host_modified = true;
905 u->params = strdupdelim (params_b, params_e);
907 u->query = strdupdelim (query_b, query_e);
909 u->fragment = strdupdelim (fragment_b, fragment_e);
911 if (opt.enable_iri || path_modified || u->fragment || host_modified || path_b == path_e)
913 /* If we suspect that a transformation has rendered what
914 url_string might return different from URL_ENCODED, rebuild
915 u->url using url_string. */
916 u->url = url_string (u, URL_AUTH_SHOW);
918 if (url_encoded != url)
919 xfree ((char *) url_encoded);
923 if (url_encoded == url)
924 u->url = xstrdup (url);
926 u->url = (char *) url_encoded;
932 /* Cleanup in case of error: */
933 if (url_encoded && url_encoded != url)
934 xfree ((char *) url_encoded);
936 /* Transmit the error code to the caller, if the caller wants to
943 /* Return the error message string from ERROR_CODE, which should have
944 been retrieved from url_parse. The error message is translated. */
947 url_error (const char *url, int error_code)
949 assert (error_code >= 0 && ((size_t) error_code) < countof (parse_errors));
951 if (error_code == PE_UNSUPPORTED_SCHEME)
954 char *scheme = xstrdup (url);
955 assert (url_has_scheme (url));
957 if ((p = strchr (scheme, ':')))
959 if (!strcasecmp (scheme, "https"))
960 error = aprintf (_("HTTPS support not compiled in"));
962 error = aprintf (_(parse_errors[error_code]), quote (scheme));
968 return xstrdup (_(parse_errors[error_code]));
971 /* Split PATH into DIR and FILE. PATH comes from the URL and is
972 expected to be URL-escaped.
974 The path is split into directory (the part up to the last slash)
975 and file (the part after the last slash), which are subsequently
979 "foo/bar/baz" "foo/bar" "baz"
980 "foo/bar/" "foo/bar" ""
982 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
984 DIR and FILE are freshly allocated. */
987 split_path (const char *path, char **dir, char **file)
989 char *last_slash = strrchr (path, '/');
993 *file = xstrdup (path);
997 *dir = strdupdelim (path, last_slash);
998 *file = xstrdup (last_slash + 1);
1000 url_unescape (*dir);
1001 url_unescape (*file);
1004 /* Note: URL's "full path" is the path with the query string and
1005 params appended. The "fragment" (#foo) is intentionally ignored,
1006 but that might be changed. For example, if the original URL was
1007 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
1008 the full path will be "/foo/bar/baz;bullshit?querystring". */
1010 /* Return the length of the full path, without the terminating
1014 full_path_length (const struct url *url)
1018 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
1029 /* Write out the full path. */
1032 full_path_write (const struct url *url, char *where)
1034 #define FROB(el, chr) do { \
1035 char *f_el = url->el; \
1037 int l = strlen (f_el); \
1039 memcpy (where, f_el, l); \
1051 /* Public function for getting the "full path". E.g. if u->path is
1052 "foo/bar" and u->query is "param=value", full_path will be
1053 "/foo/bar?param=value". */
1056 url_full_path (const struct url *url)
1058 int length = full_path_length (url);
1059 char *full_path = xmalloc (length + 1);
1061 full_path_write (url, full_path);
1062 full_path[length] = '\0';
1067 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1068 escaping of certain characters, such as "/" and ":". Returns a
1069 count of unescaped chars. */
1072 unescape_single_char (char *str, char chr)
1074 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1075 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1076 char *h = str; /* hare */
1077 char *t = str; /* tortoise */
1078 for (; *h; h++, t++)
1080 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1091 /* Escape unsafe and reserved characters, except for the slash
1095 url_escape_dir (const char *dir)
1097 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1101 unescape_single_char (newdir, '/');
1105 /* Sync u->path and u->url with u->dir and u->file. Called after
1106 u->file or u->dir have been changed, typically by the FTP code. */
1109 sync_path (struct url *u)
1111 char *newpath, *efile, *edir;
1115 /* u->dir and u->file are not escaped. URL-escape them before
1116 reassembling them into u->path. That way, if they contain
1117 separators like '?' or even if u->file contains slashes, the
1118 path will be correctly assembled. (u->file can contain slashes
1119 if the URL specifies it with %2f, or if an FTP server returns
1121 edir = url_escape_dir (u->dir);
1122 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1125 newpath = xstrdup (efile);
1128 int dirlen = strlen (edir);
1129 int filelen = strlen (efile);
1131 /* Copy "DIR/FILE" to newpath. */
1132 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1133 memcpy (p, edir, dirlen);
1136 memcpy (p, efile, filelen);
1145 if (efile != u->file)
1148 /* Regenerate u->url as well. */
1150 u->url = url_string (u, URL_AUTH_SHOW);
1153 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1154 This way we can sync u->path and u->url when they get changed. */
1157 url_set_dir (struct url *url, const char *newdir)
1160 url->dir = xstrdup (newdir);
1165 url_set_file (struct url *url, const char *newfile)
1168 url->file = xstrdup (newfile);
1173 url_free (struct url *url)
1179 xfree_null (url->params);
1180 xfree_null (url->query);
1181 xfree_null (url->fragment);
1182 xfree_null (url->user);
1183 xfree_null (url->passwd);
1191 /* Create all the necessary directories for PATH (a file). Calls
1192 make_directory internally. */
1194 mkalldirs (const char *path)
1201 p = path + strlen (path);
1202 for (; *p != '/' && p != path; p--)
1205 /* Don't create if it's just a file. */
1206 if ((p == path) && (*p != '/'))
1208 t = strdupdelim (path, p);
1210 /* Check whether the directory exists. */
1211 if ((stat (t, &st) == 0))
1213 if (S_ISDIR (st.st_mode))
1220 /* If the dir exists as a file name, remove it first. This
1221 is *only* for Wget to work with buggy old CERN http
1222 servers. Here is the scenario: When Wget tries to
1223 retrieve a directory without a slash, e.g.
1224 http://foo/bar (bar being a directory), CERN server will
1225 not redirect it too http://foo/bar/ -- it will generate a
1226 directory listing containing links to bar/file1,
1227 bar/file2, etc. Wget will lose because it saves this
1228 HTML listing to a file `bar', so it cannot create the
1229 directory. To work around this, if the file of the same
1230 name exists, we just remove it and create the directory
1232 DEBUGP (("Removing %s because of directory danger!\n", t));
1236 res = make_directory (t);
1238 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1243 /* Functions for constructing the file name out of URL components. */
1245 /* A growable string structure, used by url_file_name and friends.
1246 This should perhaps be moved to utils.c.
1248 The idea is to have a convenient and efficient way to construct a
1249 string by having various functions append data to it. Instead of
1250 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1251 functions in questions, we pass the pointer to this struct.
1253 Functions that write to the members in this struct must make sure
1254 that base remains null terminated by calling append_null().
1259 int size; /* memory allocated */
1260 int tail; /* string length */
1263 /* Ensure that the string can accept APPEND_COUNT more characters past
1264 the current TAIL position. If necessary, this will grow the string
1265 and update its allocated size. If the string is already large
1266 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1267 #define GROW(g, append_size) do { \
1268 struct growable *G_ = g; \
1269 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1272 /* Return the tail position of the string. */
1273 #define TAIL(r) ((r)->base + (r)->tail)
1275 /* Move the tail position by APPEND_COUNT characters. */
1276 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1279 /* Append NULL to DEST. */
1281 append_null (struct growable *dest)
1287 /* Append CH to DEST. */
1289 append_char (char ch, struct growable *dest)
1295 TAIL_INCR (dest, 1);
1301 /* Append the string STR to DEST. */
1303 append_string (const char *str, struct growable *dest)
1305 int l = strlen (str);
1310 memcpy (TAIL (dest), str, l);
1311 TAIL_INCR (dest, l);
1319 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1320 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1321 filechr_control = 4 /* a control character, e.g. 0-31 */
1324 #define FILE_CHAR_TEST(c, mask) \
1325 ((opt.restrict_files_nonascii && !c_isascii ((unsigned char)(c))) || \
1326 (filechr_table[(unsigned char)(c)] & (mask)))
1328 /* Shorthands for the table: */
1329 #define U filechr_not_unix
1330 #define W filechr_not_windows
1331 #define C filechr_control
1336 /* Table of characters unsafe under various conditions (see above).
1338 Arguably we could also claim `%' to be unsafe, since we use it as
1339 the escape character. If we ever want to be able to reliably
1340 translate file name back to URL, this would become important
1341 crucial. Right now, it's better to be minimal in escaping. */
1343 static const unsigned char filechr_table[256] =
1345 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1346 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1347 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1348 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1349 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1350 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1351 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1352 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1353 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1354 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1355 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1356 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1357 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1358 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1359 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1360 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1362 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1363 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1364 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1365 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1367 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1368 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1369 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1370 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1378 /* FN_PORT_SEP is the separator between host and port in file names
1379 for non-standard port numbers. On Unix this is normally ':', as in
1380 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1381 because Windows can't handle ':' in file names. */
1382 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1384 /* FN_QUERY_SEP is the separator between the file name and the URL
1385 query, normally '?'. Since Windows cannot handle '?' as part of
1386 file name, we use '@' instead there. */
1387 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1388 #define FN_QUERY_SEP_STR (opt.restrict_files_os != restrict_windows ? "?" : "@")
1390 /* Quote path element, characters in [b, e), as file name, and append
1391 the quoted string to DEST. Each character is quoted as per
1392 file_unsafe_char and the corresponding table.
1394 If ESCAPED is true, the path element is considered to be
1395 URL-escaped and will be unescaped prior to inspection. */
1398 append_uri_pathel (const char *b, const char *e, bool escaped,
1399 struct growable *dest)
1405 if (opt.restrict_files_os == restrict_unix)
1406 mask = filechr_not_unix;
1408 mask = filechr_not_windows;
1409 if (opt.restrict_files_ctrl)
1410 mask |= filechr_control;
1412 /* Copy [b, e) to PATHEL and URL-unescape it. */
1416 BOUNDED_TO_ALLOCA (b, e, unescaped);
1417 url_unescape (unescaped);
1419 e = unescaped + strlen (unescaped);
1422 /* Defang ".." when found as component of path. Remember that path
1423 comes from the URL and might contain malicious input. */
1424 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1430 /* Walk the PATHEL string and check how many characters we'll need
1433 for (p = b; p < e; p++)
1434 if (FILE_CHAR_TEST (*p, mask))
1437 /* Calculate the length of the output string. e-b is the input
1438 string length. Each quoted char introduces two additional
1439 characters in the string, hence 2*quoted. */
1440 outlen = (e - b) + (2 * quoted);
1441 GROW (dest, outlen);
1445 /* If there's nothing to quote, we can simply append the string
1446 without processing it again. */
1447 memcpy (TAIL (dest), b, outlen);
1451 char *q = TAIL (dest);
1452 for (p = b; p < e; p++)
1454 if (!FILE_CHAR_TEST (*p, mask))
1458 unsigned char ch = *p;
1460 *q++ = XNUM_TO_DIGIT (ch >> 4);
1461 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1464 assert (q - TAIL (dest) == outlen);
1467 /* Perform inline case transformation if required. */
1468 if (opt.restrict_files_case == restrict_lowercase
1469 || opt.restrict_files_case == restrict_uppercase)
1472 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1474 if (opt.restrict_files_case == restrict_lowercase)
1475 *q = c_tolower (*q);
1477 *q = c_toupper (*q);
1481 TAIL_INCR (dest, outlen);
1485 /* Append to DEST the directory structure that corresponds the
1486 directory part of URL's path. For example, if the URL is
1487 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1489 Each path element ("dir1" and "dir2" in the above example) is
1490 examined, url-unescaped, and re-escaped as file name element.
1492 Additionally, it cuts as many directories from the path as
1493 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1494 will produce "bar" for the above example. For 2 or more, it will
1497 Each component of the path is quoted for use as file name. */
1500 append_dir_structure (const struct url *u, struct growable *dest)
1502 char *pathel, *next;
1503 int cut = opt.cut_dirs;
1505 /* Go through the path components, de-URL-quote them, and quote them
1506 (if necessary) as file names. */
1509 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1514 /* Ignore empty pathels. */
1518 append_char ('/', dest);
1519 append_uri_pathel (pathel, next, true, dest);
1523 /* Return a unique file name that matches the given URL as well as
1524 possible. Does not create directories on the file system. */
1527 url_file_name (const struct url *u, char *replaced_filename)
1529 struct growable fnres; /* stands for "file name result" */
1530 struct growable temp_fnres;
1533 char *fname, *unique, *fname_len_check;
1534 const char *index_filename = "index.html"; /* The default index file is index.html */
1541 temp_fnres.base = NULL;
1542 temp_fnres.size = 0;
1543 temp_fnres.tail = 0;
1545 /* If an alternative index file was defined, change index_filename */
1546 if (opt.default_page)
1547 index_filename = opt.default_page;
1550 /* Start with the directory prefix, if specified. */
1552 append_string (opt.dir_prefix, &fnres);
1554 /* If "dirstruct" is turned on (typically the case with -r), add
1555 the host and port (unless those have been turned off) and
1556 directory structure. */
1559 if (opt.protocol_directories)
1562 append_char ('/', &fnres);
1563 append_string (supported_schemes[u->scheme].name, &fnres);
1565 if (opt.add_hostdir)
1568 append_char ('/', &fnres);
1569 if (0 != strcmp (u->host, ".."))
1570 append_string (u->host, &fnres);
1572 /* Host name can come from the network; malicious DNS may
1573 allow ".." to be resolved, causing us to write to
1574 "../<file>". Defang such host names. */
1575 append_string ("%2E%2E", &fnres);
1576 if (u->port != scheme_default_port (u->scheme))
1579 number_to_string (portstr, u->port);
1580 append_char (FN_PORT_SEP, &fnres);
1581 append_string (portstr, &fnres);
1585 append_dir_structure (u, &fnres);
1588 if (!replaced_filename)
1590 /* Create the filename. */
1591 u_file = *u->file ? u->file : index_filename;
1593 /* Append "?query" to the file name, even if empty,
1594 * and create fname_len_check. */
1596 fname_len_check = concat_strings (u_file, FN_QUERY_SEP_STR, u->query, NULL);
1598 fname_len_check = strdupdelim (u_file, u_file + strlen (u_file));
1602 u_file = replaced_filename;
1603 fname_len_check = strdupdelim (u_file, u_file + strlen (u_file));
1606 append_uri_pathel (fname_len_check,
1607 fname_len_check + strlen (fname_len_check), false, &temp_fnres);
1609 /* Zero-terminate the temporary file name. */
1610 append_char ('\0', &temp_fnres);
1612 /* Check that the length of the file name is acceptable. */
1614 if (MAX_PATH > (fnres.tail + CHOMP_BUFFER + 2))
1616 max_length = MAX_PATH - (fnres.tail + CHOMP_BUFFER + 2);
1617 /* FIXME: In Windows a filename is usually limited to 255 characters.
1618 To really be accurate you could call GetVolumeInformation() to get
1619 lpMaximumComponentLength
1621 if (max_length > 255)
1631 max_length = get_max_length (fnres.base, fnres.tail, _PC_NAME_MAX) - CHOMP_BUFFER;
1633 if (max_length > 0 && strlen (temp_fnres.base) > max_length)
1635 logprintf (LOG_NOTQUIET, "The name is too long, %lu chars total.\n",
1636 (unsigned long) strlen (temp_fnres.base));
1637 logprintf (LOG_NOTQUIET, "Trying to shorten...\n");
1639 /* Shorten the file name. */
1640 temp_fnres.base[max_length] = '\0';
1642 logprintf (LOG_NOTQUIET, "New name is %s.\n", temp_fnres.base);
1645 free (fname_len_check);
1647 /* The filename has already been 'cleaned' by append_uri_pathel() above. So,
1648 * just append it. */
1650 append_char ('/', &fnres);
1651 append_string (temp_fnres.base, &fnres);
1655 /* Make a final check that the path length is acceptable? */
1656 /* TODO: check fnres.base for path length problem */
1658 free (temp_fnres.base);
1660 /* Check the cases in which the unique extensions are not used:
1661 1) Clobbering is turned off (-nc).
1662 2) Retrieval with regetting.
1663 3) Timestamping is used.
1664 4) Hierarchy is built.
1665 5) Backups are specified.
1667 The exception is the case when file does exist and is a
1668 directory (see `mkalldirs' for explanation). */
1671 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1677 unique = unique_name (fname, true);
1678 if (unique != fname)
1682 /* On VMS, alter the name as required. */
1687 unique2 = ods_conform( unique);
1688 if (unique2 != unique)
1694 #endif /* def __VMS */
1699 /* Resolve "." and ".." elements of PATH by destructively modifying
1700 PATH and return true if PATH has been modified, false otherwise.
1702 The algorithm is in spirit similar to the one described in rfc1808,
1703 although implemented differently, in one pass. To recap, path
1704 elements containing only "." are removed, and ".." is taken to mean
1705 "back up one element". Single leading and trailing slashes are
1708 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1709 test examples are provided below. If you change anything in this
1710 function, run test_path_simplify to make sure you haven't broken a
1714 path_simplify (enum url_scheme scheme, char *path)
1716 char *h = path; /* hare */
1717 char *t = path; /* tortoise */
1719 char *end = strchr (path, '\0');
1723 /* Hare should be at the beginning of a path element. */
1725 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1730 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1732 /* Handle "../" by retreating the tortoise by one path
1733 element -- but not past beggining. */
1736 /* Move backwards until T hits the beginning of the
1737 previous path element or the beginning of path. */
1738 for (--t; t > beg && t[-1] != '/'; t--)
1741 else if (scheme == SCHEME_FTP)
1743 /* If we're at the beginning, copy the "../" literally
1744 and move the beginning so a later ".." doesn't remove
1745 it. This violates RFC 3986; but we do it for FTP
1746 anyway because there is otherwise no way to get at a
1747 parent directory, when the FTP server drops us in a
1748 non-root directory (which is not uncommon). */
1757 /* A regular path element. If H hasn't advanced past T,
1758 simply skip to the next path element. Otherwise, copy
1759 the path element until the next slash. */
1762 /* Skip the path element, including the slash. */
1763 while (h < end && *h != '/')
1770 /* Copy the path element, including the final slash. */
1771 while (h < end && *h != '/')
1785 /* Return the length of URL's path. Path is considered to be
1786 terminated by one or more of the ?query or ;params or #fragment,
1787 depending on the scheme. */
1790 path_end (const char *url)
1792 enum url_scheme scheme = url_scheme (url);
1794 if (scheme == SCHEME_INVALID)
1795 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1796 /* +2 to ignore the first two separators ':' and '/' */
1797 seps = init_seps (scheme) + 2;
1798 return strpbrk_or_eos (url, seps);
1801 /* Find the last occurrence of character C in the range [b, e), or
1802 NULL, if none are present. */
1803 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1805 /* Merge BASE with LINK and return the resulting URI.
1807 Either of the URIs may be absolute or relative, complete with the
1808 host name, or path only. This tries to reasonably handle all
1809 foreseeable cases. It only employs minimal URL parsing, without
1810 knowledge of the specifics of schemes.
1812 I briefly considered making this function call path_simplify after
1813 the merging process, as rfc1738 seems to suggest. This is a bad
1814 idea for several reasons: 1) it complexifies the code, and 2)
1815 url_parse has to simplify path anyway, so it's wasteful to boot. */
1818 uri_merge (const char *base, const char *link)
1824 if (url_has_scheme (link))
1825 return xstrdup (link);
1827 /* We may not examine BASE past END. */
1828 end = path_end (base);
1829 linklength = strlen (link);
1833 /* Empty LINK points back to BASE, query string and all. */
1834 return xstrdup (base);
1836 else if (*link == '?')
1838 /* LINK points to the same location, but changes the query
1839 string. Examples: */
1840 /* uri_merge("path", "?new") -> "path?new" */
1841 /* uri_merge("path?foo", "?new") -> "path?new" */
1842 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1843 /* uri_merge("path#foo", "?new") -> "path?new" */
1844 int baselength = end - base;
1845 merge = xmalloc (baselength + linklength + 1);
1846 memcpy (merge, base, baselength);
1847 memcpy (merge + baselength, link, linklength);
1848 merge[baselength + linklength] = '\0';
1850 else if (*link == '#')
1852 /* uri_merge("path", "#new") -> "path#new" */
1853 /* uri_merge("path#foo", "#new") -> "path#new" */
1854 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1855 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1857 const char *end1 = strchr (base, '#');
1859 end1 = base + strlen (base);
1860 baselength = end1 - base;
1861 merge = xmalloc (baselength + linklength + 1);
1862 memcpy (merge, base, baselength);
1863 memcpy (merge + baselength, link, linklength);
1864 merge[baselength + linklength] = '\0';
1866 else if (*link == '/' && *(link + 1) == '/')
1868 /* LINK begins with "//" and so is a net path: we need to
1869 replace everything after (and including) the double slash
1872 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1873 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1874 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1878 const char *start_insert;
1880 /* Look for first slash. */
1881 slash = memchr (base, '/', end - base);
1882 /* If found slash and it is a double slash, then replace
1883 from this point, else default to replacing from the
1885 if (slash && *(slash + 1) == '/')
1886 start_insert = slash;
1888 start_insert = base;
1890 span = start_insert - base;
1891 merge = xmalloc (span + linklength + 1);
1893 memcpy (merge, base, span);
1894 memcpy (merge + span, link, linklength);
1895 merge[span + linklength] = '\0';
1897 else if (*link == '/')
1899 /* LINK is an absolute path: we need to replace everything
1900 after (and including) the FIRST slash with LINK.
1902 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1903 "/qux/xyzzy", our result should be
1904 "http://host/qux/xyzzy". */
1907 const char *start_insert = NULL; /* for gcc to shut up. */
1908 const char *pos = base;
1909 bool seen_slash_slash = false;
1910 /* We're looking for the first slash, but want to ignore
1913 slash = memchr (pos, '/', end - pos);
1914 if (slash && !seen_slash_slash)
1915 if (*(slash + 1) == '/')
1918 seen_slash_slash = true;
1922 /* At this point, SLASH is the location of the first / after
1923 "//", or the first slash altogether. START_INSERT is the
1924 pointer to the location where LINK will be inserted. When
1925 examining the last two examples, keep in mind that LINK
1928 if (!slash && !seen_slash_slash)
1929 /* example: "foo" */
1931 start_insert = base;
1932 else if (!slash && seen_slash_slash)
1933 /* example: "http://foo" */
1936 else if (slash && !seen_slash_slash)
1937 /* example: "foo/bar" */
1939 start_insert = base;
1940 else if (slash && seen_slash_slash)
1941 /* example: "http://something/" */
1943 start_insert = slash;
1945 span = start_insert - base;
1946 merge = xmalloc (span + linklength + 1);
1948 memcpy (merge, base, span);
1949 memcpy (merge + span, link, linklength);
1950 merge[span + linklength] = '\0';
1954 /* LINK is a relative URL: we need to replace everything
1955 after last slash (possibly empty) with LINK.
1957 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1958 our result should be "whatever/foo/qux/xyzzy". */
1959 bool need_explicit_slash = false;
1961 const char *start_insert;
1962 const char *last_slash = find_last_char (base, end, '/');
1965 /* No slash found at all. Replace what we have with LINK. */
1966 start_insert = base;
1968 else if (last_slash && last_slash >= base + 2
1969 && last_slash[-2] == ':' && last_slash[-1] == '/')
1971 /* example: http://host" */
1973 start_insert = end + 1;
1974 need_explicit_slash = true;
1978 /* example: "whatever/foo/bar" */
1980 start_insert = last_slash + 1;
1983 span = start_insert - base;
1984 merge = xmalloc (span + linklength + 1);
1986 memcpy (merge, base, span);
1987 if (need_explicit_slash)
1988 merge[span - 1] = '/';
1989 memcpy (merge + span, link, linklength);
1990 merge[span + linklength] = '\0';
1996 #define APPEND(p, s) do { \
1997 int len = strlen (s); \
1998 memcpy (p, s, len); \
2002 /* Use this instead of password when the actual password is supposed
2003 to be hidden. We intentionally use a generic string without giving
2004 away the number of characters in the password, like previous
2006 #define HIDDEN_PASSWORD "*password*"
2008 /* Recreate the URL string from the data in URL.
2010 If HIDE is true (as it is when we're calling this on a URL we plan
2011 to print, but not when calling it to canonicalize a URL for use
2012 within the program), password will be hidden. Unsafe characters in
2013 the URL will be quoted. */
2016 url_string (const struct url *url, enum url_auth_mode auth_mode)
2020 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
2022 int scheme_port = supported_schemes[url->scheme].default_port;
2023 const char *scheme_str = supported_schemes[url->scheme].leading_string;
2024 int fplen = full_path_length (url);
2026 bool brackets_around_host;
2028 assert (scheme_str != NULL);
2030 /* Make sure the user name and password are quoted. */
2033 if (auth_mode != URL_AUTH_HIDE)
2035 quoted_user = url_escape_allow_passthrough (url->user);
2038 if (auth_mode == URL_AUTH_HIDE_PASSWD)
2039 quoted_passwd = HIDDEN_PASSWORD;
2041 quoted_passwd = url_escape_allow_passthrough (url->passwd);
2046 /* In the unlikely event that the host name contains non-printable
2047 characters, quote it for displaying to the user. */
2048 quoted_host = url_escape_allow_passthrough (url->host);
2050 /* Undo the quoting of colons that URL escaping performs. IPv6
2051 addresses may legally contain colons, and in that case must be
2052 placed in square brackets. */
2053 if (quoted_host != url->host)
2054 unescape_single_char (quoted_host, ':');
2055 brackets_around_host = strchr (quoted_host, ':') != NULL;
2057 size = (strlen (scheme_str)
2058 + strlen (quoted_host)
2059 + (brackets_around_host ? 2 : 0)
2062 if (url->port != scheme_port)
2063 size += 1 + numdigit (url->port);
2066 size += 1 + strlen (quoted_user);
2068 size += 1 + strlen (quoted_passwd);
2071 p = result = xmalloc (size);
2073 APPEND (p, scheme_str);
2076 APPEND (p, quoted_user);
2080 APPEND (p, quoted_passwd);
2085 if (brackets_around_host)
2087 APPEND (p, quoted_host);
2088 if (brackets_around_host)
2090 if (url->port != scheme_port)
2093 p = number_to_string (p, url->port);
2096 full_path_write (url, p);
2100 assert (p - result == size);
2102 if (quoted_user && quoted_user != url->user)
2103 xfree (quoted_user);
2104 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
2105 && quoted_passwd != url->passwd)
2106 xfree (quoted_passwd);
2107 if (quoted_host != url->host)
2108 xfree (quoted_host);
2113 /* Return true if scheme a is similar to scheme b.
2115 Schemes are similar if they are equal. If SSL is supported, schemes
2116 are also similar if one is http (SCHEME_HTTP) and the other is https
2119 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
2124 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
2125 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
2132 getchar_from_escaped_string (const char *str, char *c)
2134 const char *p = str;
2136 assert (str && *str);
2141 if (!c_isxdigit(p[1]) || !c_isxdigit(p[2]))
2149 return 0; /* error: invalid string */
2151 *c = X2DIGITS_TO_NUM (p[1], p[2]);
2152 if (URL_RESERVED_CHAR(*c))
2170 are_urls_equal (const char *u1, const char *u2)
2181 && (pp = getchar_from_escaped_string (p, &ch1))
2182 && (qq = getchar_from_escaped_string (q, &ch2))
2183 && (c_tolower(ch1) == c_tolower(ch2)))
2189 return (*p == 0 && *q == 0 ? true : false);
2193 /* Debugging and testing support for path_simplify. */
2196 /* Debug: run path_simplify on PATH and return the result in a new
2197 string. Useful for calling from the debugger. */
2201 char *copy = xstrdup (path);
2202 path_simplify (copy);
2208 run_test (char *test, char *expected_result, enum url_scheme scheme,
2209 bool expected_change)
2211 char *test_copy = xstrdup (test);
2212 bool modified = path_simplify (scheme, test_copy);
2214 if (0 != strcmp (test_copy, expected_result))
2216 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2217 test, expected_result, test_copy);
2220 if (modified != expected_change)
2222 if (expected_change)
2223 printf ("Expected modification with path_simplify(\"%s\").\n",
2226 printf ("Expected no modification with path_simplify(\"%s\").\n",
2230 mu_assert ("", modified == expected_change);
2235 test_path_simplify (void)
2238 char *test, *result;
2239 enum url_scheme scheme;
2242 { "", "", SCHEME_HTTP, false },
2243 { ".", "", SCHEME_HTTP, true },
2244 { "./", "", SCHEME_HTTP, true },
2245 { "..", "", SCHEME_HTTP, true },
2246 { "../", "", SCHEME_HTTP, true },
2247 { "..", "..", SCHEME_FTP, false },
2248 { "../", "../", SCHEME_FTP, false },
2249 { "foo", "foo", SCHEME_HTTP, false },
2250 { "foo/bar", "foo/bar", SCHEME_HTTP, false },
2251 { "foo///bar", "foo///bar", SCHEME_HTTP, false },
2252 { "foo/.", "foo/", SCHEME_HTTP, true },
2253 { "foo/./", "foo/", SCHEME_HTTP, true },
2254 { "foo./", "foo./", SCHEME_HTTP, false },
2255 { "foo/../bar", "bar", SCHEME_HTTP, true },
2256 { "foo/../bar/", "bar/", SCHEME_HTTP, true },
2257 { "foo/bar/..", "foo/", SCHEME_HTTP, true },
2258 { "foo/bar/../x", "foo/x", SCHEME_HTTP, true },
2259 { "foo/bar/../x/", "foo/x/", SCHEME_HTTP, true },
2260 { "foo/..", "", SCHEME_HTTP, true },
2261 { "foo/../..", "", SCHEME_HTTP, true },
2262 { "foo/../../..", "", SCHEME_HTTP, true },
2263 { "foo/../../bar/../../baz", "baz", SCHEME_HTTP, true },
2264 { "foo/../..", "..", SCHEME_FTP, true },
2265 { "foo/../../..", "../..", SCHEME_FTP, true },
2266 { "foo/../../bar/../../baz", "../../baz", SCHEME_FTP, true },
2267 { "a/b/../../c", "c", SCHEME_HTTP, true },
2268 { "./a/../b", "b", SCHEME_HTTP, true }
2272 for (i = 0; i < countof (tests); i++)
2274 const char *message;
2275 char *test = tests[i].test;
2276 char *expected_result = tests[i].result;
2277 enum url_scheme scheme = tests[i].scheme;
2278 bool expected_change = tests[i].should_modify;
2279 message = run_test (test, expected_result, scheme, expected_change);
2280 if (message) return message;
2286 test_append_uri_pathel()
2293 char *expected_result;
2295 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2298 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2300 struct growable dest;
2301 const char *p = test_array[i].input;
2303 memset (&dest, 0, sizeof (dest));
2305 append_string (test_array[i].original_url, &dest);
2306 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2308 mu_assert ("test_append_uri_pathel: wrong result",
2309 strcmp (dest.base, test_array[i].expected_result) == 0);
2316 test_are_urls_equal()
2322 bool expected_result;
2324 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2325 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2326 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2327 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2328 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2329 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2332 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2334 mu_assert ("test_are_urls_equal: wrong result",
2335 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2341 #endif /* TESTING */