2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
3 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
5 This file is part of GNU Wget.
7 GNU Wget is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or (at
10 your option) any later version.
12 GNU Wget is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with Wget. If not, see <http://www.gnu.org/licenses/>.
20 Additional permission under GNU GPL version 3 section 7
22 If you modify this program, or any covered work, by linking or
23 combining it with the OpenSSL project's OpenSSL library (or a
24 modified version of that library), containing parts covered by the
25 terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
26 grants you additional permission to convey the resulting work.
27 Corresponding Source for a non-source form of such a combination
28 shall include the source code for the parts of OpenSSL used as well
29 as that of the covered work. */
44 #include "host.h" /* for is_valid_ipv6_address */
51 scm_disabled = 1, /* for https when OpenSSL fails to init. */
52 scm_has_params = 2, /* whether scheme has ;params */
53 scm_has_query = 4, /* whether scheme has ?query */
54 scm_has_fragment = 8 /* whether scheme has #fragment */
59 /* Short name of the scheme, such as "http" or "ftp". */
61 /* Leading string that identifies the scheme, such as "https://". */
62 const char *leading_string;
63 /* Default port of the scheme when none is specified. */
69 /* Supported schemes: */
70 static struct scheme_data supported_schemes[] =
72 { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment },
74 { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment },
76 { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment },
82 /* Forward declarations: */
84 static bool path_simplify (enum url_scheme, char *);
86 /* Support for escaping and unescaping of URL strings. */
88 /* Table of "reserved" and "unsafe" characters. Those terms are
89 rfc1738-speak, as such largely obsoleted by rfc2396 and later
90 specs, but the general idea remains.
92 A reserved character is the one that you can't decode without
93 changing the meaning of the URL. For example, you can't decode
94 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
95 path components is different. Non-reserved characters can be
96 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
97 unsafe characters are loosely based on rfc1738, plus "$" and ",",
98 as recommended by rfc2396, and minus "~", which is very frequently
99 used (and sometimes unrecognized as %7E by broken servers).
101 An unsafe character is the one that should be encoded when URLs are
102 placed in foreign environments. E.g. space and newline are unsafe
103 in HTTP contexts because HTTP uses them as separator and line
104 terminator, so they must be encoded to %20 and %0A respectively.
105 "*" is unsafe in shell context, etc.
107 We determine whether a character is unsafe through static table
108 lookup. This code assumes ASCII character set and 8-bit chars. */
111 /* rfc1738 reserved chars + "$" and ",". */
114 /* rfc1738 unsafe chars, plus non-printables. */
118 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
119 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
120 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
122 /* Shorthands for the table: */
123 #define R urlchr_reserved
124 #define U urlchr_unsafe
127 static const unsigned char urlchr_table[256] =
129 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
130 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
131 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
132 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
133 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
134 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
135 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
136 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
137 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
138 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
139 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
140 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
141 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
142 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
143 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
144 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
146 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
147 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
148 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
149 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
151 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
152 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
153 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
154 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
160 /* URL-unescape the string S.
162 This is done by transforming the sequences "%HH" to the character
163 represented by the hexadecimal digits HH. If % is not followed by
164 two hexadecimal digits, it is inserted literally.
166 The transformation is done in place. If you need the original
167 string intact, make a copy before calling this function. */
170 url_unescape (char *s)
172 char *t = s; /* t - tortoise */
173 char *h = s; /* h - hare */
185 /* Do nothing if '%' is not followed by two hex digits. */
186 if (!h[1] || !h[2] || !(c_isxdigit (h[1]) && c_isxdigit (h[2])))
188 c = X2DIGITS_TO_NUM (h[1], h[2]);
189 /* Don't unescape %00 because there is no way to insert it
190 into a C string without effectively truncating it. */
200 /* The core of url_escape_* functions. Escapes the characters that
201 match the provided mask in urlchr_table.
203 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
204 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
205 allocated string will be returned in all cases. */
208 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
215 for (p1 = s; *p1; p1++)
216 if (urlchr_test (*p1, mask))
217 addition += 2; /* Two more characters (hex digits) */
220 return allow_passthrough ? (char *)s : xstrdup (s);
222 newlen = (p1 - s) + addition;
223 newstr = xmalloc (newlen + 1);
229 /* Quote the characters that match the test mask. */
230 if (urlchr_test (*p1, mask))
232 unsigned char c = *p1++;
234 *p2++ = XNUM_TO_DIGIT (c >> 4);
235 *p2++ = XNUM_TO_DIGIT (c & 0xf);
240 assert (p2 - newstr == newlen);
246 /* URL-escape the unsafe characters (see urlchr_table) in a given
247 string, returning a freshly allocated string. */
250 url_escape (const char *s)
252 return url_escape_1 (s, urlchr_unsafe, false);
255 /* URL-escape the unsafe and reserved characters (see urlchr_table) in
256 a given string, returning a freshly allocated string. */
259 url_escape_unsafe_and_reserved (const char *s)
261 return url_escape_1 (s, urlchr_unsafe|urlchr_reserved, false);
264 /* URL-escape the unsafe characters (see urlchr_table) in a given
265 string. If no characters are unsafe, S is returned. */
268 url_escape_allow_passthrough (const char *s)
270 return url_escape_1 (s, urlchr_unsafe, true);
273 /* Decide whether the char at position P needs to be encoded. (It is
274 not enough to pass a single char *P because the function may need
275 to inspect the surrounding context.)
277 Return true if the char should be escaped as %XX, false otherwise. */
280 char_needs_escaping (const char *p)
284 if (c_isxdigit (*(p + 1)) && c_isxdigit (*(p + 2)))
287 /* Garbled %.. sequence: encode `%'. */
290 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
296 /* Translate a %-escaped (but possibly non-conformant) input string S
297 into a %-escaped (and conformant) output string. If no characters
298 are encoded or decoded, return the same string S; otherwise, return
299 a freshly allocated string with the new contents.
301 After a URL has been run through this function, the protocols that
302 use `%' as the quote character can use the resulting string as-is,
303 while those that don't can use url_unescape to get to the intended
304 data. This function is stable: once the input is transformed,
305 further transformations of the result yield the same output.
307 Let's discuss why this function is needed.
309 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
310 a raw space character would mess up the HTTP request, it needs to
311 be quoted, like this:
313 GET /abc%20def HTTP/1.0
315 It would appear that the unsafe chars need to be quoted, for
316 example with url_escape. But what if we're requested to download
317 `abc%20def'? url_escape transforms "%" to "%25", which would leave
318 us with `abc%2520def'. This is incorrect -- since %-escapes are
319 part of URL syntax, "%20" is the correct way to denote a literal
320 space on the Wget command line. This leads to the conclusion that
321 in that case Wget should not call url_escape, but leave the `%20'
322 as is. This is clearly contradictory, but it only gets worse.
324 What if the requested URI is `abc%20 def'? If we call url_escape,
325 we end up with `/abc%2520%20def', which is almost certainly not
326 intended. If we don't call url_escape, we are left with the
327 embedded space and cannot complete the request. What the user
328 meant was for Wget to request `/abc%20%20def', and this is where
329 reencode_escapes kicks in.
331 Wget used to solve this by first decoding %-quotes, and then
332 encoding all the "unsafe" characters found in the resulting string.
333 This was wrong because it didn't preserve certain URL special
334 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
335 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
336 whether we considered `+' reserved (it is). One of these results
337 is inevitable because by the second step we would lose information
338 on whether the `+' was originally encoded or not. Both results
339 were wrong because in CGI parameters + means space, while %2B means
340 literal plus. reencode_escapes correctly translates the above to
341 "a%2B+b", i.e. returns the original string.
343 This function uses a modified version of the algorithm originally
344 proposed by Anon Sricharoenchai:
346 * Encode all "unsafe" characters, except those that are also
347 "reserved", to %XX. See urlchr_table for which characters are
350 * Encode the "%" characters not followed by two hex digits to
353 * Pass through all other characters and %XX escapes as-is. (Up to
354 Wget 1.10 this decoded %XX escapes corresponding to "safe"
355 characters, but that was obtrusive and broke some servers.)
359 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
361 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
365 "foo bar" -> "foo%20bar"
366 "foo%20bar" -> "foo%20bar"
367 "foo %20bar" -> "foo%20%20bar"
368 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
369 "foo%25%20bar" -> "foo%25%20bar"
370 "foo%2%20bar" -> "foo%252%20bar"
371 "foo+bar" -> "foo+bar" (plus is reserved!)
372 "foo%2b+bar" -> "foo%2b+bar" */
375 reencode_escapes (const char *s)
381 int encode_count = 0;
383 /* First pass: inspect the string to see if there's anything to do,
384 and to calculate the new length. */
385 for (p1 = s; *p1; p1++)
386 if (char_needs_escaping (p1))
390 /* The string is good as it is. */
391 return (char *) s; /* C const model sucks. */
394 /* Each encoding adds two characters (hex digits). */
395 newlen = oldlen + 2 * encode_count;
396 newstr = xmalloc (newlen + 1);
398 /* Second pass: copy the string to the destination address, encoding
399 chars when needed. */
404 if (char_needs_escaping (p1))
406 unsigned char c = *p1++;
408 *p2++ = XNUM_TO_DIGIT (c >> 4);
409 *p2++ = XNUM_TO_DIGIT (c & 0xf);
415 assert (p2 - newstr == newlen);
419 /* Returns the scheme type if the scheme is supported, or
420 SCHEME_INVALID if not. */
423 url_scheme (const char *url)
427 for (i = 0; supported_schemes[i].leading_string; i++)
428 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
429 strlen (supported_schemes[i].leading_string)))
431 if (!(supported_schemes[i].flags & scm_disabled))
432 return (enum url_scheme) i;
434 return SCHEME_INVALID;
437 return SCHEME_INVALID;
440 #define SCHEME_CHAR(ch) (c_isalnum (ch) || (ch) == '-' || (ch) == '+')
442 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
443 currently implemented, it returns true if URL begins with
447 url_has_scheme (const char *url)
451 /* The first char must be a scheme char. */
452 if (!*p || !SCHEME_CHAR (*p))
455 /* Followed by 0 or more scheme chars. */
456 while (*p && SCHEME_CHAR (*p))
458 /* Terminated by ':'. */
463 scheme_default_port (enum url_scheme scheme)
465 return supported_schemes[scheme].default_port;
469 scheme_disable (enum url_scheme scheme)
471 supported_schemes[scheme].flags |= scm_disabled;
474 /* Skip the username and password, if present in the URL. The
475 function should *not* be called with the complete URL, but with the
476 portion after the scheme.
478 If no username and password are found, return URL. */
481 url_skip_credentials (const char *url)
483 /* Look for '@' that comes before terminators, such as '/', '?',
485 const char *p = (const char *)strpbrk (url, "@/?#;");
491 /* Parse credentials contained in [BEG, END). The region is expected
492 to have come from a URL and is unescaped. */
495 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
501 return false; /* empty user name */
503 colon = memchr (beg, ':', end - beg);
505 return false; /* again empty user name */
509 *passwd = strdupdelim (colon + 1, end);
511 url_unescape (*passwd);
518 *user = strdupdelim (beg, userend);
519 url_unescape (*user);
523 /* Used by main.c: detect URLs written using the "shorthand" URL forms
524 originally popularized by Netscape and NcFTP. HTTP shorthands look
527 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
528 www.foo.com[:port] -> http://www.foo.com[:port]
530 FTP shorthands look like this:
532 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
533 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
535 If the URL needs not or cannot be rewritten, return NULL. */
538 rewrite_shorthand_url (const char *url)
543 if (url_scheme (url) != SCHEME_INVALID)
546 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
548 p = strpbrk (url, ":/");
552 /* If we're looking at "://", it means the URL uses a scheme we
553 don't support, which may include "https" when compiled without
554 SSL support. Don't bogusly rewrite such URLs. */
555 if (p && p[0] == ':' && p[1] == '/' && p[2] == '/')
560 /* Colon indicates ftp, as in foo.bar.com:path. Check for
561 special case of http port number ("localhost:10000"). */
562 int digits = strspn (p + 1, "0123456789");
563 if (digits && (p[1 + digits] == '/' || p[1 + digits] == '\0'))
566 /* Turn "foo.bar.com:path" to "ftp://foo.bar.com/path". */
567 ret = aprintf ("ftp://%s", url);
568 ret[6 + (p - url)] = '/';
573 /* Just prepend "http://" to URL. */
574 ret = aprintf ("http://%s", url);
579 static void split_path (const char *, char **, char **);
581 /* Like strpbrk, with the exception that it returns the pointer to the
582 terminating zero (end-of-string aka "eos") if no matching character
586 strpbrk_or_eos (const char *s, const char *accept)
588 char *p = strpbrk (s, accept);
590 p = strchr (s, '\0');
594 /* Turn STR into lowercase; return true if a character was actually
598 lowercase_str (char *str)
600 bool changed = false;
602 if (c_isupper (*str))
605 *str = c_tolower (*str);
611 init_seps (enum url_scheme scheme)
613 static char seps[8] = ":/";
615 int flags = supported_schemes[scheme].flags;
617 if (flags & scm_has_params)
619 if (flags & scm_has_query)
621 if (flags & scm_has_fragment)
627 static const char *parse_errors[] = {
628 #define PE_NO_ERROR 0
630 #define PE_UNSUPPORTED_SCHEME 1
631 N_("Unsupported scheme %s"),
632 #define PE_INVALID_HOST_NAME 2
633 N_("Invalid host name"),
634 #define PE_BAD_PORT_NUMBER 3
635 N_("Bad port number"),
636 #define PE_INVALID_USER_NAME 4
637 N_("Invalid user name"),
638 #define PE_UNTERMINATED_IPV6_ADDRESS 5
639 N_("Unterminated IPv6 numeric address"),
640 #define PE_IPV6_NOT_SUPPORTED 6
641 N_("IPv6 addresses not supported"),
642 #define PE_INVALID_IPV6_ADDRESS 7
643 N_("Invalid IPv6 numeric address")
648 Return a new struct url if successful, NULL on error. In case of
649 error, and if ERROR is not NULL, also set *ERROR to the appropriate
652 url_parse (const char *url, int *error, struct iri *iri)
656 bool path_modified, host_modified;
658 enum url_scheme scheme;
661 const char *uname_b, *uname_e;
662 const char *host_b, *host_e;
663 const char *path_b, *path_e;
664 const char *params_b, *params_e;
665 const char *query_b, *query_e;
666 const char *fragment_b, *fragment_e;
669 char *user = NULL, *passwd = NULL;
671 char *url_encoded = NULL, *new_url = NULL;
675 scheme = url_scheme (url);
676 if (scheme == SCHEME_INVALID)
678 error_code = PE_UNSUPPORTED_SCHEME;
682 if (iri && iri->utf8_encode)
684 url_unescape ((char *) url);
685 iri->utf8_encode = remote_to_utf8 (iri, url, (const char **) &new_url);
686 if (!iri->utf8_encode)
690 url_encoded = reencode_escapes (new_url ? new_url : url);
693 if (new_url && url_encoded != new_url)
696 p += strlen (supported_schemes[scheme].leading_string);
698 p = url_skip_credentials (p);
701 /* scheme://user:pass@host[:port]... */
704 /* We attempt to break down the URL into the components path,
705 params, query, and fragment. They are ordered like this:
707 scheme://host[:port][/path][;params][?query][#fragment] */
709 path_b = path_e = NULL;
710 params_b = params_e = NULL;
711 query_b = query_e = NULL;
712 fragment_b = fragment_e = NULL;
714 /* Initialize separators for optional parts of URL, depending on the
715 scheme. For example, FTP has params, and HTTP and HTTPS have
716 query string and fragment. */
717 seps = init_seps (scheme);
723 /* Handle IPv6 address inside square brackets. Ideally we'd
724 just look for the terminating ']', but rfc2732 mandates
725 rejecting invalid IPv6 addresses. */
727 /* The address begins after '['. */
729 host_e = strchr (host_b, ']');
733 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
738 /* Check if the IPv6 address is valid. */
739 if (!is_valid_ipv6_address(host_b, host_e))
741 error_code = PE_INVALID_IPV6_ADDRESS;
745 /* Continue parsing after the closing ']'. */
748 error_code = PE_IPV6_NOT_SUPPORTED;
752 /* The closing bracket must be followed by a separator or by the
754 /* http://[::1]... */
756 if (!strchr (seps, *p))
758 /* Trailing garbage after []-delimited IPv6 address. */
759 error_code = PE_INVALID_HOST_NAME;
765 p = strpbrk_or_eos (p, seps);
768 ++seps; /* advance to '/' */
770 if (host_b == host_e)
772 error_code = PE_INVALID_HOST_NAME;
776 port = scheme_default_port (scheme);
779 const char *port_b, *port_e, *pp;
781 /* scheme://host:port/tralala */
785 p = strpbrk_or_eos (p, seps);
788 /* Allow empty port, as per rfc2396. */
789 if (port_b != port_e)
790 for (port = 0, pp = port_b; pp < port_e; pp++)
792 if (!c_isdigit (*pp))
794 /* http://host:12randomgarbage/blah */
796 error_code = PE_BAD_PORT_NUMBER;
799 port = 10 * port + (*pp - '0');
800 /* Check for too large port numbers here, before we have
801 a chance to overflow on bogus port values. */
804 error_code = PE_BAD_PORT_NUMBER;
809 /* Advance to the first separator *after* '/' (either ';' or '?',
810 depending on the scheme). */
813 /* Get the optional parts of URL, each part being delimited by
814 current location and the position of the next separator. */
815 #define GET_URL_PART(sepchar, var) do { \
817 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
821 GET_URL_PART ('/', path);
822 if (supported_schemes[scheme].flags & scm_has_params)
823 GET_URL_PART (';', params);
824 if (supported_schemes[scheme].flags & scm_has_query)
825 GET_URL_PART ('?', query);
826 if (supported_schemes[scheme].flags & scm_has_fragment)
827 GET_URL_PART ('#', fragment);
832 if (uname_b != uname_e)
834 /* http://user:pass@host */
836 /* uname_b uname_e */
837 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
839 error_code = PE_INVALID_USER_NAME;
844 u = xnew0 (struct url);
846 u->host = strdupdelim (host_b, host_e);
851 u->path = strdupdelim (path_b, path_e);
852 path_modified = path_simplify (scheme, u->path);
853 split_path (u->path, &u->dir, &u->file);
855 host_modified = lowercase_str (u->host);
857 /* Decode %HH sequences in host name. This is important not so much
858 to support %HH sequences in host names (which other browser
859 don't), but to support binary characters (which will have been
860 converted to %HH by reencode_escapes). */
861 if (strchr (u->host, '%'))
863 url_unescape (u->host);
864 host_modified = true;
866 /* Apply IDNA regardless of iri->utf8_encode status */
867 if (opt.enable_iri && iri)
869 char *new = idn_encode (iri, u->host);
874 host_modified = true;
880 u->params = strdupdelim (params_b, params_e);
882 u->query = strdupdelim (query_b, query_e);
884 u->fragment = strdupdelim (fragment_b, fragment_e);
886 if (opt.enable_iri || path_modified || u->fragment || host_modified || path_b == path_e)
888 /* If we suspect that a transformation has rendered what
889 url_string might return different from URL_ENCODED, rebuild
890 u->url using url_string. */
891 u->url = url_string (u, URL_AUTH_SHOW);
893 if (url_encoded != url)
894 xfree ((char *) url_encoded);
898 if (url_encoded == url)
899 u->url = xstrdup (url);
901 u->url = url_encoded;
907 /* Cleanup in case of error: */
908 if (url_encoded && url_encoded != url)
911 /* Transmit the error code to the caller, if the caller wants to
918 /* Return the error message string from ERROR_CODE, which should have
919 been retrieved from url_parse. The error message is translated. */
922 url_error (const char *url, int error_code)
924 assert (error_code >= 0 && ((size_t) error_code) < countof (parse_errors));
926 if (error_code == PE_UNSUPPORTED_SCHEME)
929 char *scheme = xstrdup (url);
930 assert (url_has_scheme (url));
932 if ((p = strchr (scheme, ':')))
934 if (!strcasecmp (scheme, "https"))
935 error = aprintf (_("HTTPS support not compiled in"));
937 error = aprintf (_(parse_errors[error_code]), quote (scheme));
943 return xstrdup (_(parse_errors[error_code]));
946 /* Split PATH into DIR and FILE. PATH comes from the URL and is
947 expected to be URL-escaped.
949 The path is split into directory (the part up to the last slash)
950 and file (the part after the last slash), which are subsequently
954 "foo/bar/baz" "foo/bar" "baz"
955 "foo/bar/" "foo/bar" ""
957 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
959 DIR and FILE are freshly allocated. */
962 split_path (const char *path, char **dir, char **file)
964 char *last_slash = strrchr (path, '/');
968 *file = xstrdup (path);
972 *dir = strdupdelim (path, last_slash);
973 *file = xstrdup (last_slash + 1);
976 url_unescape (*file);
979 /* Note: URL's "full path" is the path with the query string and
980 params appended. The "fragment" (#foo) is intentionally ignored,
981 but that might be changed. For example, if the original URL was
982 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
983 the full path will be "/foo/bar/baz;bullshit?querystring". */
985 /* Return the length of the full path, without the terminating
989 full_path_length (const struct url *url)
993 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
1004 /* Write out the full path. */
1007 full_path_write (const struct url *url, char *where)
1009 #define FROB(el, chr) do { \
1010 char *f_el = url->el; \
1012 int l = strlen (f_el); \
1014 memcpy (where, f_el, l); \
1026 /* Public function for getting the "full path". E.g. if u->path is
1027 "foo/bar" and u->query is "param=value", full_path will be
1028 "/foo/bar?param=value". */
1031 url_full_path (const struct url *url)
1033 int length = full_path_length (url);
1034 char *full_path = xmalloc (length + 1);
1036 full_path_write (url, full_path);
1037 full_path[length] = '\0';
1042 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1043 escaping of certain characters, such as "/" and ":". Returns a
1044 count of unescaped chars. */
1047 unescape_single_char (char *str, char chr)
1049 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1050 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1051 char *h = str; /* hare */
1052 char *t = str; /* tortoise */
1053 for (; *h; h++, t++)
1055 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1066 /* Escape unsafe and reserved characters, except for the slash
1070 url_escape_dir (const char *dir)
1072 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1076 unescape_single_char (newdir, '/');
1080 /* Sync u->path and u->url with u->dir and u->file. Called after
1081 u->file or u->dir have been changed, typically by the FTP code. */
1084 sync_path (struct url *u)
1086 char *newpath, *efile, *edir;
1090 /* u->dir and u->file are not escaped. URL-escape them before
1091 reassembling them into u->path. That way, if they contain
1092 separators like '?' or even if u->file contains slashes, the
1093 path will be correctly assembled. (u->file can contain slashes
1094 if the URL specifies it with %2f, or if an FTP server returns
1096 edir = url_escape_dir (u->dir);
1097 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1100 newpath = xstrdup (efile);
1103 int dirlen = strlen (edir);
1104 int filelen = strlen (efile);
1106 /* Copy "DIR/FILE" to newpath. */
1107 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1108 memcpy (p, edir, dirlen);
1111 memcpy (p, efile, filelen);
1120 if (efile != u->file)
1123 /* Regenerate u->url as well. */
1125 u->url = url_string (u, URL_AUTH_SHOW);
1128 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1129 This way we can sync u->path and u->url when they get changed. */
1132 url_set_dir (struct url *url, const char *newdir)
1135 url->dir = xstrdup (newdir);
1140 url_set_file (struct url *url, const char *newfile)
1143 url->file = xstrdup (newfile);
1148 url_free (struct url *url)
1154 xfree_null (url->params);
1155 xfree_null (url->query);
1156 xfree_null (url->fragment);
1157 xfree_null (url->user);
1158 xfree_null (url->passwd);
1166 /* Create all the necessary directories for PATH (a file). Calls
1167 make_directory internally. */
1169 mkalldirs (const char *path)
1176 p = path + strlen (path);
1177 for (; *p != '/' && p != path; p--)
1180 /* Don't create if it's just a file. */
1181 if ((p == path) && (*p != '/'))
1183 t = strdupdelim (path, p);
1185 /* Check whether the directory exists. */
1186 if ((stat (t, &st) == 0))
1188 if (S_ISDIR (st.st_mode))
1195 /* If the dir exists as a file name, remove it first. This
1196 is *only* for Wget to work with buggy old CERN http
1197 servers. Here is the scenario: When Wget tries to
1198 retrieve a directory without a slash, e.g.
1199 http://foo/bar (bar being a directory), CERN server will
1200 not redirect it too http://foo/bar/ -- it will generate a
1201 directory listing containing links to bar/file1,
1202 bar/file2, etc. Wget will lose because it saves this
1203 HTML listing to a file `bar', so it cannot create the
1204 directory. To work around this, if the file of the same
1205 name exists, we just remove it and create the directory
1207 DEBUGP (("Removing %s because of directory danger!\n", t));
1211 res = make_directory (t);
1213 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1218 /* Functions for constructing the file name out of URL components. */
1220 /* A growable string structure, used by url_file_name and friends.
1221 This should perhaps be moved to utils.c.
1223 The idea is to have a convenient and efficient way to construct a
1224 string by having various functions append data to it. Instead of
1225 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1226 functions in questions, we pass the pointer to this struct. */
1234 /* Ensure that the string can accept APPEND_COUNT more characters past
1235 the current TAIL position. If necessary, this will grow the string
1236 and update its allocated size. If the string is already large
1237 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1238 #define GROW(g, append_size) do { \
1239 struct growable *G_ = g; \
1240 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1243 /* Return the tail position of the string. */
1244 #define TAIL(r) ((r)->base + (r)->tail)
1246 /* Move the tail position by APPEND_COUNT characters. */
1247 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1249 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1253 append_string (const char *str, struct growable *dest)
1255 int l = strlen (str);
1257 memcpy (TAIL (dest), str, l);
1258 TAIL_INCR (dest, l);
1261 /* Append CH to DEST. For example, append_char (0, DEST)
1262 zero-terminates DEST. */
1265 append_char (char ch, struct growable *dest)
1269 TAIL_INCR (dest, 1);
1273 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1274 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1275 filechr_control = 4 /* a control character, e.g. 0-31 */
1278 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1280 /* Shorthands for the table: */
1281 #define U filechr_not_unix
1282 #define W filechr_not_windows
1283 #define C filechr_control
1288 /* Table of characters unsafe under various conditions (see above).
1290 Arguably we could also claim `%' to be unsafe, since we use it as
1291 the escape character. If we ever want to be able to reliably
1292 translate file name back to URL, this would become important
1293 crucial. Right now, it's better to be minimal in escaping. */
1295 static const unsigned char filechr_table[256] =
1297 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1298 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1299 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1300 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1301 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1302 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1303 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1304 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1305 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1306 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1307 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1308 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1309 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1310 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1311 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1312 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1314 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1315 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1316 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1317 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1319 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1320 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1321 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1322 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1330 /* FN_PORT_SEP is the separator between host and port in file names
1331 for non-standard port numbers. On Unix this is normally ':', as in
1332 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1333 because Windows can't handle ':' in file names. */
1334 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1336 /* FN_QUERY_SEP is the separator between the file name and the URL
1337 query, normally '?'. Since Windows cannot handle '?' as part of
1338 file name, we use '@' instead there. */
1339 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1341 /* Quote path element, characters in [b, e), as file name, and append
1342 the quoted string to DEST. Each character is quoted as per
1343 file_unsafe_char and the corresponding table.
1345 If ESCAPED is true, the path element is considered to be
1346 URL-escaped and will be unescaped prior to inspection. */
1349 append_uri_pathel (const char *b, const char *e, bool escaped,
1350 struct growable *dest)
1356 if (opt.restrict_files_os == restrict_unix)
1357 mask = filechr_not_unix;
1359 mask = filechr_not_windows;
1360 if (opt.restrict_files_ctrl)
1361 mask |= filechr_control;
1363 /* Copy [b, e) to PATHEL and URL-unescape it. */
1367 BOUNDED_TO_ALLOCA (b, e, unescaped);
1368 url_unescape (unescaped);
1370 e = unescaped + strlen (unescaped);
1373 /* Defang ".." when found as component of path. Remember that path
1374 comes from the URL and might contain malicious input. */
1375 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1381 /* Walk the PATHEL string and check how many characters we'll need
1384 for (p = b; p < e; p++)
1385 if (FILE_CHAR_TEST (*p, mask))
1388 /* Calculate the length of the output string. e-b is the input
1389 string length. Each quoted char introduces two additional
1390 characters in the string, hence 2*quoted. */
1391 outlen = (e - b) + (2 * quoted);
1392 GROW (dest, outlen);
1396 /* If there's nothing to quote, we can simply append the string
1397 without processing it again. */
1398 memcpy (TAIL (dest), b, outlen);
1402 char *q = TAIL (dest);
1403 for (p = b; p < e; p++)
1405 if (!FILE_CHAR_TEST (*p, mask))
1409 unsigned char ch = *p;
1411 *q++ = XNUM_TO_DIGIT (ch >> 4);
1412 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1415 assert (q - TAIL (dest) == outlen);
1418 /* Perform inline case transformation if required. */
1419 if (opt.restrict_files_case == restrict_lowercase
1420 || opt.restrict_files_case == restrict_uppercase)
1423 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1425 if (opt.restrict_files_case == restrict_lowercase)
1426 *q = c_tolower (*q);
1428 *q = c_toupper (*q);
1432 TAIL_INCR (dest, outlen);
1435 /* Append to DEST the directory structure that corresponds the
1436 directory part of URL's path. For example, if the URL is
1437 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1439 Each path element ("dir1" and "dir2" in the above example) is
1440 examined, url-unescaped, and re-escaped as file name element.
1442 Additionally, it cuts as many directories from the path as
1443 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1444 will produce "bar" for the above example. For 2 or more, it will
1447 Each component of the path is quoted for use as file name. */
1450 append_dir_structure (const struct url *u, struct growable *dest)
1452 char *pathel, *next;
1453 int cut = opt.cut_dirs;
1455 /* Go through the path components, de-URL-quote them, and quote them
1456 (if necessary) as file names. */
1459 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1464 /* Ignore empty pathels. */
1468 append_char ('/', dest);
1469 append_uri_pathel (pathel, next, true, dest);
1473 /* Return a unique file name that matches the given URL as good as
1474 possible. Does not create directories on the file system. */
1477 url_file_name (const struct url *u)
1479 struct growable fnres; /* stands for "file name result" */
1481 const char *u_file, *u_query;
1482 char *fname, *unique;
1483 char *index_filename = "index.html"; /* The default index file is index.html */
1489 /* If an alternative index file was defined, change index_filename */
1490 if (opt.default_page)
1491 index_filename = opt.default_page;
1494 /* Start with the directory prefix, if specified. */
1496 append_string (opt.dir_prefix, &fnres);
1498 /* If "dirstruct" is turned on (typically the case with -r), add
1499 the host and port (unless those have been turned off) and
1500 directory structure. */
1503 if (opt.protocol_directories)
1506 append_char ('/', &fnres);
1507 append_string (supported_schemes[u->scheme].name, &fnres);
1509 if (opt.add_hostdir)
1512 append_char ('/', &fnres);
1513 if (0 != strcmp (u->host, ".."))
1514 append_string (u->host, &fnres);
1516 /* Host name can come from the network; malicious DNS may
1517 allow ".." to be resolved, causing us to write to
1518 "../<file>". Defang such host names. */
1519 append_string ("%2E%2E", &fnres);
1520 if (u->port != scheme_default_port (u->scheme))
1523 number_to_string (portstr, u->port);
1524 append_char (FN_PORT_SEP, &fnres);
1525 append_string (portstr, &fnres);
1529 append_dir_structure (u, &fnres);
1532 /* Add the file name. */
1534 append_char ('/', &fnres);
1535 u_file = *u->file ? u->file : index_filename;
1536 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1538 /* Append "?query" to the file name. */
1539 u_query = u->query && *u->query ? u->query : NULL;
1542 append_char (FN_QUERY_SEP, &fnres);
1543 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1546 /* Zero-terminate the file name. */
1547 append_char ('\0', &fnres);
1551 /* Check the cases in which the unique extensions are not used:
1552 1) Clobbering is turned off (-nc).
1553 2) Retrieval with regetting.
1554 3) Timestamping is used.
1555 4) Hierarchy is built.
1557 The exception is the case when file does exist and is a
1558 directory (see `mkalldirs' for explanation). */
1560 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1561 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1564 unique = unique_name (fname, true);
1565 if (unique != fname)
1570 /* Resolve "." and ".." elements of PATH by destructively modifying
1571 PATH and return true if PATH has been modified, false otherwise.
1573 The algorithm is in spirit similar to the one described in rfc1808,
1574 although implemented differently, in one pass. To recap, path
1575 elements containing only "." are removed, and ".." is taken to mean
1576 "back up one element". Single leading and trailing slashes are
1579 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1580 test examples are provided below. If you change anything in this
1581 function, run test_path_simplify to make sure you haven't broken a
1585 path_simplify (enum url_scheme scheme, char *path)
1587 char *h = path; /* hare */
1588 char *t = path; /* tortoise */
1590 char *end = strchr (path, '\0');
1594 /* Hare should be at the beginning of a path element. */
1596 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1601 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1603 /* Handle "../" by retreating the tortoise by one path
1604 element -- but not past beggining. */
1607 /* Move backwards until T hits the beginning of the
1608 previous path element or the beginning of path. */
1609 for (--t; t > beg && t[-1] != '/'; t--)
1612 else if (scheme == SCHEME_FTP)
1614 /* If we're at the beginning, copy the "../" literally
1615 and move the beginning so a later ".." doesn't remove
1616 it. This violates RFC 3986; but we do it for FTP
1617 anyway because there is otherwise no way to get at a
1618 parent directory, when the FTP server drops us in a
1619 non-root directory (which is not uncommon). */
1628 /* A regular path element. If H hasn't advanced past T,
1629 simply skip to the next path element. Otherwise, copy
1630 the path element until the next slash. */
1633 /* Skip the path element, including the slash. */
1634 while (h < end && *h != '/')
1641 /* Copy the path element, including the final slash. */
1642 while (h < end && *h != '/')
1656 /* Return the length of URL's path. Path is considered to be
1657 terminated by one or more of the ?query or ;params or #fragment,
1658 depending on the scheme. */
1661 path_end (const char *url)
1663 enum url_scheme scheme = url_scheme (url);
1665 if (scheme == SCHEME_INVALID)
1666 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1667 /* +2 to ignore the first two separators ':' and '/' */
1668 seps = init_seps (scheme) + 2;
1669 return strpbrk_or_eos (url, seps);
1672 /* Find the last occurrence of character C in the range [b, e), or
1673 NULL, if none are present. */
1674 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1676 /* Merge BASE with LINK and return the resulting URI.
1678 Either of the URIs may be absolute or relative, complete with the
1679 host name, or path only. This tries to reasonably handle all
1680 foreseeable cases. It only employs minimal URL parsing, without
1681 knowledge of the specifics of schemes.
1683 I briefly considered making this function call path_simplify after
1684 the merging process, as rfc1738 seems to suggest. This is a bad
1685 idea for several reasons: 1) it complexifies the code, and 2)
1686 url_parse has to simplify path anyway, so it's wasteful to boot. */
1689 uri_merge (const char *base, const char *link)
1695 if (url_has_scheme (link))
1696 return xstrdup (link);
1698 /* We may not examine BASE past END. */
1699 end = path_end (base);
1700 linklength = strlen (link);
1704 /* Empty LINK points back to BASE, query string and all. */
1705 return xstrdup (base);
1707 else if (*link == '?')
1709 /* LINK points to the same location, but changes the query
1710 string. Examples: */
1711 /* uri_merge("path", "?new") -> "path?new" */
1712 /* uri_merge("path?foo", "?new") -> "path?new" */
1713 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1714 /* uri_merge("path#foo", "?new") -> "path?new" */
1715 int baselength = end - base;
1716 merge = xmalloc (baselength + linklength + 1);
1717 memcpy (merge, base, baselength);
1718 memcpy (merge + baselength, link, linklength);
1719 merge[baselength + linklength] = '\0';
1721 else if (*link == '#')
1723 /* uri_merge("path", "#new") -> "path#new" */
1724 /* uri_merge("path#foo", "#new") -> "path#new" */
1725 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1726 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1728 const char *end1 = strchr (base, '#');
1730 end1 = base + strlen (base);
1731 baselength = end1 - base;
1732 merge = xmalloc (baselength + linklength + 1);
1733 memcpy (merge, base, baselength);
1734 memcpy (merge + baselength, link, linklength);
1735 merge[baselength + linklength] = '\0';
1737 else if (*link == '/' && *(link + 1) == '/')
1739 /* LINK begins with "//" and so is a net path: we need to
1740 replace everything after (and including) the double slash
1743 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1744 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1745 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1749 const char *start_insert;
1751 /* Look for first slash. */
1752 slash = memchr (base, '/', end - base);
1753 /* If found slash and it is a double slash, then replace
1754 from this point, else default to replacing from the
1756 if (slash && *(slash + 1) == '/')
1757 start_insert = slash;
1759 start_insert = base;
1761 span = start_insert - base;
1762 merge = xmalloc (span + linklength + 1);
1764 memcpy (merge, base, span);
1765 memcpy (merge + span, link, linklength);
1766 merge[span + linklength] = '\0';
1768 else if (*link == '/')
1770 /* LINK is an absolute path: we need to replace everything
1771 after (and including) the FIRST slash with LINK.
1773 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1774 "/qux/xyzzy", our result should be
1775 "http://host/qux/xyzzy". */
1778 const char *start_insert = NULL; /* for gcc to shut up. */
1779 const char *pos = base;
1780 bool seen_slash_slash = false;
1781 /* We're looking for the first slash, but want to ignore
1784 slash = memchr (pos, '/', end - pos);
1785 if (slash && !seen_slash_slash)
1786 if (*(slash + 1) == '/')
1789 seen_slash_slash = true;
1793 /* At this point, SLASH is the location of the first / after
1794 "//", or the first slash altogether. START_INSERT is the
1795 pointer to the location where LINK will be inserted. When
1796 examining the last two examples, keep in mind that LINK
1799 if (!slash && !seen_slash_slash)
1800 /* example: "foo" */
1802 start_insert = base;
1803 else if (!slash && seen_slash_slash)
1804 /* example: "http://foo" */
1807 else if (slash && !seen_slash_slash)
1808 /* example: "foo/bar" */
1810 start_insert = base;
1811 else if (slash && seen_slash_slash)
1812 /* example: "http://something/" */
1814 start_insert = slash;
1816 span = start_insert - base;
1817 merge = xmalloc (span + linklength + 1);
1819 memcpy (merge, base, span);
1820 memcpy (merge + span, link, linklength);
1821 merge[span + linklength] = '\0';
1825 /* LINK is a relative URL: we need to replace everything
1826 after last slash (possibly empty) with LINK.
1828 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1829 our result should be "whatever/foo/qux/xyzzy". */
1830 bool need_explicit_slash = false;
1832 const char *start_insert;
1833 const char *last_slash = find_last_char (base, end, '/');
1836 /* No slash found at all. Replace what we have with LINK. */
1837 start_insert = base;
1839 else if (last_slash && last_slash >= base + 2
1840 && last_slash[-2] == ':' && last_slash[-1] == '/')
1842 /* example: http://host" */
1844 start_insert = end + 1;
1845 need_explicit_slash = true;
1849 /* example: "whatever/foo/bar" */
1851 start_insert = last_slash + 1;
1854 span = start_insert - base;
1855 merge = xmalloc (span + linklength + 1);
1857 memcpy (merge, base, span);
1858 if (need_explicit_slash)
1859 merge[span - 1] = '/';
1860 memcpy (merge + span, link, linklength);
1861 merge[span + linklength] = '\0';
1867 #define APPEND(p, s) do { \
1868 int len = strlen (s); \
1869 memcpy (p, s, len); \
1873 /* Use this instead of password when the actual password is supposed
1874 to be hidden. We intentionally use a generic string without giving
1875 away the number of characters in the password, like previous
1877 #define HIDDEN_PASSWORD "*password*"
1879 /* Recreate the URL string from the data in URL.
1881 If HIDE is true (as it is when we're calling this on a URL we plan
1882 to print, but not when calling it to canonicalize a URL for use
1883 within the program), password will be hidden. Unsafe characters in
1884 the URL will be quoted. */
1887 url_string (const struct url *url, enum url_auth_mode auth_mode)
1891 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1893 int scheme_port = supported_schemes[url->scheme].default_port;
1894 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1895 int fplen = full_path_length (url);
1897 bool brackets_around_host;
1899 assert (scheme_str != NULL);
1901 /* Make sure the user name and password are quoted. */
1904 if (auth_mode != URL_AUTH_HIDE)
1906 quoted_user = url_escape_allow_passthrough (url->user);
1909 if (auth_mode == URL_AUTH_HIDE_PASSWD)
1910 quoted_passwd = HIDDEN_PASSWORD;
1912 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1917 /* In the unlikely event that the host name contains non-printable
1918 characters, quote it for displaying to the user. */
1919 quoted_host = url_escape_allow_passthrough (url->host);
1921 /* Undo the quoting of colons that URL escaping performs. IPv6
1922 addresses may legally contain colons, and in that case must be
1923 placed in square brackets. */
1924 if (quoted_host != url->host)
1925 unescape_single_char (quoted_host, ':');
1926 brackets_around_host = strchr (quoted_host, ':') != NULL;
1928 size = (strlen (scheme_str)
1929 + strlen (quoted_host)
1930 + (brackets_around_host ? 2 : 0)
1933 if (url->port != scheme_port)
1934 size += 1 + numdigit (url->port);
1937 size += 1 + strlen (quoted_user);
1939 size += 1 + strlen (quoted_passwd);
1942 p = result = xmalloc (size);
1944 APPEND (p, scheme_str);
1947 APPEND (p, quoted_user);
1951 APPEND (p, quoted_passwd);
1956 if (brackets_around_host)
1958 APPEND (p, quoted_host);
1959 if (brackets_around_host)
1961 if (url->port != scheme_port)
1964 p = number_to_string (p, url->port);
1967 full_path_write (url, p);
1971 assert (p - result == size);
1973 if (quoted_user && quoted_user != url->user)
1974 xfree (quoted_user);
1975 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
1976 && quoted_passwd != url->passwd)
1977 xfree (quoted_passwd);
1978 if (quoted_host != url->host)
1979 xfree (quoted_host);
1984 /* Return true if scheme a is similar to scheme b.
1986 Schemes are similar if they are equal. If SSL is supported, schemes
1987 are also similar if one is http (SCHEME_HTTP) and the other is https
1990 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1995 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1996 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
2003 getchar_from_escaped_string (const char *str, char *c)
2005 const char *p = str;
2007 assert (str && *str);
2012 if (!c_isxdigit(p[1]) || !c_isxdigit(p[2]))
2020 return 0; /* error: invalid string */
2022 *c = X2DIGITS_TO_NUM (p[1], p[2]);
2023 if (URL_RESERVED_CHAR(*c))
2041 are_urls_equal (const char *u1, const char *u2)
2052 && (pp = getchar_from_escaped_string (p, &ch1))
2053 && (qq = getchar_from_escaped_string (q, &ch2))
2054 && (c_tolower(ch1) == c_tolower(ch2)))
2060 return (*p == 0 && *q == 0 ? true : false);
2064 /* Debugging and testing support for path_simplify. */
2067 /* Debug: run path_simplify on PATH and return the result in a new
2068 string. Useful for calling from the debugger. */
2072 char *copy = xstrdup (path);
2073 path_simplify (copy);
2079 run_test (char *test, char *expected_result, enum url_scheme scheme,
2080 bool expected_change)
2082 char *test_copy = xstrdup (test);
2083 bool modified = path_simplify (scheme, test_copy);
2085 if (0 != strcmp (test_copy, expected_result))
2087 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2088 test, expected_result, test_copy);
2091 if (modified != expected_change)
2093 if (expected_change)
2094 printf ("Expected modification with path_simplify(\"%s\").\n",
2097 printf ("Expected no modification with path_simplify(\"%s\").\n",
2101 mu_assert ("", modified == expected_change);
2106 test_path_simplify (void)
2109 char *test, *result;
2110 enum url_scheme scheme;
2113 { "", "", SCHEME_HTTP, false },
2114 { ".", "", SCHEME_HTTP, true },
2115 { "./", "", SCHEME_HTTP, true },
2116 { "..", "", SCHEME_HTTP, true },
2117 { "../", "", SCHEME_HTTP, true },
2118 { "..", "..", SCHEME_FTP, false },
2119 { "../", "../", SCHEME_FTP, false },
2120 { "foo", "foo", SCHEME_HTTP, false },
2121 { "foo/bar", "foo/bar", SCHEME_HTTP, false },
2122 { "foo///bar", "foo///bar", SCHEME_HTTP, false },
2123 { "foo/.", "foo/", SCHEME_HTTP, true },
2124 { "foo/./", "foo/", SCHEME_HTTP, true },
2125 { "foo./", "foo./", SCHEME_HTTP, false },
2126 { "foo/../bar", "bar", SCHEME_HTTP, true },
2127 { "foo/../bar/", "bar/", SCHEME_HTTP, true },
2128 { "foo/bar/..", "foo/", SCHEME_HTTP, true },
2129 { "foo/bar/../x", "foo/x", SCHEME_HTTP, true },
2130 { "foo/bar/../x/", "foo/x/", SCHEME_HTTP, true },
2131 { "foo/..", "", SCHEME_HTTP, true },
2132 { "foo/../..", "", SCHEME_HTTP, true },
2133 { "foo/../../..", "", SCHEME_HTTP, true },
2134 { "foo/../../bar/../../baz", "baz", SCHEME_HTTP, true },
2135 { "foo/../..", "..", SCHEME_FTP, true },
2136 { "foo/../../..", "../..", SCHEME_FTP, true },
2137 { "foo/../../bar/../../baz", "../../baz", SCHEME_FTP, true },
2138 { "a/b/../../c", "c", SCHEME_HTTP, true },
2139 { "./a/../b", "b", SCHEME_HTTP, true }
2143 for (i = 0; i < countof (tests); i++)
2145 const char *message;
2146 char *test = tests[i].test;
2147 char *expected_result = tests[i].result;
2148 enum url_scheme scheme = tests[i].scheme;
2149 bool expected_change = tests[i].should_modify;
2150 message = run_test (test, expected_result, scheme, expected_change);
2151 if (message) return message;
2157 test_append_uri_pathel()
2164 char *expected_result;
2166 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2169 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2171 struct growable dest;
2172 const char *p = test_array[i].input;
2174 memset (&dest, 0, sizeof (dest));
2176 append_string (test_array[i].original_url, &dest);
2177 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2178 append_char ('\0', &dest);
2180 mu_assert ("test_append_uri_pathel: wrong result",
2181 strcmp (dest.base, test_array[i].expected_result) == 0);
2188 test_are_urls_equal()
2194 bool expected_result;
2196 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2197 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2198 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2199 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2200 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2201 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2204 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2206 mu_assert ("test_are_urls_equal: wrong result",
2207 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2213 #endif /* TESTING */