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, bool percent_encode)
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 iri->utf8_encode = remote_to_utf8 (iri, iri->orig_url ? iri->orig_url : url, (const char **) &new_url);
685 if (!iri->utf8_encode)
688 iri->orig_url = xstrdup (url);
691 /* XXX XXX Could that change introduce (security) bugs ??? XXX XXX*/
693 url_encoded = reencode_escapes (new_url ? new_url : url);
695 url_encoded = new_url ? new_url : url;
699 if (new_url && url_encoded != new_url)
702 p += strlen (supported_schemes[scheme].leading_string);
704 p = url_skip_credentials (p);
707 /* scheme://user:pass@host[:port]... */
710 /* We attempt to break down the URL into the components path,
711 params, query, and fragment. They are ordered like this:
713 scheme://host[:port][/path][;params][?query][#fragment] */
715 path_b = path_e = NULL;
716 params_b = params_e = NULL;
717 query_b = query_e = NULL;
718 fragment_b = fragment_e = NULL;
720 /* Initialize separators for optional parts of URL, depending on the
721 scheme. For example, FTP has params, and HTTP and HTTPS have
722 query string and fragment. */
723 seps = init_seps (scheme);
729 /* Handle IPv6 address inside square brackets. Ideally we'd
730 just look for the terminating ']', but rfc2732 mandates
731 rejecting invalid IPv6 addresses. */
733 /* The address begins after '['. */
735 host_e = strchr (host_b, ']');
739 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
744 /* Check if the IPv6 address is valid. */
745 if (!is_valid_ipv6_address(host_b, host_e))
747 error_code = PE_INVALID_IPV6_ADDRESS;
751 /* Continue parsing after the closing ']'. */
754 error_code = PE_IPV6_NOT_SUPPORTED;
758 /* The closing bracket must be followed by a separator or by the
760 /* http://[::1]... */
762 if (!strchr (seps, *p))
764 /* Trailing garbage after []-delimited IPv6 address. */
765 error_code = PE_INVALID_HOST_NAME;
771 p = strpbrk_or_eos (p, seps);
774 ++seps; /* advance to '/' */
776 if (host_b == host_e)
778 error_code = PE_INVALID_HOST_NAME;
782 port = scheme_default_port (scheme);
785 const char *port_b, *port_e, *pp;
787 /* scheme://host:port/tralala */
791 p = strpbrk_or_eos (p, seps);
794 /* Allow empty port, as per rfc2396. */
795 if (port_b != port_e)
796 for (port = 0, pp = port_b; pp < port_e; pp++)
798 if (!c_isdigit (*pp))
800 /* http://host:12randomgarbage/blah */
802 error_code = PE_BAD_PORT_NUMBER;
805 port = 10 * port + (*pp - '0');
806 /* Check for too large port numbers here, before we have
807 a chance to overflow on bogus port values. */
810 error_code = PE_BAD_PORT_NUMBER;
815 /* Advance to the first separator *after* '/' (either ';' or '?',
816 depending on the scheme). */
819 /* Get the optional parts of URL, each part being delimited by
820 current location and the position of the next separator. */
821 #define GET_URL_PART(sepchar, var) do { \
823 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
827 GET_URL_PART ('/', path);
828 if (supported_schemes[scheme].flags & scm_has_params)
829 GET_URL_PART (';', params);
830 if (supported_schemes[scheme].flags & scm_has_query)
831 GET_URL_PART ('?', query);
832 if (supported_schemes[scheme].flags & scm_has_fragment)
833 GET_URL_PART ('#', fragment);
838 if (uname_b != uname_e)
840 /* http://user:pass@host */
842 /* uname_b uname_e */
843 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
845 error_code = PE_INVALID_USER_NAME;
850 u = xnew0 (struct url);
852 u->host = strdupdelim (host_b, host_e);
857 u->path = strdupdelim (path_b, path_e);
858 path_modified = path_simplify (scheme, u->path);
859 split_path (u->path, &u->dir, &u->file);
861 host_modified = lowercase_str (u->host);
863 /* Decode %HH sequences in host name. This is important not so much
864 to support %HH sequences in host names (which other browser
865 don't), but to support binary characters (which will have been
866 converted to %HH by reencode_escapes). */
867 if (strchr (u->host, '%'))
869 url_unescape (u->host);
870 host_modified = true;
872 /* Apply IDNA regardless of iri->utf8_encode status */
873 if (opt.enable_iri && iri)
875 char *new = idn_encode (iri, u->host);
880 host_modified = true;
886 u->params = strdupdelim (params_b, params_e);
888 u->query = strdupdelim (query_b, query_e);
890 u->fragment = strdupdelim (fragment_b, fragment_e);
892 if (opt.enable_iri || path_modified || u->fragment || host_modified || path_b == path_e)
894 /* If we suspect that a transformation has rendered what
895 url_string might return different from URL_ENCODED, rebuild
896 u->url using url_string. */
897 u->url = url_string (u, URL_AUTH_SHOW);
899 if (url_encoded != url)
900 xfree ((char *) url_encoded);
904 if (url_encoded == url)
905 u->url = xstrdup (url);
907 u->url = url_encoded;
913 /* Cleanup in case of error: */
914 if (url_encoded && url_encoded != url)
917 /* Transmit the error code to the caller, if the caller wants to
924 /* Return the error message string from ERROR_CODE, which should have
925 been retrieved from url_parse. The error message is translated. */
928 url_error (const char *url, int error_code)
930 assert (error_code >= 0 && ((size_t) error_code) < countof (parse_errors));
932 if (error_code == PE_UNSUPPORTED_SCHEME)
935 char *scheme = xstrdup (url);
936 assert (url_has_scheme (url));
938 if ((p = strchr (scheme, ':')))
940 if (!strcasecmp (scheme, "https"))
941 error = aprintf (_("HTTPS support not compiled in"));
943 error = aprintf (_(parse_errors[error_code]), quote (scheme));
949 return xstrdup (_(parse_errors[error_code]));
952 /* Split PATH into DIR and FILE. PATH comes from the URL and is
953 expected to be URL-escaped.
955 The path is split into directory (the part up to the last slash)
956 and file (the part after the last slash), which are subsequently
960 "foo/bar/baz" "foo/bar" "baz"
961 "foo/bar/" "foo/bar" ""
963 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
965 DIR and FILE are freshly allocated. */
968 split_path (const char *path, char **dir, char **file)
970 char *last_slash = strrchr (path, '/');
974 *file = xstrdup (path);
978 *dir = strdupdelim (path, last_slash);
979 *file = xstrdup (last_slash + 1);
982 url_unescape (*file);
985 /* Note: URL's "full path" is the path with the query string and
986 params appended. The "fragment" (#foo) is intentionally ignored,
987 but that might be changed. For example, if the original URL was
988 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
989 the full path will be "/foo/bar/baz;bullshit?querystring". */
991 /* Return the length of the full path, without the terminating
995 full_path_length (const struct url *url)
999 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
1010 /* Write out the full path. */
1013 full_path_write (const struct url *url, char *where)
1015 #define FROB(el, chr) do { \
1016 char *f_el = url->el; \
1018 int l = strlen (f_el); \
1020 memcpy (where, f_el, l); \
1032 /* Public function for getting the "full path". E.g. if u->path is
1033 "foo/bar" and u->query is "param=value", full_path will be
1034 "/foo/bar?param=value". */
1037 url_full_path (const struct url *url)
1039 int length = full_path_length (url);
1040 char *full_path = xmalloc (length + 1);
1042 full_path_write (url, full_path);
1043 full_path[length] = '\0';
1048 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1049 escaping of certain characters, such as "/" and ":". Returns a
1050 count of unescaped chars. */
1053 unescape_single_char (char *str, char chr)
1055 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1056 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1057 char *h = str; /* hare */
1058 char *t = str; /* tortoise */
1059 for (; *h; h++, t++)
1061 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1072 /* Escape unsafe and reserved characters, except for the slash
1076 url_escape_dir (const char *dir)
1078 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1082 unescape_single_char (newdir, '/');
1086 /* Sync u->path and u->url with u->dir and u->file. Called after
1087 u->file or u->dir have been changed, typically by the FTP code. */
1090 sync_path (struct url *u)
1092 char *newpath, *efile, *edir;
1096 /* u->dir and u->file are not escaped. URL-escape them before
1097 reassembling them into u->path. That way, if they contain
1098 separators like '?' or even if u->file contains slashes, the
1099 path will be correctly assembled. (u->file can contain slashes
1100 if the URL specifies it with %2f, or if an FTP server returns
1102 edir = url_escape_dir (u->dir);
1103 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1106 newpath = xstrdup (efile);
1109 int dirlen = strlen (edir);
1110 int filelen = strlen (efile);
1112 /* Copy "DIR/FILE" to newpath. */
1113 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1114 memcpy (p, edir, dirlen);
1117 memcpy (p, efile, filelen);
1126 if (efile != u->file)
1129 /* Regenerate u->url as well. */
1131 u->url = url_string (u, URL_AUTH_SHOW);
1134 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1135 This way we can sync u->path and u->url when they get changed. */
1138 url_set_dir (struct url *url, const char *newdir)
1141 url->dir = xstrdup (newdir);
1146 url_set_file (struct url *url, const char *newfile)
1149 url->file = xstrdup (newfile);
1154 url_free (struct url *url)
1160 xfree_null (url->params);
1161 xfree_null (url->query);
1162 xfree_null (url->fragment);
1163 xfree_null (url->user);
1164 xfree_null (url->passwd);
1172 /* Create all the necessary directories for PATH (a file). Calls
1173 make_directory internally. */
1175 mkalldirs (const char *path)
1182 p = path + strlen (path);
1183 for (; *p != '/' && p != path; p--)
1186 /* Don't create if it's just a file. */
1187 if ((p == path) && (*p != '/'))
1189 t = strdupdelim (path, p);
1191 /* Check whether the directory exists. */
1192 if ((stat (t, &st) == 0))
1194 if (S_ISDIR (st.st_mode))
1201 /* If the dir exists as a file name, remove it first. This
1202 is *only* for Wget to work with buggy old CERN http
1203 servers. Here is the scenario: When Wget tries to
1204 retrieve a directory without a slash, e.g.
1205 http://foo/bar (bar being a directory), CERN server will
1206 not redirect it too http://foo/bar/ -- it will generate a
1207 directory listing containing links to bar/file1,
1208 bar/file2, etc. Wget will lose because it saves this
1209 HTML listing to a file `bar', so it cannot create the
1210 directory. To work around this, if the file of the same
1211 name exists, we just remove it and create the directory
1213 DEBUGP (("Removing %s because of directory danger!\n", t));
1217 res = make_directory (t);
1219 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1224 /* Functions for constructing the file name out of URL components. */
1226 /* A growable string structure, used by url_file_name and friends.
1227 This should perhaps be moved to utils.c.
1229 The idea is to have a convenient and efficient way to construct a
1230 string by having various functions append data to it. Instead of
1231 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1232 functions in questions, we pass the pointer to this struct. */
1240 /* Ensure that the string can accept APPEND_COUNT more characters past
1241 the current TAIL position. If necessary, this will grow the string
1242 and update its allocated size. If the string is already large
1243 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1244 #define GROW(g, append_size) do { \
1245 struct growable *G_ = g; \
1246 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1249 /* Return the tail position of the string. */
1250 #define TAIL(r) ((r)->base + (r)->tail)
1252 /* Move the tail position by APPEND_COUNT characters. */
1253 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1255 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1259 append_string (const char *str, struct growable *dest)
1261 int l = strlen (str);
1263 memcpy (TAIL (dest), str, l);
1264 TAIL_INCR (dest, l);
1267 /* Append CH to DEST. For example, append_char (0, DEST)
1268 zero-terminates DEST. */
1271 append_char (char ch, struct growable *dest)
1275 TAIL_INCR (dest, 1);
1279 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1280 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1281 filechr_control = 4 /* a control character, e.g. 0-31 */
1284 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1286 /* Shorthands for the table: */
1287 #define U filechr_not_unix
1288 #define W filechr_not_windows
1289 #define C filechr_control
1294 /* Table of characters unsafe under various conditions (see above).
1296 Arguably we could also claim `%' to be unsafe, since we use it as
1297 the escape character. If we ever want to be able to reliably
1298 translate file name back to URL, this would become important
1299 crucial. Right now, it's better to be minimal in escaping. */
1301 static const unsigned char filechr_table[256] =
1303 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1304 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1305 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1306 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1307 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1308 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1309 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1310 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1311 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1312 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1313 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1314 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1315 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1316 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1317 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1318 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1320 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1321 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1322 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1323 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1325 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1326 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1327 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1328 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1336 /* FN_PORT_SEP is the separator between host and port in file names
1337 for non-standard port numbers. On Unix this is normally ':', as in
1338 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1339 because Windows can't handle ':' in file names. */
1340 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1342 /* FN_QUERY_SEP is the separator between the file name and the URL
1343 query, normally '?'. Since Windows cannot handle '?' as part of
1344 file name, we use '@' instead there. */
1345 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1347 /* Quote path element, characters in [b, e), as file name, and append
1348 the quoted string to DEST. Each character is quoted as per
1349 file_unsafe_char and the corresponding table.
1351 If ESCAPED is true, the path element is considered to be
1352 URL-escaped and will be unescaped prior to inspection. */
1355 append_uri_pathel (const char *b, const char *e, bool escaped,
1356 struct growable *dest)
1362 if (opt.restrict_files_os == restrict_unix)
1363 mask = filechr_not_unix;
1365 mask = filechr_not_windows;
1366 if (opt.restrict_files_ctrl)
1367 mask |= filechr_control;
1369 /* Copy [b, e) to PATHEL and URL-unescape it. */
1373 BOUNDED_TO_ALLOCA (b, e, unescaped);
1374 url_unescape (unescaped);
1376 e = unescaped + strlen (unescaped);
1379 /* Defang ".." when found as component of path. Remember that path
1380 comes from the URL and might contain malicious input. */
1381 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1387 /* Walk the PATHEL string and check how many characters we'll need
1390 for (p = b; p < e; p++)
1391 if (FILE_CHAR_TEST (*p, mask))
1394 /* Calculate the length of the output string. e-b is the input
1395 string length. Each quoted char introduces two additional
1396 characters in the string, hence 2*quoted. */
1397 outlen = (e - b) + (2 * quoted);
1398 GROW (dest, outlen);
1402 /* If there's nothing to quote, we can simply append the string
1403 without processing it again. */
1404 memcpy (TAIL (dest), b, outlen);
1408 char *q = TAIL (dest);
1409 for (p = b; p < e; p++)
1411 if (!FILE_CHAR_TEST (*p, mask))
1415 unsigned char ch = *p;
1417 *q++ = XNUM_TO_DIGIT (ch >> 4);
1418 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1421 assert (q - TAIL (dest) == outlen);
1424 /* Perform inline case transformation if required. */
1425 if (opt.restrict_files_case == restrict_lowercase
1426 || opt.restrict_files_case == restrict_uppercase)
1429 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1431 if (opt.restrict_files_case == restrict_lowercase)
1432 *q = c_tolower (*q);
1434 *q = c_toupper (*q);
1438 TAIL_INCR (dest, outlen);
1441 /* Append to DEST the directory structure that corresponds the
1442 directory part of URL's path. For example, if the URL is
1443 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1445 Each path element ("dir1" and "dir2" in the above example) is
1446 examined, url-unescaped, and re-escaped as file name element.
1448 Additionally, it cuts as many directories from the path as
1449 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1450 will produce "bar" for the above example. For 2 or more, it will
1453 Each component of the path is quoted for use as file name. */
1456 append_dir_structure (const struct url *u, struct growable *dest)
1458 char *pathel, *next;
1459 int cut = opt.cut_dirs;
1461 /* Go through the path components, de-URL-quote them, and quote them
1462 (if necessary) as file names. */
1465 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1470 /* Ignore empty pathels. */
1474 append_char ('/', dest);
1475 append_uri_pathel (pathel, next, true, dest);
1479 /* Return a unique file name that matches the given URL as good as
1480 possible. Does not create directories on the file system. */
1483 url_file_name (const struct url *u)
1485 struct growable fnres; /* stands for "file name result" */
1487 const char *u_file, *u_query;
1488 char *fname, *unique;
1489 char *index_filename = "index.html"; /* The default index file is index.html */
1495 /* If an alternative index file was defined, change index_filename */
1496 if (opt.default_page)
1497 index_filename = opt.default_page;
1500 /* Start with the directory prefix, if specified. */
1502 append_string (opt.dir_prefix, &fnres);
1504 /* If "dirstruct" is turned on (typically the case with -r), add
1505 the host and port (unless those have been turned off) and
1506 directory structure. */
1509 if (opt.protocol_directories)
1512 append_char ('/', &fnres);
1513 append_string (supported_schemes[u->scheme].name, &fnres);
1515 if (opt.add_hostdir)
1518 append_char ('/', &fnres);
1519 if (0 != strcmp (u->host, ".."))
1520 append_string (u->host, &fnres);
1522 /* Host name can come from the network; malicious DNS may
1523 allow ".." to be resolved, causing us to write to
1524 "../<file>". Defang such host names. */
1525 append_string ("%2E%2E", &fnres);
1526 if (u->port != scheme_default_port (u->scheme))
1529 number_to_string (portstr, u->port);
1530 append_char (FN_PORT_SEP, &fnres);
1531 append_string (portstr, &fnres);
1535 append_dir_structure (u, &fnres);
1538 /* Add the file name. */
1540 append_char ('/', &fnres);
1541 u_file = *u->file ? u->file : index_filename;
1542 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1544 /* Append "?query" to the file name. */
1545 u_query = u->query && *u->query ? u->query : NULL;
1548 append_char (FN_QUERY_SEP, &fnres);
1549 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1552 /* Zero-terminate the file name. */
1553 append_char ('\0', &fnres);
1557 /* Check the cases in which the unique extensions are not used:
1558 1) Clobbering is turned off (-nc).
1559 2) Retrieval with regetting.
1560 3) Timestamping is used.
1561 4) Hierarchy is built.
1563 The exception is the case when file does exist and is a
1564 directory (see `mkalldirs' for explanation). */
1566 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1567 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1570 unique = unique_name (fname, true);
1571 if (unique != fname)
1576 /* Resolve "." and ".." elements of PATH by destructively modifying
1577 PATH and return true if PATH has been modified, false otherwise.
1579 The algorithm is in spirit similar to the one described in rfc1808,
1580 although implemented differently, in one pass. To recap, path
1581 elements containing only "." are removed, and ".." is taken to mean
1582 "back up one element". Single leading and trailing slashes are
1585 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1586 test examples are provided below. If you change anything in this
1587 function, run test_path_simplify to make sure you haven't broken a
1591 path_simplify (enum url_scheme scheme, char *path)
1593 char *h = path; /* hare */
1594 char *t = path; /* tortoise */
1596 char *end = strchr (path, '\0');
1600 /* Hare should be at the beginning of a path element. */
1602 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1607 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1609 /* Handle "../" by retreating the tortoise by one path
1610 element -- but not past beggining. */
1613 /* Move backwards until T hits the beginning of the
1614 previous path element or the beginning of path. */
1615 for (--t; t > beg && t[-1] != '/'; t--)
1618 else if (scheme == SCHEME_FTP)
1620 /* If we're at the beginning, copy the "../" literally
1621 and move the beginning so a later ".." doesn't remove
1622 it. This violates RFC 3986; but we do it for FTP
1623 anyway because there is otherwise no way to get at a
1624 parent directory, when the FTP server drops us in a
1625 non-root directory (which is not uncommon). */
1634 /* A regular path element. If H hasn't advanced past T,
1635 simply skip to the next path element. Otherwise, copy
1636 the path element until the next slash. */
1639 /* Skip the path element, including the slash. */
1640 while (h < end && *h != '/')
1647 /* Copy the path element, including the final slash. */
1648 while (h < end && *h != '/')
1662 /* Return the length of URL's path. Path is considered to be
1663 terminated by one or more of the ?query or ;params or #fragment,
1664 depending on the scheme. */
1667 path_end (const char *url)
1669 enum url_scheme scheme = url_scheme (url);
1671 if (scheme == SCHEME_INVALID)
1672 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1673 /* +2 to ignore the first two separators ':' and '/' */
1674 seps = init_seps (scheme) + 2;
1675 return strpbrk_or_eos (url, seps);
1678 /* Find the last occurrence of character C in the range [b, e), or
1679 NULL, if none are present. */
1680 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1682 /* Merge BASE with LINK and return the resulting URI.
1684 Either of the URIs may be absolute or relative, complete with the
1685 host name, or path only. This tries to reasonably handle all
1686 foreseeable cases. It only employs minimal URL parsing, without
1687 knowledge of the specifics of schemes.
1689 I briefly considered making this function call path_simplify after
1690 the merging process, as rfc1738 seems to suggest. This is a bad
1691 idea for several reasons: 1) it complexifies the code, and 2)
1692 url_parse has to simplify path anyway, so it's wasteful to boot. */
1695 uri_merge (const char *base, const char *link)
1701 if (url_has_scheme (link))
1702 return xstrdup (link);
1704 /* We may not examine BASE past END. */
1705 end = path_end (base);
1706 linklength = strlen (link);
1710 /* Empty LINK points back to BASE, query string and all. */
1711 return xstrdup (base);
1713 else if (*link == '?')
1715 /* LINK points to the same location, but changes the query
1716 string. Examples: */
1717 /* uri_merge("path", "?new") -> "path?new" */
1718 /* uri_merge("path?foo", "?new") -> "path?new" */
1719 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1720 /* uri_merge("path#foo", "?new") -> "path?new" */
1721 int baselength = end - base;
1722 merge = xmalloc (baselength + linklength + 1);
1723 memcpy (merge, base, baselength);
1724 memcpy (merge + baselength, link, linklength);
1725 merge[baselength + linklength] = '\0';
1727 else if (*link == '#')
1729 /* uri_merge("path", "#new") -> "path#new" */
1730 /* uri_merge("path#foo", "#new") -> "path#new" */
1731 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1732 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1734 const char *end1 = strchr (base, '#');
1736 end1 = base + strlen (base);
1737 baselength = end1 - base;
1738 merge = xmalloc (baselength + linklength + 1);
1739 memcpy (merge, base, baselength);
1740 memcpy (merge + baselength, link, linklength);
1741 merge[baselength + linklength] = '\0';
1743 else if (*link == '/' && *(link + 1) == '/')
1745 /* LINK begins with "//" and so is a net path: we need to
1746 replace everything after (and including) the double slash
1749 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1750 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1751 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1755 const char *start_insert;
1757 /* Look for first slash. */
1758 slash = memchr (base, '/', end - base);
1759 /* If found slash and it is a double slash, then replace
1760 from this point, else default to replacing from the
1762 if (slash && *(slash + 1) == '/')
1763 start_insert = slash;
1765 start_insert = base;
1767 span = start_insert - base;
1768 merge = xmalloc (span + linklength + 1);
1770 memcpy (merge, base, span);
1771 memcpy (merge + span, link, linklength);
1772 merge[span + linklength] = '\0';
1774 else if (*link == '/')
1776 /* LINK is an absolute path: we need to replace everything
1777 after (and including) the FIRST slash with LINK.
1779 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1780 "/qux/xyzzy", our result should be
1781 "http://host/qux/xyzzy". */
1784 const char *start_insert = NULL; /* for gcc to shut up. */
1785 const char *pos = base;
1786 bool seen_slash_slash = false;
1787 /* We're looking for the first slash, but want to ignore
1790 slash = memchr (pos, '/', end - pos);
1791 if (slash && !seen_slash_slash)
1792 if (*(slash + 1) == '/')
1795 seen_slash_slash = true;
1799 /* At this point, SLASH is the location of the first / after
1800 "//", or the first slash altogether. START_INSERT is the
1801 pointer to the location where LINK will be inserted. When
1802 examining the last two examples, keep in mind that LINK
1805 if (!slash && !seen_slash_slash)
1806 /* example: "foo" */
1808 start_insert = base;
1809 else if (!slash && seen_slash_slash)
1810 /* example: "http://foo" */
1813 else if (slash && !seen_slash_slash)
1814 /* example: "foo/bar" */
1816 start_insert = base;
1817 else if (slash && seen_slash_slash)
1818 /* example: "http://something/" */
1820 start_insert = slash;
1822 span = start_insert - base;
1823 merge = xmalloc (span + linklength + 1);
1825 memcpy (merge, base, span);
1826 memcpy (merge + span, link, linklength);
1827 merge[span + linklength] = '\0';
1831 /* LINK is a relative URL: we need to replace everything
1832 after last slash (possibly empty) with LINK.
1834 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1835 our result should be "whatever/foo/qux/xyzzy". */
1836 bool need_explicit_slash = false;
1838 const char *start_insert;
1839 const char *last_slash = find_last_char (base, end, '/');
1842 /* No slash found at all. Replace what we have with LINK. */
1843 start_insert = base;
1845 else if (last_slash && last_slash >= base + 2
1846 && last_slash[-2] == ':' && last_slash[-1] == '/')
1848 /* example: http://host" */
1850 start_insert = end + 1;
1851 need_explicit_slash = true;
1855 /* example: "whatever/foo/bar" */
1857 start_insert = last_slash + 1;
1860 span = start_insert - base;
1861 merge = xmalloc (span + linklength + 1);
1863 memcpy (merge, base, span);
1864 if (need_explicit_slash)
1865 merge[span - 1] = '/';
1866 memcpy (merge + span, link, linklength);
1867 merge[span + linklength] = '\0';
1873 #define APPEND(p, s) do { \
1874 int len = strlen (s); \
1875 memcpy (p, s, len); \
1879 /* Use this instead of password when the actual password is supposed
1880 to be hidden. We intentionally use a generic string without giving
1881 away the number of characters in the password, like previous
1883 #define HIDDEN_PASSWORD "*password*"
1885 /* Recreate the URL string from the data in URL.
1887 If HIDE is true (as it is when we're calling this on a URL we plan
1888 to print, but not when calling it to canonicalize a URL for use
1889 within the program), password will be hidden. Unsafe characters in
1890 the URL will be quoted. */
1893 url_string (const struct url *url, enum url_auth_mode auth_mode)
1897 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1899 int scheme_port = supported_schemes[url->scheme].default_port;
1900 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1901 int fplen = full_path_length (url);
1903 bool brackets_around_host;
1905 assert (scheme_str != NULL);
1907 /* Make sure the user name and password are quoted. */
1910 if (auth_mode != URL_AUTH_HIDE)
1912 quoted_user = url_escape_allow_passthrough (url->user);
1915 if (auth_mode == URL_AUTH_HIDE_PASSWD)
1916 quoted_passwd = HIDDEN_PASSWORD;
1918 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1923 /* In the unlikely event that the host name contains non-printable
1924 characters, quote it for displaying to the user. */
1925 quoted_host = url_escape_allow_passthrough (url->host);
1927 /* Undo the quoting of colons that URL escaping performs. IPv6
1928 addresses may legally contain colons, and in that case must be
1929 placed in square brackets. */
1930 if (quoted_host != url->host)
1931 unescape_single_char (quoted_host, ':');
1932 brackets_around_host = strchr (quoted_host, ':') != NULL;
1934 size = (strlen (scheme_str)
1935 + strlen (quoted_host)
1936 + (brackets_around_host ? 2 : 0)
1939 if (url->port != scheme_port)
1940 size += 1 + numdigit (url->port);
1943 size += 1 + strlen (quoted_user);
1945 size += 1 + strlen (quoted_passwd);
1948 p = result = xmalloc (size);
1950 APPEND (p, scheme_str);
1953 APPEND (p, quoted_user);
1957 APPEND (p, quoted_passwd);
1962 if (brackets_around_host)
1964 APPEND (p, quoted_host);
1965 if (brackets_around_host)
1967 if (url->port != scheme_port)
1970 p = number_to_string (p, url->port);
1973 full_path_write (url, p);
1977 assert (p - result == size);
1979 if (quoted_user && quoted_user != url->user)
1980 xfree (quoted_user);
1981 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
1982 && quoted_passwd != url->passwd)
1983 xfree (quoted_passwd);
1984 if (quoted_host != url->host)
1985 xfree (quoted_host);
1990 /* Return true if scheme a is similar to scheme b.
1992 Schemes are similar if they are equal. If SSL is supported, schemes
1993 are also similar if one is http (SCHEME_HTTP) and the other is https
1996 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
2001 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
2002 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
2009 getchar_from_escaped_string (const char *str, char *c)
2011 const char *p = str;
2013 assert (str && *str);
2018 if (!c_isxdigit(p[1]) || !c_isxdigit(p[2]))
2026 return 0; /* error: invalid string */
2028 *c = X2DIGITS_TO_NUM (p[1], p[2]);
2029 if (URL_RESERVED_CHAR(*c))
2047 are_urls_equal (const char *u1, const char *u2)
2058 && (pp = getchar_from_escaped_string (p, &ch1))
2059 && (qq = getchar_from_escaped_string (q, &ch2))
2060 && (c_tolower(ch1) == c_tolower(ch2)))
2066 return (*p == 0 && *q == 0 ? true : false);
2070 /* Debugging and testing support for path_simplify. */
2073 /* Debug: run path_simplify on PATH and return the result in a new
2074 string. Useful for calling from the debugger. */
2078 char *copy = xstrdup (path);
2079 path_simplify (copy);
2085 run_test (char *test, char *expected_result, enum url_scheme scheme,
2086 bool expected_change)
2088 char *test_copy = xstrdup (test);
2089 bool modified = path_simplify (scheme, test_copy);
2091 if (0 != strcmp (test_copy, expected_result))
2093 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2094 test, expected_result, test_copy);
2097 if (modified != expected_change)
2099 if (expected_change)
2100 printf ("Expected modification with path_simplify(\"%s\").\n",
2103 printf ("Expected no modification with path_simplify(\"%s\").\n",
2107 mu_assert ("", modified == expected_change);
2112 test_path_simplify (void)
2115 char *test, *result;
2116 enum url_scheme scheme;
2119 { "", "", SCHEME_HTTP, false },
2120 { ".", "", SCHEME_HTTP, true },
2121 { "./", "", SCHEME_HTTP, true },
2122 { "..", "", SCHEME_HTTP, true },
2123 { "../", "", SCHEME_HTTP, true },
2124 { "..", "..", SCHEME_FTP, false },
2125 { "../", "../", SCHEME_FTP, false },
2126 { "foo", "foo", SCHEME_HTTP, false },
2127 { "foo/bar", "foo/bar", SCHEME_HTTP, false },
2128 { "foo///bar", "foo///bar", SCHEME_HTTP, false },
2129 { "foo/.", "foo/", SCHEME_HTTP, true },
2130 { "foo/./", "foo/", SCHEME_HTTP, true },
2131 { "foo./", "foo./", SCHEME_HTTP, false },
2132 { "foo/../bar", "bar", SCHEME_HTTP, true },
2133 { "foo/../bar/", "bar/", SCHEME_HTTP, true },
2134 { "foo/bar/..", "foo/", SCHEME_HTTP, true },
2135 { "foo/bar/../x", "foo/x", SCHEME_HTTP, true },
2136 { "foo/bar/../x/", "foo/x/", SCHEME_HTTP, true },
2137 { "foo/..", "", SCHEME_HTTP, true },
2138 { "foo/../..", "", SCHEME_HTTP, true },
2139 { "foo/../../..", "", SCHEME_HTTP, true },
2140 { "foo/../../bar/../../baz", "baz", SCHEME_HTTP, true },
2141 { "foo/../..", "..", SCHEME_FTP, true },
2142 { "foo/../../..", "../..", SCHEME_FTP, true },
2143 { "foo/../../bar/../../baz", "../../baz", SCHEME_FTP, true },
2144 { "a/b/../../c", "c", SCHEME_HTTP, true },
2145 { "./a/../b", "b", SCHEME_HTTP, true }
2149 for (i = 0; i < countof (tests); i++)
2151 const char *message;
2152 char *test = tests[i].test;
2153 char *expected_result = tests[i].result;
2154 enum url_scheme scheme = tests[i].scheme;
2155 bool expected_change = tests[i].should_modify;
2156 message = run_test (test, expected_result, scheme, expected_change);
2157 if (message) return message;
2163 test_append_uri_pathel()
2170 char *expected_result;
2172 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2175 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2177 struct growable dest;
2178 const char *p = test_array[i].input;
2180 memset (&dest, 0, sizeof (dest));
2182 append_string (test_array[i].original_url, &dest);
2183 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2184 append_char ('\0', &dest);
2186 mu_assert ("test_append_uri_pathel: wrong result",
2187 strcmp (dest.base, test_array[i].expected_result) == 0);
2194 test_are_urls_equal()
2200 bool expected_result;
2202 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2203 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2204 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2205 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2206 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2207 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2210 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2212 mu_assert ("test_are_urls_equal: wrong result",
2213 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2219 #endif /* TESTING */