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"), /* support for format token only here */
632 #define PE_MISSING_SCHEME 2
633 N_("Scheme missing"),
634 #define PE_INVALID_HOST_NAME 3
635 N_("Invalid host name"),
636 #define PE_BAD_PORT_NUMBER 4
637 N_("Bad port number"),
638 #define PE_INVALID_USER_NAME 5
639 N_("Invalid user name"),
640 #define PE_UNTERMINATED_IPV6_ADDRESS 6
641 N_("Unterminated IPv6 numeric address"),
642 #define PE_IPV6_NOT_SUPPORTED 7
643 N_("IPv6 addresses not supported"),
644 #define PE_INVALID_IPV6_ADDRESS 8
645 N_("Invalid IPv6 numeric address")
650 Return a new struct url if successful, NULL on error. In case of
651 error, and if ERROR is not NULL, also set *ERROR to the appropriate
654 url_parse (const char *url, int *error, struct iri *iri, bool percent_encode)
658 bool path_modified, host_modified;
660 enum url_scheme scheme;
663 const char *uname_b, *uname_e;
664 const char *host_b, *host_e;
665 const char *path_b, *path_e;
666 const char *params_b, *params_e;
667 const char *query_b, *query_e;
668 const char *fragment_b, *fragment_e;
671 char *user = NULL, *passwd = NULL;
673 const char *url_encoded = NULL;
674 char *new_url = NULL;
678 scheme = url_scheme (url);
679 if (scheme == SCHEME_INVALID)
681 if (url_has_scheme (url))
682 error_code = PE_UNSUPPORTED_SCHEME;
684 error_code = PE_MISSING_SCHEME;
688 if (iri && iri->utf8_encode)
690 iri->utf8_encode = remote_to_utf8 (iri, iri->orig_url ? iri->orig_url : url, (const char **) &new_url);
691 if (!iri->utf8_encode)
694 iri->orig_url = xstrdup (url);
697 /* XXX XXX Could that change introduce (security) bugs ??? XXX XXX*/
699 url_encoded = reencode_escapes (new_url ? new_url : url);
701 url_encoded = new_url ? new_url : url;
705 if (new_url && url_encoded != new_url)
708 p += strlen (supported_schemes[scheme].leading_string);
710 p = url_skip_credentials (p);
713 /* scheme://user:pass@host[:port]... */
716 /* We attempt to break down the URL into the components path,
717 params, query, and fragment. They are ordered like this:
719 scheme://host[:port][/path][;params][?query][#fragment] */
721 path_b = path_e = NULL;
722 params_b = params_e = NULL;
723 query_b = query_e = NULL;
724 fragment_b = fragment_e = NULL;
726 /* Initialize separators for optional parts of URL, depending on the
727 scheme. For example, FTP has params, and HTTP and HTTPS have
728 query string and fragment. */
729 seps = init_seps (scheme);
735 /* Handle IPv6 address inside square brackets. Ideally we'd
736 just look for the terminating ']', but rfc2732 mandates
737 rejecting invalid IPv6 addresses. */
739 /* The address begins after '['. */
741 host_e = strchr (host_b, ']');
745 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
750 /* Check if the IPv6 address is valid. */
751 if (!is_valid_ipv6_address(host_b, host_e))
753 error_code = PE_INVALID_IPV6_ADDRESS;
757 /* Continue parsing after the closing ']'. */
760 error_code = PE_IPV6_NOT_SUPPORTED;
764 /* The closing bracket must be followed by a separator or by the
766 /* http://[::1]... */
768 if (!strchr (seps, *p))
770 /* Trailing garbage after []-delimited IPv6 address. */
771 error_code = PE_INVALID_HOST_NAME;
777 p = strpbrk_or_eos (p, seps);
780 ++seps; /* advance to '/' */
782 if (host_b == host_e)
784 error_code = PE_INVALID_HOST_NAME;
788 port = scheme_default_port (scheme);
791 const char *port_b, *port_e, *pp;
793 /* scheme://host:port/tralala */
797 p = strpbrk_or_eos (p, seps);
800 /* Allow empty port, as per rfc2396. */
801 if (port_b != port_e)
802 for (port = 0, pp = port_b; pp < port_e; pp++)
804 if (!c_isdigit (*pp))
806 /* http://host:12randomgarbage/blah */
808 error_code = PE_BAD_PORT_NUMBER;
811 port = 10 * port + (*pp - '0');
812 /* Check for too large port numbers here, before we have
813 a chance to overflow on bogus port values. */
816 error_code = PE_BAD_PORT_NUMBER;
821 /* Advance to the first separator *after* '/' (either ';' or '?',
822 depending on the scheme). */
825 /* Get the optional parts of URL, each part being delimited by
826 current location and the position of the next separator. */
827 #define GET_URL_PART(sepchar, var) do { \
829 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
833 GET_URL_PART ('/', path);
834 if (supported_schemes[scheme].flags & scm_has_params)
835 GET_URL_PART (';', params);
836 if (supported_schemes[scheme].flags & scm_has_query)
837 GET_URL_PART ('?', query);
838 if (supported_schemes[scheme].flags & scm_has_fragment)
839 GET_URL_PART ('#', fragment);
844 if (uname_b != uname_e)
846 /* http://user:pass@host */
848 /* uname_b uname_e */
849 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
851 error_code = PE_INVALID_USER_NAME;
856 u = xnew0 (struct url);
858 u->host = strdupdelim (host_b, host_e);
863 u->path = strdupdelim (path_b, path_e);
864 path_modified = path_simplify (scheme, u->path);
865 split_path (u->path, &u->dir, &u->file);
867 host_modified = lowercase_str (u->host);
869 /* Decode %HH sequences in host name. This is important not so much
870 to support %HH sequences in host names (which other browser
871 don't), but to support binary characters (which will have been
872 converted to %HH by reencode_escapes). */
873 if (strchr (u->host, '%'))
875 url_unescape (u->host);
876 host_modified = true;
878 /* Apply IDNA regardless of iri->utf8_encode status */
879 if (opt.enable_iri && iri)
881 char *new = idn_encode (iri, u->host);
886 host_modified = true;
892 u->params = strdupdelim (params_b, params_e);
894 u->query = strdupdelim (query_b, query_e);
896 u->fragment = strdupdelim (fragment_b, fragment_e);
898 if (opt.enable_iri || path_modified || u->fragment || host_modified || path_b == path_e)
900 /* If we suspect that a transformation has rendered what
901 url_string might return different from URL_ENCODED, rebuild
902 u->url using url_string. */
903 u->url = url_string (u, URL_AUTH_SHOW);
905 if (url_encoded != url)
906 xfree ((char *) url_encoded);
910 if (url_encoded == url)
911 u->url = xstrdup (url);
913 u->url = (char *) url_encoded;
919 /* Cleanup in case of error: */
920 if (url_encoded && url_encoded != url)
921 xfree ((char *) url_encoded);
923 /* Transmit the error code to the caller, if the caller wants to
930 /* Return the error message string from ERROR_CODE, which should have
931 been retrieved from url_parse. The error message is translated. */
934 url_error (const char *url, int error_code)
936 assert (error_code >= 0 && ((size_t) error_code) < countof (parse_errors));
938 if (error_code == PE_UNSUPPORTED_SCHEME)
941 char *scheme = xstrdup (url);
942 assert (url_has_scheme (url));
944 if ((p = strchr (scheme, ':')))
946 if (!strcasecmp (scheme, "https"))
947 error = aprintf (_("HTTPS support not compiled in"));
949 error = aprintf (_(parse_errors[error_code]), quote (scheme));
955 return xstrdup (_(parse_errors[error_code]));
958 /* Split PATH into DIR and FILE. PATH comes from the URL and is
959 expected to be URL-escaped.
961 The path is split into directory (the part up to the last slash)
962 and file (the part after the last slash), which are subsequently
966 "foo/bar/baz" "foo/bar" "baz"
967 "foo/bar/" "foo/bar" ""
969 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
971 DIR and FILE are freshly allocated. */
974 split_path (const char *path, char **dir, char **file)
976 char *last_slash = strrchr (path, '/');
980 *file = xstrdup (path);
984 *dir = strdupdelim (path, last_slash);
985 *file = xstrdup (last_slash + 1);
988 url_unescape (*file);
991 /* Note: URL's "full path" is the path with the query string and
992 params appended. The "fragment" (#foo) is intentionally ignored,
993 but that might be changed. For example, if the original URL was
994 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
995 the full path will be "/foo/bar/baz;bullshit?querystring". */
997 /* Return the length of the full path, without the terminating
1001 full_path_length (const struct url *url)
1005 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
1016 /* Write out the full path. */
1019 full_path_write (const struct url *url, char *where)
1021 #define FROB(el, chr) do { \
1022 char *f_el = url->el; \
1024 int l = strlen (f_el); \
1026 memcpy (where, f_el, l); \
1038 /* Public function for getting the "full path". E.g. if u->path is
1039 "foo/bar" and u->query is "param=value", full_path will be
1040 "/foo/bar?param=value". */
1043 url_full_path (const struct url *url)
1045 int length = full_path_length (url);
1046 char *full_path = xmalloc (length + 1);
1048 full_path_write (url, full_path);
1049 full_path[length] = '\0';
1054 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1055 escaping of certain characters, such as "/" and ":". Returns a
1056 count of unescaped chars. */
1059 unescape_single_char (char *str, char chr)
1061 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1062 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1063 char *h = str; /* hare */
1064 char *t = str; /* tortoise */
1065 for (; *h; h++, t++)
1067 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1078 /* Escape unsafe and reserved characters, except for the slash
1082 url_escape_dir (const char *dir)
1084 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1088 unescape_single_char (newdir, '/');
1092 /* Sync u->path and u->url with u->dir and u->file. Called after
1093 u->file or u->dir have been changed, typically by the FTP code. */
1096 sync_path (struct url *u)
1098 char *newpath, *efile, *edir;
1102 /* u->dir and u->file are not escaped. URL-escape them before
1103 reassembling them into u->path. That way, if they contain
1104 separators like '?' or even if u->file contains slashes, the
1105 path will be correctly assembled. (u->file can contain slashes
1106 if the URL specifies it with %2f, or if an FTP server returns
1108 edir = url_escape_dir (u->dir);
1109 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1112 newpath = xstrdup (efile);
1115 int dirlen = strlen (edir);
1116 int filelen = strlen (efile);
1118 /* Copy "DIR/FILE" to newpath. */
1119 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1120 memcpy (p, edir, dirlen);
1123 memcpy (p, efile, filelen);
1132 if (efile != u->file)
1135 /* Regenerate u->url as well. */
1137 u->url = url_string (u, URL_AUTH_SHOW);
1140 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1141 This way we can sync u->path and u->url when they get changed. */
1144 url_set_dir (struct url *url, const char *newdir)
1147 url->dir = xstrdup (newdir);
1152 url_set_file (struct url *url, const char *newfile)
1155 url->file = xstrdup (newfile);
1160 url_free (struct url *url)
1166 xfree_null (url->params);
1167 xfree_null (url->query);
1168 xfree_null (url->fragment);
1169 xfree_null (url->user);
1170 xfree_null (url->passwd);
1178 /* Create all the necessary directories for PATH (a file). Calls
1179 make_directory internally. */
1181 mkalldirs (const char *path)
1188 p = path + strlen (path);
1189 for (; *p != '/' && p != path; p--)
1192 /* Don't create if it's just a file. */
1193 if ((p == path) && (*p != '/'))
1195 t = strdupdelim (path, p);
1197 /* Check whether the directory exists. */
1198 if ((stat (t, &st) == 0))
1200 if (S_ISDIR (st.st_mode))
1207 /* If the dir exists as a file name, remove it first. This
1208 is *only* for Wget to work with buggy old CERN http
1209 servers. Here is the scenario: When Wget tries to
1210 retrieve a directory without a slash, e.g.
1211 http://foo/bar (bar being a directory), CERN server will
1212 not redirect it too http://foo/bar/ -- it will generate a
1213 directory listing containing links to bar/file1,
1214 bar/file2, etc. Wget will lose because it saves this
1215 HTML listing to a file `bar', so it cannot create the
1216 directory. To work around this, if the file of the same
1217 name exists, we just remove it and create the directory
1219 DEBUGP (("Removing %s because of directory danger!\n", t));
1223 res = make_directory (t);
1225 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1230 /* Functions for constructing the file name out of URL components. */
1232 /* A growable string structure, used by url_file_name and friends.
1233 This should perhaps be moved to utils.c.
1235 The idea is to have a convenient and efficient way to construct a
1236 string by having various functions append data to it. Instead of
1237 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1238 functions in questions, we pass the pointer to this struct. */
1246 /* Ensure that the string can accept APPEND_COUNT more characters past
1247 the current TAIL position. If necessary, this will grow the string
1248 and update its allocated size. If the string is already large
1249 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1250 #define GROW(g, append_size) do { \
1251 struct growable *G_ = g; \
1252 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1255 /* Return the tail position of the string. */
1256 #define TAIL(r) ((r)->base + (r)->tail)
1258 /* Move the tail position by APPEND_COUNT characters. */
1259 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1261 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1265 append_string (const char *str, struct growable *dest)
1267 int l = strlen (str);
1269 memcpy (TAIL (dest), str, l);
1270 TAIL_INCR (dest, l);
1273 /* Append CH to DEST. For example, append_char (0, DEST)
1274 zero-terminates DEST. */
1277 append_char (char ch, struct growable *dest)
1281 TAIL_INCR (dest, 1);
1285 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1286 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1287 filechr_control = 4 /* a control character, e.g. 0-31 */
1290 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1292 /* Shorthands for the table: */
1293 #define U filechr_not_unix
1294 #define W filechr_not_windows
1295 #define C filechr_control
1300 /* Table of characters unsafe under various conditions (see above).
1302 Arguably we could also claim `%' to be unsafe, since we use it as
1303 the escape character. If we ever want to be able to reliably
1304 translate file name back to URL, this would become important
1305 crucial. Right now, it's better to be minimal in escaping. */
1307 static const unsigned char filechr_table[256] =
1309 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1310 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1311 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1312 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1313 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1314 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1315 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1316 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1317 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1318 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1319 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1320 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1321 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1322 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1323 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1324 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1326 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1327 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1328 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1329 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1331 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1332 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1333 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1334 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1342 /* FN_PORT_SEP is the separator between host and port in file names
1343 for non-standard port numbers. On Unix this is normally ':', as in
1344 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1345 because Windows can't handle ':' in file names. */
1346 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1348 /* FN_QUERY_SEP is the separator between the file name and the URL
1349 query, normally '?'. Since Windows cannot handle '?' as part of
1350 file name, we use '@' instead there. */
1351 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1353 /* Quote path element, characters in [b, e), as file name, and append
1354 the quoted string to DEST. Each character is quoted as per
1355 file_unsafe_char and the corresponding table.
1357 If ESCAPED is true, the path element is considered to be
1358 URL-escaped and will be unescaped prior to inspection. */
1361 append_uri_pathel (const char *b, const char *e, bool escaped,
1362 struct growable *dest)
1368 if (opt.restrict_files_os == restrict_unix)
1369 mask = filechr_not_unix;
1371 mask = filechr_not_windows;
1372 if (opt.restrict_files_ctrl)
1373 mask |= filechr_control;
1375 /* Copy [b, e) to PATHEL and URL-unescape it. */
1379 BOUNDED_TO_ALLOCA (b, e, unescaped);
1380 url_unescape (unescaped);
1382 e = unescaped + strlen (unescaped);
1385 /* Defang ".." when found as component of path. Remember that path
1386 comes from the URL and might contain malicious input. */
1387 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1393 /* Walk the PATHEL string and check how many characters we'll need
1396 for (p = b; p < e; p++)
1397 if (FILE_CHAR_TEST (*p, mask))
1400 /* Calculate the length of the output string. e-b is the input
1401 string length. Each quoted char introduces two additional
1402 characters in the string, hence 2*quoted. */
1403 outlen = (e - b) + (2 * quoted);
1404 GROW (dest, outlen);
1408 /* If there's nothing to quote, we can simply append the string
1409 without processing it again. */
1410 memcpy (TAIL (dest), b, outlen);
1414 char *q = TAIL (dest);
1415 for (p = b; p < e; p++)
1417 if (!FILE_CHAR_TEST (*p, mask))
1421 unsigned char ch = *p;
1423 *q++ = XNUM_TO_DIGIT (ch >> 4);
1424 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1427 assert (q - TAIL (dest) == outlen);
1430 /* Perform inline case transformation if required. */
1431 if (opt.restrict_files_case == restrict_lowercase
1432 || opt.restrict_files_case == restrict_uppercase)
1435 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1437 if (opt.restrict_files_case == restrict_lowercase)
1438 *q = c_tolower (*q);
1440 *q = c_toupper (*q);
1444 TAIL_INCR (dest, outlen);
1447 /* Append to DEST the directory structure that corresponds the
1448 directory part of URL's path. For example, if the URL is
1449 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1451 Each path element ("dir1" and "dir2" in the above example) is
1452 examined, url-unescaped, and re-escaped as file name element.
1454 Additionally, it cuts as many directories from the path as
1455 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1456 will produce "bar" for the above example. For 2 or more, it will
1459 Each component of the path is quoted for use as file name. */
1462 append_dir_structure (const struct url *u, struct growable *dest)
1464 char *pathel, *next;
1465 int cut = opt.cut_dirs;
1467 /* Go through the path components, de-URL-quote them, and quote them
1468 (if necessary) as file names. */
1471 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1476 /* Ignore empty pathels. */
1480 append_char ('/', dest);
1481 append_uri_pathel (pathel, next, true, dest);
1485 /* Return a unique file name that matches the given URL as good as
1486 possible. Does not create directories on the file system. */
1489 url_file_name (const struct url *u)
1491 struct growable fnres; /* stands for "file name result" */
1493 const char *u_file, *u_query;
1494 char *fname, *unique;
1495 char *index_filename = "index.html"; /* The default index file is index.html */
1501 /* If an alternative index file was defined, change index_filename */
1502 if (opt.default_page)
1503 index_filename = opt.default_page;
1506 /* Start with the directory prefix, if specified. */
1508 append_string (opt.dir_prefix, &fnres);
1510 /* If "dirstruct" is turned on (typically the case with -r), add
1511 the host and port (unless those have been turned off) and
1512 directory structure. */
1515 if (opt.protocol_directories)
1518 append_char ('/', &fnres);
1519 append_string (supported_schemes[u->scheme].name, &fnres);
1521 if (opt.add_hostdir)
1524 append_char ('/', &fnres);
1525 if (0 != strcmp (u->host, ".."))
1526 append_string (u->host, &fnres);
1528 /* Host name can come from the network; malicious DNS may
1529 allow ".." to be resolved, causing us to write to
1530 "../<file>". Defang such host names. */
1531 append_string ("%2E%2E", &fnres);
1532 if (u->port != scheme_default_port (u->scheme))
1535 number_to_string (portstr, u->port);
1536 append_char (FN_PORT_SEP, &fnres);
1537 append_string (portstr, &fnres);
1541 append_dir_structure (u, &fnres);
1544 /* Add the file name. */
1546 append_char ('/', &fnres);
1547 u_file = *u->file ? u->file : index_filename;
1548 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1550 /* Append "?query" to the file name. */
1551 u_query = u->query && *u->query ? u->query : NULL;
1554 append_char (FN_QUERY_SEP, &fnres);
1555 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1558 /* Zero-terminate the file name. */
1559 append_char ('\0', &fnres);
1563 /* Check the cases in which the unique extensions are not used:
1564 1) Clobbering is turned off (-nc).
1565 2) Retrieval with regetting.
1566 3) Timestamping is used.
1567 4) Hierarchy is built.
1569 The exception is the case when file does exist and is a
1570 directory (see `mkalldirs' for explanation). */
1572 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1573 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1576 unique = unique_name (fname, true);
1577 if (unique != fname)
1582 /* Resolve "." and ".." elements of PATH by destructively modifying
1583 PATH and return true if PATH has been modified, false otherwise.
1585 The algorithm is in spirit similar to the one described in rfc1808,
1586 although implemented differently, in one pass. To recap, path
1587 elements containing only "." are removed, and ".." is taken to mean
1588 "back up one element". Single leading and trailing slashes are
1591 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1592 test examples are provided below. If you change anything in this
1593 function, run test_path_simplify to make sure you haven't broken a
1597 path_simplify (enum url_scheme scheme, char *path)
1599 char *h = path; /* hare */
1600 char *t = path; /* tortoise */
1602 char *end = strchr (path, '\0');
1606 /* Hare should be at the beginning of a path element. */
1608 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1613 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1615 /* Handle "../" by retreating the tortoise by one path
1616 element -- but not past beggining. */
1619 /* Move backwards until T hits the beginning of the
1620 previous path element or the beginning of path. */
1621 for (--t; t > beg && t[-1] != '/'; t--)
1624 else if (scheme == SCHEME_FTP)
1626 /* If we're at the beginning, copy the "../" literally
1627 and move the beginning so a later ".." doesn't remove
1628 it. This violates RFC 3986; but we do it for FTP
1629 anyway because there is otherwise no way to get at a
1630 parent directory, when the FTP server drops us in a
1631 non-root directory (which is not uncommon). */
1640 /* A regular path element. If H hasn't advanced past T,
1641 simply skip to the next path element. Otherwise, copy
1642 the path element until the next slash. */
1645 /* Skip the path element, including the slash. */
1646 while (h < end && *h != '/')
1653 /* Copy the path element, including the final slash. */
1654 while (h < end && *h != '/')
1668 /* Return the length of URL's path. Path is considered to be
1669 terminated by one or more of the ?query or ;params or #fragment,
1670 depending on the scheme. */
1673 path_end (const char *url)
1675 enum url_scheme scheme = url_scheme (url);
1677 if (scheme == SCHEME_INVALID)
1678 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1679 /* +2 to ignore the first two separators ':' and '/' */
1680 seps = init_seps (scheme) + 2;
1681 return strpbrk_or_eos (url, seps);
1684 /* Find the last occurrence of character C in the range [b, e), or
1685 NULL, if none are present. */
1686 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1688 /* Merge BASE with LINK and return the resulting URI.
1690 Either of the URIs may be absolute or relative, complete with the
1691 host name, or path only. This tries to reasonably handle all
1692 foreseeable cases. It only employs minimal URL parsing, without
1693 knowledge of the specifics of schemes.
1695 I briefly considered making this function call path_simplify after
1696 the merging process, as rfc1738 seems to suggest. This is a bad
1697 idea for several reasons: 1) it complexifies the code, and 2)
1698 url_parse has to simplify path anyway, so it's wasteful to boot. */
1701 uri_merge (const char *base, const char *link)
1707 if (url_has_scheme (link))
1708 return xstrdup (link);
1710 /* We may not examine BASE past END. */
1711 end = path_end (base);
1712 linklength = strlen (link);
1716 /* Empty LINK points back to BASE, query string and all. */
1717 return xstrdup (base);
1719 else if (*link == '?')
1721 /* LINK points to the same location, but changes the query
1722 string. Examples: */
1723 /* uri_merge("path", "?new") -> "path?new" */
1724 /* uri_merge("path?foo", "?new") -> "path?new" */
1725 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1726 /* uri_merge("path#foo", "?new") -> "path?new" */
1727 int baselength = end - base;
1728 merge = xmalloc (baselength + linklength + 1);
1729 memcpy (merge, base, baselength);
1730 memcpy (merge + baselength, link, linklength);
1731 merge[baselength + linklength] = '\0';
1733 else if (*link == '#')
1735 /* uri_merge("path", "#new") -> "path#new" */
1736 /* uri_merge("path#foo", "#new") -> "path#new" */
1737 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1738 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1740 const char *end1 = strchr (base, '#');
1742 end1 = base + strlen (base);
1743 baselength = end1 - base;
1744 merge = xmalloc (baselength + linklength + 1);
1745 memcpy (merge, base, baselength);
1746 memcpy (merge + baselength, link, linklength);
1747 merge[baselength + linklength] = '\0';
1749 else if (*link == '/' && *(link + 1) == '/')
1751 /* LINK begins with "//" and so is a net path: we need to
1752 replace everything after (and including) the double slash
1755 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1756 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1757 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1761 const char *start_insert;
1763 /* Look for first slash. */
1764 slash = memchr (base, '/', end - base);
1765 /* If found slash and it is a double slash, then replace
1766 from this point, else default to replacing from the
1768 if (slash && *(slash + 1) == '/')
1769 start_insert = slash;
1771 start_insert = base;
1773 span = start_insert - base;
1774 merge = xmalloc (span + linklength + 1);
1776 memcpy (merge, base, span);
1777 memcpy (merge + span, link, linklength);
1778 merge[span + linklength] = '\0';
1780 else if (*link == '/')
1782 /* LINK is an absolute path: we need to replace everything
1783 after (and including) the FIRST slash with LINK.
1785 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1786 "/qux/xyzzy", our result should be
1787 "http://host/qux/xyzzy". */
1790 const char *start_insert = NULL; /* for gcc to shut up. */
1791 const char *pos = base;
1792 bool seen_slash_slash = false;
1793 /* We're looking for the first slash, but want to ignore
1796 slash = memchr (pos, '/', end - pos);
1797 if (slash && !seen_slash_slash)
1798 if (*(slash + 1) == '/')
1801 seen_slash_slash = true;
1805 /* At this point, SLASH is the location of the first / after
1806 "//", or the first slash altogether. START_INSERT is the
1807 pointer to the location where LINK will be inserted. When
1808 examining the last two examples, keep in mind that LINK
1811 if (!slash && !seen_slash_slash)
1812 /* example: "foo" */
1814 start_insert = base;
1815 else if (!slash && seen_slash_slash)
1816 /* example: "http://foo" */
1819 else if (slash && !seen_slash_slash)
1820 /* example: "foo/bar" */
1822 start_insert = base;
1823 else if (slash && seen_slash_slash)
1824 /* example: "http://something/" */
1826 start_insert = slash;
1828 span = start_insert - base;
1829 merge = xmalloc (span + linklength + 1);
1831 memcpy (merge, base, span);
1832 memcpy (merge + span, link, linklength);
1833 merge[span + linklength] = '\0';
1837 /* LINK is a relative URL: we need to replace everything
1838 after last slash (possibly empty) with LINK.
1840 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1841 our result should be "whatever/foo/qux/xyzzy". */
1842 bool need_explicit_slash = false;
1844 const char *start_insert;
1845 const char *last_slash = find_last_char (base, end, '/');
1848 /* No slash found at all. Replace what we have with LINK. */
1849 start_insert = base;
1851 else if (last_slash && last_slash >= base + 2
1852 && last_slash[-2] == ':' && last_slash[-1] == '/')
1854 /* example: http://host" */
1856 start_insert = end + 1;
1857 need_explicit_slash = true;
1861 /* example: "whatever/foo/bar" */
1863 start_insert = last_slash + 1;
1866 span = start_insert - base;
1867 merge = xmalloc (span + linklength + 1);
1869 memcpy (merge, base, span);
1870 if (need_explicit_slash)
1871 merge[span - 1] = '/';
1872 memcpy (merge + span, link, linklength);
1873 merge[span + linklength] = '\0';
1879 #define APPEND(p, s) do { \
1880 int len = strlen (s); \
1881 memcpy (p, s, len); \
1885 /* Use this instead of password when the actual password is supposed
1886 to be hidden. We intentionally use a generic string without giving
1887 away the number of characters in the password, like previous
1889 #define HIDDEN_PASSWORD "*password*"
1891 /* Recreate the URL string from the data in URL.
1893 If HIDE is true (as it is when we're calling this on a URL we plan
1894 to print, but not when calling it to canonicalize a URL for use
1895 within the program), password will be hidden. Unsafe characters in
1896 the URL will be quoted. */
1899 url_string (const struct url *url, enum url_auth_mode auth_mode)
1903 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1905 int scheme_port = supported_schemes[url->scheme].default_port;
1906 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1907 int fplen = full_path_length (url);
1909 bool brackets_around_host;
1911 assert (scheme_str != NULL);
1913 /* Make sure the user name and password are quoted. */
1916 if (auth_mode != URL_AUTH_HIDE)
1918 quoted_user = url_escape_allow_passthrough (url->user);
1921 if (auth_mode == URL_AUTH_HIDE_PASSWD)
1922 quoted_passwd = HIDDEN_PASSWORD;
1924 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1929 /* In the unlikely event that the host name contains non-printable
1930 characters, quote it for displaying to the user. */
1931 quoted_host = url_escape_allow_passthrough (url->host);
1933 /* Undo the quoting of colons that URL escaping performs. IPv6
1934 addresses may legally contain colons, and in that case must be
1935 placed in square brackets. */
1936 if (quoted_host != url->host)
1937 unescape_single_char (quoted_host, ':');
1938 brackets_around_host = strchr (quoted_host, ':') != NULL;
1940 size = (strlen (scheme_str)
1941 + strlen (quoted_host)
1942 + (brackets_around_host ? 2 : 0)
1945 if (url->port != scheme_port)
1946 size += 1 + numdigit (url->port);
1949 size += 1 + strlen (quoted_user);
1951 size += 1 + strlen (quoted_passwd);
1954 p = result = xmalloc (size);
1956 APPEND (p, scheme_str);
1959 APPEND (p, quoted_user);
1963 APPEND (p, quoted_passwd);
1968 if (brackets_around_host)
1970 APPEND (p, quoted_host);
1971 if (brackets_around_host)
1973 if (url->port != scheme_port)
1976 p = number_to_string (p, url->port);
1979 full_path_write (url, p);
1983 assert (p - result == size);
1985 if (quoted_user && quoted_user != url->user)
1986 xfree (quoted_user);
1987 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
1988 && quoted_passwd != url->passwd)
1989 xfree (quoted_passwd);
1990 if (quoted_host != url->host)
1991 xfree (quoted_host);
1996 /* Return true if scheme a is similar to scheme b.
1998 Schemes are similar if they are equal. If SSL is supported, schemes
1999 are also similar if one is http (SCHEME_HTTP) and the other is https
2002 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
2007 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
2008 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
2015 getchar_from_escaped_string (const char *str, char *c)
2017 const char *p = str;
2019 assert (str && *str);
2024 if (!c_isxdigit(p[1]) || !c_isxdigit(p[2]))
2032 return 0; /* error: invalid string */
2034 *c = X2DIGITS_TO_NUM (p[1], p[2]);
2035 if (URL_RESERVED_CHAR(*c))
2053 are_urls_equal (const char *u1, const char *u2)
2064 && (pp = getchar_from_escaped_string (p, &ch1))
2065 && (qq = getchar_from_escaped_string (q, &ch2))
2066 && (c_tolower(ch1) == c_tolower(ch2)))
2072 return (*p == 0 && *q == 0 ? true : false);
2076 /* Debugging and testing support for path_simplify. */
2079 /* Debug: run path_simplify on PATH and return the result in a new
2080 string. Useful for calling from the debugger. */
2084 char *copy = xstrdup (path);
2085 path_simplify (copy);
2091 run_test (char *test, char *expected_result, enum url_scheme scheme,
2092 bool expected_change)
2094 char *test_copy = xstrdup (test);
2095 bool modified = path_simplify (scheme, test_copy);
2097 if (0 != strcmp (test_copy, expected_result))
2099 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2100 test, expected_result, test_copy);
2103 if (modified != expected_change)
2105 if (expected_change)
2106 printf ("Expected modification with path_simplify(\"%s\").\n",
2109 printf ("Expected no modification with path_simplify(\"%s\").\n",
2113 mu_assert ("", modified == expected_change);
2118 test_path_simplify (void)
2121 char *test, *result;
2122 enum url_scheme scheme;
2125 { "", "", SCHEME_HTTP, false },
2126 { ".", "", SCHEME_HTTP, true },
2127 { "./", "", SCHEME_HTTP, true },
2128 { "..", "", SCHEME_HTTP, true },
2129 { "../", "", SCHEME_HTTP, true },
2130 { "..", "..", SCHEME_FTP, false },
2131 { "../", "../", SCHEME_FTP, false },
2132 { "foo", "foo", SCHEME_HTTP, false },
2133 { "foo/bar", "foo/bar", SCHEME_HTTP, false },
2134 { "foo///bar", "foo///bar", SCHEME_HTTP, false },
2135 { "foo/.", "foo/", SCHEME_HTTP, true },
2136 { "foo/./", "foo/", SCHEME_HTTP, true },
2137 { "foo./", "foo./", SCHEME_HTTP, false },
2138 { "foo/../bar", "bar", SCHEME_HTTP, true },
2139 { "foo/../bar/", "bar/", SCHEME_HTTP, true },
2140 { "foo/bar/..", "foo/", SCHEME_HTTP, true },
2141 { "foo/bar/../x", "foo/x", SCHEME_HTTP, true },
2142 { "foo/bar/../x/", "foo/x/", SCHEME_HTTP, true },
2143 { "foo/..", "", SCHEME_HTTP, true },
2144 { "foo/../..", "", SCHEME_HTTP, true },
2145 { "foo/../../..", "", SCHEME_HTTP, true },
2146 { "foo/../../bar/../../baz", "baz", SCHEME_HTTP, true },
2147 { "foo/../..", "..", SCHEME_FTP, true },
2148 { "foo/../../..", "../..", SCHEME_FTP, true },
2149 { "foo/../../bar/../../baz", "../../baz", SCHEME_FTP, true },
2150 { "a/b/../../c", "c", SCHEME_HTTP, true },
2151 { "./a/../b", "b", SCHEME_HTTP, true }
2155 for (i = 0; i < countof (tests); i++)
2157 const char *message;
2158 char *test = tests[i].test;
2159 char *expected_result = tests[i].result;
2160 enum url_scheme scheme = tests[i].scheme;
2161 bool expected_change = tests[i].should_modify;
2162 message = run_test (test, expected_result, scheme, expected_change);
2163 if (message) return message;
2169 test_append_uri_pathel()
2176 char *expected_result;
2178 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2181 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2183 struct growable dest;
2184 const char *p = test_array[i].input;
2186 memset (&dest, 0, sizeof (dest));
2188 append_string (test_array[i].original_url, &dest);
2189 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2190 append_char ('\0', &dest);
2192 mu_assert ("test_append_uri_pathel: wrong result",
2193 strcmp (dest.base, test_array[i].expected_result) == 0);
2200 test_are_urls_equal()
2206 bool expected_result;
2208 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2209 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2210 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2211 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2212 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2213 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2216 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2218 mu_assert ("test_are_urls_equal: wrong result",
2219 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2225 #endif /* TESTING */