2 Copyright (C) 1996-2005 Free Software Foundation, Inc.
4 This file is part of GNU Wget.
6 GNU Wget is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or (at
9 your option) any later version.
11 GNU Wget is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with Wget; if not, write to the Free Software Foundation, Inc.,
18 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
20 In addition, as a special exception, the Free Software Foundation
21 gives permission to link the code of its release of Wget with the
22 OpenSSL project's "OpenSSL" library (or with modified versions of it
23 that use the same license as the "OpenSSL" library), and distribute
24 the linked executables. You must obey the GNU General Public License
25 in all respects for all of the code used other than "OpenSSL". If you
26 modify this file, you may extend this exception to your version of the
27 file, but you are not obligated to do so. If you do not wish to do
28 so, delete this exception statement from your version. */
44 #include "host.h" /* for is_valid_ipv6_address */
47 scm_disabled = 1, /* for https when OpenSSL fails to init. */
48 scm_has_params = 2, /* whether scheme has ;params */
49 scm_has_query = 4, /* whether scheme has ?query */
50 scm_has_fragment = 8 /* whether scheme has #fragment */
55 /* Short name of the scheme, such as "http" or "ftp". */
57 /* Leading string that identifies the scheme, such as "https://". */
58 const char *leading_string;
59 /* Default port of the scheme when none is specified. */
65 /* Supported schemes: */
66 static struct scheme_data supported_schemes[] =
68 { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment },
70 { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment },
72 { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment },
78 /* Forward declarations: */
80 static bool path_simplify (char *);
82 /* Support for escaping and unescaping of URL strings. */
84 /* Table of "reserved" and "unsafe" characters. Those terms are
85 rfc1738-speak, as such largely obsoleted by rfc2396 and later
86 specs, but the general idea remains.
88 A reserved character is the one that you can't decode without
89 changing the meaning of the URL. For example, you can't decode
90 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
91 path components is different. Non-reserved characters can be
92 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
93 unsafe characters are loosely based on rfc1738, plus "$" and ",",
94 as recommended by rfc2396, and minus "~", which is very frequently
95 used (and sometimes unrecognized as %7E by broken servers).
97 An unsafe character is the one that should be encoded when URLs are
98 placed in foreign environments. E.g. space and newline are unsafe
99 in HTTP contexts because HTTP uses them as separator and line
100 terminator, so they must be encoded to %20 and %0A respectively.
101 "*" is unsafe in shell context, etc.
103 We determine whether a character is unsafe through static table
104 lookup. This code assumes ASCII character set and 8-bit chars. */
107 /* rfc1738 reserved chars + "$" and ",". */
110 /* rfc1738 unsafe chars, plus non-printables. */
114 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
115 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
116 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
118 /* Shorthands for the table: */
119 #define R urlchr_reserved
120 #define U urlchr_unsafe
123 static const unsigned char urlchr_table[256] =
125 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
126 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
127 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
128 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
129 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
130 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
131 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
132 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
133 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
134 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
135 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
136 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
137 U, 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, U, U, U, 0, U, /* x y z { | } ~ DEL */
142 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
143 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
144 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
145 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,
150 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
156 /* URL-unescape the string S.
158 This is done by transforming the sequences "%HH" to the character
159 represented by the hexadecimal digits HH. If % is not followed by
160 two hexadecimal digits, it is inserted literally.
162 The transformation is done in place. If you need the original
163 string intact, make a copy before calling this function. */
166 url_unescape (char *s)
168 char *t = s; /* t - tortoise */
169 char *h = s; /* h - hare */
181 /* Do nothing if '%' is not followed by two hex digits. */
182 if (!h[1] || !h[2] || !(ISXDIGIT (h[1]) && ISXDIGIT (h[2])))
184 c = X2DIGITS_TO_NUM (h[1], h[2]);
185 /* Don't unescape %00 because there is no way to insert it
186 into a C string without effectively truncating it. */
196 /* The core of url_escape_* functions. Escapes the characters that
197 match the provided mask in urlchr_table.
199 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
200 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
201 allocated string will be returned in all cases. */
204 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
211 for (p1 = s; *p1; p1++)
212 if (urlchr_test (*p1, mask))
213 addition += 2; /* Two more characters (hex digits) */
216 return allow_passthrough ? (char *)s : xstrdup (s);
218 newlen = (p1 - s) + addition;
219 newstr = xmalloc (newlen + 1);
225 /* Quote the characters that match the test mask. */
226 if (urlchr_test (*p1, mask))
228 unsigned char c = *p1++;
230 *p2++ = XNUM_TO_DIGIT (c >> 4);
231 *p2++ = XNUM_TO_DIGIT (c & 0xf);
236 assert (p2 - newstr == newlen);
242 /* URL-escape the unsafe characters (see urlchr_table) in a given
243 string, returning a freshly allocated string. */
246 url_escape (const char *s)
248 return url_escape_1 (s, urlchr_unsafe, false);
251 /* URL-escape the unsafe characters (see urlchr_table) in a given
252 string. If no characters are unsafe, S is returned. */
255 url_escape_allow_passthrough (const char *s)
257 return url_escape_1 (s, urlchr_unsafe, true);
260 /* Decide whether the char at position P needs to be encoded. (It is
261 not enough to pass a single char *P because the function may need
262 to inspect the surrounding context.)
264 Return true if the char should be escaped as %XX, false otherwise. */
267 char_needs_escaping (const char *p)
271 if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2)))
274 /* Garbled %.. sequence: encode `%'. */
277 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
283 /* Translate a %-escaped (but possibly non-conformant) input string S
284 into a %-escaped (and conformant) output string. If no characters
285 are encoded or decoded, return the same string S; otherwise, return
286 a freshly allocated string with the new contents.
288 After a URL has been run through this function, the protocols that
289 use `%' as the quote character can use the resulting string as-is,
290 while those that don't can use url_unescape to get to the intended
291 data. This function is stable: once the input is transformed,
292 further transformations of the result yield the same output.
294 Let's discuss why this function is needed.
296 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
297 a raw space character would mess up the HTTP request, it needs to
298 be quoted, like this:
300 GET /abc%20def HTTP/1.0
302 It would appear that the unsafe chars need to be quoted, for
303 example with url_escape. But what if we're requested to download
304 `abc%20def'? url_escape transforms "%" to "%25", which would leave
305 us with `abc%2520def'. This is incorrect -- since %-escapes are
306 part of URL syntax, "%20" is the correct way to denote a literal
307 space on the Wget command line. This leads to the conclusion that
308 in that case Wget should not call url_escape, but leave the `%20'
309 as is. This is clearly contradictory, but it only gets worse.
311 What if the requested URI is `abc%20 def'? If we call url_escape,
312 we end up with `/abc%2520%20def', which is almost certainly not
313 intended. If we don't call url_escape, we are left with the
314 embedded space and cannot complete the request. What the user
315 meant was for Wget to request `/abc%20%20def', and this is where
316 reencode_escapes kicks in.
318 Wget used to solve this by first decoding %-quotes, and then
319 encoding all the "unsafe" characters found in the resulting string.
320 This was wrong because it didn't preserve certain URL special
321 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
322 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
323 whether we considered `+' reserved (it is). One of these results
324 is inevitable because by the second step we would lose information
325 on whether the `+' was originally encoded or not. Both results
326 were wrong because in CGI parameters + means space, while %2B means
327 literal plus. reencode_escapes correctly translates the above to
328 "a%2B+b", i.e. returns the original string.
330 This function uses a modified version of the algorithm originally
331 proposed by Anon Sricharoenchai:
333 * Encode all "unsafe" characters, except those that are also
334 "reserved", to %XX. See urlchr_table for which characters are
337 * Encode the "%" characters not followed by two hex digits to
340 * Pass through all other characters and %XX escapes as-is. (Up to
341 Wget 1.10 this decoded %XX escapes corresponding to "safe"
342 characters, but that was obtrusive and broke some servers.)
346 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
348 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
352 "foo bar" -> "foo%20bar"
353 "foo%20bar" -> "foo%20bar"
354 "foo %20bar" -> "foo%20%20bar"
355 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
356 "foo%25%20bar" -> "foo%25%20bar"
357 "foo%2%20bar" -> "foo%252%20bar"
358 "foo+bar" -> "foo+bar" (plus is reserved!)
359 "foo%2b+bar" -> "foo%2b+bar" */
362 reencode_escapes (const char *s)
368 int encode_count = 0;
370 /* First pass: inspect the string to see if there's anything to do,
371 and to calculate the new length. */
372 for (p1 = s; *p1; p1++)
373 if (char_needs_escaping (p1))
377 /* The string is good as it is. */
378 return (char *) s; /* C const model sucks. */
381 /* Each encoding adds two characters (hex digits). */
382 newlen = oldlen + 2 * encode_count;
383 newstr = xmalloc (newlen + 1);
385 /* Second pass: copy the string to the destination address, encoding
386 chars when needed. */
391 if (char_needs_escaping (p1))
393 unsigned char c = *p1++;
395 *p2++ = XNUM_TO_DIGIT (c >> 4);
396 *p2++ = XNUM_TO_DIGIT (c & 0xf);
402 assert (p2 - newstr == newlen);
406 /* Returns the scheme type if the scheme is supported, or
407 SCHEME_INVALID if not. */
410 url_scheme (const char *url)
414 for (i = 0; supported_schemes[i].leading_string; i++)
415 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
416 strlen (supported_schemes[i].leading_string)))
418 if (!(supported_schemes[i].flags & scm_disabled))
419 return (enum url_scheme) i;
421 return SCHEME_INVALID;
424 return SCHEME_INVALID;
427 #define SCHEME_CHAR(ch) (ISALNUM (ch) || (ch) == '-' || (ch) == '+')
429 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
430 currently implemented, it returns true if URL begins with
434 url_has_scheme (const char *url)
438 /* The first char must be a scheme char. */
439 if (!*p || !SCHEME_CHAR (*p))
442 /* Followed by 0 or more scheme chars. */
443 while (*p && SCHEME_CHAR (*p))
445 /* Terminated by ':'. */
450 scheme_default_port (enum url_scheme scheme)
452 return supported_schemes[scheme].default_port;
456 scheme_disable (enum url_scheme scheme)
458 supported_schemes[scheme].flags |= scm_disabled;
461 /* Skip the username and password, if present in the URL. The
462 function should *not* be called with the complete URL, but with the
463 portion after the scheme.
465 If no username and password are found, return URL. */
468 url_skip_credentials (const char *url)
470 /* Look for '@' that comes before terminators, such as '/', '?',
472 const char *p = (const char *)strpbrk (url, "@/?#;");
478 /* Parse credentials contained in [BEG, END). The region is expected
479 to have come from a URL and is unescaped. */
482 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
488 return false; /* empty user name */
490 colon = memchr (beg, ':', end - beg);
492 return false; /* again empty user name */
496 *passwd = strdupdelim (colon + 1, end);
498 url_unescape (*passwd);
505 *user = strdupdelim (beg, userend);
506 url_unescape (*user);
510 /* Used by main.c: detect URLs written using the "shorthand" URL forms
511 originally popularized by Netscape and NcFTP. HTTP shorthands look
514 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
515 www.foo.com[:port] -> http://www.foo.com[:port]
517 FTP shorthands look like this:
519 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
520 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
522 If the URL needs not or cannot be rewritten, return NULL. */
525 rewrite_shorthand_url (const char *url)
530 if (url_scheme (url) != SCHEME_INVALID)
533 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
535 p = strpbrk (url, ":/");
539 /* If we're looking at "://", it means the URL uses a scheme we
540 don't support, which may include "https" when compiled without
541 SSL support. Don't bogusly rewrite such URLs. */
542 if (p && p[0] == ':' && p[1] == '/' && p[2] == '/')
547 /* Colon indicates ftp, as in foo.bar.com:path. Check for
548 special case of http port number ("localhost:10000"). */
549 int digits = strspn (p + 1, "0123456789");
550 if (digits && (p[1 + digits] == '/' || p[1 + digits] == '\0'))
553 /* Turn "foo.bar.com:path" to "ftp://foo.bar.com/path". */
554 ret = aprintf ("ftp://%s", url);
555 ret[6 + (p - url)] = '/';
560 /* Just prepend "http://" to URL. */
561 ret = aprintf ("http://%s", url);
566 static void split_path (const char *, char **, char **);
568 /* Like strpbrk, with the exception that it returns the pointer to the
569 terminating zero (end-of-string aka "eos") if no matching character
573 strpbrk_or_eos (const char *s, const char *accept)
575 char *p = strpbrk (s, accept);
577 p = strchr (s, '\0');
581 /* Turn STR into lowercase; return true if a character was actually
585 lowercase_str (char *str)
587 bool changed = false;
592 *str = TOLOWER (*str);
598 init_seps (enum url_scheme scheme)
600 static char seps[8] = ":/";
602 int flags = supported_schemes[scheme].flags;
604 if (flags & scm_has_params)
606 if (flags & scm_has_query)
608 if (flags & scm_has_fragment)
614 static const char *parse_errors[] = {
615 #define PE_NO_ERROR 0
617 #define PE_UNSUPPORTED_SCHEME 1
618 N_("Unsupported scheme"),
619 #define PE_INVALID_HOST_NAME 2
620 N_("Invalid host name"),
621 #define PE_BAD_PORT_NUMBER 3
622 N_("Bad port number"),
623 #define PE_INVALID_USER_NAME 4
624 N_("Invalid user name"),
625 #define PE_UNTERMINATED_IPV6_ADDRESS 5
626 N_("Unterminated IPv6 numeric address"),
627 #define PE_IPV6_NOT_SUPPORTED 6
628 N_("IPv6 addresses not supported"),
629 #define PE_INVALID_IPV6_ADDRESS 7
630 N_("Invalid IPv6 numeric address")
635 Return a new struct url if successful, NULL on error. In case of
636 error, and if ERROR is not NULL, also set *ERROR to the appropriate
639 url_parse (const char *url, int *error)
643 bool path_modified, host_modified;
645 enum url_scheme scheme;
648 const char *uname_b, *uname_e;
649 const char *host_b, *host_e;
650 const char *path_b, *path_e;
651 const char *params_b, *params_e;
652 const char *query_b, *query_e;
653 const char *fragment_b, *fragment_e;
656 char *user = NULL, *passwd = NULL;
658 char *url_encoded = NULL;
662 scheme = url_scheme (url);
663 if (scheme == SCHEME_INVALID)
665 error_code = PE_UNSUPPORTED_SCHEME;
669 url_encoded = reencode_escapes (url);
672 p += strlen (supported_schemes[scheme].leading_string);
674 p = url_skip_credentials (p);
677 /* scheme://user:pass@host[:port]... */
680 /* We attempt to break down the URL into the components path,
681 params, query, and fragment. They are ordered like this:
683 scheme://host[:port][/path][;params][?query][#fragment] */
685 path_b = path_e = NULL;
686 params_b = params_e = NULL;
687 query_b = query_e = NULL;
688 fragment_b = fragment_e = NULL;
690 /* Initialize separators for optional parts of URL, depending on the
691 scheme. For example, FTP has params, and HTTP and HTTPS have
692 query string and fragment. */
693 seps = init_seps (scheme);
699 /* Handle IPv6 address inside square brackets. Ideally we'd
700 just look for the terminating ']', but rfc2732 mandates
701 rejecting invalid IPv6 addresses. */
703 /* The address begins after '['. */
705 host_e = strchr (host_b, ']');
709 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
714 /* Check if the IPv6 address is valid. */
715 if (!is_valid_ipv6_address(host_b, host_e))
717 error_code = PE_INVALID_IPV6_ADDRESS;
721 /* Continue parsing after the closing ']'. */
724 error_code = PE_IPV6_NOT_SUPPORTED;
728 /* The closing bracket must be followed by a separator or by the
730 /* http://[::1]... */
732 if (!strchr (seps, *p))
734 /* Trailing garbage after []-delimited IPv6 address. */
735 error_code = PE_INVALID_HOST_NAME;
741 p = strpbrk_or_eos (p, seps);
744 ++seps; /* advance to '/' */
746 if (host_b == host_e)
748 error_code = PE_INVALID_HOST_NAME;
752 port = scheme_default_port (scheme);
755 const char *port_b, *port_e, *pp;
757 /* scheme://host:port/tralala */
761 p = strpbrk_or_eos (p, seps);
764 /* Allow empty port, as per rfc2396. */
765 if (port_b != port_e)
766 for (port = 0, pp = port_b; pp < port_e; pp++)
770 /* http://host:12randomgarbage/blah */
772 error_code = PE_BAD_PORT_NUMBER;
775 port = 10 * port + (*pp - '0');
776 /* Check for too large port numbers here, before we have
777 a chance to overflow on bogus port values. */
780 error_code = PE_BAD_PORT_NUMBER;
785 /* Advance to the first separator *after* '/' (either ';' or '?',
786 depending on the scheme). */
789 /* Get the optional parts of URL, each part being delimited by
790 current location and the position of the next separator. */
791 #define GET_URL_PART(sepchar, var) do { \
793 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
797 GET_URL_PART ('/', path);
798 if (supported_schemes[scheme].flags & scm_has_params)
799 GET_URL_PART (';', params);
800 if (supported_schemes[scheme].flags & scm_has_query)
801 GET_URL_PART ('?', query);
802 if (supported_schemes[scheme].flags & scm_has_fragment)
803 GET_URL_PART ('#', fragment);
808 if (uname_b != uname_e)
810 /* http://user:pass@host */
812 /* uname_b uname_e */
813 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
815 error_code = PE_INVALID_USER_NAME;
820 u = xnew0 (struct url);
822 u->host = strdupdelim (host_b, host_e);
827 u->path = strdupdelim (path_b, path_e);
828 path_modified = path_simplify (u->path);
829 split_path (u->path, &u->dir, &u->file);
831 host_modified = lowercase_str (u->host);
833 /* Decode %HH sequences in host name. This is important not so much
834 to support %HH sequences in host names (which other browser
835 don't), but to support binary characters (which will have been
836 converted to %HH by reencode_escapes). */
837 if (strchr (u->host, '%'))
839 url_unescape (u->host);
840 host_modified = true;
844 u->params = strdupdelim (params_b, params_e);
846 u->query = strdupdelim (query_b, query_e);
848 u->fragment = strdupdelim (fragment_b, fragment_e);
850 if (path_modified || u->fragment || host_modified || path_b == path_e)
852 /* If we suspect that a transformation has rendered what
853 url_string might return different from URL_ENCODED, rebuild
854 u->url using url_string. */
855 u->url = url_string (u, false);
857 if (url_encoded != url)
858 xfree ((char *) url_encoded);
862 if (url_encoded == url)
863 u->url = xstrdup (url);
865 u->url = url_encoded;
871 /* Cleanup in case of error: */
872 if (url_encoded && url_encoded != url)
875 /* Transmit the error code to the caller, if the caller wants to
882 /* Return the error message string from ERROR_CODE, which should have
883 been retrieved from url_parse. The error message is translated. */
886 url_error (int error_code)
888 assert (error_code >= 0 && error_code < countof (parse_errors));
889 return _(parse_errors[error_code]);
892 /* Split PATH into DIR and FILE. PATH comes from the URL and is
893 expected to be URL-escaped.
895 The path is split into directory (the part up to the last slash)
896 and file (the part after the last slash), which are subsequently
900 "foo/bar/baz" "foo/bar" "baz"
901 "foo/bar/" "foo/bar" ""
903 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
905 DIR and FILE are freshly allocated. */
908 split_path (const char *path, char **dir, char **file)
910 char *last_slash = strrchr (path, '/');
914 *file = xstrdup (path);
918 *dir = strdupdelim (path, last_slash);
919 *file = xstrdup (last_slash + 1);
922 url_unescape (*file);
925 /* Note: URL's "full path" is the path with the query string and
926 params appended. The "fragment" (#foo) is intentionally ignored,
927 but that might be changed. For example, if the original URL was
928 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
929 the full path will be "/foo/bar/baz;bullshit?querystring". */
931 /* Return the length of the full path, without the terminating
935 full_path_length (const struct url *url)
939 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
950 /* Write out the full path. */
953 full_path_write (const struct url *url, char *where)
955 #define FROB(el, chr) do { \
956 char *f_el = url->el; \
958 int l = strlen (f_el); \
960 memcpy (where, f_el, l); \
972 /* Public function for getting the "full path". E.g. if u->path is
973 "foo/bar" and u->query is "param=value", full_path will be
974 "/foo/bar?param=value". */
977 url_full_path (const struct url *url)
979 int length = full_path_length (url);
980 char *full_path = xmalloc (length + 1);
982 full_path_write (url, full_path);
983 full_path[length] = '\0';
988 /* Unescape CHR in an otherwise escaped STR. Used to selectively
989 escaping of certain characters, such as "/" and ":". Returns a
990 count of unescaped chars. */
993 unescape_single_char (char *str, char chr)
995 const char c1 = XNUM_TO_DIGIT (chr >> 4);
996 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
997 char *h = str; /* hare */
998 char *t = str; /* tortoise */
1001 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1012 /* Escape unsafe and reserved characters, except for the slash
1016 url_escape_dir (const char *dir)
1018 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1022 unescape_single_char (newdir, '/');
1026 /* Sync u->path and u->url with u->dir and u->file. Called after
1027 u->file or u->dir have been changed, typically by the FTP code. */
1030 sync_path (struct url *u)
1032 char *newpath, *efile, *edir;
1036 /* u->dir and u->file are not escaped. URL-escape them before
1037 reassembling them into u->path. That way, if they contain
1038 separators like '?' or even if u->file contains slashes, the
1039 path will be correctly assembled. (u->file can contain slashes
1040 if the URL specifies it with %2f, or if an FTP server returns
1042 edir = url_escape_dir (u->dir);
1043 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1046 newpath = xstrdup (efile);
1049 int dirlen = strlen (edir);
1050 int filelen = strlen (efile);
1052 /* Copy "DIR/FILE" to newpath. */
1053 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1054 memcpy (p, edir, dirlen);
1057 memcpy (p, efile, filelen);
1066 if (efile != u->file)
1069 /* Regenerate u->url as well. */
1071 u->url = url_string (u, false);
1074 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1075 This way we can sync u->path and u->url when they get changed. */
1078 url_set_dir (struct url *url, const char *newdir)
1081 url->dir = xstrdup (newdir);
1086 url_set_file (struct url *url, const char *newfile)
1089 url->file = xstrdup (newfile);
1094 url_free (struct url *url)
1100 xfree_null (url->params);
1101 xfree_null (url->query);
1102 xfree_null (url->fragment);
1103 xfree_null (url->user);
1104 xfree_null (url->passwd);
1112 /* Create all the necessary directories for PATH (a file). Calls
1113 make_directory internally. */
1115 mkalldirs (const char *path)
1122 p = path + strlen (path);
1123 for (; *p != '/' && p != path; p--)
1126 /* Don't create if it's just a file. */
1127 if ((p == path) && (*p != '/'))
1129 t = strdupdelim (path, p);
1131 /* Check whether the directory exists. */
1132 if ((stat (t, &st) == 0))
1134 if (S_ISDIR (st.st_mode))
1141 /* If the dir exists as a file name, remove it first. This
1142 is *only* for Wget to work with buggy old CERN http
1143 servers. Here is the scenario: When Wget tries to
1144 retrieve a directory without a slash, e.g.
1145 http://foo/bar (bar being a directory), CERN server will
1146 not redirect it too http://foo/bar/ -- it will generate a
1147 directory listing containing links to bar/file1,
1148 bar/file2, etc. Wget will lose because it saves this
1149 HTML listing to a file `bar', so it cannot create the
1150 directory. To work around this, if the file of the same
1151 name exists, we just remove it and create the directory
1153 DEBUGP (("Removing %s because of directory danger!\n", t));
1157 res = make_directory (t);
1159 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1164 /* Functions for constructing the file name out of URL components. */
1166 /* A growable string structure, used by url_file_name and friends.
1167 This should perhaps be moved to utils.c.
1169 The idea is to have a convenient and efficient way to construct a
1170 string by having various functions append data to it. Instead of
1171 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1172 functions in questions, we pass the pointer to this struct. */
1180 /* Ensure that the string can accept APPEND_COUNT more characters past
1181 the current TAIL position. If necessary, this will grow the string
1182 and update its allocated size. If the string is already large
1183 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1184 #define GROW(g, append_size) do { \
1185 struct growable *G_ = g; \
1186 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1189 /* Return the tail position of the string. */
1190 #define TAIL(r) ((r)->base + (r)->tail)
1192 /* Move the tail position by APPEND_COUNT characters. */
1193 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1195 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1199 append_string (const char *str, struct growable *dest)
1201 int l = strlen (str);
1203 memcpy (TAIL (dest), str, l);
1204 TAIL_INCR (dest, l);
1207 /* Append CH to DEST. For example, append_char (0, DEST)
1208 zero-terminates DEST. */
1211 append_char (char ch, struct growable *dest)
1215 TAIL_INCR (dest, 1);
1219 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1220 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1221 filechr_control = 4 /* a control character, e.g. 0-31 */
1224 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1226 /* Shorthands for the table: */
1227 #define U filechr_not_unix
1228 #define W filechr_not_windows
1229 #define C filechr_control
1234 /* Table of characters unsafe under various conditions (see above).
1236 Arguably we could also claim `%' to be unsafe, since we use it as
1237 the escape character. If we ever want to be able to reliably
1238 translate file name back to URL, this would become important
1239 crucial. Right now, it's better to be minimal in escaping. */
1241 static const unsigned char filechr_table[256] =
1243 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1244 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1245 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1246 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1247 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1248 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1249 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1250 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1251 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1252 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1253 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1254 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1255 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1256 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1257 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1258 0, 0, 0, 0, 0, 0, 0, 0, /* x y z { | } ~ DEL */
1260 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1261 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1262 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1263 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1265 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1266 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1267 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1268 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1276 /* FN_PORT_SEP is the separator between host and port in file names
1277 for non-standard port numbers. On Unix this is normally ':', as in
1278 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1279 because Windows can't handle ':' in file names. */
1280 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1282 /* FN_QUERY_SEP is the separator between the file name and the URL
1283 query, normally '?'. Since Windows cannot handle '?' as part of
1284 file name, we use '@' instead there. */
1285 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1287 /* Quote path element, characters in [b, e), as file name, and append
1288 the quoted string to DEST. Each character is quoted as per
1289 file_unsafe_char and the corresponding table.
1291 If ESCAPED is true, the path element is considered to be
1292 URL-escaped and will be unescaped prior to inspection. */
1295 append_uri_pathel (const char *b, const char *e, bool escaped,
1296 struct growable *dest)
1302 if (opt.restrict_files_os == restrict_unix)
1303 mask = filechr_not_unix;
1305 mask = filechr_not_windows;
1306 if (opt.restrict_files_ctrl)
1307 mask |= filechr_control;
1309 /* Copy [b, e) to PATHEL and URL-unescape it. */
1313 BOUNDED_TO_ALLOCA (b, e, unescaped);
1314 url_unescape (unescaped);
1316 e = unescaped + strlen (unescaped);
1319 /* Defang ".." when found as component of path. Remember that path
1320 comes from the URL and might contain malicious input. */
1321 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1327 /* Walk the PATHEL string and check how many characters we'll need
1330 for (p = b; p < e; p++)
1331 if (FILE_CHAR_TEST (*p, mask))
1334 /* Calculate the length of the output string. e-b is the input
1335 string length. Each quoted char introduces two additional
1336 characters in the string, hence 2*quoted. */
1337 outlen = (e - b) + (2 * quoted);
1338 GROW (dest, outlen);
1342 /* If there's nothing to quote, we can simply append the string
1343 without processing it again. */
1344 memcpy (TAIL (dest), b, outlen);
1348 char *q = TAIL (dest);
1349 for (p = b; p < e; p++)
1351 if (!FILE_CHAR_TEST (*p, mask))
1355 unsigned char ch = *p;
1357 *q++ = XNUM_TO_DIGIT (ch >> 4);
1358 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1361 assert (q - TAIL (dest) == outlen);
1363 TAIL_INCR (dest, outlen);
1366 /* Append to DEST the directory structure that corresponds the
1367 directory part of URL's path. For example, if the URL is
1368 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1370 Each path element ("dir1" and "dir2" in the above example) is
1371 examined, url-unescaped, and re-escaped as file name element.
1373 Additionally, it cuts as many directories from the path as
1374 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1375 will produce "bar" for the above example. For 2 or more, it will
1378 Each component of the path is quoted for use as file name. */
1381 append_dir_structure (const struct url *u, struct growable *dest)
1383 char *pathel, *next;
1384 int cut = opt.cut_dirs;
1386 /* Go through the path components, de-URL-quote them, and quote them
1387 (if necessary) as file names. */
1390 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1395 /* Ignore empty pathels. */
1399 append_char ('/', dest);
1400 append_uri_pathel (pathel, next, true, dest);
1404 /* Return a unique file name that matches the given URL as good as
1405 possible. Does not create directories on the file system. */
1408 url_file_name (const struct url *u)
1410 struct growable fnres; /* stands for "file name result" */
1412 const char *u_file, *u_query;
1413 char *fname, *unique;
1419 /* Start with the directory prefix, if specified. */
1421 append_string (opt.dir_prefix, &fnres);
1423 /* If "dirstruct" is turned on (typically the case with -r), add
1424 the host and port (unless those have been turned off) and
1425 directory structure. */
1428 if (opt.protocol_directories)
1431 append_char ('/', &fnres);
1432 append_string (supported_schemes[u->scheme].name, &fnres);
1434 if (opt.add_hostdir)
1437 append_char ('/', &fnres);
1438 if (0 != strcmp (u->host, ".."))
1439 append_string (u->host, &fnres);
1441 /* Host name can come from the network; malicious DNS may
1442 allow ".." to be resolved, causing us to write to
1443 "../<file>". Defang such host names. */
1444 append_string ("%2E%2E", &fnres);
1445 if (u->port != scheme_default_port (u->scheme))
1448 number_to_string (portstr, u->port);
1449 append_char (FN_PORT_SEP, &fnres);
1450 append_string (portstr, &fnres);
1454 append_dir_structure (u, &fnres);
1457 /* Add the file name. */
1459 append_char ('/', &fnres);
1460 u_file = *u->file ? u->file : "index.html";
1461 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1463 /* Append "?query" to the file name. */
1464 u_query = u->query && *u->query ? u->query : NULL;
1467 append_char (FN_QUERY_SEP, &fnres);
1468 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1471 /* Zero-terminate the file name. */
1472 append_char ('\0', &fnres);
1476 /* Check the cases in which the unique extensions are not used:
1477 1) Clobbering is turned off (-nc).
1478 2) Retrieval with regetting.
1479 3) Timestamping is used.
1480 4) Hierarchy is built.
1482 The exception is the case when file does exist and is a
1483 directory (see `mkalldirs' for explanation). */
1485 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1486 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1489 unique = unique_name (fname, true);
1490 if (unique != fname)
1495 /* Resolve "." and ".." elements of PATH by destructively modifying
1496 PATH and return true if PATH has been modified, false otherwise.
1498 The algorithm is in spirit similar to the one described in rfc1808,
1499 although implemented differently, in one pass. To recap, path
1500 elements containing only "." are removed, and ".." is taken to mean
1501 "back up one element". Single leading and trailing slashes are
1504 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1505 test examples are provided below. If you change anything in this
1506 function, run test_path_simplify to make sure you haven't broken a
1510 path_simplify (char *path)
1512 char *h = path; /* hare */
1513 char *t = path; /* tortoise */
1514 char *beg = path; /* boundary for backing the tortoise */
1515 char *end = path + strlen (path);
1519 /* Hare should be at the beginning of a path element. */
1521 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1526 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1528 /* Handle "../" by retreating the tortoise by one path
1529 element -- but not past beggining. */
1532 /* Move backwards until T hits the beginning of the
1533 previous path element or the beginning of path. */
1534 for (--t; t > beg && t[-1] != '/'; t--)
1539 /* If we're at the beginning, copy the "../" literally
1540 move the beginning so a later ".." doesn't remove
1550 /* A regular path element. If H hasn't advanced past T,
1551 simply skip to the next path element. Otherwise, copy
1552 the path element until the next slash. */
1555 /* Skip the path element, including the slash. */
1556 while (h < end && *h != '/')
1563 /* Copy the path element, including the final slash. */
1564 while (h < end && *h != '/')
1578 /* Return the length of URL's path. Path is considered to be
1579 terminated by one or more of the ?query or ;params or #fragment,
1580 depending on the scheme. */
1583 path_end (const char *url)
1585 enum url_scheme scheme = url_scheme (url);
1587 if (scheme == SCHEME_INVALID)
1588 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1589 /* +2 to ignore the first two separators ':' and '/' */
1590 seps = init_seps (scheme) + 2;
1591 return strpbrk_or_eos (url, seps);
1594 /* Find the last occurrence of character C in the range [b, e), or
1595 NULL, if none are present. We might want to use memrchr (a GNU
1596 extension) under GNU libc. */
1599 find_last_char (const char *b, const char *e, char c)
1607 /* Merge BASE with LINK and return the resulting URI.
1609 Either of the URIs may be absolute or relative, complete with the
1610 host name, or path only. This tries to reasonably handle all
1611 foreseeable cases. It only employs minimal URL parsing, without
1612 knowledge of the specifics of schemes.
1614 I briefly considered making this function call path_simplify after
1615 the merging process, as rfc1738 seems to suggest. This is a bad
1616 idea for several reasons: 1) it complexifies the code, and 2)
1617 url_parse has to simplify path anyway, so it's wasteful to boot. */
1620 uri_merge (const char *base, const char *link)
1626 if (url_has_scheme (link))
1627 return xstrdup (link);
1629 /* We may not examine BASE past END. */
1630 end = path_end (base);
1631 linklength = strlen (link);
1635 /* Empty LINK points back to BASE, query string and all. */
1636 return xstrdup (base);
1638 else if (*link == '?')
1640 /* LINK points to the same location, but changes the query
1641 string. Examples: */
1642 /* uri_merge("path", "?new") -> "path?new" */
1643 /* uri_merge("path?foo", "?new") -> "path?new" */
1644 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1645 /* uri_merge("path#foo", "?new") -> "path?new" */
1646 int baselength = end - base;
1647 merge = xmalloc (baselength + linklength + 1);
1648 memcpy (merge, base, baselength);
1649 memcpy (merge + baselength, link, linklength);
1650 merge[baselength + linklength] = '\0';
1652 else if (*link == '#')
1654 /* uri_merge("path", "#new") -> "path#new" */
1655 /* uri_merge("path#foo", "#new") -> "path#new" */
1656 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1657 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1659 const char *end1 = strchr (base, '#');
1661 end1 = base + strlen (base);
1662 baselength = end1 - base;
1663 merge = xmalloc (baselength + linklength + 1);
1664 memcpy (merge, base, baselength);
1665 memcpy (merge + baselength, link, linklength);
1666 merge[baselength + linklength] = '\0';
1668 else if (*link == '/' && *(link + 1) == '/')
1670 /* LINK begins with "//" and so is a net path: we need to
1671 replace everything after (and including) the double slash
1674 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1675 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1676 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1680 const char *start_insert;
1682 /* Look for first slash. */
1683 slash = memchr (base, '/', end - base);
1684 /* If found slash and it is a double slash, then replace
1685 from this point, else default to replacing from the
1687 if (slash && *(slash + 1) == '/')
1688 start_insert = slash;
1690 start_insert = base;
1692 span = start_insert - base;
1693 merge = xmalloc (span + linklength + 1);
1695 memcpy (merge, base, span);
1696 memcpy (merge + span, link, linklength);
1697 merge[span + linklength] = '\0';
1699 else if (*link == '/')
1701 /* LINK is an absolute path: we need to replace everything
1702 after (and including) the FIRST slash with LINK.
1704 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1705 "/qux/xyzzy", our result should be
1706 "http://host/qux/xyzzy". */
1709 const char *start_insert = NULL; /* for gcc to shut up. */
1710 const char *pos = base;
1711 bool seen_slash_slash = false;
1712 /* We're looking for the first slash, but want to ignore
1715 slash = memchr (pos, '/', end - pos);
1716 if (slash && !seen_slash_slash)
1717 if (*(slash + 1) == '/')
1720 seen_slash_slash = true;
1724 /* At this point, SLASH is the location of the first / after
1725 "//", or the first slash altogether. START_INSERT is the
1726 pointer to the location where LINK will be inserted. When
1727 examining the last two examples, keep in mind that LINK
1730 if (!slash && !seen_slash_slash)
1731 /* example: "foo" */
1733 start_insert = base;
1734 else if (!slash && seen_slash_slash)
1735 /* example: "http://foo" */
1738 else if (slash && !seen_slash_slash)
1739 /* example: "foo/bar" */
1741 start_insert = base;
1742 else if (slash && seen_slash_slash)
1743 /* example: "http://something/" */
1745 start_insert = slash;
1747 span = start_insert - base;
1748 merge = xmalloc (span + linklength + 1);
1750 memcpy (merge, base, span);
1751 memcpy (merge + span, link, linklength);
1752 merge[span + linklength] = '\0';
1756 /* LINK is a relative URL: we need to replace everything
1757 after last slash (possibly empty) with LINK.
1759 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1760 our result should be "whatever/foo/qux/xyzzy". */
1761 bool need_explicit_slash = false;
1763 const char *start_insert;
1764 const char *last_slash = find_last_char (base, end, '/');
1767 /* No slash found at all. Replace what we have with LINK. */
1768 start_insert = base;
1770 else if (last_slash && last_slash >= base + 2
1771 && last_slash[-2] == ':' && last_slash[-1] == '/')
1773 /* example: http://host" */
1775 start_insert = end + 1;
1776 need_explicit_slash = true;
1780 /* example: "whatever/foo/bar" */
1782 start_insert = last_slash + 1;
1785 span = start_insert - base;
1786 merge = xmalloc (span + linklength + 1);
1788 memcpy (merge, base, span);
1789 if (need_explicit_slash)
1790 merge[span - 1] = '/';
1791 memcpy (merge + span, link, linklength);
1792 merge[span + linklength] = '\0';
1798 #define APPEND(p, s) do { \
1799 int len = strlen (s); \
1800 memcpy (p, s, len); \
1804 /* Use this instead of password when the actual password is supposed
1805 to be hidden. We intentionally use a generic string without giving
1806 away the number of characters in the password, like previous
1808 #define HIDDEN_PASSWORD "*password*"
1810 /* Recreate the URL string from the data in URL.
1812 If HIDE is true (as it is when we're calling this on a URL we plan
1813 to print, but not when calling it to canonicalize a URL for use
1814 within the program), password will be hidden. Unsafe characters in
1815 the URL will be quoted. */
1818 url_string (const struct url *url, bool hide_password)
1822 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1824 int scheme_port = supported_schemes[url->scheme].default_port;
1825 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1826 int fplen = full_path_length (url);
1828 bool brackets_around_host;
1830 assert (scheme_str != NULL);
1832 /* Make sure the user name and password are quoted. */
1835 quoted_user = url_escape_allow_passthrough (url->user);
1839 quoted_passwd = HIDDEN_PASSWORD;
1841 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1845 /* In the unlikely event that the host name contains non-printable
1846 characters, quote it for displaying to the user. */
1847 quoted_host = url_escape_allow_passthrough (url->host);
1849 /* Undo the quoting of colons that URL escaping performs. IPv6
1850 addresses may legally contain colons, and in that case must be
1851 placed in square brackets. */
1852 if (quoted_host != url->host)
1853 unescape_single_char (quoted_host, ':');
1854 brackets_around_host = strchr (quoted_host, ':') != NULL;
1856 size = (strlen (scheme_str)
1857 + strlen (quoted_host)
1858 + (brackets_around_host ? 2 : 0)
1861 if (url->port != scheme_port)
1862 size += 1 + numdigit (url->port);
1865 size += 1 + strlen (quoted_user);
1867 size += 1 + strlen (quoted_passwd);
1870 p = result = xmalloc (size);
1872 APPEND (p, scheme_str);
1875 APPEND (p, quoted_user);
1879 APPEND (p, quoted_passwd);
1884 if (brackets_around_host)
1886 APPEND (p, quoted_host);
1887 if (brackets_around_host)
1889 if (url->port != scheme_port)
1892 p = number_to_string (p, url->port);
1895 full_path_write (url, p);
1899 assert (p - result == size);
1901 if (quoted_user && quoted_user != url->user)
1902 xfree (quoted_user);
1903 if (quoted_passwd && !hide_password && quoted_passwd != url->passwd)
1904 xfree (quoted_passwd);
1905 if (quoted_host != url->host)
1906 xfree (quoted_host);
1911 /* Return true if scheme a is similar to scheme b.
1913 Schemes are similar if they are equal. If SSL is supported, schemes
1914 are also similar if one is http (SCHEME_HTTP) and the other is https
1917 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1922 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1923 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1930 /* Debugging and testing support for path_simplify. */
1932 /* Debug: run path_simplify on PATH and return the result in a new
1933 string. Useful for calling from the debugger. */
1937 char *copy = xstrdup (path);
1938 path_simplify (copy);
1943 run_test (char *test, char *expected_result, bool expected_change)
1945 char *test_copy = xstrdup (test);
1946 bool modified = path_simplify (test_copy);
1948 if (0 != strcmp (test_copy, expected_result))
1950 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
1951 test, expected_result, test_copy);
1953 if (modified != expected_change)
1955 if (expected_change)
1956 printf ("Expected modification with path_simplify(\"%s\").\n",
1959 printf ("Expected no modification with path_simplify(\"%s\").\n",
1966 test_path_simplify (void)
1969 char *test, *result;
1975 { "..", "..", false },
1976 { "../", "../", false },
1977 { "foo", "foo", false },
1978 { "foo/bar", "foo/bar", false },
1979 { "foo///bar", "foo///bar", false },
1980 { "foo/.", "foo/", true },
1981 { "foo/./", "foo/", true },
1982 { "foo./", "foo./", false },
1983 { "foo/../bar", "bar", true },
1984 { "foo/../bar/", "bar/", true },
1985 { "foo/bar/..", "foo/", true },
1986 { "foo/bar/../x", "foo/x", true },
1987 { "foo/bar/../x/", "foo/x/", true },
1988 { "foo/..", "", true },
1989 { "foo/../..", "..", true },
1990 { "foo/../../..", "../..", true },
1991 { "foo/../../bar/../../baz", "../../baz", true },
1992 { "a/b/../../c", "c", true },
1993 { "./a/../b", "b", true }
1997 for (i = 0; i < countof (tests); i++)
1999 char *test = tests[i].test;
2000 char *expected_result = tests[i].result;
2001 bool expected_change = tests[i].should_modify;
2002 run_test (test, expected_result, expected_change);