2 Copyright (C) 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
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, 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 */
49 const char *leading_string;
54 /* Supported schemes: */
55 static struct scheme_data supported_schemes[] =
57 { "http", "http://", DEFAULT_HTTP_PORT, 1 },
59 { "https", "https://", DEFAULT_HTTPS_PORT, 1 },
61 { "ftp", "ftp://", DEFAULT_FTP_PORT, 1 },
67 /* Forward declarations: */
69 static bool path_simplify (char *);
71 /* Support for escaping and unescaping of URL strings. */
73 /* Table of "reserved" and "unsafe" characters. Those terms are
74 rfc1738-speak, as such largely obsoleted by rfc2396 and later
75 specs, but the general idea remains.
77 A reserved character is the one that you can't decode without
78 changing the meaning of the URL. For example, you can't decode
79 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
80 path components is different. Non-reserved characters can be
81 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
82 unsafe characters are loosely based on rfc1738, plus "$" and ",",
83 as recommended by rfc2396, and minus "~", which is very frequently
84 used (and sometimes unrecognized as %7E by broken servers).
86 An unsafe character is the one that should be encoded when URLs are
87 placed in foreign environments. E.g. space and newline are unsafe
88 in HTTP contexts because HTTP uses them as separator and line
89 terminator, so they must be encoded to %20 and %0A respectively.
90 "*" is unsafe in shell context, etc.
92 We determine whether a character is unsafe through static table
93 lookup. This code assumes ASCII character set and 8-bit chars. */
96 /* rfc1738 reserved chars + "$" and ",". */
99 /* rfc1738 unsafe chars, plus non-printables. */
103 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
104 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
105 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
107 /* Shorthands for the table: */
108 #define R urlchr_reserved
109 #define U urlchr_unsafe
112 static const unsigned char urlchr_table[256] =
114 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
115 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
116 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
117 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
118 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
119 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
120 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
121 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
122 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
123 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
124 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
125 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
126 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
127 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
128 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
129 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
131 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
132 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
133 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
134 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
136 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
137 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
138 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
139 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
145 /* URL-unescape the string S.
147 This is done by transforming the sequences "%HH" to the character
148 represented by the hexadecimal digits HH. If % is not followed by
149 two hexadecimal digits, it is inserted literally.
151 The transformation is done in place. If you need the original
152 string intact, make a copy before calling this function. */
155 url_unescape (char *s)
157 char *t = s; /* t - tortoise */
158 char *h = s; /* h - hare */
170 /* Do nothing if '%' is not followed by two hex digits. */
171 if (!h[1] || !h[2] || !(ISXDIGIT (h[1]) && ISXDIGIT (h[2])))
173 c = X2DIGITS_TO_NUM (h[1], h[2]);
174 /* Don't unescape %00 because there is no way to insert it
175 into a C string without effectively truncating it. */
185 /* The core of url_escape_* functions. Escapes the characters that
186 match the provided mask in urlchr_table.
188 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
189 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
190 allocated string will be returned in all cases. */
193 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
200 for (p1 = s; *p1; p1++)
201 if (urlchr_test (*p1, mask))
202 addition += 2; /* Two more characters (hex digits) */
205 return allow_passthrough ? (char *)s : xstrdup (s);
207 newlen = (p1 - s) + addition;
208 newstr = xmalloc (newlen + 1);
214 /* Quote the characters that match the test mask. */
215 if (urlchr_test (*p1, mask))
217 unsigned char c = *p1++;
219 *p2++ = XNUM_TO_DIGIT (c >> 4);
220 *p2++ = XNUM_TO_DIGIT (c & 0xf);
225 assert (p2 - newstr == newlen);
231 /* URL-escape the unsafe characters (see urlchr_table) in a given
232 string, returning a freshly allocated string. */
235 url_escape (const char *s)
237 return url_escape_1 (s, urlchr_unsafe, false);
240 /* URL-escape the unsafe characters (see urlchr_table) in a given
241 string. If no characters are unsafe, S is returned. */
244 url_escape_allow_passthrough (const char *s)
246 return url_escape_1 (s, urlchr_unsafe, true);
249 /* Decide whether the char at position P needs to be encoded. (It is
250 not enough to pass a single char *P because the function may need
251 to inspect the surrounding context.)
253 Return true if the char should be escaped as %XX, false otherwise. */
256 char_needs_escaping (const char *p)
260 if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2)))
263 /* Garbled %.. sequence: encode `%'. */
266 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
272 /* Translate a %-escaped (but possibly non-conformant) input string S
273 into a %-escaped (and conformant) output string. If no characters
274 are encoded or decoded, return the same string S; otherwise, return
275 a freshly allocated string with the new contents.
277 After a URL has been run through this function, the protocols that
278 use `%' as the quote character can use the resulting string as-is,
279 while those that don't can use url_unescape to get to the intended
280 data. This function is stable: once the input is transformed,
281 further transformations of the result yield the same output.
283 Let's discuss why this function is needed.
285 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
286 a raw space character would mess up the HTTP request, it needs to
287 be quoted, like this:
289 GET /abc%20def HTTP/1.0
291 It would appear that the unsafe chars need to be quoted, for
292 example with url_escape. But what if we're requested to download
293 `abc%20def'? url_escape transforms "%" to "%25", which would leave
294 us with `abc%2520def'. This is incorrect -- since %-escapes are
295 part of URL syntax, "%20" is the correct way to denote a literal
296 space on the Wget command line. This leads to the conclusion that
297 in that case Wget should not call url_escape, but leave the `%20'
298 as is. This is clearly contradictory, but it only gets worse.
300 What if the requested URI is `abc%20 def'? If we call url_escape,
301 we end up with `/abc%2520%20def', which is almost certainly not
302 intended. If we don't call url_escape, we are left with the
303 embedded space and cannot complete the request. What the user
304 meant was for Wget to request `/abc%20%20def', and this is where
305 reencode_escapes kicks in.
307 Wget used to solve this by first decoding %-quotes, and then
308 encoding all the "unsafe" characters found in the resulting string.
309 This was wrong because it didn't preserve certain URL special
310 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
311 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
312 whether we considered `+' reserved (it is). One of these results
313 is inevitable because by the second step we would lose information
314 on whether the `+' was originally encoded or not. Both results
315 were wrong because in CGI parameters + means space, while %2B means
316 literal plus. reencode_escapes correctly translates the above to
317 "a%2B+b", i.e. returns the original string.
319 This function uses a modified version of the algorithm originally
320 proposed by Anon Sricharoenchai:
322 * Encode all "unsafe" characters, except those that are also
323 "reserved", to %XX. See urlchr_table for which characters are
326 * Encode the "%" characters not followed by two hex digits to
329 * Pass through all other characters and %XX escapes as-is. (Up to
330 Wget 1.10 this decoded %XX escapes corresponding to "safe"
331 characters, but that was obtrusive and broke some servers.)
335 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
337 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
341 "foo bar" -> "foo%20bar"
342 "foo%20bar" -> "foo%20bar"
343 "foo %20bar" -> "foo%20%20bar"
344 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
345 "foo%25%20bar" -> "foo%25%20bar"
346 "foo%2%20bar" -> "foo%252%20bar"
347 "foo+bar" -> "foo+bar" (plus is reserved!)
348 "foo%2b+bar" -> "foo%2b+bar" */
351 reencode_escapes (const char *s)
357 int encode_count = 0;
359 /* First pass: inspect the string to see if there's anything to do,
360 and to calculate the new length. */
361 for (p1 = s; *p1; p1++)
362 if (char_needs_escaping (p1))
366 /* The string is good as it is. */
367 return (char *) s; /* C const model sucks. */
370 /* Each encoding adds two characters (hex digits). */
371 newlen = oldlen + 2 * encode_count;
372 newstr = xmalloc (newlen + 1);
374 /* Second pass: copy the string to the destination address, encoding
375 chars when needed. */
380 if (char_needs_escaping (p1))
382 unsigned char c = *p1++;
384 *p2++ = XNUM_TO_DIGIT (c >> 4);
385 *p2++ = XNUM_TO_DIGIT (c & 0xf);
391 assert (p2 - newstr == newlen);
395 /* Returns the scheme type if the scheme is supported, or
396 SCHEME_INVALID if not. */
399 url_scheme (const char *url)
403 for (i = 0; supported_schemes[i].leading_string; i++)
404 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
405 strlen (supported_schemes[i].leading_string)))
407 if (supported_schemes[i].enabled)
408 return (enum url_scheme) i;
410 return SCHEME_INVALID;
413 return SCHEME_INVALID;
416 #define SCHEME_CHAR(ch) (ISALNUM (ch) || (ch) == '-' || (ch) == '+')
418 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
419 currently implemented, it returns true if URL begins with
423 url_has_scheme (const char *url)
427 /* The first char must be a scheme char. */
428 if (!*p || !SCHEME_CHAR (*p))
431 /* Followed by 0 or more scheme chars. */
432 while (*p && SCHEME_CHAR (*p))
434 /* Terminated by ':'. */
439 scheme_default_port (enum url_scheme scheme)
441 return supported_schemes[scheme].default_port;
445 scheme_disable (enum url_scheme scheme)
447 supported_schemes[scheme].enabled = false;
450 /* Skip the username and password, if present in the URL. The
451 function should *not* be called with the complete URL, but with the
452 portion after the scheme.
454 If no username and password are found, return URL. */
457 url_skip_credentials (const char *url)
459 /* Look for '@' that comes before terminators, such as '/', '?',
461 const char *p = (const char *)strpbrk (url, "@/?#;");
467 /* Parse credentials contained in [BEG, END). The region is expected
468 to have come from a URL and is unescaped. */
471 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
477 return false; /* empty user name */
479 colon = memchr (beg, ':', end - beg);
481 return false; /* again empty user name */
485 *passwd = strdupdelim (colon + 1, end);
487 url_unescape (*passwd);
494 *user = strdupdelim (beg, userend);
495 url_unescape (*user);
499 /* Used by main.c: detect URLs written using the "shorthand" URL forms
500 popularized by Netscape and NcFTP. HTTP shorthands look like this:
502 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
503 www.foo.com[:port] -> http://www.foo.com[:port]
505 FTP shorthands look like this:
507 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
508 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
510 If the URL needs not or cannot be rewritten, return NULL. */
513 rewrite_shorthand_url (const char *url)
517 if (url_scheme (url) != SCHEME_INVALID)
520 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
522 for (p = url; *p && *p != ':' && *p != '/'; p++)
528 /* If we're looking at "://", it means the URL uses a scheme we
529 don't support, which may include "https" when compiled without
530 SSL support. Don't bogusly rewrite such URLs. */
531 if (p[0] == ':' && p[1] == '/' && p[2] == '/')
538 /* If the characters after the colon and before the next slash
539 or end of string are all digits, it's HTTP. */
541 for (pp = p + 1; ISDIGIT (*pp); pp++)
543 if (digits > 0 && (*pp == '/' || *pp == '\0'))
546 /* Prepend "ftp://" to the entire URL... */
547 res = xmalloc (6 + strlen (url) + 1);
548 sprintf (res, "ftp://%s", url);
549 /* ...and replace ':' with '/'. */
550 res[6 + (p - url)] = '/';
557 /* Just prepend "http://" to what we have. */
558 res = xmalloc (7 + strlen (url) + 1);
559 sprintf (res, "http://%s", url);
564 static void split_path (const char *, char **, char **);
566 /* Like strpbrk, with the exception that it returns the pointer to the
567 terminating zero (end-of-string aka "eos") if no matching character
570 Although I normally balk at Gcc-specific optimizations, it probably
571 makes sense here: glibc has optimizations that detect strpbrk being
572 called with literal string as ACCEPT and inline the search. That
573 optimization is defeated if strpbrk is hidden within the call to
574 another function. (And no, making strpbrk_or_eos inline doesn't
575 help because the check for literal accept is in the
578 #if defined(__GNUC__) && __GNUC__ >= 3
580 #define strpbrk_or_eos(s, accept) ({ \
581 char *SOE_p = strpbrk (s, accept); \
583 SOE_p = strchr (s, '\0'); \
587 #else /* not __GNUC__ or old gcc */
590 strpbrk_or_eos (const char *s, const char *accept)
592 char *p = strpbrk (s, accept);
594 p = strchr (s, '\0');
597 #endif /* not __GNUC__ or old gcc */
599 /* Turn STR into lowercase; return true if a character was actually
603 lowercase_str (char *str)
605 bool changed = false;
610 *str = TOLOWER (*str);
615 static const char *parse_errors[] = {
616 #define PE_NO_ERROR 0
618 #define PE_UNSUPPORTED_SCHEME 1
619 N_("Unsupported scheme"),
620 #define PE_EMPTY_HOST 2
622 #define PE_BAD_PORT_NUMBER 3
623 N_("Bad port number"),
624 #define PE_INVALID_USER_NAME 4
625 N_("Invalid user name"),
626 #define PE_UNTERMINATED_IPV6_ADDRESS 5
627 N_("Unterminated IPv6 numeric address"),
628 #define PE_IPV6_NOT_SUPPORTED 6
629 N_("IPv6 addresses not supported"),
630 #define PE_INVALID_IPV6_ADDRESS 7
631 N_("Invalid IPv6 numeric address")
636 Return a new struct url if successful, NULL on error. In case of
637 error, and if ERROR is not NULL, also set *ERROR to the appropriate
640 url_parse (const char *url, int *error)
644 bool path_modified, host_modified;
646 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 params_b = params_e = NULL;
686 query_b = query_e = NULL;
687 fragment_b = fragment_e = NULL;
693 /* Handle IPv6 address inside square brackets. Ideally we'd
694 just look for the terminating ']', but rfc2732 mandates
695 rejecting invalid IPv6 addresses. */
697 /* The address begins after '['. */
699 host_e = strchr (host_b, ']');
703 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
708 /* Check if the IPv6 address is valid. */
709 if (!is_valid_ipv6_address(host_b, host_e))
711 error_code = PE_INVALID_IPV6_ADDRESS;
715 /* Continue parsing after the closing ']'. */
718 error_code = PE_IPV6_NOT_SUPPORTED;
724 p = strpbrk_or_eos (p, ":/;?#");
728 if (host_b == host_e)
730 error_code = PE_EMPTY_HOST;
734 port = scheme_default_port (scheme);
737 const char *port_b, *port_e, *pp;
739 /* scheme://host:port/tralala */
743 p = strpbrk_or_eos (p, "/;?#");
746 /* Allow empty port, as per rfc2396. */
747 if (port_b != port_e)
749 for (port = 0, pp = port_b; pp < port_e; pp++)
753 /* http://host:12randomgarbage/blah */
755 error_code = PE_BAD_PORT_NUMBER;
758 port = 10 * port + (*pp - '0');
759 /* Check for too large port numbers here, before we have
760 a chance to overflow on bogus port values. */
763 error_code = PE_BAD_PORT_NUMBER;
774 p = strpbrk_or_eos (p, ";?#");
779 /* Path is not allowed not to exist. */
787 p = strpbrk_or_eos (p, "?#");
794 p = strpbrk_or_eos (p, "#");
797 /* Hack that allows users to use '?' (a wildcard character) in
798 FTP URLs without it being interpreted as a query string
800 if (scheme == SCHEME_FTP)
802 query_b = query_e = NULL;
815 if (uname_b != uname_e)
817 /* http://user:pass@host */
819 /* uname_b uname_e */
820 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
822 error_code = PE_INVALID_USER_NAME;
827 u = xnew0 (struct url);
829 u->host = strdupdelim (host_b, host_e);
834 u->path = strdupdelim (path_b, path_e);
835 path_modified = path_simplify (u->path);
836 split_path (u->path, &u->dir, &u->file);
838 host_modified = lowercase_str (u->host);
840 /* Decode %HH sequences in host name. This is important not so much
841 to support %HH sequences in host names (which other browser
842 don't), but to support binary characters (which will have been
843 converted to %HH by reencode_escapes). */
844 if (strchr (u->host, '%'))
846 url_unescape (u->host);
847 host_modified = true;
851 u->params = strdupdelim (params_b, params_e);
853 u->query = strdupdelim (query_b, query_e);
855 u->fragment = strdupdelim (fragment_b, fragment_e);
857 if (path_modified || u->fragment || host_modified || path_b == path_e)
859 /* If we suspect that a transformation has rendered what
860 url_string might return different from URL_ENCODED, rebuild
861 u->url using url_string. */
862 u->url = url_string (u, false);
864 if (url_encoded != url)
865 xfree ((char *) url_encoded);
869 if (url_encoded == url)
870 u->url = xstrdup (url);
872 u->url = url_encoded;
878 /* Cleanup in case of error: */
879 if (url_encoded && url_encoded != url)
882 /* Transmit the error code to the caller, if the caller wants to
889 /* Return the error message string from ERROR_CODE, which should have
890 been retrieved from url_parse. The error message is translated. */
893 url_error (int error_code)
895 assert (error_code >= 0 && error_code < countof (parse_errors));
896 return _(parse_errors[error_code]);
899 /* Split PATH into DIR and FILE. PATH comes from the URL and is
900 expected to be URL-escaped.
902 The path is split into directory (the part up to the last slash)
903 and file (the part after the last slash), which are subsequently
907 "foo/bar/baz" "foo/bar" "baz"
908 "foo/bar/" "foo/bar" ""
910 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
912 DIR and FILE are freshly allocated. */
915 split_path (const char *path, char **dir, char **file)
917 char *last_slash = strrchr (path, '/');
921 *file = xstrdup (path);
925 *dir = strdupdelim (path, last_slash);
926 *file = xstrdup (last_slash + 1);
929 url_unescape (*file);
932 /* Note: URL's "full path" is the path with the query string and
933 params appended. The "fragment" (#foo) is intentionally ignored,
934 but that might be changed. For example, if the original URL was
935 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
936 the full path will be "/foo/bar/baz;bullshit?querystring". */
938 /* Return the length of the full path, without the terminating
942 full_path_length (const struct url *url)
946 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
957 /* Write out the full path. */
960 full_path_write (const struct url *url, char *where)
962 #define FROB(el, chr) do { \
963 char *f_el = url->el; \
965 int l = strlen (f_el); \
967 memcpy (where, f_el, l); \
979 /* Public function for getting the "full path". E.g. if u->path is
980 "foo/bar" and u->query is "param=value", full_path will be
981 "/foo/bar?param=value". */
984 url_full_path (const struct url *url)
986 int length = full_path_length (url);
987 char *full_path = xmalloc (length + 1);
989 full_path_write (url, full_path);
990 full_path[length] = '\0';
995 /* Unescape CHR in an otherwise escaped STR. Used to selectively
996 escaping of certain characters, such as "/" and ":". Returns a
997 count of unescaped chars. */
1000 unescape_single_char (char *str, char chr)
1002 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1003 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1004 char *h = str; /* hare */
1005 char *t = str; /* tortoise */
1006 for (; *h; h++, t++)
1008 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1019 /* Escape unsafe and reserved characters, except for the slash
1023 url_escape_dir (const char *dir)
1025 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1029 unescape_single_char (newdir, '/');
1033 /* Sync u->path and u->url with u->dir and u->file. Called after
1034 u->file or u->dir have been changed, typically by the FTP code. */
1037 sync_path (struct url *u)
1039 char *newpath, *efile, *edir;
1043 /* u->dir and u->file are not escaped. URL-escape them before
1044 reassembling them into u->path. That way, if they contain
1045 separators like '?' or even if u->file contains slashes, the
1046 path will be correctly assembled. (u->file can contain slashes
1047 if the URL specifies it with %2f, or if an FTP server returns
1049 edir = url_escape_dir (u->dir);
1050 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1053 newpath = xstrdup (efile);
1056 int dirlen = strlen (edir);
1057 int filelen = strlen (efile);
1059 /* Copy "DIR/FILE" to newpath. */
1060 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1061 memcpy (p, edir, dirlen);
1064 memcpy (p, efile, filelen);
1073 if (efile != u->file)
1076 /* Regenerate u->url as well. */
1078 u->url = url_string (u, false);
1081 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1082 This way we can sync u->path and u->url when they get changed. */
1085 url_set_dir (struct url *url, const char *newdir)
1088 url->dir = xstrdup (newdir);
1093 url_set_file (struct url *url, const char *newfile)
1096 url->file = xstrdup (newfile);
1101 url_free (struct url *url)
1107 xfree_null (url->params);
1108 xfree_null (url->query);
1109 xfree_null (url->fragment);
1110 xfree_null (url->user);
1111 xfree_null (url->passwd);
1119 /* Create all the necessary directories for PATH (a file). Calls
1120 make_directory internally. */
1122 mkalldirs (const char *path)
1129 p = path + strlen (path);
1130 for (; *p != '/' && p != path; p--)
1133 /* Don't create if it's just a file. */
1134 if ((p == path) && (*p != '/'))
1136 t = strdupdelim (path, p);
1138 /* Check whether the directory exists. */
1139 if ((stat (t, &st) == 0))
1141 if (S_ISDIR (st.st_mode))
1148 /* If the dir exists as a file name, remove it first. This
1149 is *only* for Wget to work with buggy old CERN http
1150 servers. Here is the scenario: When Wget tries to
1151 retrieve a directory without a slash, e.g.
1152 http://foo/bar (bar being a directory), CERN server will
1153 not redirect it too http://foo/bar/ -- it will generate a
1154 directory listing containing links to bar/file1,
1155 bar/file2, etc. Wget will lose because it saves this
1156 HTML listing to a file `bar', so it cannot create the
1157 directory. To work around this, if the file of the same
1158 name exists, we just remove it and create the directory
1160 DEBUGP (("Removing %s because of directory danger!\n", t));
1164 res = make_directory (t);
1166 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1171 /* Functions for constructing the file name out of URL components. */
1173 /* A growable string structure, used by url_file_name and friends.
1174 This should perhaps be moved to utils.c.
1176 The idea is to have a convenient and efficient way to construct a
1177 string by having various functions append data to it. Instead of
1178 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1179 functions in questions, we pass the pointer to this struct. */
1187 /* Ensure that the string can accept APPEND_COUNT more characters past
1188 the current TAIL position. If necessary, this will grow the string
1189 and update its allocated size. If the string is already large
1190 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1191 #define GROW(g, append_size) do { \
1192 struct growable *G_ = g; \
1193 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1196 /* Return the tail position of the string. */
1197 #define TAIL(r) ((r)->base + (r)->tail)
1199 /* Move the tail position by APPEND_COUNT characters. */
1200 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1202 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1206 append_string (const char *str, struct growable *dest)
1208 int l = strlen (str);
1210 memcpy (TAIL (dest), str, l);
1211 TAIL_INCR (dest, l);
1214 /* Append CH to DEST. For example, append_char (0, DEST)
1215 zero-terminates DEST. */
1218 append_char (char ch, struct growable *dest)
1222 TAIL_INCR (dest, 1);
1226 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1227 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1228 filechr_control = 4 /* a control character, e.g. 0-31 */
1231 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1233 /* Shorthands for the table: */
1234 #define U filechr_not_unix
1235 #define W filechr_not_windows
1236 #define C filechr_control
1241 /* Table of characters unsafe under various conditions (see above).
1243 Arguably we could also claim `%' to be unsafe, since we use it as
1244 the escape character. If we ever want to be able to reliably
1245 translate file name back to URL, this would become important
1246 crucial. Right now, it's better to be minimal in escaping. */
1248 static const unsigned char filechr_table[256] =
1250 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1251 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1252 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1253 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1254 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1255 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1256 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1257 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1258 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1259 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1260 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1261 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1262 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1263 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1264 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1265 0, 0, 0, 0, 0, 0, 0, 0, /* x y z { | } ~ DEL */
1267 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1268 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1269 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1270 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1272 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1273 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1274 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1275 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1283 /* FN_PORT_SEP is the separator between host and port in file names
1284 for non-standard port numbers. On Unix this is normally ':', as in
1285 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1286 because Windows can't handle ':' in file names. */
1287 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1289 /* FN_QUERY_SEP is the separator between the file name and the URL
1290 query, normally '?'. Since Windows cannot handle '?' as part of
1291 file name, we use '@' instead there. */
1292 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1294 /* Quote path element, characters in [b, e), as file name, and append
1295 the quoted string to DEST. Each character is quoted as per
1296 file_unsafe_char and the corresponding table.
1298 If ESCAPED is true, the path element is considered to be
1299 URL-escaped and will be unescaped prior to inspection. */
1302 append_uri_pathel (const char *b, const char *e, bool escaped,
1303 struct growable *dest)
1309 if (opt.restrict_files_os == restrict_unix)
1310 mask = filechr_not_unix;
1312 mask = filechr_not_windows;
1313 if (opt.restrict_files_ctrl)
1314 mask |= filechr_control;
1316 /* Copy [b, e) to PATHEL and URL-unescape it. */
1320 BOUNDED_TO_ALLOCA (b, e, unescaped);
1321 url_unescape (unescaped);
1323 e = unescaped + strlen (unescaped);
1326 /* Defang ".." when found as component of path. Remember that path
1327 comes from the URL and might contain malicious input. */
1328 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1334 /* Walk the PATHEL string and check how many characters we'll need
1337 for (p = b; p < e; p++)
1338 if (FILE_CHAR_TEST (*p, mask))
1341 /* Calculate the length of the output string. e-b is the input
1342 string length. Each quoted char introduces two additional
1343 characters in the string, hence 2*quoted. */
1344 outlen = (e - b) + (2 * quoted);
1345 GROW (dest, outlen);
1349 /* If there's nothing to quote, we can simply append the string
1350 without processing it again. */
1351 memcpy (TAIL (dest), b, outlen);
1355 char *q = TAIL (dest);
1356 for (p = b; p < e; p++)
1358 if (!FILE_CHAR_TEST (*p, mask))
1362 unsigned char ch = *p;
1364 *q++ = XNUM_TO_DIGIT (ch >> 4);
1365 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1368 assert (q - TAIL (dest) == outlen);
1370 TAIL_INCR (dest, outlen);
1373 /* Append to DEST the directory structure that corresponds the
1374 directory part of URL's path. For example, if the URL is
1375 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1377 Each path element ("dir1" and "dir2" in the above example) is
1378 examined, url-unescaped, and re-escaped as file name element.
1380 Additionally, it cuts as many directories from the path as
1381 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1382 will produce "bar" for the above example. For 2 or more, it will
1385 Each component of the path is quoted for use as file name. */
1388 append_dir_structure (const struct url *u, struct growable *dest)
1390 char *pathel, *next;
1391 int cut = opt.cut_dirs;
1393 /* Go through the path components, de-URL-quote them, and quote them
1394 (if necessary) as file names. */
1397 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1402 /* Ignore empty pathels. */
1406 append_char ('/', dest);
1407 append_uri_pathel (pathel, next, true, dest);
1411 /* Return a unique file name that matches the given URL as good as
1412 possible. Does not create directories on the file system. */
1415 url_file_name (const struct url *u)
1417 struct growable fnres; /* stands for "file name result" */
1419 const char *u_file, *u_query;
1420 char *fname, *unique;
1426 /* Start with the directory prefix, if specified. */
1428 append_string (opt.dir_prefix, &fnres);
1430 /* If "dirstruct" is turned on (typically the case with -r), add
1431 the host and port (unless those have been turned off) and
1432 directory structure. */
1435 if (opt.protocol_directories)
1438 append_char ('/', &fnres);
1439 append_string (supported_schemes[u->scheme].name, &fnres);
1441 if (opt.add_hostdir)
1444 append_char ('/', &fnres);
1445 if (0 != strcmp (u->host, ".."))
1446 append_string (u->host, &fnres);
1448 /* Host name can come from the network; malicious DNS may
1449 allow ".." to be resolved, causing us to write to
1450 "../<file>". Defang such host names. */
1451 append_string ("%2E%2E", &fnres);
1452 if (u->port != scheme_default_port (u->scheme))
1455 number_to_string (portstr, u->port);
1456 append_char (FN_PORT_SEP, &fnres);
1457 append_string (portstr, &fnres);
1461 append_dir_structure (u, &fnres);
1464 /* Add the file name. */
1466 append_char ('/', &fnres);
1467 u_file = *u->file ? u->file : "index.html";
1468 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1470 /* Append "?query" to the file name. */
1471 u_query = u->query && *u->query ? u->query : NULL;
1474 append_char (FN_QUERY_SEP, &fnres);
1475 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1478 /* Zero-terminate the file name. */
1479 append_char ('\0', &fnres);
1483 /* Check the cases in which the unique extensions are not used:
1484 1) Clobbering is turned off (-nc).
1485 2) Retrieval with regetting.
1486 3) Timestamping is used.
1487 4) Hierarchy is built.
1489 The exception is the case when file does exist and is a
1490 directory (see `mkalldirs' for explanation). */
1492 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1493 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1496 unique = unique_name (fname, true);
1497 if (unique != fname)
1502 /* Resolve "." and ".." elements of PATH by destructively modifying
1503 PATH and return true if PATH has been modified, false otherwise.
1505 The algorithm is in spirit similar to the one described in rfc1808,
1506 although implemented differently, in one pass. To recap, path
1507 elements containing only "." are removed, and ".." is taken to mean
1508 "back up one element". Single leading and trailing slashes are
1511 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1512 test examples are provided below. If you change anything in this
1513 function, run test_path_simplify to make sure you haven't broken a
1517 path_simplify (char *path)
1519 char *h = path; /* hare */
1520 char *t = path; /* tortoise */
1521 char *beg = path; /* boundary for backing the tortoise */
1522 char *end = path + strlen (path);
1526 /* Hare should be at the beginning of a path element. */
1528 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1533 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1535 /* Handle "../" by retreating the tortoise by one path
1536 element -- but not past beggining. */
1539 /* Move backwards until T hits the beginning of the
1540 previous path element or the beginning of path. */
1541 for (--t; t > beg && t[-1] != '/'; t--)
1546 /* If we're at the beginning, copy the "../" literally
1547 move the beginning so a later ".." doesn't remove
1557 /* A regular path element. If H hasn't advanced past T,
1558 simply skip to the next path element. Otherwise, copy
1559 the path element until the next slash. */
1562 /* Skip the path element, including the slash. */
1563 while (h < end && *h != '/')
1570 /* Copy the path element, including the final slash. */
1571 while (h < end && *h != '/')
1585 /* Return the length of URL's path. Path is considered to be
1586 terminated by one of '?', ';', '#', or by the end of the
1590 path_length (const char *url)
1592 const char *q = strpbrk_or_eos (url, "?;#");
1596 /* Find the last occurrence of character C in the range [b, e), or
1597 NULL, if none are present. We might want to use memrchr (a GNU
1598 extension) under GNU libc. */
1601 find_last_char (const char *b, const char *e, char c)
1609 /* Merge BASE with LINK and return the resulting URI.
1611 Either of the URIs may be absolute or relative, complete with the
1612 host name, or path only. This tries to reasonably handle all
1613 foreseeable cases. It only employs minimal URL parsing, without
1614 knowledge of the specifics of schemes.
1616 I briefly considered making this function call path_simplify after
1617 the merging process, as rfc1738 seems to suggest. This is a bad
1618 idea for several reasons: 1) it complexifies the code, and 2)
1619 url_parse has to simplify path anyway, so it's wasteful to boot. */
1622 uri_merge (const char *base, const char *link)
1628 if (url_has_scheme (link))
1629 return xstrdup (link);
1631 /* We may not examine BASE past END. */
1632 end = base + path_length (base);
1633 linklength = strlen (link);
1637 /* Empty LINK points back to BASE, query string and all. */
1638 return xstrdup (base);
1640 else if (*link == '?')
1642 /* LINK points to the same location, but changes the query
1643 string. Examples: */
1644 /* uri_merge("path", "?new") -> "path?new" */
1645 /* uri_merge("path?foo", "?new") -> "path?new" */
1646 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1647 /* uri_merge("path#foo", "?new") -> "path?new" */
1648 int baselength = end - base;
1649 merge = xmalloc (baselength + linklength + 1);
1650 memcpy (merge, base, baselength);
1651 memcpy (merge + baselength, link, linklength);
1652 merge[baselength + linklength] = '\0';
1654 else if (*link == '#')
1656 /* uri_merge("path", "#new") -> "path#new" */
1657 /* uri_merge("path#foo", "#new") -> "path#new" */
1658 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1659 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1661 const char *end1 = strchr (base, '#');
1663 end1 = base + strlen (base);
1664 baselength = end1 - base;
1665 merge = xmalloc (baselength + linklength + 1);
1666 memcpy (merge, base, baselength);
1667 memcpy (merge + baselength, link, linklength);
1668 merge[baselength + linklength] = '\0';
1670 else if (*link == '/' && *(link + 1) == '/')
1672 /* LINK begins with "//" and so is a net path: we need to
1673 replace everything after (and including) the double slash
1676 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1677 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1678 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1682 const char *start_insert;
1684 /* Look for first slash. */
1685 slash = memchr (base, '/', end - base);
1686 /* If found slash and it is a double slash, then replace
1687 from this point, else default to replacing from the
1689 if (slash && *(slash + 1) == '/')
1690 start_insert = slash;
1692 start_insert = base;
1694 span = start_insert - base;
1695 merge = xmalloc (span + linklength + 1);
1697 memcpy (merge, base, span);
1698 memcpy (merge + span, link, linklength);
1699 merge[span + linklength] = '\0';
1701 else if (*link == '/')
1703 /* LINK is an absolute path: we need to replace everything
1704 after (and including) the FIRST slash with LINK.
1706 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1707 "/qux/xyzzy", our result should be
1708 "http://host/qux/xyzzy". */
1711 const char *start_insert = NULL; /* for gcc to shut up. */
1712 const char *pos = base;
1713 bool seen_slash_slash = false;
1714 /* We're looking for the first slash, but want to ignore
1717 slash = memchr (pos, '/', end - pos);
1718 if (slash && !seen_slash_slash)
1719 if (*(slash + 1) == '/')
1722 seen_slash_slash = true;
1726 /* At this point, SLASH is the location of the first / after
1727 "//", or the first slash altogether. START_INSERT is the
1728 pointer to the location where LINK will be inserted. When
1729 examining the last two examples, keep in mind that LINK
1732 if (!slash && !seen_slash_slash)
1733 /* example: "foo" */
1735 start_insert = base;
1736 else if (!slash && seen_slash_slash)
1737 /* example: "http://foo" */
1740 else if (slash && !seen_slash_slash)
1741 /* example: "foo/bar" */
1743 start_insert = base;
1744 else if (slash && seen_slash_slash)
1745 /* example: "http://something/" */
1747 start_insert = slash;
1749 span = start_insert - base;
1750 merge = xmalloc (span + linklength + 1);
1752 memcpy (merge, base, span);
1753 memcpy (merge + span, link, linklength);
1754 merge[span + linklength] = '\0';
1758 /* LINK is a relative URL: we need to replace everything
1759 after last slash (possibly empty) with LINK.
1761 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1762 our result should be "whatever/foo/qux/xyzzy". */
1763 bool need_explicit_slash = false;
1765 const char *start_insert;
1766 const char *last_slash = find_last_char (base, end, '/');
1769 /* No slash found at all. Replace what we have with LINK. */
1770 start_insert = base;
1772 else if (last_slash && last_slash >= base + 2
1773 && last_slash[-2] == ':' && last_slash[-1] == '/')
1775 /* example: http://host" */
1777 start_insert = end + 1;
1778 need_explicit_slash = true;
1782 /* example: "whatever/foo/bar" */
1784 start_insert = last_slash + 1;
1787 span = start_insert - base;
1788 merge = xmalloc (span + linklength + 1);
1790 memcpy (merge, base, span);
1791 if (need_explicit_slash)
1792 merge[span - 1] = '/';
1793 memcpy (merge + span, link, linklength);
1794 merge[span + linklength] = '\0';
1800 #define APPEND(p, s) do { \
1801 int len = strlen (s); \
1802 memcpy (p, s, len); \
1806 /* Use this instead of password when the actual password is supposed
1807 to be hidden. We intentionally use a generic string without giving
1808 away the number of characters in the password, like previous
1810 #define HIDDEN_PASSWORD "*password*"
1812 /* Recreate the URL string from the data in URL.
1814 If HIDE is true (as it is when we're calling this on a URL we plan
1815 to print, but not when calling it to canonicalize a URL for use
1816 within the program), password will be hidden. Unsafe characters in
1817 the URL will be quoted. */
1820 url_string (const struct url *url, bool hide_password)
1824 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1826 int scheme_port = supported_schemes[url->scheme].default_port;
1827 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1828 int fplen = full_path_length (url);
1830 bool brackets_around_host;
1832 assert (scheme_str != NULL);
1834 /* Make sure the user name and password are quoted. */
1837 quoted_user = url_escape_allow_passthrough (url->user);
1841 quoted_passwd = HIDDEN_PASSWORD;
1843 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1847 /* In the unlikely event that the host name contains non-printable
1848 characters, quote it for displaying to the user. */
1849 quoted_host = url_escape_allow_passthrough (url->host);
1851 /* Undo the quoting of colons that URL escaping performs. IPv6
1852 addresses may legally contain colons, and in that case must be
1853 placed in square brackets. */
1854 if (quoted_host != url->host)
1855 unescape_single_char (quoted_host, ':');
1856 brackets_around_host = strchr (quoted_host, ':') != NULL;
1858 size = (strlen (scheme_str)
1859 + strlen (quoted_host)
1860 + (brackets_around_host ? 2 : 0)
1863 if (url->port != scheme_port)
1864 size += 1 + numdigit (url->port);
1867 size += 1 + strlen (quoted_user);
1869 size += 1 + strlen (quoted_passwd);
1872 p = result = xmalloc (size);
1874 APPEND (p, scheme_str);
1877 APPEND (p, quoted_user);
1881 APPEND (p, quoted_passwd);
1886 if (brackets_around_host)
1888 APPEND (p, quoted_host);
1889 if (brackets_around_host)
1891 if (url->port != scheme_port)
1894 p = number_to_string (p, url->port);
1897 full_path_write (url, p);
1901 assert (p - result == size);
1903 if (quoted_user && quoted_user != url->user)
1904 xfree (quoted_user);
1905 if (quoted_passwd && !hide_password && quoted_passwd != url->passwd)
1906 xfree (quoted_passwd);
1907 if (quoted_host != url->host)
1908 xfree (quoted_host);
1913 /* Return true if scheme a is similar to scheme b.
1915 Schemes are similar if they are equal. If SSL is supported, schemes
1916 are also similar if one is http (SCHEME_HTTP) and the other is https
1919 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1924 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1925 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1932 /* Debugging and testing support for path_simplify. */
1934 /* Debug: run path_simplify on PATH and return the result in a new
1935 string. Useful for calling from the debugger. */
1939 char *copy = xstrdup (path);
1940 path_simplify (copy);
1945 run_test (char *test, char *expected_result, bool expected_change)
1947 char *test_copy = xstrdup (test);
1948 bool modified = path_simplify (test_copy);
1950 if (0 != strcmp (test_copy, expected_result))
1952 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
1953 test, expected_result, test_copy);
1955 if (modified != expected_change)
1957 if (expected_change)
1958 printf ("Expected modification with path_simplify(\"%s\").\n",
1961 printf ("Expected no modification with path_simplify(\"%s\").\n",
1968 test_path_simplify (void)
1971 char *test, *result;
1977 { "..", "..", false },
1978 { "../", "../", false },
1979 { "foo", "foo", false },
1980 { "foo/bar", "foo/bar", false },
1981 { "foo///bar", "foo///bar", false },
1982 { "foo/.", "foo/", true },
1983 { "foo/./", "foo/", true },
1984 { "foo./", "foo./", false },
1985 { "foo/../bar", "bar", true },
1986 { "foo/../bar/", "bar/", true },
1987 { "foo/bar/..", "foo/", true },
1988 { "foo/bar/../x", "foo/x", true },
1989 { "foo/bar/../x/", "foo/x/", true },
1990 { "foo/..", "", true },
1991 { "foo/../..", "..", true },
1992 { "foo/../../..", "../..", true },
1993 { "foo/../../bar/../../baz", "../../baz", true },
1994 { "a/b/../../c", "c", true },
1995 { "./a/../b", "b", true }
1999 for (i = 0; i < countof (tests); i++)
2001 char *test = tests[i].test;
2002 char *expected_result = tests[i].result;
2003 bool expected_change = tests[i].should_modify;
2004 run_test (test, expected_result, expected_change);