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. */
39 #include <sys/types.h>
49 #include "host.h" /* for is_valid_ipv6_address */
58 const char *leading_string;
63 /* Supported schemes: */
64 static struct scheme_data supported_schemes[] =
66 { "http", "http://", DEFAULT_HTTP_PORT, 1 },
68 { "https", "https://", DEFAULT_HTTPS_PORT, 1 },
70 { "ftp", "ftp://", DEFAULT_FTP_PORT, 1 },
76 /* Forward declarations: */
78 static int path_simplify PARAMS ((char *));
80 /* Support for escaping and unescaping of URL strings. */
82 /* Table of "reserved" and "unsafe" characters. Those terms are
83 rfc1738-speak, as such largely obsoleted by rfc2396 and later
84 specs, but the general idea remains.
86 A reserved character is the one that you can't decode without
87 changing the meaning of the URL. For example, you can't decode
88 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
89 path components is different. Non-reserved characters can be
90 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
91 unsafe characters are loosely based on rfc1738, plus "$" and ",",
92 as recommended by rfc2396, and minus "~", which is very frequently
93 used (and sometimes unrecognized as %7E by broken servers).
95 An unsafe character is the one that should be encoded when URLs are
96 placed in foreign environments. E.g. space and newline are unsafe
97 in HTTP contexts because HTTP uses them as separator and line
98 terminator, so they must be encoded to %20 and %0A respectively.
99 "*" is unsafe in shell context, etc.
101 We determine whether a character is unsafe through static table
102 lookup. This code assumes ASCII character set and 8-bit chars. */
105 /* rfc1738 reserved chars + "$" and ",". */
108 /* rfc1738 unsafe chars, plus non-printables. */
112 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
113 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
114 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
116 /* Shorthands for the table: */
117 #define R urlchr_reserved
118 #define U urlchr_unsafe
121 static const unsigned char urlchr_table[256] =
123 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
124 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
125 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
126 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
127 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
128 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
129 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
130 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
131 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
132 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
133 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
134 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
135 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
136 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
137 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
138 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
140 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
141 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
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,
145 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
146 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
147 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
148 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
154 /* URL-unescape the string S.
156 This is done by transforming the sequences "%HH" to the character
157 represented by the hexadecimal digits HH. If % is not followed by
158 two hexadecimal digits, it is inserted literally.
160 The transformation is done in place. If you need the original
161 string intact, make a copy before calling this function. */
164 url_unescape (char *s)
166 char *t = s; /* t - tortoise */
167 char *h = s; /* h - hare */
179 /* Do nothing if '%' is not followed by two hex digits. */
180 if (!h[1] || !h[2] || !(ISXDIGIT (h[1]) && ISXDIGIT (h[2])))
182 c = X2DIGITS_TO_NUM (h[1], h[2]);
183 /* Don't unescape %00 because there is no way to insert it
184 into a C string without effectively truncating it. */
194 /* The core of url_escape_* functions. Escapes the characters that
195 match the provided mask in urlchr_table.
197 If ALLOW_PASSTHROUGH is non-zero, a string with no unsafe chars
198 will be returned unchanged. If ALLOW_PASSTHROUGH is zero, a
199 freshly allocated string will be returned in all cases. */
202 url_escape_1 (const char *s, unsigned char mask, int allow_passthrough)
209 for (p1 = s; *p1; p1++)
210 if (urlchr_test (*p1, mask))
211 addition += 2; /* Two more characters (hex digits) */
214 return allow_passthrough ? (char *)s : xstrdup (s);
216 newlen = (p1 - s) + addition;
217 newstr = (char *)xmalloc (newlen + 1);
223 /* Quote the characters that match the test mask. */
224 if (urlchr_test (*p1, mask))
226 unsigned char c = *p1++;
228 *p2++ = XNUM_TO_DIGIT (c >> 4);
229 *p2++ = XNUM_TO_DIGIT (c & 0xf);
234 assert (p2 - newstr == newlen);
240 /* URL-escape the unsafe characters (see urlchr_table) in a given
241 string, returning a freshly allocated string. */
244 url_escape (const char *s)
246 return url_escape_1 (s, urlchr_unsafe, 0);
249 /* URL-escape the unsafe characters (see urlchr_table) in a given
250 string. If no characters are unsafe, S is returned. */
253 url_escape_allow_passthrough (const char *s)
255 return url_escape_1 (s, urlchr_unsafe, 1);
258 /* Decide whether the char at position P needs to be encoded. (It is
259 not enough to pass a single char *P because the function may need
260 to inspect the surrounding context.)
262 Return 1 if the char should be escaped as %XX, 0 otherwise. */
265 char_needs_escaping (const char *p)
269 if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2)))
272 /* Garbled %.. sequence: encode `%'. */
275 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
281 /* Translate a %-escaped (but possibly non-conformant) input string S
282 into a %-escaped (and conformant) output string. If no characters
283 are encoded or decoded, return the same string S; otherwise, return
284 a freshly allocated string with the new contents.
286 After a URL has been run through this function, the protocols that
287 use `%' as the quote character can use the resulting string as-is,
288 while those that don't can use url_unescape to get to the intended
289 data. This function is stable: once the input is transformed,
290 further transformations of the result yield the same output.
292 Let's discuss why this function is needed.
294 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
295 a raw space character would mess up the HTTP request, it needs to
296 be quoted, like this:
298 GET /abc%20def HTTP/1.0
300 It appears that the unsafe chars need to be quoted, for example
301 with url_escape. But what if we're requested to download
302 `abc%20def'? url_escape transforms "%" to "%25", which would leave
303 us with `abc%2520def'. This is incorrect -- since %-escapes are
304 part of URL syntax, "%20" is the correct way to denote a literal
305 space on the Wget command line. This leaves us in the conclusion
306 that in that case Wget should not call url_escape, but leave the
307 `%20' as is. This is clearly contradictory, but it only gets
310 What if the requested URI is `abc%20 def'? If we call url_escape,
311 we end up with `/abc%2520%20def', which is almost certainly not
312 intended. If we don't call url_escape, we are left with the
313 embedded space and cannot complete the request. What the user
314 meant was for Wget to request `/abc%20%20def', and this is where
315 reencode_escapes kicks in.
317 Wget used to solve this by first decoding %-quotes, and then
318 encoding all the "unsafe" characters found in the resulting string.
319 This was wrong because it didn't preserve certain URL special
320 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
321 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
322 whether we considered `+' reserved (it is). One of these results
323 is inevitable because by the second step we would lose information
324 on whether the `+' was originally encoded or not. Both results
325 were wrong because in CGI parameters + means space, while %2B means
326 literal plus. reencode_escapes correctly translates the above to
327 "a%2B+b", i.e. returns the original string.
329 This function uses a modified version of the algorithm originally
330 proposed by Anon Sricharoenchai:
332 * Encode all "unsafe" characters, except those that are also
333 "reserved", to %XX. See urlchr_table for which characters are
336 * Encode the "%" characters not followed by two hex digits to
339 * Pass through all other characters and %XX escapes as-is. (Up to
340 Wget 1.10 this decoded %XX escapes corresponding to "safe"
341 characters, but that was obtrusive and broke some servers.)
345 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
347 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
351 "foo bar" -> "foo%20bar"
352 "foo%20bar" -> "foo%20bar"
353 "foo %20bar" -> "foo%20%20bar"
354 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
355 "foo%25%20bar" -> "foo%25%20bar"
356 "foo%2%20bar" -> "foo%252%20bar"
357 "foo+bar" -> "foo+bar" (plus is reserved!)
358 "foo%2b+bar" -> "foo%2b+bar" */
361 reencode_escapes (const char *s)
367 int encode_count = 0;
369 /* First pass: inspect the string to see if there's anything to do,
370 and to calculate the new length. */
371 for (p1 = s; *p1; p1++)
372 if (char_needs_escaping (p1))
376 /* The string is good as it is. */
377 return (char *) s; /* C const model sucks. */
380 /* Each encoding adds two characters (hex digits). */
381 newlen = oldlen + 2 * encode_count;
382 newstr = xmalloc (newlen + 1);
384 /* Second pass: copy the string to the destination address, encoding
385 chars when needed. */
390 if (char_needs_escaping (p1))
392 unsigned char c = *p1++;
394 *p2++ = XNUM_TO_DIGIT (c >> 4);
395 *p2++ = XNUM_TO_DIGIT (c & 0xf);
401 assert (p2 - newstr == newlen);
405 /* Returns the scheme type if the scheme is supported, or
406 SCHEME_INVALID if not. */
409 url_scheme (const char *url)
413 for (i = 0; supported_schemes[i].leading_string; i++)
414 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
415 strlen (supported_schemes[i].leading_string)))
417 if (supported_schemes[i].enabled)
418 return (enum url_scheme) i;
420 return SCHEME_INVALID;
423 return SCHEME_INVALID;
426 #define SCHEME_CHAR(ch) (ISALNUM (ch) || (ch) == '-' || (ch) == '+')
428 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
429 currently implemented, it returns true if URL begins with
433 url_has_scheme (const char *url)
437 /* The first char must be a scheme char. */
438 if (!*p || !SCHEME_CHAR (*p))
441 /* Followed by 0 or more scheme chars. */
442 while (*p && SCHEME_CHAR (*p))
444 /* Terminated by ':'. */
449 scheme_default_port (enum url_scheme scheme)
451 return supported_schemes[scheme].default_port;
455 scheme_disable (enum url_scheme scheme)
457 supported_schemes[scheme].enabled = 0;
460 /* Skip the username and password, if present in the URL. The
461 function should *not* be called with the complete URL, but with the
462 portion after the scheme.
464 If no username and password are found, return URL. */
467 url_skip_credentials (const char *url)
469 /* Look for '@' that comes before terminators, such as '/', '?',
471 const char *p = (const char *)strpbrk (url, "@/?#;");
477 /* Parse credentials contained in [BEG, END). The region is expected
478 to have come from a URL and is unescaped. */
481 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
487 return 0; /* empty user name */
489 colon = memchr (beg, ':', end - beg);
491 return 0; /* again empty user name */
495 *passwd = strdupdelim (colon + 1, end);
497 url_unescape (*passwd);
504 *user = strdupdelim (beg, userend);
505 url_unescape (*user);
509 /* Used by main.c: detect URLs written using the "shorthand" URL forms
510 popularized by Netscape and NcFTP. HTTP shorthands look like this:
512 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
513 www.foo.com[:port] -> http://www.foo.com[:port]
515 FTP shorthands look like this:
517 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
518 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
520 If the URL needs not or cannot be rewritten, return NULL. */
523 rewrite_shorthand_url (const char *url)
527 if (url_scheme (url) != SCHEME_INVALID)
530 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
532 for (p = url; *p && *p != ':' && *p != '/'; p++)
542 /* If the characters after the colon and before the next slash
543 or end of string are all digits, it's HTTP. */
545 for (pp = p + 1; ISDIGIT (*pp); pp++)
547 if (digits > 0 && (*pp == '/' || *pp == '\0'))
550 /* Prepend "ftp://" to the entire URL... */
551 res = xmalloc (6 + strlen (url) + 1);
552 sprintf (res, "ftp://%s", url);
553 /* ...and replace ':' with '/'. */
554 res[6 + (p - url)] = '/';
561 /* Just prepend "http://" to what we have. */
562 res = xmalloc (7 + strlen (url) + 1);
563 sprintf (res, "http://%s", url);
568 static void split_path PARAMS ((const char *, char **, char **));
570 /* Like strpbrk, with the exception that it returns the pointer to the
571 terminating zero (end-of-string aka "eos") if no matching character
574 Although I normally balk at Gcc-specific optimizations, it probably
575 makes sense here: glibc has optimizations that detect strpbrk being
576 called with literal string as ACCEPT and inline the search. That
577 optimization is defeated if strpbrk is hidden within the call to
578 another function. (And no, making strpbrk_or_eos inline doesn't
579 help because the check for literal accept is in the
584 #define strpbrk_or_eos(s, accept) ({ \
585 char *SOE_p = strpbrk (s, accept); \
587 SOE_p = (char *)s + strlen (s); \
591 #else /* not __GNUC__ */
594 strpbrk_or_eos (const char *s, const char *accept)
596 char *p = strpbrk (s, accept);
598 p = (char *)s + strlen (s);
603 /* Turn STR into lowercase; return non-zero if a character was
607 lowercase_str (char *str)
614 *str = TOLOWER (*str);
619 static const char *parse_errors[] = {
620 #define PE_NO_ERROR 0
622 #define PE_UNSUPPORTED_SCHEME 1
623 N_("Unsupported scheme"),
624 #define PE_EMPTY_HOST 2
626 #define PE_BAD_PORT_NUMBER 3
627 N_("Bad port number"),
628 #define PE_INVALID_USER_NAME 4
629 N_("Invalid user name"),
630 #define PE_UNTERMINATED_IPV6_ADDRESS 5
631 N_("Unterminated IPv6 numeric address"),
632 #define PE_IPV6_NOT_SUPPORTED 6
633 N_("IPv6 addresses not supported"),
634 #define PE_INVALID_IPV6_ADDRESS 7
635 N_("Invalid IPv6 numeric address")
640 Return a new struct url if successful, NULL on error. In case of
641 error, and if ERROR is not NULL, also set *ERROR to the appropriate
644 url_parse (const char *url, int *error)
648 int path_modified, host_modified;
650 enum url_scheme scheme;
652 const char *uname_b, *uname_e;
653 const char *host_b, *host_e;
654 const char *path_b, *path_e;
655 const char *params_b, *params_e;
656 const char *query_b, *query_e;
657 const char *fragment_b, *fragment_e;
660 char *user = NULL, *passwd = NULL;
662 char *url_encoded = NULL;
666 scheme = url_scheme (url);
667 if (scheme == SCHEME_INVALID)
669 error_code = PE_UNSUPPORTED_SCHEME;
673 url_encoded = reencode_escapes (url);
676 p += strlen (supported_schemes[scheme].leading_string);
678 p = url_skip_credentials (p);
681 /* scheme://user:pass@host[:port]... */
684 /* We attempt to break down the URL into the components path,
685 params, query, and fragment. They are ordered like this:
687 scheme://host[:port][/path][;params][?query][#fragment] */
689 params_b = params_e = NULL;
690 query_b = query_e = NULL;
691 fragment_b = fragment_e = NULL;
697 /* Handle IPv6 address inside square brackets. Ideally we'd
698 just look for the terminating ']', but rfc2732 mandates
699 rejecting invalid IPv6 addresses. */
701 /* The address begins after '['. */
703 host_e = strchr (host_b, ']');
707 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
712 /* Check if the IPv6 address is valid. */
713 if (!is_valid_ipv6_address(host_b, host_e))
715 error_code = PE_INVALID_IPV6_ADDRESS;
719 /* Continue parsing after the closing ']'. */
722 error_code = PE_IPV6_NOT_SUPPORTED;
728 p = strpbrk_or_eos (p, ":/;?#");
732 if (host_b == host_e)
734 error_code = PE_EMPTY_HOST;
738 port = scheme_default_port (scheme);
741 const char *port_b, *port_e, *pp;
743 /* scheme://host:port/tralala */
747 p = strpbrk_or_eos (p, "/;?#");
750 /* Allow empty port, as per rfc2396. */
751 if (port_b != port_e)
753 for (port = 0, pp = port_b; pp < port_e; pp++)
757 /* http://host:12randomgarbage/blah */
759 error_code = PE_BAD_PORT_NUMBER;
762 port = 10 * port + (*pp - '0');
763 /* Check for too large port numbers here, before we have
764 a chance to overflow on bogus port values. */
767 error_code = PE_BAD_PORT_NUMBER;
778 p = strpbrk_or_eos (p, ";?#");
783 /* Path is not allowed not to exist. */
791 p = strpbrk_or_eos (p, "?#");
798 p = strpbrk_or_eos (p, "#");
801 /* Hack that allows users to use '?' (a wildcard character) in
802 FTP URLs without it being interpreted as a query string
804 if (scheme == SCHEME_FTP)
806 query_b = query_e = NULL;
819 if (uname_b != uname_e)
821 /* http://user:pass@host */
823 /* uname_b uname_e */
824 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
826 error_code = PE_INVALID_USER_NAME;
831 u = xnew0 (struct url);
833 u->host = strdupdelim (host_b, host_e);
838 u->path = strdupdelim (path_b, path_e);
839 path_modified = path_simplify (u->path);
840 split_path (u->path, &u->dir, &u->file);
842 host_modified = lowercase_str (u->host);
844 /* Decode %HH sequences in host name. This is important not so much
845 to support %HH sequences in host names (which other browser
846 don't), but to support binary characters (which will have been
847 converted to %HH by reencode_escapes). */
848 if (strchr (u->host, '%'))
850 url_unescape (u->host);
855 u->params = strdupdelim (params_b, params_e);
857 u->query = strdupdelim (query_b, query_e);
859 u->fragment = strdupdelim (fragment_b, fragment_e);
861 if (path_modified || u->fragment || host_modified || path_b == path_e)
863 /* If we suspect that a transformation has rendered what
864 url_string might return different from URL_ENCODED, rebuild
865 u->url using url_string. */
866 u->url = url_string (u, 0);
868 if (url_encoded != url)
869 xfree ((char *) url_encoded);
873 if (url_encoded == url)
874 u->url = xstrdup (url);
876 u->url = url_encoded;
883 /* Cleanup in case of error: */
884 if (url_encoded && url_encoded != url)
887 /* Transmit the error code to the caller, if the caller wants to
894 /* Return the error message string from ERROR_CODE, which should have
895 been retrieved from url_parse. The error message is translated. */
898 url_error (int error_code)
900 assert (error_code >= 0 && error_code < countof (parse_errors));
901 return _(parse_errors[error_code]);
904 /* Split PATH into DIR and FILE. PATH comes from the URL and is
905 expected to be URL-escaped.
907 The path is split into directory (the part up to the last slash)
908 and file (the part after the last slash), which are subsequently
912 "foo/bar/baz" "foo/bar" "baz"
913 "foo/bar/" "foo/bar" ""
915 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
917 DIR and FILE are freshly allocated. */
920 split_path (const char *path, char **dir, char **file)
922 char *last_slash = strrchr (path, '/');
926 *file = xstrdup (path);
930 *dir = strdupdelim (path, last_slash);
931 *file = xstrdup (last_slash + 1);
934 url_unescape (*file);
937 /* Note: URL's "full path" is the path with the query string and
938 params appended. The "fragment" (#foo) is intentionally ignored,
939 but that might be changed. For example, if the original URL was
940 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
941 the full path will be "/foo/bar/baz;bullshit?querystring". */
943 /* Return the length of the full path, without the terminating
947 full_path_length (const struct url *url)
951 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
962 /* Write out the full path. */
965 full_path_write (const struct url *url, char *where)
967 #define FROB(el, chr) do { \
968 char *f_el = url->el; \
970 int l = strlen (f_el); \
972 memcpy (where, f_el, l); \
984 /* Public function for getting the "full path". E.g. if u->path is
985 "foo/bar" and u->query is "param=value", full_path will be
986 "/foo/bar?param=value". */
989 url_full_path (const struct url *url)
991 int length = full_path_length (url);
992 char *full_path = (char *) xmalloc (length + 1);
994 full_path_write (url, full_path);
995 full_path[length] = '\0';
1000 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1001 escaping of certain characters, such as "/" and ":". Returns a
1002 count of unescaped chars. */
1005 unescape_single_char (char *str, char chr)
1007 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1008 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1009 char *h = str; /* hare */
1010 char *t = str; /* tortoise */
1011 for (; *h; h++, t++)
1013 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1024 /* Escape unsafe and reserved characters, except for the slash
1028 url_escape_dir (const char *dir)
1030 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1034 unescape_single_char (newdir, '/');
1038 /* Sync u->path and u->url with u->dir and u->file. Called after
1039 u->file or u->dir have been changed, typically by the FTP code. */
1042 sync_path (struct url *u)
1044 char *newpath, *efile, *edir;
1048 /* u->dir and u->file are not escaped. URL-escape them before
1049 reassembling them into u->path. That way, if they contain
1050 separators like '?' or even if u->file contains slashes, the
1051 path will be correctly assembled. (u->file can contain slashes
1052 if the URL specifies it with %2f, or if an FTP server returns
1054 edir = url_escape_dir (u->dir);
1055 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1058 newpath = xstrdup (efile);
1061 int dirlen = strlen (edir);
1062 int filelen = strlen (efile);
1064 /* Copy "DIR/FILE" to newpath. */
1065 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1066 memcpy (p, edir, dirlen);
1069 memcpy (p, efile, filelen);
1078 if (efile != u->file)
1081 /* Regenerate u->url as well. */
1083 u->url = url_string (u, 0);
1086 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1087 This way we can sync u->path and u->url when they get changed. */
1090 url_set_dir (struct url *url, const char *newdir)
1093 url->dir = xstrdup (newdir);
1098 url_set_file (struct url *url, const char *newfile)
1101 url->file = xstrdup (newfile);
1106 url_free (struct url *url)
1112 xfree_null (url->params);
1113 xfree_null (url->query);
1114 xfree_null (url->fragment);
1115 xfree_null (url->user);
1116 xfree_null (url->passwd);
1124 /* Create all the necessary directories for PATH (a file). Calls
1125 mkdirhier() internally. */
1127 mkalldirs (const char *path)
1134 p = path + strlen (path);
1135 for (; *p != '/' && p != path; p--)
1138 /* Don't create if it's just a file. */
1139 if ((p == path) && (*p != '/'))
1141 t = strdupdelim (path, p);
1143 /* Check whether the directory exists. */
1144 if ((stat (t, &st) == 0))
1146 if (S_ISDIR (st.st_mode))
1153 /* If the dir exists as a file name, remove it first. This
1154 is *only* for Wget to work with buggy old CERN http
1155 servers. Here is the scenario: When Wget tries to
1156 retrieve a directory without a slash, e.g.
1157 http://foo/bar (bar being a directory), CERN server will
1158 not redirect it too http://foo/bar/ -- it will generate a
1159 directory listing containing links to bar/file1,
1160 bar/file2, etc. Wget will lose because it saves this
1161 HTML listing to a file `bar', so it cannot create the
1162 directory. To work around this, if the file of the same
1163 name exists, we just remove it and create the directory
1165 DEBUGP (("Removing %s because of directory danger!\n", t));
1169 res = make_directory (t);
1171 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1176 /* Functions for constructing the file name out of URL components. */
1178 /* A growable string structure, used by url_file_name and friends.
1179 This should perhaps be moved to utils.c.
1181 The idea is to have a convenient and efficient way to construct a
1182 string by having various functions append data to it. Instead of
1183 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1184 functions in questions, we pass the pointer to this struct. */
1192 /* Ensure that the string can accept APPEND_COUNT more characters past
1193 the current TAIL position. If necessary, this will grow the string
1194 and update its allocated size. If the string is already large
1195 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1196 #define GROW(g, append_size) do { \
1197 struct growable *G_ = g; \
1198 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1201 /* Return the tail position of the string. */
1202 #define TAIL(r) ((r)->base + (r)->tail)
1204 /* Move the tail position by APPEND_COUNT characters. */
1205 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1207 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1211 append_string (const char *str, struct growable *dest)
1213 int l = strlen (str);
1215 memcpy (TAIL (dest), str, l);
1216 TAIL_INCR (dest, l);
1219 /* Append CH to DEST. For example, append_char (0, DEST)
1220 zero-terminates DEST. */
1223 append_char (char ch, struct growable *dest)
1227 TAIL_INCR (dest, 1);
1231 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1232 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1233 filechr_control = 4 /* a control character, e.g. 0-31 */
1236 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1238 /* Shorthands for the table: */
1239 #define U filechr_not_unix
1240 #define W filechr_not_windows
1241 #define C filechr_control
1246 /* Table of characters unsafe under various conditions (see above).
1248 Arguably we could also claim `%' to be unsafe, since we use it as
1249 the escape character. If we ever want to be able to reliably
1250 translate file name back to URL, this would become important
1251 crucial. Right now, it's better to be minimal in escaping. */
1253 static const unsigned char filechr_table[256] =
1255 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1256 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1257 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1258 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1259 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1260 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1261 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1262 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1263 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1264 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1265 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1266 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1267 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1268 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1269 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1270 0, 0, 0, 0, 0, 0, 0, 0, /* x y z { | } ~ DEL */
1272 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1273 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
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,
1277 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1278 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1279 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1280 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1288 /* FN_PORT_SEP is the separator between host and port in file names
1289 for non-standard port numbers. On Unix this is normally ':', as in
1290 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1291 because Windows can't handle ':' in file names. */
1292 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1294 /* FN_QUERY_SEP is the separator between the file name and the URL
1295 query, normally '?'. Since Windows cannot handle '?' as part of
1296 file name, we use '@' instead there. */
1297 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1299 /* Quote path element, characters in [b, e), as file name, and append
1300 the quoted string to DEST. Each character is quoted as per
1301 file_unsafe_char and the corresponding table.
1303 If ESCAPED_P is non-zero, the path element is considered to be
1304 URL-escaped and will be unescaped prior to inspection. */
1307 append_uri_pathel (const char *b, const char *e, int escaped_p,
1308 struct growable *dest)
1314 if (opt.restrict_files_os == restrict_unix)
1315 mask = filechr_not_unix;
1317 mask = filechr_not_windows;
1318 if (opt.restrict_files_ctrl)
1319 mask |= filechr_control;
1321 /* Copy [b, e) to PATHEL and URL-unescape it. */
1325 BOUNDED_TO_ALLOCA (b, e, unescaped);
1326 url_unescape (unescaped);
1328 e = unescaped + strlen (unescaped);
1331 /* Defang ".." when found as component of path. Remember that path
1332 comes from the URL and might contain malicious input. */
1333 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1339 /* Walk the PATHEL string and check how many characters we'll need
1342 for (p = b; p < e; p++)
1343 if (FILE_CHAR_TEST (*p, mask))
1346 /* Calculate the length of the output string. e-b is the input
1347 string length. Each quoted char introduces two additional
1348 characters in the string, hence 2*quoted. */
1349 outlen = (e - b) + (2 * quoted);
1350 GROW (dest, outlen);
1354 /* If there's nothing to quote, we can simply append the string
1355 without processing it again. */
1356 memcpy (TAIL (dest), b, outlen);
1360 char *q = TAIL (dest);
1361 for (p = b; p < e; p++)
1363 if (!FILE_CHAR_TEST (*p, mask))
1367 unsigned char ch = *p;
1369 *q++ = XNUM_TO_DIGIT (ch >> 4);
1370 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1373 assert (q - TAIL (dest) == outlen);
1375 TAIL_INCR (dest, outlen);
1378 /* Append to DEST the directory structure that corresponds the
1379 directory part of URL's path. For example, if the URL is
1380 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1382 Each path element ("dir1" and "dir2" in the above example) is
1383 examined, url-unescaped, and re-escaped as file name element.
1385 Additionally, it cuts as many directories from the path as
1386 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1387 will produce "bar" for the above example. For 2 or more, it will
1390 Each component of the path is quoted for use as file name. */
1393 append_dir_structure (const struct url *u, struct growable *dest)
1395 char *pathel, *next;
1396 int cut = opt.cut_dirs;
1398 /* Go through the path components, de-URL-quote them, and quote them
1399 (if necessary) as file names. */
1402 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1407 /* Ignore empty pathels. */
1411 append_char ('/', dest);
1412 append_uri_pathel (pathel, next, 1, dest);
1416 /* Return a unique file name that matches the given URL as good as
1417 possible. Does not create directories on the file system. */
1420 url_file_name (const struct url *u)
1422 struct growable fnres; /* stands for "file name result" */
1424 const char *u_file, *u_query;
1425 char *fname, *unique;
1431 /* Start with the directory prefix, if specified. */
1433 append_string (opt.dir_prefix, &fnres);
1435 /* If "dirstruct" is turned on (typically the case with -r), add
1436 the host and port (unless those have been turned off) and
1437 directory structure. */
1440 if (opt.protocol_directories)
1443 append_char ('/', &fnres);
1444 append_string (supported_schemes[u->scheme].name, &fnres);
1446 if (opt.add_hostdir)
1449 append_char ('/', &fnres);
1450 if (0 != strcmp (u->host, ".."))
1451 append_string (u->host, &fnres);
1453 /* Host name can come from the network; malicious DNS may
1454 allow ".." to be resolved, causing us to write to
1455 "../<file>". Defang such host names. */
1456 append_string ("%2E%2E", &fnres);
1457 if (u->port != scheme_default_port (u->scheme))
1460 number_to_string (portstr, u->port);
1461 append_char (FN_PORT_SEP, &fnres);
1462 append_string (portstr, &fnres);
1466 append_dir_structure (u, &fnres);
1469 /* Add the file name. */
1471 append_char ('/', &fnres);
1472 u_file = *u->file ? u->file : "index.html";
1473 append_uri_pathel (u_file, u_file + strlen (u_file), 0, &fnres);
1475 /* Append "?query" to the file name. */
1476 u_query = u->query && *u->query ? u->query : NULL;
1479 append_char (FN_QUERY_SEP, &fnres);
1480 append_uri_pathel (u_query, u_query + strlen (u_query), 1, &fnres);
1483 /* Zero-terminate the file name. */
1484 append_char ('\0', &fnres);
1488 /* Check the cases in which the unique extensions are not used:
1489 1) Clobbering is turned off (-nc).
1490 2) Retrieval with regetting.
1491 3) Timestamping is used.
1492 4) Hierarchy is built.
1494 The exception is the case when file does exist and is a
1495 directory (see `mkalldirs' for explanation). */
1497 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1498 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1501 unique = unique_name (fname, 1);
1502 if (unique != fname)
1507 /* Resolve "." and ".." elements of PATH by destructively modifying
1508 PATH and return non-zero if PATH has been modified, zero otherwise.
1510 The algorithm is in spirit similar to the one described in rfc1808,
1511 although implemented differently, in one pass. To recap, path
1512 elements containing only "." are removed, and ".." is taken to mean
1513 "back up one element". Single leading and trailing slashes are
1516 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1517 test examples are provided below. If you change anything in this
1518 function, run test_path_simplify to make sure you haven't broken a
1522 path_simplify (char *path)
1524 char *h = path; /* hare */
1525 char *t = path; /* tortoise */
1526 char *beg = path; /* boundary for backing the tortoise */
1527 char *end = path + strlen (path);
1531 /* Hare should be at the beginning of a path element. */
1533 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1538 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1540 /* Handle "../" by retreating the tortoise by one path
1541 element -- but not past beggining. */
1544 /* Move backwards until T hits the beginning of the
1545 previous path element or the beginning of path. */
1546 for (--t; t > beg && t[-1] != '/'; t--)
1551 /* If we're at the beginning, copy the "../" literally
1552 move the beginning so a later ".." doesn't remove
1562 /* A regular path element. If H hasn't advanced past T,
1563 simply skip to the next path element. Otherwise, copy
1564 the path element until the next slash. */
1567 /* Skip the path element, including the slash. */
1568 while (h < end && *h != '/')
1575 /* Copy the path element, including the final slash. */
1576 while (h < end && *h != '/')
1590 /* Return the length of URL's path. Path is considered to be
1591 terminated by one of '?', ';', '#', or by the end of the
1595 path_length (const char *url)
1597 const char *q = strpbrk_or_eos (url, "?;#");
1601 /* Find the last occurrence of character C in the range [b, e), or
1602 NULL, if none are present. We might want to use memrchr (a GNU
1603 extension) under GNU libc. */
1606 find_last_char (const char *b, const char *e, char c)
1614 /* Merge BASE with LINK and return the resulting URI.
1616 Either of the URIs may be absolute or relative, complete with the
1617 host name, or path only. This tries to reasonably handle all
1618 foreseeable cases. It only employs minimal URL parsing, without
1619 knowledge of the specifics of schemes.
1621 I briefly considered making this function call path_simplify after
1622 the merging process, as rfc1738 seems to suggest. This is a bad
1623 idea for several reasons: 1) it complexifies the code, and 2)
1624 url_parse has to simplify path anyway, so it's wasteful to boot. */
1627 uri_merge (const char *base, const char *link)
1633 if (url_has_scheme (link))
1634 return xstrdup (link);
1636 /* We may not examine BASE past END. */
1637 end = base + path_length (base);
1638 linklength = strlen (link);
1642 /* Empty LINK points back to BASE, query string and all. */
1643 return xstrdup (base);
1645 else if (*link == '?')
1647 /* LINK points to the same location, but changes the query
1648 string. Examples: */
1649 /* uri_merge("path", "?new") -> "path?new" */
1650 /* uri_merge("path?foo", "?new") -> "path?new" */
1651 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1652 /* uri_merge("path#foo", "?new") -> "path?new" */
1653 int baselength = end - base;
1654 merge = xmalloc (baselength + linklength + 1);
1655 memcpy (merge, base, baselength);
1656 memcpy (merge + baselength, link, linklength);
1657 merge[baselength + linklength] = '\0';
1659 else if (*link == '#')
1661 /* uri_merge("path", "#new") -> "path#new" */
1662 /* uri_merge("path#foo", "#new") -> "path#new" */
1663 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1664 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1666 const char *end1 = strchr (base, '#');
1668 end1 = base + strlen (base);
1669 baselength = end1 - base;
1670 merge = xmalloc (baselength + linklength + 1);
1671 memcpy (merge, base, baselength);
1672 memcpy (merge + baselength, link, linklength);
1673 merge[baselength + linklength] = '\0';
1675 else if (*link == '/' && *(link + 1) == '/')
1677 /* LINK begins with "//" and so is a net path: we need to
1678 replace everything after (and including) the double slash
1681 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1682 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1683 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1687 const char *start_insert;
1689 /* Look for first slash. */
1690 slash = memchr (base, '/', end - base);
1691 /* If found slash and it is a double slash, then replace
1692 from this point, else default to replacing from the
1694 if (slash && *(slash + 1) == '/')
1695 start_insert = slash;
1697 start_insert = base;
1699 span = start_insert - base;
1700 merge = (char *)xmalloc (span + linklength + 1);
1702 memcpy (merge, base, span);
1703 memcpy (merge + span, link, linklength);
1704 merge[span + linklength] = '\0';
1706 else if (*link == '/')
1708 /* LINK is an absolute path: we need to replace everything
1709 after (and including) the FIRST slash with LINK.
1711 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1712 "/qux/xyzzy", our result should be
1713 "http://host/qux/xyzzy". */
1716 const char *start_insert = NULL; /* for gcc to shut up. */
1717 const char *pos = base;
1718 int seen_slash_slash = 0;
1719 /* We're looking for the first slash, but want to ignore
1722 slash = memchr (pos, '/', end - pos);
1723 if (slash && !seen_slash_slash)
1724 if (*(slash + 1) == '/')
1727 seen_slash_slash = 1;
1731 /* At this point, SLASH is the location of the first / after
1732 "//", or the first slash altogether. START_INSERT is the
1733 pointer to the location where LINK will be inserted. When
1734 examining the last two examples, keep in mind that LINK
1737 if (!slash && !seen_slash_slash)
1738 /* example: "foo" */
1740 start_insert = base;
1741 else if (!slash && seen_slash_slash)
1742 /* example: "http://foo" */
1745 else if (slash && !seen_slash_slash)
1746 /* example: "foo/bar" */
1748 start_insert = base;
1749 else if (slash && seen_slash_slash)
1750 /* example: "http://something/" */
1752 start_insert = slash;
1754 span = start_insert - base;
1755 merge = (char *)xmalloc (span + linklength + 1);
1757 memcpy (merge, base, span);
1758 memcpy (merge + span, link, linklength);
1759 merge[span + linklength] = '\0';
1763 /* LINK is a relative URL: we need to replace everything
1764 after last slash (possibly empty) with LINK.
1766 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1767 our result should be "whatever/foo/qux/xyzzy". */
1768 int need_explicit_slash = 0;
1770 const char *start_insert;
1771 const char *last_slash = find_last_char (base, end, '/');
1774 /* No slash found at all. Replace what we have with LINK. */
1775 start_insert = base;
1777 else if (last_slash && last_slash >= base + 2
1778 && last_slash[-2] == ':' && last_slash[-1] == '/')
1780 /* example: http://host" */
1782 start_insert = end + 1;
1783 need_explicit_slash = 1;
1787 /* example: "whatever/foo/bar" */
1789 start_insert = last_slash + 1;
1792 span = start_insert - base;
1793 merge = (char *)xmalloc (span + linklength + 1);
1795 memcpy (merge, base, span);
1796 if (need_explicit_slash)
1797 merge[span - 1] = '/';
1798 memcpy (merge + span, link, linklength);
1799 merge[span + linklength] = '\0';
1805 #define APPEND(p, s) do { \
1806 int len = strlen (s); \
1807 memcpy (p, s, len); \
1811 /* Use this instead of password when the actual password is supposed
1812 to be hidden. We intentionally use a generic string without giving
1813 away the number of characters in the password, like previous
1815 #define HIDDEN_PASSWORD "*password*"
1817 /* Recreate the URL string from the data in URL.
1819 If HIDE is non-zero (as it is when we're calling this on a URL we
1820 plan to print, but not when calling it to canonicalize a URL for
1821 use within the program), password will be hidden. Unsafe
1822 characters in the URL will be quoted. */
1825 url_string (const struct url *url, int hide_password)
1829 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1831 int scheme_port = supported_schemes[url->scheme].default_port;
1832 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1833 int fplen = full_path_length (url);
1835 int brackets_around_host;
1837 assert (scheme_str != NULL);
1839 /* Make sure the user name and password are quoted. */
1842 quoted_user = url_escape_allow_passthrough (url->user);
1846 quoted_passwd = HIDDEN_PASSWORD;
1848 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1852 /* In the unlikely event that the host name contains non-printable
1853 characters, quote it for displaying to the user. */
1854 quoted_host = url_escape_allow_passthrough (url->host);
1856 /* Undo the quoting of colons that URL escaping performs. IPv6
1857 addresses may legally contain colons, and in that case must be
1858 placed in square brackets. */
1859 if (quoted_host != url->host)
1860 unescape_single_char (quoted_host, ':');
1861 brackets_around_host = strchr (quoted_host, ':') != NULL;
1863 size = (strlen (scheme_str)
1864 + strlen (quoted_host)
1865 + (brackets_around_host ? 2 : 0)
1868 if (url->port != scheme_port)
1869 size += 1 + numdigit (url->port);
1872 size += 1 + strlen (quoted_user);
1874 size += 1 + strlen (quoted_passwd);
1877 p = result = xmalloc (size);
1879 APPEND (p, scheme_str);
1882 APPEND (p, quoted_user);
1886 APPEND (p, quoted_passwd);
1891 if (brackets_around_host)
1893 APPEND (p, quoted_host);
1894 if (brackets_around_host)
1896 if (url->port != scheme_port)
1899 p = number_to_string (p, url->port);
1902 full_path_write (url, p);
1906 assert (p - result == size);
1908 if (quoted_user && quoted_user != url->user)
1909 xfree (quoted_user);
1910 if (quoted_passwd && !hide_password && quoted_passwd != url->passwd)
1911 xfree (quoted_passwd);
1912 if (quoted_host != url->host)
1913 xfree (quoted_host);
1918 /* Return non-zero if scheme a is similar to scheme b.
1920 Schemes are similar if they are equal. If SSL is supported, schemes
1921 are also similar if one is http (SCHEME_HTTP) and the other is https
1924 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1929 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1930 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1937 /* Debugging and testing support for path_simplify. */
1939 /* Debug: run path_simplify on PATH and return the result in a new
1940 string. Useful for calling from the debugger. */
1944 char *copy = xstrdup (path);
1945 path_simplify (copy);
1950 run_test (char *test, char *expected_result, int expected_change)
1952 char *test_copy = xstrdup (test);
1953 int modified = path_simplify (test_copy);
1955 if (0 != strcmp (test_copy, expected_result))
1957 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
1958 test, expected_result, test_copy);
1960 if (modified != expected_change)
1962 if (expected_change == 1)
1963 printf ("Expected modification with path_simplify(\"%s\").\n",
1966 printf ("Expected no modification with path_simplify(\"%s\").\n",
1973 test_path_simplify (void)
1976 char *test, *result;
1983 { "../", "../", 0 },
1984 { "foo", "foo", 0 },
1985 { "foo/bar", "foo/bar", 0 },
1986 { "foo///bar", "foo///bar", 0 },
1987 { "foo/.", "foo/", 1 },
1988 { "foo/./", "foo/", 1 },
1989 { "foo./", "foo./", 0 },
1990 { "foo/../bar", "bar", 1 },
1991 { "foo/../bar/", "bar/", 1 },
1992 { "foo/bar/..", "foo/", 1 },
1993 { "foo/bar/../x", "foo/x", 1 },
1994 { "foo/bar/../x/", "foo/x/", 1 },
1995 { "foo/..", "", 1 },
1996 { "foo/../..", "..", 1 },
1997 { "foo/../../..", "../..", 1 },
1998 { "foo/../../bar/../../baz", "../../baz", 1 },
1999 { "a/b/../../c", "c", 1 },
2000 { "./a/../b", "b", 1 }
2004 for (i = 0; i < countof (tests); i++)
2006 char *test = tests[i].test;
2007 char *expected_result = tests[i].result;
2008 int expected_change = tests[i].should_modify;
2009 run_test (test, expected_result, expected_change);