2 Copyright (C) 1995, 1996, 1997, 2000, 2001, 2003, 2003
3 Free Software Foundation, Inc.
5 This file is part of GNU Wget.
7 GNU Wget is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 2 of the License, or (at
10 your option) any later version.
12 GNU Wget is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with Wget; if not, write to the Free Software
19 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
21 In addition, as a special exception, the Free Software Foundation
22 gives permission to link the code of its release of Wget with the
23 OpenSSL project's "OpenSSL" library (or with modified versions of it
24 that use the same license as the "OpenSSL" library), and distribute
25 the linked executables. You must obey the GNU General Public License
26 in all respects for all of the code used other than "OpenSSL". If you
27 modify this file, you may extend this exception to your version of the
28 file, but you are not obligated to do so. If you do not wish to do
29 so, delete this exception statement from your version. */
40 #include <sys/types.h>
62 /* Supported schemes: */
63 static struct scheme_data supported_schemes[] =
65 { "http://", DEFAULT_HTTP_PORT, 1 },
67 { "https://", DEFAULT_HTTPS_PORT, 1 },
69 { "ftp://", DEFAULT_FTP_PORT, 1 },
75 /* Forward declarations: */
77 static int path_simplify PARAMS ((char *));
79 /* Support for encoding and decoding of URL strings. We determine
80 whether a character is unsafe through static table lookup. This
81 code assumes ASCII character set and 8-bit chars. */
84 /* rfc1738 reserved chars, preserved from encoding. */
87 /* rfc1738 unsafe chars, plus some more. */
91 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
92 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
93 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
95 /* Shorthands for the table: */
96 #define R urlchr_reserved
97 #define U urlchr_unsafe
100 const static unsigned char urlchr_table[256] =
102 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
103 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
104 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
105 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
106 U, 0, U, RU, 0, U, R, 0, /* SP ! " # $ % & ' */
107 0, 0, 0, R, 0, 0, 0, R, /* ( ) * + , - . / */
108 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
109 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
110 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
111 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
112 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
113 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
114 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
115 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
116 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
117 0, 0, 0, U, U, U, U, U, /* x y z { | } ~ DEL */
119 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
120 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
121 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
122 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
124 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
125 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
126 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
127 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
133 /* URL-unescape the string S.
135 This is done by transforming the sequences "%HH" to the character
136 represented by the hexadecimal digits HH. If % is not followed by
137 two hexadecimal digits, it is inserted literally.
139 The transformation is done in place. If you need the original
140 string intact, make a copy before calling this function. */
143 url_unescape (char *s)
145 char *t = s; /* t - tortoise */
146 char *h = s; /* h - hare */
157 /* Do nothing if '%' is not followed by two hex digits. */
158 if (!h[1] || !h[2] || !(ISXDIGIT (h[1]) && ISXDIGIT (h[2])))
160 *t = X2DIGITS_TO_NUM (h[1], h[2]);
167 /* The core of url_escape_* functions. Escapes the characters that
168 match the provided mask in urlchr_table.
170 If ALLOW_PASSTHROUGH is non-zero, a string with no unsafe chars
171 will be returned unchanged. If ALLOW_PASSTHROUGH is zero, a
172 freshly allocated string will be returned in all cases. */
175 url_escape_1 (const char *s, unsigned char mask, int allow_passthrough)
182 for (p1 = s; *p1; p1++)
183 if (urlchr_test (*p1, mask))
184 addition += 2; /* Two more characters (hex digits) */
187 return allow_passthrough ? (char *)s : xstrdup (s);
189 newlen = (p1 - s) + addition;
190 newstr = (char *)xmalloc (newlen + 1);
196 /* Quote the characters that match the test mask. */
197 if (urlchr_test (*p1, mask))
199 unsigned char c = *p1++;
201 *p2++ = XNUM_TO_DIGIT (c >> 4);
202 *p2++ = XNUM_TO_DIGIT (c & 0xf);
207 assert (p2 - newstr == newlen);
213 /* URL-escape the unsafe characters (see urlchr_table) in a given
214 string, returning a freshly allocated string. */
217 url_escape (const char *s)
219 return url_escape_1 (s, urlchr_unsafe, 0);
222 /* URL-escape the unsafe characters (see urlchr_table) in a given
223 string. If no characters are unsafe, S is returned. */
226 url_escape_allow_passthrough (const char *s)
228 return url_escape_1 (s, urlchr_unsafe, 1);
231 enum copy_method { CM_DECODE, CM_ENCODE, CM_PASSTHROUGH };
233 /* Decide whether to encode, decode, or pass through the char at P.
234 This used to be a macro, but it got a little too convoluted. */
235 static inline enum copy_method
236 decide_copy_method (const char *p)
240 if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2)))
242 /* %xx sequence: decode it, unless it would decode to an
243 unsafe or a reserved char; in that case, leave it as
245 char preempt = X2DIGITS_TO_NUM (*(p + 1), *(p + 2));
246 if (URL_UNSAFE_CHAR (preempt) || URL_RESERVED_CHAR (preempt))
247 return CM_PASSTHROUGH;
252 /* Garbled %.. sequence: encode `%'. */
255 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
258 return CM_PASSTHROUGH;
261 /* Translate a %-escaped (but possibly non-conformant) input string S
262 into a %-escaped (and conformant) output string. If no characters
263 are encoded or decoded, return the same string S; otherwise, return
264 a freshly allocated string with the new contents.
266 After a URL has been run through this function, the protocols that
267 use `%' as the quote character can use the resulting string as-is,
268 while those that don't call url_unescape() to get to the intended
269 data. This function is also stable: after an input string is
270 transformed the first time, all further transformations of the
271 result yield the same result string.
273 Let's discuss why this function is needed.
275 Imagine Wget is to retrieve `http://abc.xyz/abc def'. Since a raw
276 space character would mess up the HTTP request, it needs to be
279 GET /abc%20def HTTP/1.0
281 It appears that the unsafe chars need to be quoted, for example
282 with url_escape. But what if we're requested to download
283 `abc%20def'? url_escape transforms "%" to "%25", which would leave
284 us with `abc%2520def'. This is incorrect -- since %-escapes are
285 part of URL syntax, "%20" is the correct way to denote a literal
286 space on the Wget command line. This leaves us in the conclusion
287 that in that case Wget should not call url_escape, but leave the
290 And what if the requested URI is `abc%20 def'? If we call
291 url_escape, we end up with `/abc%2520%20def', which is almost
292 certainly not intended. If we don't call url_escape, we are left
293 with the embedded space and cannot complete the request. What the
294 user meant was for Wget to request `/abc%20%20def', and this is
295 where reencode_escapes kicks in.
297 Wget used to solve this by first decoding %-quotes, and then
298 encoding all the "unsafe" characters found in the resulting string.
299 This was wrong because it didn't preserve certain URL special
300 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
301 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
302 whether we considered `+' reserved (it is). One of these results
303 is inevitable because by the second step we would lose information
304 on whether the `+' was originally encoded or not. Both results
305 were wrong because in CGI parameters + means space, while %2B means
306 literal plus. reencode_escapes correctly translates the above to
307 "a%2B+b", i.e. returns the original string.
309 This function uses an algorithm proposed by Anon Sricharoenchai:
311 1. Encode all URL_UNSAFE and the "%" that are not followed by 2
314 2. Decode all "%XX" except URL_UNSAFE, URL_RESERVED (";/?:@=&") and
317 ...except that this code conflates the two steps, and decides
318 whether to encode, decode, or pass through each character in turn.
319 The function still uses two passes, but their logic is the same --
320 the first pass exists merely for the sake of allocation. Another
321 small difference is that we include `+' to URL_RESERVED.
325 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
327 "http://abc.xyz/%20%3F%2561%25aa%25%20a?a=a+a%2Ba&b=b%26c%3Dc"
331 "foo bar" -> "foo%20bar"
332 "foo%20bar" -> "foo%20bar"
333 "foo %20bar" -> "foo%20%20bar"
334 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
335 "foo%25%20bar" -> "foo%25%20bar"
336 "foo%2%20bar" -> "foo%252%20bar"
337 "foo+bar" -> "foo+bar" (plus is reserved!)
338 "foo%2b+bar" -> "foo%2b+bar" */
341 reencode_escapes (const char *s)
347 int encode_count = 0;
348 int decode_count = 0;
350 /* First, pass through the string to see if there's anything to do,
351 and to calculate the new length. */
352 for (p1 = s; *p1; p1++)
354 switch (decide_copy_method (p1))
367 if (!encode_count && !decode_count)
368 /* The string is good as it is. */
369 return (char *)s; /* C const model sucks. */
372 /* Each encoding adds two characters (hex digits), while each
373 decoding removes two characters. */
374 newlen = oldlen + 2 * (encode_count - decode_count);
375 newstr = xmalloc (newlen + 1);
382 switch (decide_copy_method (p1))
386 unsigned char c = *p1++;
388 *p2++ = XNUM_TO_DIGIT (c >> 4);
389 *p2++ = XNUM_TO_DIGIT (c & 0xf);
393 *p2++ = X2DIGITS_TO_NUM (p1[1], p1[2]);
394 p1 += 3; /* skip %xx */
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 here. The function
461 should *not* be called with the complete URL, but with the part
462 right after the scheme.
464 If no username and password are found, return 0. */
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_has_scheme (url))
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 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")
639 /* The following two functions were adapted from glibc. */
642 is_valid_ipv4_address (const char *str, const char *end)
652 if (ch >= '0' && ch <= '9')
654 val = val * 10 + (ch - '0');
665 else if (ch == '.' && saw_digit == 1)
682 is_valid_ipv6_address (const char *str, const char *end)
702 /* Leading :: requires some special handling. */
706 if (str == end || *str != ':')
718 /* if ch is a number, add it to val. */
722 val |= XDIGIT_TO_NUM (ch);
729 /* if ch is a colon ... */
742 if (tp > NS_IN6ADDRSZ - NS_INT16SZ)
750 /* if ch is a dot ... */
751 if (ch == '.' && (tp <= NS_IN6ADDRSZ - NS_INADDRSZ)
752 && is_valid_ipv4_address (curtok, end) == 1)
764 if (tp > NS_IN6ADDRSZ - NS_INT16SZ)
771 if (tp == NS_IN6ADDRSZ)
776 if (tp != NS_IN6ADDRSZ)
785 Return a new struct url if successful, NULL on error. In case of
786 error, and if ERROR is not NULL, also set *ERROR to the appropriate
789 url_parse (const char *url, int *error)
793 int path_modified, host_modified;
795 enum url_scheme scheme;
797 const char *uname_b, *uname_e;
798 const char *host_b, *host_e;
799 const char *path_b, *path_e;
800 const char *params_b, *params_e;
801 const char *query_b, *query_e;
802 const char *fragment_b, *fragment_e;
805 char *user = NULL, *passwd = NULL;
807 char *url_encoded = NULL;
811 scheme = url_scheme (url);
812 if (scheme == SCHEME_INVALID)
814 error_code = PE_UNSUPPORTED_SCHEME;
818 url_encoded = reencode_escapes (url);
821 p += strlen (supported_schemes[scheme].leading_string);
823 p += url_skip_credentials (p);
826 /* scheme://user:pass@host[:port]... */
829 /* We attempt to break down the URL into the components path,
830 params, query, and fragment. They are ordered like this:
832 scheme://host[:port][/path][;params][?query][#fragment] */
834 params_b = params_e = NULL;
835 query_b = query_e = NULL;
836 fragment_b = fragment_e = NULL;
842 /* Handle IPv6 address inside square brackets. Ideally we'd
843 just look for the terminating ']', but rfc2732 mandates
844 rejecting invalid IPv6 addresses. */
846 /* The address begins after '['. */
848 host_e = strchr (host_b, ']');
852 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
857 /* Check if the IPv6 address is valid. */
858 if (!is_valid_ipv6_address(host_b, host_e))
860 error_code = PE_INVALID_IPV6_ADDRESS;
864 /* Continue parsing after the closing ']'. */
867 error_code = PE_IPV6_NOT_SUPPORTED;
873 p = strpbrk_or_eos (p, ":/;?#");
877 if (host_b == host_e)
879 error_code = PE_EMPTY_HOST;
883 port = scheme_default_port (scheme);
886 const char *port_b, *port_e, *pp;
888 /* scheme://host:port/tralala */
892 p = strpbrk_or_eos (p, "/;?#");
895 if (port_b == port_e)
897 /* http://host:/whatever */
899 error_code = PE_BAD_PORT_NUMBER;
903 for (port = 0, pp = port_b; pp < port_e; pp++)
907 /* http://host:12randomgarbage/blah */
909 error_code = PE_BAD_PORT_NUMBER;
913 port = 10 * port + (*pp - '0');
921 p = strpbrk_or_eos (p, ";?#");
926 /* Path is not allowed not to exist. */
934 p = strpbrk_or_eos (p, "?#");
941 p = strpbrk_or_eos (p, "#");
944 /* Hack that allows users to use '?' (a wildcard character) in
945 FTP URLs without it being interpreted as a query string
947 if (scheme == SCHEME_FTP)
949 query_b = query_e = NULL;
962 if (uname_b != uname_e)
964 /* http://user:pass@host */
966 /* uname_b uname_e */
967 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
969 error_code = PE_INVALID_USER_NAME;
974 u = xnew0 (struct url);
976 u->host = strdupdelim (host_b, host_e);
981 u->path = strdupdelim (path_b, path_e);
982 path_modified = path_simplify (u->path);
983 split_path (u->path, &u->dir, &u->file);
985 host_modified = lowercase_str (u->host);
988 u->params = strdupdelim (params_b, params_e);
990 u->query = strdupdelim (query_b, query_e);
992 u->fragment = strdupdelim (fragment_b, fragment_e);
994 if (path_modified || u->fragment || host_modified || path_b == path_e)
996 /* If we suspect that a transformation has rendered what
997 url_string might return different from URL_ENCODED, rebuild
998 u->url using url_string. */
999 u->url = url_string (u, 0);
1001 if (url_encoded != url)
1002 xfree ((char *) url_encoded);
1006 if (url_encoded == url)
1007 u->url = xstrdup (url);
1009 u->url = url_encoded;
1016 /* Cleanup in case of error: */
1017 if (url_encoded && url_encoded != url)
1018 xfree (url_encoded);
1020 /* Transmit the error code to the caller, if the caller wants to
1023 *error = error_code;
1027 /* Return the error message string from ERROR_CODE, which should have
1028 been retrieved from url_parse. The error message is translated. */
1031 url_error (int error_code)
1033 assert (error_code >= 0 && error_code < countof (parse_errors));
1034 return _(parse_errors[error_code]);
1037 /* Split PATH into DIR and FILE. PATH comes from the URL and is
1038 expected to be URL-escaped.
1040 The path is split into directory (the part up to the last slash)
1041 and file (the part after the last slash), which are subsequently
1042 unescaped. Examples:
1045 "foo/bar/baz" "foo/bar" "baz"
1046 "foo/bar/" "foo/bar" ""
1048 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
1050 DIR and FILE are freshly allocated. */
1053 split_path (const char *path, char **dir, char **file)
1055 char *last_slash = strrchr (path, '/');
1058 *dir = xstrdup ("");
1059 *file = xstrdup (path);
1063 *dir = strdupdelim (path, last_slash);
1064 *file = xstrdup (last_slash + 1);
1066 url_unescape (*dir);
1067 url_unescape (*file);
1070 /* Note: URL's "full path" is the path with the query string and
1071 params appended. The "fragment" (#foo) is intentionally ignored,
1072 but that might be changed. For example, if the original URL was
1073 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
1074 the full path will be "/foo/bar/baz;bullshit?querystring". */
1076 /* Return the length of the full path, without the terminating
1080 full_path_length (const struct url *url)
1084 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
1095 /* Write out the full path. */
1098 full_path_write (const struct url *url, char *where)
1100 #define FROB(el, chr) do { \
1101 char *f_el = url->el; \
1103 int l = strlen (f_el); \
1105 memcpy (where, f_el, l); \
1117 /* Public function for getting the "full path". E.g. if u->path is
1118 "foo/bar" and u->query is "param=value", full_path will be
1119 "/foo/bar?param=value". */
1122 url_full_path (const struct url *url)
1124 int length = full_path_length (url);
1125 char *full_path = (char *)xmalloc(length + 1);
1127 full_path_write (url, full_path);
1128 full_path[length] = '\0';
1133 /* Escape unsafe and reserved characters, except for the slash
1137 url_escape_dir (const char *dir)
1139 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1144 /* Unescape slashes in NEWDIR. */
1146 h = newdir; /* hare */
1147 t = newdir; /* tortoise */
1149 for (; *h; h++, t++)
1151 /* url_escape_1 having converted '/' to "%2F" exactly. */
1152 if (*h == '%' && h[1] == '2' && h[2] == 'F')
1165 /* Sync u->path and u->url with u->dir and u->file. Called after
1166 u->file or u->dir have been changed, typically by the FTP code. */
1169 sync_path (struct url *u)
1171 char *newpath, *efile, *edir;
1175 /* u->dir and u->file are not escaped. URL-escape them before
1176 reassembling them into u->path. That way, if they contain
1177 separators like '?' or even if u->file contains slashes, the
1178 path will be correctly assembled. (u->file can contain slashes
1179 if the URL specifies it with %2f, or if an FTP server returns
1181 edir = url_escape_dir (u->dir);
1182 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1185 newpath = xstrdup (efile);
1188 int dirlen = strlen (edir);
1189 int filelen = strlen (efile);
1191 /* Copy "DIR/FILE" to newpath. */
1192 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1193 memcpy (p, edir, dirlen);
1196 memcpy (p, efile, filelen);
1205 if (efile != u->file)
1208 /* Regenerate u->url as well. */
1210 u->url = url_string (u, 0);
1213 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1214 This way we can sync u->path and u->url when they get changed. */
1217 url_set_dir (struct url *url, const char *newdir)
1220 url->dir = xstrdup (newdir);
1225 url_set_file (struct url *url, const char *newfile)
1228 url->file = xstrdup (newfile);
1233 url_free (struct url *url)
1239 xfree_null (url->params);
1240 xfree_null (url->query);
1241 xfree_null (url->fragment);
1242 xfree_null (url->user);
1243 xfree_null (url->passwd);
1251 /* Create all the necessary directories for PATH (a file). Calls
1252 mkdirhier() internally. */
1254 mkalldirs (const char *path)
1261 p = path + strlen (path);
1262 for (; *p != '/' && p != path; p--)
1265 /* Don't create if it's just a file. */
1266 if ((p == path) && (*p != '/'))
1268 t = strdupdelim (path, p);
1270 /* Check whether the directory exists. */
1271 if ((stat (t, &st) == 0))
1273 if (S_ISDIR (st.st_mode))
1280 /* If the dir exists as a file name, remove it first. This
1281 is *only* for Wget to work with buggy old CERN http
1282 servers. Here is the scenario: When Wget tries to
1283 retrieve a directory without a slash, e.g.
1284 http://foo/bar (bar being a directory), CERN server will
1285 not redirect it too http://foo/bar/ -- it will generate a
1286 directory listing containing links to bar/file1,
1287 bar/file2, etc. Wget will lose because it saves this
1288 HTML listing to a file `bar', so it cannot create the
1289 directory. To work around this, if the file of the same
1290 name exists, we just remove it and create the directory
1292 DEBUGP (("Removing %s because of directory danger!\n", t));
1296 res = make_directory (t);
1298 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1303 /* Functions for constructing the file name out of URL components. */
1305 /* A growable string structure, used by url_file_name and friends.
1306 This should perhaps be moved to utils.c.
1308 The idea is to have a convenient and efficient way to construct a
1309 string by having various functions append data to it. Instead of
1310 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1311 functions in questions, we pass the pointer to this struct. */
1319 /* Ensure that the string can accept APPEND_COUNT more characters past
1320 the current TAIL position. If necessary, this will grow the string
1321 and update its allocated size. If the string is already large
1322 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1323 #define GROW(g, append_size) do { \
1324 struct growable *G_ = g; \
1325 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1328 /* Return the tail position of the string. */
1329 #define TAIL(r) ((r)->base + (r)->tail)
1331 /* Move the tail position by APPEND_COUNT characters. */
1332 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1334 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1338 append_string (const char *str, struct growable *dest)
1340 int l = strlen (str);
1342 memcpy (TAIL (dest), str, l);
1343 TAIL_INCR (dest, l);
1346 /* Append CH to DEST. For example, append_char (0, DEST)
1347 zero-terminates DEST. */
1350 append_char (char ch, struct growable *dest)
1354 TAIL_INCR (dest, 1);
1358 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1359 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1360 filechr_control = 4 /* a control character, e.g. 0-31 */
1363 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1365 /* Shorthands for the table: */
1366 #define U filechr_not_unix
1367 #define W filechr_not_windows
1368 #define C filechr_control
1373 /* Table of characters unsafe under various conditions (see above).
1375 Arguably we could also claim `%' to be unsafe, since we use it as
1376 the escape character. If we ever want to be able to reliably
1377 translate file name back to URL, this would become important
1378 crucial. Right now, it's better to be minimal in escaping. */
1380 const static unsigned char filechr_table[256] =
1382 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1383 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1384 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1385 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1386 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1387 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1388 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1389 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1390 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1391 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1392 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1393 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1394 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1395 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1396 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1397 0, 0, 0, 0, 0, 0, 0, 0, /* x y z { | } ~ DEL */
1399 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1400 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1401 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1402 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1404 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1405 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1406 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1407 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1415 /* FN_PORT_SEP is the separator between host and port in file names
1416 for non-standard port numbers. On Unix this is normally ':', as in
1417 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1418 because Windows can't handle ':' in file names. */
1419 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1421 /* FN_QUERY_SEP is the separator between the file name and the URL
1422 query, normally '?'. Since Windows cannot handle '?' as part of
1423 file name, we use '@' instead there. */
1424 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1426 /* Quote path element, characters in [b, e), as file name, and append
1427 the quoted string to DEST. Each character is quoted as per
1428 file_unsafe_char and the corresponding table.
1430 If ESCAPED_P is non-zero, the path element is considered to be
1431 URL-escaped and will be unescaped prior to inspection. */
1434 append_uri_pathel (const char *b, const char *e, int escaped_p,
1435 struct growable *dest)
1441 if (opt.restrict_files_os == restrict_unix)
1442 mask = filechr_not_unix;
1444 mask = filechr_not_windows;
1445 if (opt.restrict_files_ctrl)
1446 mask |= filechr_control;
1448 /* Copy [b, e) to PATHEL and URL-unescape it. */
1452 BOUNDED_TO_ALLOCA (b, e, unescaped);
1453 url_unescape (unescaped);
1455 e = unescaped + strlen (unescaped);
1458 /* Walk the PATHEL string and check how many characters we'll need
1459 to add for file quoting. */
1461 for (p = b; p < e; p++)
1462 if (FILE_CHAR_TEST (*p, mask))
1465 /* e-b is the string length. Each quoted char means two additional
1466 characters in the string, hence 2*quoted. */
1467 outlen = (e - b) + (2 * quoted);
1468 GROW (dest, outlen);
1472 /* If there's nothing to quote, we don't need to go through the
1473 string the second time. */
1474 memcpy (TAIL (dest), b, outlen);
1478 char *q = TAIL (dest);
1479 for (p = b; p < e; p++)
1481 if (!FILE_CHAR_TEST (*p, mask))
1485 unsigned char ch = *p;
1487 *q++ = XNUM_TO_DIGIT (ch >> 4);
1488 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1491 assert (q - TAIL (dest) == outlen);
1493 TAIL_INCR (dest, outlen);
1496 /* Append to DEST the directory structure that corresponds the
1497 directory part of URL's path. For example, if the URL is
1498 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1500 Each path element ("dir1" and "dir2" in the above example) is
1501 examined, url-unescaped, and re-escaped as file name element.
1503 Additionally, it cuts as many directories from the path as
1504 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1505 will produce "bar" for the above example. For 2 or more, it will
1508 Each component of the path is quoted for use as file name. */
1511 append_dir_structure (const struct url *u, struct growable *dest)
1513 char *pathel, *next;
1514 int cut = opt.cut_dirs;
1516 /* Go through the path components, de-URL-quote them, and quote them
1517 (if necessary) as file names. */
1520 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1525 /* Ignore empty pathels. */
1529 append_char ('/', dest);
1530 append_uri_pathel (pathel, next, 1, dest);
1534 /* Return a unique file name that matches the given URL as good as
1535 possible. Does not create directories on the file system. */
1538 url_file_name (const struct url *u)
1540 struct growable fnres;
1542 char *u_file, *u_query;
1543 char *fname, *unique;
1549 /* Start with the directory prefix, if specified. */
1551 append_string (opt.dir_prefix, &fnres);
1553 /* If "dirstruct" is turned on (typically the case with -r), add
1554 the host and port (unless those have been turned off) and
1555 directory structure. */
1558 if (opt.add_hostdir)
1561 append_char ('/', &fnres);
1562 append_string (u->host, &fnres);
1563 if (u->port != scheme_default_port (u->scheme))
1566 number_to_string (portstr, u->port);
1567 append_char (FN_PORT_SEP, &fnres);
1568 append_string (portstr, &fnres);
1572 append_dir_structure (u, &fnres);
1575 /* Add the file name. */
1577 append_char ('/', &fnres);
1578 u_file = *u->file ? u->file : "index.html";
1579 append_uri_pathel (u_file, u_file + strlen (u_file), 0, &fnres);
1581 /* Append "?query" to the file name. */
1582 u_query = u->query && *u->query ? u->query : NULL;
1585 append_char (FN_QUERY_SEP, &fnres);
1586 append_uri_pathel (u_query, u_query + strlen (u_query), 1, &fnres);
1589 /* Zero-terminate the file name. */
1590 append_char ('\0', &fnres);
1594 /* Check the cases in which the unique extensions are not used:
1595 1) Clobbering is turned off (-nc).
1596 2) Retrieval with regetting.
1597 3) Timestamping is used.
1598 4) Hierarchy is built.
1600 The exception is the case when file does exist and is a
1601 directory (see `mkalldirs' for explanation). */
1603 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1604 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1607 unique = unique_name (fname, 1);
1608 if (unique != fname)
1613 /* Return the length of URL's path. Path is considered to be
1614 terminated by one of '?', ';', '#', or by the end of the
1617 path_length (const char *url)
1619 const char *q = strpbrk_or_eos (url, "?;#");
1623 /* Find the last occurrence of character C in the range [b, e), or
1624 NULL, if none are present. This is equivalent to strrchr(b, c),
1625 except that it accepts an END argument instead of requiring the
1626 string to be zero-terminated. Why is there no memrchr()? */
1628 find_last_char (const char *b, const char *e, char c)
1636 /* Resolve "." and ".." elements of PATH by destructively modifying
1637 PATH and return non-zero if PATH has been modified, zero otherwise.
1639 The algorithm is in spirit similar to the one described in rfc1808,
1640 although implemented differently, in one pass. To recap, path
1641 elements containing only "." are removed, and ".." is taken to mean
1642 "back up one element". Single leading and trailing slashes are
1645 This function does not handle URL escapes explicitly. If you're
1646 passing paths from URLs, make sure to unquote "%2e" and "%2E" to
1647 ".", so that this function can find the dots. (Wget's URL parser
1648 calls reencode_escapes, which see.)
1650 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1651 test examples are provided below. If you change anything in this
1652 function, run test_path_simplify to make sure you haven't broken a
1656 path_simplify (char *path)
1660 /* Preserve the leading '/'. */
1664 h = path; /* hare */
1665 t = path; /* tortoise */
1666 end = path + strlen (path);
1670 /* Hare should be at the beginning of a path element. */
1672 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1677 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1679 /* Handle "../" by retreating the tortoise by one path
1680 element -- but not past beggining of PATH. */
1683 /* Move backwards until T hits the beginning of the
1684 previous path element or the beginning of path. */
1685 for (--t; t > path && t[-1] != '/'; t--)
1692 /* Ignore empty path elements. Supporting them well is hard
1693 (where do you save "http://x.com///y.html"?), and they
1694 don't bring any practical gain. Plus, they break our
1695 filesystem-influenced assumptions: allowing them would
1696 make "x/y//../z" simplify to "x/y/z", whereas most people
1697 would expect "x/z". */
1702 /* A regular path element. If H hasn't advanced past T,
1703 simply skip to the next path element. Otherwise, copy
1704 the path element until the next slash. */
1707 /* Skip the path element, including the slash. */
1708 while (h < end && *h != '/')
1715 /* Copy the path element, including the final slash. */
1716 while (h < end && *h != '/')
1730 /* Merge BASE with LINK and return the resulting URI.
1732 Either of the URIs may be absolute or relative, complete with the
1733 host name, or path only. This tries to reasonably handle all
1734 foreseeable cases. It only employs minimal URL parsing, without
1735 knowledge of the specifics of schemes.
1737 Perhaps this function should call path_simplify so that the callers
1738 don't have to call url_parse unconditionally. */
1741 uri_merge (const char *base, const char *link)
1747 if (url_has_scheme (link))
1748 return xstrdup (link);
1750 /* We may not examine BASE past END. */
1751 end = base + path_length (base);
1752 linklength = strlen (link);
1756 /* Empty LINK points back to BASE, query string and all. */
1757 return xstrdup (base);
1759 else if (*link == '?')
1761 /* LINK points to the same location, but changes the query
1762 string. Examples: */
1763 /* uri_merge("path", "?new") -> "path?new" */
1764 /* uri_merge("path?foo", "?new") -> "path?new" */
1765 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1766 /* uri_merge("path#foo", "?new") -> "path?new" */
1767 int baselength = end - base;
1768 merge = xmalloc (baselength + linklength + 1);
1769 memcpy (merge, base, baselength);
1770 memcpy (merge + baselength, link, linklength);
1771 merge[baselength + linklength] = '\0';
1773 else if (*link == '#')
1775 /* uri_merge("path", "#new") -> "path#new" */
1776 /* uri_merge("path#foo", "#new") -> "path#new" */
1777 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1778 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1780 const char *end1 = strchr (base, '#');
1782 end1 = base + strlen (base);
1783 baselength = end1 - base;
1784 merge = xmalloc (baselength + linklength + 1);
1785 memcpy (merge, base, baselength);
1786 memcpy (merge + baselength, link, linklength);
1787 merge[baselength + linklength] = '\0';
1789 else if (*link == '/' && *(link + 1) == '/')
1791 /* LINK begins with "//" and so is a net path: we need to
1792 replace everything after (and including) the double slash
1795 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1796 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1797 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1801 const char *start_insert;
1803 /* Look for first slash. */
1804 slash = memchr (base, '/', end - base);
1805 /* If found slash and it is a double slash, then replace
1806 from this point, else default to replacing from the
1808 if (slash && *(slash + 1) == '/')
1809 start_insert = slash;
1811 start_insert = base;
1813 span = start_insert - base;
1814 merge = (char *)xmalloc (span + linklength + 1);
1816 memcpy (merge, base, span);
1817 memcpy (merge + span, link, linklength);
1818 merge[span + linklength] = '\0';
1820 else if (*link == '/')
1822 /* LINK is an absolute path: we need to replace everything
1823 after (and including) the FIRST slash with LINK.
1825 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1826 "/qux/xyzzy", our result should be
1827 "http://host/qux/xyzzy". */
1830 const char *start_insert = NULL; /* for gcc to shut up. */
1831 const char *pos = base;
1832 int seen_slash_slash = 0;
1833 /* We're looking for the first slash, but want to ignore
1836 slash = memchr (pos, '/', end - pos);
1837 if (slash && !seen_slash_slash)
1838 if (*(slash + 1) == '/')
1841 seen_slash_slash = 1;
1845 /* At this point, SLASH is the location of the first / after
1846 "//", or the first slash altogether. START_INSERT is the
1847 pointer to the location where LINK will be inserted. When
1848 examining the last two examples, keep in mind that LINK
1851 if (!slash && !seen_slash_slash)
1852 /* example: "foo" */
1854 start_insert = base;
1855 else if (!slash && seen_slash_slash)
1856 /* example: "http://foo" */
1859 else if (slash && !seen_slash_slash)
1860 /* example: "foo/bar" */
1862 start_insert = base;
1863 else if (slash && seen_slash_slash)
1864 /* example: "http://something/" */
1866 start_insert = slash;
1868 span = start_insert - base;
1869 merge = (char *)xmalloc (span + linklength + 1);
1871 memcpy (merge, base, span);
1872 memcpy (merge + span, link, linklength);
1873 merge[span + linklength] = '\0';
1877 /* LINK is a relative URL: we need to replace everything
1878 after last slash (possibly empty) with LINK.
1880 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1881 our result should be "whatever/foo/qux/xyzzy". */
1882 int need_explicit_slash = 0;
1884 const char *start_insert;
1885 const char *last_slash = find_last_char (base, end, '/');
1888 /* No slash found at all. Append LINK to what we have,
1889 but we'll need a slash as a separator.
1891 Example: if base == "foo" and link == "qux/xyzzy", then
1892 we cannot just append link to base, because we'd get
1893 "fooqux/xyzzy", whereas what we want is
1896 To make sure the / gets inserted, we set
1897 need_explicit_slash to 1. We also set start_insert
1898 to end + 1, so that the length calculations work out
1899 correctly for one more (slash) character. Accessing
1900 that character is fine, since it will be the
1901 delimiter, '\0' or '?'. */
1902 /* example: "foo?..." */
1903 /* ^ ('?' gets changed to '/') */
1904 start_insert = end + 1;
1905 need_explicit_slash = 1;
1907 else if (last_slash && last_slash >= base + 2
1908 && last_slash[-2] == ':' && last_slash[-1] == '/')
1910 /* example: http://host" */
1912 start_insert = end + 1;
1913 need_explicit_slash = 1;
1917 /* example: "whatever/foo/bar" */
1919 start_insert = last_slash + 1;
1922 span = start_insert - base;
1923 merge = (char *)xmalloc (span + linklength + 1);
1925 memcpy (merge, base, span);
1926 if (need_explicit_slash)
1927 merge[span - 1] = '/';
1928 memcpy (merge + span, link, linklength);
1929 merge[span + linklength] = '\0';
1935 #define APPEND(p, s) do { \
1936 int len = strlen (s); \
1937 memcpy (p, s, len); \
1941 /* Use this instead of password when the actual password is supposed
1942 to be hidden. We intentionally use a generic string without giving
1943 away the number of characters in the password, like previous
1945 #define HIDDEN_PASSWORD "*password*"
1947 /* Recreate the URL string from the data in URL.
1949 If HIDE is non-zero (as it is when we're calling this on a URL we
1950 plan to print, but not when calling it to canonicalize a URL for
1951 use within the program), password will be hidden. Unsafe
1952 characters in the URL will be quoted. */
1955 url_string (const struct url *url, int hide_password)
1959 char *quoted_user = NULL, *quoted_passwd = NULL;
1961 int scheme_port = supported_schemes[url->scheme].default_port;
1962 char *scheme_str = supported_schemes[url->scheme].leading_string;
1963 int fplen = full_path_length (url);
1965 int brackets_around_host = 0;
1967 assert (scheme_str != NULL);
1969 /* Make sure the user name and password are quoted. */
1972 quoted_user = url_escape_allow_passthrough (url->user);
1976 quoted_passwd = HIDDEN_PASSWORD;
1978 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1982 if (strchr (url->host, ':'))
1983 brackets_around_host = 1;
1985 size = (strlen (scheme_str)
1986 + strlen (url->host)
1987 + (brackets_around_host ? 2 : 0)
1990 if (url->port != scheme_port)
1991 size += 1 + numdigit (url->port);
1994 size += 1 + strlen (quoted_user);
1996 size += 1 + strlen (quoted_passwd);
1999 p = result = xmalloc (size);
2001 APPEND (p, scheme_str);
2004 APPEND (p, quoted_user);
2008 APPEND (p, quoted_passwd);
2013 if (brackets_around_host)
2015 APPEND (p, url->host);
2016 if (brackets_around_host)
2018 if (url->port != scheme_port)
2021 p = number_to_string (p, url->port);
2024 full_path_write (url, p);
2028 assert (p - result == size);
2030 if (quoted_user && quoted_user != url->user)
2031 xfree (quoted_user);
2032 if (quoted_passwd && !hide_password
2033 && quoted_passwd != url->passwd)
2034 xfree (quoted_passwd);
2039 /* Return non-zero if scheme a is similar to scheme b.
2041 Schemes are similar if they are equal. If SSL is supported, schemes
2042 are also similar if one is http (SCHEME_HTTP) and the other is https
2045 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
2050 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
2051 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
2058 /* Debugging and testing support for path_simplify. */
2060 /* Debug: run path_simplify on PATH and return the result in a new
2061 string. Useful for calling from the debugger. */
2065 char *copy = xstrdup (path);
2066 path_simplify (copy);
2071 run_test (char *test, char *expected_result, int expected_change)
2073 char *test_copy = xstrdup (test);
2074 int modified = path_simplify (test_copy);
2076 if (0 != strcmp (test_copy, expected_result))
2078 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2079 test, expected_result, test_copy);
2081 if (modified != expected_change)
2083 if (expected_change == 1)
2084 printf ("Expected no modification with path_simplify(\"%s\").\n",
2087 printf ("Expected modification with path_simplify(\"%s\").\n",
2094 test_path_simplify (void)
2097 char *test, *result;
2103 { "foo", "foo", 0 },
2104 { "foo/bar", "foo/bar", 0 },
2105 { "foo///bar", "foo/bar", 1 },
2106 { "foo/.", "foo/", 1 },
2107 { "foo/./", "foo/", 1 },
2108 { "foo./", "foo./", 0 },
2109 { "foo/../bar", "bar", 1 },
2110 { "foo/../bar/", "bar/", 1 },
2111 { "foo/bar/..", "foo/", 1 },
2112 { "foo/bar/../x", "foo/x", 1 },
2113 { "foo/bar/../x/", "foo/x/", 1 },
2114 { "foo/..", "", 1 },
2115 { "foo/../..", "", 1 },
2116 { "a/b/../../c", "c", 1 },
2117 { "./a/../b", "b", 1 }
2121 for (i = 0; i < countof (tests); i++)
2123 char *test = tests[i].test;
2124 char *expected_result = tests[i].result;
2125 int expected_change = tests[i].should_modify;
2126 run_test (test, expected_result, expected_change);
2129 /* Now run all the tests with a leading slash before the test case,
2130 to prove that the slash is being preserved. */
2131 for (i = 0; i < countof (tests); i++)
2133 char *test, *expected_result;
2134 int expected_change = tests[i].should_modify;
2136 test = xmalloc (1 + strlen (tests[i].test) + 1);
2137 sprintf (test, "/%s", tests[i].test);
2139 expected_result = xmalloc (1 + strlen (tests[i].result) + 1);
2140 sprintf (expected_result, "/%s", tests[i].result);
2142 run_test (test, expected_result, expected_change);
2145 xfree (expected_result);