2 Copyright (C) 1996-2005 Free Software Foundation, Inc.
4 This file is part of GNU Wget.
6 GNU Wget is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or (at
9 your option) any later version.
11 GNU Wget is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with Wget; if not, write to the Free Software Foundation, Inc.,
18 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
20 In addition, as a special exception, the Free Software Foundation
21 gives permission to link the code of its release of Wget with the
22 OpenSSL project's "OpenSSL" library (or with modified versions of it
23 that use the same license as the "OpenSSL" library), and distribute
24 the linked executables. You must obey the GNU General Public License
25 in all respects for all of the code used other than "OpenSSL". If you
26 modify this file, you may extend this exception to your version of the
27 file, but you are not obligated to do so. If you do not wish to do
28 so, delete this exception statement from your version. */
44 #include "host.h" /* for is_valid_ipv6_address */
47 scm_disabled = 1, /* for https when OpenSSL fails to init. */
48 scm_has_params = 2, /* whether scheme has ;params */
49 scm_has_query = 4, /* whether scheme has ?query */
50 scm_has_fragment = 8 /* whether scheme has #fragment */
55 /* Short name of the scheme, such as "http" or "ftp". */
57 /* Leading string that identifies the scheme, such as "https://". */
58 const char *leading_string;
59 /* Default port of the scheme when none is specified. */
65 /* Supported schemes: */
66 static struct scheme_data supported_schemes[] =
68 { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment },
70 { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment },
72 { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment },
78 /* Forward declarations: */
80 static bool path_simplify (char *);
82 /* Support for escaping and unescaping of URL strings. */
84 /* Table of "reserved" and "unsafe" characters. Those terms are
85 rfc1738-speak, as such largely obsoleted by rfc2396 and later
86 specs, but the general idea remains.
88 A reserved character is the one that you can't decode without
89 changing the meaning of the URL. For example, you can't decode
90 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
91 path components is different. Non-reserved characters can be
92 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
93 unsafe characters are loosely based on rfc1738, plus "$" and ",",
94 as recommended by rfc2396, and minus "~", which is very frequently
95 used (and sometimes unrecognized as %7E by broken servers).
97 An unsafe character is the one that should be encoded when URLs are
98 placed in foreign environments. E.g. space and newline are unsafe
99 in HTTP contexts because HTTP uses them as separator and line
100 terminator, so they must be encoded to %20 and %0A respectively.
101 "*" is unsafe in shell context, etc.
103 We determine whether a character is unsafe through static table
104 lookup. This code assumes ASCII character set and 8-bit chars. */
107 /* rfc1738 reserved chars + "$" and ",". */
110 /* rfc1738 unsafe chars, plus non-printables. */
114 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
115 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
116 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
118 /* Shorthands for the table: */
119 #define R urlchr_reserved
120 #define U urlchr_unsafe
123 static const unsigned char urlchr_table[256] =
125 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
126 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
127 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
128 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
129 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
130 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
131 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
132 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
133 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
134 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
135 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
136 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
137 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
138 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
139 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
140 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
142 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
143 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
144 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
145 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
147 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
148 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
149 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
150 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
156 /* URL-unescape the string S.
158 This is done by transforming the sequences "%HH" to the character
159 represented by the hexadecimal digits HH. If % is not followed by
160 two hexadecimal digits, it is inserted literally.
162 The transformation is done in place. If you need the original
163 string intact, make a copy before calling this function. */
166 url_unescape (char *s)
168 char *t = s; /* t - tortoise */
169 char *h = s; /* h - hare */
181 /* Do nothing if '%' is not followed by two hex digits. */
182 if (!h[1] || !h[2] || !(ISXDIGIT (h[1]) && ISXDIGIT (h[2])))
184 c = X2DIGITS_TO_NUM (h[1], h[2]);
185 /* Don't unescape %00 because there is no way to insert it
186 into a C string without effectively truncating it. */
196 /* The core of url_escape_* functions. Escapes the characters that
197 match the provided mask in urlchr_table.
199 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
200 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
201 allocated string will be returned in all cases. */
204 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
211 for (p1 = s; *p1; p1++)
212 if (urlchr_test (*p1, mask))
213 addition += 2; /* Two more characters (hex digits) */
216 return allow_passthrough ? (char *)s : xstrdup (s);
218 newlen = (p1 - s) + addition;
219 newstr = xmalloc (newlen + 1);
225 /* Quote the characters that match the test mask. */
226 if (urlchr_test (*p1, mask))
228 unsigned char c = *p1++;
230 *p2++ = XNUM_TO_DIGIT (c >> 4);
231 *p2++ = XNUM_TO_DIGIT (c & 0xf);
236 assert (p2 - newstr == newlen);
242 /* URL-escape the unsafe characters (see urlchr_table) in a given
243 string, returning a freshly allocated string. */
246 url_escape (const char *s)
248 return url_escape_1 (s, urlchr_unsafe, false);
251 /* URL-escape the unsafe characters (see urlchr_table) in a given
252 string. If no characters are unsafe, S is returned. */
255 url_escape_allow_passthrough (const char *s)
257 return url_escape_1 (s, urlchr_unsafe, true);
260 /* Decide whether the char at position P needs to be encoded. (It is
261 not enough to pass a single char *P because the function may need
262 to inspect the surrounding context.)
264 Return true if the char should be escaped as %XX, false otherwise. */
267 char_needs_escaping (const char *p)
271 if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2)))
274 /* Garbled %.. sequence: encode `%'. */
277 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
283 /* Translate a %-escaped (but possibly non-conformant) input string S
284 into a %-escaped (and conformant) output string. If no characters
285 are encoded or decoded, return the same string S; otherwise, return
286 a freshly allocated string with the new contents.
288 After a URL has been run through this function, the protocols that
289 use `%' as the quote character can use the resulting string as-is,
290 while those that don't can use url_unescape to get to the intended
291 data. This function is stable: once the input is transformed,
292 further transformations of the result yield the same output.
294 Let's discuss why this function is needed.
296 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
297 a raw space character would mess up the HTTP request, it needs to
298 be quoted, like this:
300 GET /abc%20def HTTP/1.0
302 It would appear that the unsafe chars need to be quoted, for
303 example with url_escape. But what if we're requested to download
304 `abc%20def'? url_escape transforms "%" to "%25", which would leave
305 us with `abc%2520def'. This is incorrect -- since %-escapes are
306 part of URL syntax, "%20" is the correct way to denote a literal
307 space on the Wget command line. This leads to the conclusion that
308 in that case Wget should not call url_escape, but leave the `%20'
309 as is. This is clearly contradictory, but it only gets worse.
311 What if the requested URI is `abc%20 def'? If we call url_escape,
312 we end up with `/abc%2520%20def', which is almost certainly not
313 intended. If we don't call url_escape, we are left with the
314 embedded space and cannot complete the request. What the user
315 meant was for Wget to request `/abc%20%20def', and this is where
316 reencode_escapes kicks in.
318 Wget used to solve this by first decoding %-quotes, and then
319 encoding all the "unsafe" characters found in the resulting string.
320 This was wrong because it didn't preserve certain URL special
321 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
322 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
323 whether we considered `+' reserved (it is). One of these results
324 is inevitable because by the second step we would lose information
325 on whether the `+' was originally encoded or not. Both results
326 were wrong because in CGI parameters + means space, while %2B means
327 literal plus. reencode_escapes correctly translates the above to
328 "a%2B+b", i.e. returns the original string.
330 This function uses a modified version of the algorithm originally
331 proposed by Anon Sricharoenchai:
333 * Encode all "unsafe" characters, except those that are also
334 "reserved", to %XX. See urlchr_table for which characters are
337 * Encode the "%" characters not followed by two hex digits to
340 * Pass through all other characters and %XX escapes as-is. (Up to
341 Wget 1.10 this decoded %XX escapes corresponding to "safe"
342 characters, but that was obtrusive and broke some servers.)
346 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
348 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
352 "foo bar" -> "foo%20bar"
353 "foo%20bar" -> "foo%20bar"
354 "foo %20bar" -> "foo%20%20bar"
355 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
356 "foo%25%20bar" -> "foo%25%20bar"
357 "foo%2%20bar" -> "foo%252%20bar"
358 "foo+bar" -> "foo+bar" (plus is reserved!)
359 "foo%2b+bar" -> "foo%2b+bar" */
362 reencode_escapes (const char *s)
368 int encode_count = 0;
370 /* First pass: inspect the string to see if there's anything to do,
371 and to calculate the new length. */
372 for (p1 = s; *p1; p1++)
373 if (char_needs_escaping (p1))
377 /* The string is good as it is. */
378 return (char *) s; /* C const model sucks. */
381 /* Each encoding adds two characters (hex digits). */
382 newlen = oldlen + 2 * encode_count;
383 newstr = xmalloc (newlen + 1);
385 /* Second pass: copy the string to the destination address, encoding
386 chars when needed. */
391 if (char_needs_escaping (p1))
393 unsigned char c = *p1++;
395 *p2++ = XNUM_TO_DIGIT (c >> 4);
396 *p2++ = XNUM_TO_DIGIT (c & 0xf);
402 assert (p2 - newstr == newlen);
406 /* Returns the scheme type if the scheme is supported, or
407 SCHEME_INVALID if not. */
410 url_scheme (const char *url)
414 for (i = 0; supported_schemes[i].leading_string; i++)
415 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
416 strlen (supported_schemes[i].leading_string)))
418 if (!(supported_schemes[i].flags & scm_disabled))
419 return (enum url_scheme) i;
421 return SCHEME_INVALID;
424 return SCHEME_INVALID;
427 #define SCHEME_CHAR(ch) (ISALNUM (ch) || (ch) == '-' || (ch) == '+')
429 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
430 currently implemented, it returns true if URL begins with
434 url_has_scheme (const char *url)
438 /* The first char must be a scheme char. */
439 if (!*p || !SCHEME_CHAR (*p))
442 /* Followed by 0 or more scheme chars. */
443 while (*p && SCHEME_CHAR (*p))
445 /* Terminated by ':'. */
450 scheme_default_port (enum url_scheme scheme)
452 return supported_schemes[scheme].default_port;
456 scheme_disable (enum url_scheme scheme)
458 supported_schemes[scheme].flags |= scm_disabled;
461 /* Skip the username and password, if present in the URL. The
462 function should *not* be called with the complete URL, but with the
463 portion after the scheme.
465 If no username and password are found, return URL. */
468 url_skip_credentials (const char *url)
470 /* Look for '@' that comes before terminators, such as '/', '?',
472 const char *p = (const char *)strpbrk (url, "@/?#;");
478 /* Parse credentials contained in [BEG, END). The region is expected
479 to have come from a URL and is unescaped. */
482 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
488 return false; /* empty user name */
490 colon = memchr (beg, ':', end - beg);
492 return false; /* again empty user name */
496 *passwd = strdupdelim (colon + 1, end);
498 url_unescape (*passwd);
505 *user = strdupdelim (beg, userend);
506 url_unescape (*user);
510 /* Used by main.c: detect URLs written using the "shorthand" URL forms
511 popularized by Netscape and NcFTP. HTTP shorthands look like this:
513 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
514 www.foo.com[:port] -> http://www.foo.com[:port]
516 FTP shorthands look like this:
518 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
519 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
521 If the URL needs not or cannot be rewritten, return NULL. */
524 rewrite_shorthand_url (const char *url)
528 if (url_scheme (url) != SCHEME_INVALID)
531 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
533 for (p = url; *p && *p != ':' && *p != '/'; p++)
539 /* If we're looking at "://", it means the URL uses a scheme we
540 don't support, which may include "https" when compiled without
541 SSL support. Don't bogusly rewrite such URLs. */
542 if (p[0] == ':' && p[1] == '/' && p[2] == '/')
549 /* If the characters after the colon and before the next slash
550 or end of string are all digits, it's HTTP. */
552 for (pp = p + 1; ISDIGIT (*pp); pp++)
554 if (digits > 0 && (*pp == '/' || *pp == '\0'))
557 /* Prepend "ftp://" to the entire URL... */
558 res = xmalloc (6 + strlen (url) + 1);
559 sprintf (res, "ftp://%s", url);
560 /* ...and replace ':' with '/'. */
561 res[6 + (p - url)] = '/';
568 /* Just prepend "http://" to what we have. */
569 res = xmalloc (7 + strlen (url) + 1);
570 sprintf (res, "http://%s", url);
575 static void split_path (const char *, char **, char **);
577 /* Like strpbrk, with the exception that it returns the pointer to the
578 terminating zero (end-of-string aka "eos") if no matching character
582 strpbrk_or_eos (const char *s, const char *accept)
584 char *p = strpbrk (s, accept);
586 p = strchr (s, '\0');
590 /* Turn STR into lowercase; return true if a character was actually
594 lowercase_str (char *str)
596 bool changed = false;
601 *str = TOLOWER (*str);
607 init_seps (enum url_scheme scheme)
609 static char seps[8] = ":/";
611 int flags = supported_schemes[scheme].flags;
613 if (flags & scm_has_params)
615 if (flags & scm_has_query)
617 if (flags & scm_has_fragment)
623 static const char *parse_errors[] = {
624 #define PE_NO_ERROR 0
626 #define PE_UNSUPPORTED_SCHEME 1
627 N_("Unsupported scheme"),
628 #define PE_INVALID_HOST_NAME 2
629 N_("Invalid host name"),
630 #define PE_BAD_PORT_NUMBER 3
631 N_("Bad port number"),
632 #define PE_INVALID_USER_NAME 4
633 N_("Invalid user name"),
634 #define PE_UNTERMINATED_IPV6_ADDRESS 5
635 N_("Unterminated IPv6 numeric address"),
636 #define PE_IPV6_NOT_SUPPORTED 6
637 N_("IPv6 addresses not supported"),
638 #define PE_INVALID_IPV6_ADDRESS 7
639 N_("Invalid IPv6 numeric address")
644 Return a new struct url if successful, NULL on error. In case of
645 error, and if ERROR is not NULL, also set *ERROR to the appropriate
648 url_parse (const char *url, int *error)
652 bool path_modified, host_modified;
654 enum url_scheme scheme;
657 const char *uname_b, *uname_e;
658 const char *host_b, *host_e;
659 const char *path_b, *path_e;
660 const char *params_b, *params_e;
661 const char *query_b, *query_e;
662 const char *fragment_b, *fragment_e;
665 char *user = NULL, *passwd = NULL;
667 char *url_encoded = NULL;
671 scheme = url_scheme (url);
672 if (scheme == SCHEME_INVALID)
674 error_code = PE_UNSUPPORTED_SCHEME;
678 url_encoded = reencode_escapes (url);
681 p += strlen (supported_schemes[scheme].leading_string);
683 p = url_skip_credentials (p);
686 /* scheme://user:pass@host[:port]... */
689 /* We attempt to break down the URL into the components path,
690 params, query, and fragment. They are ordered like this:
692 scheme://host[:port][/path][;params][?query][#fragment] */
694 path_b = path_e = NULL;
695 params_b = params_e = NULL;
696 query_b = query_e = NULL;
697 fragment_b = fragment_e = NULL;
699 /* Initialize separators for optional parts of URL, depending on the
700 scheme. For example, FTP has params, and HTTP and HTTPS have
701 query string and fragment. */
702 seps = init_seps (scheme);
708 /* Handle IPv6 address inside square brackets. Ideally we'd
709 just look for the terminating ']', but rfc2732 mandates
710 rejecting invalid IPv6 addresses. */
712 /* The address begins after '['. */
714 host_e = strchr (host_b, ']');
718 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
723 /* Check if the IPv6 address is valid. */
724 if (!is_valid_ipv6_address(host_b, host_e))
726 error_code = PE_INVALID_IPV6_ADDRESS;
730 /* Continue parsing after the closing ']'. */
733 error_code = PE_IPV6_NOT_SUPPORTED;
737 /* The closing bracket must be followed by a separator or by the
739 /* http://[::1]... */
741 if (!strchr (seps, *p))
743 /* Trailing garbage after []-delimited IPv6 address. */
744 error_code = PE_INVALID_HOST_NAME;
750 p = strpbrk_or_eos (p, seps);
753 ++seps; /* advance to '/' */
755 if (host_b == host_e)
757 error_code = PE_INVALID_HOST_NAME;
761 port = scheme_default_port (scheme);
764 const char *port_b, *port_e, *pp;
766 /* scheme://host:port/tralala */
770 p = strpbrk_or_eos (p, seps);
773 /* Allow empty port, as per rfc2396. */
774 if (port_b != port_e)
775 for (port = 0, pp = port_b; pp < port_e; pp++)
779 /* http://host:12randomgarbage/blah */
781 error_code = PE_BAD_PORT_NUMBER;
784 port = 10 * port + (*pp - '0');
785 /* Check for too large port numbers here, before we have
786 a chance to overflow on bogus port values. */
789 error_code = PE_BAD_PORT_NUMBER;
794 /* Advance to the first separator *after* '/' (either ';' or '?',
795 depending on the scheme). */
798 /* Get the optional parts of URL, each part being delimited by
799 current location and the position of the next separator. */
800 #define GET_URL_PART(sepchar, var) do { \
802 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
806 GET_URL_PART ('/', path);
807 if (supported_schemes[scheme].flags & scm_has_params)
808 GET_URL_PART (';', params);
809 if (supported_schemes[scheme].flags & scm_has_query)
810 GET_URL_PART ('?', query);
811 if (supported_schemes[scheme].flags & scm_has_fragment)
812 GET_URL_PART ('#', fragment);
817 if (uname_b != uname_e)
819 /* http://user:pass@host */
821 /* uname_b uname_e */
822 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
824 error_code = PE_INVALID_USER_NAME;
829 u = xnew0 (struct url);
831 u->host = strdupdelim (host_b, host_e);
836 u->path = strdupdelim (path_b, path_e);
837 path_modified = path_simplify (u->path);
838 split_path (u->path, &u->dir, &u->file);
840 host_modified = lowercase_str (u->host);
842 /* Decode %HH sequences in host name. This is important not so much
843 to support %HH sequences in host names (which other browser
844 don't), but to support binary characters (which will have been
845 converted to %HH by reencode_escapes). */
846 if (strchr (u->host, '%'))
848 url_unescape (u->host);
849 host_modified = true;
853 u->params = strdupdelim (params_b, params_e);
855 u->query = strdupdelim (query_b, query_e);
857 u->fragment = strdupdelim (fragment_b, fragment_e);
859 if (path_modified || u->fragment || host_modified || path_b == path_e)
861 /* If we suspect that a transformation has rendered what
862 url_string might return different from URL_ENCODED, rebuild
863 u->url using url_string. */
864 u->url = url_string (u, false);
866 if (url_encoded != url)
867 xfree ((char *) url_encoded);
871 if (url_encoded == url)
872 u->url = xstrdup (url);
874 u->url = url_encoded;
880 /* Cleanup in case of error: */
881 if (url_encoded && url_encoded != url)
884 /* Transmit the error code to the caller, if the caller wants to
891 /* Return the error message string from ERROR_CODE, which should have
892 been retrieved from url_parse. The error message is translated. */
895 url_error (int error_code)
897 assert (error_code >= 0 && error_code < countof (parse_errors));
898 return _(parse_errors[error_code]);
901 /* Split PATH into DIR and FILE. PATH comes from the URL and is
902 expected to be URL-escaped.
904 The path is split into directory (the part up to the last slash)
905 and file (the part after the last slash), which are subsequently
909 "foo/bar/baz" "foo/bar" "baz"
910 "foo/bar/" "foo/bar" ""
912 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
914 DIR and FILE are freshly allocated. */
917 split_path (const char *path, char **dir, char **file)
919 char *last_slash = strrchr (path, '/');
923 *file = xstrdup (path);
927 *dir = strdupdelim (path, last_slash);
928 *file = xstrdup (last_slash + 1);
931 url_unescape (*file);
934 /* Note: URL's "full path" is the path with the query string and
935 params appended. The "fragment" (#foo) is intentionally ignored,
936 but that might be changed. For example, if the original URL was
937 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
938 the full path will be "/foo/bar/baz;bullshit?querystring". */
940 /* Return the length of the full path, without the terminating
944 full_path_length (const struct url *url)
948 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
959 /* Write out the full path. */
962 full_path_write (const struct url *url, char *where)
964 #define FROB(el, chr) do { \
965 char *f_el = url->el; \
967 int l = strlen (f_el); \
969 memcpy (where, f_el, l); \
981 /* Public function for getting the "full path". E.g. if u->path is
982 "foo/bar" and u->query is "param=value", full_path will be
983 "/foo/bar?param=value". */
986 url_full_path (const struct url *url)
988 int length = full_path_length (url);
989 char *full_path = xmalloc (length + 1);
991 full_path_write (url, full_path);
992 full_path[length] = '\0';
997 /* Unescape CHR in an otherwise escaped STR. Used to selectively
998 escaping of certain characters, such as "/" and ":". Returns a
999 count of unescaped chars. */
1002 unescape_single_char (char *str, char chr)
1004 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1005 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1006 char *h = str; /* hare */
1007 char *t = str; /* tortoise */
1008 for (; *h; h++, t++)
1010 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1021 /* Escape unsafe and reserved characters, except for the slash
1025 url_escape_dir (const char *dir)
1027 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1031 unescape_single_char (newdir, '/');
1035 /* Sync u->path and u->url with u->dir and u->file. Called after
1036 u->file or u->dir have been changed, typically by the FTP code. */
1039 sync_path (struct url *u)
1041 char *newpath, *efile, *edir;
1045 /* u->dir and u->file are not escaped. URL-escape them before
1046 reassembling them into u->path. That way, if they contain
1047 separators like '?' or even if u->file contains slashes, the
1048 path will be correctly assembled. (u->file can contain slashes
1049 if the URL specifies it with %2f, or if an FTP server returns
1051 edir = url_escape_dir (u->dir);
1052 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1055 newpath = xstrdup (efile);
1058 int dirlen = strlen (edir);
1059 int filelen = strlen (efile);
1061 /* Copy "DIR/FILE" to newpath. */
1062 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1063 memcpy (p, edir, dirlen);
1066 memcpy (p, efile, filelen);
1075 if (efile != u->file)
1078 /* Regenerate u->url as well. */
1080 u->url = url_string (u, false);
1083 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1084 This way we can sync u->path and u->url when they get changed. */
1087 url_set_dir (struct url *url, const char *newdir)
1090 url->dir = xstrdup (newdir);
1095 url_set_file (struct url *url, const char *newfile)
1098 url->file = xstrdup (newfile);
1103 url_free (struct url *url)
1109 xfree_null (url->params);
1110 xfree_null (url->query);
1111 xfree_null (url->fragment);
1112 xfree_null (url->user);
1113 xfree_null (url->passwd);
1121 /* Create all the necessary directories for PATH (a file). Calls
1122 make_directory internally. */
1124 mkalldirs (const char *path)
1131 p = path + strlen (path);
1132 for (; *p != '/' && p != path; p--)
1135 /* Don't create if it's just a file. */
1136 if ((p == path) && (*p != '/'))
1138 t = strdupdelim (path, p);
1140 /* Check whether the directory exists. */
1141 if ((stat (t, &st) == 0))
1143 if (S_ISDIR (st.st_mode))
1150 /* If the dir exists as a file name, remove it first. This
1151 is *only* for Wget to work with buggy old CERN http
1152 servers. Here is the scenario: When Wget tries to
1153 retrieve a directory without a slash, e.g.
1154 http://foo/bar (bar being a directory), CERN server will
1155 not redirect it too http://foo/bar/ -- it will generate a
1156 directory listing containing links to bar/file1,
1157 bar/file2, etc. Wget will lose because it saves this
1158 HTML listing to a file `bar', so it cannot create the
1159 directory. To work around this, if the file of the same
1160 name exists, we just remove it and create the directory
1162 DEBUGP (("Removing %s because of directory danger!\n", t));
1166 res = make_directory (t);
1168 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1173 /* Functions for constructing the file name out of URL components. */
1175 /* A growable string structure, used by url_file_name and friends.
1176 This should perhaps be moved to utils.c.
1178 The idea is to have a convenient and efficient way to construct a
1179 string by having various functions append data to it. Instead of
1180 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1181 functions in questions, we pass the pointer to this struct. */
1189 /* Ensure that the string can accept APPEND_COUNT more characters past
1190 the current TAIL position. If necessary, this will grow the string
1191 and update its allocated size. If the string is already large
1192 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1193 #define GROW(g, append_size) do { \
1194 struct growable *G_ = g; \
1195 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1198 /* Return the tail position of the string. */
1199 #define TAIL(r) ((r)->base + (r)->tail)
1201 /* Move the tail position by APPEND_COUNT characters. */
1202 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1204 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1208 append_string (const char *str, struct growable *dest)
1210 int l = strlen (str);
1212 memcpy (TAIL (dest), str, l);
1213 TAIL_INCR (dest, l);
1216 /* Append CH to DEST. For example, append_char (0, DEST)
1217 zero-terminates DEST. */
1220 append_char (char ch, struct growable *dest)
1224 TAIL_INCR (dest, 1);
1228 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1229 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1230 filechr_control = 4 /* a control character, e.g. 0-31 */
1233 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1235 /* Shorthands for the table: */
1236 #define U filechr_not_unix
1237 #define W filechr_not_windows
1238 #define C filechr_control
1243 /* Table of characters unsafe under various conditions (see above).
1245 Arguably we could also claim `%' to be unsafe, since we use it as
1246 the escape character. If we ever want to be able to reliably
1247 translate file name back to URL, this would become important
1248 crucial. Right now, it's better to be minimal in escaping. */
1250 static const unsigned char filechr_table[256] =
1252 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1253 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1254 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1255 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1256 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1257 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1258 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1259 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1260 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1261 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1262 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1263 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1264 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1265 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1266 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1267 0, 0, 0, 0, 0, 0, 0, 0, /* x y z { | } ~ DEL */
1269 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1270 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1271 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1272 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
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,
1276 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,
1285 /* FN_PORT_SEP is the separator between host and port in file names
1286 for non-standard port numbers. On Unix this is normally ':', as in
1287 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1288 because Windows can't handle ':' in file names. */
1289 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1291 /* FN_QUERY_SEP is the separator between the file name and the URL
1292 query, normally '?'. Since Windows cannot handle '?' as part of
1293 file name, we use '@' instead there. */
1294 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1296 /* Quote path element, characters in [b, e), as file name, and append
1297 the quoted string to DEST. Each character is quoted as per
1298 file_unsafe_char and the corresponding table.
1300 If ESCAPED is true, the path element is considered to be
1301 URL-escaped and will be unescaped prior to inspection. */
1304 append_uri_pathel (const char *b, const char *e, bool escaped,
1305 struct growable *dest)
1311 if (opt.restrict_files_os == restrict_unix)
1312 mask = filechr_not_unix;
1314 mask = filechr_not_windows;
1315 if (opt.restrict_files_ctrl)
1316 mask |= filechr_control;
1318 /* Copy [b, e) to PATHEL and URL-unescape it. */
1322 BOUNDED_TO_ALLOCA (b, e, unescaped);
1323 url_unescape (unescaped);
1325 e = unescaped + strlen (unescaped);
1328 /* Defang ".." when found as component of path. Remember that path
1329 comes from the URL and might contain malicious input. */
1330 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1336 /* Walk the PATHEL string and check how many characters we'll need
1339 for (p = b; p < e; p++)
1340 if (FILE_CHAR_TEST (*p, mask))
1343 /* Calculate the length of the output string. e-b is the input
1344 string length. Each quoted char introduces two additional
1345 characters in the string, hence 2*quoted. */
1346 outlen = (e - b) + (2 * quoted);
1347 GROW (dest, outlen);
1351 /* If there's nothing to quote, we can simply append the string
1352 without processing it again. */
1353 memcpy (TAIL (dest), b, outlen);
1357 char *q = TAIL (dest);
1358 for (p = b; p < e; p++)
1360 if (!FILE_CHAR_TEST (*p, mask))
1364 unsigned char ch = *p;
1366 *q++ = XNUM_TO_DIGIT (ch >> 4);
1367 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1370 assert (q - TAIL (dest) == outlen);
1372 TAIL_INCR (dest, outlen);
1375 /* Append to DEST the directory structure that corresponds the
1376 directory part of URL's path. For example, if the URL is
1377 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1379 Each path element ("dir1" and "dir2" in the above example) is
1380 examined, url-unescaped, and re-escaped as file name element.
1382 Additionally, it cuts as many directories from the path as
1383 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1384 will produce "bar" for the above example. For 2 or more, it will
1387 Each component of the path is quoted for use as file name. */
1390 append_dir_structure (const struct url *u, struct growable *dest)
1392 char *pathel, *next;
1393 int cut = opt.cut_dirs;
1395 /* Go through the path components, de-URL-quote them, and quote them
1396 (if necessary) as file names. */
1399 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1404 /* Ignore empty pathels. */
1408 append_char ('/', dest);
1409 append_uri_pathel (pathel, next, true, dest);
1413 /* Return a unique file name that matches the given URL as good as
1414 possible. Does not create directories on the file system. */
1417 url_file_name (const struct url *u)
1419 struct growable fnres; /* stands for "file name result" */
1421 const char *u_file, *u_query;
1422 char *fname, *unique;
1428 /* Start with the directory prefix, if specified. */
1430 append_string (opt.dir_prefix, &fnres);
1432 /* If "dirstruct" is turned on (typically the case with -r), add
1433 the host and port (unless those have been turned off) and
1434 directory structure. */
1437 if (opt.protocol_directories)
1440 append_char ('/', &fnres);
1441 append_string (supported_schemes[u->scheme].name, &fnres);
1443 if (opt.add_hostdir)
1446 append_char ('/', &fnres);
1447 if (0 != strcmp (u->host, ".."))
1448 append_string (u->host, &fnres);
1450 /* Host name can come from the network; malicious DNS may
1451 allow ".." to be resolved, causing us to write to
1452 "../<file>". Defang such host names. */
1453 append_string ("%2E%2E", &fnres);
1454 if (u->port != scheme_default_port (u->scheme))
1457 number_to_string (portstr, u->port);
1458 append_char (FN_PORT_SEP, &fnres);
1459 append_string (portstr, &fnres);
1463 append_dir_structure (u, &fnres);
1466 /* Add the file name. */
1468 append_char ('/', &fnres);
1469 u_file = *u->file ? u->file : "index.html";
1470 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1472 /* Append "?query" to the file name. */
1473 u_query = u->query && *u->query ? u->query : NULL;
1476 append_char (FN_QUERY_SEP, &fnres);
1477 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1480 /* Zero-terminate the file name. */
1481 append_char ('\0', &fnres);
1485 /* Check the cases in which the unique extensions are not used:
1486 1) Clobbering is turned off (-nc).
1487 2) Retrieval with regetting.
1488 3) Timestamping is used.
1489 4) Hierarchy is built.
1491 The exception is the case when file does exist and is a
1492 directory (see `mkalldirs' for explanation). */
1494 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1495 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1498 unique = unique_name (fname, true);
1499 if (unique != fname)
1504 /* Resolve "." and ".." elements of PATH by destructively modifying
1505 PATH and return true if PATH has been modified, false otherwise.
1507 The algorithm is in spirit similar to the one described in rfc1808,
1508 although implemented differently, in one pass. To recap, path
1509 elements containing only "." are removed, and ".." is taken to mean
1510 "back up one element". Single leading and trailing slashes are
1513 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1514 test examples are provided below. If you change anything in this
1515 function, run test_path_simplify to make sure you haven't broken a
1519 path_simplify (char *path)
1521 char *h = path; /* hare */
1522 char *t = path; /* tortoise */
1523 char *beg = path; /* boundary for backing the tortoise */
1524 char *end = path + strlen (path);
1528 /* Hare should be at the beginning of a path element. */
1530 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1535 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1537 /* Handle "../" by retreating the tortoise by one path
1538 element -- but not past beggining. */
1541 /* Move backwards until T hits the beginning of the
1542 previous path element or the beginning of path. */
1543 for (--t; t > beg && t[-1] != '/'; t--)
1548 /* If we're at the beginning, copy the "../" literally
1549 move the beginning so a later ".." doesn't remove
1559 /* A regular path element. If H hasn't advanced past T,
1560 simply skip to the next path element. Otherwise, copy
1561 the path element until the next slash. */
1564 /* Skip the path element, including the slash. */
1565 while (h < end && *h != '/')
1572 /* Copy the path element, including the final slash. */
1573 while (h < end && *h != '/')
1587 /* Return the length of URL's path. Path is considered to be
1588 terminated by one or more of the ?query or ;params or #fragment,
1589 depending on the scheme. */
1592 path_end (const char *url)
1594 enum url_scheme scheme = url_scheme (url);
1596 if (scheme == SCHEME_INVALID)
1597 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1598 /* +2 to ignore the first two separators ':' and '/' */
1599 seps = init_seps (scheme) + 2;
1600 return strpbrk_or_eos (url, seps);
1603 /* Find the last occurrence of character C in the range [b, e), or
1604 NULL, if none are present. We might want to use memrchr (a GNU
1605 extension) under GNU libc. */
1608 find_last_char (const char *b, const char *e, char c)
1616 /* Merge BASE with LINK and return the resulting URI.
1618 Either of the URIs may be absolute or relative, complete with the
1619 host name, or path only. This tries to reasonably handle all
1620 foreseeable cases. It only employs minimal URL parsing, without
1621 knowledge of the specifics of schemes.
1623 I briefly considered making this function call path_simplify after
1624 the merging process, as rfc1738 seems to suggest. This is a bad
1625 idea for several reasons: 1) it complexifies the code, and 2)
1626 url_parse has to simplify path anyway, so it's wasteful to boot. */
1629 uri_merge (const char *base, const char *link)
1635 if (url_has_scheme (link))
1636 return xstrdup (link);
1638 /* We may not examine BASE past END. */
1639 end = path_end (base);
1640 linklength = strlen (link);
1644 /* Empty LINK points back to BASE, query string and all. */
1645 return xstrdup (base);
1647 else if (*link == '?')
1649 /* LINK points to the same location, but changes the query
1650 string. Examples: */
1651 /* uri_merge("path", "?new") -> "path?new" */
1652 /* uri_merge("path?foo", "?new") -> "path?new" */
1653 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1654 /* uri_merge("path#foo", "?new") -> "path?new" */
1655 int baselength = end - base;
1656 merge = xmalloc (baselength + linklength + 1);
1657 memcpy (merge, base, baselength);
1658 memcpy (merge + baselength, link, linklength);
1659 merge[baselength + linklength] = '\0';
1661 else if (*link == '#')
1663 /* uri_merge("path", "#new") -> "path#new" */
1664 /* uri_merge("path#foo", "#new") -> "path#new" */
1665 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1666 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1668 const char *end1 = strchr (base, '#');
1670 end1 = base + strlen (base);
1671 baselength = end1 - base;
1672 merge = xmalloc (baselength + linklength + 1);
1673 memcpy (merge, base, baselength);
1674 memcpy (merge + baselength, link, linklength);
1675 merge[baselength + linklength] = '\0';
1677 else if (*link == '/' && *(link + 1) == '/')
1679 /* LINK begins with "//" and so is a net path: we need to
1680 replace everything after (and including) the double slash
1683 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1684 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1685 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1689 const char *start_insert;
1691 /* Look for first slash. */
1692 slash = memchr (base, '/', end - base);
1693 /* If found slash and it is a double slash, then replace
1694 from this point, else default to replacing from the
1696 if (slash && *(slash + 1) == '/')
1697 start_insert = slash;
1699 start_insert = base;
1701 span = start_insert - base;
1702 merge = xmalloc (span + linklength + 1);
1704 memcpy (merge, base, span);
1705 memcpy (merge + span, link, linklength);
1706 merge[span + linklength] = '\0';
1708 else if (*link == '/')
1710 /* LINK is an absolute path: we need to replace everything
1711 after (and including) the FIRST slash with LINK.
1713 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1714 "/qux/xyzzy", our result should be
1715 "http://host/qux/xyzzy". */
1718 const char *start_insert = NULL; /* for gcc to shut up. */
1719 const char *pos = base;
1720 bool seen_slash_slash = false;
1721 /* We're looking for the first slash, but want to ignore
1724 slash = memchr (pos, '/', end - pos);
1725 if (slash && !seen_slash_slash)
1726 if (*(slash + 1) == '/')
1729 seen_slash_slash = true;
1733 /* At this point, SLASH is the location of the first / after
1734 "//", or the first slash altogether. START_INSERT is the
1735 pointer to the location where LINK will be inserted. When
1736 examining the last two examples, keep in mind that LINK
1739 if (!slash && !seen_slash_slash)
1740 /* example: "foo" */
1742 start_insert = base;
1743 else if (!slash && seen_slash_slash)
1744 /* example: "http://foo" */
1747 else if (slash && !seen_slash_slash)
1748 /* example: "foo/bar" */
1750 start_insert = base;
1751 else if (slash && seen_slash_slash)
1752 /* example: "http://something/" */
1754 start_insert = slash;
1756 span = start_insert - base;
1757 merge = xmalloc (span + linklength + 1);
1759 memcpy (merge, base, span);
1760 memcpy (merge + span, link, linklength);
1761 merge[span + linklength] = '\0';
1765 /* LINK is a relative URL: we need to replace everything
1766 after last slash (possibly empty) with LINK.
1768 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1769 our result should be "whatever/foo/qux/xyzzy". */
1770 bool need_explicit_slash = false;
1772 const char *start_insert;
1773 const char *last_slash = find_last_char (base, end, '/');
1776 /* No slash found at all. Replace what we have with LINK. */
1777 start_insert = base;
1779 else if (last_slash && last_slash >= base + 2
1780 && last_slash[-2] == ':' && last_slash[-1] == '/')
1782 /* example: http://host" */
1784 start_insert = end + 1;
1785 need_explicit_slash = true;
1789 /* example: "whatever/foo/bar" */
1791 start_insert = last_slash + 1;
1794 span = start_insert - base;
1795 merge = xmalloc (span + linklength + 1);
1797 memcpy (merge, base, span);
1798 if (need_explicit_slash)
1799 merge[span - 1] = '/';
1800 memcpy (merge + span, link, linklength);
1801 merge[span + linklength] = '\0';
1807 #define APPEND(p, s) do { \
1808 int len = strlen (s); \
1809 memcpy (p, s, len); \
1813 /* Use this instead of password when the actual password is supposed
1814 to be hidden. We intentionally use a generic string without giving
1815 away the number of characters in the password, like previous
1817 #define HIDDEN_PASSWORD "*password*"
1819 /* Recreate the URL string from the data in URL.
1821 If HIDE is true (as it is when we're calling this on a URL we plan
1822 to print, but not when calling it to canonicalize a URL for use
1823 within the program), password will be hidden. Unsafe characters in
1824 the URL will be quoted. */
1827 url_string (const struct url *url, bool hide_password)
1831 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1833 int scheme_port = supported_schemes[url->scheme].default_port;
1834 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1835 int fplen = full_path_length (url);
1837 bool brackets_around_host;
1839 assert (scheme_str != NULL);
1841 /* Make sure the user name and password are quoted. */
1844 quoted_user = url_escape_allow_passthrough (url->user);
1848 quoted_passwd = HIDDEN_PASSWORD;
1850 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1854 /* In the unlikely event that the host name contains non-printable
1855 characters, quote it for displaying to the user. */
1856 quoted_host = url_escape_allow_passthrough (url->host);
1858 /* Undo the quoting of colons that URL escaping performs. IPv6
1859 addresses may legally contain colons, and in that case must be
1860 placed in square brackets. */
1861 if (quoted_host != url->host)
1862 unescape_single_char (quoted_host, ':');
1863 brackets_around_host = strchr (quoted_host, ':') != NULL;
1865 size = (strlen (scheme_str)
1866 + strlen (quoted_host)
1867 + (brackets_around_host ? 2 : 0)
1870 if (url->port != scheme_port)
1871 size += 1 + numdigit (url->port);
1874 size += 1 + strlen (quoted_user);
1876 size += 1 + strlen (quoted_passwd);
1879 p = result = xmalloc (size);
1881 APPEND (p, scheme_str);
1884 APPEND (p, quoted_user);
1888 APPEND (p, quoted_passwd);
1893 if (brackets_around_host)
1895 APPEND (p, quoted_host);
1896 if (brackets_around_host)
1898 if (url->port != scheme_port)
1901 p = number_to_string (p, url->port);
1904 full_path_write (url, p);
1908 assert (p - result == size);
1910 if (quoted_user && quoted_user != url->user)
1911 xfree (quoted_user);
1912 if (quoted_passwd && !hide_password && quoted_passwd != url->passwd)
1913 xfree (quoted_passwd);
1914 if (quoted_host != url->host)
1915 xfree (quoted_host);
1920 /* Return true if scheme a is similar to scheme b.
1922 Schemes are similar if they are equal. If SSL is supported, schemes
1923 are also similar if one is http (SCHEME_HTTP) and the other is https
1926 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1931 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1932 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1939 /* Debugging and testing support for path_simplify. */
1941 /* Debug: run path_simplify on PATH and return the result in a new
1942 string. Useful for calling from the debugger. */
1946 char *copy = xstrdup (path);
1947 path_simplify (copy);
1952 run_test (char *test, char *expected_result, bool expected_change)
1954 char *test_copy = xstrdup (test);
1955 bool modified = path_simplify (test_copy);
1957 if (0 != strcmp (test_copy, expected_result))
1959 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
1960 test, expected_result, test_copy);
1962 if (modified != expected_change)
1964 if (expected_change)
1965 printf ("Expected modification with path_simplify(\"%s\").\n",
1968 printf ("Expected no modification with path_simplify(\"%s\").\n",
1975 test_path_simplify (void)
1978 char *test, *result;
1984 { "..", "..", false },
1985 { "../", "../", false },
1986 { "foo", "foo", false },
1987 { "foo/bar", "foo/bar", false },
1988 { "foo///bar", "foo///bar", false },
1989 { "foo/.", "foo/", true },
1990 { "foo/./", "foo/", true },
1991 { "foo./", "foo./", false },
1992 { "foo/../bar", "bar", true },
1993 { "foo/../bar/", "bar/", true },
1994 { "foo/bar/..", "foo/", true },
1995 { "foo/bar/../x", "foo/x", true },
1996 { "foo/bar/../x/", "foo/x/", true },
1997 { "foo/..", "", true },
1998 { "foo/../..", "..", true },
1999 { "foo/../../..", "../..", true },
2000 { "foo/../../bar/../../baz", "../../baz", true },
2001 { "a/b/../../c", "c", true },
2002 { "./a/../b", "b", true }
2006 for (i = 0; i < countof (tests); i++)
2008 char *test = tests[i].test;
2009 char *expected_result = tests[i].result;
2010 bool expected_change = tests[i].should_modify;
2011 run_test (test, expected_result, expected_change);