2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
3 2004, 2005, 2006, 2007, 2008 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 3 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, see <http://www.gnu.org/licenses/>.
20 Additional permission under GNU GPL version 3 section 7
22 If you modify this program, or any covered work, by linking or
23 combining it with the OpenSSL project's OpenSSL library (or a
24 modified version of that library), containing parts covered by the
25 terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
26 grants you additional permission to convey the resulting work.
27 Corresponding Source for a non-source form of such a combination
28 shall include the source code for the parts of OpenSSL used as well
29 as that of the covered work. */
44 #include "host.h" /* for is_valid_ipv6_address */
51 scm_disabled = 1, /* for https when OpenSSL fails to init. */
52 scm_has_params = 2, /* whether scheme has ;params */
53 scm_has_query = 4, /* whether scheme has ?query */
54 scm_has_fragment = 8 /* whether scheme has #fragment */
59 /* Short name of the scheme, such as "http" or "ftp". */
61 /* Leading string that identifies the scheme, such as "https://". */
62 const char *leading_string;
63 /* Default port of the scheme when none is specified. */
69 /* Supported schemes: */
70 static struct scheme_data supported_schemes[] =
72 { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment },
74 { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment },
76 { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment },
82 /* Forward declarations: */
84 static bool path_simplify (enum url_scheme, char *);
86 /* Support for escaping and unescaping of URL strings. */
88 /* Table of "reserved" and "unsafe" characters. Those terms are
89 rfc1738-speak, as such largely obsoleted by rfc2396 and later
90 specs, but the general idea remains.
92 A reserved character is the one that you can't decode without
93 changing the meaning of the URL. For example, you can't decode
94 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
95 path components is different. Non-reserved characters can be
96 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
97 unsafe characters are loosely based on rfc1738, plus "$" and ",",
98 as recommended by rfc2396, and minus "~", which is very frequently
99 used (and sometimes unrecognized as %7E by broken servers).
101 An unsafe character is the one that should be encoded when URLs are
102 placed in foreign environments. E.g. space and newline are unsafe
103 in HTTP contexts because HTTP uses them as separator and line
104 terminator, so they must be encoded to %20 and %0A respectively.
105 "*" is unsafe in shell context, etc.
107 We determine whether a character is unsafe through static table
108 lookup. This code assumes ASCII character set and 8-bit chars. */
111 /* rfc1738 reserved chars + "$" and ",". */
114 /* rfc1738 unsafe chars, plus non-printables. */
118 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
119 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
120 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
122 /* Shorthands for the table: */
123 #define R urlchr_reserved
124 #define U urlchr_unsafe
127 static const unsigned char urlchr_table[256] =
129 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
130 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
131 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
132 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
133 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
134 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
135 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
136 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
137 RU, 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, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
141 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
142 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
143 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
144 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
146 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
147 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
148 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
149 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
151 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
152 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
153 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
154 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
160 /* URL-unescape the string S.
162 This is done by transforming the sequences "%HH" to the character
163 represented by the hexadecimal digits HH. If % is not followed by
164 two hexadecimal digits, it is inserted literally.
166 The transformation is done in place. If you need the original
167 string intact, make a copy before calling this function. */
170 url_unescape (char *s)
172 char *t = s; /* t - tortoise */
173 char *h = s; /* h - hare */
185 /* Do nothing if '%' is not followed by two hex digits. */
186 if (!h[1] || !h[2] || !(c_isxdigit (h[1]) && c_isxdigit (h[2])))
188 c = X2DIGITS_TO_NUM (h[1], h[2]);
189 /* Don't unescape %00 because there is no way to insert it
190 into a C string without effectively truncating it. */
200 /* The core of url_escape_* functions. Escapes the characters that
201 match the provided mask in urlchr_table.
203 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
204 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
205 allocated string will be returned in all cases. */
208 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
215 for (p1 = s; *p1; p1++)
216 if (urlchr_test (*p1, mask))
217 addition += 2; /* Two more characters (hex digits) */
220 return allow_passthrough ? (char *)s : xstrdup (s);
222 newlen = (p1 - s) + addition;
223 newstr = xmalloc (newlen + 1);
229 /* Quote the characters that match the test mask. */
230 if (urlchr_test (*p1, mask))
232 unsigned char c = *p1++;
234 *p2++ = XNUM_TO_DIGIT (c >> 4);
235 *p2++ = XNUM_TO_DIGIT (c & 0xf);
240 assert (p2 - newstr == newlen);
246 /* URL-escape the unsafe characters (see urlchr_table) in a given
247 string, returning a freshly allocated string. */
250 url_escape (const char *s)
252 return url_escape_1 (s, urlchr_unsafe, false);
255 /* URL-escape the unsafe characters (see urlchr_table) in a given
256 string. If no characters are unsafe, S is returned. */
259 url_escape_allow_passthrough (const char *s)
261 return url_escape_1 (s, urlchr_unsafe, true);
264 /* Decide whether the char at position P needs to be encoded. (It is
265 not enough to pass a single char *P because the function may need
266 to inspect the surrounding context.)
268 Return true if the char should be escaped as %XX, false otherwise. */
271 char_needs_escaping (const char *p)
275 if (c_isxdigit (*(p + 1)) && c_isxdigit (*(p + 2)))
278 /* Garbled %.. sequence: encode `%'. */
281 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
287 /* Translate a %-escaped (but possibly non-conformant) input string S
288 into a %-escaped (and conformant) output string. If no characters
289 are encoded or decoded, return the same string S; otherwise, return
290 a freshly allocated string with the new contents.
292 After a URL has been run through this function, the protocols that
293 use `%' as the quote character can use the resulting string as-is,
294 while those that don't can use url_unescape to get to the intended
295 data. This function is stable: once the input is transformed,
296 further transformations of the result yield the same output.
298 Let's discuss why this function is needed.
300 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
301 a raw space character would mess up the HTTP request, it needs to
302 be quoted, like this:
304 GET /abc%20def HTTP/1.0
306 It would appear that the unsafe chars need to be quoted, for
307 example with url_escape. But what if we're requested to download
308 `abc%20def'? url_escape transforms "%" to "%25", which would leave
309 us with `abc%2520def'. This is incorrect -- since %-escapes are
310 part of URL syntax, "%20" is the correct way to denote a literal
311 space on the Wget command line. This leads to the conclusion that
312 in that case Wget should not call url_escape, but leave the `%20'
313 as is. This is clearly contradictory, but it only gets worse.
315 What if the requested URI is `abc%20 def'? If we call url_escape,
316 we end up with `/abc%2520%20def', which is almost certainly not
317 intended. If we don't call url_escape, we are left with the
318 embedded space and cannot complete the request. What the user
319 meant was for Wget to request `/abc%20%20def', and this is where
320 reencode_escapes kicks in.
322 Wget used to solve this by first decoding %-quotes, and then
323 encoding all the "unsafe" characters found in the resulting string.
324 This was wrong because it didn't preserve certain URL special
325 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
326 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
327 whether we considered `+' reserved (it is). One of these results
328 is inevitable because by the second step we would lose information
329 on whether the `+' was originally encoded or not. Both results
330 were wrong because in CGI parameters + means space, while %2B means
331 literal plus. reencode_escapes correctly translates the above to
332 "a%2B+b", i.e. returns the original string.
334 This function uses a modified version of the algorithm originally
335 proposed by Anon Sricharoenchai:
337 * Encode all "unsafe" characters, except those that are also
338 "reserved", to %XX. See urlchr_table for which characters are
341 * Encode the "%" characters not followed by two hex digits to
344 * Pass through all other characters and %XX escapes as-is. (Up to
345 Wget 1.10 this decoded %XX escapes corresponding to "safe"
346 characters, but that was obtrusive and broke some servers.)
350 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
352 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
356 "foo bar" -> "foo%20bar"
357 "foo%20bar" -> "foo%20bar"
358 "foo %20bar" -> "foo%20%20bar"
359 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
360 "foo%25%20bar" -> "foo%25%20bar"
361 "foo%2%20bar" -> "foo%252%20bar"
362 "foo+bar" -> "foo+bar" (plus is reserved!)
363 "foo%2b+bar" -> "foo%2b+bar" */
366 reencode_escapes (const char *s)
372 int encode_count = 0;
374 /* First pass: inspect the string to see if there's anything to do,
375 and to calculate the new length. */
376 for (p1 = s; *p1; p1++)
377 if (char_needs_escaping (p1))
381 /* The string is good as it is. */
382 return (char *) s; /* C const model sucks. */
385 /* Each encoding adds two characters (hex digits). */
386 newlen = oldlen + 2 * encode_count;
387 newstr = xmalloc (newlen + 1);
389 /* Second pass: copy the string to the destination address, encoding
390 chars when needed. */
395 if (char_needs_escaping (p1))
397 unsigned char c = *p1++;
399 *p2++ = XNUM_TO_DIGIT (c >> 4);
400 *p2++ = XNUM_TO_DIGIT (c & 0xf);
406 assert (p2 - newstr == newlen);
410 /* Returns the scheme type if the scheme is supported, or
411 SCHEME_INVALID if not. */
414 url_scheme (const char *url)
418 for (i = 0; supported_schemes[i].leading_string; i++)
419 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
420 strlen (supported_schemes[i].leading_string)))
422 if (!(supported_schemes[i].flags & scm_disabled))
423 return (enum url_scheme) i;
425 return SCHEME_INVALID;
428 return SCHEME_INVALID;
431 #define SCHEME_CHAR(ch) (c_isalnum (ch) || (ch) == '-' || (ch) == '+')
433 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
434 currently implemented, it returns true if URL begins with
438 url_has_scheme (const char *url)
442 /* The first char must be a scheme char. */
443 if (!*p || !SCHEME_CHAR (*p))
446 /* Followed by 0 or more scheme chars. */
447 while (*p && SCHEME_CHAR (*p))
449 /* Terminated by ':'. */
454 scheme_default_port (enum url_scheme scheme)
456 return supported_schemes[scheme].default_port;
460 scheme_disable (enum url_scheme scheme)
462 supported_schemes[scheme].flags |= scm_disabled;
465 /* Skip the username and password, if present in the URL. The
466 function should *not* be called with the complete URL, but with the
467 portion after the scheme.
469 If no username and password are found, return URL. */
472 url_skip_credentials (const char *url)
474 /* Look for '@' that comes before terminators, such as '/', '?',
476 const char *p = (const char *)strpbrk (url, "@/?#;");
482 /* Parse credentials contained in [BEG, END). The region is expected
483 to have come from a URL and is unescaped. */
486 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
492 return false; /* empty user name */
494 colon = memchr (beg, ':', end - beg);
496 return false; /* again empty user name */
500 *passwd = strdupdelim (colon + 1, end);
502 url_unescape (*passwd);
509 *user = strdupdelim (beg, userend);
510 url_unescape (*user);
514 /* Used by main.c: detect URLs written using the "shorthand" URL forms
515 originally popularized by Netscape and NcFTP. HTTP shorthands look
518 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
519 www.foo.com[:port] -> http://www.foo.com[:port]
521 FTP shorthands look like this:
523 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
524 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
526 If the URL needs not or cannot be rewritten, return NULL. */
529 rewrite_shorthand_url (const char *url)
534 if (url_scheme (url) != SCHEME_INVALID)
537 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
539 p = strpbrk (url, ":/");
543 /* If we're looking at "://", it means the URL uses a scheme we
544 don't support, which may include "https" when compiled without
545 SSL support. Don't bogusly rewrite such URLs. */
546 if (p && p[0] == ':' && p[1] == '/' && p[2] == '/')
551 /* Colon indicates ftp, as in foo.bar.com:path. Check for
552 special case of http port number ("localhost:10000"). */
553 int digits = strspn (p + 1, "0123456789");
554 if (digits && (p[1 + digits] == '/' || p[1 + digits] == '\0'))
557 /* Turn "foo.bar.com:path" to "ftp://foo.bar.com/path". */
558 ret = aprintf ("ftp://%s", url);
559 ret[6 + (p - url)] = '/';
564 /* Just prepend "http://" to URL. */
565 ret = aprintf ("http://%s", url);
570 static void split_path (const char *, char **, char **);
572 /* Like strpbrk, with the exception that it returns the pointer to the
573 terminating zero (end-of-string aka "eos") if no matching character
577 strpbrk_or_eos (const char *s, const char *accept)
579 char *p = strpbrk (s, accept);
581 p = strchr (s, '\0');
585 /* Turn STR into lowercase; return true if a character was actually
589 lowercase_str (char *str)
591 bool changed = false;
593 if (c_isupper (*str))
596 *str = c_tolower (*str);
602 init_seps (enum url_scheme scheme)
604 static char seps[8] = ":/";
606 int flags = supported_schemes[scheme].flags;
608 if (flags & scm_has_params)
610 if (flags & scm_has_query)
612 if (flags & scm_has_fragment)
618 static const char *parse_errors[] = {
619 #define PE_NO_ERROR 0
621 #define PE_UNSUPPORTED_SCHEME 1
622 N_("Unsupported scheme %s"),
623 #define PE_INVALID_HOST_NAME 2
624 N_("Invalid host name"),
625 #define PE_BAD_PORT_NUMBER 3
626 N_("Bad port number"),
627 #define PE_INVALID_USER_NAME 4
628 N_("Invalid user name"),
629 #define PE_UNTERMINATED_IPV6_ADDRESS 5
630 N_("Unterminated IPv6 numeric address"),
631 #define PE_IPV6_NOT_SUPPORTED 6
632 N_("IPv6 addresses not supported"),
633 #define PE_INVALID_IPV6_ADDRESS 7
634 N_("Invalid IPv6 numeric address")
639 Return a new struct url if successful, NULL on error. In case of
640 error, and if ERROR is not NULL, also set *ERROR to the appropriate
643 url_parse (const char *url, int *error, struct iri *iri)
647 bool path_modified, host_modified;
649 enum url_scheme scheme;
652 const char *uname_b, *uname_e;
653 const char *host_b, *host_e;
654 const char *path_b, *path_e;
655 const char *params_b, *params_e;
656 const char *query_b, *query_e;
657 const char *fragment_b, *fragment_e;
660 char *user = NULL, *passwd = NULL;
662 char *url_encoded = NULL, *new_url = NULL;
666 scheme = url_scheme (url);
667 if (scheme == SCHEME_INVALID)
669 error_code = PE_UNSUPPORTED_SCHEME;
673 if (iri && iri->utf8_encode)
675 url_unescape ((char *) url);
676 iri->utf8_encode = remote_to_utf8 (iri, url, (const char **) &new_url);
677 if (!iri->utf8_encode)
681 url_encoded = reencode_escapes (new_url ? new_url : url);
684 if (new_url && url_encoded != new_url)
687 p += strlen (supported_schemes[scheme].leading_string);
689 p = url_skip_credentials (p);
692 /* scheme://user:pass@host[:port]... */
695 /* We attempt to break down the URL into the components path,
696 params, query, and fragment. They are ordered like this:
698 scheme://host[:port][/path][;params][?query][#fragment] */
700 path_b = path_e = NULL;
701 params_b = params_e = NULL;
702 query_b = query_e = NULL;
703 fragment_b = fragment_e = NULL;
705 /* Initialize separators for optional parts of URL, depending on the
706 scheme. For example, FTP has params, and HTTP and HTTPS have
707 query string and fragment. */
708 seps = init_seps (scheme);
714 /* Handle IPv6 address inside square brackets. Ideally we'd
715 just look for the terminating ']', but rfc2732 mandates
716 rejecting invalid IPv6 addresses. */
718 /* The address begins after '['. */
720 host_e = strchr (host_b, ']');
724 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
729 /* Check if the IPv6 address is valid. */
730 if (!is_valid_ipv6_address(host_b, host_e))
732 error_code = PE_INVALID_IPV6_ADDRESS;
736 /* Continue parsing after the closing ']'. */
739 error_code = PE_IPV6_NOT_SUPPORTED;
743 /* The closing bracket must be followed by a separator or by the
745 /* http://[::1]... */
747 if (!strchr (seps, *p))
749 /* Trailing garbage after []-delimited IPv6 address. */
750 error_code = PE_INVALID_HOST_NAME;
756 p = strpbrk_or_eos (p, seps);
759 ++seps; /* advance to '/' */
761 if (host_b == host_e)
763 error_code = PE_INVALID_HOST_NAME;
767 port = scheme_default_port (scheme);
770 const char *port_b, *port_e, *pp;
772 /* scheme://host:port/tralala */
776 p = strpbrk_or_eos (p, seps);
779 /* Allow empty port, as per rfc2396. */
780 if (port_b != port_e)
781 for (port = 0, pp = port_b; pp < port_e; pp++)
783 if (!c_isdigit (*pp))
785 /* http://host:12randomgarbage/blah */
787 error_code = PE_BAD_PORT_NUMBER;
790 port = 10 * port + (*pp - '0');
791 /* Check for too large port numbers here, before we have
792 a chance to overflow on bogus port values. */
795 error_code = PE_BAD_PORT_NUMBER;
800 /* Advance to the first separator *after* '/' (either ';' or '?',
801 depending on the scheme). */
804 /* Get the optional parts of URL, each part being delimited by
805 current location and the position of the next separator. */
806 #define GET_URL_PART(sepchar, var) do { \
808 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
812 GET_URL_PART ('/', path);
813 if (supported_schemes[scheme].flags & scm_has_params)
814 GET_URL_PART (';', params);
815 if (supported_schemes[scheme].flags & scm_has_query)
816 GET_URL_PART ('?', query);
817 if (supported_schemes[scheme].flags & scm_has_fragment)
818 GET_URL_PART ('#', fragment);
823 if (uname_b != uname_e)
825 /* http://user:pass@host */
827 /* uname_b uname_e */
828 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
830 error_code = PE_INVALID_USER_NAME;
835 u = xnew0 (struct url);
837 u->host = strdupdelim (host_b, host_e);
842 u->path = strdupdelim (path_b, path_e);
843 path_modified = path_simplify (scheme, u->path);
844 split_path (u->path, &u->dir, &u->file);
846 host_modified = lowercase_str (u->host);
848 /* Decode %HH sequences in host name. This is important not so much
849 to support %HH sequences in host names (which other browser
850 don't), but to support binary characters (which will have been
851 converted to %HH by reencode_escapes). */
852 if (strchr (u->host, '%'))
854 url_unescape (u->host);
855 host_modified = true;
857 /* Apply IDNA regardless of iri->utf8_encode status */
858 if (opt.enable_iri && iri)
860 char *new = idn_encode (iri, u->host);
865 host_modified = true;
871 u->params = strdupdelim (params_b, params_e);
873 u->query = strdupdelim (query_b, query_e);
875 u->fragment = strdupdelim (fragment_b, fragment_e);
877 if (opt.enable_iri || path_modified || u->fragment || host_modified || path_b == path_e)
879 /* If we suspect that a transformation has rendered what
880 url_string might return different from URL_ENCODED, rebuild
881 u->url using url_string. */
882 u->url = url_string (u, URL_AUTH_SHOW);
884 if (url_encoded != url)
885 xfree ((char *) url_encoded);
889 if (url_encoded == url)
890 u->url = xstrdup (url);
892 u->url = url_encoded;
898 /* Cleanup in case of error: */
899 if (url_encoded && url_encoded != url)
902 /* Transmit the error code to the caller, if the caller wants to
909 /* Return the error message string from ERROR_CODE, which should have
910 been retrieved from url_parse. The error message is translated. */
913 url_error (const char *url, int error_code)
915 assert (error_code >= 0 && ((size_t) error_code) < countof (parse_errors));
917 if (error_code == PE_UNSUPPORTED_SCHEME)
920 char *scheme = xstrdup (url);
921 assert (url_has_scheme (url));
923 if ((p = strchr (scheme, ':')))
925 if (!strcasecmp (scheme, "https"))
926 asprintf (&error, _("HTTPS support not compiled in"));
928 asprintf (&error, _(parse_errors[error_code]), quote (scheme));
934 return xstrdup (_(parse_errors[error_code]));
937 /* Split PATH into DIR and FILE. PATH comes from the URL and is
938 expected to be URL-escaped.
940 The path is split into directory (the part up to the last slash)
941 and file (the part after the last slash), which are subsequently
945 "foo/bar/baz" "foo/bar" "baz"
946 "foo/bar/" "foo/bar" ""
948 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
950 DIR and FILE are freshly allocated. */
953 split_path (const char *path, char **dir, char **file)
955 char *last_slash = strrchr (path, '/');
959 *file = xstrdup (path);
963 *dir = strdupdelim (path, last_slash);
964 *file = xstrdup (last_slash + 1);
967 url_unescape (*file);
970 /* Note: URL's "full path" is the path with the query string and
971 params appended. The "fragment" (#foo) is intentionally ignored,
972 but that might be changed. For example, if the original URL was
973 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
974 the full path will be "/foo/bar/baz;bullshit?querystring". */
976 /* Return the length of the full path, without the terminating
980 full_path_length (const struct url *url)
984 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
995 /* Write out the full path. */
998 full_path_write (const struct url *url, char *where)
1000 #define FROB(el, chr) do { \
1001 char *f_el = url->el; \
1003 int l = strlen (f_el); \
1005 memcpy (where, f_el, l); \
1017 /* Public function for getting the "full path". E.g. if u->path is
1018 "foo/bar" and u->query is "param=value", full_path will be
1019 "/foo/bar?param=value". */
1022 url_full_path (const struct url *url)
1024 int length = full_path_length (url);
1025 char *full_path = xmalloc (length + 1);
1027 full_path_write (url, full_path);
1028 full_path[length] = '\0';
1033 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1034 escaping of certain characters, such as "/" and ":". Returns a
1035 count of unescaped chars. */
1038 unescape_single_char (char *str, char chr)
1040 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1041 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1042 char *h = str; /* hare */
1043 char *t = str; /* tortoise */
1044 for (; *h; h++, t++)
1046 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1057 /* Escape unsafe and reserved characters, except for the slash
1061 url_escape_dir (const char *dir)
1063 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1067 unescape_single_char (newdir, '/');
1071 /* Sync u->path and u->url with u->dir and u->file. Called after
1072 u->file or u->dir have been changed, typically by the FTP code. */
1075 sync_path (struct url *u)
1077 char *newpath, *efile, *edir;
1081 /* u->dir and u->file are not escaped. URL-escape them before
1082 reassembling them into u->path. That way, if they contain
1083 separators like '?' or even if u->file contains slashes, the
1084 path will be correctly assembled. (u->file can contain slashes
1085 if the URL specifies it with %2f, or if an FTP server returns
1087 edir = url_escape_dir (u->dir);
1088 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1091 newpath = xstrdup (efile);
1094 int dirlen = strlen (edir);
1095 int filelen = strlen (efile);
1097 /* Copy "DIR/FILE" to newpath. */
1098 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1099 memcpy (p, edir, dirlen);
1102 memcpy (p, efile, filelen);
1111 if (efile != u->file)
1114 /* Regenerate u->url as well. */
1116 u->url = url_string (u, URL_AUTH_SHOW);
1119 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1120 This way we can sync u->path and u->url when they get changed. */
1123 url_set_dir (struct url *url, const char *newdir)
1126 url->dir = xstrdup (newdir);
1131 url_set_file (struct url *url, const char *newfile)
1134 url->file = xstrdup (newfile);
1139 url_free (struct url *url)
1145 xfree_null (url->params);
1146 xfree_null (url->query);
1147 xfree_null (url->fragment);
1148 xfree_null (url->user);
1149 xfree_null (url->passwd);
1157 /* Create all the necessary directories for PATH (a file). Calls
1158 make_directory internally. */
1160 mkalldirs (const char *path)
1167 p = path + strlen (path);
1168 for (; *p != '/' && p != path; p--)
1171 /* Don't create if it's just a file. */
1172 if ((p == path) && (*p != '/'))
1174 t = strdupdelim (path, p);
1176 /* Check whether the directory exists. */
1177 if ((stat (t, &st) == 0))
1179 if (S_ISDIR (st.st_mode))
1186 /* If the dir exists as a file name, remove it first. This
1187 is *only* for Wget to work with buggy old CERN http
1188 servers. Here is the scenario: When Wget tries to
1189 retrieve a directory without a slash, e.g.
1190 http://foo/bar (bar being a directory), CERN server will
1191 not redirect it too http://foo/bar/ -- it will generate a
1192 directory listing containing links to bar/file1,
1193 bar/file2, etc. Wget will lose because it saves this
1194 HTML listing to a file `bar', so it cannot create the
1195 directory. To work around this, if the file of the same
1196 name exists, we just remove it and create the directory
1198 DEBUGP (("Removing %s because of directory danger!\n", t));
1202 res = make_directory (t);
1204 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1209 /* Functions for constructing the file name out of URL components. */
1211 /* A growable string structure, used by url_file_name and friends.
1212 This should perhaps be moved to utils.c.
1214 The idea is to have a convenient and efficient way to construct a
1215 string by having various functions append data to it. Instead of
1216 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1217 functions in questions, we pass the pointer to this struct. */
1225 /* Ensure that the string can accept APPEND_COUNT more characters past
1226 the current TAIL position. If necessary, this will grow the string
1227 and update its allocated size. If the string is already large
1228 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1229 #define GROW(g, append_size) do { \
1230 struct growable *G_ = g; \
1231 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1234 /* Return the tail position of the string. */
1235 #define TAIL(r) ((r)->base + (r)->tail)
1237 /* Move the tail position by APPEND_COUNT characters. */
1238 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1240 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1244 append_string (const char *str, struct growable *dest)
1246 int l = strlen (str);
1248 memcpy (TAIL (dest), str, l);
1249 TAIL_INCR (dest, l);
1252 /* Append CH to DEST. For example, append_char (0, DEST)
1253 zero-terminates DEST. */
1256 append_char (char ch, struct growable *dest)
1260 TAIL_INCR (dest, 1);
1264 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1265 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1266 filechr_control = 4 /* a control character, e.g. 0-31 */
1269 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1271 /* Shorthands for the table: */
1272 #define U filechr_not_unix
1273 #define W filechr_not_windows
1274 #define C filechr_control
1279 /* Table of characters unsafe under various conditions (see above).
1281 Arguably we could also claim `%' to be unsafe, since we use it as
1282 the escape character. If we ever want to be able to reliably
1283 translate file name back to URL, this would become important
1284 crucial. Right now, it's better to be minimal in escaping. */
1286 static const unsigned char filechr_table[256] =
1288 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1289 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1290 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1291 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1292 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1293 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1294 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1295 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1296 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1297 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1298 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1299 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1300 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1301 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1302 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1303 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1305 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1306 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1307 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1308 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1310 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1311 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1312 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1313 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1321 /* FN_PORT_SEP is the separator between host and port in file names
1322 for non-standard port numbers. On Unix this is normally ':', as in
1323 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1324 because Windows can't handle ':' in file names. */
1325 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1327 /* FN_QUERY_SEP is the separator between the file name and the URL
1328 query, normally '?'. Since Windows cannot handle '?' as part of
1329 file name, we use '@' instead there. */
1330 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1332 /* Quote path element, characters in [b, e), as file name, and append
1333 the quoted string to DEST. Each character is quoted as per
1334 file_unsafe_char and the corresponding table.
1336 If ESCAPED is true, the path element is considered to be
1337 URL-escaped and will be unescaped prior to inspection. */
1340 append_uri_pathel (const char *b, const char *e, bool escaped,
1341 struct growable *dest)
1347 if (opt.restrict_files_os == restrict_unix)
1348 mask = filechr_not_unix;
1350 mask = filechr_not_windows;
1351 if (opt.restrict_files_ctrl)
1352 mask |= filechr_control;
1354 /* Copy [b, e) to PATHEL and URL-unescape it. */
1358 BOUNDED_TO_ALLOCA (b, e, unescaped);
1359 url_unescape (unescaped);
1361 e = unescaped + strlen (unescaped);
1364 /* Defang ".." when found as component of path. Remember that path
1365 comes from the URL and might contain malicious input. */
1366 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1372 /* Walk the PATHEL string and check how many characters we'll need
1375 for (p = b; p < e; p++)
1376 if (FILE_CHAR_TEST (*p, mask))
1379 /* Calculate the length of the output string. e-b is the input
1380 string length. Each quoted char introduces two additional
1381 characters in the string, hence 2*quoted. */
1382 outlen = (e - b) + (2 * quoted);
1383 GROW (dest, outlen);
1387 /* If there's nothing to quote, we can simply append the string
1388 without processing it again. */
1389 memcpy (TAIL (dest), b, outlen);
1393 char *q = TAIL (dest);
1394 for (p = b; p < e; p++)
1396 if (!FILE_CHAR_TEST (*p, mask))
1400 unsigned char ch = *p;
1402 *q++ = XNUM_TO_DIGIT (ch >> 4);
1403 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1406 assert (q - TAIL (dest) == outlen);
1409 /* Perform inline case transformation if required. */
1410 if (opt.restrict_files_case == restrict_lowercase
1411 || opt.restrict_files_case == restrict_uppercase)
1414 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1416 if (opt.restrict_files_case == restrict_lowercase)
1417 *q = c_tolower (*q);
1419 *q = c_toupper (*q);
1423 TAIL_INCR (dest, outlen);
1426 /* Append to DEST the directory structure that corresponds the
1427 directory part of URL's path. For example, if the URL is
1428 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1430 Each path element ("dir1" and "dir2" in the above example) is
1431 examined, url-unescaped, and re-escaped as file name element.
1433 Additionally, it cuts as many directories from the path as
1434 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1435 will produce "bar" for the above example. For 2 or more, it will
1438 Each component of the path is quoted for use as file name. */
1441 append_dir_structure (const struct url *u, struct growable *dest)
1443 char *pathel, *next;
1444 int cut = opt.cut_dirs;
1446 /* Go through the path components, de-URL-quote them, and quote them
1447 (if necessary) as file names. */
1450 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1455 /* Ignore empty pathels. */
1459 append_char ('/', dest);
1460 append_uri_pathel (pathel, next, true, dest);
1464 /* Return a unique file name that matches the given URL as good as
1465 possible. Does not create directories on the file system. */
1468 url_file_name (const struct url *u)
1470 struct growable fnres; /* stands for "file name result" */
1472 const char *u_file, *u_query;
1473 char *fname, *unique;
1474 char *index_filename = "index.html"; /* The default index file is index.html */
1480 /* If an alternative index file was defined, change index_filename */
1481 if (opt.default_page)
1482 index_filename = opt.default_page;
1485 /* Start with the directory prefix, if specified. */
1487 append_string (opt.dir_prefix, &fnres);
1489 /* If "dirstruct" is turned on (typically the case with -r), add
1490 the host and port (unless those have been turned off) and
1491 directory structure. */
1494 if (opt.protocol_directories)
1497 append_char ('/', &fnres);
1498 append_string (supported_schemes[u->scheme].name, &fnres);
1500 if (opt.add_hostdir)
1503 append_char ('/', &fnres);
1504 if (0 != strcmp (u->host, ".."))
1505 append_string (u->host, &fnres);
1507 /* Host name can come from the network; malicious DNS may
1508 allow ".." to be resolved, causing us to write to
1509 "../<file>". Defang such host names. */
1510 append_string ("%2E%2E", &fnres);
1511 if (u->port != scheme_default_port (u->scheme))
1514 number_to_string (portstr, u->port);
1515 append_char (FN_PORT_SEP, &fnres);
1516 append_string (portstr, &fnres);
1520 append_dir_structure (u, &fnres);
1523 /* Add the file name. */
1525 append_char ('/', &fnres);
1526 u_file = *u->file ? u->file : index_filename;
1527 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1529 /* Append "?query" to the file name. */
1530 u_query = u->query && *u->query ? u->query : NULL;
1533 append_char (FN_QUERY_SEP, &fnres);
1534 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1537 /* Zero-terminate the file name. */
1538 append_char ('\0', &fnres);
1542 /* Check the cases in which the unique extensions are not used:
1543 1) Clobbering is turned off (-nc).
1544 2) Retrieval with regetting.
1545 3) Timestamping is used.
1546 4) Hierarchy is built.
1548 The exception is the case when file does exist and is a
1549 directory (see `mkalldirs' for explanation). */
1551 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1552 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1555 unique = unique_name (fname, true);
1556 if (unique != fname)
1561 /* Resolve "." and ".." elements of PATH by destructively modifying
1562 PATH and return true if PATH has been modified, false otherwise.
1564 The algorithm is in spirit similar to the one described in rfc1808,
1565 although implemented differently, in one pass. To recap, path
1566 elements containing only "." are removed, and ".." is taken to mean
1567 "back up one element". Single leading and trailing slashes are
1570 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1571 test examples are provided below. If you change anything in this
1572 function, run test_path_simplify to make sure you haven't broken a
1576 path_simplify (enum url_scheme scheme, char *path)
1578 char *h = path; /* hare */
1579 char *t = path; /* tortoise */
1581 char *end = strchr (path, '\0');
1585 /* Hare should be at the beginning of a path element. */
1587 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1592 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1594 /* Handle "../" by retreating the tortoise by one path
1595 element -- but not past beggining. */
1598 /* Move backwards until T hits the beginning of the
1599 previous path element or the beginning of path. */
1600 for (--t; t > beg && t[-1] != '/'; t--)
1603 else if (scheme == SCHEME_FTP)
1605 /* If we're at the beginning, copy the "../" literally
1606 and move the beginning so a later ".." doesn't remove
1607 it. This violates RFC 3986; but we do it for FTP
1608 anyway because there is otherwise no way to get at a
1609 parent directory, when the FTP server drops us in a
1610 non-root directory (which is not uncommon). */
1619 /* A regular path element. If H hasn't advanced past T,
1620 simply skip to the next path element. Otherwise, copy
1621 the path element until the next slash. */
1624 /* Skip the path element, including the slash. */
1625 while (h < end && *h != '/')
1632 /* Copy the path element, including the final slash. */
1633 while (h < end && *h != '/')
1647 /* Return the length of URL's path. Path is considered to be
1648 terminated by one or more of the ?query or ;params or #fragment,
1649 depending on the scheme. */
1652 path_end (const char *url)
1654 enum url_scheme scheme = url_scheme (url);
1656 if (scheme == SCHEME_INVALID)
1657 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1658 /* +2 to ignore the first two separators ':' and '/' */
1659 seps = init_seps (scheme) + 2;
1660 return strpbrk_or_eos (url, seps);
1663 /* Find the last occurrence of character C in the range [b, e), or
1664 NULL, if none are present. */
1665 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1667 /* Merge BASE with LINK and return the resulting URI.
1669 Either of the URIs may be absolute or relative, complete with the
1670 host name, or path only. This tries to reasonably handle all
1671 foreseeable cases. It only employs minimal URL parsing, without
1672 knowledge of the specifics of schemes.
1674 I briefly considered making this function call path_simplify after
1675 the merging process, as rfc1738 seems to suggest. This is a bad
1676 idea for several reasons: 1) it complexifies the code, and 2)
1677 url_parse has to simplify path anyway, so it's wasteful to boot. */
1680 uri_merge (const char *base, const char *link)
1686 if (url_has_scheme (link))
1687 return xstrdup (link);
1689 /* We may not examine BASE past END. */
1690 end = path_end (base);
1691 linklength = strlen (link);
1695 /* Empty LINK points back to BASE, query string and all. */
1696 return xstrdup (base);
1698 else if (*link == '?')
1700 /* LINK points to the same location, but changes the query
1701 string. Examples: */
1702 /* uri_merge("path", "?new") -> "path?new" */
1703 /* uri_merge("path?foo", "?new") -> "path?new" */
1704 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1705 /* uri_merge("path#foo", "?new") -> "path?new" */
1706 int baselength = end - base;
1707 merge = xmalloc (baselength + linklength + 1);
1708 memcpy (merge, base, baselength);
1709 memcpy (merge + baselength, link, linklength);
1710 merge[baselength + linklength] = '\0';
1712 else if (*link == '#')
1714 /* uri_merge("path", "#new") -> "path#new" */
1715 /* uri_merge("path#foo", "#new") -> "path#new" */
1716 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1717 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1719 const char *end1 = strchr (base, '#');
1721 end1 = base + strlen (base);
1722 baselength = end1 - base;
1723 merge = xmalloc (baselength + linklength + 1);
1724 memcpy (merge, base, baselength);
1725 memcpy (merge + baselength, link, linklength);
1726 merge[baselength + linklength] = '\0';
1728 else if (*link == '/' && *(link + 1) == '/')
1730 /* LINK begins with "//" and so is a net path: we need to
1731 replace everything after (and including) the double slash
1734 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1735 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1736 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1740 const char *start_insert;
1742 /* Look for first slash. */
1743 slash = memchr (base, '/', end - base);
1744 /* If found slash and it is a double slash, then replace
1745 from this point, else default to replacing from the
1747 if (slash && *(slash + 1) == '/')
1748 start_insert = slash;
1750 start_insert = base;
1752 span = start_insert - base;
1753 merge = xmalloc (span + linklength + 1);
1755 memcpy (merge, base, span);
1756 memcpy (merge + span, link, linklength);
1757 merge[span + linklength] = '\0';
1759 else if (*link == '/')
1761 /* LINK is an absolute path: we need to replace everything
1762 after (and including) the FIRST slash with LINK.
1764 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1765 "/qux/xyzzy", our result should be
1766 "http://host/qux/xyzzy". */
1769 const char *start_insert = NULL; /* for gcc to shut up. */
1770 const char *pos = base;
1771 bool seen_slash_slash = false;
1772 /* We're looking for the first slash, but want to ignore
1775 slash = memchr (pos, '/', end - pos);
1776 if (slash && !seen_slash_slash)
1777 if (*(slash + 1) == '/')
1780 seen_slash_slash = true;
1784 /* At this point, SLASH is the location of the first / after
1785 "//", or the first slash altogether. START_INSERT is the
1786 pointer to the location where LINK will be inserted. When
1787 examining the last two examples, keep in mind that LINK
1790 if (!slash && !seen_slash_slash)
1791 /* example: "foo" */
1793 start_insert = base;
1794 else if (!slash && seen_slash_slash)
1795 /* example: "http://foo" */
1798 else if (slash && !seen_slash_slash)
1799 /* example: "foo/bar" */
1801 start_insert = base;
1802 else if (slash && seen_slash_slash)
1803 /* example: "http://something/" */
1805 start_insert = slash;
1807 span = start_insert - base;
1808 merge = xmalloc (span + linklength + 1);
1810 memcpy (merge, base, span);
1811 memcpy (merge + span, link, linklength);
1812 merge[span + linklength] = '\0';
1816 /* LINK is a relative URL: we need to replace everything
1817 after last slash (possibly empty) with LINK.
1819 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1820 our result should be "whatever/foo/qux/xyzzy". */
1821 bool need_explicit_slash = false;
1823 const char *start_insert;
1824 const char *last_slash = find_last_char (base, end, '/');
1827 /* No slash found at all. Replace what we have with LINK. */
1828 start_insert = base;
1830 else if (last_slash && last_slash >= base + 2
1831 && last_slash[-2] == ':' && last_slash[-1] == '/')
1833 /* example: http://host" */
1835 start_insert = end + 1;
1836 need_explicit_slash = true;
1840 /* example: "whatever/foo/bar" */
1842 start_insert = last_slash + 1;
1845 span = start_insert - base;
1846 merge = xmalloc (span + linklength + 1);
1848 memcpy (merge, base, span);
1849 if (need_explicit_slash)
1850 merge[span - 1] = '/';
1851 memcpy (merge + span, link, linklength);
1852 merge[span + linklength] = '\0';
1858 #define APPEND(p, s) do { \
1859 int len = strlen (s); \
1860 memcpy (p, s, len); \
1864 /* Use this instead of password when the actual password is supposed
1865 to be hidden. We intentionally use a generic string without giving
1866 away the number of characters in the password, like previous
1868 #define HIDDEN_PASSWORD "*password*"
1870 /* Recreate the URL string from the data in URL.
1872 If HIDE is true (as it is when we're calling this on a URL we plan
1873 to print, but not when calling it to canonicalize a URL for use
1874 within the program), password will be hidden. Unsafe characters in
1875 the URL will be quoted. */
1878 url_string (const struct url *url, enum url_auth_mode auth_mode)
1882 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1884 int scheme_port = supported_schemes[url->scheme].default_port;
1885 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1886 int fplen = full_path_length (url);
1888 bool brackets_around_host;
1890 assert (scheme_str != NULL);
1892 /* Make sure the user name and password are quoted. */
1895 if (auth_mode != URL_AUTH_HIDE)
1897 quoted_user = url_escape_allow_passthrough (url->user);
1900 if (auth_mode == URL_AUTH_HIDE_PASSWD)
1901 quoted_passwd = HIDDEN_PASSWORD;
1903 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1908 /* In the unlikely event that the host name contains non-printable
1909 characters, quote it for displaying to the user. */
1910 quoted_host = url_escape_allow_passthrough (url->host);
1912 /* Undo the quoting of colons that URL escaping performs. IPv6
1913 addresses may legally contain colons, and in that case must be
1914 placed in square brackets. */
1915 if (quoted_host != url->host)
1916 unescape_single_char (quoted_host, ':');
1917 brackets_around_host = strchr (quoted_host, ':') != NULL;
1919 size = (strlen (scheme_str)
1920 + strlen (quoted_host)
1921 + (brackets_around_host ? 2 : 0)
1924 if (url->port != scheme_port)
1925 size += 1 + numdigit (url->port);
1928 size += 1 + strlen (quoted_user);
1930 size += 1 + strlen (quoted_passwd);
1933 p = result = xmalloc (size);
1935 APPEND (p, scheme_str);
1938 APPEND (p, quoted_user);
1942 APPEND (p, quoted_passwd);
1947 if (brackets_around_host)
1949 APPEND (p, quoted_host);
1950 if (brackets_around_host)
1952 if (url->port != scheme_port)
1955 p = number_to_string (p, url->port);
1958 full_path_write (url, p);
1962 assert (p - result == size);
1964 if (quoted_user && quoted_user != url->user)
1965 xfree (quoted_user);
1966 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
1967 && quoted_passwd != url->passwd)
1968 xfree (quoted_passwd);
1969 if (quoted_host != url->host)
1970 xfree (quoted_host);
1975 /* Return true if scheme a is similar to scheme b.
1977 Schemes are similar if they are equal. If SSL is supported, schemes
1978 are also similar if one is http (SCHEME_HTTP) and the other is https
1981 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
1986 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
1987 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
1994 getchar_from_escaped_string (const char *str, char *c)
1996 const char *p = str;
1998 assert (str && *str);
2003 if (!c_isxdigit(p[1]) || !c_isxdigit(p[2]))
2011 return 0; /* error: invalid string */
2013 *c = X2DIGITS_TO_NUM (p[1], p[2]);
2014 if (URL_RESERVED_CHAR(*c))
2032 are_urls_equal (const char *u1, const char *u2)
2043 && (pp = getchar_from_escaped_string (p, &ch1))
2044 && (qq = getchar_from_escaped_string (q, &ch2))
2045 && (c_tolower(ch1) == c_tolower(ch2)))
2051 return (*p == 0 && *q == 0 ? true : false);
2055 /* Debugging and testing support for path_simplify. */
2058 /* Debug: run path_simplify on PATH and return the result in a new
2059 string. Useful for calling from the debugger. */
2063 char *copy = xstrdup (path);
2064 path_simplify (copy);
2070 run_test (char *test, char *expected_result, enum url_scheme scheme,
2071 bool expected_change)
2073 char *test_copy = xstrdup (test);
2074 bool modified = path_simplify (scheme, 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);
2082 if (modified != expected_change)
2084 if (expected_change)
2085 printf ("Expected modification with path_simplify(\"%s\").\n",
2088 printf ("Expected no modification with path_simplify(\"%s\").\n",
2092 mu_assert ("", modified == expected_change);
2097 test_path_simplify (void)
2100 char *test, *result;
2101 enum url_scheme scheme;
2104 { "", "", SCHEME_HTTP, false },
2105 { ".", "", SCHEME_HTTP, true },
2106 { "./", "", SCHEME_HTTP, true },
2107 { "..", "", SCHEME_HTTP, true },
2108 { "../", "", SCHEME_HTTP, true },
2109 { "..", "..", SCHEME_FTP, false },
2110 { "../", "../", SCHEME_FTP, false },
2111 { "foo", "foo", SCHEME_HTTP, false },
2112 { "foo/bar", "foo/bar", SCHEME_HTTP, false },
2113 { "foo///bar", "foo///bar", SCHEME_HTTP, false },
2114 { "foo/.", "foo/", SCHEME_HTTP, true },
2115 { "foo/./", "foo/", SCHEME_HTTP, true },
2116 { "foo./", "foo./", SCHEME_HTTP, false },
2117 { "foo/../bar", "bar", SCHEME_HTTP, true },
2118 { "foo/../bar/", "bar/", SCHEME_HTTP, true },
2119 { "foo/bar/..", "foo/", SCHEME_HTTP, true },
2120 { "foo/bar/../x", "foo/x", SCHEME_HTTP, true },
2121 { "foo/bar/../x/", "foo/x/", SCHEME_HTTP, true },
2122 { "foo/..", "", SCHEME_HTTP, true },
2123 { "foo/../..", "", SCHEME_HTTP, true },
2124 { "foo/../../..", "", SCHEME_HTTP, true },
2125 { "foo/../../bar/../../baz", "baz", SCHEME_HTTP, true },
2126 { "foo/../..", "..", SCHEME_FTP, true },
2127 { "foo/../../..", "../..", SCHEME_FTP, true },
2128 { "foo/../../bar/../../baz", "../../baz", SCHEME_FTP, true },
2129 { "a/b/../../c", "c", SCHEME_HTTP, true },
2130 { "./a/../b", "b", SCHEME_HTTP, true }
2134 for (i = 0; i < countof (tests); i++)
2136 const char *message;
2137 char *test = tests[i].test;
2138 char *expected_result = tests[i].result;
2139 enum url_scheme scheme = tests[i].scheme;
2140 bool expected_change = tests[i].should_modify;
2141 message = run_test (test, expected_result, scheme, expected_change);
2142 if (message) return message;
2148 test_append_uri_pathel()
2155 char *expected_result;
2157 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2160 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2162 struct growable dest;
2163 const char *p = test_array[i].input;
2165 memset (&dest, 0, sizeof (dest));
2167 append_string (test_array[i].original_url, &dest);
2168 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2169 append_char ('\0', &dest);
2171 mu_assert ("test_append_uri_pathel: wrong result",
2172 strcmp (dest.base, test_array[i].expected_result) == 0);
2179 test_are_urls_equal()
2185 bool expected_result;
2187 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2188 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2189 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2190 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2191 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2192 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2195 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2197 mu_assert ("test_are_urls_equal: wrong result",
2198 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2204 #endif /* TESTING */