X-Git-Url: http://sjero.net/git/?p=wget;a=blobdiff_plain;f=src%2Furl.c;h=31148735dc9fbd0fa13cbbcb45643bf0db43abb6;hp=d27a902c0b82b441feacf5fb7bb259c573999d6d;hb=38a7829dcb4eb5dba28dbf0f05c6a80fea9217f8;hpb=dadde97838e5e739d8c41022b65fea56aae0e99f diff --git a/src/url.c b/src/url.c index d27a902c..1edfbae3 100644 --- a/src/url.c +++ b/src/url.c @@ -1,12 +1,13 @@ /* URL handling. - Copyright (C) 1995, 1996, 1997, 2000, 2001, 2003, 2003 - Free Software Foundation, Inc. + Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, + 2005, 2006, 2007, 2008, 2009, 2010, 2011 Free Software Foundation, + Inc. This file is part of GNU Wget. GNU Wget is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by -the Free Software Foundation; either version 2 of the License, or (at +the Free Software Foundation; either version 3 of the License, or (at your option) any later version. GNU Wget is distributed in the hope that it will be useful, @@ -15,76 +16,105 @@ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License -along with Wget; if not, write to the Free Software -Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. +along with Wget. If not, see . -In addition, as a special exception, the Free Software Foundation -gives permission to link the code of its release of Wget with the -OpenSSL project's "OpenSSL" library (or with modified versions of it -that use the same license as the "OpenSSL" library), and distribute -the linked executables. You must obey the GNU General Public License -in all respects for all of the code used other than "OpenSSL". If you -modify this file, you may extend this exception to your version of the -file, but you are not obligated to do so. If you do not wish to do -so, delete this exception statement from your version. */ +Additional permission under GNU GPL version 3 section 7 -#include +If you modify this program, or any covered work, by linking or +combining it with the OpenSSL project's OpenSSL library (or a +modified version of that library), containing parts covered by the +terms of the OpenSSL or SSLeay licenses, the Free Software Foundation +grants you additional permission to convey the resulting work. +Corresponding Source for a non-source form of such a combination +shall include the source code for the parts of OpenSSL used as well +as that of the covered work. */ + +#include "wget.h" #include #include -#ifdef HAVE_STRING_H -# include -#else -# include -#endif -#include -#ifdef HAVE_UNISTD_H -# include -#endif +#include +#include #include #include -#include "wget.h" #include "utils.h" #include "url.h" +#include "host.h" /* for is_valid_ipv6_address */ + +#ifdef __VMS +#include "vms.h" +#endif /* def __VMS */ -#ifndef errno -extern int errno; +#ifdef TESTING +#include "test.h" #endif +enum { + scm_disabled = 1, /* for https when OpenSSL fails to init. */ + scm_has_params = 2, /* whether scheme has ;params */ + scm_has_query = 4, /* whether scheme has ?query */ + scm_has_fragment = 8 /* whether scheme has #fragment */ +}; + struct scheme_data { - char *leading_string; + /* Short name of the scheme, such as "http" or "ftp". */ + const char *name; + /* Leading string that identifies the scheme, such as "https://". */ + const char *leading_string; + /* Default port of the scheme when none is specified. */ int default_port; - int enabled; + /* Various flags. */ + int flags; }; /* Supported schemes: */ static struct scheme_data supported_schemes[] = { - { "http://", DEFAULT_HTTP_PORT, 1 }, + { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment }, #ifdef HAVE_SSL - { "https://", DEFAULT_HTTPS_PORT, 1 }, + { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment }, #endif - { "ftp://", DEFAULT_FTP_PORT, 1 }, + { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment }, /* SCHEME_INVALID */ - { NULL, -1, 0 } + { NULL, NULL, -1, 0 } }; /* Forward declarations: */ -static int path_simplify PARAMS ((char *)); +static bool path_simplify (enum url_scheme, char *); -/* Support for encoding and decoding of URL strings. We determine - whether a character is unsafe through static table lookup. This - code assumes ASCII character set and 8-bit chars. */ +/* Support for escaping and unescaping of URL strings. */ + +/* Table of "reserved" and "unsafe" characters. Those terms are + rfc1738-speak, as such largely obsoleted by rfc2396 and later + specs, but the general idea remains. + + A reserved character is the one that you can't decode without + changing the meaning of the URL. For example, you can't decode + "/foo/%2f/bar" into "/foo///bar" because the number and contents of + path components is different. Non-reserved characters can be + changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The + unsafe characters are loosely based on rfc1738, plus "$" and ",", + as recommended by rfc2396, and minus "~", which is very frequently + used (and sometimes unrecognized as %7E by broken servers). + + An unsafe character is the one that should be encoded when URLs are + placed in foreign environments. E.g. space and newline are unsafe + in HTTP contexts because HTTP uses them as separator and line + terminator, so they must be encoded to %20 and %0A respectively. + "*" is unsafe in shell context, etc. + + We determine whether a character is unsafe through static table + lookup. This code assumes ASCII character set and 8-bit chars. */ enum { - /* rfc1738 reserved chars, preserved from encoding. */ + /* rfc1738 reserved chars + "$" and ",". */ urlchr_reserved = 1, - /* rfc1738 unsafe chars, plus some more. */ + /* rfc1738 unsafe chars, plus non-printables. */ urlchr_unsafe = 2 }; @@ -97,14 +127,14 @@ enum { #define U urlchr_unsafe #define RU R|U -const static unsigned char urlchr_table[256] = +static const unsigned char urlchr_table[256] = { U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */ U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */ U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */ U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */ - U, 0, U, RU, 0, U, R, 0, /* SP ! " # $ % & ' */ - 0, 0, 0, R, 0, 0, 0, R, /* ( ) * + , - . / */ + U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */ + 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */ 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */ 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */ RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */ @@ -114,7 +144,7 @@ const static unsigned char urlchr_table[256] = U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */ 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */ 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */ - 0, 0, 0, U, U, U, U, U, /* x y z { | } ~ DEL */ + 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */ U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, @@ -139,27 +169,33 @@ const static unsigned char urlchr_table[256] = The transformation is done in place. If you need the original string intact, make a copy before calling this function. */ -static void +void url_unescape (char *s) { - char *t = s; /* t - tortoise */ - char *h = s; /* h - hare */ + char *t = s; /* t - tortoise */ + char *h = s; /* h - hare */ for (; *h; h++, t++) { if (*h != '%') - { - copychar: - *t = *h; - } + { + copychar: + *t = *h; + } else - { - /* Do nothing if '%' is not followed by two hex digits. */ - if (!h[1] || !h[2] || !(ISXDIGIT (h[1]) && ISXDIGIT (h[2]))) - goto copychar; - *t = X2DIGITS_TO_NUM (h[1], h[2]); - h += 2; - } + { + char c; + /* Do nothing if '%' is not followed by two hex digits. */ + if (!h[1] || !h[2] || !(c_isxdigit (h[1]) && c_isxdigit (h[2]))) + goto copychar; + c = X2DIGITS_TO_NUM (h[1], h[2]); + /* Don't unescape %00 because there is no way to insert it + into a C string without effectively truncating it. */ + if (c == '\0') + goto copychar; + *t = c; + h += 2; + } } *t = '\0'; } @@ -167,12 +203,12 @@ url_unescape (char *s) /* The core of url_escape_* functions. Escapes the characters that match the provided mask in urlchr_table. - If ALLOW_PASSTHROUGH is non-zero, a string with no unsafe chars - will be returned unchanged. If ALLOW_PASSTHROUGH is zero, a - freshly allocated string will be returned in all cases. */ + If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be + returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly + allocated string will be returned in all cases. */ static char * -url_escape_1 (const char *s, unsigned char mask, int allow_passthrough) +url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough) { const char *p1; char *p2, *newstr; @@ -181,13 +217,13 @@ url_escape_1 (const char *s, unsigned char mask, int allow_passthrough) for (p1 = s; *p1; p1++) if (urlchr_test (*p1, mask)) - addition += 2; /* Two more characters (hex digits) */ + addition += 2; /* Two more characters (hex digits) */ if (!addition) return allow_passthrough ? (char *)s : xstrdup (s); newlen = (p1 - s) + addition; - newstr = (char *)xmalloc (newlen + 1); + newstr = xmalloc (newlen + 1); p1 = s; p2 = newstr; @@ -195,14 +231,14 @@ url_escape_1 (const char *s, unsigned char mask, int allow_passthrough) { /* Quote the characters that match the test mask. */ if (urlchr_test (*p1, mask)) - { - unsigned char c = *p1++; - *p2++ = '%'; - *p2++ = XNUM_TO_digit (c >> 4); - *p2++ = XNUM_TO_digit (c & 0xf); - } + { + unsigned char c = *p1++; + *p2++ = '%'; + *p2++ = XNUM_TO_DIGIT (c >> 4); + *p2++ = XNUM_TO_DIGIT (c & 0xf); + } else - *p2++ = *p1++; + *p2++ = *p1++; } assert (p2 - newstr == newlen); *p2 = '\0'; @@ -216,7 +252,16 @@ url_escape_1 (const char *s, unsigned char mask, int allow_passthrough) char * url_escape (const char *s) { - return url_escape_1 (s, urlchr_unsafe, 0); + return url_escape_1 (s, urlchr_unsafe, false); +} + +/* URL-escape the unsafe and reserved characters (see urlchr_table) in + a given string, returning a freshly allocated string. */ + +char * +url_escape_unsafe_and_reserved (const char *s) +{ + return url_escape_1 (s, urlchr_unsafe|urlchr_reserved, false); } /* URL-escape the unsafe characters (see urlchr_table) in a given @@ -225,37 +270,30 @@ url_escape (const char *s) static char * url_escape_allow_passthrough (const char *s) { - return url_escape_1 (s, urlchr_unsafe, 1); + return url_escape_1 (s, urlchr_unsafe, true); } -enum copy_method { CM_DECODE, CM_ENCODE, CM_PASSTHROUGH }; +/* Decide whether the char at position P needs to be encoded. (It is + not enough to pass a single char *P because the function may need + to inspect the surrounding context.) -/* Decide whether to encode, decode, or pass through the char at P. - This used to be a macro, but it got a little too convoluted. */ -static inline enum copy_method -decide_copy_method (const char *p) + Return true if the char should be escaped as %XX, false otherwise. */ + +static inline bool +char_needs_escaping (const char *p) { if (*p == '%') { - if (ISXDIGIT (*(p + 1)) && ISXDIGIT (*(p + 2))) - { - /* %xx sequence: decode it, unless it would decode to an - unsafe or a reserved char; in that case, leave it as - is. */ - char preempt = X2DIGITS_TO_NUM (*(p + 1), *(p + 2)); - if (URL_UNSAFE_CHAR (preempt) || URL_RESERVED_CHAR (preempt)) - return CM_PASSTHROUGH; - else - return CM_DECODE; - } + if (c_isxdigit (*(p + 1)) && c_isxdigit (*(p + 2))) + return false; else - /* Garbled %.. sequence: encode `%'. */ - return CM_ENCODE; + /* Garbled %.. sequence: encode `%'. */ + return true; } else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p)) - return CM_ENCODE; + return true; else - return CM_PASSTHROUGH; + return false; } /* Translate a %-escaped (but possibly non-conformant) input string S @@ -265,34 +303,33 @@ decide_copy_method (const char *p) After a URL has been run through this function, the protocols that use `%' as the quote character can use the resulting string as-is, - while those that don't call url_unescape() to get to the intended - data. This function is also stable: after an input string is - transformed the first time, all further transformations of the - result yield the same result string. + while those that don't can use url_unescape to get to the intended + data. This function is stable: once the input is transformed, + further transformations of the result yield the same output. Let's discuss why this function is needed. - Imagine Wget is to retrieve `http://abc.xyz/abc def'. Since a raw - space character would mess up the HTTP request, it needs to be - quoted, like this: + Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since + a raw space character would mess up the HTTP request, it needs to + be quoted, like this: GET /abc%20def HTTP/1.0 - It appears that the unsafe chars need to be quoted, for example - with url_escape. But what if we're requested to download + It would appear that the unsafe chars need to be quoted, for + example with url_escape. But what if we're requested to download `abc%20def'? url_escape transforms "%" to "%25", which would leave us with `abc%2520def'. This is incorrect -- since %-escapes are part of URL syntax, "%20" is the correct way to denote a literal - space on the Wget command line. This leaves us in the conclusion - that in that case Wget should not call url_escape, but leave the - `%20' as is. + space on the Wget command line. This leads to the conclusion that + in that case Wget should not call url_escape, but leave the `%20' + as is. This is clearly contradictory, but it only gets worse. - And what if the requested URI is `abc%20 def'? If we call - url_escape, we end up with `/abc%2520%20def', which is almost - certainly not intended. If we don't call url_escape, we are left - with the embedded space and cannot complete the request. What the - user meant was for Wget to request `/abc%20%20def', and this is - where reencode_escapes kicks in. + What if the requested URI is `abc%20 def'? If we call url_escape, + we end up with `/abc%2520%20def', which is almost certainly not + intended. If we don't call url_escape, we are left with the + embedded space and cannot complete the request. What the user + meant was for Wget to request `/abc%20%20def', and this is where + reencode_escapes kicks in. Wget used to solve this by first decoding %-quotes, and then encoding all the "unsafe" characters found in the resulting string. @@ -306,25 +343,25 @@ decide_copy_method (const char *p) literal plus. reencode_escapes correctly translates the above to "a%2B+b", i.e. returns the original string. - This function uses an algorithm proposed by Anon Sricharoenchai: + This function uses a modified version of the algorithm originally + proposed by Anon Sricharoenchai: - 1. Encode all URL_UNSAFE and the "%" that are not followed by 2 - hexdigits. + * Encode all "unsafe" characters, except those that are also + "reserved", to %XX. See urlchr_table for which characters are + unsafe and reserved. - 2. Decode all "%XX" except URL_UNSAFE, URL_RESERVED (";/?:@=&") and - "+". + * Encode the "%" characters not followed by two hex digits to + "%25". - ...except that this code conflates the two steps, and decides - whether to encode, decode, or pass through each character in turn. - The function still uses two passes, but their logic is the same -- - the first pass exists merely for the sake of allocation. Another - small difference is that we include `+' to URL_RESERVED. + * Pass through all other characters and %XX escapes as-is. (Up to + Wget 1.10 this decoded %XX escapes corresponding to "safe" + characters, but that was obtrusive and broke some servers.) Anon's test case: "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc" -> - "http://abc.xyz/%20%3F%2561%25aa%25%20a?a=a+a%2Ba&b=b%26c%3Dc" + "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc" Simpler test cases: @@ -345,58 +382,38 @@ reencode_escapes (const char *s) int oldlen, newlen; int encode_count = 0; - int decode_count = 0; - /* First, pass through the string to see if there's anything to do, + /* First pass: inspect the string to see if there's anything to do, and to calculate the new length. */ for (p1 = s; *p1; p1++) - { - switch (decide_copy_method (p1)) - { - case CM_ENCODE: - ++encode_count; - break; - case CM_DECODE: - ++decode_count; - break; - case CM_PASSTHROUGH: - break; - } - } + if (char_needs_escaping (p1)) + ++encode_count; - if (!encode_count && !decode_count) + if (!encode_count) /* The string is good as it is. */ - return (char *)s; /* C const model sucks. */ + return (char *) s; /* C const model sucks. */ oldlen = p1 - s; - /* Each encoding adds two characters (hex digits), while each - decoding removes two characters. */ - newlen = oldlen + 2 * (encode_count - decode_count); + /* Each encoding adds two characters (hex digits). */ + newlen = oldlen + 2 * encode_count; newstr = xmalloc (newlen + 1); + /* Second pass: copy the string to the destination address, encoding + chars when needed. */ p1 = s; p2 = newstr; while (*p1) - { - switch (decide_copy_method (p1)) - { - case CM_ENCODE: - { - unsigned char c = *p1++; - *p2++ = '%'; - *p2++ = XNUM_TO_DIGIT (c >> 4); - *p2++ = XNUM_TO_DIGIT (c & 0xf); - } - break; - case CM_DECODE: - *p2++ = X2DIGITS_TO_NUM (p1[1], p1[2]); - p1 += 3; /* skip %xx */ - break; - case CM_PASSTHROUGH: - *p2++ = *p1++; - } - } + if (char_needs_escaping (p1)) + { + unsigned char c = *p1++; + *p2++ = '%'; + *p2++ = XNUM_TO_DIGIT (c >> 4); + *p2++ = XNUM_TO_DIGIT (c & 0xf); + } + else + *p2++ = *p1++; + *p2 = '\0'; assert (p2 - newstr == newlen); return newstr; @@ -412,31 +429,31 @@ url_scheme (const char *url) for (i = 0; supported_schemes[i].leading_string; i++) if (0 == strncasecmp (url, supported_schemes[i].leading_string, - strlen (supported_schemes[i].leading_string))) + strlen (supported_schemes[i].leading_string))) { - if (supported_schemes[i].enabled) - return (enum url_scheme) i; - else - return SCHEME_INVALID; + if (!(supported_schemes[i].flags & scm_disabled)) + return (enum url_scheme) i; + else + return SCHEME_INVALID; } return SCHEME_INVALID; } -#define SCHEME_CHAR(ch) (ISALNUM (ch) || (ch) == '-' || (ch) == '+') +#define SCHEME_CHAR(ch) (c_isalnum (ch) || (ch) == '-' || (ch) == '+') /* Return 1 if the URL begins with any "scheme", 0 otherwise. As currently implemented, it returns true if URL begins with [-+a-zA-Z0-9]+: . */ -int +bool url_has_scheme (const char *url) { const char *p = url; /* The first char must be a scheme char. */ if (!*p || !SCHEME_CHAR (*p)) - return 0; + return false; ++p; /* Followed by 0 or more scheme chars. */ while (*p && SCHEME_CHAR (*p)) @@ -445,6 +462,13 @@ url_has_scheme (const char *url) return *p == ':'; } +bool +url_valid_scheme (const char *url) +{ + enum url_scheme scheme = url_scheme (url); + return scheme != SCHEME_INVALID; +} + int scheme_default_port (enum url_scheme scheme) { @@ -454,41 +478,41 @@ scheme_default_port (enum url_scheme scheme) void scheme_disable (enum url_scheme scheme) { - supported_schemes[scheme].enabled = 0; + supported_schemes[scheme].flags |= scm_disabled; } -/* Skip the username and password, if present here. The function - should *not* be called with the complete URL, but with the part - right after the scheme. +/* Skip the username and password, if present in the URL. The + function should *not* be called with the complete URL, but with the + portion after the scheme. - If no username and password are found, return 0. */ + If no username and password are found, return URL. */ -static int +static const char * url_skip_credentials (const char *url) { /* Look for '@' that comes before terminators, such as '/', '?', '#', or ';'. */ const char *p = (const char *)strpbrk (url, "@/?#;"); if (!p || *p != '@') - return 0; - return p + 1 - url; + return url; + return p + 1; } /* Parse credentials contained in [BEG, END). The region is expected to have come from a URL and is unescaped. */ -static int +static bool parse_credentials (const char *beg, const char *end, char **user, char **passwd) { char *colon; const char *userend; if (beg == end) - return 0; /* empty user name */ + return false; /* empty user name */ colon = memchr (beg, ':', end - beg); if (colon == beg) - return 0; /* again empty user name */ + return false; /* again empty user name */ if (colon) { @@ -503,11 +527,12 @@ parse_credentials (const char *beg, const char *end, char **user, char **passwd) } *user = strdupdelim (beg, userend); url_unescape (*user); - return 1; + return true; } /* Used by main.c: detect URLs written using the "shorthand" URL forms - popularized by Netscape and NcFTP. HTTP shorthands look like this: + originally popularized by Netscape and NcFTP. HTTP shorthands look + like this: www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file www.foo.com[:port] -> http://www.foo.com[:port] @@ -523,258 +548,112 @@ char * rewrite_shorthand_url (const char *url) { const char *p; + char *ret; - if (url_has_scheme (url)) + if (url_scheme (url) != SCHEME_INVALID) return NULL; /* Look for a ':' or '/'. The former signifies NcFTP syntax, the latter Netscape. */ - for (p = url; *p && *p != ':' && *p != '/'; p++) - ; - + p = strpbrk (url, ":/"); if (p == url) return NULL; - if (*p == ':') + /* If we're looking at "://", it means the URL uses a scheme we + don't support, which may include "https" when compiled without + SSL support. Don't bogusly rewrite such URLs. */ + if (p && p[0] == ':' && p[1] == '/' && p[2] == '/') + return NULL; + + if (p && *p == ':') { - const char *pp; - char *res; - /* If the characters after the colon and before the next slash - or end of string are all digits, it's HTTP. */ - int digits = 0; - for (pp = p + 1; ISDIGIT (*pp); pp++) - ++digits; - if (digits > 0 && (*pp == '/' || *pp == '\0')) - goto http; - - /* Prepend "ftp://" to the entire URL... */ - res = xmalloc (6 + strlen (url) + 1); - sprintf (res, "ftp://%s", url); - /* ...and replace ':' with '/'. */ - res[6 + (p - url)] = '/'; - return res; + /* Colon indicates ftp, as in foo.bar.com:path. Check for + special case of http port number ("localhost:10000"). */ + int digits = strspn (p + 1, "0123456789"); + if (digits && (p[1 + digits] == '/' || p[1 + digits] == '\0')) + goto http; + + /* Turn "foo.bar.com:path" to "ftp://foo.bar.com/path". */ + ret = aprintf ("ftp://%s", url); + ret[6 + (p - url)] = '/'; } else { - char *res; http: - /* Just prepend "http://" to what we have. */ - res = xmalloc (7 + strlen (url) + 1); - sprintf (res, "http://%s", url); - return res; + /* Just prepend "http://" to URL. */ + ret = aprintf ("http://%s", url); } + return ret; } -static void split_path PARAMS ((const char *, char **, char **)); +static void split_path (const char *, char **, char **); /* Like strpbrk, with the exception that it returns the pointer to the terminating zero (end-of-string aka "eos") if no matching character - is found. - - Although I normally balk at Gcc-specific optimizations, it probably - makes sense here: glibc has optimizations that detect strpbrk being - called with literal string as ACCEPT and inline the search. That - optimization is defeated if strpbrk is hidden within the call to - another function. (And no, making strpbrk_or_eos inline doesn't - help because the check for literal accept is in the - preprocessor.) */ - -#ifdef __GNUC__ - -#define strpbrk_or_eos(s, accept) ({ \ - char *SOE_p = strpbrk (s, accept); \ - if (!SOE_p) \ - SOE_p = (char *)s + strlen (s); \ - SOE_p; \ -}) + is found. */ -#else /* not __GNUC__ */ - -static char * +static inline char * strpbrk_or_eos (const char *s, const char *accept) { char *p = strpbrk (s, accept); if (!p) - p = (char *)s + strlen (s); + p = strchr (s, '\0'); return p; } -#endif -/* Turn STR into lowercase; return non-zero if a character was - actually changed. */ +/* Turn STR into lowercase; return true if a character was actually + changed. */ -static int +static bool lowercase_str (char *str) { - int change = 0; + bool changed = false; for (; *str; str++) - if (ISUPPER (*str)) + if (c_isupper (*str)) { - change = 1; - *str = TOLOWER (*str); + changed = true; + *str = c_tolower (*str); } - return change; + return changed; } -static char *parse_errors[] = { -#define PE_NO_ERROR 0 - "No error", -#define PE_UNSUPPORTED_SCHEME 1 - "Unsupported scheme", -#define PE_EMPTY_HOST 2 - "Empty host", -#define PE_BAD_PORT_NUMBER 3 - "Bad port number", -#define PE_INVALID_USER_NAME 4 - "Invalid user name", -#define PE_UNTERMINATED_IPV6_ADDRESS 5 - "Unterminated IPv6 numeric address", -#define PE_IPV6_NOT_SUPPORTED 6 - "IPv6 addresses not supported", -#define PE_INVALID_IPV6_ADDRESS 7 - "Invalid IPv6 numeric address" -}; - -#define SETERR(p, v) do { \ - if (p) \ - *(p) = (v); \ -} while (0) - -#ifdef ENABLE_IPV6 -/* The following two functions were adapted from glibc. */ - -static int -is_valid_ipv4_address (const char *str, const char *end) +static const char * +init_seps (enum url_scheme scheme) { - int saw_digit, octets; - int val; - - saw_digit = 0; - octets = 0; - val = 0; - - while (str < end) { - int ch = *str++; - - if (ch >= '0' && ch <= '9') { - val = val * 10 + (ch - '0'); - - if (val > 255) - return 0; - if (saw_digit == 0) { - if (++octets > 4) - return 0; - saw_digit = 1; - } - } else if (ch == '.' && saw_digit == 1) { - if (octets == 4) - return 0; - val = 0; - saw_digit = 0; - } else - return 0; - } - if (octets < 4) - return 0; - - return 1; + static char seps[8] = ":/"; + char *p = seps + 2; + int flags = supported_schemes[scheme].flags; + + if (flags & scm_has_params) + *p++ = ';'; + if (flags & scm_has_query) + *p++ = '?'; + if (flags & scm_has_fragment) + *p++ = '#'; + *p = '\0'; + return seps; } -static const int NS_INADDRSZ = 4; -static const int NS_IN6ADDRSZ = 16; -static const int NS_INT16SZ = 2; - -static int -is_valid_ipv6_address (const char *str, const char *end) -{ - static const char xdigits[] = "0123456789abcdef"; - const char *curtok; - int tp; - const char *colonp; - int saw_xdigit; - unsigned int val; - - tp = 0; - colonp = NULL; - - if (str == end) - return 0; - - /* Leading :: requires some special handling. */ - if (*str == ':') - { - ++str; - if (str == end || *str != ':') - return 0; - } - - curtok = str; - saw_xdigit = 0; - val = 0; - - while (str < end) { - int ch = *str++; - const char *pch; - - /* if ch is a number, add it to val. */ - pch = strchr(xdigits, ch); - if (pch != NULL) { - val <<= 4; - val |= (pch - xdigits); - if (val > 0xffff) - return 0; - saw_xdigit = 1; - continue; - } - - /* if ch is a colon ... */ - if (ch == ':') { - curtok = str; - if (saw_xdigit == 0) { - if (colonp != NULL) - return 0; - colonp = str + tp; - continue; - } else if (str == end) { - return 0; - } - if (tp > NS_IN6ADDRSZ - NS_INT16SZ) - return 0; - tp += NS_INT16SZ; - saw_xdigit = 0; - val = 0; - continue; - } - - /* if ch is a dot ... */ - if (ch == '.' && (tp <= NS_IN6ADDRSZ - NS_INADDRSZ) && - is_valid_ipv4_address(curtok, end) == 1) { - tp += NS_INADDRSZ; - saw_xdigit = 0; - break; - } - - return 0; - } - - if (saw_xdigit == 1) { - if (tp > NS_IN6ADDRSZ - NS_INT16SZ) - return 0; - tp += NS_INT16SZ; - } - - if (colonp != NULL) { - if (tp == NS_IN6ADDRSZ) - return 0; - tp = NS_IN6ADDRSZ; - } - - if (tp != NS_IN6ADDRSZ) - return 0; - - return 1; -} -#endif +static const char *parse_errors[] = { +#define PE_NO_ERROR 0 + N_("No error"), +#define PE_UNSUPPORTED_SCHEME 1 + N_("Unsupported scheme %s"), /* support for format token only here */ +#define PE_MISSING_SCHEME 2 + N_("Scheme missing"), +#define PE_INVALID_HOST_NAME 3 + N_("Invalid host name"), +#define PE_BAD_PORT_NUMBER 4 + N_("Bad port number"), +#define PE_INVALID_USER_NAME 5 + N_("Invalid user name"), +#define PE_UNTERMINATED_IPV6_ADDRESS 6 + N_("Unterminated IPv6 numeric address"), +#define PE_IPV6_NOT_SUPPORTED 7 + N_("IPv6 addresses not supported"), +#define PE_INVALID_IPV6_ADDRESS 8 + N_("Invalid IPv6 numeric address") +}; /* Parse a URL. @@ -782,13 +661,14 @@ is_valid_ipv6_address (const char *str, const char *end) error, and if ERROR is not NULL, also set *ERROR to the appropriate error code. */ struct url * -url_parse (const char *url, int *error) +url_parse (const char *url, int *error, struct iri *iri, bool percent_encode) { struct url *u; const char *p; - int path_modified, host_modified; + bool path_modified, host_modified; enum url_scheme scheme; + const char *seps; const char *uname_b, *uname_e; const char *host_b, *host_e; @@ -800,21 +680,47 @@ url_parse (const char *url, int *error) int port; char *user = NULL, *passwd = NULL; - char *url_encoded; + const char *url_encoded = NULL; + char *new_url = NULL; + + int error_code; scheme = url_scheme (url); if (scheme == SCHEME_INVALID) { - SETERR (error, PE_UNSUPPORTED_SCHEME); - return NULL; + if (url_has_scheme (url)) + error_code = PE_UNSUPPORTED_SCHEME; + else + error_code = PE_MISSING_SCHEME; + goto error; + } + + if (iri && iri->utf8_encode) + { + iri->utf8_encode = remote_to_utf8 (iri, iri->orig_url ? iri->orig_url : url, (const char **) &new_url); + if (!iri->utf8_encode) + new_url = NULL; + else + { + iri->orig_url = xstrdup (url); + percent_encode = true; + } } - url_encoded = reencode_escapes (url); + /* XXX XXX Could that change introduce (security) bugs ??? XXX XXX*/ + if (percent_encode) + url_encoded = reencode_escapes (new_url ? new_url : url); + else + url_encoded = new_url ? new_url : url; + p = url_encoded; + if (new_url && url_encoded != new_url) + xfree (new_url); + p += strlen (supported_schemes[scheme].leading_string); uname_b = p; - p += url_skip_credentials (p); + p = url_skip_credentials (p); uname_e = p; /* scheme://user:pass@host[:port]... */ @@ -825,53 +731,71 @@ url_parse (const char *url, int *error) scheme://host[:port][/path][;params][?query][#fragment] */ + path_b = path_e = NULL; params_b = params_e = NULL; query_b = query_e = NULL; fragment_b = fragment_e = NULL; + /* Initialize separators for optional parts of URL, depending on the + scheme. For example, FTP has params, and HTTP and HTTPS have + query string and fragment. */ + seps = init_seps (scheme); + host_b = p; if (*p == '[') { /* Handle IPv6 address inside square brackets. Ideally we'd - just look for the terminating ']', but rfc2732 mandates - rejecting invalid IPv6 addresses. */ + just look for the terminating ']', but rfc2732 mandates + rejecting invalid IPv6 addresses. */ /* The address begins after '['. */ host_b = p + 1; host_e = strchr (host_b, ']'); if (!host_e) - { - SETERR (error, PE_UNTERMINATED_IPV6_ADDRESS); - return NULL; - } + { + error_code = PE_UNTERMINATED_IPV6_ADDRESS; + goto error; + } #ifdef ENABLE_IPV6 /* Check if the IPv6 address is valid. */ if (!is_valid_ipv6_address(host_b, host_e)) - { - SETERR (error, PE_INVALID_IPV6_ADDRESS); - return NULL; - } + { + error_code = PE_INVALID_IPV6_ADDRESS; + goto error; + } /* Continue parsing after the closing ']'. */ p = host_e + 1; #else - SETERR (error, PE_IPV6_NOT_SUPPORTED); - return NULL; + error_code = PE_IPV6_NOT_SUPPORTED; + goto error; #endif + + /* The closing bracket must be followed by a separator or by the + null char. */ + /* http://[::1]... */ + /* ^ */ + if (!strchr (seps, *p)) + { + /* Trailing garbage after []-delimited IPv6 address. */ + error_code = PE_INVALID_HOST_NAME; + goto error; + } } else { - p = strpbrk_or_eos (p, ":/;?#"); + p = strpbrk_or_eos (p, seps); host_e = p; } + ++seps; /* advance to '/' */ if (host_b == host_e) { - SETERR (error, PE_EMPTY_HOST); - return NULL; + error_code = PE_INVALID_HOST_NAME; + goto error; } port = scheme_default_port (scheme); @@ -883,74 +807,51 @@ url_parse (const char *url, int *error) /* ^ */ ++p; port_b = p; - p = strpbrk_or_eos (p, "/;?#"); + p = strpbrk_or_eos (p, seps); port_e = p; - if (port_b == port_e) - { - /* http://host:/whatever */ - /* ^ */ - SETERR (error, PE_BAD_PORT_NUMBER); - return NULL; - } - - for (port = 0, pp = port_b; pp < port_e; pp++) - { - if (!ISDIGIT (*pp)) - { - /* http://host:12randomgarbage/blah */ - /* ^ */ - SETERR (error, PE_BAD_PORT_NUMBER); - return NULL; - } - - port = 10 * port + (*pp - '0'); - } + /* Allow empty port, as per rfc2396. */ + if (port_b != port_e) + for (port = 0, pp = port_b; pp < port_e; pp++) + { + if (!c_isdigit (*pp)) + { + /* http://host:12randomgarbage/blah */ + /* ^ */ + error_code = PE_BAD_PORT_NUMBER; + goto error; + } + port = 10 * port + (*pp - '0'); + /* Check for too large port numbers here, before we have + a chance to overflow on bogus port values. */ + if (port > 0xffff) + { + error_code = PE_BAD_PORT_NUMBER; + goto error; + } + } } + /* Advance to the first separator *after* '/' (either ';' or '?', + depending on the scheme). */ + ++seps; + + /* Get the optional parts of URL, each part being delimited by + current location and the position of the next separator. */ +#define GET_URL_PART(sepchar, var) do { \ + if (*p == sepchar) \ + var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \ + ++seps; \ +} while (0) - if (*p == '/') - { - ++p; - path_b = p; - p = strpbrk_or_eos (p, ";?#"); - path_e = p; - } - else - { - /* Path is not allowed not to exist. */ - path_b = path_e = p; - } + GET_URL_PART ('/', path); + if (supported_schemes[scheme].flags & scm_has_params) + GET_URL_PART (';', params); + if (supported_schemes[scheme].flags & scm_has_query) + GET_URL_PART ('?', query); + if (supported_schemes[scheme].flags & scm_has_fragment) + GET_URL_PART ('#', fragment); - if (*p == ';') - { - ++p; - params_b = p; - p = strpbrk_or_eos (p, "?#"); - params_e = p; - } - if (*p == '?') - { - ++p; - query_b = p; - p = strpbrk_or_eos (p, "#"); - query_e = p; - - /* Hack that allows users to use '?' (a wildcard character) in - FTP URLs without it being interpreted as a query string - delimiter. */ - if (scheme == SCHEME_FTP) - { - query_b = query_e = NULL; - path_e = p; - } - } - if (*p == '#') - { - ++p; - fragment_b = p; - p += strlen (p); - fragment_e = p; - } +#undef GET_URL_PART assert (*p == 0); if (uname_b != uname_e) @@ -959,15 +860,13 @@ url_parse (const char *url, int *error) /* ^ ^ */ /* uname_b uname_e */ if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd)) - { - SETERR (error, PE_INVALID_USER_NAME); - return NULL; - } + { + error_code = PE_INVALID_USER_NAME; + goto error; + } } - u = (struct url *)xmalloc (sizeof (struct url)); - memset (u, 0, sizeof (*u)); - + u = xnew0 (struct url); u->scheme = scheme; u->host = strdupdelim (host_b, host_e); u->port = port; @@ -975,11 +874,33 @@ url_parse (const char *url, int *error) u->passwd = passwd; u->path = strdupdelim (path_b, path_e); - path_modified = path_simplify (u->path); + path_modified = path_simplify (scheme, u->path); split_path (u->path, &u->dir, &u->file); host_modified = lowercase_str (u->host); + /* Decode %HH sequences in host name. This is important not so much + to support %HH sequences in host names (which other browser + don't), but to support binary characters (which will have been + converted to %HH by reencode_escapes). */ + if (strchr (u->host, '%')) + { + url_unescape (u->host); + host_modified = true; + + /* Apply IDNA regardless of iri->utf8_encode status */ + if (opt.enable_iri && iri) + { + char *new = idn_encode (iri, u->host); + if (new) + { + xfree (u->host); + u->host = new; + host_modified = true; + } + } + } + if (params_b) u->params = strdupdelim (params_b, params_e); if (query_b) @@ -987,33 +908,64 @@ url_parse (const char *url, int *error) if (fragment_b) u->fragment = strdupdelim (fragment_b, fragment_e); - if (path_modified || u->fragment || host_modified || path_b == path_e) + if (opt.enable_iri || path_modified || u->fragment || host_modified || path_b == path_e) { /* If we suspect that a transformation has rendered what - url_string might return different from URL_ENCODED, rebuild - u->url using url_string. */ - u->url = url_string (u, 0); + url_string might return different from URL_ENCODED, rebuild + u->url using url_string. */ + u->url = url_string (u, URL_AUTH_SHOW); if (url_encoded != url) - xfree ((char *) url_encoded); + xfree ((char *) url_encoded); } else { if (url_encoded == url) - u->url = xstrdup (url); + u->url = xstrdup (url); else - u->url = url_encoded; + u->url = (char *) url_encoded; } - url_encoded = NULL; return u; + + error: + /* Cleanup in case of error: */ + if (url_encoded && url_encoded != url) + xfree ((char *) url_encoded); + + /* Transmit the error code to the caller, if the caller wants to + know. */ + if (error) + *error = error_code; + return NULL; } -const char * -url_error (int error_code) +/* Return the error message string from ERROR_CODE, which should have + been retrieved from url_parse. The error message is translated. */ + +char * +url_error (const char *url, int error_code) { - assert (error_code >= 0 && error_code < countof (parse_errors)); - return parse_errors[error_code]; + assert (error_code >= 0 && ((size_t) error_code) < countof (parse_errors)); + + if (error_code == PE_UNSUPPORTED_SCHEME) + { + char *error, *p; + char *scheme = xstrdup (url); + assert (url_has_scheme (url)); + + if ((p = strchr (scheme, ':'))) + *p = '\0'; + if (!strcasecmp (scheme, "https")) + error = aprintf (_("HTTPS support not compiled in")); + else + error = aprintf (_(parse_errors[error_code]), quote (scheme)); + xfree (scheme); + + return error; + } + else + return xstrdup (_(parse_errors[error_code])); } /* Split PATH into DIR and FILE. PATH comes from the URL and is @@ -1079,14 +1031,14 @@ full_path_length (const struct url *url) static void full_path_write (const struct url *url, char *where) { -#define FROB(el, chr) do { \ - char *f_el = url->el; \ - if (f_el) { \ - int l = strlen (f_el); \ - *where++ = chr; \ - memcpy (where, f_el, l); \ - where += l; \ - } \ +#define FROB(el, chr) do { \ + char *f_el = url->el; \ + if (f_el) { \ + int l = strlen (f_el); \ + *where++ = chr; \ + memcpy (where, f_el, l); \ + where += l; \ + } \ } while (0) FROB (path, '/'); @@ -1104,7 +1056,7 @@ char * url_full_path (const struct url *url) { int length = full_path_length (url); - char *full_path = (char *)xmalloc(length + 1); + char *full_path = xmalloc (length + 1); full_path_write (url, full_path); full_path[length] = '\0'; @@ -1112,6 +1064,30 @@ url_full_path (const struct url *url) return full_path; } +/* Unescape CHR in an otherwise escaped STR. Used to selectively + escaping of certain characters, such as "/" and ":". Returns a + count of unescaped chars. */ + +static void +unescape_single_char (char *str, char chr) +{ + const char c1 = XNUM_TO_DIGIT (chr >> 4); + const char c2 = XNUM_TO_DIGIT (chr & 0xf); + char *h = str; /* hare */ + char *t = str; /* tortoise */ + for (; *h; h++, t++) + { + if (h[0] == '%' && h[1] == c1 && h[2] == c2) + { + *t = chr; + h += 2; + } + else + *t = *h; + } + *t = '\0'; +} + /* Escape unsafe and reserved characters, except for the slash characters. */ @@ -1119,27 +1095,10 @@ static char * url_escape_dir (const char *dir) { char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1); - char *h, *t; if (newdir == dir) return (char *)dir; - /* Unescape slashes in NEWDIR. */ - - h = newdir; /* hare */ - t = newdir; /* tortoise */ - - for (; *h; h++, t++) - { - if (*h == '%' && h[1] == '2' && h[2] == 'F') - { - *t = '/'; - h += 2; - } - else - *t = *h; - } - *t = '\0'; - + unescape_single_char (newdir, '/'); return newdir; } @@ -1176,7 +1135,7 @@ sync_path (struct url *u) *p++ = '/'; memcpy (p, efile, filelen); p += filelen; - *p++ = '\0'; + *p = '\0'; } u->path = newpath; @@ -1188,7 +1147,7 @@ sync_path (struct url *u) /* Regenerate u->url as well. */ xfree (u->url); - u->url = url_string (u, 0); + u->url = url_string (u, URL_AUTH_SHOW); } /* Mutators. Code in ftp.c insists on changing u->dir and u->file. @@ -1217,11 +1176,11 @@ url_free (struct url *url) xfree (url->path); xfree (url->url); - FREE_MAYBE (url->params); - FREE_MAYBE (url->query); - FREE_MAYBE (url->fragment); - FREE_MAYBE (url->user); - FREE_MAYBE (url->passwd); + xfree_null (url->params); + xfree_null (url->query); + xfree_null (url->fragment); + xfree_null (url->user); + xfree_null (url->passwd); xfree (url->dir); xfree (url->file); @@ -1230,13 +1189,13 @@ url_free (struct url *url) } /* Create all the necessary directories for PATH (a file). Calls - mkdirhier() internally. */ + make_directory internally. */ int mkalldirs (const char *path) { const char *p; char *t; - struct stat st; + struct_stat st; int res; p = path + strlen (path); @@ -1252,27 +1211,27 @@ mkalldirs (const char *path) if ((stat (t, &st) == 0)) { if (S_ISDIR (st.st_mode)) - { - xfree (t); - return 0; - } + { + xfree (t); + return 0; + } else - { - /* If the dir exists as a file name, remove it first. This - is *only* for Wget to work with buggy old CERN http - servers. Here is the scenario: When Wget tries to - retrieve a directory without a slash, e.g. - http://foo/bar (bar being a directory), CERN server will - not redirect it too http://foo/bar/ -- it will generate a - directory listing containing links to bar/file1, - bar/file2, etc. Wget will lose because it saves this - HTML listing to a file `bar', so it cannot create the - directory. To work around this, if the file of the same - name exists, we just remove it and create the directory - anyway. */ - DEBUGP (("Removing %s because of directory danger!\n", t)); - unlink (t); - } + { + /* If the dir exists as a file name, remove it first. This + is *only* for Wget to work with buggy old CERN http + servers. Here is the scenario: When Wget tries to + retrieve a directory without a slash, e.g. + http://foo/bar (bar being a directory), CERN server will + not redirect it too http://foo/bar/ -- it will generate a + directory listing containing links to bar/file1, + bar/file2, etc. Wget will lose because it saves this + HTML listing to a file `bar', so it cannot create the + directory. To work around this, if the file of the same + name exists, we just remove it and create the directory + anyway. */ + DEBUGP (("Removing %s because of directory danger!\n", t)); + unlink (t); + } } res = make_directory (t); if (res != 0) @@ -1289,21 +1248,25 @@ mkalldirs (const char *path) The idea is to have a convenient and efficient way to construct a string by having various functions append data to it. Instead of passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the - functions in questions, we pass the pointer to this struct. */ + functions in questions, we pass the pointer to this struct. + + Functions that write to the members in this struct must make sure + that base remains null terminated by calling append_null(). + */ struct growable { char *base; - int size; - int tail; + int size; /* memory allocated */ + int tail; /* string length */ }; /* Ensure that the string can accept APPEND_COUNT more characters past the current TAIL position. If necessary, this will grow the string and update its allocated size. If the string is already large enough to take TAIL+APPEND_COUNT characters, this does nothing. */ -#define GROW(g, append_size) do { \ - struct growable *G_ = g; \ - DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \ +#define GROW(g, append_size) do { \ + struct growable *G_ = g; \ + DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \ } while (0) /* Return the tail position of the string. */ @@ -1312,36 +1275,55 @@ struct growable { /* Move the tail position by APPEND_COUNT characters. */ #define TAIL_INCR(r, append_count) ((r)->tail += append_count) -/* Append the string STR to DEST. NOTICE: the string in DEST is not - terminated. */ +/* Append NULL to DEST. */ static void -append_string (const char *str, struct growable *dest) +append_null (struct growable *dest) { - int l = strlen (str); - GROW (dest, l); - memcpy (TAIL (dest), str, l); - TAIL_INCR (dest, l); + GROW (dest, 1); + *TAIL (dest) = 0; } -/* Append CH to DEST. For example, append_char (0, DEST) - zero-terminates DEST. */ - +/* Append CH to DEST. */ static void append_char (char ch, struct growable *dest) { - GROW (dest, 1); - *TAIL (dest) = ch; - TAIL_INCR (dest, 1); + if (ch) + { + GROW (dest, 1); + *TAIL (dest) = ch; + TAIL_INCR (dest, 1); + } + + append_null (dest); +} + +/* Append the string STR to DEST. */ +static void +append_string (const char *str, struct growable *dest) +{ + int l = strlen (str); + + if (l) + { + GROW (dest, l); + memcpy (TAIL (dest), str, l); + TAIL_INCR (dest, l); + } + + append_null (dest); } + enum { - filechr_not_unix = 1, /* unusable on Unix, / and \0 */ - filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */ - filechr_control = 4, /* a control character, e.g. 0-31 */ + filechr_not_unix = 1, /* unusable on Unix, / and \0 */ + filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */ + filechr_control = 4 /* a control character, e.g. 0-31 */ }; -#define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask)) +#define FILE_CHAR_TEST(c, mask) \ + ((opt.restrict_files_nonascii && !c_isascii ((unsigned char)(c))) || \ + (filechr_table[(unsigned char)(c)] & (mask))) /* Shorthands for the table: */ #define U filechr_not_unix @@ -1358,7 +1340,7 @@ enum { translate file name back to URL, this would become important crucial. Right now, it's better to be minimal in escaping. */ -const static unsigned char filechr_table[256] = +static const unsigned char filechr_table[256] = { UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */ C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */ @@ -1375,7 +1357,7 @@ UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */ 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */ 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */ 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */ - 0, 0, 0, 0, 0, 0, 0, 0, /* x y z { | } ~ DEL */ + 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */ C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */ C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */ @@ -1403,17 +1385,19 @@ UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */ query, normally '?'. Since Windows cannot handle '?' as part of file name, we use '@' instead there. */ #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@') +#define FN_QUERY_SEP_STR (opt.restrict_files_os != restrict_windows ? "?" : "@") /* Quote path element, characters in [b, e), as file name, and append the quoted string to DEST. Each character is quoted as per - file_unsafe_char and the corresponding table. */ + file_unsafe_char and the corresponding table. + + If ESCAPED is true, the path element is considered to be + URL-escaped and will be unescaped prior to inspection. */ static void -append_uri_pathel (const char *b, const char *e, struct growable *dest) +append_uri_pathel (const char *b, const char *e, bool escaped, + struct growable *dest) { - char *pathel; - int pathlen; - const char *p; int quoted, outlen; @@ -1426,46 +1410,76 @@ append_uri_pathel (const char *b, const char *e, struct growable *dest) mask |= filechr_control; /* Copy [b, e) to PATHEL and URL-unescape it. */ - BOUNDED_TO_ALLOCA (b, e, pathel); - url_unescape (pathel); - pathlen = strlen (pathel); + if (escaped) + { + char *unescaped; + BOUNDED_TO_ALLOCA (b, e, unescaped); + url_unescape (unescaped); + b = unescaped; + e = unescaped + strlen (unescaped); + } - /* Go through PATHEL and check how many characters we'll need to - add for file quoting. */ + /* Defang ".." when found as component of path. Remember that path + comes from the URL and might contain malicious input. */ + if (e - b == 2 && b[0] == '.' && b[1] == '.') + { + b = "%2E%2E"; + e = b + 6; + } + + /* Walk the PATHEL string and check how many characters we'll need + to quote. */ quoted = 0; - for (p = pathel; *p; p++) + for (p = b; p < e; p++) if (FILE_CHAR_TEST (*p, mask)) ++quoted; - /* p - pathel is the string length. Each quoted char means two - additional characters in the string, hence 2*quoted. */ - outlen = (p - pathel) + (2 * quoted); + /* Calculate the length of the output string. e-b is the input + string length. Each quoted char introduces two additional + characters in the string, hence 2*quoted. */ + outlen = (e - b) + (2 * quoted); GROW (dest, outlen); if (!quoted) { - /* If there's nothing to quote, we don't need to go through the - string the second time. */ - memcpy (TAIL (dest), pathel, outlen); + /* If there's nothing to quote, we can simply append the string + without processing it again. */ + memcpy (TAIL (dest), b, outlen); } else { char *q = TAIL (dest); - for (p = pathel; *p; p++) - { - if (!FILE_CHAR_TEST (*p, mask)) - *q++ = *p; - else - { - unsigned char ch = *p; - *q++ = '%'; - *q++ = XNUM_TO_DIGIT (ch >> 4); - *q++ = XNUM_TO_DIGIT (ch & 0xf); - } - } + for (p = b; p < e; p++) + { + if (!FILE_CHAR_TEST (*p, mask)) + *q++ = *p; + else + { + unsigned char ch = *p; + *q++ = '%'; + *q++ = XNUM_TO_DIGIT (ch >> 4); + *q++ = XNUM_TO_DIGIT (ch & 0xf); + } + } assert (q - TAIL (dest) == outlen); } + + /* Perform inline case transformation if required. */ + if (opt.restrict_files_case == restrict_lowercase + || opt.restrict_files_case == restrict_uppercase) + { + char *q; + for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q) + { + if (opt.restrict_files_case == restrict_lowercase) + *q = c_tolower (*q); + else + *q = c_toupper (*q); + } + } + TAIL_INCR (dest, outlen); + append_null (dest); } /* Append to DEST the directory structure that corresponds the @@ -1495,34 +1509,44 @@ append_dir_structure (const struct url *u, struct growable *dest) for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1) { if (cut-- > 0) - continue; + continue; if (pathel == next) - /* Ignore empty pathels. path_simplify should remove - occurrences of "//" from the path, but it has special cases - for starting / which generates an empty pathel here. */ - continue; + /* Ignore empty pathels. */ + continue; if (dest->tail) - append_char ('/', dest); - append_uri_pathel (pathel, next, dest); + append_char ('/', dest); + append_uri_pathel (pathel, next, true, dest); } } -/* Return a unique file name that matches the given URL as good as +/* Return a unique file name that matches the given URL as well as possible. Does not create directories on the file system. */ char * -url_file_name (const struct url *u) +url_file_name (const struct url *u, char *replaced_filename) { - struct growable fnres; + struct growable fnres; /* stands for "file name result" */ + struct growable temp_fnres; - char *u_file, *u_query; - char *fname, *unique; + const char *u_file; + char *fname, *unique, *fname_len_check; + const char *index_filename = "index.html"; /* The default index file is index.html */ + size_t max_length; fnres.base = NULL; fnres.size = 0; fnres.tail = 0; + temp_fnres.base = NULL; + temp_fnres.size = 0; + temp_fnres.tail = 0; + + /* If an alternative index file was defined, change index_filename */ + if (opt.default_page) + index_filename = opt.default_page; + + /* Start with the directory prefix, if specified. */ if (opt.dir_prefix) append_string (opt.dir_prefix, &fnres); @@ -1532,206 +1556,252 @@ url_file_name (const struct url *u) directory structure. */ if (opt.dirstruct) { + if (opt.protocol_directories) + { + if (fnres.tail) + append_char ('/', &fnres); + append_string (supported_schemes[u->scheme].name, &fnres); + } if (opt.add_hostdir) - { - if (fnres.tail) - append_char ('/', &fnres); - append_string (u->host, &fnres); - if (u->port != scheme_default_port (u->scheme)) - { - char portstr[24]; - number_to_string (portstr, u->port); - append_char (FN_PORT_SEP, &fnres); - append_string (portstr, &fnres); - } - } + { + if (fnres.tail) + append_char ('/', &fnres); + if (0 != strcmp (u->host, "..")) + append_string (u->host, &fnres); + else + /* Host name can come from the network; malicious DNS may + allow ".." to be resolved, causing us to write to + "../". Defang such host names. */ + append_string ("%2E%2E", &fnres); + if (u->port != scheme_default_port (u->scheme)) + { + char portstr[24]; + number_to_string (portstr, u->port); + append_char (FN_PORT_SEP, &fnres); + append_string (portstr, &fnres); + } + } append_dir_structure (u, &fnres); } - /* Add the file name. */ - if (fnres.tail) - append_char ('/', &fnres); - u_file = *u->file ? u->file : "index.html"; - append_uri_pathel (u_file, u_file + strlen (u_file), &fnres); + if (!replaced_filename) + { + /* Create the filename. */ + u_file = *u->file ? u->file : index_filename; + + /* Append "?query" to the file name, even if empty, + * and create fname_len_check. */ + if (u->query) + fname_len_check = concat_strings (u_file, FN_QUERY_SEP_STR, u->query, NULL); + else + fname_len_check = strdupdelim (u_file, u_file + strlen (u_file)); + } + else + { + u_file = replaced_filename; + fname_len_check = strdupdelim (u_file, u_file + strlen (u_file)); + } + + append_uri_pathel (fname_len_check, + fname_len_check + strlen (fname_len_check), false, &temp_fnres); + + /* Zero-terminate the temporary file name. */ + append_char ('\0', &temp_fnres); - /* Append "?query" to the file name. */ - u_query = u->query && *u->query ? u->query : NULL; - if (u_query) + /* Check that the length of the file name is acceptable. */ +#ifdef WINDOWS + if (MAX_PATH > (fnres.tail + CHOMP_BUFFER + 2)) { - append_char (FN_QUERY_SEP, &fnres); - append_uri_pathel (u_query, u_query + strlen (u_query), &fnres); + max_length = MAX_PATH - (fnres.tail + CHOMP_BUFFER + 2); + /* FIXME: In Windows a filename is usually limited to 255 characters. + To really be accurate you could call GetVolumeInformation() to get + lpMaximumComponentLength + */ + if (max_length > 255) + { + max_length = 255; + } + } + else + { + max_length = 0; + } +#else + max_length = get_max_length (fnres.base, fnres.tail, _PC_NAME_MAX) - CHOMP_BUFFER; +#endif + if (max_length > 0 && strlen (temp_fnres.base) > max_length) + { + logprintf (LOG_NOTQUIET, "The name is too long, %lu chars total.\n", + (unsigned long) strlen (temp_fnres.base)); + logprintf (LOG_NOTQUIET, "Trying to shorten...\n"); + + /* Shorten the file name. */ + temp_fnres.base[max_length] = '\0'; + + logprintf (LOG_NOTQUIET, "New name is %s.\n", temp_fnres.base); } - /* Zero-terminate the file name. */ - append_char ('\0', &fnres); + free (fname_len_check); + + /* The filename has already been 'cleaned' by append_uri_pathel() above. So, + * just append it. */ + if (fnres.tail) + append_char ('/', &fnres); + append_string (temp_fnres.base, &fnres); fname = fnres.base; + /* Make a final check that the path length is acceptable? */ + /* TODO: check fnres.base for path length problem */ + + free (temp_fnres.base); + /* Check the cases in which the unique extensions are not used: 1) Clobbering is turned off (-nc). 2) Retrieval with regetting. 3) Timestamping is used. 4) Hierarchy is built. + 5) Backups are specified. The exception is the case when file does exist and is a directory (see `mkalldirs' for explanation). */ - if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct) + if (ALLOW_CLOBBER && !(file_exists_p (fname) && !file_non_directory_p (fname))) - return fname; + { + unique = fname; + } + else + { + unique = unique_name (fname, true); + if (unique != fname) + xfree (fname); + } - unique = unique_name (fname, 1); - if (unique != fname) - xfree (fname); - return unique; -} +/* On VMS, alter the name as required. */ +#ifdef __VMS + { + char *unique2; -/* Return the length of URL's path. Path is considered to be - terminated by one of '?', ';', '#', or by the end of the - string. */ -static int -path_length (const char *url) -{ - const char *q = strpbrk_or_eos (url, "?;#"); - return q - url; -} + unique2 = ods_conform( unique); + if (unique2 != unique) + { + xfree (unique); + unique = unique2; + } + } +#endif /* def __VMS */ -/* Find the last occurrence of character C in the range [b, e), or - NULL, if none are present. This is equivalent to strrchr(b, c), - except that it accepts an END argument instead of requiring the - string to be zero-terminated. Why is there no memrchr()? */ -static const char * -find_last_char (const char *b, const char *e, char c) -{ - for (; e > b; e--) - if (*e == c) - return e; - return NULL; + return unique; } /* Resolve "." and ".." elements of PATH by destructively modifying - PATH. "." is resolved by removing that path element, and ".." is - resolved by removing the preceding path element. Leading and - trailing slashes are preserved. + PATH and return true if PATH has been modified, false otherwise. - Return non-zero if any changes have been made. + The algorithm is in spirit similar to the one described in rfc1808, + although implemented differently, in one pass. To recap, path + elements containing only "." are removed, and ".." is taken to mean + "back up one element". Single leading and trailing slashes are + preserved. For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive test examples are provided below. If you change anything in this function, run test_path_simplify to make sure you haven't broken a - test case. + test case. */ - A previous version of this function was based on path_simplify() - from GNU Bash, but it has been rewritten for Wget 1.8.1. */ - -static int -path_simplify (char *path) +static bool +path_simplify (enum url_scheme scheme, char *path) { - int change = 0; - char *p, *end; - - if (path[0] == '/') - ++path; /* preserve the leading '/'. */ + char *h = path; /* hare */ + char *t = path; /* tortoise */ + char *beg = path; + char *end = strchr (path, '\0'); - p = path; - end = p + strlen (p) + 1; /* position past the terminating zero. */ - - while (1) + while (h < end) { - again: - /* P should point to the beginning of a path element. */ - - if (*p == '.' && (*(p + 1) == '/' || *(p + 1) == '\0')) - { - /* Handle "./foo" by moving "foo" two characters to the - left. */ - if (*(p + 1) == '/') - { - change = 1; - memmove (p, p + 2, end - (p + 2)); - end -= 2; - goto again; - } - else - { - change = 1; - *p = '\0'; - break; - } - } - else if (*p == '.' && *(p + 1) == '.' - && (*(p + 2) == '/' || *(p + 2) == '\0')) - { - /* Handle "../foo" by moving "foo" one path element to the - left. */ - char *b = p; /* not p-1 because P can equal PATH */ - - /* Backtrack by one path element, but not past the beginning - of PATH. */ - - /* foo/bar/../baz */ - /* ^ p */ - /* ^ b */ - - if (b > path) - { - /* Move backwards until B hits the beginning of the - previous path element or the beginning of path. */ - for (--b; b > path && *(b - 1) != '/'; b--) - ; - } - - change = 1; - if (*(p + 2) == '/') - { - memmove (b, p + 3, end - (p + 3)); - end -= (p + 3) - b; - p = b; - } - else - { - *b = '\0'; - break; - } - - goto again; - } - else if (*p == '/') - { - /* Remove empty path elements. Not mandated by rfc1808 et - al, but it seems like a good idea to get rid of them. - Supporting them properly is hard (in which directory do - you save http://x.com///y.html?) and they don't seem to - bring much gain. */ - char *q = p; - while (*q == '/') - ++q; - change = 1; - if (*q == '\0') - { - *p = '\0'; - break; - } - memmove (p, q, end - q); - end -= q - p; - goto again; - } - - /* Skip to the next path element. */ - while (*p && *p != '/') - ++p; - if (*p == '\0') - break; - - /* Make sure P points to the beginning of the next path element, - which is location after the slash. */ - ++p; + /* Hare should be at the beginning of a path element. */ + + if (h[0] == '.' && (h[1] == '/' || h[1] == '\0')) + { + /* Ignore "./". */ + h += 2; + } + else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0')) + { + /* Handle "../" by retreating the tortoise by one path + element -- but not past beggining. */ + if (t > beg) + { + /* Move backwards until T hits the beginning of the + previous path element or the beginning of path. */ + for (--t; t > beg && t[-1] != '/'; t--) + ; + } + else if (scheme == SCHEME_FTP) + { + /* If we're at the beginning, copy the "../" literally + and move the beginning so a later ".." doesn't remove + it. This violates RFC 3986; but we do it for FTP + anyway because there is otherwise no way to get at a + parent directory, when the FTP server drops us in a + non-root directory (which is not uncommon). */ + beg = t + 3; + goto regular; + } + h += 3; + } + else + { + regular: + /* A regular path element. If H hasn't advanced past T, + simply skip to the next path element. Otherwise, copy + the path element until the next slash. */ + if (t == h) + { + /* Skip the path element, including the slash. */ + while (h < end && *h != '/') + t++, h++; + if (h < end) + t++, h++; + } + else + { + /* Copy the path element, including the final slash. */ + while (h < end && *h != '/') + *t++ = *h++; + if (h < end) + *t++ = *h++; + } + } } - return change; + if (t != h) + *t = '\0'; + + return t != h; } +/* Return the length of URL's path. Path is considered to be + terminated by one or more of the ?query or ;params or #fragment, + depending on the scheme. */ + +static const char * +path_end (const char *url) +{ + enum url_scheme scheme = url_scheme (url); + const char *seps; + if (scheme == SCHEME_INVALID) + scheme = SCHEME_HTTP; /* use http semantics for rel links */ + /* +2 to ignore the first two separators ':' and '/' */ + seps = init_seps (scheme) + 2; + return strpbrk_or_eos (url, seps); +} + +/* Find the last occurrence of character C in the range [b, e), or + NULL, if none are present. */ +#define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b)) + /* Merge BASE with LINK and return the resulting URI. Either of the URIs may be absolute or relative, complete with the @@ -1739,8 +1809,10 @@ path_simplify (char *path) foreseeable cases. It only employs minimal URL parsing, without knowledge of the specifics of schemes. - Perhaps this function should call path_simplify so that the callers - don't have to call url_parse unconditionally. */ + I briefly considered making this function call path_simplify after + the merging process, as rfc1738 seems to suggest. This is a bad + idea for several reasons: 1) it complexifies the code, and 2) + url_parse has to simplify path anyway, so it's wasteful to boot. */ char * uri_merge (const char *base, const char *link) @@ -1753,7 +1825,7 @@ uri_merge (const char *base, const char *link) return xstrdup (link); /* We may not examine BASE past END. */ - end = base + path_length (base); + end = path_end (base); linklength = strlen (link); if (!*link) @@ -1764,7 +1836,7 @@ uri_merge (const char *base, const char *link) else if (*link == '?') { /* LINK points to the same location, but changes the query - string. Examples: */ + string. Examples: */ /* uri_merge("path", "?new") -> "path?new" */ /* uri_merge("path?foo", "?new") -> "path?new" */ /* uri_merge("path?foo#bar", "?new") -> "path?new" */ @@ -1784,7 +1856,7 @@ uri_merge (const char *base, const char *link) int baselength; const char *end1 = strchr (base, '#'); if (!end1) - end1 = base + strlen (base); + end1 = base + strlen (base); baselength = end1 - base; merge = xmalloc (baselength + linklength + 1); memcpy (merge, base, baselength); @@ -1794,8 +1866,8 @@ uri_merge (const char *base, const char *link) else if (*link == '/' && *(link + 1) == '/') { /* LINK begins with "//" and so is a net path: we need to - replace everything after (and including) the double slash - with LINK. */ + replace everything after (and including) the double slash + with LINK. */ /* uri_merge("foo", "//new/bar") -> "//new/bar" */ /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */ @@ -1808,128 +1880,112 @@ uri_merge (const char *base, const char *link) /* Look for first slash. */ slash = memchr (base, '/', end - base); /* If found slash and it is a double slash, then replace - from this point, else default to replacing from the - beginning. */ + from this point, else default to replacing from the + beginning. */ if (slash && *(slash + 1) == '/') - start_insert = slash; + start_insert = slash; else - start_insert = base; + start_insert = base; span = start_insert - base; - merge = (char *)xmalloc (span + linklength + 1); + merge = xmalloc (span + linklength + 1); if (span) - memcpy (merge, base, span); + memcpy (merge, base, span); memcpy (merge + span, link, linklength); merge[span + linklength] = '\0'; } else if (*link == '/') { /* LINK is an absolute path: we need to replace everything - after (and including) the FIRST slash with LINK. + after (and including) the FIRST slash with LINK. - So, if BASE is "http://host/whatever/foo/bar", and LINK is - "/qux/xyzzy", our result should be - "http://host/qux/xyzzy". */ + So, if BASE is "http://host/whatever/foo/bar", and LINK is + "/qux/xyzzy", our result should be + "http://host/qux/xyzzy". */ int span; const char *slash; const char *start_insert = NULL; /* for gcc to shut up. */ const char *pos = base; - int seen_slash_slash = 0; + bool seen_slash_slash = false; /* We're looking for the first slash, but want to ignore - double slash. */ + double slash. */ again: slash = memchr (pos, '/', end - pos); if (slash && !seen_slash_slash) - if (*(slash + 1) == '/') - { - pos = slash + 2; - seen_slash_slash = 1; - goto again; - } + if (*(slash + 1) == '/') + { + pos = slash + 2; + seen_slash_slash = true; + goto again; + } /* At this point, SLASH is the location of the first / after - "//", or the first slash altogether. START_INSERT is the - pointer to the location where LINK will be inserted. When - examining the last two examples, keep in mind that LINK - begins with '/'. */ + "//", or the first slash altogether. START_INSERT is the + pointer to the location where LINK will be inserted. When + examining the last two examples, keep in mind that LINK + begins with '/'. */ if (!slash && !seen_slash_slash) - /* example: "foo" */ - /* ^ */ - start_insert = base; + /* example: "foo" */ + /* ^ */ + start_insert = base; else if (!slash && seen_slash_slash) - /* example: "http://foo" */ - /* ^ */ - start_insert = end; + /* example: "http://foo" */ + /* ^ */ + start_insert = end; else if (slash && !seen_slash_slash) - /* example: "foo/bar" */ - /* ^ */ - start_insert = base; + /* example: "foo/bar" */ + /* ^ */ + start_insert = base; else if (slash && seen_slash_slash) - /* example: "http://something/" */ - /* ^ */ - start_insert = slash; + /* example: "http://something/" */ + /* ^ */ + start_insert = slash; span = start_insert - base; - merge = (char *)xmalloc (span + linklength + 1); + merge = xmalloc (span + linklength + 1); if (span) - memcpy (merge, base, span); + memcpy (merge, base, span); memcpy (merge + span, link, linklength); merge[span + linklength] = '\0'; } else { /* LINK is a relative URL: we need to replace everything - after last slash (possibly empty) with LINK. + after last slash (possibly empty) with LINK. - So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy", - our result should be "whatever/foo/qux/xyzzy". */ - int need_explicit_slash = 0; + So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy", + our result should be "whatever/foo/qux/xyzzy". */ + bool need_explicit_slash = false; int span; const char *start_insert; const char *last_slash = find_last_char (base, end, '/'); if (!last_slash) - { - /* No slash found at all. Append LINK to what we have, - but we'll need a slash as a separator. - - Example: if base == "foo" and link == "qux/xyzzy", then - we cannot just append link to base, because we'd get - "fooqux/xyzzy", whereas what we want is - "foo/qux/xyzzy". - - To make sure the / gets inserted, we set - need_explicit_slash to 1. We also set start_insert - to end + 1, so that the length calculations work out - correctly for one more (slash) character. Accessing - that character is fine, since it will be the - delimiter, '\0' or '?'. */ - /* example: "foo?..." */ - /* ^ ('?' gets changed to '/') */ - start_insert = end + 1; - need_explicit_slash = 1; - } + { + /* No slash found at all. Replace what we have with LINK. */ + start_insert = base; + } else if (last_slash && last_slash >= base + 2 - && last_slash[-2] == ':' && last_slash[-1] == '/') - { - /* example: http://host" */ - /* ^ */ - start_insert = end + 1; - need_explicit_slash = 1; - } + && last_slash[-2] == ':' && last_slash[-1] == '/') + { + /* example: http://host" */ + /* ^ */ + start_insert = end + 1; + need_explicit_slash = true; + } else - { - /* example: "whatever/foo/bar" */ - /* ^ */ - start_insert = last_slash + 1; - } + { + /* example: "whatever/foo/bar" */ + /* ^ */ + start_insert = last_slash + 1; + } span = start_insert - base; - merge = (char *)xmalloc (span + linklength + 1); + merge = xmalloc (span + linklength + 1); if (span) - memcpy (merge, base, span); + memcpy (merge, base, span); if (need_explicit_slash) - merge[span - 1] = '/'; + merge[span - 1] = '/'; memcpy (merge + span, link, linklength); merge[span + linklength] = '\0'; } @@ -1937,10 +1993,10 @@ uri_merge (const char *base, const char *link) return merge; } -#define APPEND(p, s) do { \ - int len = strlen (s); \ - memcpy (p, s, len); \ - p += len; \ +#define APPEND(p, s) do { \ + int len = strlen (s); \ + memcpy (p, s, len); \ + p += len; \ } while (0) /* Use this instead of password when the actual password is supposed @@ -1951,54 +2007,65 @@ uri_merge (const char *base, const char *link) /* Recreate the URL string from the data in URL. - If HIDE is non-zero (as it is when we're calling this on a URL we - plan to print, but not when calling it to canonicalize a URL for - use within the program), password will be hidden. Unsafe - characters in the URL will be quoted. */ + If HIDE is true (as it is when we're calling this on a URL we plan + to print, but not when calling it to canonicalize a URL for use + within the program), password will be hidden. Unsafe characters in + the URL will be quoted. */ char * -url_string (const struct url *url, int hide_password) +url_string (const struct url *url, enum url_auth_mode auth_mode) { int size; char *result, *p; - char *quoted_user = NULL, *quoted_passwd = NULL; + char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL; - int scheme_port = supported_schemes[url->scheme].default_port; - char *scheme_str = supported_schemes[url->scheme].leading_string; + int scheme_port = supported_schemes[url->scheme].default_port; + const char *scheme_str = supported_schemes[url->scheme].leading_string; int fplen = full_path_length (url); - int brackets_around_host = 0; + bool brackets_around_host; assert (scheme_str != NULL); /* Make sure the user name and password are quoted. */ if (url->user) { - quoted_user = url_escape_allow_passthrough (url->user); - if (url->passwd) - { - if (hide_password) - quoted_passwd = HIDDEN_PASSWORD; - else - quoted_passwd = url_escape_allow_passthrough (url->passwd); - } + if (auth_mode != URL_AUTH_HIDE) + { + quoted_user = url_escape_allow_passthrough (url->user); + if (url->passwd) + { + if (auth_mode == URL_AUTH_HIDE_PASSWD) + quoted_passwd = (char *) HIDDEN_PASSWORD; + else + quoted_passwd = url_escape_allow_passthrough (url->passwd); + } + } } - if (strchr (url->host, ':')) - brackets_around_host = 1; + /* In the unlikely event that the host name contains non-printable + characters, quote it for displaying to the user. */ + quoted_host = url_escape_allow_passthrough (url->host); + + /* Undo the quoting of colons that URL escaping performs. IPv6 + addresses may legally contain colons, and in that case must be + placed in square brackets. */ + if (quoted_host != url->host) + unescape_single_char (quoted_host, ':'); + brackets_around_host = strchr (quoted_host, ':') != NULL; size = (strlen (scheme_str) - + strlen (url->host) - + (brackets_around_host ? 2 : 0) - + fplen - + 1); + + strlen (quoted_host) + + (brackets_around_host ? 2 : 0) + + fplen + + 1); if (url->port != scheme_port) size += 1 + numdigit (url->port); if (quoted_user) { size += 1 + strlen (quoted_user); if (quoted_passwd) - size += 1 + strlen (quoted_passwd); + size += 1 + strlen (quoted_passwd); } p = result = xmalloc (size); @@ -2008,16 +2075,16 @@ url_string (const struct url *url, int hide_password) { APPEND (p, quoted_user); if (quoted_passwd) - { - *p++ = ':'; - APPEND (p, quoted_passwd); - } + { + *p++ = ':'; + APPEND (p, quoted_passwd); + } *p++ = '@'; } if (brackets_around_host) *p++ = '['; - APPEND (p, url->host); + APPEND (p, quoted_host); if (brackets_around_host) *p++ = ']'; if (url->port != scheme_port) @@ -2034,34 +2101,98 @@ url_string (const struct url *url, int hide_password) if (quoted_user && quoted_user != url->user) xfree (quoted_user); - if (quoted_passwd && !hide_password + if (quoted_passwd && auth_mode == URL_AUTH_SHOW && quoted_passwd != url->passwd) xfree (quoted_passwd); + if (quoted_host != url->host) + xfree (quoted_host); return result; } -/* Return non-zero if scheme a is similar to scheme b. - +/* Return true if scheme a is similar to scheme b. + Schemes are similar if they are equal. If SSL is supported, schemes are also similar if one is http (SCHEME_HTTP) and the other is https (SCHEME_HTTPS). */ -int +bool schemes_are_similar_p (enum url_scheme a, enum url_scheme b) { if (a == b) - return 1; + return true; #ifdef HAVE_SSL if ((a == SCHEME_HTTP && b == SCHEME_HTTPS) || (a == SCHEME_HTTPS && b == SCHEME_HTTP)) - return 1; + return true; #endif - return 0; + return false; } -#if 0 +static int +getchar_from_escaped_string (const char *str, char *c) +{ + const char *p = str; + + assert (str && *str); + assert (c); + + if (p[0] == '%') + { + if (!c_isxdigit(p[1]) || !c_isxdigit(p[2])) + { + *c = '%'; + return 1; + } + else + { + if (p[2] == 0) + return 0; /* error: invalid string */ + + *c = X2DIGITS_TO_NUM (p[1], p[2]); + if (URL_RESERVED_CHAR(*c)) + { + *c = '%'; + return 1; + } + else + return 3; + } + } + else + { + *c = p[0]; + } + + return 1; +} + +bool +are_urls_equal (const char *u1, const char *u2) +{ + const char *p, *q; + int pp, qq; + char ch1, ch2; + assert(u1 && u2); + + p = u1; + q = u2; + + while (*p && *q + && (pp = getchar_from_escaped_string (p, &ch1)) + && (qq = getchar_from_escaped_string (q, &ch2)) + && (c_tolower(ch1) == c_tolower(ch2))) + { + p += pp; + q += qq; + } + + return (*p == 0 && *q == 0 ? true : false); +} + +#ifdef TESTING /* Debugging and testing support for path_simplify. */ +#if 0 /* Debug: run path_simplify on PATH and return the result in a new string. Useful for calling from the debugger. */ static char * @@ -2071,83 +2202,146 @@ ps (char *path) path_simplify (copy); return copy; } +#endif -static void -run_test (char *test, char *expected_result, int expected_change) +static const char * +run_test (const char *test, const char *expected_result, enum url_scheme scheme, + bool expected_change) { char *test_copy = xstrdup (test); - int modified = path_simplify (test_copy); + bool modified = path_simplify (scheme, test_copy); if (0 != strcmp (test_copy, expected_result)) { printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n", - test, expected_result, test_copy); + test, expected_result, test_copy); + mu_assert ("", 0); } if (modified != expected_change) { - if (expected_change == 1) - printf ("Expected no modification with path_simplify(\"%s\").\n", - test); + if (expected_change) + printf ("Expected modification with path_simplify(\"%s\").\n", + test); else - printf ("Expected modification with path_simplify(\"%s\").\n", - test); + printf ("Expected no modification with path_simplify(\"%s\").\n", + test); } xfree (test_copy); + mu_assert ("", modified == expected_change); + return NULL; } -static void +const char * test_path_simplify (void) { - static struct { - char *test, *result; - int should_modify; + static const struct { + const char *test, *result; + enum url_scheme scheme; + bool should_modify; } tests[] = { - { "", "", 0 }, - { ".", "", 1 }, - { "..", "", 1 }, - { "foo", "foo", 0 }, - { "foo/bar", "foo/bar", 0 }, - { "foo///bar", "foo/bar", 1 }, - { "foo/.", "foo/", 1 }, - { "foo/./", "foo/", 1 }, - { "foo./", "foo./", 0 }, - { "foo/../bar", "bar", 1 }, - { "foo/../bar/", "bar/", 1 }, - { "foo/bar/..", "foo/", 1 }, - { "foo/bar/../x", "foo/x", 1 }, - { "foo/bar/../x/", "foo/x/", 1 }, - { "foo/..", "", 1 }, - { "foo/../..", "", 1 }, - { "a/b/../../c", "c", 1 }, - { "./a/../b", "b", 1 } + { "", "", SCHEME_HTTP, false }, + { ".", "", SCHEME_HTTP, true }, + { "./", "", SCHEME_HTTP, true }, + { "..", "", SCHEME_HTTP, true }, + { "../", "", SCHEME_HTTP, true }, + { "..", "..", SCHEME_FTP, false }, + { "../", "../", SCHEME_FTP, false }, + { "foo", "foo", SCHEME_HTTP, false }, + { "foo/bar", "foo/bar", SCHEME_HTTP, false }, + { "foo///bar", "foo///bar", SCHEME_HTTP, false }, + { "foo/.", "foo/", SCHEME_HTTP, true }, + { "foo/./", "foo/", SCHEME_HTTP, true }, + { "foo./", "foo./", SCHEME_HTTP, false }, + { "foo/../bar", "bar", SCHEME_HTTP, true }, + { "foo/../bar/", "bar/", SCHEME_HTTP, true }, + { "foo/bar/..", "foo/", SCHEME_HTTP, true }, + { "foo/bar/../x", "foo/x", SCHEME_HTTP, true }, + { "foo/bar/../x/", "foo/x/", SCHEME_HTTP, true }, + { "foo/..", "", SCHEME_HTTP, true }, + { "foo/../..", "", SCHEME_HTTP, true }, + { "foo/../../..", "", SCHEME_HTTP, true }, + { "foo/../../bar/../../baz", "baz", SCHEME_HTTP, true }, + { "foo/../..", "..", SCHEME_FTP, true }, + { "foo/../../..", "../..", SCHEME_FTP, true }, + { "foo/../../bar/../../baz", "../../baz", SCHEME_FTP, true }, + { "a/b/../../c", "c", SCHEME_HTTP, true }, + { "./a/../b", "b", SCHEME_HTTP, true } }; - int i; + unsigned i; for (i = 0; i < countof (tests); i++) { - char *test = tests[i].test; - char *expected_result = tests[i].result; - int expected_change = tests[i].should_modify; - run_test (test, expected_result, expected_change); + const char *message; + const char *test = tests[i].test; + const char *expected_result = tests[i].result; + enum url_scheme scheme = tests[i].scheme; + bool expected_change = tests[i].should_modify; + + message = run_test (test, expected_result, scheme, expected_change); + if (message) return message; } + return NULL; +} - /* Now run all the tests with a leading slash before the test case, - to prove that the slash is being preserved. */ - for (i = 0; i < countof (tests); i++) +const char * +test_append_uri_pathel(void) +{ + unsigned i; + static const struct { + const char *original_url; + const char *input; + bool escaped; + const char *expected_result; + } test_array[] = { + { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" }, + }; + + for (i = 0; i < countof(test_array); ++i) { - char *test, *expected_result; - int expected_change = tests[i].should_modify; + struct growable dest; + const char *p = test_array[i].input; + + memset (&dest, 0, sizeof (dest)); - test = xmalloc (1 + strlen (tests[i].test) + 1); - sprintf (test, "/%s", tests[i].test); + append_string (test_array[i].original_url, &dest); + append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest); - expected_result = xmalloc (1 + strlen (tests[i].result) + 1); - sprintf (expected_result, "/%s", tests[i].result); + mu_assert ("test_append_uri_pathel: wrong result", + strcmp (dest.base, test_array[i].expected_result) == 0); + } - run_test (test, expected_result, expected_change); + return NULL; +} - xfree (test); - xfree (expected_result); +const char * +test_are_urls_equal(void) +{ + unsigned i; + static const struct { + const char *url1; + const char *url2; + bool expected_result; + } test_array[] = { + { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true }, + { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false }, + { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false }, + { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true }, + { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false }, + { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false }, + }; + + for (i = 0; i < countof(test_array); ++i) + { + mu_assert ("test_are_urls_equal: wrong result", + are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result); } + + return NULL; } -#endif + +#endif /* TESTING */ + +/* + * vim: et ts=2 sw=2 + */ +