X-Git-Url: http://sjero.net/git/?a=blobdiff_plain;f=src%2Furl.c;h=e1685ab9aa8554093b0ba6d7bd6e817f1b951324;hb=d5be8ecca466601bda9b81c28a79077fbda6ccde;hp=c077f000250598ef028ce96395ccf2b593146258;hpb=b0b1c815c15e49c9172f59428810713097a65e37;p=wget diff --git a/src/url.c b/src/url.c index c077f000..e1685ab9 100644 --- a/src/url.c +++ b/src/url.c @@ -1,20 +1,20 @@ /* URL handling. - Copyright (C) 1995, 1996, 1997, 2000 Free Software Foundation, Inc. + Copyright (C) 1995, 1996, 1997, 2000, 2001 Free Software Foundation, Inc. -This file is part of Wget. +This file is part of GNU Wget. -This program is free software; you can redistribute it and/or modify +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 your option) any later version. -This program is distributed in the hope that it will be useful, +GNU Wget is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of 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 this program; if not, write to the Free Software +along with Wget; if not, write to the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ #include @@ -26,7 +26,6 @@ Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ #else # include #endif -#include #include #ifdef HAVE_UNISTD_H # include @@ -43,165 +42,87 @@ Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ extern int errno; #endif -/* Default port definitions */ -#define DEFAULT_HTTP_PORT 80 -#define DEFAULT_FTP_PORT 21 - -/* Table of Unsafe chars. This is intialized in - init_unsafe_char_table. */ - -static char unsafe_char_table[256]; - -#define UNSAFE_CHAR(c) (unsafe_char_table[(unsigned char)(c)]) - -/* If S contains unsafe characters, free it and replace it with a - version that doesn't. */ -#define URL_CLEANSE(s) do \ -{ \ - if (contains_unsafe (s)) \ - { \ - char *uc_tmp = encode_string (s); \ - free (s); \ - (s) = uc_tmp; \ - } \ -} while (0) - -/* Is a directory "."? */ +/* Is X "."? */ #define DOTP(x) ((*(x) == '.') && (!*(x + 1))) -/* Is a directory ".."? */ +/* Is X ".."? */ #define DDOTP(x) ((*(x) == '.') && (*(x + 1) == '.') && (!*(x + 2))) -#if 0 -static void path_simplify_with_kludge PARAMS ((char *)); -#endif static int urlpath_length PARAMS ((const char *)); -/* NULL-terminated list of strings to be recognized as prototypes (URL - schemes). Note that recognized doesn't mean supported -- only HTTP - and FTP are currently supported. - - However, a string that does not match anything in the list will be - considered a relative URL. Thus it's important that this list has - anything anyone could think of being legal. - - There are wild things here. :-) Take a look at - for more - fun. */ -static char *protostrings[] = +struct scheme_data { - "cid:", - "clsid:", - "file:", - "finger:", - "ftp:", - "gopher:", - "hdl:", - "http:", - "https:", - "ilu:", - "ior:", - "irc:", - "java:", - "javascript:", - "lifn:", - "mailto:", - "mid:", - "news:", - "nntp:", - "path:", - "prospero:", - "rlogin:", - "service:", - "shttp:", - "snews:", - "stanf:", - "telnet:", - "tn3270:", - "wais:", - "whois++:", - NULL + char *leading_string; + int default_port; }; -struct proto +/* Supported schemes: */ +static struct scheme_data supported_schemes[] = { - char *name; - uerr_t ind; - unsigned short port; -}; + { "http://", DEFAULT_HTTP_PORT }, +#ifdef HAVE_SSL + { "https://", DEFAULT_HTTPS_PORT }, +#endif + { "ftp://", DEFAULT_FTP_PORT }, -/* Similar to former, but for supported protocols: */ -static struct proto sup_protos[] = -{ - { "http://", URLHTTP, DEFAULT_HTTP_PORT }, - { "ftp://", URLFTP, DEFAULT_FTP_PORT }, - /*{ "file://", URLFILE, DEFAULT_FTP_PORT },*/ + /* SCHEME_INVALID */ + { NULL, -1 } }; -static void parse_dir PARAMS ((const char *, char **, char **)); -static uerr_t parse_uname PARAMS ((const char *, char **, char **)); -static char *construct PARAMS ((const char *, const char *, int , int)); static char *construct_relative PARAMS ((const char *, const char *)); -static char process_ftp_type PARAMS ((char *)); -/* Returns the number of characters to be skipped if the first thing - in a URL is URL: (which is 0 or 4+). The optional spaces after - URL: are also skipped. */ -int -skip_url (const char *url) -{ - int i; +/* 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. */ - if (TOUPPER (url[0]) == 'U' - && TOUPPER (url[1]) == 'R' - && TOUPPER (url[2]) == 'L' - && url[3] == ':') - { - /* Skip blanks. */ - for (i = 4; url[i] && ISSPACE (url[i]); i++); - return i; - } - else - return 0; -} +enum { + urlchr_reserved = 1, + urlchr_unsafe = 2 +}; -/* Unsafe chars: - - anything <= 32; - - stuff from rfc1738 ("<>\"#%{}|\\^~[]`"); - - @ and :, for user/password encoding. - - everything over 127 (but we don't bother with recording those. */ -void -init_unsafe_char_table (void) -{ - int i; - for (i = 0; i < 256; i++) - if (i < 32 || i >= 127 - || i == '<' - || i == '>' - || i == '\"' - || i == '#' - || i == '%' - || i == '{' - || i == '}' - || i == '|' - || i == '\\' - || i == '^' - || i == '~' - || i == '[' - || i == ']' - || i == '`') - unsafe_char_table[i] = 1; -} +#define R urlchr_reserved +#define U urlchr_unsafe +#define RU R|U -/* Returns 1 if the string contains unsafe characters, 0 otherwise. */ -int -contains_unsafe (const char *s) +#define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask)) + +/* rfc1738 reserved chars, preserved from encoding. */ + +#define RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved) + +/* rfc1738 unsafe chars, plus some more. */ + +#define UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe) + +const static unsigned char urlchr_table[256] = { - for (; *s; s++) - if (UNSAFE_CHAR (*s)) - return 1; - return 0; -} + 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, /* ( ) * + , - . / */ + 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 */ + 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, 0, /* X Y Z [ \ ] ^ _ */ + 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 */ + + 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, + 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, + + 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, + 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, +}; /* Decodes the forms %xy in a URL to the character the hexadecimal code of which is xy. xy are hexadecimal digits from @@ -212,574 +133,804 @@ contains_unsafe (const char *s) static void decode_string (char *s) { - char *p = s; + char *t = s; /* t - tortoise */ + char *h = s; /* h - hare */ - for (; *s; s++, p++) + for (; *h; h++, t++) { - if (*s != '%') - *p = *s; + if (*h != '%') + { + copychar: + *t = *h; + } else { - /* Do nothing if at the end of the string, or if the chars - are not hex-digits. */ - if (!*(s + 1) || !*(s + 2) - || !(ISXDIGIT (*(s + 1)) && ISXDIGIT (*(s + 2)))) - { - *p = *s; - continue; - } - *p = (ASC2HEXD (*(s + 1)) << 4) + ASC2HEXD (*(s + 2)); - s += 2; + /* Do nothing if '%' is not followed by two hex digits. */ + if (!*(h + 1) || !*(h + 2) + || !(ISXDIGIT (*(h + 1)) && ISXDIGIT (*(h + 2)))) + goto copychar; + *t = (XCHAR_TO_XDIGIT (*(h + 1)) << 4) + XCHAR_TO_XDIGIT (*(h + 2)); + h += 2; } } - *p = '\0'; + *t = '\0'; } -/* Encode the unsafe characters (as determined by URL_UNSAFE) in a +/* Like encode_string, but return S if there are no unsafe chars. */ + +static char * +encode_string_maybe (const char *s) +{ + const char *p1; + char *p2, *newstr; + int newlen; + int addition = 0; + + for (p1 = s; *p1; p1++) + if (UNSAFE_CHAR (*p1)) + addition += 2; /* Two more characters (hex digits) */ + + if (!addition) + return (char *)s; + + newlen = (p1 - s) + addition; + newstr = (char *)xmalloc (newlen + 1); + + p1 = s; + p2 = newstr; + while (*p1) + { + if (UNSAFE_CHAR (*p1)) + { + const unsigned char c = *p1++; + *p2++ = '%'; + *p2++ = XDIGIT_TO_XCHAR (c >> 4); + *p2++ = XDIGIT_TO_XCHAR (c & 0xf); + } + else + *p2++ = *p1++; + } + *p2 = '\0'; + assert (p2 - newstr == newlen); + + return newstr; +} + +/* Encode the unsafe characters (as determined by UNSAFE_CHAR) in a given string, returning a malloc-ed %XX encoded string. */ + char * encode_string (const char *s) { - const char *b; - char *p, *res; - int i; - - b = s; - for (i = 0; *s; s++, i++) - if (UNSAFE_CHAR (*s)) - i += 2; /* Two more characters (hex digits) */ - res = (char *)xmalloc (i + 1); - s = b; - for (p = res; *s; s++) - if (UNSAFE_CHAR (*s)) - { - const unsigned char c = *s; - *p++ = '%'; - *p++ = HEXD2ASC (c >> 4); - *p++ = HEXD2ASC (c & 0xf); - } - else - *p++ = *s; - *p = '\0'; - return res; + char *encoded = encode_string_maybe (s); + if (encoded != s) + return encoded; + else + return xstrdup (s); } + +/* Encode unsafe characters in PTR to %xx. If such encoding is done, + the old value of PTR is freed and PTR is made to point to the newly + allocated storage. */ + +#define ENCODE(ptr) do { \ + char *e_new = encode_string_maybe (ptr); \ + if (e_new != ptr) \ + { \ + xfree (ptr); \ + ptr = e_new; \ + } \ +} while (0) -/* Returns the proto-type if URL's protocol is supported, or - URLUNKNOWN if not. */ -uerr_t -urlproto (const char *url) -{ - int i; +enum copy_method { CM_DECODE, CM_ENCODE, CM_PASSTHROUGH }; - url += skip_url (url); - for (i = 0; i < ARRAY_SIZE (sup_protos); i++) - if (!strncasecmp (url, sup_protos[i].name, strlen (sup_protos[i].name))) - return sup_protos[i].ind; - for (i = 0; url[i] && url[i] != ':' && url[i] != '/'; i++); - if (url[i] == ':') +/* 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) +{ + if (*p == '%') { - for (++i; url[i] && url[i] != '/'; i++) - if (!ISDIGIT (url[i])) - return URLBADPORT; - if (url[i - 1] == ':') - return URLFTP; + 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 = (XCHAR_TO_XDIGIT (*(p + 1)) << 4) + + XCHAR_TO_XDIGIT (*(p + 2)); + + if (UNSAFE_CHAR (preempt) || RESERVED_CHAR (preempt)) + return CM_PASSTHROUGH; + else + return CM_DECODE; + } else - return URLHTTP; + /* Garbled %.. sequence: encode `%'. */ + return CM_ENCODE; } + else if (UNSAFE_CHAR (*p) && !RESERVED_CHAR (*p)) + return CM_ENCODE; else - return URLHTTP; + return CM_PASSTHROUGH; } -/* Skip the protocol part of the URL, e.g. `http://'. If no protocol - part is found, returns 0. */ +/* Translate a %-quoting (but possibly non-conformant) input string S + into a %-quoting (and conformant) output string. If no characters + are encoded or decoded, return the same string S; otherwise, return + a freshly allocated string with the new contents. + + 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 decode_string() 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. + + 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: + + GET /abc%20def HTTP/1.0 + + So it appears that the unsafe chars need to be quoted, as with + encode_string. But what if we're requested to download + `abc%20def'? Remember that %-encoding is valid URL syntax, so what + the user meant was a literal space, and he was kind enough to quote + it. In that case, Wget should obviously leave the `%20' as is, and + send the same request as above. So in this case we may not call + encode_string. + + But what if the requested URI is `abc%20 def'? If we call + encode_string, we end up with `/abc%2520%20def', which is almost + certainly not intended. If we don't call encode_string, we are + left with the embedded space and cannot send the request. What the + user meant was for Wget to request `/abc%20%20def', and this is + where reencode_string kicks in. + + Wget used to solve this by first decoding %-quotes, and then + encoding all the "unsafe" characters found in the resulting string. + This was wrong because it didn't preserve certain URL special + (reserved) characters. For instance, URI containing "a%2B+b" (0x2b + == '+') would get translated to "a%2B%2Bb" or "a++b" depending on + whether we considered `+' reserved (it is). One of these results + is inevitable because by the second step we would lose information + on whether the `+' was originally encoded or not. Both results + were wrong because in CGI parameters + means space, while %2B means + literal plus. reencode_string correctly translates the above to + "a%2B+b", i.e. returns the original string. + + This function uses an algorithm proposed by Anon Sricharoenchai: + + 1. Encode all URL_UNSAFE and the "%" that are not followed by 2 + hexdigits. + + 2. Decode all "%XX" except URL_UNSAFE, URL_RESERVED (";/?:@=&") and + "+". + + ...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. + + 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" + + Simpler test cases: + + "foo bar" -> "foo%20bar" + "foo%20bar" -> "foo%20bar" + "foo %20bar" -> "foo%20%20bar" + "foo%%20bar" -> "foo%25%20bar" (0x25 == '%') + "foo%25%20bar" -> "foo%25%20bar" + "foo%2%20bar" -> "foo%252%20bar" + "foo+bar" -> "foo+bar" (plus is reserved!) + "foo%2b+bar" -> "foo%2b+bar" */ + +char * +reencode_string (const char *s) +{ + const char *p1; + char *newstr, *p2; + int oldlen, newlen; + + int encode_count = 0; + int decode_count = 0; + + /* First, pass through 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 (!encode_count && !decode_count) + /* The string is good as it is. */ + 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); + newstr = xmalloc (newlen + 1); + + p1 = s; + p2 = newstr; + + while (*p1) + { + switch (decide_copy_method (p1)) + { + case CM_ENCODE: + { + char c = *p1++; + *p2++ = '%'; + *p2++ = XDIGIT_TO_XCHAR (c >> 4); + *p2++ = XDIGIT_TO_XCHAR (c & 0xf); + } + break; + case CM_DECODE: + *p2++ = ((XCHAR_TO_XDIGIT (*(p1 + 1)) << 4) + + (XCHAR_TO_XDIGIT (*(p1 + 2)))); + p1 += 3; /* skip %xx */ + break; + case CM_PASSTHROUGH: + *p2++ = *p1++; + } + } + *p2 = '\0'; + assert (p2 - newstr == newlen); + return newstr; +} + +/* Run PTR_VAR through reencode_string. If a new string is consed, + free PTR_VAR and make it point to the new storage. Obviously, + PTR_VAR needs to be an lvalue. */ + +#define REENCODE(ptr_var) do { \ + char *rf_new = reencode_string (ptr_var); \ + if (rf_new != ptr_var) \ + { \ + xfree (ptr_var); \ + ptr_var = rf_new; \ + } \ +} while (0) + +/* Returns the scheme type if the scheme is supported, or + SCHEME_INVALID if not. */ +enum url_scheme +url_scheme (const char *url) +{ + int i; + + for (i = 0; supported_schemes[i].leading_string; i++) + if (!strncasecmp (url, supported_schemes[i].leading_string, + strlen (supported_schemes[i].leading_string))) + return (enum url_scheme)i; + return SCHEME_INVALID; +} + +/* Return the number of characters needed to skip the scheme part of + the URL, e.g. `http://'. If no scheme is found, returns 0. */ int -skip_proto (const char *url) +url_skip_scheme (const char *url) { - char **s; - int l; + const char *p = url; - for (s = protostrings; *s; s++) - if (!strncasecmp (*s, url, strlen (*s))) - break; - if (!*s) + /* Skip the scheme name. We allow `-' and `+' because of `whois++', + etc. */ + while (ISALNUM (*p) || *p == '-' || *p == '+') + ++p; + if (*p != ':') return 0; - l = strlen (*s); - /* HTTP and FTP protocols are expected to yield exact host names - (i.e. the `//' part must be skipped, too). */ - if (!strcmp (*s, "http:") || !strcmp (*s, "ftp:")) - l += 2; - return l; + /* Skip ':'. */ + ++p; + + /* Skip "//" if found. */ + if (*p == '/' && *(p + 1) == '/') + p += 2; + + return p - url; } -/* Returns 1 if the URL begins with a protocol (supported or +/* Returns 1 if the URL begins with a scheme (supported or unsupported), 0 otherwise. */ int -has_proto (const char *url) +url_has_scheme (const char *url) { - char **s; + const char *p = url; + while (ISALNUM (*p) || *p == '-' || *p == '+') + ++p; + return *p == ':'; +} - url += skip_url (url); - for (s = protostrings; *s; s++) - if (strncasecmp (url, *s, strlen (*s)) == 0) - return 1; - return 0; +int +scheme_default_port (enum url_scheme scheme) +{ + return supported_schemes[scheme].default_port; } /* Skip the username and password, if present here. The function should be called *not* with the complete URL, but with the part - right after the protocol. + right after the scheme. If no username and password are found, return 0. */ int -skip_uname (const char *url) +url_skip_uname (const char *url) { const char *p; - for (p = url; *p && *p != '/'; p++) - if (*p == '@') - break; - /* If a `@' was found before the first occurrence of `/', skip - it. */ - if (*p == '@') - return p - url + 1; - else + + /* Look for '@' that comes before '/' or '?'. */ + p = (const char *)strpbrk (url, "/?@"); + if (!p || *p != '@') return 0; -} - -/* Allocate a new urlinfo structure, fill it with default values and - return a pointer to it. */ -struct urlinfo * -newurl (void) -{ - struct urlinfo *u; - u = (struct urlinfo *)xmalloc (sizeof (struct urlinfo)); - memset (u, 0, sizeof (*u)); - u->proto = URLUNKNOWN; - return u; + return p - url + 1; } -/* Perform a "deep" free of the urlinfo structure. The structure - should have been created with newurl, but need not have been used. - If free_pointer is non-0, free the pointer itself. */ -void -freeurl (struct urlinfo *u, int complete) -{ - assert (u != NULL); - FREE_MAYBE (u->url); - FREE_MAYBE (u->host); - FREE_MAYBE (u->path); - FREE_MAYBE (u->file); - FREE_MAYBE (u->dir); - FREE_MAYBE (u->user); - FREE_MAYBE (u->passwd); - FREE_MAYBE (u->local); - FREE_MAYBE (u->referer); - if (u->proxy) - freeurl (u->proxy, 1); - if (complete) - free (u); - return; -} - -/* Extract the given URL of the form - (http:|ftp:)// (user (:password)?@)?hostname (:port)? (/path)? - 1. hostname (terminated with `/' or `:') - 2. port number (terminated with `/'), or chosen for the protocol - 3. dirname (everything after hostname) - Most errors are handled. No allocation is done, you must supply - pointers to allocated memory. - ...and a host of other stuff :-) - - - Recognizes hostname:dir/file for FTP and - hostname (:portnum)?/dir/file for HTTP. - - Parses the path to yield directory and file - - Parses the URL to yield the username and passwd (if present) - - Decodes the strings, in case they contain "forbidden" characters - - Writes the result to struct urlinfo - - If the argument STRICT is set, it recognizes only the canonical - form. */ -uerr_t -parseurl (const char *url, struct urlinfo *u, int strict) +static int +parse_uname (const char *str, int len, char **user, char **passwd) { - int i, l, abs_ftp; - int recognizable; /* Recognizable URL is the one where - the protocol name was explicitly - named, i.e. it wasn't deduced from - the URL format. */ - uerr_t type; - - DEBUGP (("parseurl (\"%s\") -> ", url)); - url += skip_url (url); - recognizable = has_proto (url); - if (strict && !recognizable) - return URLUNKNOWN; - for (i = 0, l = 0; i < ARRAY_SIZE (sup_protos); i++) + char *colon; + + if (len == 0) + /* Empty user name not allowed. */ + return 0; + + colon = memchr (str, ':', len); + if (colon == str) + /* Empty user name again. */ + return 0; + + if (colon) { - l = strlen (sup_protos[i].name); - if (!strncasecmp (sup_protos[i].name, url, l)) - break; + int pwlen = len - (colon + 1 - str); + *passwd = xmalloc (pwlen + 1); + memcpy (*passwd, colon + 1, pwlen); + (*passwd)[pwlen] = '\0'; + len -= pwlen + 1; } - /* If protocol is recognizable, but unsupported, bail out, else - suppose unknown. */ - if (recognizable && i == ARRAY_SIZE (sup_protos)) - return URLUNKNOWN; - else if (i == ARRAY_SIZE (sup_protos)) - type = URLUNKNOWN; else - u->proto = type = sup_protos[i].ind; - - if (type == URLUNKNOWN) - l = 0; - /* Allow a username and password to be specified (i.e. just skip - them for now). */ - if (recognizable) - l += skip_uname (url + l); - for (i = l; url[i] && url[i] != ':' && url[i] != '/'; i++); - if (i == l) - return URLBADHOST; - /* Get the hostname. */ - u->host = strdupdelim (url + l, url + i); - DEBUGP (("host %s -> ", u->host)); - - /* Assume no port has been given. */ - u->port = 0; - if (url[i] == ':') - { - /* We have a colon delimiting the hostname. It could mean that - a port number is following it, or a directory. */ - if (ISDIGIT (url[++i])) /* A port number */ - { - if (type == URLUNKNOWN) - u->proto = type = URLHTTP; - for (; url[i] && url[i] != '/'; i++) - if (ISDIGIT (url[i])) - u->port = 10 * u->port + (url[i] - '0'); - else - return URLBADPORT; - if (!u->port) - return URLBADPORT; - DEBUGP (("port %hu -> ", u->port)); - } - else if (type == URLUNKNOWN) /* or a directory */ - u->proto = type = URLFTP; - else /* or just a misformed port number */ - return URLBADPORT; - } - else if (type == URLUNKNOWN) - u->proto = type = URLHTTP; - if (!u->port) + *passwd = NULL; + + *user = xmalloc (len + 1); + memcpy (*user, str, len); + (*user)[len] = '\0'; + + return 1; +} + +/* Used by main.c: detect URLs written using the "shorthand" URL forms + 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] + + FTP shorthands look like this: + + foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file + foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file + + If the URL needs not or cannot be rewritten, return NULL. */ +char * +rewrite_shorthand_url (const char *url) +{ + const char *p; + + if (url_has_scheme (url)) + return NULL; + + /* Look for a ':' or '/'. The former signifies NcFTP syntax, the + latter Netscape. */ + for (p = url; *p && *p != ':' && *p != '/'; p++) + ; + + if (p == url) + return NULL; + + if (*p == ':') { - int i; - for (i = 0; i < ARRAY_SIZE (sup_protos); i++) - if (sup_protos[i].ind == type) - break; - if (i == ARRAY_SIZE (sup_protos)) - return URLUNKNOWN; - u->port = sup_protos[i].port; + const char *pp, *path; + 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... */ + path = p + 1; + res = xmalloc (6 + strlen (url) + 1); + sprintf (res, "ftp://%s", url); + /* ...and replace ':' with '/'. */ + res[6 + (p - url)] = '/'; + return res; } - /* Some delimiter troubles... */ - if (url[i] == '/' && url[i - 1] != ':') - ++i; - if (type == URLHTTP) - while (url[i] && url[i] == '/') - ++i; - u->path = (char *)xmalloc (strlen (url + i) + 8); - strcpy (u->path, url + i); - if (type == URLFTP) + else { - u->ftp_type = process_ftp_type (u->path); - /* #### We don't handle type `d' correctly yet. */ - if (!u->ftp_type || TOUPPER (u->ftp_type) == 'D') - u->ftp_type = 'I'; + char *res; + http: + /* Just prepend "http://" to what we have. */ + res = xmalloc (7 + strlen (url) + 1); + sprintf (res, "http://%s", url); + return res; } - DEBUGP (("opath %s -> ", u->path)); - /* Parse the username and password (if existing). */ - parse_uname (url, &u->user, &u->passwd); - /* Decode the strings, as per RFC 1738. */ - decode_string (u->host); - decode_string (u->path); - if (u->user) - decode_string (u->user); - if (u->passwd) - decode_string (u->passwd); - /* Parse the directory. */ - parse_dir (u->path, &u->dir, &u->file); - DEBUGP (("dir %s -> file %s -> ", u->dir, u->file)); - /* Simplify the directory. */ - path_simplify (u->dir); - /* Remove the leading `/' in HTTP. */ - if (type == URLHTTP && *u->dir == '/') - strcpy (u->dir, u->dir + 1); - DEBUGP (("ndir %s\n", u->dir)); - /* Strip trailing `/'. */ - l = strlen (u->dir); - if (l && u->dir[l - 1] == '/') - u->dir[l - 1] = '\0'; - /* Re-create the path: */ - abs_ftp = (u->proto == URLFTP && *u->dir == '/'); - /* sprintf (u->path, "%s%s%s%s", abs_ftp ? "%2F": "/", - abs_ftp ? (u->dir + 1) : u->dir, *u->dir ? "/" : "", u->file); */ - strcpy (u->path, abs_ftp ? "%2F" : "/"); - strcat (u->path, abs_ftp ? (u->dir + 1) : u->dir); - strcat (u->path, *u->dir ? "/" : ""); - strcat (u->path, u->file); - URL_CLEANSE (u->path); - DEBUGP (("newpath: %s\n", u->path)); - /* Create the clean URL. */ - u->url = str_url (u, 0); - return URLOK; } -/* Special versions of DOTP and DDOTP for parse_dir(). */ +static void parse_path PARAMS ((const char *, char **, char **)); -#define PD_DOTP(x) ((*(x) == '.') && (!*((x) + 1) || *((x) + 1) == '?')) -#define PD_DDOTP(x) ((*(x) == '.') && (*(x) == '.') \ - && (!*((x) + 2) || *((x) + 2) == '?')) +static char * +strpbrk_or_eos (const char *s, const char *accept) +{ + char *p = strpbrk (s, accept); + if (!p) + p = (char *)s + strlen (s); + return p; +} -/* Build the directory and filename components of the path. Both - components are *separately* malloc-ed strings! It does not change - the contents of path. +static char *parse_errors[] = { +#define PE_NO_ERROR 0 + "No error", +#define PE_UNRECOGNIZED_SCHEME 1 + "Unrecognized 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" +}; - If the path ends with "." or "..", they are (correctly) counted as - directories. */ -static void -parse_dir (const char *path, char **dir, char **file) +#define SETERR(p, v) do { \ + if (p) \ + *(p) = (v); \ +} while (0) + +/* Parse a URL. + + Return a new struct url if successful, NULL on error. In case of + error, and if ERROR is not NULL, also set *ERROR to the appropriate + error code. */ +struct url * +url_parse (const char *url, int *error) { - int i, l; + struct url *u; + const char *p; + + enum url_scheme scheme; + + const char *uname_b, *uname_e; + const char *host_b, *host_e; + const char *path_b, *path_e; + const char *params_b, *params_e; + const char *query_b, *query_e; + const char *fragment_b, *fragment_e; - l = urlpath_length (path); - for (i = l; i && path[i] != '/'; i--); + int port; + char *user = NULL, *passwd = NULL; - if (!i && *path != '/') /* Just filename */ + const char *url_orig = url; + + p = url = reencode_string (url); + + scheme = url_scheme (url); + if (scheme == SCHEME_INVALID) { - if (PD_DOTP (path) || PD_DDOTP (path)) - { - *dir = strdupdelim (path, path + l); - *file = xstrdup (path + l); /* normally empty, but could - contain ?... */ - } - else - { - *dir = xstrdup (""); /* This is required because of FTP */ - *file = xstrdup (path); - } + SETERR (error, PE_UNRECOGNIZED_SCHEME); + return NULL; } - else if (!i) /* /filename */ + + p += strlen (supported_schemes[scheme].leading_string); + uname_b = p; + p += url_skip_uname (p); + uname_e = p; + + /* scheme://user:pass@host[:port]... */ + /* ^ */ + + /* We attempt to break down the URL into the components path, + params, query, and fragment. They are ordered like this: + + scheme://host[:port][/path][;params][?query][#fragment] */ + + params_b = params_e = NULL; + query_b = query_e = NULL; + fragment_b = fragment_e = NULL; + + host_b = p; + p = strpbrk_or_eos (p, ":/;?#"); + host_e = p; + + if (host_b == host_e) { - if (PD_DOTP (path + 1) || PD_DDOTP (path + 1)) - { - *dir = strdupdelim (path, path + l); - *file = xstrdup (path + l); /* normally empty, but could - contain ?... */ - } - else - { - *dir = xstrdup ("/"); - *file = xstrdup (path + 1); - } + SETERR (error, PE_EMPTY_HOST); + return NULL; } - else /* Nonempty directory with or without a filename */ + + port = scheme_default_port (scheme); + if (*p == ':') { - if (PD_DOTP (path + i + 1) || PD_DDOTP (path + i + 1)) - { - *dir = strdupdelim (path, path + l); - *file = xstrdup (path + l); /* normally empty, but could - contain ?... */ - } - else + const char *port_b, *port_e, *pp; + + /* scheme://host:port/tralala */ + /* ^ */ + ++p; + port_b = p; + p = strpbrk_or_eos (p, "/;?#"); + port_e = p; + + if (port_b == port_e) { - *dir = strdupdelim (path, path + i); - *file = xstrdup (path + i + 1); + /* http://host:/whatever */ + /* ^ */ + SETERR (error, PE_BAD_PORT_NUMBER); + return NULL; } - } -} -/* Find the optional username and password within the URL, as per - RFC1738. The returned user and passwd char pointers are - malloc-ed. */ -static uerr_t -parse_uname (const char *url, char **user, char **passwd) -{ - int l; - const char *p, *col; - char **where; - - *user = NULL; - *passwd = NULL; - url += skip_url (url); - /* Look for end of protocol string. */ - l = skip_proto (url); - if (!l) - return URLUNKNOWN; - /* Add protocol offset. */ - url += l; - /* Is there an `@' character? */ - for (p = url; *p && *p != '/'; p++) - if (*p == '@') - break; - /* If not, return. */ - if (*p != '@') - return URLOK; - /* Else find the username and password. */ - for (p = col = url; *p != '@'; p++) - { - if (*p == ':' && !*user) + for (port = 0, pp = port_b; pp < port_e; pp++) { - *user = (char *)xmalloc (p - url + 1); - memcpy (*user, url, p - url); - (*user)[p - url] = '\0'; - col = p + 1; + if (!ISDIGIT (*pp)) + { + /* http://host:12randomgarbage/blah */ + /* ^ */ + SETERR (error, PE_BAD_PORT_NUMBER); + return NULL; + } + port = 10 * port + (*pp - '0'); } } - /* Decide whether you have only the username or both. */ - where = *user ? passwd : user; - *where = (char *)xmalloc (p - col + 1); - memcpy (*where, col, p - col); - (*where)[p - col] = '\0'; - return URLOK; -} - -/* If PATH ends with `;type=X', return the character X. */ -static char -process_ftp_type (char *path) -{ - int len = strlen (path); - if (len >= 7 - && !memcmp (path + len - 7, ";type=", 6)) + if (*p == '/') { - path[len - 7] = '\0'; - return path[len - 1]; + ++p; + path_b = p; + p = strpbrk_or_eos (p, ";?#"); + path_e = p; } else - return '\0'; -} - -/* Return the URL as fine-formed string, with a proper protocol, - optional port number, directory and optional user/password. If - HIDE is non-zero, password will be hidden. The forbidden - characters in the URL will be cleansed. */ -char * -str_url (const struct urlinfo *u, int hide) -{ - char *res, *host, *user, *passwd, *proto_name, *dir, *file; - int i, l, ln, lu, lh, lp, lf, ld; - unsigned short proto_default_port; + { + /* Path is not allowed not to exist. */ + path_b = path_e = p; + } - /* Look for the protocol name. */ - for (i = 0; i < ARRAY_SIZE (sup_protos); i++) - if (sup_protos[i].ind == u->proto) - break; - if (i == ARRAY_SIZE (sup_protos)) - return NULL; - proto_name = sup_protos[i].name; - proto_default_port = sup_protos[i].port; - host = CLEANDUP (u->host); - dir = CLEANDUP (u->dir); - file = CLEANDUP (u->file); - user = passwd = NULL; - if (u->user) - user = CLEANDUP (u->user); - if (u->passwd) + if (*p == ';') { - int i; - passwd = CLEANDUP (u->passwd); - if (hide) - for (i = 0; passwd[i]; i++) - passwd[i] = 'x'; + ++p; + params_b = p; + p = strpbrk_or_eos (p, "?#"); + params_e = p; } - if (u->proto == URLFTP && *dir == '/') + if (*p == '?') { - char *tmp = (char *)xmalloc (strlen (dir) + 3); - /*sprintf (tmp, "%%2F%s", dir + 1);*/ - tmp[0] = '%'; - tmp[1] = '2'; - tmp[2] = 'F'; - strcpy (tmp + 3, dir + 1); - free (dir); - dir = tmp; + ++p; + query_b = p; + p = strpbrk_or_eos (p, "#"); + query_e = p; } + if (*p == '#') + { + ++p; + fragment_b = p; + p += strlen (p); + fragment_e = p; + } + assert (*p == 0); - ln = strlen (proto_name); - lu = user ? strlen (user) : 0; - lp = passwd ? strlen (passwd) : 0; - lh = strlen (host); - ld = strlen (dir); - lf = strlen (file); - res = (char *)xmalloc (ln + lu + lp + lh + ld + lf + 20); /* safe sex */ - /* sprintf (res, "%s%s%s%s%s%s:%d/%s%s%s", proto_name, - (user ? user : ""), (passwd ? ":" : ""), - (passwd ? passwd : ""), (user ? "@" : ""), - host, u->port, dir, *dir ? "/" : "", file); */ - l = 0; - memcpy (res, proto_name, ln); - l += ln; - if (user) + if (uname_b != uname_e) { - memcpy (res + l, user, lu); - l += lu; - if (passwd) + /* http://user:pass@host */ + /* ^ ^ */ + /* uname_b uname_e */ + if (!parse_uname (uname_b, uname_e - uname_b - 1, &user, &passwd)) { - res[l++] = ':'; - memcpy (res + l, passwd, lp); - l += lp; + SETERR (error, PE_INVALID_USER_NAME); + return NULL; } - res[l++] = '@'; } - memcpy (res + l, host, lh); - l += lh; - if (u->port != proto_default_port) + + u = (struct url *)xmalloc (sizeof (struct url)); + memset (u, 0, sizeof (*u)); + + if (url == url_orig) + u->url = xstrdup (url); + else + u->url = (char *)url; + + u->scheme = scheme; + u->host = strdupdelim (host_b, host_e); + u->port = port; + u->user = user; + u->passwd = passwd; + + u->path = strdupdelim (path_b, path_e); + path_simplify (u->path); + + if (params_b) + u->params = strdupdelim (params_b, params_e); + if (query_b) + u->query = strdupdelim (query_b, query_e); + if (fragment_b) + u->fragment = strdupdelim (fragment_b, fragment_e); + + parse_path (u->path, &u->dir, &u->file); + + return u; +} + +const char * +url_error (int error_code) +{ + assert (error_code >= 0 && error_code < ARRAY_SIZE (parse_errors)); + return parse_errors[error_code]; +} + +static void +parse_path (const char *quoted_path, char **dir, char **file) +{ + char *path, *last_slash; + + STRDUP_ALLOCA (path, quoted_path); + decode_string (path); + + last_slash = strrchr (path, '/'); + if (!last_slash) { - res[l++] = ':'; - long_to_string (res + l, (long)u->port); - l += numdigit (u->port); + *dir = xstrdup (""); + *file = xstrdup (path); + } + else + { + *dir = strdupdelim (path, last_slash); + *file = xstrdup (last_slash + 1); } - res[l++] = '/'; - memcpy (res + l, dir, ld); - l += ld; - if (*dir) - res[l++] = '/'; - strcpy (res + l, file); - free (host); - free (dir); - free (file); - FREE_MAYBE (user); - FREE_MAYBE (passwd); - return res; } -/* Check whether two URL-s are equivalent, i.e. pointing to the same - location. Uses parseurl to parse them, and compares the canonical - forms. +/* Note: URL's "full path" is the path with the query string and + params appended. The "fragment" (#foo) is intentionally ignored, + but that might be changed. For example, if the original URL was + "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment", + the full path will be "/foo/bar/baz;bullshit?querystring". */ - Returns 1 if the URL1 is equivalent to URL2, 0 otherwise. Also - return 0 on error. */ -int -url_equal (const char *url1, const char *url2) +/* Return the length of the full path, without the terminating + zero. */ + +static int +full_path_length (const struct url *url) { - struct urlinfo *u1, *u2; - uerr_t err; - int res; + int len = 0; + +#define FROB(el) if (url->el) len += 1 + strlen (url->el) - u1 = newurl (); - err = parseurl (url1, u1, 0); - if (err != URLOK) + FROB (path); + FROB (params); + FROB (query); + +#undef FROB + + return len; +} + +/* Write out the full path. */ + +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; \ + } \ +} while (0) + + FROB (path, '/'); + FROB (params, ';'); + FROB (query, '?'); + +#undef FROB +} + +/* Public function for getting the "full path". */ +char * +url_full_path (const struct url *url) +{ + int length = full_path_length (url); + char *full_path = (char *)xmalloc(length + 1); + + full_path_write (url, full_path); + full_path[length] = '\0'; + + return full_path; +} + +/* Sync u->path and u->url with u->dir and u->file. */ +static void +sync_path (struct url *url) +{ + char *newpath; + + xfree (url->path); + + if (!*url->dir) { - freeurl (u1, 1); - return 0; + newpath = xstrdup (url->file); + REENCODE (newpath); } - u2 = newurl (); - err = parseurl (url2, u2, 0); - if (err != URLOK) + else { - freeurl (u2, 1); - return 0; + int dirlen = strlen (url->dir); + int filelen = strlen (url->file); + + newpath = xmalloc (dirlen + 1 + filelen + 1); + memcpy (newpath, url->dir, dirlen); + newpath[dirlen] = '/'; + memcpy (newpath + dirlen + 1, url->file, filelen); + newpath[dirlen + 1 + filelen] = '\0'; + REENCODE (newpath); } - res = !strcmp (u1->url, u2->url); - freeurl (u1, 1); - freeurl (u2, 1); - return res; + + url->path = newpath; + + /* Synchronize u->url. */ + xfree (url->url); + url->url = url_string (url, 0); +} + +/* Mutators. Code in ftp.c insists on changing u->dir and u->file. + This way we can sync u->path and u->url when they get changed. */ + +void +url_set_dir (struct url *url, const char *newdir) +{ + xfree (url->dir); + url->dir = xstrdup (newdir); + sync_path (url); +} + +void +url_set_file (struct url *url, const char *newfile) +{ + xfree (url->file); + url->file = xstrdup (newfile); + sync_path (url); +} + +void +url_free (struct url *url) +{ + xfree (url->host); + 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); + FREE_MAYBE (url->dir); + FREE_MAYBE (url->file); + + xfree (url); } urlpos * @@ -839,9 +990,9 @@ free_urlpos (urlpos *l) while (l) { urlpos *next = l->next; - free (l->url); + xfree (l->url); FREE_MAYBE (l->local_name); - free (l); + xfree (l); l = next; } } @@ -894,7 +1045,7 @@ mkalldirs (const char *path) { if (S_ISDIR (st.st_mode)) { - free (t); + xfree (t); return 0; } else @@ -918,7 +1069,7 @@ mkalldirs (const char *path) res = make_directory (t); if (res != 0) logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno)); - free (t); + xfree (t); return res; } @@ -935,14 +1086,11 @@ count_slashes (const char *s) /* Return the path name of the URL-equivalent file name, with a remote-like structure of directories. */ static char * -mkstruct (const struct urlinfo *u) +mkstruct (const struct url *u) { char *host, *dir, *file, *res, *dirpref; int l; - assert (u->dir != NULL); - assert (u->host != NULL); - if (opt.cut_dirs) { char *ptr = u->dir + (*u->dir == '/'); @@ -962,7 +1110,7 @@ mkstruct (const struct urlinfo *u) if (opt.add_hostdir && !opt.simple_check) { char *nhost = realhost (host); - free (host); + xfree (host); host = nhost; } /* Add dir_prefix and hostname (if required) to the beginning of @@ -985,7 +1133,7 @@ mkstruct (const struct urlinfo *u) else dirpref = ""; } - free (host); + xfree (host); /* If there is a prefix, prepend it. */ if (*dirpref) @@ -994,8 +1142,7 @@ mkstruct (const struct urlinfo *u) sprintf (newdir, "%s%s%s", dirpref, *dir == '/' ? "" : "/", dir); dir = newdir; } - dir = xstrdup (dir); - URL_CLEANSE (dir); + dir = encode_string (dir); l = strlen (dir); if (l && dir[l - 1] == '/') dir[l - 1] = '\0'; @@ -1008,14 +1155,71 @@ mkstruct (const struct urlinfo *u) /* Finally, construct the full name. */ res = (char *)xmalloc (strlen (dir) + 1 + strlen (file) + 1); sprintf (res, "%s%s%s", dir, *dir ? "/" : "", file); - free (dir); + xfree (dir); return res; } +/* Compose a file name out of BASE, an unescaped file name, and QUERY, + an escaped query string. The trick is to make sure that unsafe + characters in BASE are escaped, and that slashes in QUERY are also + escaped. */ + +static char * +compose_file_name (char *base, char *query) +{ + char result[256]; + char *from; + char *to = result; + + /* Copy BASE to RESULT and encode all unsafe characters. */ + from = base; + while (*from && to - result < sizeof (result)) + { + if (UNSAFE_CHAR (*from)) + { + const unsigned char c = *from++; + *to++ = '%'; + *to++ = XDIGIT_TO_XCHAR (c >> 4); + *to++ = XDIGIT_TO_XCHAR (c & 0xf); + } + else + *to++ = *from++; + } + + if (query && to - result < sizeof (result)) + { + *to++ = '?'; + + /* Copy QUERY to RESULT and encode all '/' characters. */ + from = query; + while (*from && to - result < sizeof (result)) + { + if (*from == '/') + { + *to++ = '%'; + *to++ = '2'; + *to++ = 'F'; + ++from; + } + else + *to++ = *from++; + } + } + + if (to - result < sizeof (result)) + *to = '\0'; + else + /* Truncate input which is too long, presumably due to a huge + query string. */ + result[sizeof (result) - 1] = '\0'; + + return xstrdup (result); +} + /* Create a unique filename, corresponding to a given URL. Calls mkstruct if necessary. Does *not* actually create any directories. */ char * -url_filename (const struct urlinfo *u) +url_filename (const struct url *u) { char *file, *name; int have_prefix = 0; /* whether we must prepend opt.dir_prefix */ @@ -1027,10 +1231,9 @@ url_filename (const struct urlinfo *u) } else { - if (!*u->file) - file = xstrdup ("index.html"); - else - file = xstrdup (u->file); + char *base = *u->file ? u->file : "index.html"; + char *query = u->query && *u->query ? u->query : NULL; + file = compose_file_name (base, query); } if (!have_prefix) @@ -1042,7 +1245,7 @@ url_filename (const struct urlinfo *u) char *nfile = (char *)xmalloc (strlen (opt.dir_prefix) + 1 + strlen (file) + 1); sprintf (nfile, "%s/%s", opt.dir_prefix, file); - free (file); + xfree (file); file = nfile; } } @@ -1071,7 +1274,7 @@ url_filename (const struct urlinfo *u) /* Find a unique name. */ name = unique_name (file); - free (file); + xfree (file); return name; } @@ -1098,37 +1301,46 @@ find_last_char (const char *b, const char *e, char c) return NULL; } -/* Construct a URL by concatenating an absolute URL and a path, which - may or may not be absolute. This tries to behave "reasonably" in - all foreseeable cases. It employs little specific knowledge about - protocols or URL-specific stuff -- it just works on strings. */ +/* Resolve the result of "linking" a base URI (BASE) to a + link-specified URI (LINK). + + Either of the URIs may be absolute or relative, complete with the + host name, or path only. This tries to behave "reasonably" in all + foreseeable cases. It employs little specific knowledge about + schemes or URL-specific stuff -- it just works on strings. + + The parameters LINKLENGTH is useful if LINK is not zero-terminated. + See uri_merge for a gentler interface to this functionality. + + #### This function should handle `./' and `../' so that the evil + path_simplify can go. */ static char * -construct (const char *url, const char *sub, int subsize, int no_proto) +uri_merge_1 (const char *base, const char *link, int linklength, int no_scheme) { char *constr; - if (no_proto) + if (no_scheme) { - const char *end = url + urlpath_length (url); + const char *end = base + urlpath_length (base); - if (*sub != '/') + if (*link != '/') { - /* SUB is a relative URL: we need to replace everything - after last slash (possibly empty) with SUB. + /* LINK is a relative URL: we need to replace everything + after last slash (possibly empty) with LINK. - So, if URL is "whatever/foo/bar", and SUB is "qux/xyzzy", + 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; int span; const char *start_insert; - const char *last_slash = find_last_char (url, end, '/'); /* the last slash. */ + const char *last_slash = find_last_char (base, end, '/'); if (!last_slash) { - /* No slash found at all. Append SUB to what we have, + /* No slash found at all. Append LINK to what we have, but we'll need a slash as a separator. - Example: if url == "foo" and sub == "qux/xyzzy", then - we cannot just append sub to url, because we'd get + 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". @@ -1143,7 +1355,7 @@ construct (const char *url, const char *sub, int subsize, int no_proto) start_insert = end + 1; need_explicit_slash = 1; } - else if (last_slash && last_slash != url && *(last_slash - 1) == '/') + else if (last_slash && last_slash != base && *(last_slash - 1) == '/') { /* example: http://host" */ /* ^ */ @@ -1157,28 +1369,28 @@ construct (const char *url, const char *sub, int subsize, int no_proto) start_insert = last_slash + 1; } - span = start_insert - url; - constr = (char *)xmalloc (span + subsize + 1); + span = start_insert - base; + constr = (char *)xmalloc (span + linklength + 1); if (span) - memcpy (constr, url, span); + memcpy (constr, base, span); if (need_explicit_slash) constr[span - 1] = '/'; - if (subsize) - memcpy (constr + span, sub, subsize); - constr[span + subsize] = '\0'; + if (linklength) + memcpy (constr + span, link, linklength); + constr[span + linklength] = '\0'; } - else /* *sub == `/' */ + else /* *link == `/' */ { - /* SUB is an absolute path: we need to replace everything - after (and including) the FIRST slash with SUB. + /* LINK is an absolute path: we need to replace everything + after (and including) the FIRST slash with LINK. - So, if URL is "http://host/whatever/foo/bar", and SUB is + 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 = url; + const char *pos = base; int seen_slash_slash = 0; /* We're looking for the first slash, but want to ignore double slash. */ @@ -1194,14 +1406,14 @@ construct (const char *url, const char *sub, int subsize, int no_proto) /* 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 SUB will be inserted. When - examining the last two examples, keep in mind that SUB + 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 = url; + start_insert = base; else if (!slash && seen_slash_slash) /* example: "http://foo" */ /* ^ */ @@ -1209,86 +1421,169 @@ construct (const char *url, const char *sub, int subsize, int no_proto) else if (slash && !seen_slash_slash) /* example: "foo/bar" */ /* ^ */ - start_insert = url; + start_insert = base; else if (slash && seen_slash_slash) /* example: "http://something/" */ /* ^ */ start_insert = slash; - span = start_insert - url; - constr = (char *)xmalloc (span + subsize + 1); + span = start_insert - base; + constr = (char *)xmalloc (span + linklength + 1); if (span) - memcpy (constr, url, span); - if (subsize) - memcpy (constr + span, sub, subsize); - constr[span + subsize] = '\0'; + memcpy (constr, base, span); + if (linklength) + memcpy (constr + span, link, linklength); + constr[span + linklength] = '\0'; } } - else /* !no_proto */ + else /* !no_scheme */ { - constr = strdupdelim (sub, sub + subsize); + constr = strdupdelim (link, link + linklength); } return constr; } -/* Like the function above, but with a saner caller interface. */ +/* Merge BASE with LINK and return the resulting URI. This is an + interface to uri_merge_1 that assumes that LINK is a + zero-terminated string. */ char * -url_concat (const char *base_url, const char *new_url) +uri_merge (const char *base, const char *link) { - return construct (base_url, new_url, strlen (new_url), !has_proto (new_url)); + return uri_merge_1 (base, link, strlen (link), !url_has_scheme (link)); } -/* Optimize URL by host, destructively replacing u->host with realhost - (u->host). Do this regardless of opt.simple_check. */ -void -opt_url (struct urlinfo *u) -{ - /* Find the "true" host. */ - char *host = realhost (u->host); - free (u->host); - u->host = host; - assert (u->dir != NULL); /* the URL must have been parsed */ - /* Refresh the printed representation. */ - free (u->url); - u->url = str_url (u, 0); -} +#define APPEND(p, s) do { \ + int len = strlen (s); \ + memcpy (p, s, len); \ + p += len; \ +} while (0) -/* This beautiful kludge is fortunately not needed, as I've made - parse_dir do the (almost) right thing, so that a query can never - become a part of directory. */ -#if 0 -/* Call path_simplify, but make sure that the part after the - question-mark, if any, is not destroyed by path_simplify's - "optimizations". */ -void -path_simplify_with_kludge (char *path) +/* Use this instead of password when the actual password is supposed + to be hidden. We intentionally use a generic string without giving + away the number of characters in the password, like previous + versions did. */ +#define HIDDEN_PASSWORD "*password*" + +/* 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. */ + +char * +url_string (const struct url *url, int hide_password) { - char *query = strchr (path, '?'); - if (query) - /* path_simplify also works destructively, so we also have the - license to write. */ - *query = '\0'; - path_simplify (path); - if (query) + int size; + char *result, *p; + char *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 fplen = full_path_length (url); + + assert (scheme_str != NULL); + + /* Make sure the user name and password are quoted. */ + if (url->user) + { + quoted_user = encode_string_maybe (url->user); + if (url->passwd) + { + if (hide_password) + quoted_passwd = HIDDEN_PASSWORD; + else + quoted_passwd = encode_string_maybe (url->passwd); + } + } + + size = (strlen (scheme_str) + + strlen (url->host) + + 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); + } + + p = result = xmalloc (size); + + APPEND (p, scheme_str); + if (quoted_user) + { + APPEND (p, quoted_user); + if (quoted_passwd) + { + *p++ = ':'; + APPEND (p, quoted_passwd); + } + *p++ = '@'; + } + + APPEND (p, url->host); + if (url->port != scheme_port) { - char *newend = path + strlen (path); - *query = '?'; - if (newend != query) - memmove (newend, query, strlen (query) + 1); + *p++ = ':'; + long_to_string (p, url->port); + p += strlen (p); } + + full_path_write (url, p); + p += fplen; + *p++ = '\0'; + + assert (p - result == size); + + if (quoted_user && quoted_user != url->user) + xfree (quoted_user); + if (quoted_passwd && !hide_password + && quoted_passwd != url->passwd) + xfree (quoted_passwd); + + return result; } -#endif -/* Returns proxy host address, in accordance with PROTO. */ +/* Returns proxy host address, in accordance with SCHEME. */ char * -getproxy (uerr_t proto) +getproxy (enum url_scheme scheme) { - if (proto == URLHTTP) - return opt.http_proxy ? opt.http_proxy : getenv ("http_proxy"); - else if (proto == URLFTP) - return opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy"); - else + char *proxy = NULL; + char *rewritten_url; + static char rewritten_storage[1024]; + + switch (scheme) + { + case SCHEME_HTTP: + proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy"); + break; +#ifdef HAVE_SSL + case SCHEME_HTTPS: + proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy"); + break; +#endif + case SCHEME_FTP: + proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy"); + break; + case SCHEME_INVALID: + break; + } + if (!proxy || !*proxy) return NULL; + + /* Handle shorthands. */ + rewritten_url = rewrite_shorthand_url (proxy); + if (rewritten_url) + { + strncpy (rewritten_storage, rewritten_url, sizeof(rewritten_storage)); + rewritten_storage[sizeof (rewritten_storage) - 1] = '\0'; + proxy = rewritten_storage; + } + + return proxy; } /* Should a host be accessed through proxy, concerning no_proxy? */ @@ -1302,6 +1597,7 @@ no_proxy_match (const char *host, const char **no_proxy) } static void write_backup_file PARAMS ((const char *, downloaded_file_t)); +static void replace_attr PARAMS ((const char **, int, FILE *, const char *)); /* Change the links in an HTML document. Accepts a structure that defines the positions of all the links. */ @@ -1310,9 +1606,11 @@ convert_links (const char *file, urlpos *l) { struct file_memory *fm; FILE *fp; - char *p; + const char *p; downloaded_file_t downloaded_file_return; + logprintf (LOG_VERBOSE, _("Converting %s... "), file); + { /* First we do a "dry run": go through the list L and see whether any URL needs to be converted in the first place. If not, just @@ -1320,18 +1618,15 @@ convert_links (const char *file, urlpos *l) int count = 0; urlpos *dry = l; for (dry = l; dry; dry = dry->next) - if (dry->flags & (UABS2REL | UREL2ABS)) + if (dry->convert != CO_NOCONVERT) ++count; if (!count) { - logprintf (LOG_VERBOSE, _("Nothing to do while converting %s.\n"), - file); + logputs (LOG_VERBOSE, _("nothing to do.\n")); return; } } - logprintf (LOG_VERBOSE, _("Converting %s... "), file); - fm = read_file (file); if (!fm) { @@ -1370,16 +1665,16 @@ convert_links (const char *file, urlpos *l) for (; l; l = l->next) { char *url_start = fm->content + l->pos; + if (l->pos >= fm->length) { DEBUGP (("Something strange is going on. Please investigate.")); break; } /* If the URL is not to be converted, skip it. */ - if (!(l->flags & (UABS2REL | UREL2ABS))) + if (l->convert == CO_NOCONVERT) { - DEBUGP (("Skipping %s at position %d (flags %d).\n", l->url, - l->pos, l->flags)); + DEBUGP (("Skipping %s at position %d.\n", l->url, l->pos)); continue; } @@ -1387,30 +1682,26 @@ convert_links (const char *file, urlpos *l) quote, to the outfile. */ fwrite (p, 1, url_start - p, fp); p = url_start; - if (l->flags & UABS2REL) + if (l->convert == CO_CONVERT_TO_RELATIVE) { /* Convert absolute URL to relative. */ char *newname = construct_relative (file, l->local_name); - putc (*p, fp); /* quoting char */ - fputs (newname, fp); - p += l->size - 1; - putc (*p, fp); /* close quote */ - ++p; - DEBUGP (("ABS2REL: %s to %s at position %d in %s.\n", + char *quoted_newname = html_quote_string (newname); + replace_attr (&p, l->size, fp, quoted_newname); + DEBUGP (("TO_RELATIVE: %s to %s at position %d in %s.\n", l->url, newname, l->pos, file)); - free (newname); + xfree (newname); + xfree (quoted_newname); } - else if (l->flags & UREL2ABS) + else if (l->convert == CO_CONVERT_TO_COMPLETE) { /* Convert the link to absolute URL. */ char *newlink = l->url; - putc (*p, fp); /* quoting char */ - fputs (newlink, fp); - p += l->size - 1; - putc (*p, fp); /* close quote */ - ++p; - DEBUGP (("REL2ABS: to %s at position %d in %s.\n", + char *quoted_newlink = html_quote_string (newlink); + replace_attr (&p, l->size, fp, quoted_newlink); + DEBUGP (("TO_COMPLETE: to %s at position %d in %s.\n", newlink, l->pos, file)); + xfree (quoted_newlink); } } /* Output the rest of the file. */ @@ -1555,7 +1846,21 @@ write_backup_file (const char *file, downloaded_file_t downloaded_file_return) rather than making a copy of the string... Another note is that I thought I could just add a field to the urlpos structure saying that we'd written a .orig file for this URL, but that didn't work, - so I had to make this separate list. */ + so I had to make this separate list. + -- Dan Harkless + + This [adding a field to the urlpos structure] didn't work + because convert_file() is called twice: once after all its + sublinks have been retrieved in recursive_retrieve(), and + once at the end of the day in convert_all_links(). The + original linked list collected in recursive_retrieve() is + lost after the first invocation of convert_links(), and + convert_all_links() makes a new one (it calls get_urls_html() + for each file it covers.) That's why your first approach didn't + work. The way to make it work is perhaps to make this flag a + field in the `urls_html' list. + -- Hrvoje Niksic + */ converted_file_ptr = xmalloc(sizeof(*converted_file_ptr)); converted_file_ptr->string = xstrdup(file); /* die on out-of-mem. */ converted_file_ptr->next = converted_files; @@ -1563,6 +1868,87 @@ write_backup_file (const char *file, downloaded_file_t downloaded_file_return) } } +static int find_fragment PARAMS ((const char *, int, const char **, + const char **)); + +static void +replace_attr (const char **pp, int raw_size, FILE *fp, const char *new_str) +{ + const char *p = *pp; + int quote_flag = 0; + int size = raw_size; + char quote_char = '\"'; + const char *frag_beg, *frag_end; + + /* Structure of our string is: + "...old-contents..." + <--- l->size ---> (with quotes) + OR: + ...old-contents... + <--- l->size --> (no quotes) */ + + if (*p == '\"' || *p == '\'') + { + quote_char = *p; + quote_flag = 1; + ++p; + size -= 2; /* disregard opening and closing quote */ + } + putc (quote_char, fp); + fputs (new_str, fp); + + /* Look for fragment identifier, if any. */ + if (find_fragment (p, size, &frag_beg, &frag_end)) + fwrite (frag_beg, 1, frag_end - frag_beg, fp); + p += size; + if (quote_flag) + ++p; + putc (quote_char, fp); + *pp = p; +} + +/* Find the first occurrence of '#' in [BEG, BEG+SIZE) that is not + preceded by '&'. If the character is not found, return zero. If + the character is found, return 1 and set BP and EP to point to the + beginning and end of the region. + + This is used for finding the fragment indentifiers in URLs. */ + +static int +find_fragment (const char *beg, int size, const char **bp, const char **ep) +{ + const char *end = beg + size; + int saw_amp = 0; + for (; beg < end; beg++) + { + switch (*beg) + { + case '&': + saw_amp = 1; + break; + case '#': + if (!saw_amp) + { + *bp = beg; + *ep = end; + return 1; + } + /* fallthrough */ + default: + saw_amp = 0; + } + } + return 0; +} + +typedef struct _downloaded_file_list { + char* file; + downloaded_file_t download_type; + struct _downloaded_file_list* next; +} downloaded_file_list; + +static downloaded_file_list *downloaded_files; + /* Remembers which files have been downloaded. In the standard case, should be called with mode == FILE_DOWNLOADED_NORMALLY for each file we actually download successfully (i.e. not for ones we have failures on or that we skip @@ -1578,15 +1964,7 @@ write_backup_file (const char *file, downloaded_file_t downloaded_file_return) downloaded_file_t downloaded_file (downloaded_file_t mode, const char* file) { - typedef struct _downloaded_file_list - { - char* file; - downloaded_file_t download_type; - struct _downloaded_file_list* next; - } downloaded_file_list; - boolean found_file = FALSE; - static downloaded_file_list* downloaded_files = NULL; downloaded_file_list* rover = downloaded_files; while (rover != NULL) @@ -1614,10 +1992,16 @@ downloaded_file (downloaded_file_t mode, const char* file) return FILE_NOT_ALREADY_DOWNLOADED; } } - -/* Initialization of static stuff. */ + void -url_init (void) +downloaded_files_free (void) { - init_unsafe_char_table (); + downloaded_file_list* rover = downloaded_files; + while (rover) + { + downloaded_file_list *next = rover->next; + xfree (rover->file); + xfree (rover); + rover = next; + } }