/* URL handling.
- Copyright (C) 1995, 1996, 1997, 2000, 2001, 2003, 2003
- Free Software Foundation, Inc.
+ Copyright (C) 2005 Free Software Foundation, Inc.
This file is part of GNU Wget.
#include <stdio.h>
#include <stdlib.h>
-#ifdef HAVE_STRING_H
-# include <string.h>
-#else
-# include <strings.h>
-#endif
-#include <sys/types.h>
+#include <string.h>
#ifdef HAVE_UNISTD_H
# include <unistd.h>
#endif
#include "url.h"
#include "host.h" /* for is_valid_ipv6_address */
-#ifndef errno
-extern int errno;
-#endif
-
struct scheme_data
{
const char *name;
/* Forward declarations: */
-static int path_simplify PARAMS ((char *));
+static int path_simplify (char *);
\f
/* Support for escaping and unescaping of URL strings. */
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". Wget
- uses the rfc1738 set of reserved characters, plus "$" and ",", as
- recommended by rfc2396.
-
- An unsafe characters is the one that should be encoded when URLs
- are placed in foreign environments. E.g. space and newline are
- unsafe in HTTP contexts because HTTP uses them as separator and
+ changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
+ unsafe characters are loosely based on rfc1738, plus "$" 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.
#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, 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,
}
else
{
+ char c;
/* 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]);
+ 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;
}
}
return url_escape_1 (s, urlchr_unsafe, 1);
}
\f
-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.)
+
+ Return 1 if the char should be escaped as %XX, 0 otherwise. */
-/* 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)
+static inline int
+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;
- }
+ return 0;
else
/* Garbled %.. sequence: encode `%'. */
- return CM_ENCODE;
+ return 1;
}
else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
- return CM_ENCODE;
+ return 1;
else
- return CM_PASSTHROUGH;
+ return 0;
}
/* Translate a %-escaped (but possibly non-conformant) input string S
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.
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:
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;
{
const char *p;
- if (url_has_scheme (url))
+ if (url_scheme (url) != SCHEME_INVALID)
return NULL;
/* Look for a ':' or '/'. The former signifies NcFTP syntax, the
if (p == url)
return NULL;
+ /* 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[0] == ':' && p[1] == '/' && p[2] == '/')
+ return NULL;
+
if (*p == ':')
{
const char *pp;
}
}
\f
-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
help because the check for literal accept is in the
preprocessor.) */
-#ifdef __GNUC__
+#if defined(__GNUC__) && __GNUC__ >= 3
#define strpbrk_or_eos(s, accept) ({ \
char *SOE_p = strpbrk (s, accept); \
if (!SOE_p) \
- SOE_p = (char *)s + strlen (s); \
+ SOE_p = strchr (s, '\0'); \
SOE_p; \
})
-#else /* not __GNUC__ */
+#else /* not __GNUC__ or old gcc */
-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
+#endif /* not __GNUC__ or old gcc */
/* Turn STR into lowercase; return non-zero if a character was
actually changed. */
host_modified = lowercase_str (u->host);
/* Decode %HH sequences in host name. This is important not so much
- to support %HH sequences, but to support binary characters (which
- will have been converted to %HH by reencode_escapes). */
+ 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);
else
u->url = url_encoded;
}
- url_encoded = NULL;
return u;
url_full_path (const struct url *url)
{
int length = full_path_length (url);
- char *full_path = (char *)xmalloc(length + 1);
+ char *full_path = (char *) xmalloc (length + 1);
full_path_write (url, full_path);
full_path[length] = '\0';
return full_path;
}
-/* Escape unsafe and reserved characters, except for the slash
- characters. */
+/* Unescape CHR in an otherwise escaped STR. Used to selectively
+ escaping of certain characters, such as "/" and ":". Returns a
+ count of unescaped chars. */
-static char *
-url_escape_dir (const char *dir)
+static void
+unescape_single_char (char *str, char chr)
{
- 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 */
-
+ 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++)
{
- /* url_escape_1 having converted '/' to "%2F" exactly. */
- if (*h == '%' && h[1] == '2' && h[2] == 'F')
+ if (h[0] == '%' && h[1] == c1 && h[2] == c2)
{
- *t = '/';
+ *t = chr;
h += 2;
}
else
*t = *h;
}
*t = '\0';
+}
+
+/* Escape unsafe and reserved characters, except for the slash
+ characters. */
+static char *
+url_escape_dir (const char *dir)
+{
+ char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
+ if (newdir == dir)
+ return (char *)dir;
+
+ unescape_single_char (newdir, '/');
return newdir;
}
*p++ = '/';
memcpy (p, efile, filelen);
p += filelen;
- *p++ = '\0';
+ *p = '\0';
}
u->path = newpath;
}
\f
/* Create all the necessary directories for PATH (a file). Calls
- mkdirhier() internally. */
+ make_directory internally. */
int
mkalldirs (const char *path)
{
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 */
"back up one element". Single leading and trailing slashes are
preserved.
- This function does not handle URL escapes explicitly. If you're
- passing paths from URLs, make sure to unquote "%2e" and "%2E" to
- ".", so that this function can find the dots. (Wget's URL parser
- calls reencode_escapes, which see.)
-
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
{
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;
const char *scheme_str = supported_schemes[url->scheme].leading_string;
}
}
- /* Numeric IPv6 addresses can contain ':' and need to be quoted with
- brackets. */
- brackets_around_host = strchr (url->host, ':') != NULL;
+ /* 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)
+ + strlen (quoted_host)
+ (brackets_around_host ? 2 : 0)
+ fplen
+ 1);
if (brackets_around_host)
*p++ = '[';
- APPEND (p, url->host);
+ APPEND (p, quoted_host);
if (brackets_around_host)
*p++ = ']';
if (url->port != scheme_port)
if (quoted_user && quoted_user != url->user)
xfree (quoted_user);
- if (quoted_passwd && !hide_password
- && quoted_passwd != url->passwd)
+ if (quoted_passwd && !hide_password && quoted_passwd != url->passwd)
xfree (quoted_passwd);
+ if (quoted_host != url->host)
+ xfree (quoted_host);
return result;
}
projects / wget / blobdiff