struct scheme_data
{
- char *leading_string;
+ const char *name;
+ const char *leading_string;
int default_port;
int enabled;
};
/* Supported schemes: */
static struct scheme_data supported_schemes[] =
{
- { "http://", DEFAULT_HTTP_PORT, 1 },
+ { "http", "http://", DEFAULT_HTTP_PORT, 1 },
#ifdef HAVE_SSL
- { "https://", DEFAULT_HTTPS_PORT, 1 },
+ { "https", "https://", DEFAULT_HTTPS_PORT, 1 },
#endif
- { "ftp://", DEFAULT_FTP_PORT, 1 },
+ { "ftp", "ftp://", DEFAULT_FTP_PORT, 1 },
/* SCHEME_INVALID */
- { NULL, -1, 0 }
+ { NULL, NULL, -1, 0 }
};
/* Forward declarations: */
static int path_simplify PARAMS ((char *));
\f
-/* 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". Wget
+ uses the rfc1738 set of reserved characters, plus "$" and ",", as
+ recommended by rfc2396.
+
+ An unsafe characters is the one that should be encoded when URLs
+ are placed in foreign environments. E.g. space and newline are
+ unsafe in HTTP contexts because HTTP uses them as separator and
+ terminator, so they must be encoded to %20 and %0A respectively.
+ "*" is unsafe in shell context, etc.
+
+ We determine whether a character is unsafe through static table
+ lookup. This code assumes ASCII character set and 8-bit chars. */
enum {
- /* rfc1738 reserved chars, preserved from encoding. */
+ /* rfc1738 reserved chars + "$" and ",". */
urlchr_reserved = 1,
- /* rfc1738 unsafe chars, plus some more. */
+ /* rfc1738 unsafe chars, plus non-printables. */
urlchr_unsafe = 2
};
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 */
return change;
}
-static char *parse_errors[] = {
+static const char *parse_errors[] = {
#define PE_NO_ERROR 0
N_("No error"),
#define PE_UNSUPPORTED_SCHEME 1
static int
is_valid_ipv4_address (const char *str, const char *end)
{
- int saw_digit, octets;
- int val;
+ int saw_digit = 0;
+ int octets = 0;
+ int val = 0;
- saw_digit = 0;
- octets = 0;
- val = 0;
-
- while (str < end) {
- int ch = *str++;
+ while (str < end)
+ {
+ int ch = *str++;
- if (ch >= '0' && ch <= '9') {
- val = val * 10 + (ch - '0');
+ 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 (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 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";
+ enum {
+ NS_INADDRSZ = 4,
+ NS_IN6ADDRSZ = 16,
+ NS_INT16SZ = 2
+ };
+
const char *curtok;
int tp;
const char *colonp;
saw_xdigit = 0;
val = 0;
- while (str < end) {
- int ch = *str++;
- const char *pch;
+ while (str < end)
+ {
+ int ch = *str++;
- /* 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 number, add it to val. */
+ if (ISXDIGIT (ch))
+ {
+ val <<= 4;
+ val |= XDIGIT_TO_NUM (ch);
+ 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 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)
+ if (saw_xdigit == 1)
+ {
+ 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;
+ if (colonp != NULL)
+ {
+ if (tp == NS_IN6ADDRSZ)
+ return 0;
+ tp = NS_IN6ADDRSZ;
}
-
- 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;
p = strpbrk_or_eos (p, "/;?#");
port_e = p;
- if (port_b == port_e)
+ /* Allow empty port, as per rfc2396. */
+ if (port_b != port_e)
{
- /* http://host:/whatever */
- /* ^ */
- error_code = PE_BAD_PORT_NUMBER;
- goto error;
- }
-
- for (port = 0, pp = port_b; pp < port_e; pp++)
- {
- if (!ISDIGIT (*pp))
+ for (port = 0, pp = port_b; pp < port_e; pp++)
{
- /* http://host:12randomgarbage/blah */
- /* ^ */
- error_code = PE_BAD_PORT_NUMBER;
- goto error;
+ if (!ISDIGIT (*pp))
+ {
+ /* http://host:12randomgarbage/blah */
+ /* ^ */
+ error_code = PE_BAD_PORT_NUMBER;
+ goto error;
+ }
+ port = 10 * port + (*pp - '0');
}
-
- port = 10 * port + (*pp - '0');
}
}
}
}
- 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;
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);
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_control = 4 /* a control character, e.g. 0-31 */
};
#define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
/* 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_P is non-zero, 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, int escaped_p,
+ struct growable *dest)
{
- char *pathel;
- int pathlen;
-
const char *p;
int quoted, outlen;
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_p)
+ {
+ char *unescaped;
+ BOUNDED_TO_ALLOCA (b, e, unescaped);
+ url_unescape (unescaped);
+ b = unescaped;
+ e = unescaped + strlen (unescaped);
+ }
+
+ /* 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;
+ }
- /* Go through PATHEL and check how many characters we'll need to
- add for file quoting. */
+ /* 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++)
+ for (p = b; p < e; p++)
{
if (!FILE_CHAR_TEST (*p, mask))
*q++ = *p;
if (cut-- > 0)
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. */
+ /* Ignore empty pathels. */
continue;
if (dest->tail)
append_char ('/', dest);
- append_uri_pathel (pathel, next, dest);
+ append_uri_pathel (pathel, next, 1, dest);
}
}
{
struct growable fnres;
- char *u_file, *u_query;
+ const char *u_file, *u_query;
char *fname, *unique;
fnres.base = NULL;
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)
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);
+ append_uri_pathel (u_file, u_file + strlen (u_file), 0, &fnres);
/* Append "?query" to the file name. */
u_query = u->query && *u->query ? u->query : NULL;
if (u_query)
{
append_char (FN_QUERY_SEP, &fnres);
- append_uri_pathel (u_query, u_query + strlen (u_query), &fnres);
+ append_uri_pathel (u_query, u_query + strlen (u_query), 1, &fnres);
}
/* Zero-terminate the file name. */
xfree (fname);
return unique;
}
-
-/* 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;
-}
-
-/* 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;
-}
\f
/* Resolve "." and ".." elements of PATH by destructively modifying
PATH and return non-zero if PATH has been modified, zero otherwise.
static int
path_simplify (char *path)
{
- char *h, *t, *end;
-
- /* Preserve the leading '/'. */
- if (path[0] == '/')
- ++path;
-
- h = path; /* hare */
- t = path; /* tortoise */
- end = path + strlen (path);
+ char *h = path; /* hare */
+ char *t = path; /* tortoise */
+ char *beg = path; /* boundary for backing the tortoise */
+ char *end = path + strlen (path);
while (h < end)
{
else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
{
/* Handle "../" by retreating the tortoise by one path
- element -- but not past beggining of PATH. */
- if (t > path)
+ element -- but not past beggining. */
+ if (t > beg)
{
- /* Move backwards until B hits the beginning of the
+ /* Move backwards until T hits the beginning of the
previous path element or the beginning of path. */
- for (--t; t > path && t[-1] != '/'; t--)
+ for (--t; t > beg && t[-1] != '/'; t--)
;
}
+ else
+ {
+ /* If we're at the beginning, copy the "../" literally
+ move the beginning so a later ".." doesn't remove
+ it. */
+ beg = t + 3;
+ goto regular;
+ }
h += 3;
}
- else if (*h == '/')
- {
- /* Ignore empty path elements. Supporting them well is hard
- (where do you save "http://x.com///y.html"?), and they
- don't bring any practical gain. Plus, they break our
- filesystem-influenced assumptions: allowing them would
- make "x/y//../z" simplify to "x/y/z", whereas most people
- would expect "x/z". */
- ++h;
- }
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. */
return t != h;
}
\f
+/* 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;
+}
+
+/* Find the last occurrence of character C in the range [b, e), or
+ NULL, if none are present. We might want to use memrchr (a GNU
+ extension) under GNU libc. */
+
+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;
+}
+
/* Merge BASE with LINK and return the resulting URI.
Either of the URIs may be absolute or relative, complete with the
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)
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] == '/')
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;
+ const char *scheme_str = supported_schemes[url->scheme].leading_string;
int fplen = full_path_length (url);
- int brackets_around_host = 0;
+ int brackets_around_host;
assert (scheme_str != NULL);
}
}
- if (strchr (url->host, ':'))
- brackets_around_host = 1;
+ /* Numeric IPv6 addresses can contain ':' and need to be quoted with
+ brackets. */
+ brackets_around_host = strchr (url->host, ':') != NULL;
size = (strlen (scheme_str)
+ strlen (url->host)
if (modified != expected_change)
{
if (expected_change == 1)
- printf ("Expected no modification with path_simplify(\"%s\").\n",
+ printf ("Expected modification with path_simplify(\"%s\").\n",
test);
else
- printf ("Expected modification with path_simplify(\"%s\").\n",
+ printf ("Expected no modification with path_simplify(\"%s\").\n",
test);
}
xfree (test_copy);
char *test, *result;
int 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 }
+ { "", "", 0 },
+ { ".", "", 1 },
+ { "./", "", 1 },
+ { "..", "..", 0 },
+ { "../", "../", 0 },
+ { "foo", "foo", 0 },
+ { "foo/bar", "foo/bar", 0 },
+ { "foo///bar", "foo///bar", 0 },
+ { "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 },
+ { "foo/../../..", "../..", 1 },
+ { "foo/../../bar/../../baz", "../../baz", 1 },
+ { "a/b/../../c", "c", 1 },
+ { "./a/../b", "b", 1 }
};
int i;
int expected_change = tests[i].should_modify;
run_test (test, expected_result, expected_change);
}
-
- /* 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++)
- {
- char *test, *expected_result;
- int expected_change = tests[i].should_modify;
-
- test = xmalloc (1 + strlen (tests[i].test) + 1);
- sprintf (test, "/%s", tests[i].test);
-
- expected_result = xmalloc (1 + strlen (tests[i].result) + 1);
- sprintf (expected_result, "/%s", tests[i].result);
-
- run_test (test, expected_result, expected_change);
-
- xfree (test);
- xfree (expected_result);
- }
}
#endif