/* Is X ".."? */
#define DDOTP(x) ((*(x) == '.') && (*(x + 1) == '.') && (!*(x + 2)))
-static int urlpath_length PARAMS ((const char *));
-
struct scheme_data
{
char *leading_string;
int default_port;
+ int enabled;
};
/* Supported schemes: */
static struct scheme_data supported_schemes[] =
{
- { "http://", DEFAULT_HTTP_PORT },
+ { "http://", DEFAULT_HTTP_PORT, 1 },
#ifdef HAVE_SSL
- { "https://", DEFAULT_HTTPS_PORT },
+ { "https://", DEFAULT_HTTPS_PORT, 1 },
#endif
- { "ftp://", DEFAULT_FTP_PORT },
+ { "ftp://", DEFAULT_FTP_PORT, 1 },
/* SCHEME_INVALID */
- { NULL, -1 }
+ { NULL, -1, 0 }
};
+/* Forward declarations: */
+
static char *construct_relative PARAMS ((const char *, const char *));
+static int path_simplify PARAMS ((char *));
+
\f
/* Support for encoding and decoding of URL strings. We determine
"foo+bar" -> "foo+bar" (plus is reserved!)
"foo%2b+bar" -> "foo%2b+bar" */
-char *
+static char *
reencode_string (const char *s)
{
const char *p1;
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;
+ if (0 == strncasecmp (url, supported_schemes[i].leading_string,
+ strlen (supported_schemes[i].leading_string)))
+ {
+ if (supported_schemes[i].enabled)
+ return (enum url_scheme) i;
+ else
+ return SCHEME_INVALID;
+ }
+
return SCHEME_INVALID;
}
return supported_schemes[scheme].default_port;
}
+void
+scheme_disable (enum url_scheme scheme)
+{
+ supported_schemes[scheme].enabled = 0;
+}
+
/* 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 scheme.
if (*p == ':')
{
- const char *pp, *path;
+ const char *pp;
char *res;
/* If the characters after the colon and before the next slash
or end of string are all digits, it's HTTP. */
int digits = 0;
for (pp = p + 1; ISDIGIT (*pp); pp++)
++digits;
- if (digits > 0
- && (*pp == '/' || *pp == '\0'))
+ 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 '/'. */
{
int change = 0;
for (; *str; str++)
- if (!ISLOWER (*str))
+ if (ISUPPER (*str))
{
change = 1;
*str = TOLOWER (*str);
static char *parse_errors[] = {
#define PE_NO_ERROR 0
"No error",
-#define PE_UNRECOGNIZED_SCHEME 1
- "Unrecognized scheme",
+#define PE_UNSUPPORTED_SCHEME 1
+ "Unsupported scheme",
#define PE_EMPTY_HOST 2
"Empty host",
#define PE_BAD_PORT_NUMBER 3
scheme = url_scheme (url);
if (scheme == SCHEME_INVALID)
{
- SETERR (error, PE_UNRECOGNIZED_SCHEME);
+ SETERR (error, PE_UNSUPPORTED_SCHEME);
return NULL;
}
if (fragment_b)
u->fragment = strdupdelim (fragment_b, fragment_e);
-
- if (path_modified || u->fragment || host_modified)
+ if (path_modified || u->fragment || host_modified || path_b == path_e)
{
- /* If path_simplify modified the path, or if a fragment is
- present, or if the original host name had caps in it, make
- sure that u->url is equivalent to what would be printed by
- url_string. */
+ /* If we suspect that a transformation has rendered what
+ url_string might return different from URL_ENCODED, rebuild
+ u->url using url_string. */
u->url = url_string (u, 0);
if (url_encoded != url)
#undef FROB
}
-/* Public function for getting the "full path". */
+/* Public function for getting the "full path". E.g. if u->path is
+ "foo/bar" and u->query is "param=value", full_path will be
+ "/foo/bar?param=value". */
+
char *
url_full_path (const struct url *url)
{
}
/* Sync u->path and u->url with u->dir and u->file. */
+
static void
sync_path (struct url *url)
{
return NULL;
}
DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
+
head = tail = NULL;
text = fm->content;
text_end = fm->content + fm->length;
else
++line_end;
text = line_end;
- while (line_beg < line_end
- && ISSPACE (*line_beg))
+
+ /* Strip whitespace from the beginning and end of line. */
+ while (line_beg < line_end && ISSPACE (*line_beg))
++line_beg;
- while (line_end > line_beg + 1
- && ISSPACE (*(line_end - 1)))
+ while (line_end > line_beg && ISSPACE (*(line_end - 1)))
--line_end;
+
if (line_end > line_beg)
{
+ /* URL is in the [line_beg, line_end) region. */
+
int up_error_code;
char *url_text;
struct urlpos *entry;
struct url *url;
- /* We must copy the URL to a zero-terminated string. *sigh*. */
+ /* We must copy the URL to a zero-terminated string, and we
+ can't use alloca because we're in a loop. *sigh*. */
url_text = strdupdelim (line_beg, line_end);
+
+ if (opt.base_href)
+ {
+ /* Merge opt.base_href with URL. */
+ char *merged = uri_merge (opt.base_href, url_text);
+ xfree (url_text);
+ url_text = merged;
+ }
+
url = url_parse (url_text, &up_error_code);
if (!url)
{
{
int len = strlen (dirpref);
dirpref[len] = ':';
- long_to_string (dirpref + len + 1, u->port);
+ number_to_string (dirpref + len + 1, u->port);
}
}
else /* not add_hostdir */
return name;
}
-/* Like strlen(), but allow the URL to be ended with '?'. */
+/* Return the langth of URL's path. Path is considered to be
+ terminated by one of '?', ';', '#', or by the end of the
+ string. */
static int
-urlpath_length (const char *url)
+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 almost completely equivalent to
- { *e = '\0'; return strrchr(b); }, except that it doesn't change
- the contents of the string. */
+ 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)
{
return e;
return NULL;
}
+\f
+/* Resolve "." and ".." elements of PATH by destructively modifying
+ PATH. "." is resolved by removing that path element, and ".." is
+ resolved by removing the preceding path element. Leading and
+ trailing slashes are preserved.
+
+ Return non-zero if any changes have been made.
+
+ For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
+ test examples are provided below. If you change anything in this
+ function, run test_path_simplify to make sure you haven't broken a
+ test case.
+
+ A previous version of this function was based on path_simplify()
+ from GNU Bash, but it has been rewritten for Wget 1.8.1. */
+
+static int
+path_simplify (char *path)
+{
+ int change = 0;
+ char *p, *end;
+
+ if (path[0] == '/')
+ ++path; /* preserve the leading '/'. */
+
+ p = path;
+ end = p + strlen (p) + 1; /* position past the terminating zero. */
+
+ while (1)
+ {
+ again:
+ /* P should point to the beginning of a path element. */
+
+ if (*p == '.' && (*(p + 1) == '/' || *(p + 1) == '\0'))
+ {
+ /* Handle "./foo" by moving "foo" two characters to the
+ left. */
+ if (*(p + 1) == '/')
+ {
+ change = 1;
+ memmove (p, p + 2, end - p);
+ end -= 2;
+ goto again;
+ }
+ else
+ {
+ change = 1;
+ *p = '\0';
+ break;
+ }
+ }
+ else if (*p == '.' && *(p + 1) == '.'
+ && (*(p + 2) == '/' || *(p + 2) == '\0'))
+ {
+ /* Handle "../foo" by moving "foo" one path element to the
+ left. */
+ char *b = p; /* not p-1 because P can equal PATH */
+
+ /* Backtrack by one path element, but not past the beginning
+ of PATH. */
+
+ /* foo/bar/../baz */
+ /* ^ p */
+ /* ^ b */
+
+ if (b > path)
+ {
+ /* Move backwards until B hits the beginning of the
+ previous path element or the beginning of path. */
+ for (--b; b > path && *(b - 1) != '/'; b--)
+ ;
+ }
+
+ change = 1;
+ if (*(p + 2) == '/')
+ {
+ memmove (b, p + 3, end - (p + 3));
+ end -= (p + 3) - b;
+ p = b;
+ }
+ else
+ {
+ *b = '\0';
+ break;
+ }
+ goto again;
+ }
+ else if (*p == '/')
+ {
+ /* Remove empty path elements. Not mandated by rfc1808 et
+ al, but empty path elements are not all that useful, and
+ the rest of Wget might not deal with them well. */
+ char *q = p;
+ while (*q == '/')
+ ++q;
+ change = 1;
+ if (*q == '\0')
+ {
+ *p = '\0';
+ break;
+ }
+ memmove (p, q, end - q);
+ end -= q - p;
+ goto again;
+ }
+
+ /* Skip to the next path element. */
+ while (*p && *p != '/')
+ ++p;
+ if (*p == '\0')
+ break;
+
+ /* Make sure P points to the beginning of the next path element,
+ which is location after the slash. */
+ ++p;
+ }
+
+ return change;
+}
+\f
/* Resolve the result of "linking" a base URI (BASE) to a
link-specified URI (LINK).
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. */
+ Perhaps this function should call path_simplify so that the callers
+ don't have to call url_parse unconditionally. */
static char *
uri_merge_1 (const char *base, const char *link, int linklength, int no_scheme)
{
if (no_scheme)
{
- const char *end = base + urlpath_length (base);
+ const char *end = base + path_length (base);
if (!*link)
{
if (url->port != scheme_port)
{
*p++ = ':';
- long_to_string (p, url->port);
- p += strlen (p);
+ p = number_to_string (p, url->port);
}
full_path_write (url, p);
return !sufmatch (no_proxy, host);
}
\f
-static void write_backup_file PARAMS ((const char *, downloaded_file_t));
-static void replace_attr PARAMS ((const char **, int, FILE *, const char *));
+/* Support for converting links for local viewing in downloaded HTML
+ files. This should be moved to another file, because it has
+ nothing to do with processing URLs. */
-/* Change the links in an HTML document. Accepts a structure that
- defines the positions of all the links. */
+static void write_backup_file PARAMS ((const char *, downloaded_file_t));
+static const char *replace_attr PARAMS ((const char *, int, FILE *,
+ const char *));
+static const char *replace_attr_refresh_hack PARAMS ((const char *, int, FILE *,
+ const char *, int));
+static char *local_quote_string PARAMS ((const char *));
+
+/* Change the links in one HTML file. LINKS is a list of links in the
+ document, along with their positions and the desired direction of
+ the conversion. */
void
-convert_links (const char *file, struct urlpos *l)
+convert_links (const char *file, struct urlpos *links)
{
struct file_memory *fm;
- FILE *fp;
- const char *p;
- downloaded_file_t downloaded_file_return;
+ FILE *fp;
+ const char *p;
+ downloaded_file_t downloaded_file_return;
+
+ struct urlpos *link;
int to_url_count = 0, to_file_count = 0;
logprintf (LOG_VERBOSE, _("Converting %s... "), file);
any URL needs to be converted in the first place. If not, just
leave the file alone. */
int dry_count = 0;
- struct urlpos *dry = l;
- for (dry = l; dry; dry = dry->next)
+ struct urlpos *dry = links;
+ for (dry = links; dry; dry = dry->next)
if (dry->convert != CO_NOCONVERT)
++dry_count;
if (!dry_count)
read_file_free (fm);
return;
}
+
/* Here we loop through all the URLs in file, replacing those of
them that are downloaded with relative references. */
p = fm->content;
- for (; l; l = l->next)
+ for (link = links; link; link = link->next)
{
- char *url_start = fm->content + l->pos;
+ char *url_start = fm->content + link->pos;
- if (l->pos >= fm->length)
+ if (link->pos >= fm->length)
{
DEBUGP (("Something strange is going on. Please investigate."));
break;
}
/* If the URL is not to be converted, skip it. */
- if (l->convert == CO_NOCONVERT)
+ if (link->convert == CO_NOCONVERT)
{
- DEBUGP (("Skipping %s at position %d.\n", l->url->url, l->pos));
+ DEBUGP (("Skipping %s at position %d.\n", link->url->url, link->pos));
continue;
}
quote, to the outfile. */
fwrite (p, 1, url_start - p, fp);
p = url_start;
- if (l->convert == CO_CONVERT_TO_RELATIVE)
+
+ switch (link->convert)
{
+ case CO_CONVERT_TO_RELATIVE:
/* Convert absolute URL to relative. */
- char *newname = construct_relative (file, l->local_name);
- 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->url, newname, l->pos, file));
- xfree (newname);
- xfree (quoted_newname);
- ++to_file_count;
- }
- else if (l->convert == CO_CONVERT_TO_COMPLETE)
- {
+ {
+ char *newname = construct_relative (file, link->local_name);
+ char *quoted_newname = local_quote_string (newname);
+
+ if (!link->link_refresh_p)
+ p = replace_attr (p, link->size, fp, quoted_newname);
+ else
+ p = replace_attr_refresh_hack (p, link->size, fp, quoted_newname,
+ link->refresh_timeout);
+
+ DEBUGP (("TO_RELATIVE: %s to %s at position %d in %s.\n",
+ link->url->url, newname, link->pos, file));
+ xfree (newname);
+ xfree (quoted_newname);
+ ++to_file_count;
+ break;
+ }
+ case CO_CONVERT_TO_COMPLETE:
/* Convert the link to absolute URL. */
- char *newlink = l->url->url;
- char *quoted_newlink = html_quote_string (newlink);
- replace_attr (&p, l->size, fp, quoted_newlink);
- DEBUGP (("TO_COMPLETE: <something> to %s at position %d in %s.\n",
- newlink, l->pos, file));
- xfree (quoted_newlink);
- ++to_url_count;
+ {
+ char *newlink = link->url->url;
+ char *quoted_newlink = html_quote_string (newlink);
+
+ if (!link->link_refresh_p)
+ p = replace_attr (p, link->size, fp, quoted_newlink);
+ else
+ p = replace_attr_refresh_hack (p, link->size, fp, quoted_newlink,
+ link->refresh_timeout);
+
+ DEBUGP (("TO_COMPLETE: <something> to %s at position %d in %s.\n",
+ newlink, link->pos, file));
+ xfree (quoted_newlink);
+ ++to_url_count;
+ break;
+ }
+ case CO_NULLIFY_BASE:
+ /* Change the base href to "". */
+ p = replace_attr (p, link->size, fp, "");
+ break;
+ case CO_NOCONVERT:
+ abort ();
+ break;
}
}
+
/* Output the rest of the file. */
if (p - fm->content < fm->length)
fwrite (p, 1, fm->length - (p - fm->content), fp);
fclose (fp);
read_file_free (fm);
- logprintf (LOG_VERBOSE,
- _("%d-%d\n"), to_file_count, to_url_count);
+
+ logprintf (LOG_VERBOSE, "%d-%d\n", to_file_count, to_url_count);
}
/* Construct and return a malloced copy of the relative link from two
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)
+/* Replace an attribute's original text with NEW_TEXT. */
+
+static const char *
+replace_attr (const char *p, int size, FILE *fp, const char *new_text)
{
- const char *p = *pp;
int quote_flag = 0;
- int size = raw_size;
- char quote_char = '\"';
+ char quote_char = '\"'; /* use "..." for quoting, unless the
+ original value is quoted, in which
+ case reuse its quoting char. */
const char *frag_beg, *frag_end;
/* Structure of our string is:
"...old-contents..."
- <--- l->size ---> (with quotes)
+ <--- size ---> (with quotes)
OR:
...old-contents...
- <--- l->size --> (no quotes) */
+ <--- size --> (no quotes) */
if (*p == '\"' || *p == '\'')
{
size -= 2; /* disregard opening and closing quote */
}
putc (quote_char, fp);
- fputs (new_str, fp);
+ fputs (new_text, fp);
/* Look for fragment identifier, if any. */
if (find_fragment (p, size, &frag_beg, &frag_end))
if (quote_flag)
++p;
putc (quote_char, fp);
- *pp = p;
+
+ return p;
+}
+
+/* The same as REPLACE_ATTR, but used when replacing
+ <meta http-equiv=refresh content="new_text"> because we need to
+ append "timeout_value; URL=" before the next_text. */
+
+static const char *
+replace_attr_refresh_hack (const char *p, int size, FILE *fp,
+ const char *new_text, int timeout)
+{
+ /* "0; URL=..." */
+ char *new_with_timeout = (char *)alloca (numdigit (timeout)
+ + 6 /* "; URL=" */
+ + strlen (new_text)
+ + 1);
+ sprintf (new_with_timeout, "%d; URL=%s", timeout, new_text);
+
+ return replace_attr (p, size, fp, new_with_timeout);
}
/* Find the first occurrence of '#' in [BEG, BEG+SIZE) that is not
return 0;
}
+/* Quote FILE for use as local reference to an HTML file.
+
+ We quote ? as %3F to avoid passing part of the file name as the
+ parameter when browsing the converted file through HTTP. However,
+ it is safe to do this only when `--html-extension' is turned on.
+ This is because converting "index.html?foo=bar" to
+ "index.html%3Ffoo=bar" would break local browsing, as the latter
+ isn't even recognized as an HTML file! However, converting
+ "index.html?foo=bar.html" to "index.html%3Ffoo=bar.html" should be
+ safe for both local and HTTP-served browsing. */
+
+static char *
+local_quote_string (const char *file)
+{
+ const char *file_sans_qmark;
+ int qm;
+
+ if (!opt.html_extension)
+ return html_quote_string (file);
+
+ qm = count_char (file, '?');
+
+ if (qm)
+ {
+ const char *from = file;
+ char *to, *newname;
+
+ /* qm * 2 because we replace each question mark with "%3F",
+ i.e. replace one char with three, hence two more. */
+ int fsqlen = strlen (file) + qm * 2;
+
+ to = newname = (char *)alloca (fsqlen + 1);
+ for (; *from; from++)
+ {
+ if (*from != '?')
+ *to++ = *from;
+ else
+ {
+ *to++ = '%';
+ *to++ = '3';
+ *to++ = 'F';
+ }
+ }
+ assert (to - newname == fsqlen);
+ *to = '\0';
+
+ file_sans_qmark = newname;
+ }
+ else
+ file_sans_qmark = file;
+
+ return html_quote_string (file_sans_qmark);
+}
+
/* We're storing "modes" of type downloaded_file_t in the hash table.
However, our hash tables only accept pointers for keys and values.
So when we need a pointer, we use the address of a
downloaded_files_hash = NULL;
}
}
+\f
+#if 0
+/* Debugging and testing support for path_simplify. */
+
+/* Debug: run path_simplify on PATH and return the result in a new
+ string. Useful for calling from the debugger. */
+static char *
+ps (char *path)
+{
+ char *copy = xstrdup (path);
+ path_simplify (copy);
+ return copy;
+}
+
+static void
+run_test (char *test, char *expected_result, int expected_change)
+{
+ char *test_copy = xstrdup (test);
+ int modified = path_simplify (test_copy);
+
+ if (0 != strcmp (test_copy, expected_result))
+ {
+ printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
+ test, expected_result, test_copy);
+ }
+ if (modified != expected_change)
+ {
+ if (expected_change == 1)
+ printf ("Expected no modification with path_simplify(\"%s\").\n",
+ test);
+ else
+ printf ("Expected modification with path_simplify(\"%s\").\n",
+ test);
+ }
+ xfree (test_copy);
+}
+
+static void
+test_path_simplify (void)
+{
+ static struct {
+ 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 }
+ };
+ int i;
+
+ for (i = 0; i < ARRAY_SIZE (tests); i++)
+ {
+ char *test = tests[i].test;
+ char *expected_result = tests[i].result;
+ int expected_change = tests[i].should_modify;
+ run_test (test, expected_result, expected_change);
+ }
+
+ /* 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 < ARRAY_SIZE (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