Refactor adblock simplify parsing and to honor matching options

There is a known flaw in that URL rules may erroneously match domains.

The code path for WebKitGTK+ < 1.1.14 is now crashing.
This commit is contained in:
Alexander Butenko 2009-11-20 01:00:48 +01:00 committed by Christian Dywan
parent 5580b62e00
commit 2d514bfb6c

View file

@ -26,6 +26,13 @@
(__filter && (g_str_has_prefix (__filter, "http") \ (__filter && (g_str_has_prefix (__filter, "http") \
|| g_str_has_prefix (__filter, "file"))) || g_str_has_prefix (__filter, "file")))
typedef struct
{
const gchar* page_uri;
const gchar* uri;
const gchar* query;
} Matcher;
static GHashTable* pattern; static GHashTable* pattern;
static gchar* blockcss = NULL; static gchar* blockcss = NULL;
static gchar* blockcssprivate = NULL; static gchar* blockcssprivate = NULL;
@ -50,12 +57,11 @@ adblock_build_js (const gchar* style,
" break;" " break;"
" }}" " }}"
" public += ' {display: none !important;}';" " public += ' {display: none !important;}';"
"var mystyle = document.createElement(\"style\");" " var mystyle = document.createElement('style');"
"mystyle.setAttribute(\"type\", \"text/css\");" " mystyle.setAttribute('type', 'text/css');"
" mystyle.appendChild(document.createTextNode(public));" " mystyle.appendChild(document.createTextNode(public));"
"var head = document.getElementsByTagName(\"head\")[0];" " var head = document.getElementsByTagName('head')[0];"
" if (head) head.appendChild(mystyle);" " if (head) head.appendChild(mystyle);"
"else document.documentElement.insertBefore(mystyle, document.documentElement.firstChild);"
"}, true);", "}, true);",
private, private,
style); style);
@ -73,10 +79,6 @@ adblock_fixup_regexp (gchar* src)
/* FIXME: Avoid always allocating twice the string */ /* FIXME: Avoid always allocating twice the string */
s = dst = g_malloc (strlen (src) * 2); s = dst = g_malloc (strlen (src) * 2);
/* |http:// means ^http:// */
if (src[0] == '|')
src[0] = '^';
while (*src) while (*src)
{ {
switch (*src) switch (*src)
@ -93,6 +95,15 @@ adblock_fixup_regexp (gchar* src)
case '|': case '|':
*s++ = '\\'; *s++ = '\\';
break; break;
case '/':
*s++ = '\\';
break;
/* FIXME: We actually need to match :[0-9]+ or '/'. Sign means
"here could be port number or nothing". So bla.com^ will match
bla.com/ or bla.com:8080/ but not bla.com.au/ */
case '^':
*src = '?';
break;
} }
*s++ = *src; *s++ = *src;
src++; src++;
@ -518,11 +529,37 @@ adblock_browser_populate_tool_menu_cb (MidoriBrowser* browser,
} }
static gboolean static gboolean
adblock_is_matched (const gchar* patt, adblock_is_matched (const gchar* opts,
const GRegex* regex, const GRegex* regex,
const gchar* uri) Matcher* data)
{ {
return g_regex_match_full (regex, uri, -1, 0, 0, NULL, NULL); gchar* patt;
if (g_regex_match_simple ("type=fulluri,", opts, G_REGEX_UNGREEDY, G_REGEX_MATCH_NOTEMPTY))
patt = g_strdup (data->uri);
else
patt = g_strdup (data->query);
if (g_regex_match_full (regex, patt, -1, 0, 0, NULL, NULL))
{
if (g_regex_match_simple (",third-party", opts,
G_REGEX_CASELESS, G_REGEX_MATCH_NOTEMPTY))
{
if (data->page_uri && g_regex_match_full (regex, data->page_uri, -1, 0, 0, NULL, NULL))
{
g_free (patt);
return FALSE;
}
}
/* TODO: Domain opt check */
g_free (patt);
return TRUE;
}
else
{
g_free (patt);
return FALSE;
}
} }
#if HAVE_WEBKIT_RESOURCE_REQUEST #if HAVE_WEBKIT_RESOURCE_REQUEST
@ -534,10 +571,38 @@ adblock_resource_request_starting_cb (WebKitWebView* web_view,
WebKitNetworkResponse* response, WebKitNetworkResponse* response,
GtkWidget* image) GtkWidget* image)
{ {
const gchar* uri = webkit_network_request_get_uri (request); Matcher data;
if (!strncmp (uri, "data", 4)) const char *page_uri;
const gchar* uri;
SoupMessage* msg;
SoupURI* soup_uri;
uri = webkit_network_request_get_uri (request);
if (!strncmp (uri, "data", 4) || !strncmp (uri, "file", 4))
return; return;
if (g_hash_table_find (pattern, (GHRFunc) adblock_is_matched, (char*)uri))
msg = webkit_network_request_get_message (request);
if (!msg)
return;
if (msg->method && !strncmp (msg->method, "POST", 4))
return;
soup_uri = soup_uri_new (uri);
if (soup_uri->query)
data.query = g_strdup_printf ("%s?%s", soup_uri->path, soup_uri->query);
else
data.query = g_strdup (soup_uri->path);
soup_uri_free (soup_uri);
data.uri = uri;
page_uri = webkit_web_view_get_uri (web_view);
if (!page_uri || !strcmp (page_uri, "about:blank"))
page_uri = uri;
data.page_uri = page_uri;
if (g_hash_table_find (pattern, (GHRFunc) adblock_is_matched, &data))
{ {
#if 0 #if 0
gchar* text; gchar* text;
@ -558,9 +623,32 @@ static void
adblock_session_request_queued_cb (SoupSession* session, adblock_session_request_queued_cb (SoupSession* session,
SoupMessage* msg) SoupMessage* msg)
{ {
SoupURI* soup_uri = soup_message_get_uri (msg); Matcher data;
gchar* uri = soup_uri ? soup_uri_to_string (soup_uri, FALSE) : g_strdup (""); SoupURI* soup_uri;
if (g_hash_table_find (pattern, (GHRFunc) adblock_is_matched, uri)) gchar* uri;
gchar* page_uri;
if (msg->method && !strncmp (msg->method, "POST", 4))
return;
/* FIXME: There is a crasher somewhere introduced with the refactoring */
soup_uri = soup_message_get_uri (msg);
uri = soup_uri_to_string (soup_uri, FALSE);
if (soup_uri->query)
data.query = g_strdup_printf ("%s?%s", soup_uri->path, soup_uri->query);
else
data.query = g_strdup (soup_uri->path);
soup_uri_free (soup_uri);
data.uri = uri;
page_uri = NULL; /* FIXME */
if (!page_uri || !strcmp (page_uri, "about:blank"))
page_uri = uri;
data.page_uri = page_uri;
if (g_hash_table_find (pattern, (GHRFunc) adblock_is_matched, &data))
{ {
/* FIXME: Update image tooltip */ /* FIXME: Update image tooltip */
@ -635,6 +723,27 @@ adblock_app_add_browser_cb (MidoriApp* app,
g_object_unref (statusbar); g_object_unref (statusbar);
} }
static void
adblock_compile_regexp (GHashTable* tbl,
gchar* patt,
gchar* opts)
{
GRegex* regex;
GError* error = NULL;
/* TODO: Play with optimization flags */
regex = g_regex_new (patt, G_REGEX_OPTIMIZE,
G_REGEX_MATCH_NOTEMPTY, &error);
if (!error)
g_hash_table_insert (tbl, opts, regex);
else
{
g_warning ("%s: %s", G_STRFUNC, error->message);
g_error_free (error);
}
}
static void static void
adblock_frame_add (gchar* line) adblock_frame_add (gchar* line)
{ {
@ -657,6 +766,7 @@ adblock_frame_add_private (gchar* line)
{ {
gchar** domains; gchar** domains;
gint max, i; gint max, i;
domains = g_strsplit (data[0], ",", -1); domains = g_strsplit (data[0], ",", -1);
for (max = i = 0; domains[i]; i++) for (max = i = 0; domains[i]; i++)
{ {
@ -675,6 +785,56 @@ adblock_frame_add_private (gchar* line)
g_strfreev (data); g_strfreev (data);
} }
static void
adblock_add_url_pattern (gchar* line)
{
gchar* opts;
gchar** data;
gchar* patt;
gchar* parsed;
if (line[0] == '|' && line[1] == '|' )
{
(void)*line++;
(void)*line++;
data = g_strsplit (line, "$", 2);
parsed = adblock_fixup_regexp (data[0]);
patt = g_strdup_printf ("^https?://([a-z0-9\\.]+)?%s", parsed);
if (data[1])
opts = g_strdup_printf ("type=fulluri,regexp=%s,%s", patt, data[1]);
else
opts = g_strdup_printf ("type=fulluri,regexp=%s", patt);
g_strfreev (data);
g_free (parsed);
}
else if (line[0] == '|')
{
(void)*line++;
data = g_strsplit (line, "$", 2);
parsed = adblock_fixup_regexp (data[0]);
patt = g_strdup_printf ("^%s", parsed);
if (data[1])
opts = g_strdup_printf ("type=fulluri,regexp=%s,%s", patt, data[1]);
else
opts = g_strdup_printf ("type=fulluri,regexp=%s", patt);
g_strfreev (data);
g_free (parsed);
}
else
{
patt = adblock_fixup_regexp (line);
opts = g_strdup_printf ("regexp=%s", patt);
}
/* g_debug ("got: %s opts %s", patt, opts); */
adblock_compile_regexp (pattern, patt, opts);
g_free (patt);
}
static gchar* static gchar*
adblock_parse_line (gchar* line) adblock_parse_line (gchar* line)
{ {
@ -687,63 +847,42 @@ adblock_parse_line (gchar* line)
/* FIXME: No support for whitelisting */ /* FIXME: No support for whitelisting */
if (line[0] == '@' && line[1] == '@') if (line[0] == '@' && line[1] == '@')
return NULL; return NULL;
/* FIXME: What is this? */ /* FIXME: No support for [include] and [exclude] tags */
if (line[0] == '|' && line[1] == '|') if (line[0] == '[')
return NULL; return NULL;
/* ditto */
if (strstr (line,"$")) /* Got CSS block hider */
return NULL; if (line[0] == '#' && line[1] == '#' )
/* Got block hider */
if (line[0] == '#' && line[1] == '#' && (line[2] == '.' || line[2] == '#'
|| line[2] == 'A' || line[2] == 'a' || line[2] == 'D' || line[2] == 'U'))
{ {
adblock_frame_add (line); adblock_frame_add (line);
return NULL; return NULL;
} }
/* FIXME: Do we have smth else starting with ##? */ /* Some crazy lists do this */
if (line[0] == '#' && line[1] == '#') if (line[0] == '#')
return NULL; return NULL;
/* Got per domain CSS hider rule */
if (strstr (line,"##")) if (strstr (line,"##"))
{ {
adblock_frame_add_private (line); adblock_frame_add_private (line);
return NULL; return NULL;
} }
/* FIXME: No support for [include] and [exclude] tags */
if (line[0] == '[') /* Got URL blocker rule */
return NULL; adblock_add_url_pattern (line);
return adblock_fixup_regexp (line); return line;
} }
static void static void
adblock_parse_file (gchar* path) adblock_parse_file (gchar* path)
{ {
FILE* file; FILE* file;
gchar line[500];
if ((file = g_fopen (path, "r"))) if ((file = g_fopen (path, "r")))
{ {
gchar line[500];
GRegex* regex;
while (fgets (line, 500, file)) while (fgets (line, 500, file))
{ adblock_parse_line (line);
GError* error = NULL;
gchar* parsed;
parsed = adblock_parse_line (line);
if (!parsed)
continue;
regex = g_regex_new (parsed, G_REGEX_OPTIMIZE,
G_REGEX_MATCH_NOTEMPTY, &error);
if (error)
{
g_warning ("%s: %s", G_STRFUNC, error->message);
g_error_free (error);
g_free (parsed);
}
else
g_hash_table_insert (pattern, parsed, regex);
}
fclose (file); fclose (file);
} }
} }
@ -846,6 +985,7 @@ test_adblock_pattern (void)
temp = g_file_open_tmp ("midori_adblock_match_test_XXXXXX", &filename, NULL); temp = g_file_open_tmp ("midori_adblock_match_test_XXXXXX", &filename, NULL);
/* TODO: Update some tests and add new ones. */
g_file_set_contents (filename, g_file_set_contents (filename,
"*ads.foo.bar*\n" "*ads.foo.bar*\n"
"*ads.bogus.name*\n" "*ads.bogus.name*\n"