summaryrefslogtreecommitdiffstats
path: root/WebKit/android/jni/CookieManager.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'WebKit/android/jni/CookieManager.cpp')
-rw-r--r--WebKit/android/jni/CookieManager.cpp10
1 files changed, 10 insertions, 0 deletions
diff --git a/WebKit/android/jni/CookieManager.cpp b/WebKit/android/jni/CookieManager.cpp
index f16ffd7..34e2cb6 100644
--- a/WebKit/android/jni/CookieManager.cpp
+++ b/WebKit/android/jni/CookieManager.cpp
@@ -46,6 +46,16 @@ static bool useChromiumHttpStack(JNIEnv*, jobject) {
static void removeAllCookie(JNIEnv*, jobject) {
#if USE(CHROME_NETWORK_STACK)
+ // Though WebRequestContext::get() is threadsafe, the context itself, in
+ // general, is not. The context is generally used on the HTTP stack IO
+ // thread, but this call is on the UI thread. However, the cookie_store()
+ // getter just returns a pointer which is only set when the context is first
+ // constructed. The CookieMonster is threadsafe, so the call below is safe
+ // overall.
+ //
+ // TODO: It's possible that this call is made before the WebKit thread has
+ // started up and the settings have been synced to the BrowserFrame. This
+ // will cause creation of the context to fail.
WebRequestContext::get(false)->cookie_store()->GetCookieMonster()->DeleteAllCreatedAfter(Time(), true);
// This will lazily create a new private browsing context. However, if the
// context doesn't already exist, there's no need to create it, as cookies