[Webkit-unassigned] [Bug 48758] [Qt] DRT sideeffect revealed by r71004

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=48758





--- Comment #3 from Robert Hogan <robert at webkit.org>  2010-11-01 13:12:55 PST ---
Yes, this is related to: https://bugs.webkit.org/show_bug.cgi?id=37725

Specifically the code that adds the LayoutTestController object:

void QWebFrame::addToJavaScriptWindowObject(const QString &name, QObject *object, QScriptEngine::ValueOwnership ownership)
{
    if (!page()->settings()->testAttribute(QWebSettings::JavascriptEnabled))
        return;
#if USE(JSC)
    JSC::JSLock lock(JSC::SilenceAssertionsOnly);
    JSDOMWindow* window = toJSDOMWindow(d->frame, mainThreadNormalWorld());
    JSC::Bindings::RootObject* root;
    if (ownership == QScriptEngine::QtOwnership)
        root = d->frame->script()->cacheableBindingRootObject();
    else
        root = d->frame->script()->bindingRootObject();

At the moment LayoutTestController gets added as a cacheableBindingRootObject() so that it survives page navigations. This object will only get destroyed once the page has been destroyed. 

If you change it so that LayoutTestController gets added as a bindingRootObject() instead it will get destroyed on the page navigation implicit in loading a new test and node-from-rect-basic.html passes when run as the following tests.

So there is something wrong with the state of the Qt root object after loading an initial page, then loading a new one. This problem does not arise with a 'fresh' root object.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.