[Webkit-unassigned] [Bug 62808] [Qt] ASSERTION FAILED in ResourceHandle::setDefersLoading causes crash

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=62808





--- Comment #51 from Luiz Agostini <luiz at webkit.org>  2011-07-13 07:57:48 PST ---
(In reply to comment #49)
> (In reply to comment #48)
> > (In reply to comment #46)
> > > Created an attachment (id=100537)
 --> (https://bugs.webkit.org/attachment.cgi?id=100537&action=review) [details] [details] [details]
> > > Fix the layout test
> > > 
> > > If QNetworkReplyHandler is set to synchronous load, it has to deliver content synchronously. Otherwise, it causes problems in some cases. e.g.
> > > 
> > > void ResourceHandle::loadResourceSynchronously(...) {
> > >     ... ...
> > >     RefPtr<ResourceHandle> handle = adoptRef(new ResourceHandle(request, &syncLoader, true, false));
> > >     ... ...
> > >     d->m_job = new QNetworkReplyHandler(handle.get(), QNetworkReplyHandler::SynchronousLoad, true);
> > >     d->m_job->setLoadingDeferred(false);
> > 
> > This was done this way for d->m_job to be assigned before actually sending the request. Otherwise some callbacks could come before d->m_job could get assigned. setLoadingDeferred was abused a bit but the other options were not that better.
> > 
> > As the queue is owned by the QNetworkReplyHandler object an option would be to add a 'synchronous' flag to the class QNetworkReplyHandlerCallQueue and make it resume synchronously when needed.
> 
> Luiz, do u agree with Benjamin that we should explicitly flush the reply handler in the synchronous case?

Sure. I do. But I am not sure about what would be the API.

I think that we could have different classes for the synchronous and asynchronous cases. Then the asynchronous one could have setLoadingDeferred and the synchronous could have its own different interface. I was thinking about it at the time that I was working in this code, but I am not sure if it's worth.

I think that we could solve this problem here without changes to the interfaces and then open new bugs for further refactorings.

> 
> The latest patch I made is using QNetworkReplyHandler's 'synchronous' flag to determine whether it should resume synchronously. And I have run the layout test on it and seems it works fine.

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