WebKit Bugzilla
New
Browse
Log In
×
Sign in with GitHub
or
Remember my login
Create Account
·
Forgot Password
Forgotten password account recovery
RESOLVED FIXED
Bug 51934
[Qt] QtWebKit crashes when using XMLHttpRequest from the unload event
https://bugs.webkit.org/show_bug.cgi?id=51934
Summary
[Qt] QtWebKit crashes when using XMLHttpRequest from the unload event
Yael
Reported
2011-01-05 10:52:30 PST
To reproduce the crash, open QtTestBrowser, load www.orbitz.com and close QtTestBrowser. www.orbitz.com sends XMLHTTPRequest during the unload event and that is causing the crash. callstack: 0 QObject::thread qobject.cpp 1387 0x04482945 1 WebCore::QtNetworkReplyThreadSafeProxy::QtNetworkReplyThreadSafeProxy QtNAMThreadSafeProxy.cpp 80 0x01e49fff 2 WebCore::QNetworkReplyHandler::start QNetworkReplyHandler.cpp 509 0x01e5098b 3 WebCore::QNetworkReplyHandler::QNetworkReplyHandler QNetworkReplyHandler.cpp 227 0x01e4eaa3 4 WebCore::ResourceHandle::start ResourceHandleQt.cpp 138 0x01e47809 5 WebCore::ResourceHandle::create ResourceHandle.cpp 70 0x01c512a4 6 WebCore::ResourceLoader::start ResourceLoader.cpp 163 0x01b18635 7 WebCore::ResourceLoadScheduler::servePendingRequests ResourceLoadScheduler.cpp 199 0x01b1b6c6 8 WebCore::ResourceLoadScheduler::scheduleLoad ResourceLoadScheduler.cpp 122 0x01b1b0a9 9 WebCore::ResourceLoadScheduler::scheduleSubresourceLoad ResourceLoadScheduler.cpp 90 0x01b1add6 10 WebCore::DocumentThreadableLoader::loadRequest DocumentThreadableLoader.cpp 323 0x01ae41f0 11 WebCore::DocumentThreadableLoader::DocumentThreadableLoader DocumentThreadableLoader.cpp 76 0x01ae2109 12 WebCore::DocumentThreadableLoader::create DocumentThreadableLoader.cpp 59 0x01ae185b 13 WebCore::ThreadableLoader::create ThreadableLoader.cpp 54 0x01b26734 14 WebCore::XMLHttpRequest::createRequest XMLHttpRequest.cpp 663 0x01e05335 15 WebCore::XMLHttpRequest::send XMLHttpRequest.cpp 543 0x01e04866 16 WebCore::JSXMLHttpRequest::send JSXMLHttpRequestCustom.cpp 132 0x01636136 17 WebCore::jsXMLHttpRequestPrototypeFunctionSend JSXMLHttpRequest.cpp 486 0x0152d35a 18 ?? 0 0x08047c66 19 JSC::JITCode::execute(JSC::RegisterFile*, JSC::ExecState*, JSC::JSGlobalData*) /home/yael/webkit/ws1/WebKitBuild/Debug/bin/../lib/libQtWebKit.so.4 0 0x02195520 20 JSC::Interpreter::executeCall Interpreter.cpp 849 0x0219279a 21 JSC::call CallData.cpp 38 0x021bc8d6 22 WebCore::JSMainThreadExecState::call JSMainThreadExecState.h 48 0x015eb02a 23 WebCore::JSEventListener::handleEvent JSEventListener.cpp 124 0x0161dd81 24 WebCore::EventTarget::fireEventListeners EventTarget.cpp 342 0x0180c084 25 WebCore::EventTarget::fireEventListeners EventTarget.cpp 311 0x0180bf0b 26 WebCore::DOMWindow::dispatchEvent DOMWindow.cpp 1549 0x01b612d3 27 WebCore::FrameLoader::stopLoading FrameLoader.cpp 394 0x01aec33d 28 WebCore::FrameLoader::closeURL FrameLoader.cpp 466 0x01aec794 29 WebCore::FrameLoader::detachFromParent FrameLoader.cpp 2558 0x01af669f 30 QWebPage::~QWebPage qwebpage.cpp 1942 0x01eb7ced 31 WebPage::~WebPage webpage.h 44 0x0807cacc 32 QObjectPrivate::deleteChildren qobject.cpp 1949 0x04483796 33 QWidget::~QWidget qwidget.cpp 1589 0x037dd046 34 QMainWindow::~QMainWindow qmainwindow.cpp 329 0x03cb5bc1 35 MainWindow::~MainWindow mainwindow.h 41 0x0806f09a 36 LauncherWindow::~LauncherWindow launcherwindow.cpp 59 0x08067ddb 37 qDeleteInEventHandler qobject.cpp 3980 0x04487af4 38 QObject::event qobject.cpp 1194 0x044824e2 39 QWidget::event qwidget.cpp 8646 0x037efc20 40 QMainWindow::event qmainwindow.cpp 1417 0x03cb7e03 41 QApplicationPrivate::notify_helper qapplication.cpp 4396 0x037817b6 42 QApplication::notify qapplication.cpp 4361 0x03781622 43 QCoreApplication::notifyInternal qcoreapplication.cpp 732 0x0446962f 44 QCoreApplication::sendEvent qcoreapplication.h 215 0x0806e631 45 QCoreApplicationPrivate::sendPostedEvents qcoreapplication.cpp 1373 0x0446a6f9 46 QCoreApplication::sendPostedEvents qcoreapplication.cpp 1266 0x0446a3b1 47 QCoreApplication::sendPostedEvents qcoreapplication.h 220 0x0384fea2 48 postEventSourceDispatch qeventdispatcher_glib.cpp 277 0x044a22e0 49 g_main_context_dispatch /lib/libglib-2.0.so.0 0 0x0534f5e5 50 ?? /lib/libglib-2.0.so.0 0 0x053532d8 51 g_main_context_iteration /lib/libglib-2.0.so.0 0 0x053534b8 52 QEventDispatcherGlib::processEvents qeventdispatcher_glib.cpp 415 0x044a3312 53 QGuiEventDispatcherGlib::processEvents qguieventdispatcher_glib.cpp 204 0x0385c5fe 54 QEventLoop::processEvents qeventloop.cpp 149 0x044669ef 55 QEventLoop::exec qeventloop.cpp 201 0x04466b34 56 QCoreApplication::exec qcoreapplication.cpp 1009 0x04469d21 57 QApplication::exec qapplication.cpp 3672 0x0377ec50 58 launcherMain main.cpp 41 0x080707ff 59 main main.cpp 255 0x0807293f
Attachments
Test case
(425 bytes, text/html)
2011-01-08 09:07 PST
,
Benjamin Poulain
no flags
Details
Patch
(2.67 KB, patch)
2011-01-08 10:02 PST
,
Benjamin Poulain
no flags
Details
Formatted Diff
Diff
Patch
(2.69 KB, patch)
2011-01-08 10:12 PST
,
Benjamin Poulain
no flags
Details
Formatted Diff
Diff
Show Obsolete
(1)
View All
Add attachment
proposed patch, testcase, etc.
Benjamin Poulain
Comment 1
2011-01-08 09:07:14 PST
Created
attachment 78311
[details]
Test case Reduction.
Benjamin Poulain
Comment 2
2011-01-08 10:02:10 PST
Created
attachment 78315
[details]
Patch
Benjamin Poulain
Comment 3
2011-01-08 10:02:56 PST
It is not as bad as I thought. It is just a little but in QtTestBrowser, not an issue with WebKit.
Kenneth Rohde Christiansen
Comment 4
2011-01-08 10:06:54 PST
Comment on
attachment 78315
[details]
Patch View in context:
https://bugs.webkit.org/attachment.cgi?id=78315&action=review
> Tools/ChangeLog:12 > + When accessing the network from the destructor of QWebPage, the network access manager > + was already deleted because the object WebPage was already destructed. > + > + This solve the problem by linking the lifetime of the object to the web page. The object > + is now destructed in the destructor of QObject.
This is a bit confusing... when you say "web page" it is hard to know wether you refer to QWebPage or WebPage.
> Tools/QtTestBrowser/webpage.cpp:131 > - m_qnamThread.reset(new QtNAMThread); > + if (!m_qnamThread) > + m_qnamThread = new QtNAMThread(this);
Wouldn't it be easier to keep the QScopedPointer and just check !m_quamThread->data() or so?
Benjamin Poulain
Comment 5
2011-01-08 10:12:46 PST
Created
attachment 78317
[details]
Patch
Benjamin Poulain
Comment 6
2011-01-08 10:15:29 PST
(In reply to
comment #4
)
> > + This solve the problem by linking the lifetime of the object to the web page. The object > > + is now destructed in the destructor of QObject. > > This is a bit confusing... when you say "web page" it is hard to know wether you refer to QWebPage or WebPage.
Funny, I had the feeling the changelog wasn't great but couldn't find what is the problem. Is it better now?
> > Tools/QtTestBrowser/webpage.cpp:131 > > - m_qnamThread.reset(new QtNAMThread); > > + if (!m_qnamThread) > > + m_qnamThread = new QtNAMThread(this); > > Wouldn't it be easier to keep the QScopedPointer and just check !m_quamThread->data() or so?
QScopedPointer is the problem. Because they exist in WebPage. The order of destructor is: WebPage->QWebPage->QObject. If you destroy the qnam in WebPage, you have problems when accessing it in QWebPage. I could use QWeakPointer, but I don't think that would improve anything here.
Kenneth Rohde Christiansen
Comment 7
2011-01-08 10:23:00 PST
Comment on
attachment 78317
[details]
Patch Why does QWebPage need to access it in its destructor? It seems that the nam is already set to 0 then: setNetworkAccessManager(0); in the patch
Benjamin Poulain
Comment 8
2011-01-08 10:26:41 PST
(In reply to
comment #7
)
> (From update of
attachment 78317
[details]
) > Why does QWebPage need to access it in its destructor?
Look at the backtrace from Yael or at the test case. It is simply that the page send some XHR in response to the unload event.
> It seems that the nam is already set to 0 then: setNetworkAccessManager(0); in the patch
It is set to 0 when you call WebPage::setQnamThreaded(false). I am pretty sure that cannot work but it was in the original code so I kept the logic.
Yael
Comment 9
2011-01-08 11:51:41 PST
(In reply to
comment #3
)
> It is not as bad as I thought. It is just a little but in QtTestBrowser, not an issue with WebKit.
I am glad to hear that. Thanks for fixing it anyways.
WebKit Commit Bot
Comment 10
2011-01-08 18:29:44 PST
Comment on
attachment 78317
[details]
Patch Clearing flags on attachment: 78317 Committed
r75340
: <
http://trac.webkit.org/changeset/75340
>
WebKit Commit Bot
Comment 11
2011-01-08 18:29:51 PST
All reviewed patches have been landed. Closing bug.
Note
You need to
log in
before you can comment on or make changes to this bug.
Top of Page
Format For Printing
XML
Clone This Bug