Bug 41697 – [Chromium] Crash when re-entering message loop

• Component: WebKit Misc.
• Status: RESOLVED FIXED

[Alexander Pavlov (apavlov)]

ASSERT(!pageGroupLoadDeferrer) fails in void WebView::willEnterModalLoop() when re-entering message loop (e.g. when evaluating properties that require message loop to be run (e.g. document.cookie) in Web Inspector, on a breakpoint.)

[Alexander Pavlov (apavlov)]

Created attachment 60633 [details]
[PATCH] Suggested solution

Enable re-entrancy of the message loop by tracking the nesting level.

[Yury Semikhatsky]

Comment on attachment 60633 [details]
[PATCH] Suggested solution

WebKit/chromium/src/WebViewImpl.cpp:141
 +  static PageGroupLoadDeferrer* pageGroupLoadDeferrer = 0;
You don't need explicit initialization here.

WebKit/chromium/ChangeLog:5
 +          [Chromium] Crash when re-entering message loop
Please provide more detailed comment including the description of the drawbacks of this solution.

WebKit/chromium/src/WebViewImpl.cpp:191
 +          ASSERT(pageGroupLoadDeferrer);
You may want to check in debug mode that PageGroup hasn't changed since previous call to this method

WebKit/chromium/src/js/Tests.js:1862
 +      WebInspector.console.visible = true;
Use  WebInspector.showConsole instead.

WebKit/chromium/src/js/Tests.js:1896
 +          function(callFrames) {
Please use evaluateInConsole_ instead of this custom code.

WebKit/chromium/src/js/Tests.js:1872
 +                      function(commandResult) {
Use evaluateInConsole_ instead.

[Alexander Pavlov (apavlov)]

Created attachment 60736 [details]
[PATCH] Comments addressed

[Alexander Pavlov (apavlov)]

Darin, can you please have a look at this change? You seem to know those ropes...

[Darin Fisher (:fishd, Google)]

I think it would be better to create a stack of PageGroupLoadDeferrer objects.  It is safe to nest instances of those, and that's what we should actually do if we need to support multiple levels of nesting.

[Alexander Pavlov (apavlov)]

Created attachment 60752 [details]
[PATCH] Implemented a stack of PageGroupLoadDeferrers, as suggested by fishd

[Alexander Pavlov (apavlov)]

Committing to http://svn.webkit.org/repository/webkit/trunk ...
        M       WebCore/ChangeLog
        M       WebCore/page/PageGroupLoadDeferrer.cpp
        M       WebCore/page/PageGroupLoadDeferrer.h
        M       WebKit/chromium/ChangeLog
        M       WebKit/chromium/src/WebViewImpl.cpp
        M       WebKit/chromium/src/js/Tests.js
Committed r62774