Bug 87723 – CRASH() in Heap::markRoots

• Status: RESOLVED
• Resolution: DUPLICATE
• Component: JavaScriptCore

[Dongseong Hwang]

This bug maybe arose between r116279 and r118267.

Heap::markRoots can call indirectly Heap::markRoots again.
It is why CRASH() can be fired in Heap::markRoots's following code.

    if (m_operationInProgress != NoOperation)
        CRASH();


I attached an example callstack.
JSC::Heap::markRoots() at Heap.cpp:428 0x7ffff5a2fe79	
JSC::Heap::collect() at Heap.cpp:683 0x7ffff5a307b5	
JSC::CopiedSpace::getFreshBlock() at CopiedSpace.cpp:240 0x7ffff5a25944	
JSC::CopiedSpace::addNewBlock() at CopiedSpaceInlineMethods.h:107 0x7ffff5a26859	
JSC::CopiedSpace::doneCopying() at CopiedSpace.cpp:226 0x7ffff5a257ed	
JSC::Heap::markRoots() at Heap.cpp:588 0x7ffff5a303ae	
JSC::Heap::collect() at Heap.cpp:683 0x7ffff5a307b5	
JSC::Heap::reportExtraMemoryCostSlowCase() at Heap.cpp:305 0x7ffff5a2f5f6	
JSC::Heap::reportExtraMemoryCost() at Heap.h:322 0x7ffff463ffb3	
WebCore::HTMLImageLoader::notifyFinished() at HTMLImageLoader.cpp:86 0x7ffff4afa6bd	


I tested " http://www.dorothybrowser.com/test/webkitTest/imgdecode/bgimage-png/test.html " using WebKitQt.
If you go into the site and push a back button and push a go button, you can encounter CRASH().

I don't know JSC' GC, so I want for JSC experts to fix it.

[Oliver Hunt]

Is this fixed in ToT Mark?  We shouldn't be triggering GC during GC...

[Mark Hahnenberg]

(In reply to comment #1)
> Is this fixed in ToT Mark?  We shouldn't be triggering GC during GC...

I thought it was, but when I looked, bug 86538 is still waiting for a review...

[Mark Hahnenberg]

*** This bug has been marked as a duplicate of bug 86538 ***