Implement offscreenCanvas.getContext('webgl') on a worker thread

third_party/WebKit/Source/platform/heap/ThreadState.cpp

Index: third_party/WebKit/Source/platform/heap/ThreadState.cpp
diff --git a/third_party/WebKit/Source/platform/heap/ThreadState.cpp b/third_party/WebKit/Source/platform/heap/ThreadState.cpp
index 5e5bd450d6c5a777f7cc93eb5eceb087ed63c79f..333cdb7b630e1b8f25677f4885629017ece8fed0 100644
--- a/third_party/WebKit/Source/platform/heap/ThreadState.cpp
+++ b/third_party/WebKit/Source/platform/heap/ThreadState.cpp
@@ -226,6 +226,8 @@ void ThreadState::runTerminationGC()
     // pointers into the heap owned by this thread.
     m_isTerminating = true;
 
+    releaseStaticPersistentNodes();

[xidachen, 2016/05/07 17:07:49]

sof@, haraken@: without this line, the layout test hits the
ASSERT(!currentCount);

[sof, 2016/05/07 17:11:23]

Really, isn't the releaseStaticPersistentNodes() call below sufficient?

[xidachen, 2016/05/07 17:28:30]

I just tried something, without this line here, and change the while loop to be

while(currentCount != oldCount || currentCount)

in that case, we will for sure not hit that ASSERT.

WDYT?

[sof, 2016/05/07 17:33:32]

Not a good idea, as you will be perpetually stuck in the loop if there are
persistents alive that aren't released.

The loop will be executed at least once, so the claim that a test is failing
doesn't make sense to me.

+
     // Set the terminate flag on all heap pages of this thread. This is used to
     // ensure we don't trace pages on other threads that are not part of the
     // thread local GC.