Changeset 190589 in webkit for trunk/Source/JavaScriptCore/heap/CodeBlockSet.h

Timestamp:

Oct 5, 2015, 4:31:53 PM (10 years ago)

Author:

[email protected]

Message:

Unreviewed, rolling back in r190450
​https://bugs.webkit.org/show_bug.cgi?id=149727

The cause of the leak was VM shutdown, which happens in workers.

The fix is for CodeBlockSet to participate in lastChanceToFinalize,
since it's responsible for running CodeBlock destructors.

I ran the leaks tests locally and did not see any CodeBlock-related leaks.

Restored changesets:

"CodeBlock should be a GC object"
​https://bugs.webkit.org/show_bug.cgi?id=149727
​http://trac.webkit.org/changeset/190450

File:

• trunk/Source/JavaScriptCore/heap/CodeBlockSet.h (modified) (3 diffs)

trunk/Source/JavaScriptCore/heap/CodeBlockSet.h

@@ -52 +52 @@
     CodeBlockSet();
     ~CodeBlockSet();
+
+    void lastChanceToFinalize();
     
     // Add a CodeBlock. This is only called by CodeBlock constructors.
-    void add(PassRefPtr<CodeBlock>);
+    void add(CodeBlock*);
     
-    // Clear mark bits for certain CodeBlocks depending on the type of collection.
-    void clearMarksForEdenCollection(const Vector<const JSCell*>&);
-
     // Clear all mark bits for all CodeBlocks.
     void clearMarksForFullCollection();
@@ -73 +72 @@
     void remove(CodeBlock*);
     
-    // Trace all marked code blocks. The CodeBlock is free to make use of
-    // mayBeExecuting.
-    void traceMarked(SlotVisitor&);
-
     // Add all currently executing CodeBlocks to the remembered set to be 
     // re-scanned during the next collection.
-    void rememberCurrentlyExecutingCodeBlocks(Heap*);
+    void writeBarrierCurrentlyExecutingCodeBlocks(Heap*);
 
     // Visits each CodeBlock in the heap until the visitor function returns true
@@ -102 +97 @@
 
 private:
-    void clearMarksForCodeBlocksInRememberedExecutables(const Vector<const JSCell*>&);
     void promoteYoungCodeBlocks();
 
-    // This is not a set of RefPtr<CodeBlock> because we need to be able to find
-    // arbitrary bogus pointers. I could have written a thingy that had peek types
-    // and all, but that seemed like overkill.
     HashSet<CodeBlock*> m_oldCodeBlocks;
     HashSet<CodeBlock*> m_newCodeBlocks;
-    HashSet<RefPtr<CodeBlock>> m_currentlyExecuting;
-    HashSet<RefPtr<CodeBlock>> m_remembered;
+    HashSet<CodeBlock*> m_currentlyExecuting;
 };