diff options
| author | Feng Qian <fqian@google.com> | 2009-05-05 11:52:10 -0700 |
|---|---|---|
| committer | Feng Qian <fqian@google.com> | 2009-05-06 11:29:33 -0700 |
| commit | e8f342c7034cddf274ba5b9e18c51dd3e29c7db9 (patch) | |
| tree | ab0a7960fef47b124d786bb5c29c4fbd30882244 /V8Binding/v8 | |
| parent | 8d56b03700596845df1a6ba952aa7ad4b1cdccca (diff) | |
| download | external_webkit-e8f342c7034cddf274ba5b9e18c51dd3e29c7db9.zip external_webkit-e8f342c7034cddf274ba5b9e18c51dd3e29c7db9.tar.gz external_webkit-e8f342c7034cddf274ba5b9e18c51dd3e29c7db9.tar.bz2 | |
Fix memory leaks in Java binding and some code cleanup.
WebCoreFrameBridge.cpp needs to release NPObject to get proper ref count
because the way V8 binding works.
JObjectWraper holds a weak reference to the Java object to break cycles.
Also fix ScriptController::haveWindowShell to match KJS build, this should be
upstreamed to Chrome project (I am going to do it).
The binding does not use weak reference to hold Java objects, currently
it does not create cycles between Java and JavaScript objects. I didn't see
it is an issue here. A future work is to add logging/debugging code to
make sure global Java references are dropped.
Diffstat (limited to 'V8Binding/v8')
| -rw-r--r-- | V8Binding/v8/ScriptController.h | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/V8Binding/v8/ScriptController.h b/V8Binding/v8/ScriptController.h index d26dbcc..40534fb 100644 --- a/V8Binding/v8/ScriptController.h +++ b/V8Binding/v8/ScriptController.h @@ -164,7 +164,7 @@ public: // JSC has a WindowShell object, but for V8, the ScriptController // is the WindowShell. - bool haveWindowShell() const { return true; } + bool haveWindowShell() const { return m_proxy->ContextInitialized(); } // Masquerade 'this' as the windowShell. // This is a bit of a hack, but provides reasonable compatibility |