Reland: WTF: Simplify RefPtr::operator bool.

third_party/WebKit/Source/wtf/PassRefPtr.h

Index: third_party/WebKit/Source/wtf/PassRefPtr.h
diff --git a/third_party/WebKit/Source/wtf/PassRefPtr.h b/third_party/WebKit/Source/wtf/PassRefPtr.h
index 48538cd788a66d22bbbe934ef426f31801288f61..ad34bcb19adef5992f6f05271fb03ca001705b93 100644
--- a/third_party/WebKit/Source/wtf/PassRefPtr.h
+++ b/third_party/WebKit/Source/wtf/PassRefPtr.h
@@ -82,10 +82,7 @@ public:
     T* operator->() const { return m_ptr; }
 
     bool operator!() const { return !m_ptr; }
-
-    // TODO(jbroman): Simplifying this in the obvious way causes a massive
-    // regression in a perf test on ARM. http://crbug.com/607208
-    explicit operator bool() const { return m_ptr ? &PassRefPtr::m_ptr : 0; }
+    explicit operator bool() const { return !!m_ptr; }

[Nico, 2016/06/23 16:02:06]

no !! pls

[jbroman, 2016/06/23 16:05:44]

Windows warns (with our flags, errors) about "return m_ptr;"; that's the reason
the previous CL was reverted.
Which pattern would you prefer? I could do "return m_ptr != nullptr;", if that's
better, but this seemed reasonably natural, comparing to the operator!
implementation above.

 
     friend PassRefPtr adoptRef<T>(T*);