[Webkit-unassigned] [Bug 43276] Implements IDBKeyPath extractor.

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=43276





--- Comment #18 from hans at chromium.org  2010-08-10 10:17:34 PST ---
(In reply to comment #16)
> Created an attachment (id=64017)
 --> (https://bugs.webkit.org/attachment.cgi?id=64017) [details]
> Patch

I'm only commenting on the WebPrivateOwnPtr part.

In WebKit/chromium/public/WebPrivateOwnPtr.h:
+#if WEBKIT_IMPLEMENTATION
+#include <wtf/OwnPtr.h>
+#include <wtf/PassOwnPtr.h>
+#endif

Those classes don't seem to be used.

The class should probably be Noncopyable, so hide copy ctor and assignment operator in the private section.

When WEBKIT_IMPLEMENTATION is not defined, code should probably not be allowed to construct objects of the class. To prevent this, maybe the WebPrivateOwnPtr() contructor should be declared in the private: section.

+    void reset()
+    {
+        // we own m_ptr.
+        delete m_ptr;
+        m_ptr = 0;
+    }
There is a subtle point to not defining reset() inline. If it is inline, then T must be defined when WebPrivatePtr<T> is instantiated, because it will try to bind the "delete m_ptr" call. If reset() is not inline, it's enough to have T forward-declared when instantiating the template, which is a good thing. OwnPtr does this too, for the same reason AFAIK.

Also, I thought you would do this in Bug 43709. If not, at least please update that when this goes in.

This is definitely a nice class to have, I think :)

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.