[webkit-reviews] review denied: [Bug 96474] Add status events on <link rel=prerender> elements. : [Attachment 173064] Patch

[bugzilla-daemon at webkit.org]

Adam Barth <abarth at webkit.org> has denied Gavin Peters <gavinp at chromium.org>'s
request for review:
Bug 96474: Add status events on <link rel=prerender> elements.
https://bugs.webkit.org/show_bug.cgi?id=96474

Attachment 173064: Patch
https://bugs.webkit.org/attachment.cgi?id=173064&action=review

------- Additional Comments from Adam Barth <abarth at webkit.org>
View in context: https://bugs.webkit.org/attachment.cgi?id=173064&action=review


The general approach looks good.  We need some tests and for the patch to
compile.  :)

> Source/WebCore/platform/PrerenderClient.h:35
> +#if ENABLE(LINK_PRERENDER)

If you like, you can move the ENABLE(LINK_PRERENDER) inside the class
definition.  That way you can remove the ifdef in the LinkLoader parent class
definition.  The compiler will optimize away the empty base class.

> Source/WebCore/platform/chromium/Prerender.cpp:96
> +    if (m_client)

Is m_client ever 0?  It looks like you initialize it in the constructor and
never clear it.

> Source/WebCore/platform/chromium/Prerender.h:85
> +    PrerenderClient* m_client;

Is the lifetime of m_client always longer than that of this class?  This class
is RefCounted, which makes me worry that it might outlive m_client...  In fact,
it looks like the embedder can take a reference to Prerender via WebPrerender
and keep it alive for an arbitrary length of time.

We probably need some way to 0 out this pointer when the PrerenderClient dies. 
Maybe you have that already and I missed it?