[V8] REGRESSION(94783): calling the binding script during V8 context creation slows down page loads

• Component: WebGL
• Status: RESOLVED FIXED

[Ulan Degenbaev]

See http://code.google.com/p/chromium/issues/detail?id=96354 .

A possible fix is to call the script lazily on the first run of the 'set()' method of a typed array.

[Ulan Degenbaev]

Created attachment 107631 [details]
Call the binding script lazily

[Kenneth Russell]

Comment on attachment 107631 [details]
Call the binding script lazily

View in context: https://bugs.webkit.org/attachment.cgi?id=107631&action=review

This looks good overall, thanks. I have one comment about the robustness of the code, but because this is such a high priority issue in Chromium I'm going to mark this r+ / cq+ and we can deal with the robustness issue later.

> Source/WebCore/bindings/v8/custom/V8ArrayBufferViewCustom.cpp:42
> +    v8::Handle<v8::Object> arrayBufferView = prototype->GetPrototype().As<v8::Object>();

Is there a more robust way to either find the ArrayBufferView prototype or check that we've found it with this walk? In a future version of the typed array spec a subtype of Uint8Array may be added (see http://www.khronos.org/registry/typedarray/specs/latest/#7.1 ) and this code would break if so.

[WebKit Review Bot]

Comment on attachment 107631 [details]
Call the binding script lazily

Clearing flags on attachment: 107631

Committed r95354: <http://trac.webkit.org/changeset/95354>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.