[webkit-reviews] review denied: [Bug 48719] HTML5 TreeBuilder regressed a Peacekeeper DOM test by 25% (was 40%) : [Attachment 83022] Patch

Thu Feb 24 22:55:24 PST 2011

Eric Seidel <eric at webkit.org> has denied Andy Estes <aestes at apple.com>'s
request for review:
Bug 48719: HTML5 TreeBuilder regressed a Peacekeeper DOM test by 25% (was 40%)
https://bugs.webkit.org/show_bug.cgi?id=48719

Attachment 83022: Patch
https://bugs.webkit.org/attachment.cgi?id=83022&action=review

------- Additional Comments from Eric Seidel <eric at webkit.org>
I don't like the solution, so I've been working on better ones.  As seen in bug
55011, bug 55091, bug 55005, bug 55204 and bug 55127.

If I can't make up the difference in other ways then I'll look more seriously
at Andy's proposal again.

Andy'r proposal is narrow:
- It removes a single malloc (creation of the <html> element) and makes us not
have to move nodes (which is needlessly expensive due to Node::willRemove() and
some other related junk).
I also think it reduces hackability of the parser some.

I'm not over-my-dead-body opposed to his change as-written, but I've spent
yesterday and today and plan to spend tomorrow working on better, more broadly
applicable solutions.