|Summary:||The new tree builder needs to call attach() on elements it attaches to the DOM|
|Product:||WebKit||Reporter:||Adam Barth <abarth>|
|Component:||New Bugs||Assignee:||Adam Barth <abarth>|
|Severity:||Normal||CC:||commit-queue, eric, ggaren, hyatt, jamesr|
|Version:||528+ (Nightly build)|
|OS:||OS X 10.5|
|Bug Depends on:|
Description Adam Barth 2010-06-28 12:27:39 PDT
The new tree builder needs to call attach() on elements it attaches to the DOM
Comment 2 Eric Seidel (no email) 2010-06-28 12:42:50 PDT
Comment on attachment 59923 [details] Patch Seems like some of this basic stuff we could steal from the odl tree builder.
Comment 3 Adam Barth 2010-06-28 12:47:12 PDT
Comment on attachment 59923 [details] Patch We'd have to refactor the old tree builder. This stuff is in big monolithic functions with manual ref counting...
Comment 4 Geoffrey Garen 2010-06-28 12:52:03 PDT
It would really be more efficient to build the whole tree and then attach() it from its root. That's a pretty big refactoring task, but if you're rewriting tree building anyway, maybe it's natural to make that change now.
Comment 5 Adam Barth 2010-06-28 12:59:08 PDT
Oh, we can do that. You just call attach once at the root and it does all the children for you?
Comment 6 Eric Seidel (no email) 2010-06-28 13:03:33 PDT
(In reply to comment #3) > (From update of attachment 59923 [details]) > We'd have to refactor the old tree builder. This stuff is in big monolithic functions with manual ref counting... Totally doable. I wonder how many assumptions in other parts of WebCore that could break.
Comment 7 WebKit Commit Bot 2010-06-28 13:05:40 PDT
Comment on attachment 59923 [details] Patch Clearing flags on attachment: 59923 Committed r62028: <http://trac.webkit.org/changeset/62028>
Comment 8 WebKit Commit Bot 2010-06-28 13:05:45 PDT
All reviewed patches have been landed. Closing bug.
Comment 9 Geoffrey Garen 2010-06-28 13:11:58 PDT
Comment 10 Eric Seidel (no email) 2010-06-28 13:20:48 PDT
(In reply to comment #4) > It would really be more efficient to build the whole tree and then attach() it from its root. I believe you. But could you clarify what aspects would make it more efficient? Would that cut out layouts/style resolves?
Comment 11 Eric Seidel (no email) 2010-06-28 13:21:09 PDT
Thank you for bringing this up Geoff. This is exactly the time to do so!
Comment 12 James Robinson 2010-06-28 13:41:00 PDT
Comment 13 Geoffrey Garen 2010-06-28 13:57:43 PDT
Comment 14 Adam Barth 2010-06-28 14:21:29 PDT
Can one of you expert-types file a new bug about this and marking it as blocking 41123? We have a good handle on when scripts execute in the new design. I'd rather not do this until we're able to actually run the LayoutTests so we can get some sort of handle on what we're breaking.