Of course! Currently during recalc style, we both create our ComputedStyle objects and our LayoutObjects. One can imagine triggering a bunch of style recalcs without a layout, in which case the layout tree construction may be unnecessary work.
If we ended up building a hybrid Layout Tree whereby certain content uses a new system, and other parts the legacy one, we'd ideally be able to make decisions about what type of LayoutObject (new or old) to make using the entire DOM tree's ComputedStyles. We can currently only reason about our ancestors, not our descendants, during Layout Tree construction.
This would necessitate keeping track of what nodes are marked for Reattach during style resolution, an additional Lifecycle state where we do that reattach, and logic for when to run it.
My presumption is this would go to the Style team?
Update: This last patch that landed successfully pulls out all layout tree construction from style resolution. See the design document linked at the top (under Squad section 1) for more implementation details.
nainar: To help us when we review the quarterly bugs, can you briefly indicate under what conditions we should consider bringing this back up in priority? I guess linking to your retrospective with its list of future work is appropriate, but you might also have other points to add here.
Hi folks, Since there is no active work on this I am going to close this CL. It's huge anyway and any further progress should be tracked in a child CL. Will make sure to cc everyone mentioned here when I do open a new bug with further progress.
Comment 1 by dglazkov@chromium.org
, Mar 29 2016