Currently we don't have a test stressing only detaching. When tested, it is usually done in combination with other phases such as layout without thinking about the interactions. This is important as our current detaching algorithm is inefficient, leading to several traversal to the root as we generate multiple repaints. See bug 92697 for example where Eric came to the same conclusion. I would like to change how we detach thus the need for some performance coverage.
Created attachment 166923 [details] Proposed test, welcome suggestions on how to better stress detaching.
Comment on attachment 166923 [details] Proposed test, welcome suggestions on how to better stress detaching. SGTM. Might make more sense to add this with your first perf fix though. :)
Comment on attachment 166923 [details] Proposed test, welcome suggestions on how to better stress detaching. > SGTM. Might make more sense to add this with your first perf fix though. :) Thanks Eric, the idea is to see the improvements as we go. Also it increases our coverage regardless of whether we follow the optimizations to their end.
Comment on attachment 166923 [details] Proposed test, welcome suggestions on how to better stress detaching. Clearing flags on attachment: 166923 Committed r130341: <http://trac.webkit.org/changeset/130341>
All reviewed patches have been landed. Closing bug.