[Webkit-unassigned] [Bug 59571] [JSC] Nodes in shadow DOM should not be GCed while their hosts are alive

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=59571





--- Comment #5 from Dominic Cooney <dominicc at chromium.org>  2011-04-27 15:49:36 PST ---
re: Geoffrey’s r-

My choice of words—“advisory”—was not right. I’ve done some debugging and the test was erroneously passing on Mac because as written, the shadow root wrapper was still alive in the machine thread roots.

Peturbing the test causes it drop out of the root set, for example replacing this line:

GCController.collect();

with this:

if (window.GCController)
    GCController.collect();

is enough to make the shadow root wrapper disappear from machineThreadRoots.

What’s your advice here? Don’t allocate the objects (a given), use this if test, and check in failures for Mac? What do you recommend as a robust way to make the thread roots reflect the state of variables in the script?

re: Sam’s comment—why are we doing this

Shadow roots aren’t exposed to script. But at some point down the road they should be, as part of the component model. This change is some incremental progress towards that goal. Making the change now makes sense because it means we can keep pushing on component model plumbing in layout tests, without exposing any component model API prematurely.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.