Bug 19377 – REGRESSION (r34109): "Inspect Element" on a node in a subframe doesn't highlight that node

• Component: Web Inspector (Deprecated)
• Status: RESOLVED FIXED

[Adam Roben (:aroben)]

Choosing "Inspect Element" from the context menu on a node in a subframe doesn't highlight that node in the Elements panel.

To repro:
1. Go to http://tivofaq.com/
2. Right-click on the "ILoveMyTiVo" link at the top of the right frame
3. Choose "Inspect Element"

Results:
The Inspector appears, but the <frameset> element is highlighted, not the link you right-clicked on.

[Adam Roben (:aroben)]

I think this was caused by r31890. What's happening is we're ending up with two JSInspectedObjectWrappers for the same JSHTMLHtmlElement. The two wrappers have different global objects. Presumably one of them is the subframe's global object and one of them is the main frame's. The effect of this is that the two wrappers fail a strict equality check (they'd fail a non-strict equality check, too).

[Adam Roben (:aroben)]

(In reply to comment #1)
> Presumably one of them is the subframe's global
> object and one of them is the main frame's.

This guess turns out to be correct.

[Adam Roben (:aroben)]

Actually, I suspect that r34109 was the cause of this bug. That change caused us to start caching wrappers per-global object. Prior to that we only had one wrapper for each JS object.

[Timothy Hatcher]

Created attachment 22430 [details]
Proposed patch

[Adam Roben (:aroben)]

Comment on attachment 22430 [details]
Proposed patch

"sameObjects" is a somewhat awkward name for an equality function. I think it needs some version of "to be" in it, like "areSameObject" or "objectsAreSame" or "objectsAreEqual". It also seems like we might as well call up to C++ to do the comparison on the unwrapped objects for real, if we're sending everything through a single bottleneck anyway.

r=me, but do consider renaming sameObjects.

[Timothy Hatcher]

Landed in r35315.