Ted pointed out that I should implement a fake remote window proxy, akin to the fake window proxy, so that the transplanting changes I'm making can be fuzzed.

I guess the question here is what needs to be tested on the SpiderMonkey side. Most of the trickiness of bug 1510760 lives inside the Gecko callbacks.

Here's what the patch for that does to js/src:

• Transplant no longer requires same class for both objects.
• Can end up with origobj not being a CCW.
• In RemapWrapper, if rewrap() returns a new object, call into remapSwap() with it.
• Also in RW, the object it returns doesn't have to be a CCW, but if it isn't, it must be a DOMRemoteProxyHandler. That's more of a constraint on the callback.

I guess it also might make sense for the shell implementation to more closely mimic what Gecko does, so we can see if SpiderMonkey deals with it correctly.