Honestly, I don't particularly care about toString, it's just the only builtin I could think off the top of my head that could be tested for this. Because right now we handle it in exec_source, which is only for interpreted functions... but it's a good test case of such behavior.

Do you think a fix like in 2/2 is more appropriate (also for quirks modes, but with null instead of undefined)? And perhaps fix whatever is causing toString to fail, which seems unrelated I guess.

That might mean moving it out of exec_source though.

Same story there, why do you think it's somehow related to detached scopes? The attached test fails with your patch.

I attached a wrong test, here is the right one.

Jacek