586168 - can't implement Function.caller with function proxies

Status: RESOLVED DUPLICATE of bug 892903

(Core :: JavaScript Engine, defect)

Description

[Dave Herman [:dherman]]

I can't see any way to write a function proxy that implements Function.caller (docs at attached URL). I tried the following:

    function call() { saved = fproxy.caller }
    function trap(receiver, name) {
        if (name === "caller")
            return trap.caller;
    }
    fproxy = Proxy.createFunction({ get: trap }, call, call)
    function main() { fproxy() }
    main()

If you read the saved global, you get the call function. If you change trap to return trap.caller.caller and rerun the program, the saved global is null.

Dave

Comment 1

[Jeff Walden [:Waldo]]

Function.caller is poison in ES5 (which can easily be emulated); isn't not implementing it desirable, even a feature perhaps?

Comment 2

[Andreas Gal :gal]

We already censor .caller across wrappers, for what its worth.

Comment 3

[Brendan Eich [:brendan]]

Narcissus may want to implement guest .caller but it won't be based on the host .caller I hope.

Re: comment 1, .caller is poison only in ES5 strict, not ES5 non-strict, but of course ES5 strict is what Harmony is based on, so the deck is stacked.

The main interest in supporting .caller in Narcissus is to pass most of the tests that SpiderMonkey and Rhino pass.

/be

Comment 4

[Jeff Walden [:Waldo]]

(In reply to comment #3)
> Re: comment 1, .caller is poison only in ES5 strict, not ES5 non-strict, but of
> course ES5 strict is what Harmony is based on, so the deck is stacked.

Oops, yes, I was imprecise.  What I meant is that, in strict mode and where new functions can be created, fun.caller is poison: fun.bind().caller is poison, for example.

Comment 5

[Brendan Eich [:brendan]]

Is this bug WONTFIX? Those tests should be excluded from Narcissus test runs, as a badge of honor :-|.

/be

Comment 6

[André Bargull [:anba]]

Proxy.createFunction() was removed in bug 892903.