Object.freezing proxies should freeze or throw?

Claude Pache claude.pache at gmail.com
Tue Aug 9 22:19:51 UTC 2016


By reviewing more carefully the Proxy internal methods, I have found the following glitch in [[OwnPropertyKeys]]. 
https://tc39.github.io/ecma262/#sec-proxy-object-internal-methods-and-internal-slots-ownpropertykeys <https://tc39.github.io/ecma262/#sec-proxy-object-internal-methods-and-internal-slots-ownpropertykeys>

Testcase:

```js
var base = Object.freeze({x: 1})
var target = new Proxy(base, {
    ownKeys(o) {
        return ['x', 'x']
    }
})

Object.keys(target) // ['x', 'x'] // ok; according to ECMA262, Chrome, and Safari TP; buggy in Firefox

var proxy = new Proxy(target, {
    ownKeys(o) {
        return ['x']
    }
})

Object.keys(proxy) // !!! throws in ECMA262, Chrome, and Safari TP; should be fixed so that it returns ['x']
```

The issue is in step 17:

  17. Repeat, for each key that is an element of targetNonconfigurableKeys,
        a. If key is not an element of uncheckedResultKeys, throw a TypeError exception.
        b. Remove all occurrences of key from uncheckedResultKeys.

because targetNonconfigurableKeys contains a list with twice the value 'x', and, at the second iteration, a TypeError is thrown in step a.

The solution is simple: step 14 should guard against duplicate keys when constructing targetNonconfigurableKeys and targetConfigurableKeys. Additional text marked in bold:

  14. Repeat, for each element key of targetKeys,
      I. If key is not already an element of targetNonconfigurableKeys or targetConfigurableKeys, then 
            a. Let desc be ? target.[[GetOwnProperty]](key).
            b. If desc is not undefined and desc.[[Configurable]] is false, then
                  i. Append key as an element of targetNonconfigurableKeys.
            c. Else,
                ii. Append key as an element of targetConfigurableKeys.


―Claude

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.mozilla.org/pipermail/es-discuss/attachments/20160810/21c1525b/attachment-0001.html>


More information about the es-discuss mailing list