Tom S [:evilpie] Assignee
	Description • 10 years ago

This is special case that allows going through a Proxy (ScriptedDirectProxy in our case) to check for arrays. Used in some places like Array.isArray, JSON.stringify etc.

	Tom S [:evilpie] Assignee
	Comment 2 • 10 years ago

Missing tests. The changes to JSON.stringify aren't really useful for proxies until bug 895223 is fixed.

	ziyunfei
	Comment 3 • 10 years ago

Comment on attachment 8539530 [details] [diff] [review]
is-array

Review of attachment 8539530 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/jsobj.cpp
@@ +3650,5 @@
> +    if (obj->is<ProxyObject>()) {
> +        if (obj->as<ProxyObject>().handler()->isES6Proxy()) {
> +            RootedObject target(cx, obj->as<ProxyObject>().target());
> +            if (!target) {
> +                JS_ReportErrorNumber(cx, js_GetErrorMessage, nullptr, JSMSG_PROXY_REVOKED);

Rev 30 has changed this to return false instead of throwing an exception.

	Tom S [:evilpie] Assignee
	Comment 4 • 10 years ago

Awesome \o/. Thank you ziyunfei! This means we can actually simplify this function again, because we don't really need the context parameter and can just return true/false.

There is however still one problem I am aware of. Something like this is currently not working (actually after fixing bug 1115361): Array.isArray(new (newGlobal().Proxy([], {})). We could fix this by introducing a new ESClass_Array value that follows the semantics of IsArray.

	Tom S [:evilpie] Assignee
	Comment 5 • 9 years ago

I fixed this, but we need to land some tests.

	Tom S [:evilpie] Assignee
	Comment 7 • 9 years ago

We will add some test for JSON.stringify in bug 1111604.

	Jeff Walden [:Waldo]
	Comment 8 • 9 years ago

Comment on attachment 8556106 [details] [diff] [review]
Some tests

Review of attachment 8556106 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/tests/ecma_6/Array/isArray.js
@@ +18,5 @@
> +}
> +
> +revocable.revoke();
> +assertEq(Array.isArray(revocable.proxy), false);
> +assertEq(Array.isArray(proxy), false);

Bah.  I am not really a fan of these semantics.  It seems to me that touching a revoked proxy or examining anything about it, at all, should always throw.  Just as Object.isExtensible(revocable.proxy) throws, rather than returning true or false.  What do you think about getting the spec changed on the question?

Fine to land what's here for now, tho, if/until we get that changed.

@@ +23,5 @@
> +
> +var global = newGlobal();
> +var array = global.Array();
> +assertEq(Array.isArray(array), true);
> +assertEq(Array.isArray(new Proxy(array, {})), true);

assertEq(Array.isArray(new global.Proxy(array, {})), true);

	Tom S [:evilpie] Assignee
	Comment 9 • 9 years ago

Revoked proxies actually used to throw. Waldo said he would open a bug.
https://hg.mozilla.org/integration/mozilla-inbound/rev/3538586606e4

	Carsten Book [:Tomcat]
	Comment 10 • 9 years ago

https://hg.mozilla.org/mozilla-central/rev/3538586606e4

	Jeff Walden [:Waldo]
	Comment 11 • 9 years ago

Filed https://bugs.ecmascript.org/show_bug.cgi?id=3840 for the spec issue, which applies to various other predicates in the spec as well.