Expose Value::IsArgumentsObject in V8 API.

Expose Value::IsArgumentsObject in V8 API.

R=yangguo@chromium.org, svenpanne@chromium.org, yangguo

Committed: https://code.google.com/p/v8/source/detail?r=23105

[Yang, 2014-08-12 10:53:46]

https://codereview.chromium.org/460333002/diff/1/src/api.cc
File src/api.cc (right):

https://codereview.chromium.org/460333002/diff/1/src/api.cc#newcode2390
src/api.cc:2390: return
obj->HasSpecificClassOf(isolate->heap()->Arguments_string());
I don't think this is fool-proof. You could always just create a constructor
named "Arguments".

How about checking against the sloppy_arguments_map or the strict_arguments_map?

And I noticed how Heap::Arguments_string is not even used. We could remove that
root entry.

[aandrey, 2014-08-12 11:17:16]

https://codereview.chromium.org/460333002/diff/1/src/api.cc
File src/api.cc (right):

https://codereview.chromium.org/460333002/diff/1/src/api.cc#newcode3307
src/api.cc:3307: if
(class_name->IsOneByteEqualTo(STATIC_ASCII_VECTOR("Arguments"))) {
I think Arguments_string() should also be used here, as well as in
bootstrapper.cc line 1184.

[aandrey, 2014-08-12 11:34:12]

> How about checking against the sloppy_arguments_map or the
strict_arguments_map?

But this would catch only the Arguments object that is currently bound to
native_context, right?

[Yang, 2014-08-12 11:53:24]

On 2014/08/12 11:34:12, aandrey wrote:
> > How about checking against the sloppy_arguments_map or the
> strict_arguments_map?
> 
> But this would catch only the Arguments object that is currently bound to
> native_context, right?

Hm. You are right.

[Yang, 2014-08-12 13:44:18]

On 2014/08/12 11:53:24, Yang wrote:
> On 2014/08/12 11:34:12, aandrey wrote:
> > > How about checking against the sloppy_arguments_map or the
> > strict_arguments_map?
> > 
> > But this would catch only the Arguments object that is currently bound to
> > native_context, right?
> 
> Hm. You are right.

You can use JSObject::GetCreationContext to get the context from which you can
get those maps.

[aandrey, 2014-08-13 12:07:14]

On 2014/08/12 13:44:18, Yang wrote:
> On 2014/08/12 11:53:24, Yang wrote:
> > On 2014/08/12 11:34:12, aandrey wrote:
> > > > How about checking against the sloppy_arguments_map or the
> > > strict_arguments_map?
> > > 
> > > But this would catch only the Arguments object that is currently bound to
> > > native_context, right?
> > 
> > Hm. You are right.
> 
> You can use JSObject::GetCreationContext to get the context from which you can
> get those maps.

I tried and this doesn't work. The following fails:

  v8::Handle<Value> arguments_object = CompileRun("var out = 0; (function(){ out
= arguments; out._dummy = 42; })(1,2); out;");
  CHECK(arguments_object->IsArgumentsObject());

Why are you against HasSpecificClassOf("Arguments")?
- actually this is exactly what I need, otherwise I would have used
JSObject::class_name() directly if it was exposed
- in this way implemented many v8::Value::Is* checkers
- in this way implemented v8::Object::ObjectProtoToString() and I do want to be
consistent with it
- macro IS_ARGUMENTS(arg) = (%_ClassOf(arg) === 'Arguments');

Any other method seems to be inconsistent with the above.

[Yang, 2014-08-13 12:12:06]

On 2014/08/13 12:07:14, aandrey wrote:
> On 2014/08/12 13:44:18, Yang wrote:
> > On 2014/08/12 11:53:24, Yang wrote:
> > > On 2014/08/12 11:34:12, aandrey wrote:
> > > > > How about checking against the sloppy_arguments_map or the
> > > > strict_arguments_map?
> > > > 
> > > > But this would catch only the Arguments object that is currently bound
to
> > > > native_context, right?
> > > 
> > > Hm. You are right.
> > 
> > You can use JSObject::GetCreationContext to get the context from which you
can
> > get those maps.
> 
> I tried and this doesn't work. The following fails:
> 
>   v8::Handle<Value> arguments_object = CompileRun("var out = 0; (function(){
out
> = arguments; out._dummy = 42; })(1,2); out;");
>   CHECK(arguments_object->IsArgumentsObject());
> 
> Why are you against HasSpecificClassOf("Arguments")?
> - actually this is exactly what I need, otherwise I would have used
> JSObject::class_name() directly if it was exposed
> - in this way implemented many v8::Value::Is* checkers
> - in this way implemented v8::Object::ObjectProtoToString() and I do want to
be
> consistent with it
> - macro IS_ARGUMENTS(arg) = (%_ClassOf(arg) === 'Arguments');
> 
> Any other method seems to be inconsistent with the above.

Alright. I just had to convince myself that it's not possible to fake an
arguments object somehow. LGTM.

[aandrey, 2014-08-13 12:13:12]

Thanks. Do I also need an API owner stamp, or it can be landed?

[Yang, 2014-08-13 12:25:59]

On 2014/08/13 12:13:12, aandrey wrote:
> Thanks. Do I also need an API owner stamp, or it can be landed?


I think you can land it.

[aandrey, 2014-08-13 13:06:41]

Committed patchset #1 manually as 23105 (presubmit successful).

[jochen (gone - plz use gerrit), 2014-08-13 14:25:56]

please don't depend on this API from M38. It won't be in the V8 branch.