further improve 'arguments' support

[rbri]

Use the existing ScriptRuntime.typeErrorThrower(cx) instead of having a separate ThrowTypeError in Arguments.

This also fixes the setup of the ThrowTypeError returned by ScriptRuntime.typeErrorThrower(cx).

[aardvark179]

This looks good, but it feels like it's only partway there, but not quite. If you're altering ScriptRuntime.typeErrorThrower() anyway, then let's do it properly:

1. %TypeErrorThrower% is per realm, so add it to the list of things on TopLevel (effectively our realm object).
2. Change ScriptRuntime.typeError to accept a scope (the no args version can use cx.topCallScope), and get it from the TopLevel object.

I am also not quite clear on why we need another subclass of BaseError rather than just a LambdaFunction, but maybe I'm missing something.

Error handling feels like an area we should come up with a plan to tidy up. At the moment we throw internal errors and record their type based on a name, but all those constructors are per realm, so we should probably come up with a plan to fix that. Since the ScriptRuntime functions are used everywhere we'd better also do this in a way that allows us to introduces any new scope argument, or whatever, gradually, otherwise we'll have a really PR.

[rbri]

This looks good, but it feels like it's only partway there, but not quite.

@aardvark179 hehe it is a draft... ;-)

will try to finish my changes first, and then i will have a look at your suggestions. As in many other areas we have to go a long way to reach the spec

[gbrail]

This looks good to me -- thanks. I'm idly curious whether the "read only arguments" behavior could have been implemented by making the object's properties non-configurable and making the object non-extensible. Are there other JS objects that are supposed to silently ignore attempts to set properties like this one, or is arguments special?

[rbri]

I'm idly curious whether the "read only arguments" behavior could have been implemented by making the object's properties non-configurable

at least the property descriptor shows that they are configurable (see the last added test)

[gbrail]

Thanks for the progress on this!