1260782 - constants not visible in window

Status: RESOLVED INVALID

(Core :: JavaScript Engine, defect)

Description

[Daniel Glazman (:glazou) (not active in Mozilla any more)]

While a |foo| variable is visible through a call to |window.foo|,
a |bar| const is always undefined through a call to |window.bar|.
This was not the case before the changes introduced for let and
const for ES6. Since I see nothing in ES6 that makes this mandatory
(this is more a conformance with Chrome), I'm filing this bug.

Codepen: http://codepen.io/anon/pen/RaZyMX

FWIW, a feature of BlueGriffon was completely horked because of that...

Comment 1

[Jason Orendorff [:jorendorff]]

The standard does require this. Otherwise we would never have implemented it, as it horked a ton of stuff for us throughout Firefox. Ask :shu about it sometime.

I don't blame you for being unable to find this in the standard, which is practically impenetrable even for people who have lived in there for years. But here's the summary:

*   ECMA-262 15.1.11 GlobalDeclarationInstantiation
    <https://tc39.github.io/ecma262/#sec-globaldeclarationinstantiation> specifies how global
    variable are set up; this happens just before the code of a script starts to run.
    Step 16.b.i.1 creates `const` bindings by calling CreateImmutableBinding.
    Step 18.a creates `var` bindings by calling CreateGlobalVarBinding.

*   CreateGlobalVarBinding is defined in 8.1.1.4.17
    <https://tc39.github.io/ecma262/#sec-createglobalvarbinding>. It creates a binding
    in envRec.[[ObjectRecord]] which entails defining a property on the global object.

*   CreateImmutableBinding for global environments is defined in 8.1.1.4.3
    <https://tc39.github.io/ecma262/#sec-global-environment-records-createimmutablebinding-n-s>.
    It creates an entry in envRec.[[DeclarativeRecord]], which is the "global lexical scope".
    No objects are touched.

I don't like the global lexical scope. Never did. It's awful. But, we follow the spec.

Not sure if this will help, but there is also
<https://hacks.mozilla.org/2015/07/es6-in-depth-let-and-const/>
which talks about all this nonsense in less technical terms.

Comment 2

[Daniel Glazman (:glazou) (not active in Mozilla any more)]

Jason, thanks for the detailed info, and sorry then for the spam.
If I completely understand that following the spec is needed,
please allow me to say that part of the spec is stupid and impossible
to understand from a JS author's perspective who's not used to dive
into Ecma specs. I find this totally ridiculous and would definitely
and strongly support Mozilla if it brings the issue to Ecma (or bring
it again, more probably). Anyway, thanks again.