Bug 152844 – ES6: Implement lexical scoping for function definitions in strict mode

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Saam Barati]

The current flavor of programs break because of function hoisting:
```

let cond = true
if (cond) {
    let h = 20;
    function bar() { 
        return h; // ReferenceError because bar is hoisted outside of "if" statement scope
    }
    baz();
}
```

[Saam Barati]

Lets start with just implementing this in strict mode.
Then I think there is some algorithm defined in the spec
where where we can make the example listed here work
as long as we don't break backwards compatibility.

[Saam Barati]

Created attachment 273374 [details]
WIP

I think most of the code is done.
Just need to write new tests and edit old tests.

[Saam Barati]

Created attachment 273480 [details]
patch

[WebKit Commit Bot]

Attachment 273480 [details] did not pass style-queue:


ERROR: Source/JavaScriptCore/bytecompiler/BytecodeGenerator.cpp:1886:  Missing space before {  [whitespace/braces] [5]
Total errors found: 1 in 17 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Geoffrey Garen]

Comment on attachment 273480 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=273480&action=review

r=me

> Source/JavaScriptCore/bytecompiler/BytecodeGenerator.cpp:1897
> +     * We can transform block scoped function declarations in strict mode like so:

This is more like a must than a can.

> Source/JavaScriptCore/bytecompiler/BytecodeGenerator.cpp:2638
> +    HashSet<UniquedStringImpl*> functionsToRemove;

This should be a Vector with inline capacity.

[Saam Barati]

Comment on attachment 273480 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=273480&action=review

>> Source/JavaScriptCore/bytecompiler/BytecodeGenerator.cpp:1897
>> +     * We can transform block scoped function declarations in strict mode like so:
> 
> This is more like a must than a can.

Indeed. Will change.

>> Source/JavaScriptCore/bytecompiler/BytecodeGenerator.cpp:2638
>> +    HashSet<UniquedStringImpl*> functionsToRemove;
> 
> This should be a Vector with inline capacity.

Yeah of course. I don't know why I thought there could be duplicates.

[Saam Barati]

landed in:
http://trac.webkit.org/changeset/197915