It's not clear in JS.md what values processing the Elements list creates during instantiation

[saambarati]

For example, lets consider the following JS program:

module = new WebAssembly.Module({
    ;; single function (at index 0)
    ;; export function at index 0 as "foo"
    ;; import a table
    ;; elements section with a single element:
    ;;      element = { offset 0, function indices: [0] }
    (all of this is in the binary of course)
});

let table = new WebAssembly.Table({element: "anyfunc", initial: 1});
let instance = new WebAssembly.Instance(module,  {import: {table}});

instance.exports.foo === table.get(0) // What is this? true or false?

• Are the Web Assembly Function objects that get stored into table supposed to be exactly the same object that gets stored into the exports dictionary?

[rossberg]

Yes, they have to be the same. The wording of the algorithm in JS.md is pretty explicit on that, note the construction and querying of the funcs list.

[saambarati]

@rossberg-chromium I'm specifically talking about the elements section. There is hardly any text inside Instance construction talking about the Element section, and it doesn't reference the issue I stated above:
https://github.com/WebAssembly/design/blob/master/JS.md#webassemblyinstance-constructor

The querying you're talking about is done for function imports/exports and table exports. It's not talked about with respect to the Elements section.

If we want the behavior you describe, JS.md should have some text about it. FWIW, I agree that this should be the specced behavior.

[rossberg]

@saambarati, I don't quite understand, your example does test function exports against table exports, so AFAICT, its behaviour is fully specified. And other than that, elements don't have any observable identity, so what is missing?

[saambarati]

@rossberg-chromium My example doesn't export a table. It imports a table and has the element section write an imported function into index zero in the table

[rossberg]

@saambarati, ah, yes, you're right, thanks! Probably the fix is to include all elements of imported tables in funcsand later run over the table to update it, independent of whether it is exported. I'll try to prepare a PR.

But to clarify the question in the abstract, the overarching idea is that every Wasm function should be wrapped into a JS function at most once.

[rossberg]

See #911.

[rossberg]

This is fixed.