Support ES2019 Function.toString() revision

[p-bakker]

See:

• https://github.com/tc39/Function-prototype-toString-revision
• https://tc39.es/Function-prototype-toString-revision/

Some examples:
Currently:

=> Math.pow
function pow() { [native code for Math.pow, arity=2] }

Expected:

=> Math.pow.toString()
function pow() { [native code] }

Currently:

=> (function(){}).bind({}).toString()
function () {
	[native code, arity=0]
}

Expected:

=> (function(){}).bind({}).toString()
function () { [native code] }

The spec revision also details that the .toString() of user-defined function inside JavaScript must return the source exactly as authored, which I'm not sure we're compliant with

[p-bakker]

The part about .toString() returning the source code as authored might be fixed by #1188

[tuchida]

It would be possible to handle this by removing the code that adds arity= to the result of Function.prototype.toString.
However, since #939(#326) is not implemented, the harness for test262 does not work.

// nativeFunctionMatcher.js
  const eatWhitespace = () => {
    while (pos < source.length) {
      const c = source[pos]; // `c` does not change. There are three similar codes.
      if (isWhitespace(c) || isNewline(c)) {
        pos += 1;
        continue;
      }

This works by replacing const with let.

Besides native code, getter/setter test was failing.

[p-bakker]

But besides the tests for this in test262 failing, code in Rhino needs to be fixed still to make it functionally work?

[tuchida]

As for arity=, the following codes needs to be removed.

rhino/rhino/src/main/java/org/mozilla/javascript/BaseFunction.java

Lines 464 to 466 in 955af15

	sb.append("[native code, arity=");
	sb.append(getArity());
	sb.append("]\n");

rhino/rhino/src/main/java/org/mozilla/javascript/IdFunctionObject.java

Lines 115 to 117 in 955af15

	sb.append(", arity=");
	sb.append(getArity());
	sb.append(justbody ? "]\n" : "] }\n");

I have not looked too closely at the getter/setter.

[tuchida]

This issue is not yet fixed.

[p-bakker]

To clarify: #1520 fixed the .toString() behaviour on user-authored JavaScript functions, but not yet on native (Java) methods

[tuchida]

What should the Java method be?

https://262.ecma-international.org/15.0/index.html#sec-function.prototype.tostring

4. If func is an Object and IsCallable(func) is true, return an implementation-defined String source code representation of func. The representation must have the syntax of a NativeFunction.

I think this sentence expects [native code].
Now Rhino returns a string like this.

js> java.math.BigInteger.valueOf.toString()
function valueOf() {/*
java.math.BigInteger valueOf(long)
*/}

In GraalJS, this is what returned.

> Java.type('java.math.BigInteger').valueOf.toString()
function valueOf() { [native code] }

[p-bakker]

[native code] is correct afaik

[rbri]

switching back to only [native code] will be great - have done some adjustments for the change in HtmlUnit to be browser compatible (HtmlUnit/htmlunit@0c18c77)

Hopefully i can remove more code and only insert the line break that ff has in the output...