Uglification of React 16.7.0-alpha (with hooks) seems to be broken

[tshen8]

Bug report or feature request?
Bug report

Uglify version (uglifyjs -V)
uglify-js@3.4.9 (via uglify-js-webpack-plugin@2.0.1)

The following is React-DOM's 16.7.0-alpha code for pushing updates to their Component Update Queue (in React Fibre):

JavaScript input

function Gg(a, b, c, d) {
	a = {
		tag: a,
		create: b,
		destroy: c,
		inputs: d,
		next: null
	};
	null === R ? (R = {
		callbackList: null,
		lastEffect: null
	}, R.lastEffect = a.next = a) : (b = R.lastEffect, null === b ? R.lastEffect = a.next = a : (c = b.next, b.next = a, a.next = c, R.lastEffect = a));
	return a
}

The uglifyjs CLI command executed or minify() options used.
All the default options

JavaScript output or error produced.

  function Vo(e, t, n, r) {
    return e = {
      tag: e,
      create: t,
      destroy: n,
      inputs: r,
      next: null
    },
    jo.lastEffect = null === jo ? (jo = {
      callbackList: null,
      lastEffect: null
    },
    e.next = e) : null === (t = jo.lastEffect) ? e.next = e : (n = t.next,
    (t.next = e).next = n,
    e),
    e
  }

The error is trying to set lastEffect on the jo object, which has not yet been defined. Perhaps a result of UglifyJS trying to reduce the return statement?

Related Issues:
facebook/react#14014
webpack-contrib/uglifyjs-webpack-plugin#374

[kzc]

Does it work with terser?

[z4o4z]

Hi @kzc, yep, it works with terser!

[kzc]

Unable to independently confirm that. Cannot reproduce the issue in uglify-js@3.4.9 with input provided and default minify options.

[tshen8]

@z4o4z terser worked great for me, thank you!

[tshen8]

@kzc The uglification was done through uglifyjs-webpack-plugin with default minify options on the entire React-DOM 16.7.0-alpha code.

Using uglifyjs directly on just the function doesn't seem to produce this issue. Hmm, I guess it must be something else in the file...

[cherscarlett]

Have this issue @starbucks as well, will be moving to terser.

[mgol]

@alexlamsl Why did you close it?

[alexlamsl]

@mgol the non-reproducible test case looks like #3271 etc., which is fixed by #3329

[krasimir]

I can confirm that terser has no this issue.