[CS2] Destructuring invalid LHS generates invalid JS

[connec]

See http://coffeescript.org/v2/#try:%7B%20%7Bb...%7D...%20%7D%20%3D%20c

{ {b...}... } = c

Compiles to:

({} = c);
Object.assign({}, b) = objectWithoutKeys(c, []);

I suspect we're missing something when verifying that an object is a valid LHS. Specifically, when a splat is present it's name must be a valid LHS.

[GeoffreyBooth]

@zdenko?

[connec]

I ended up fixing this whilst playing with an alternative fix for #4651 - 2...connec:fix-4651-alt

It will need rebased once #4652 is resolved.

Whoops, never mind - I was looking at a different example.

{ a: {b...} = {c...} } = z

[GeoffreyBooth]

@connec It’s actually a little worse than your initial comment. Look at the var declarations line:

var Object.assign({}, b), b,
  objectWithoutKeys = function(o, ks) { var res = {}; for (var k in o) ([].indexOf.call(ks, k) < 0 && {}.hasOwnProperty.call(o, k)) && (res[k] = o[k]); return res; };

({} = c);
Object.assign({}, b) = objectWithoutKeys(c, []);

var Object.assign({}, b), b?

[helixbass]

Should be fixed by #4683. See its description, I think @connec's example here actually should compile, as eg ({...{...b}} = c) seems to be valid ES6 (works in Chrome console)

[GeoffreyBooth]

Fixed by #4683.