Incorrect DCE of assignment to destructered parameter field with ES_2015 or higher

[WearyMonkey]

Code sample:

function installBaa({ obj }) {
    obj.baa = 'foo';
}

const obj = {};
installBaa({ obj });
console.log(obj.baa);

Command:

java -jar compiler.jar --js test.js -O ADVANCED --language_out ECMASCRIPT_2015

Version: v20191027

Expected output: (as output by --language_out ECMASCRIPT5)

console.log({a:"foo"}.a);

Actual output:

'use strict';console.log({}.a);

It looks like CC is losing the assignment when the target language is ECMASCRIPT_2015 or higher.

The following non-destructered parameter works as expected:

function installRoutes(x) {
    x.obj.baa = 'foo';
}

[lauraharker]

Created internal Google issue http://b/143889127

[lauraharker]

Here's a smaller repro:

const obj = {};

(function() {
  const {obj: objLocal} = {obj};
  objLocal.baa = "foo";
})();

console.log(obj.baa);

becomes with the FOLD_CONSTANTS and COMPUTE_FUNCTION_SIDE_EFFECTS flags:

const obj = {};
console.log(obj.baa);

For whatever reason the peephole passes think the IIFE is a no-op and are deleting it.

[WearyMonkey]

@lauraharker

Any update on this issue?

It forces us to introduce Babel into our pipeline to transpile spread operators before passing Closure Compiler.

[lauraharker]

Hi, I'm afraid I don't have an update :( This is on our hotlist of issues to prioritize but no one is working on it at the moment.

I'll try to take a look this week.

[WearyMonkey]

Thank you!

[lauraharker]

Update: I have a fix and am in the process of internal code review. This should be fixed in the next release.

[WearyMonkey]

@lauraharker I just tested my sample code in v20200719 and it's still not working?

Did your fix make it into the release?

[WearyMonkey]

@lauraharker it looks like your smaller repro is fixed, but not my original one.

[lauraharker]

Sorry about that, I'll reopen this issue.

link to your original repro