Bugzilla

The test log has a JS stack.
```
openWindow/</window.onmessage@https://web-platform.test:8443/worklets/resources/referrer-tests.js:6:22
EventHandlerNonNull*openWindow/<@https://web-platform.test:8443/worklets/resources/referrer-tests.js:5:7
openWindow@https://web-platform.test:8443/worklets/resources/referrer-tests.js:2:10
runReferrerTest@https://web-platform.test:8443/worklets/resources/referrer-tests.js:25:10
runReferrerTests/<@https://web-platform.test:8443/worklets/resources/referrer-tests.js:176:12
```
The `onmessage` handler should not be calling an `assert_` function without using `test.step_function()`.  The simpler way, without an explicit `step_function()` is to `resolve(win, e)` from the `onmessage` handler, and then `assert_equals()` in the `onFulfilled` parameter to `then()`.

Doing that would identify which subtests are failing the assert.  I'd then edit `runReferrerTests()` so that only a single test runs, to reduce noise.

Tests that import a script from a "page" are expected to pass.  The tests that import a script from a "script" are expected to fail with the message "The operation was aborted."

I don't see where the `Object` is coming from.  Perhaps some `console.log()`s sprinkled around the `postMessage()` calls in that file and in referrer-tests.js might provide some clues.

The test doesn't use `AudioWorkletNode.port.postMessage()` but `window.postMessage()`.  The `addModule()` call requires a control message to successfully run on the MTG.

Is https://phabricator.services.mozilla.com/D55451 intended to merely remove an unused `mStartBlocking` variable?  If its previous functionality is being replaced by something else in the same patch, then can you describe in the commit message, together with any other intended logic changes, please?

The test log has a JS stack.
```
openWindow/</window.onmessage@https://web-platform.test:8443/worklets/resources/referrer-tests.js:6:22
EventHandlerNonNull*openWindow/<@https://web-platform.test:8443/worklets/resources/referrer-tests.js:5:7
openWindow@https://web-platform.test:8443/worklets/resources/referrer-tests.js:2:10
runReferrerTest@https://web-platform.test:8443/worklets/resources/referrer-tests.js:25:10
runReferrerTests/<@https://web-platform.test:8443/worklets/resources/referrer-tests.js:176:12
```
The `onmessage` handler should not be calling an `assert_` function without using `test.step_function()`.  The simpler way, without an explicit `step_function()` is to `resolve({win: win, e: e})` from the `onmessage` handler, and then `assert_equals()` in the `onFulfilled` parameter to `then()`.

Doing that would identify which subtests are failing the assert.  I'd then edit `runReferrerTests()` so that only a single test runs, to reduce noise.

Tests that import a script from a "page" are expected to pass.  The tests that import a script from a "script" are expected to fail with the message "The operation was aborted."

I don't see where the `Object` is coming from.  Perhaps some `console.log()`s sprinkled around the `postMessage()` calls in that file and in referrer-tests.js might provide some clues.

The test doesn't use `AudioWorkletNode.port.postMessage()` but `window.postMessage()`.  The `addModule()` call requires a control message to successfully run on the MTG.

Is https://phabricator.services.mozilla.com/D55451 intended to merely remove an unused `mStartBlocking` variable?  If its previous functionality is being replaced by something else in the same patch, then can you describe in the commit message, together with any other intended logic changes, please?