worker.onmessage is seeing non-empty event.origin

[pshaughn]

This happens in WPT workers/interfaces/DedicatedWorkerGlobalScope/postMessage/message-event.html. I don't see anything explicit in the standard saying to set origin to an empty string here, but I'm also not seeing anything in the worker or message port steps saying to set it to a real origin, so the implication must be to leave it default, and browsers are passing this test.

[kunalmohan]

Hi! I want to try and solve this issue, if it's not a problem.
I see that the function to be edited is this:

servo/components/script/dom/messageevent.rs

Lines 78 to 85 in 7281336

	pub fn new_initialized(
	global: &GlobalScope,
	data: HandleValue,
	origin: DOMString,
	source: Option<&WindowProxyOrMessagePortOrServiceWorker>,
	lastEventId: DOMString,
	ports: Vec<DomRoot<MessagePort>>,
	) -> DomRoot<MessageEvent> {

Should I completely remove the origin parameter it is accepting here (and make the required changes in all of its calls) or should it be left unutilized?

[pshaughn]

MessageEvent is used in many different ways; some of them are specified as explicitly setting origin, and some of them aren't. If you click on the bolded word "origin" in https://html.spec.whatwg.org/multipage/comms.html#dom-messageevent-origin you should get a popup of all the places the HTML specification references it.

[kunalmohan]

Oh, my mistake. I was not aware of all the different specs that it supported. It will take me some time, but I'll try to understand the usage (especially the posting messages section) and the serialization of origin, and will update my status soon.

[jdm]

servo/components/script/dom/messageevent.rs

Lines 148 to 162 in 7281336

	let messageevent = MessageEvent::new(
	scope,
	atom!("message"),
	false,
	false,
	message,
	DOMString::from(origin.unwrap_or("")),
	source
	.map(|source| {
	WindowProxyOrMessagePortOrServiceWorker::WindowProxy(DomRoot::from_ref(source))
	})
	.as_ref(),
	DOMString::new(),
	ports,
	);

is the code that creates the MessageEvent that is used for worker.onmessage. Looks like we're explicitly passing a non-empty origin value there; we should always pass an empty string instead.

[jdm]

That was the wrong code link - I meant to link

servo/components/script/dom/worker.rs

Lines 158 to 166 in 7281336

	if let Ok(ports) = structuredclone::read(&global, data, message.handle_mut()) {
	MessageEvent::dispatch_jsval(
	target,
	&global,
	message.handle(),
	Some(&origin),
	None,
	ports,
	);

, where we should pass a None value instead of the origin.

[jdm]

In fact, we should remove the origin argument from Worker::handle_message entirely.

[kunalmohan]

Thanks for pointing that out. That will save a lot of time :)

Just out of curiosity, while I was looking around, I found this:

servo/components/url/origin.rs

Lines 79 to 82 in d47f0ad

	pub fn ascii_serialization(&self) -> String {
	self.clone().into_url_origin().ascii_serialization()
	}
	}

I cannot understand how this works. The recursive call (I could not find any other declaration of ascii_serialization, so I figured it must be recursive) is confusing me a bit. Can you explain this briefly?

[jdm]

into_url_origin returns a url::Origin, which comes from the rust-url crate (which is different than components/url).

[kunalmohan]

Oh, I see. Now I get it. Thanks:)