Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Cross origin workers should fail to fetch #13671

Merged
merged 4 commits into from Oct 23, 2018

Conversation

domfarolino
Copy link
Member

/cc @domenic @bzbarsky @nhiroki

Tackle low-hanging fruit in #13426.

Prior to this PR, testing that cross-origin workers fail was pretty grim, in that the test basically broke for everyone but Firefox, who seems to be the only conforming implementation here.

It is true that cross-origin workers fail to fetch in all or most implementations, but it seems that every failure is some synchronous security error, as opposed to letting the Fetch naturally result in a failure, and going through the normal means of firing "an event named error".

This PR fixes the test for non-Firefox-like impls, and extends it to include module workers. I guess I should follow-up with vendors though, to see if they'd be willing to remove their immediate-erroring behavior in favor of letting Fetch naturally fail the request. Is it possible that vendors see cross-origin workers as a big enough security concern such that it's too big a risk to Fetch the doomed request, lest it somehow slip through the cracks and got sent? I just think its odd that a lot of impls are immediately failing cross-origin worker construction, when the spec does never mentions this. Here's where we do it in Chrome.

@domfarolino
Copy link
Member Author

Clearly the original author of this test anticipated cross-origin Worker construction to fail in some implementations, hence the try {} catch() {}. The reason this test was timing out is because the step_func_done in catch was never being invoked. I've changed this to an assert_unreached, because though this means the cross-origin worker will not be fetched, it does not error out in the way the spec mandates.

@bzbarsky
Copy link
Contributor

The original test just tested for a sync exception. The async bit was added in https://bugzilla.mozilla.org/show_bug.cgi?id=1218433 and it looks like the "dangling step_func_done" problem also affects the changes made to workers/constructors/SharedWorker/same-origin.html and workers/constructors/Worker/same-origin.html in there. Those tests are also doing the timeout thing in non-Firefox browsers, for the most part. Would you mind fixing them too while you're here?

The discussion about what the behavior should be, such as it was, was https://bugzilla.mozilla.org/show_bug.cgi?id=1218433#c16 and https://bugzilla.mozilla.org/show_bug.cgi?id=1218433#c19 and I'm not sure what the spec said at the time....

Copy link
Contributor

@bzbarsky bzbarsky left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good, but it would be good to fix the two same-origin.html tests too.

assert_true(e instanceof Event);
});
} catch (e) {
assert_unreached("Constructing a cross-origin module worker should not synchronously fail. Its error even should be fired instead.");
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

s/even/event/

@bzbarsky
Copy link
Contributor

So in terms of the spec... https://html.spec.whatwg.org/multipage/workers.html#dom-worker step 1 says:

The user agent may throw a "SecurityError" DOMException if the request violates a policy decision (e.g. if the user agent is configured to not allow the page to start dedicated workers).

which seems like the clause the other UAs are using; "policy decision" is pretty broad.

In theory, if implementing the spec word for word one would have to do a bunch of work including setting up the worker global before discovering that the load failed and erroring. In practice, of course, one can just queue a task to fire error instead of throwing and not do anything else.

@domfarolino
Copy link
Member Author

Good catch @bzbarsky. I've updated the tests to still pass in the case where cross-origin workers fail to construct, given the spec text you found. I've also fixed the {Worker, SharedWorker}/same-origin.html tests. PTAL

try {
var worker = new SharedWorker(script, '');
var worker = new Worker(script);
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I also noticed that the Worker constructor test was still testing the SharedWorker constructor, probably as a result of copy paste. This meant all of these tests failed in Safari, a UA that doesn't support SharedWorkers. I've tested these updates and they look good in Chrome, Firefox, and Safari.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yikes. Good catch!

try {
var worker = new SharedWorker(script, '');
var worker = new Worker(script);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yikes. Good catch!

@bzbarsky bzbarsky merged commit f3e2b41 into master Oct 23, 2018
@domfarolino domfarolino deleted the domfarolino/cross-origin-workers branch October 23, 2018 22:52
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

4 participants