Re: [html-tests] Extend Test Coverage for `top`, `parent` HTML 7.1.1 Nested Browsing Contexts (#5105)

Good morrow, @annevk ! Turns out there were some interesting things going on beyond stylistic updates (though those are done, too) with the results, and I wanted to take a deeper look at those.

I wanted to understand the failures that were introduced by my new tests (i.e. I didn't take a deep look at pre-existing failures beyond validating that my small changes to those test files didn't _change_ results).

Regarding the Firefox timeouts, this appears to be a manifestation of [this open bug](https://bugzilla.mozilla.org/show_bug.cgi?id=1229707) in which the `load` event on nested `iframe` elements does not fire if the `iframe` does not have a `src`. That's a real bug, but, after deliberation, I decided to add a `src` to the affected `iframe`s (`/common/blank.html`) in my tests and commented them as workarounds. At the time it seemed like a better situation than having most of the tests fail here because of a tangential issue. I am not sure if that's the way you'd like to see this handled, so let me know: I'm more than happy to back that out and leave the tests failing for Firefox. After that change, all of the new tests are passing in Firefox.

However, there are some interesting failures in Safari (and also latest webkit nightly). It would seem that Webkit is holding onto references of `window.parent` even after one of the nested browsing containers (i.e. either the parent or the child) involved is removed from the document. From my understanding of the spec, this should not happen, as the parent relationship with browsing contexts requires both parties involved to be connected and active—`window.parent` should return `null` here. Webkit does not display this behavior with `window.top`; those tests have results I would expect in Webkit. If my interpretation of the [spec language](https://html.spec.whatwg.org/#nested-browsing-contexts) seems off to you, let me know, but otherwise I think I'll open a bug on Webkit this morning.

Finally, note that the history of this branch has become a bit chattery with the updates. When we're getting closer to merge, I can rebase to something more manageable if you'd like!

View on GitHub: https://github.com/w3c/web-platform-tests/pull/5105#issuecomment-285680031

Received on Friday, 10 March 2017 14:19:36 UTC