Support iframe viewport tracking

[samouri]

summary Implements iframe viewport tracking for the polyfill, as described in https://github.com/w3c/IntersectionObserver/issues/372#issuecomment-575287674.

The following tasks have been completed:

• [ ] Modified Web platform tests (link to pull request)

Implementation already completed:

• [x] WebKit (https://bugs.webkit.org/show_bug.cgi?id=208047)
• [x] Chromium (https://bugs.chromium.org/p/chromium/issues/detail?id=1015183)
• [x] Gecko (https://bugzilla.mozilla.org/show_bug.cgi?id=1617154)

cc @dvoytenko

[samouri]

open question: do we want to do feature detection for iframe viewport tracking and replace native with the polyfill if support is missing?

I'm leaning towards no and allowing consumers to detect it themselves. If they prefer the polyfill they can run delete InOb from window before running the polyfill.

[dvoytenko]

I think this is really close now.

[samouri]

Darn. I believe I found a webkit bug.

Essentially, Chrome and FF ensure that rootBounds when root:document are relative to the specified document (top is always 0). Safari gives the top based on current scroll position within the iframe.

See: https://output.jsbin.com/jutegazuxe

[samouri]

Current status:

• Using the polyfill: all tests pass in IE11, FF, Chrome, and Safari.
• Using native InOb:
  • FF: tests fail. I'm seeing the same behavior on master so I've ignored this.
  • Safari: A single newly introduced test fails. I've opened a webkit bug since I believe it is an error there: https://bugs.webkit.org/show_bug.cgi?id=219495
  • Chrome: A different newly introduced test fails. Chrome is allowing for cross-document tracking which is explicitly against spec right now. See https://bugs.chromium.org/p/chromium/issues/detail?id=1149713#c4

---

@dvoytenko: PTAL

[dvoytenko]

• Chrome: A different newly introduced test fails. Chrome is allowing for cross-document tracking which is explicitly against spec right now.

Can we link the crbug issue here as well?