Specify event.signal

[domenic]

Detailed semantics/tests to write:

• Mainline cases:
  • Doing pushState/replaceState, or anything that goes through HTML's "navigate", will synchronously cause the signal to fire abort. This is true even if you're in the middle of a navigate event handler (perhaps one of many event handlers); it doesn't wait until the end.
  • Doing HTML's "traverse the history" will in theory asynchronously cause the event to fire because of how the spec is written, but I don't think I like this very much, and all the history traversal stuff is going to be rewritten for navigate events on traversal #114 / Specify traversal and overhaul ongoing navigation tracking #109 anyway, so I dunno.
  • window.stop() will synchronously cause the signal to fire abort.
  • Clicking the stop button will cause the signal to fire abort, since that cancels ongoing navigations.
  • In all of the above cases, the abort happens synchronously and immediately, but other consequences such as navigateerror happen asynchronously at the bottom of the "inner navigate event firing algorithm"
• Degenerate cases:
  • Disconnecting an iframe during any navigate event handler will cause the signal to fire abort
  • Calling event.preventDefault() in any navigate event handler will cause the signal to fire abort
  • In both of the above cases, the abort fires after all event handlers have run (not immediately upon the action), but synchronously. Then, after a microtask, navigateerror fires on appHistory.
  • document.open() cancels navigations and should have a similar impact to window.stop(), synchronously. Although Clarify how to cancel a navigation whatwg/html#3447 makes me a bit suspicious.

---

Preview | Diff