Use DOM's post-connection steps for <iframe> elements

[domfarolino]

In whatwg/dom#1261, the concept of DOM post-connection steps were introduced, as a checkpoint to run scripts and create browsing contexts synchronously during insertion, but after all DOM mutations are complete. After that, #10188 landed to make use of the post-connection steps for <script> elements, but no such PR was made for <iframe> elements, despite 2/3 browser engines implementing iframe insertion with post-connection steps, and no strong objections from the third (Gecko). This PR updates the iframe processing model to create its nested navigable/browsing context during the element's post-connection steps, instead of the normal insertion steps. Fixes #10828.

@smaug---- Are you on board with this? Is it fair to say there are no strong objections for using this timing for <iframe> elements, as Chromium and WebKit do?

• At least two implementers are interested (and none opposed):
  • Chromium (already implemented)
  • Safari (already implemented)
• Tests are written and can be reviewed and commented upon at:
  • https://wpt.fyi/results/dom/nodes/insertion-removing-steps/insertion-removing-steps-iframe.window.html?label=experimental&label=master&aligned
• Implementation bugs are filed:
  • Gecko: https://bugzilla.mozilla.org/show_bug.cgi?id=1907731
• Corresponding HTML AAM & ARIA in HTML issues & PRs: Seems not necessary.
• MDN issue is filed: Seems not necessary.
• The top of this comment includes a clear commit message to use.

---

/iframe-embed-object.html ( diff )

[smaug----]

I'm a bit confused. How does Gecko not have post-connection steps? I mean, conceptually scriptrunners do that.
Or has something else landed to the spec which I'm not aware of?
"2/3 browser engines implementing" hints that Gecko is somehow different, but I don't know how.

[domfarolino]

@smaug---- I’m just going off the WPT results linked above—Firefox fails the insertion test. Gecko may well be capable of implementing post-connection timing via ScriptRunners, but it seems to not use that capability for iframe insertion.

[smaug----]

Ah, but that is just about the load event, no?

[domfarolino]

Well the load event is the only way an earlier-inserted frame can observe that a later-inserted iframe’s content window is null, and has not yet been created yet, because the nested document is created during the post-connection steps, at the same timing as the load event. So it is about the load event, but also about nested document creation.

[smaug----]

One could mix script elements and iframes in the test, no? Earlier script accessing iframe. (though I think Gecko would synchronously create a content window when accessed first time.)

[domfarolino]

Right. See also: https://wpt.fyi/results/dom/nodes/insertion-removing-steps/Node-appendChild-script-and-iframe.tentative.html?label=experimental&label=master&aligned