Only restore popover focus if focus is in the popover

[josepharhar]

Popovers will restore focus to the element which was focued before opening the popover when the popover is closed. However, if a user opens a popover and then tab-navigates away from the popover, focus should not be changed when the popover is closed.

More context:
https://chromium-review.googlesource.com/c/chromium/src/+/4322607

Fixes #9169

• At least two implementers are interested (and none opposed):
  • Chrome
  • WebKit
• Tests are written and can be reviewed and commented upon at:
  • Only return focus if focus was inside the popover before it closed web-platform-tests/wpt#38922
• Implementation bugs are filed:
  • Chromium: N/A
  • Gecko: https://bugzilla.mozilla.org/show_bug.cgi?id=1832870
  • WebKit: N/A

(See WHATWG Working Mode: Changes for more details.)

---

/popover.html ( diff )

[mfreed7]

I can't quite tell from the spec if <span>focused area of the document</span> is retargeted to the document, or if it might be contained in a descendent shadow root. If the former, then this looks good. If the latter, then <span>descendant</span> below should be "shadow including inclusive descendant" I think.

[josepharhar]

Whether its retargeted or not, shadow-including inclusive descendant should give us the behavior we want I think so it seems like a safer option to me. I changed it to shadow-including inclusive descendant.

[nt1m]

WebKit is onboard with this change.

[smaug----]

Is this consistent with dialog?

[nt1m]

Is this consistent with dialog?

It probably makes sense to do the same change for non-modal dialogs too.

[nt1m]

Modal dialogs however make everything else inert, so it makes sense to restore focus unconditionally.

[domenic]

On the question of non-model dialogs, #8904 is related. It would be ideal if we fixed both at once.

[josepharhar]

On the question of non-model dialogs, #8904 is related. It would be ideal if we fixed both at once.

I opened a PR here for the dialog issue: #9178

[domenic]

Editorially LGTM, and it looks like we have two-implementer interest and tests, but I'm not 100% sure discussion is settled yet (including discussion in #9178) so I'll give it another day or two before merging?