If https://github.com/w3c/selection-api/issues/53 is accepted, then we need to handle input nodes in that way. This patch should handle using them in this way.

Updated.

Comment on attachment 8655049 [details] [diff] [review] Fire selectstart and selectionchange events on the input node when the selection in that editor changes Review of attachment 8655049 [details] [diff] [review]: ----------------------------------------------------------------- ::: dom/html/nsTextEditorState.cpp @@ +92,5 @@ > if (!mTextEditorState) { > return NS_OK; > } > > + mozilla::dom::AutoHideSelectionChanges hideSelectionChanges You don't need mozilla::dom, the translation unit is using that namespace. @@ +1250,5 @@ > // Do not initialize the editor multiple times. > return NS_OK; > } > > + mozilla::dom::AutoHideSelectionChanges hideSelectionChanges(GetConstFrameSelection()); Same. ::: layout/forms/nsTextControlFrame.cpp @@ +260,5 @@ > // Make sure that editor init doesn't do things that would kill us off > // (especially off the script blockers it'll create for its DOM mutations). > { > + nsCOMPtr<nsITextControlElement> txtCtrl = do_QueryInterface(GetContent()); > + NS_ASSERTION(txtCtrl, "Content not a text control element"); Nit: please change this to MOZ_ASSERT while you're here. @@ +264,5 @@ > + NS_ASSERTION(txtCtrl, "Content not a text control element"); > + > + // Hide selection changes during the initialization, as webpages should not > + // be aware of these initializations > + mozilla::dom::AutoHideSelectionChanges hideSelectionChanges(txtCtrl->GetConstFrameSelection()); Nit: dom:: ::: layout/generic/Selection.h @@ +324,5 @@ > bool mApplyUserSelectStyle; > + > + // Non-zero if we don't want any changes we make to the selection to be > + // visible to content. If non-zero, content won't be notified about changes. > + uint32_t mHideChanges; Hmm, how about a slightly better name such as mSelectionChangeBlockerCount, and AddSelectionChangeBlocker(), RemoveSelectionChangeBlocker() and IsBlockingSelectionChangeEvents()? @@ +354,5 @@ > +{ > +private: > + nsRefPtr<Selection> mSelection; > +public: > + explicit AutoHideSelectionChanges(const nsFrameSelection* aFrame); Please use the MFBT GuardObject.h macros for this. @@ +358,5 @@ > + explicit AutoHideSelectionChanges(const nsFrameSelection* aFrame); > + > + explicit AutoHideSelectionChanges(Selection* aSelection) > + { > + mSelection = aSelection; Nit: this goes in the initializer list. ::: layout/generic/nsSelection.cpp @@ +6030,5 @@ > + mHideChanges--; > +} > + > +bool > +Selection::IsHidingChanges() Nit: const.

This got backed out in https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&revision=1cec2d7bf26b because of TEST-UNEXPECTED-FAIL | /eventsource/eventsource-close.htm | EventSource: close(), test events - assert_unreached: Dunno what to do with message number 3 Reached unreachable code

Adding dev-doc-needed to make this explicit in the doc

> - [Pref="dom.select_events.enabled"] > - attribute EventHandler onselectionchange; I'm trying to document this (by updated the selectionchange and selectstart event documentation), but I don't see the rationale to remove Document.onselectionchange property that this patch also does. Does that mean that this property has been removed (and will therefore never reach a release version of Gecko)?

That looks like a mistake to me... I don't have time to look over it right now, but onselectionchange should definitely be available on document I would think. Leaving the ni? so that I'll get around to this next week.

I've filed bug 1231193 to fix this, thanks!