selectionStart/End/Direction on input and textarea should not be nullable

[foolip]

https://html.spec.whatwg.org/multipage/forms.html#the-input-element
https://html.spec.whatwg.org/multipage/forms.html#the-textarea-element

For both elements, the spec currently says:

  attribute unsigned long? selectionStart;
  attribute unsigned long? selectionEnd;
  attribute DOMString? selectionDirection;

However, the prose never return null:
https://html.spec.whatwg.org/multipage/forms.html#dom-textarea/input-selectionstart
https://html.spec.whatwg.org/multipage/forms.html#dom-textarea/input-selectionend
https://html.spec.whatwg.org/multipage/forms.html#dom-textarea/input-selectiondirection

http://codepen.io/anon/pen/VmEgww is a test case for this. Chrome, Edge, Firefox and Safari all return 0 as the intial values for selectionStart and selectionEnd

There's some interesting variation for selectionDirection, but it's never null.

Spec change is trivial, but need to update web-platform-tests.

Found by @LoonyBean while investigating why this is nullable in spec but not in Blink.

[domenic]

They do return null:

For input elements, calling these methods while they don't apply, and setting these attributes while they don't apply, must throw an "InvalidStateError" DOMException; and getting these attributes while they don't apply must return null. Otherwise, they must act as described below

There are previous bugs on this that changed it from throwing to returning null. Web platform tests were even updated :)

[foolip]

Oh, so I started by testing textarea, and didn't think about <input type=color> and such. I take that for textarea they actually can't ever return null, and we could change the IDL to reflect that?

[foolip]

Previous changes were in #1006

[annevk]

So we have a section "APIs for the text control selections" where there is an IDL fragment that lists some things shared between input and textarea elements. This change breaks that snippet as they no longer match. How do you want to fix that?

[foolip]

Sorry, I didn't notice that. Would it suffice to append // nullable only for input for the three attributes?

[annevk]

Maybe we should just remove the extract?

[zcorpan]

Removing SGTM.