[Webkit] Missing case for filler char in collapsed selection

[Garry Yao]

There's still some logic hole in filler char creating detection, which breaks the following TC:

​http://ckeditor.t/tt/7912/1.html

[Garry Yao]

The patch expanded the filler creation check, in considering of hidden element.

Besides two commonly requested traversal utility are proposed in range.js.

[Garry Yao]

The above patch breaks some current TCs.

Considering that it's not easy to enumerate all possible (see code comments) DOM structure where an empty fix is required, so the proposal takes a new approach - rely on range::moveToElementEditStart/End to check whether the filler is needed, and I've opened t/9386@test that cited all known cases for now.

[Frederico Caldeira Knabben]

I'm against the changes on CKEDITOR.dom.walker.invisible. They sound like a hack to achieve the fix here, at the price of making the function counterintuituve.

How should I expect an empty text node not to be invisible just because it may eventually be inside a visible element?... something like this: <p>Text<br>[Empty Text Node]</p>

[Garry Yao]

The current walker::invisible impl has taken into consideration for your cited case, a text node's visibility can be considered as determinated by it's parent.

In order to simply the review and test run, I would combine the #9384 and #9386 patches, which are considered as revalent.

[Garry Yao]

Fixed with [7624].