Unify internal representation of empty anchors using fake objects

[Piotrek Koszuliński]

Currently we use fake objects on Webkits only. This is driven by ​http://docs.ckeditor.com/#!/api/CKEDITOR.plugins.link-property-fakeAnchor property.

Since we've got problems with DnD of empty anchors (#11140) and branched code means harder maintenance and bigger code size, we could simplify all that by using fake objects on every browser.

Possible threats

• Empty anchor represented by fake object may not look good depending on font size of a surrounding text. A sample with different text styling should be created for easy testing.
• New issues may be introduced while simplifying code. To be tested:
  • data processing,
  • creating anchor and selection after creation,
  • discovering anchors for options in link dialog (with link type set to anchor),
  • removing anchor,
  • options in context menu,
  • editing anchors (+dbclick),
  • and finally DnD.

Fixes

• #11140

[Piotrek Koszuliński]

I pushed ​t/11377 with a quick prototype. It looks that visually everything is ok. I found one issue though - on IEs anchor is not selected after it's created.

In any case - further test are needed.

[Piotrek Koszuliński]

PS. if first commit in t/11377 is not apparent you can check first to what values those three properties have been unified: ​https://github.com/cksource/ckeditor-dev/commit/098f1c48c545c7b1ab2ae42ff181bb7fd597579c#diff-d9e496f4ce47ad70bef0ba672713d3a1L265

[Marek Lewandowski]

Conflicts merged, rebased to master at ​t/11377 dev and created t/11377 at tests. Added fixes for IE selection, and small update in docs. One TC needed to be adjusted, and be aware that IE11 puts extra br in particular TC.

For this ticket i did testing in IE8 - IE11 all of them works ok with fake anchors. There were two major issues at the begining:

IE (general): When anchor with innertext were fully selected (i.e. <a name="foo">[bar]</a>, then edit dialog would not fetch its name. IE (general): After updating anchor name, caret was placed at the very beginning of the document.

[Piotrek Koszuliński]

Note that fake object wasn't used on FF too, so it should be tested as well.

[Marek Lewandowski]

My bad, i did testing for chrome, ff26, ie8-ie11. I've skipped only opera (since it's still blink), and safari.

[Marek Lewandowski]

Once again, conflicts merged and rebased to major at ​t/11377 dev and tests.

[Olek Nowodziński]

I rebased branches once again, added a single commit to dev and force-pushed all stuff to remotes.

I checked Chrome, FF, IE(8,9,11) - manually and automatically. Everything seems OK for me. If something's broken we're gonna figure it out before release. Hopefully :P

I'd squash those two commits

"Moved variable initialization."
"Code simplification."

since they don't introduce any new logic. They should be merged with corresponding ancestors before the branch is majorized.

[Marek Lewandowski]

Majorized with ​git:5875583254 at dev and 7f8d249449ec10 at tests.