Inserting Page break is introducing new paragraph

[Garry Yao]

Reproducing Procedures

1. Open the replace by class example page in FF;
2. Click on 'New Page' to empty document content;
3. Click on 'Insert Page Break for Printing' to insert a page break;
4. Click on the newly introduced page break fake element;
   • Expected Result: A new paragraph is created before it.

     <div style="page-break-after: always;">
     		<span style="display: none;">&nbsp;</span></div>
     

   • Actual Result: Javascript error thrown.

     <p>
        <br /></p>
     <p>
     <div style="page-break-after: always;">
     		<span style="display: none;">&nbsp;</span></div>
     </p>
     
     

[Garry Yao]

Oops, fix wrong description.

[Garry Yao]

This's supposed to be fixed by #3657.

[Garry Yao]

After some investigation, the fixing is not only limited to #3657, here I'm proposing a way of avoiding those fake elements been incorrectly positioned due to 'fixForBody' option of the parser.

[Garry Yao]

Though the patch works, the parser should have no knowledge of the fake object system, so the transformation should go into fakeobject plugin instead.

[Garry Yao]

Updates contained in the new patch:

1.The parser now is having a duo-pass parsing on the output html, which make fake elements have no more incorrect document position:

• The first pass for running the htmlfilter through, without formatter;
• The second pass for fixBody and also formatter, without any filtering; 1.The 'Page Break' plugin is now simply 'inserting' the fake element into the document, without worry about schema validation, which will be then taken care by the output parser.

[Garry Yao]

This patch's having a dependency on the latest patch from #3609 until that's get fixed.

[Frederico Caldeira Knabben]

Double parsing is not an option. A different solution needs to be found for it.

[Garry Yao]

So I would like to revert to the previous patch.

[Frederico Caldeira Knabben]

Please add the changelog entry when committing.

[Garry Yao]

The correct way of retrieving real element name of fakeelement should be the following instead:

 element.attributes[ '_cke_real_element_type' ]

[Frederico Caldeira Knabben]

Oh yes, that's much better.

[Garry Yao]

Fixed with [3859].