Font and Size controls are not working correctly

[Senthil]

Replication steps:

1. Select a paragraph of text from the editor area.
   
2. Use the Size control to make the text larger (select 16 in the control). Notice that size 16 indicated in the control.
   
3. Use the Size control and select size 16 again. The text size becomes smaller again. Notice that no size is indicated in the control.
   

Same thing is happening with Font control also.

[Garry Yao]

This is related to #3351.

[Garry Yao]

The Font/Size combo is designed to be toggle-able, not sure if something else is not working as expected besides the marking problem at #3351.

[Martin Kou]

The font/size toggle is actually a feature rather than a bug - I made a mistake when explaining it to Senthil yesterday because I didn't see the toggle highlight.

So closing this bug as invalid.

[Martin Kou]

Just talked with Senthil today, the combos are still not working correctly even when taking into account the toggle behavior. To reproduce the bug:

1. Open replacebyclass.html in IE.
2. Select "some sample text".
3. Change the font to "Comic Sans MS".
4. Change the font to "Courier New"
5. Now we end up with both Comic Sans MS and Courier New mixed up in the selected text.

A similar bug also happens with the font size combo.

[Martin Kou]

Ok.. took me a whole day to figure this out. It's caused by a rather hideous browser bug in IE.

Turns out the (sibling.$.offsetWidth > 0) check we use in CKEDITOR.dom.range::enlarge() doesn't really filter out the bookmark nodes we're adding to the document. IE is reporting an offsetWidth of 3 even though the bookmark nodes have display: none in their CSS styles. This caused the enlarge() function to ignore the existing <span> style tags when applying font and font sizes.

The solution is to add a check for display: none in the element's computed style.

[Garry Yao]

Replying to martinkou: This' great to see yet another bookmark ignorance problem be dug out, but I wonder if we can use:

sibling.$.offsetWidth > 0 && !sibling.getAttribute( '_fck_bookmark' )

Instead of

sibling.$.offsetWidth > 0 && sibling.getComputedStyle( 'display' ) != 'none'

I'm thinking of the getComputedStyle slowness when invoking on large amounts.

[Martin Kou]

Yes, that's a good idea. It shouldn't happen often that we have display: none elements.

[Martin Kou]

Fixed with [3408].

Click here for more info about our SVN system.