Range: checkStartOfBlock not forgiving bogus node

[Arne]

checkStartOfBlock returns "false" for the following case:

<p><br>^</p>

From the code perspective the cursor might not be at the start of the block, but visually it is.

Pertains to non-IE browsers.

[Garry Yao]

Looks important, but did you tracked this from some real world bug?

[Arne]

Yes. The case I had was a table cell

<td><br></td>

with the cursor blinking in the cell. Visually the cursor is both at the start and end of the cell. checkStartOfBlock reported false when I retrieved the current Range object. I wrote test cases to verify it and found that it's the same for

<p><br></p>

I have currently made a custom patch on our system which seems to work so far. Can't predict the possible side affects of such a change. But in the evaluator function getCheckStartEndBlockEvalFunction in range.js I changed the line

if ( !isStart && !CKEDITOR.env.ie && node.getName() == 'br' && !hadBr )

to

if (!CKEDITOR.env.ie && node.is('br') && !hadBr )

[Arne]

From the Range object's perspective the cursor is probably located after the <br> element which causes the evaluator function to return false...

[Wiktor Walc]

(setting version to the lowest stable of CKEditor, not sure if issue exists since 3.0 though)