Re: moveToWidget

On Tue, Aug 09, 2011 at 04:16:13AM +0000, corvid wrote:
> 
> It seems that things still aren't quite right.

  Yes.

> With my original example that went
> 
> <div>                                                                           
>   <a href="http://www.dillo.org">I am a link. How exciting!</a>                 
> 
>   <div>a div. Move down here and the link url is still shown.</div>             
>   Ordinary text outside of the div. Status output is blanked when moving down he
> re.                                                                             
> </div>      
> 
> the problem is now that I move up to the link, and the status output is
> shown, then I move down into the child div, and the status is correctly
> cleared, then I move up into the link again, and the status is still blank.

  The  example  shows  clearly the situation is much more complex
than  what  was  anticipated;  and  this  explains why there were
redundant  workaround-code  calls  sprinkled  in different parts,
instead of a clean solution.

  AFAIS,  the new moveToWidget() implementation has helped to get
closer to a "clean" way of handling it.

  The situation is this:

  .----------------------------.
  | TextBlock (tb1)            |
  | .-------------------.      |
  | | TextBlock (tb2)   |      |
  | |                   |      |
  | '-------------------'      |
  '----------------------------'

  When going from tb2 to tb1 with tb1 as the common ancestor, tb1
is not "enterNotified()" because it was already entered.

  Fortunately  in  the  case  of  an  elder  common ancestor, the
just-entered  textblock  gets the enterNotify() call, so it seems
adding this corner case to the dispatcher may be the solution.

  My thus patched tree works OK, so a commit will go soon.

  It  looks like only testing will tell us whether there're other
corner cases lurking around.


-- 
  Cheers
  Jorge.-