Re: DnD improvements

[Adam Fedor]

Willem Rein Oudshoorn wrote:

> A cursory debugging session of _DPSsetcursor:  showed that 
>
>         generic.currentMouseWindow;
>    
> is None almost all of the time.  (That is not GOOD)        
>   
> You know probably better than I do how cursor are supposed to work,
> and I would like to fix the code.  But I do not understand at all
> what the idea is behind the current code.  


Well I did qualify my last statement about tha cursor patch :-). I think 
it is better than it was, even if it doesn't look like it. The problem 
is that OpenStep assumes cursors are global and work everywhere, while X 
cursors are only valid in the window it was set in. Trying to mesh the 
two is very difficult - particularly in DnD sessions when X events seem 
to be very counterintuitive, almost counterproductive.

What's supposed to be happening is that when you set an NScursor, the 
backend should get the window that the mouse is currently in and set the 
 X cursor in that window. Every time you move the mouse to a different 
window, you should get an EnterNotify event, and the backend should make 
sure the current cursor is set in that window as well.

Drag events are a particular problem since the mouse is over the drag 
window most of the time, but occationally gets EnterNotify events when 
it crosses into another window. I'm still not sure how to make this more 
reliable.


> > Next question is about the caching of the cursors.  
> When I implemented the code I briefly thought about caching
> them.  I did not make a cache, because in my opninion caching
> the cursors is a job of the NSCursor class.  Looking at 
> the documentation of the NSCursor class did not reveal
> any method that makes this impossible.  


That's probably a good idea. As the code was, almost every drag event 
created a new X cursor (I'm not sure how much of a drain on X resources 
that was).


> Besides that, why do you create a new `ArrowCursor'?
> The says that it is done because we color it.  But
> the coloring is done in the call
>
>         DPSsetcursorcolor (....., [newCursor _cid]);
>
> So I don't see the reason for the duplicate.  


Well that's the problem with caching cursors. When I cached the default 
arrow cursor in NSCursor, and its colored in the drag session, there's 
no cooresponding code to uncolor the arrow at the end, so often the 
default arrow cursor ends up being green for the rest of the time.

--
Adam Fedor, Digital Optics            | Fudd's law of opposition: Push
fedor@doc.com  http://www.doc.com     | something hard enough, and it
fedor@gnu.org  http://www.gnustep.org | will fall over.