Re: [Gnash-commit] gnash ChangeLog testsuite/misc-ming.all/key_eve...

[zou lunkai]

> > Just wondering: does it mean that in the above case the user
> > has NO way to release the resources associated with that object ?

Ah, I think I found a way to access that object now.

Test:

obj = new Object();
obj.x = 1;
obj.onKeyDown = function ()  {  _root.objRef = this };
Key.addListener(obj);
delete obj;

Observed behaviour:

after deleting obj,  if no key has been pressed, then _root.objRef is undefined;
if any key has been pressed, then we could get a new reference
'objRef', and both
objRef and objRef.x are defined.


On 7/10/07, zou lunkai <address@hidden> wrote:
> > Just wondering: does it mean that in the above case the user
> > has NO way to release the resources associated with that object ?
>
> I guess so.
>
> Test1:
>
>  obj = new Object();
>  obj.onKeyDown = function ()  { };
>  Key.addListener(obj);
>  Key.removeListener(obj);
>  delete obj;
>
> everything has been deleted as expected.
>
>
> Test2:
>
>  obj = new Object();
>  obj.onKeyDown = function ()  { };
>  Key.addListener(obj);
>  delete obj;   // onKeyDown could be invoked after this, 'obj' is undefined
>  Key.removeListener(obj); // has not effect,  onKeyDown could still be
> invoked after this.
>
>
>
> On 7/10/07, Sandro Santilli <address@hidden> wrote:
> > On Tue, Jul 10, 2007 at 11:07:00AM +0800, zou lunkai wrote:
> > > The above testcase does something like this:
> > >
> > > obj = new Object();
> > > obj.onKeyDown = function ()  { };
> > > Key.addListener(obj);
> > > delete obj;
> > >
> > > The observed behavious is:
> > > After deleting 'obj', the onKeyDown event handler could still be
> > > invoked.  So I think the created object is still alive, but without an
> > > explicit reference.
> > >
> > > I'd like to know how GC would deal with this.  Knowing this would
> > > help me find a general solution to fix bug #20317.
> >
> > Ok. Currently we do NOT keep key listeners alive, assuming they'd
> > deregister themselves on destruction. Since you're reporting this
> > is NOT the case, the movie_root::markReachableResources (or the 
> Key::markReachableResources
> > in case it relies on the existance of the global Key object) must
> > be changed to keep listeners alive (mark them as reachable).
> >
> > See also: http://www.gnashdev.org/wiki/index.php/MemoryManagement#Listeners
> >
> > Just wondering: does it mean that in the above case the user
> > has NO way to release the resources associated with that object ?
> >
> > --strk;
> >