External Events/DSO Open Source Offering

Stephan Rudlof sr at evolgo.de
Thu Jun 29 13:53:50 UTC 2000


"David L. Caster" wrote:
> 
> Stephan,

<snipped>

> > > Anyway, I'll trim the archive down to the really useful
> > > stuff over the next day or two.  If you have something
> > > specific that you need, let me know now because I can't
> > > promise to have a lot of attention to pay to this after
> the
> > > end of the week.
> > >
> > > David L. Caster
> >
> > It's difficult for me to say what I'm interested in,
> because I haven't
> > seen your stuff...
> >
> > The motivation to look onto it has come from the -
> currently vague idea
> > - to switch Morphic (and possibly MVC) from polling to
> real events. And
> > your work seems to go into this direction, isn't it?
> >
> 
> Yes, that is correct.  In fact, getting Squeak to the point
> where it could actually sleep (if there was nothing else to
> do) pending an OS event was one of the original goals
> prompted by the long standing need to get Squeak off of
> polling the platform.  There is more to this than meets the
> eye obviously, but yes, you are correct.
> 
> Anyway, I'll work at making the stuff a little cleaner (more
> compact) as I am able.

Could you send a post to the Squeak ML, if you are finished?
Then I would like to take a look at it ;-)

>  If you are a VM guy,

Yes.

> take a look at
> squeak27/vm/src/sqExternalMessage to get a feel for the
> hoops one has to jump through just to insure a modicum of
> confidence in proper synchronization between an external
> thread and the VM.

Here platform compatibility is an issue. Did you keep this in mind?

>  You then might want to take a look at
> (Squeak image) xmdclinux.image, and in particular the
> category External Events (which BTW is right in front of you
> when you start the image!)
> 
> C'

Currently I have the idea to look for the Morphic part of improving
event handling:
- simulating real events and
- trying to be compatible with the existing stuff;
by writing an EventSensor class, which
- uses the current polling mechanism and
- buffers simulated events to reach compatibility with polling mouse
statii in Controller classes.

I hope to solve the mouse 'event' eating problem with this approach
without the must to have real events (what would be better, of course).
Possibly it could help to have a co-existence of polling and event based
VM's at different platforms for some intermediate time.

This is a first approach (I'm not an event specialist so far), which
comes from top - the GUI - and should meet down - VM - at a defined
point.

What do you think about?

Greetings,

Stephan
-- 
Stephan Rudlof (sr at evolgo.de)
   "Genius doesn't work on an assembly line basis.
    You can't simply say, 'Today I will be brilliant.'"
    -- Kirk, "The Ultimate Computer", stardate 4731.3





More information about the Squeak-dev mailing list