[gambit-list] Error handler in 3.0

Eduardo Cavazos wayo.cavazos at gmail.com
Mon Feb 14 15:51:57 EST 2005


> Interesting...  What are the features of this window manager and do
> you expect to release it?

Here's what it looks like right now:

http://proteus.freeshell.org/atomic/screens/anarchy.jpg

>From left to right you see... a simple file browser, a couple of
rudimentary REPL's, a text pager, and the main menu. Everything you
see, except for the xterm, is running in the same Gambit image. The
REPL's are able to introspect into the running gambit image.

Technically, it's already released. Some time ago I posted on the
gambit mailing list about a collection of code for gambit called
"atomic". The window manager part has been updated since then and I'll
probably make another release soon.

As far as features go, it isn't very featureful yet. For example, it
doesn't have themes.

It isn't robust either. I.e. there has been no effort towards making
it support gnome or kde.

> By the way, why are you still using 3.0?  In 4.0 you could use
> the documented with-exception-catcher.

At one point I tried to move all my code to 4.0, but I ran into
trouble with compiling it. The gsi/gsc command line interface
differences between 3.0/4.0 had not yet been documented. I was having
trouble figuring out the new interface. My gsc invocations aren't
covered in the examples in 4.0. So I decided to roll back to 3.0 and
work on it for a bit.

I am VERY interested in porting all of my atomic library code to 4.0.
I'd like to stabilize the code before porting it.

I will have to make some decisions about how to redesign the X11 event
handling so as to better utilize the thread support in 4.0. The
threads are the main reason why I would want to port to 4.0. In my
current system, there is one event loop that reads events. There is a
simple alist that maps X11 windows to "widget objects" that know how
to handle events. The event handler dispatches events according to the
table. When a new widget is created, it registers itself with that
table. With 4.0 threads, I wonder if every object that expects events
should run in it's own thread and have it's own event loop. Then I
could eliminate the global event dispatcher.

Ed


More information about the Gambit-list mailing list