[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
More sane class hierarchies
From: |
Andy Wingo |
Subject: |
More sane class hierarchies |
Date: |
Fri, 24 Oct 2003 12:02:41 +0200 |
User-agent: |
Mutt/1.5.4i |
These changes actually recede my previous email, but because since
Monday I haven't got to the internet you'll get these all at the same
time. Ah well.
I made the class hierarchies a bit more sane. Things work much more like
you would expect them to: <gtype-class-meta> is gone; <gtype-class> is
actually a metaclass now; GObjectClasses (like <gtk-entry>, etc)
actually have just one direct super now, not the whole hierarchy (much
cleaner); GObjectClasses aren't metaclasses any more (they were never
intended to be); Only the signals and properties of the class in
question are stored in the <gtype-class> now, not the whole hierarchy
(fixes some bugs, don't know when this behavior was introduced); there's
a new metaclass <set-once-class> that implements the #:set-once slot
allocations; <set-once-class> is the parent class *and* metaclass of
<gtype-class> (see primitives.scm for details); and perhaps most
importantly, <ginterface>'s are now part of the class-precedence-list of
a class if they are implemented. This makes methods work on <ginterface>
implementations.
Also, signals are supported on interfaces now. This is a good thing.
A lot of the macros had to be tweaked to make this fix. Some of the
gobject-signal stuff was renamed to operate on gtype-instances, as it's
not gobject-specific.
Just wanted to send along an update. I don't know when this will go out,
cause the high winds last night might have broken some wireless towers.
Hope to get this in soon.
Regards,
wingo.
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- More sane class hierarchies,
Andy Wingo <=