[kaffe] Kaffe 1.1 & AWT backends

[Dalibor Topic]

Salut Gwenole,

--- Gwenole Beauchesne <gbeauchesne@mandrakesoft.com> wrote:
> Hi,
> 
> Since current CVS code for 1.1 (2003/05/15) pass all tests on both IA-32
> and AMD64 platforms, I have included it into the main MDK distribution
> which should come up as a 9.2 around September.

that's excellent! Thank you very much for fixing the problems and for the 64
bit ports.

> Concerning AWT implemenations and since there is a libawt*.so around which
> is dynamically dlopen()'ed through libtool, I wonder if it is possible to
> ship with different AWT implementations and select the right one at
> runtime.

Sun has a 'standard' mechanism using system property awt.toolkit for that. See
http://www.kaffe.org/pipermail/kaffe/2002-October/028038.html and
http://www.eteks.com/pja/doc/com/eteks/awt/PJAToolkit.html for example.

for a simple implementation it should be enough to:
a) make java.awt.Toolkit use some 'kaffe.awt.toolkit' property to pick a
toolkit implementation.
b) write kaffe.awt.toolkit.XlibToolkit and kaffe.awt.toolkit.QtToolkit wrappers
around native functions.
c) prefix symbols in clib/awt/* accordingly (qt_, xlib_).
d) adapt build machinery to allow building several toolkits at once.

eventually:
e) factor out common functionality in AWT toolkits into a common library, like
PocketLinux did [1].

e) would be nice to have anyway. ;)

I wouldn't use system property awt.toolkit as long as kaffe's awt remains
peerless.

what do you think?

cheers,
dalibor topic

[1] http://www.kaffe.org/cgi-bin/viewcvs.cgi/pocketlinux/XOE_1.0/src/kaffe-all/src/libraries/clib/awt/

__________________________________
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.
http://search.yahoo.com