Ian Piumarta wrote:
> The only complications to using the framebuffer directly are the cursor
> (since we have to draw it ourselves) and reading the mouse.  The first is
> just code and the second is not too difficult on (linux) kernels with the
> new input layer (if your /dev/mouse is a link to /dev/input/mice then
> reading the mouse is almost as easy as reading the terminal keyboard).  
> On kernels without the input layer gpm can be used to do protocol
> transation from the hardware mouse to something more easily digested by
> portable software (cf, the "-t" and "-R" options).
> 
> If anyone is serious about using the framebuffer directly to do something
> interesting with Squeak then I'm prepared to spend a couple of hours
> adding support for it.

As I mentioned elsewhere (and assuming I ever get the time), I'll be 
doing pretty much that for the Zaurus Squeak VM - any help in that 
matter would very welcome.  I already read and filter the touchpad 
myself and the mouse cursor of DirectFB is inappropriate in a stylus 
setting.  Incidently, it also seems to have some screen corruption issues.

Laying Squeak on top of a pretty much barebone Linux kernel makes a lot 
of sense - it definitely would be nice for a PDA.  What's status on FFI 
these day, especially system calls?  Many of the unix utilities are 
pretty mundane once you strip away option and configuration parsing and 
reporting.

Cheers,

   Tommy