[cells-devel] New Cells User Feedback (includes patch!)

[James Bielman]

On 22 Apr 2005, Kenny Tilton wrote:

> This is great feedback, and I agree with all of it. The only problem
> is that there is little active interest in Cells, other than my own
> use of it, so I am not worrying much these days about making the
> Cells project approachable to others.

Fair enough.

> I suggest you poke around and see if you want to use Cells for
> anything, and if so we can sort out the mess in CVS.

Well, for kicks I wrote a little 6502 LAP-style macro assembler using
Cells as my internal representation of instructions and labels, which
worked out really well until I ran into a cycle I couldn't get around.

Briefly, the problem is that defining a forward-referenced label may
cause prior instructions to change size, which causes the label to be
redefined at a new location, which causes prior instructions to be
rewritten again to reflect the final label location.

So, there is a cycle, but I have domain-specific knowledge (that an
instruction can only change size once) that ensures the cycle will
terminate.

If I #-(and) out the cycle check in MD-SLOT-VALUE-ASSUME (I think),
everything works okay, so I either need a way to refactor my model to
eliminate the cycle or some way to tell Cells that a particular cycle
is OK.

Anyway, I'd never written an assembler before but I had a blast
writing this---Cells is pretty cool.  The code is up at:

        http://jamesjb.com/~jamesjb/darcs/cl-6502/

if anyone is interested.  (Assembling examples/hello.l65 will error
with a cycle unless you hack Cells to ignore cycles...)


James