key-words are ignored as labels

[t-kl]

Hi
There's this guide Moonforth to build a Forth for the DCPU-16. You can run the code interactively but it's not as nice as using your debugger. Thus I used the code from the first example and everything was working great.
But then I noticed that the ADD-command wasn't executed. I compared the generated object code from both assemblers and lo and behold: your assembler appears to ignore this label. I imagined that the reason could be that ADD is a key-word for the assembler and changed the label to ADDER. With this workaround everything works fine.
I'm not familiar with the specification of the DCPU-16 and thus it might be correct as it's currently implemented. On the other hand I assume that there shouldn't be a problem using a key-word as a label.

I've included the aforementioned code below. However I assume there is a minor bug in the original. A comma is missing in this assignment instruction:

:LOOP_XTO
  SET PC, LOOP_XTO

Kind regards

:START SET PC, MAIN ; JUMP MAIN

:NEXT
:WORD_FROM_CODE
  SET J, [I]   ; WORD-R <- *CODE-R
:NEXT_WORD
  ADD I, 0x01  ; CODE-R++
:JUMP_INTO_XTO
  SET PC, [J]  ; JUMP *WORD-R

:ENTER
:R_PUSH
  SET [Z], I   ; *RET-R <- CODE-R (PUSH PART 1)
  ADD Z, 0x01  ; RET-R++ (PUSH PART 2)
:WORD_AFTER_XTO
  SET I, J     ; CODE-R <- WORD-R
  ADD I, 0x01  ; CODE-R++
  SET PC, NEXT ; JUMP NEXT

:EXIT
  DAT EXIT_XTO
:EXIT_XTO
:R_POP
  SUB Z, 0x01  ; RET-R-- (POP PART 1)
  SET I, [Z]   ; CODE-R <- *RET-R (POP PART 2)
  SET PC, NEXT ; JUMP NEXT


:DOUBLE
  DAT ENTER DUP ADD EXIT

:DUP
  DAT DUP_XTO
:DUP_XTO
  SET PUSH, PEEK ; STACK DUP
  SET PC, NEXT   ; JUMP NEXT

:ADD
  DAT ADD_XTO
:ADD_XTO
  SET X, POP   ; STACK POP
  ADD PEEK, X  ; ADD X to top of stack
  SET PC, NEXT ; JUMP NEXT

:LOOP
  DAT LOOP_XTO
:LOOP_XTO
  SET PC LOOP_XTO

:CODE_1 DAT DOUBLE
:CODE_2 DAT DUP
:CODE_3 DAT LOOP

:MAIN
  SET I, CODE_1  ; CODE-R <- CODE_1
  SET PUSH, 0x05 ; STACK PUSH 0x05
  SET PC, NEXT   ; JUMP NEXT

[dangermccann]

Thanks for this, I'll have a look. At first blush it appears that the tokenizer is attempting to tokenize reserved words before labels, which is incorrect. I'll try to get to the bottom of it...

[dangermccann]

After looking at this a bit I found that this is issue is due to a bug in my tokenizer. The problem is that the tokenizer will match commands and reserved words before label references. However, fixing this would require quite a big refactoring effort in the tokenizer code to convert it from a single pass algorithm to a double pass algorithm that first identifies / matches labels before all other tokens. Considering that this issue has a relatively straightforward workaround I'm going to defer fixing until a later time.

I did modify the code to identify this particular condition (a non-supported label is referenced in a DAT block) and added some error handling to let the developer know. So now at least it won't fail silently.