absolute vs zero page bug

[TobyLobster]

The assembler chooses 'zero page,X' when it should choose 'absolute,X'. e.g. the "ORA absolute,X" instruction is assembled as "ORA zp,X" even though it isn't, when using a label thats defined later.

i.e. I think this will not produce the correct bytes:

"""
ORG $200
ORA label,X

label:
DB $ff
"""

[dj-on-github]

Confirmed..
root@deadhat:/home/dj/src/py6502/src# python2 small_example.py
LISTING
1 0000 :
2 0100 : org $100
3 0100 : start:
4 0100 : A9 10 lda #$10
5 0102 : A2 00 ldx #$00
6 0104 : 15 0F ora alab,x
7 0106 : loop:
8 0106 : 9D 00 10 sta $1000,x
9 0109 : E8 inx
10 010A : E9 01 sbc #$01
11 010C : 10 F8 bpl loop
12 010E : 60 rts
13 010F : alab:
14 010F : db $55

I'll look into it.
Thanks

[dj-on-github]

It's not just ORA

LISTING
1 0000 :
2 0300 : org $300
3 0300 : start:
4 0300 : A9 10 lda #$10
5 0302 : A2 00 ldx #$00
6 0304 : 35 0F and alab,x
7 0306 : loop:
8 0306 : 9D 00 10 sta $1000,x
9 0309 : E8 inx
10 030A : E9 01 sbc #$01
11 030C : 10 F8 bpl loop
12 030E : 60 rts
13 030F : alab:
14 030F : db $55

[nealcrook]

FYI I'm working on a fix to this..

[nealcrook]

Fixed by my latest pull request. Now, where a choice is available:

• ZP is used when the operand is a numeric literal <0x100 or when the operand is a backward-referenced label that resolves to an address <0x100

• Otherwise absolute addressing is used

This is only sub-optimal in the case where the operand is a forward-referenced label that eventually resolves to an address <0x100. This would only happen if you were executing code from page0. In general this would be a silly/wasteful thing to do on a 6502!