for next with local symbol can fail

[stefandrissen]

I'll add more details as I dig further.

I have a source that has multiple for next loops using the same @channel symbol.

; block 3

@channel: equ for 6

    ; some repeated code
    call foo{@channel}

next @channel

; some other code

; block 4

@channel: equ for 6

    foo{@channel}: 
    ; more repeated code
    ret

next @channel

When using local symbols, a minus can be used to force pyz80 to look backwards instead of finding the nearest local symbol. When using a next I would expect that look backwards is implied.

But... in this source if I remove some lines of comment, one of the @channel loops manages to reach value 6 causing a failure:

pass  1 ...
pass  2 ...
src\furplayer.s:941 error: Error in expression otherEffects4_7Ch{@channel}: Undefined symbol otherEffects4_7Ch6
      ld (otherEffects4_7Ch{@channel}),a

When I restore a comment line all is well again:

 pass  1 ...
 pass  2 ...
 Finished

The for | next lines for @channel are:

• 494 | 498
• 858 | 862
• 891 | 966 - contains ld (otherEffects4_7Ch{@channel}),a
• 991 | 1797 - contains otherEffects4_7Ch{@channel}: defining the symbol

The number of lines between the 3rd and 4th block trigger the 3rd block iterating past its boundary.

[stefandrissen]

And after whittling the source down to a minimum reproducible, I see the error. It is not that the for / next loops are causing a problem, it is the references to them ld (otherEffects4_7Ch{@channel}),a is picking up the closest channel counter and not the value of the channel loop that it is in.

So all my {@channel} references in a block should be {@-channel}.

@foo: equ for 2

    ;
    ; delete and line 1 is closer

    ld (bar{@foo}),a ; <--- fails, @foo on line 10 is closer

next @foo

@foo: equ for 2

   bar{@foo}:

next @foo

The local symbol look ups are respecting the general rule, that nearest definition applies. When using a local loop symbol it would make more sense, IMHO, to always use the value of the current loop.