#fill raises 'expected expression' error

[nicolandu]

#fill complains about there being a missing expression, but is satisfied with any dummy expression. This is contrary to the documentation (wiki), available at https://github.com/hlorenzi/customasm/wiki/Working-with-banks-%E2%80%94-%23bankdef%2C-%23bank#fill-attribute

[hlorenzi]

Hmm, in what case exactly are you encountering the hard error? I have a few test cases for #fill, but none of them are catching it.

About it accepting any dummy expression, it's from a not-yet-implemented feature where you could provide a custom fill pattern. I should add an error to it while the feature isn't complete!

[nicolandu]

I needed to include a dummy expression in this case: #bankdef header { #size 0x10, #outp 8 * 0x0, #fill dummy }, as well as similar banks.

The error first occurred when I switched from an older version to v0.13.2. I found it odd that the error 'missing expression' was undocumented, so I tried adding #fill true, but false, 1 and 0 all seemed to silence the error as well. This led me to think that any dummy expression was fine.

[thenorili]

@nicolandu could you please provide an asm file that reproduces the problem? I copied your example and removed dummy, but it compiles with no error.

[nicolandu]

@nicolandu could you please provide an asm file that reproduces the problem? I copied your example and removed dummy, but it compiles with no error.

This is what I had done (using Windows 10 and customasm 0.13.2):

PRG_BANK_COUNT = 2 ; PRG ROM size (x * 16 kiB)
CHR_BANK_COUNT = 1 ; CHR ROM size (y * 8 kiB)

#bankdef header   {               #size 0x10,         #outp 8 * 0x0,                          #fill dummy }
#bankdef prg      { #addr 0x8000, #size 0x8000 - 0x6, #outp 8 * 0x10,                         #fill dummy }
#bankdef vectors  { #addr 0xfffa, #size 0x6,          #outp 8 * 0x10 + 8 * 0x8000 - 8 * 0x06, #fill dummy }
#bankdef chr      {               #size 0x2000,       #outp 8 * 0x10 + 8 * 0x8000,            #fill dummy }

#bankdef zeropage { #addr 0x0,    #size 0x100 }
#bankdef oam      { #addr 0x200,  #size 0x100 }
#bankdef ram      { #addr 0x300,  #size 0x500 }

#bank header

; magic number
#d "NES", 0x1a

#d8 2 ; 2 * 16 kiB PRG ROM
#d8 1 ; 1 * 8 kiB CHR ROM
#d4 0 ; low nybble of mapper id
#d1 0
#d1 0 ; trainer presence
#d1 0 ; SRAM presence
#d1 1 ; mirroring (0:H, 1:V in memory)
#d4 0 ; high nybble of mapper id
#d4 0
#d8 0
#d8 0
#d2 0
#d1 0 ; bus conflict presence
#d1 0 ; extra RAM presence
#d2 0
#d2 0 ; region

#bank vectors
#d16 le(nmi`16)
#d16 le(reset`16)
#d16 le(irq`16)

EDIT: I don't know whether that changes anything, but this file was #included along with a few other files in my main program, if that can help.

[thenorili]

@nicolandu the program you've provided doesn't compile, 'nmi', 'reset', and 'irq' aren't defined in this file -- probably deferred resolution until they were included. Removing those commands does not result in a compilation error.

[nicolandu]

@nicolandu the program you've provided doesn't compile, 'nmi', 'reset', and 'irq' aren't defined in this file -- probably deferred resolution until they were included. Removing those commands does not result in a compilation error.

Indeed, these reset/interrupt vectors are defined in the file which includes this one. I think the issue can be closed for the time being, as it seems to be okay on your side and I don't have access to the exact setup when the error occurred. I'll keep you updated in case any similar issues arise in the future.

Thank you for your support!

[nicolandu]

It seems that the problem doesn't appear in multi-line form: whereas

#bankdef header { #size $10,  #outp 8 * $0, #fill }

doesn't work (by itself), the equivalent snippet

#bankdef header {
    #size $10
    #outp 8 * $0                        
    #fill
}

does work (by itself). I wonder if that can be of any help.

→ Tested using customasm v0.13.5, both native and web.