[epi]

Prerequisite for stuff like opt 'cpu=65c02' (#16) or opt 'format=sdx' (#17).

[pfusik]

We don't have to provide long aliases to all the short options right away. The long options should serve future expansions not 1:1 mapping of short options.

The obvious ones are listing=on, listing=off, cpu=65c02, cpu=6502.

[pfusik]

How about combining three short options into one long:

 opt o-     = opt 'output=off'
 opt o+h-f- = opt 'output=raw'
 opt o+h-f+ = opt 'output=rom'
 opt o+h+f- = opt 'output=ataridos' ; default
 opt o+h+f+ ; not useful, no long alias

[epi]

Looks good. I'd add a few restrictions:

• setting output is an error after anything has been emitted to object,
• setting output twice is an error,
• setting o, h, or f after output has been set explicitly is an error.

Without output, the single-letter options o, h and f work as previously.

[epi]

Maybe the above are too restrictive – one may want to get label table and/or listing without the object bytes, and opt 'output=spartadosx'; opt o- would be a quite self-descriptive way to spell it, even if not the most elegant.

Currently o- is allowed even mid-block, which I consider a bug because it generates broken binaries. Not sure if there's any code out there that depends on it.

I'd disallow opt o- completely after the first object byte is emitted, unless there is a good reason to keep it allowed between blocks.

[epi]

It may also make sense to get o out of output (so the latter would only set h and f), and provide 'object=on|off' so that it can be set in one opt with output.

[epi]

Ignore the last two comments. I don't think there's a use case for o- after output=....

[pfusik]

LGTM!