Eldred Habert

This is not what it currently does; though this would be better continued there.

Due to many complications with removing the lexer hack, revolving around "lexer-sensitive" directives being input at EOF (see cd072d5e6a9031ae326e9d95c7680bde94a88d81 and tests introduced by #750), and a feature-freeze so that testing...

RGBLINK supports 32-bit section IDs, so a special-case for 512 banks is not viable. Trying to change every byte on each iteration would be closer, but doesn't play nice with...

Then how should they be handled? Conditionally evaluating all blocks won't work either for example due to macro arg expansion... (As a side note, I feel like we're fighting an...

That's because `failedOnMissingInclude` causes `YYACCEPT` in parser.y

> Unless you have to deal with rgbds macros and equs. The counter-argument is "we don't use those, so our tool doesn't account for them". This is valid, as "this...

I've just explained that it is not possible to assemble the input without the dependencies in the general case. To do so in one go, RGBASM would need to be...

Parsing continues after a failed `INCBIN` as "current state known to be incorrect, let's try reporting more errors for this batch"; `-MG` instead tries to avoid being incorrect, so it...

> This could be improved while staying relatively simple and safe: only abort on the first conditional block or fatal error after a missing dep, which would at least account...

> Ah, that first suggestion on aborting on a fatal error after a missing deb most likely solve most common cases. You could still craft very specific examples where this...