[Coco] [CoCo] NitrOS-9 on real CoCo (Was DriveWire 4)

[William Astle]

On 2018-01-02 03:01 PM, Allen Huffman wrote:
> When I was getting my CoCo set back up, I recall discussing with William about the possibility of porting lwtools to OS-9. I forget the details, but it did not sound doable.
> 
> However, if “ansifront” can turn ANSI-C to K&R C, I wonder if all we need is a front-filter that would change the code back, generating labels that fit the 8 character limit on demand?

Yeah. Porting lwtools to OS9 isn't really doable. Just the code for 
lwasm itself compiles to 135K using gcc6809 with optimization enabled. 
Obviously that can't work since it's bigger than the CPU's entire 
address space. That should give you some idea why porting would be 
problematic. That leaves aside completely the runtime data required to 
actually assemble anything. In principle, some substantial rewrites of 
the code under the hood could make it more practical, the time required 
to do that is nontrivial.

In theory, a "filter" that converted source format from what lwasm (and 
before that, mamou, etc.) understands to something an assembler running 
on OS9 understands can be written. Indeed, it would probably do a much 
better job than "ansifront" could possibly do with C since assembly is 
much more precise. Such a process would handle things like "include" 
directives and the like, as well. Exactly how that process would work 
depends on how the OS9 assembler operates.

There are a few cases where you might end up with code that is a byte or 
two different in size depending on how well the OS9 assembler chooses 
addressing modes. For most modules, that won't matter. For some, it might.