[dmd-internals] OMF format Comment Record (0x88) issues

[Walter Bright]

On 2/11/2012 3:42 PM, Yao Gómez wrote:
> Walter, just a quick question.
>
> I was experimenting with a OMF object file dumper that I'm programming, and 
> then comparing the results with the OMF spec. But there's a record that I 
> can't managed to 'decipher' (so to speak).
>
> Specifically, I'm having issues with the Comment Record (0x88) with the class 
> 0x9d, that according to the spec represents the Memory Model. This record is 
> almost always the second one to appear in the object files, and its contents, 
> according to the spec are:
>
>     "The byte string consists of from one to three ASCII
>      characters and indicates the following:"
>
>      0, 1, 2, or 3    - 8086, 80186, 80286, or 80386
>                         instructions generated,
>                         respectively
>
>      O                - Optimization performed on code
>
>      s, m, c, l, or h - Small, medium, compact, large, or
>                         huge model
>
>      A, B, C, D       - 68000, 68010, 68020, or 68030
>                         instructions generated, respectively
>
> However, when I extract the aforementioned byte string, I get '7nO'. Now, I 
> suppose that the 'O' means that the code is optimized, but what about '7' and 
> 'n'. They aren't specified on the spec and don't know what they mean. Can you 
> shed some light on this?
>
> I know that you have a tool to read this kind of information (objdump or 
> something like that), but I would like to understand a little bit what's going 
> on. It's the object file I'm analyzing corrupt? Is there something different 
> between the spec and how the OMF files that DMD generates?
>

The 7 is for the 786 (!), just the next CPU. 'n' was originally the 'Windows NT' 
memory model, which was not in the original OMF spec.