Re: [hatari-devel] Symbol loading for upcoming MINT+ELF program format

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On 28.8.2023 19.10, Thorsten Otto wrote:
> although not yet officially released, i think the newly defined format is
> mature enough to deserve support in Hatari. Attached is a patch to do so.

Thanks for the patch!

Please use the same (linux style) formatting as rest of the file.

Elf header processing in symbols_load_binary() is large enough that 
contents for the "/* new MiNT+ELF */" block should be in a separate 
function.

I'm also wondering whether bulk of handling the different symbol formats 
(ELF, a.out, PC) should be split to their own files, and keep only 
generic parts in symbols-common.c (which e.g. includes the others). 
Comments?

(If it's in a separate file, then it can have its own formatting style. 
At least as long it's one of the styles already used in Hatari, not new 
one.)


> Note that currently the p_res1 member of the GEMDOS header contains the value
> "ELF(". "(" here happens to be the decimal value 40, which is the offset to
> the ELF header. That byte is masked out for the tests, should there be any
> future extensions.

Ok, good to know!


> A few definitions might by duplicate (also used in src/lilo.c), but i did not
> bother to move them out there.

That's fine, there's not really overlap.

Overlapping definitions come only with your NetBSD  support (its symbol 
handling), but I haven't merged that series because it does not work 
well enough in my testing (NetBSD boot fails earlier than without it).


	- Eero