Re: [Freemint-discuss] m68k alignment change in Debian/Gentoo

[Miro K. <mir...@gm...>]

On Fri, 13 Jun 2025 at 20:20, Eero Tamminen <oa...@he...> wrote:

> On 13.6.2025 17.49, Thorsten Otto via Freemint-discuss wrote:
> > On Freitag, 13. Juni 2025 16:39:45 CEST Eero Tamminen wrote:
> >> What would break / need fixing if MiNT GCC & related tools would be
> >> built / using 4-byte alignment?   Or should everything "just work"?
> >
> > No, it would not work. You can try mixing code by using -malign-int
> (which
> > basically does the same) and get different structure layouts.
>
> Wouldn't "just" recompiling everything (including TOS?), and using
> different library paths take care of that?
>
Out of curiosity I looked it up in our discussions about mintelf toolchain.
This post by Vincent summarises it well:
https://github.com/freemint/m68k-atari-mint-gcc/issues/20#issuecomment-1697651610.
So answer to my previous question is that no, mintelf is not using 4-byte
alignment.

TL; DR is that both TOS and FreeMiNT/XaAES use internal structures which
are passed to user code and expect 16-bit alignment. Changing that would be
a massive effort, if not straight impossible (due to TOS).


> > IMHO a very bad decision by the debian people.
>
> And apparently Gentoo maintainer too.
>
>
> > I think a lot of people misunderstand the difference between sizeof() and
> > alignof(). And if projects fall into this trap, those projects should be
> > fixed, not the compiler.
>
I have glanced over the discussion, I can understand why they are doing it.
I think this post: https://lists.debian.org/debian-68k/2025/05/msg00051.html
explains it very well. Convincing maintainers to fix alignment bugs is
really painful, I can tell from my own experience.

Too bad for us in any case.

-- 
http://mikro.atari.org