Re: Log for attempted build of encfs_1.8.1-2 on m68k (dist=unstable)
Eduard Bloch dixit:
>That's what an internet search gets you looking for "m68k emulator".
Mh sure. There’s more than one though, and for Linux, you need an MMU.
>Anyhow, I tried aranym with your image and instructions from
>https://wiki.debian.org/Aranym/Quick#preview and had mixed feelings. See
Hm okay.
>> (but I don’t speak C++ and can’t help with it) after I enabled it
>> to test, as Andreas said it worked for him.
>
>That's not a C++ issue, it fails on basic bootstrapping. Your trouble
Yes, but I still don’t speak C++ ;-)
>starts in function.c:233:
>
>static void check_alignment( STACK * s )
>{
> assert( (size_t)s->data % sizeof( LIST * ) == 0 );
That’s trivial. sizeof(LIST *) is almost certainly 4, as that is
the native pointer size, and ((size_t)s->data) is almost certainly
a multiple of 2 here, which is the minimum required alignment for
not-bytes on m68k.
My suspection here is that the data type behind s->data uses some
implicit padding assumptions, like here:
struct foo { char c; int i; };
i386: [c][p][p][p][i][i][i][i]
m68k: [c][p][i][i][i][i]
On both architectures, implicit padding is used, just the amount
differs, i.e. you hit C implementation-defined (or even undefined?)
behaviour. The correct fix here is to make the assumption explicit:
struct foo { char c; char padding1[3]; int i; }
This makes it look like i386 on all architectures.
>PS, Aranym feedback:
>
>Doesn't work: apt complains about missing https transport in
I think the image still uses the naturalnet mirror, which has
gone away. I’ll update the Wiki page. (The naturalnet server
operator has decided to force https access and redirects, hence
the trouble from apt.)
>The console is also kinda buggy, I cannot scroll with
>Shift-PageUp/Down. I often lost focus while returning to the emulator.
Mh, these are all emulator issues. We generally ssh into the system,
it makes for a much, much smoother feeling.
>So, after all, installing openssh-server and ssh console was the only
>sane solution.
Correct. I think the README and/or Wiki page says so, too?
John Paul Adrian Glaubitz dixit:
> b) There is no demand for running Linux on a real i386
>
>Both these cases do not apply to m68k. And, furthermore, Linus
Besides, the i386 *platform* still exists, just not support for
an 80386SX CPU. Same for m68k: we currently run on 68040 upwards,
the 68000 was never supported except by µClinux, and I have no
idea whether Linux can run on the 68020, but the contemporary
machines e.g. have much more RAM, I/O, etc. as well.
bye,
//mirabilos
--
<igli> exceptions: a truly awful implementation of quite a nice idea.
<igli> just about the worst way you could do something like that, afaic.
<igli> it's like anti-design. <mirabilos> that too… may I quote you on that?
<igli> sure, tho i doubt anyone will listen ;)
Reply to: