Re: gcc-8086

[Eduardo Pereira Habkost <ehabkost@conectiva.com.br>]

[-- Attachment #1: Type: text/plain, Size: 3726 bytes --]

Hi, Miguel,

On Thu, May 13, 2004 at 10:02:50AM -0600, Miguel Bolanos wrote:
<snip>
> 
> > I would like to know if there is anyone interested on the gcc-8086 work,
> > here, or that have worked on it before. I am starting to look at ELKS
> > code, and I plan to test their binaries using ELKS.
> > 
> 
> I haven't get to test it yet... but yeah i'd be interest to see how
> useful it can be for us. ATM we use bcc.

When I have time (I hope that it will be soon), I'll do some tests
using ELKS, and report them. First, I should try to boot ELKS on my only
8086 machine (an uncommon one, I guess: a hp 200lx palmtop).

> 
> > Another question: is there any interest from the ELKS project in using
> > GCC to build ELKS? Wouldn't it make easier the work of porting existing
> > Linux Kernel code to ELKS, for example?
> > 
> 
> Porting the existing kernel code... u mean making a fork for elks?

I don't know if I would call it a "fork".

I mean trying to make the 2.6 code work on 16-bit. Sure a separated tree
will be needed for this work, as most of the Linux kernel projects. Some
of the work could be used for ELKS, and lots of experience on ELKS,
and even code, would help. If it works, ELKS and Linux would become more
and more similar, but there is a long way ahead before this dream come
true.

(Okay, I am going back to the Real World. I was just wondering if that
could be possible :)


> 
> I have been making my own elks kernel based on linux 2.6 kernel.. but i
> haven't progress that much due to time availability.. but this have been
> more a fun personal project, than anything else.

I've started to do the same thing: playing with the 2.6 code, seeing
how hard would be doing that port. Then I decided look at ELKS first,
then check if it will be worth doing, and if someone haven't thought of
it yet. So here I am.

All these gcc and elks work are being a fun personal project to me, too.

> 
> > I am wondering how painful would be making the actual Linux Kernel code
> > work on a 16-bit arch, once we have a working gcc-8086. The Linux code
> > seems to be "at-least-32-bit dependant" in many parts that are supposed
> > to be arch-independant. But once those parts are changed, there are some
> > strong reasons for not doing it?
> > 
> 
> The first question that comes to my mind is... do u mean to put all the
> linux kernel on a 10mb hard disk (if u are lucky to find one this big
> for an 8086), even though u know that 99.5% of the code is useless for
> ur box?

No, I mean to build only the needed parts, as on the 2.6 code, even
more parts of the kernel are becoming optional at build time,
and there are more that could be done. In a perfect world, we wouldn't
need a separated project for linux-8086, just disable the features you
on't need.

But I didn't tested how small the kernel could be made, even on i386,
removing all unneeded parts. If it is still too big, a lot of the work
would be making unnecessary parts optional. This would help not only
8086 people, but other people that works on embedded systems. If it is
still TOO big, then the work will be even more painful.  8)

BTW, I currently have a 5MB flash disk on my "8086 machine", and should
receive a 256MB ata flash disk for it, soon.

Anyway, my "dream" would be building, from the same Linux kernel tree,
with different options, a kernel as small as ELKS, and the kernel for
my PC or to a server. It seems to be a long way.  :)

> 
> best wishes
> 
> Mike

Just sharing my thoughts. I think that I should show something more
concrete, now, and not only wondering "how nice would it be if we have
foobar ported to 8086".

-- 
Eduardo

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]