Rafal Boni wrote:
> 
> In message <20010105002506.A13678@gnyf.wheel.dk>, you write:
> 
> -> On Fri, Jan 05, 2001 at 12:19:28AM +0100, Gerald Heinig wrote:
> -> > Before you complain about "wrong list, we're not Linux":
> -> > I'm aware of that. I couldn't find a NetBSD ecoff kernel
> ->
> -> There is elf2ecoff(1)..
> 
> Though this fails pretty badly if you're trying to do it on an
> other-endian box, as it reads all the header fields and expects
> them to be in native byte order.  I started frobbing this and
> then just found a linux port that did the conversion in the
> right cases (I think it did a run-time endian-ness check and
> then used the magic # to determine endian-ness of the binary).

But if I build my own kernel, I shouldn't need this, should I? I'm assuming the cross compiler will generate ecoff
directly, with correct byte order on x86.
If it doesn't, it might be worthwhile installing NetBSD/Sparc.

> 
> I've got this sitting in my sgimips/stand tree on my x86 build
> box if you need it.  It seems to have worked pretty well so far.
> 
> [...Or we need a BFD based elf2ecoff(1), which should just take
>  care of all the details, but I know very little about BFD...]

What's BFD? 'scuse the ignorance :-)

> 
> I still haven't figured out why (1) my ARCS seems to return a bogus
> memory map, (2) why ARCS fails TFTP'ing images [it gets first packet
> and then goes catatonic, (3) why IDE [the boot-time diag environment]
> can load and run some things that neither ARCS nor SASH can.

I got my Indigo to download and boot a kernel yesterday using Soren's tips, so I seem to be OK on that one now.

> 
> I've had all the above problems for probably three months, but I
> haven't really had the time to set aside a few hours play with the
> I2 that is having these problems.  I did get a newer version of SASH
> to see if that helps, but haven't yet installed it :-<

Cheers,

Gerald