>Although not quite; the goal of the bfd toolchain is to understand all
>formats, even `historic' ones.  Yes, the bootloader needs fixing, and I'd
>say it probably should understand ECOFF, since that's what the PROM wants as
>well.

Sure. The thing with a.out on mips is it *is* historical: we've been
using ELF for years and years. It's time to drop the backward compatiblity.


>I ran into trouble here, too; while elf2ecoff works, and is nifty, it would
>be even more nice to be able to tell ld to `do it directly' (which it should
>be able to do), or use objcopy.  I haven't yet gotten far to understanding
>why the resultant kernels don't boot - more news to come.

There's constraints on section order, aligment, and padding which
objcopy doesnt grok. I supopse we could write Yet Another Linkscript...

>I did this, but I haven't completely verified its working order.  Wanna try
>it?  :)

Yes, please. I dont have user-profililng data (there's a PR on that, we need
to patch the glue code gcc/egcs's uses to call _mcount), but I have
some kernel profiling data I can run against the original ELF kernel.