In message <Pine.NEB.4.05.9903021700040.4373-100000@duhnet.net>Todd Vierling wr
ites
>On Tue, 2 Mar 1999, Jonathan Stone wrote:
>
>: Just for the record: We've tried *very* a hard to keep the toolchain
>: and the backend ``the same'' on both big- and little-endian mips
>: machines.  We build support for both EL and EB into the compiler and
>: the rest of the toolchain.
>: 
>: Speaking with the portmaster hat on, there really should not be *ANY*
>: netbsd-mipse[lb].[ch] files.
>
>This case is special: it is an a.out backend for mipsel, intended to read
>pmax kernels.  I don't think mipseb has an equivalent (the a.out magic
>number in question here is M_NETBSD_PMAX).

Todd,

The static a.out format we used in early days is the a.out format used
by 4.4bsd on mips boxes: both pmax *and* newsmips. Ted Lemon did a bfd
backend for binutils 2.6 or thereabouts.  if we support this -- which
I really really *DONT* want, Simon is already booting ELF kernels with
a two-stage loader -- then we should support a.out format for both
mipsel and mipseb targets.

Its possible that 4.4BSD used the same a.out MID for both mipsel and
mipseb: since they dont store the MID in network byteorder they can
use the endian-ness to disambiguate.


>No.  It's my own blunder, and I intend to fix it (I forgot to commit some
>changes).  The fact that mipselnetbsd.c exists is quite deliberate.

I see where you're coming from, but if we are going to support this
historical format at all, I think we should support it for both ports.