Subject: Re: NetBSD/Mac68k and X11 Windows
To: None <cgd@alpha.bostic.com, macbsd-general@sun-lamp.cs.berkeley.edu>
From: Richard Todd <rmtodd@servalan.servalan.com>
List: macbsd-general
Date: 09/10/1994 23:28:00
In servalan.mailinglist.macbsd cgd writes:
>unless i'm mistaken, A/UX uses COFF (or maybe ECOFF) as its binary
>type...  it shouldn't be that hard to get an exec loader working for
>it, and get some simple binaries running...

It's COFF.  

>In fact, to get emulation for the 'simpler' syscalls is probably
>_really_ easy.  it was only a couple of days work to get most
>statically linked binaries for the OS not named above running,
>including things like 'telnet', and the OS in question is in some ways
>_radically_ different than anything NetBSD had ever seen before...

Emulating the syscalls is one thing.  Getting the various other bits of the
environment to work will be a bit more of a hassle (all the files that programs
need that aren't in the same places as they are in NetBSD.)  Supporting 
executables that aren't statically linked on A/UX is going to be messy,
since the A/UX (SVR3-ish) shared library scheme doesn't use dynamic run-time
linking; it just mmaps (well, not mmap, but a call that does the same sort of
thing) a big statically-linked blob-of-library into the process's addr space,
so you'll have to provide a statically-linked NetBSD library that's compatible
at a low level with what the A/UX shlib-ed executables expect. Or you could
just steal a copy of /shlib/libc_s* off of an A/UX system. :-) 

Of course, to get to the point where you could actually run a X11 server that
was built on A/UX, you'd also have to hack /dev/console to implement all the
A/UX-specific ioctls you'd need.  
--
Richard Todd	rmtodd@mailhost.ecn.uoknor.edu        rmtodd@servalan.uucp
	  New Improved Domain: rmtodd@servalan.servalan.com