Subject: Executable file format
To: None <port-pc532@sun-lamp.cs.berkeley.edu>
From: Ian Dall <DALL@hfrd.dsto.gov.au>
List: port-pc532
Date: 05/23/1994 12:34:26
% ====== RFC 822 headers potentially mangled by stupid mail11 gateway ======
% From: Ian Dall <dall@hfrd.dsto.gov.au>
% To: port-pc532@sun-lamp.cs.berkeley.edu
Apologies to anyone reading this more than once, but it is hard to know
what will catch everyone.
I've pretty much completed the port of gas2.2 and binutils2.3 for the
pc532-mach target. I believe this should be compatable for the NetBSD
people. I am under the impression that the NetBSD ld and as are based
on older versions of binutils and gas. My ports don't include dynamic
library support which I believe has been done for NetBSD by others.
This may be of limited interest (if you have a working assembler,
there is not much incentive to upgrade), but I think it is desirable
to get these changes fed back to the FSF so we are not always having
to merge in our own changes to new releases. Particularly in the case
of binutils, ns32k support has been done many times but despite
efforts (by me at least) has never been included into the official FSF
source. Now binutils looks like it is being supported better and is
making a serious attempt at multiple target support I think there is a
better chance for feeding this back.
I can build gas both with and without BFD_ASSEMBLER and have it work.
There is unfortunatly a noticable performance hit in the latter case.
I have also fixed a couple of more long standing gas bugs (like the
jump table kludge not working for the 32k).
I can build a substantial body of code (the mach micro kernel) using
the new binutils and gas and it works.
I am just cleaning up around the edges. I am wondering if anyone has
any ideas about what machine info number we should use (for the second
half word in a.out) to indicate our machines. The files created by the
old tools had zero here (which the new tools take to mean "assume the
default"), so there is no problem with reading the old files whatever
we use, but if we create files with something in this field, we need
to be consistent. The only real constraint is that we should avoid
numbers used by other machines. Any suggestions?
Ian
------------------------------------------------------------------------------