A few more thoughts,

The PPC version of MacOS was originally compiled on AIX, and some Apple
Servers ran a modified version of AIX, hence also the legacy of XCOFF on
MacOS. The AIX also spawned the use of TOC relative global data access.

Darwin uses an entirely different runtime model, although MacOS X supports
CFM for Carbon, the native format is something called 'dyld'? 

Has anybody made a comparison with the ABI specs for Darwin?

Regards,

Roger


----------
>From: Thor Lancelot Simon <tls@rek.tjls.com>
>To: "David A. Gatwood" <dgatwood@deepspace.mklinux.org>
>Cc: port-macppc@netbsd.org
>Subject: Re: char is unsigned... wow
>Date: Tue, Aug 1, 2000, 4:57 pm
>

>On Mon, Jul 31, 2000 at 07:23:04PM -0700, David A. Gatwood wrote:
>> On Mon, 31 Jul 2000, Cliff Neighbors wrote:
>> 
>> > being new to PPC I must admit that I was slightly astonished
>> > to see that by default the C data type "char" is equivalent to "unsigned
>> > char".
>> 
>> Heh.  Yeah, it's a curious thing.
>> 
>> 
>> > this is not only true for the gcc "specs" in netbsd-macppc, but
>> > moreover the "-fsigned-char" flag appears not to be used in the
>> > kernel builds.
>> 
>> Moreover, it's a standard on all PowerPC OSes handed down from on high
>> (IBMoto, IIRC).
>
>I believe it's consistent between the two ABIs, but it's important to
>understand that the *are* *two* POWER/PowerPC ABIs "handed down from on
>high".  So, if you're really interested in what was "handed down from on
>high" you have to look at both of them.
>
>Unfortunately, the free OSes all use the SVR4 one, which is obviously broken 
>in at least one way (relating to varargs in the kernel).  It's probably
>much too late to change that now.
>
>AIX and MacOS (IIRC) use the original AIX ABI, which is *not* the same
>as what we use.
>
>Thor