Re: fix x86 darwin

[Mike Stump <mrs at apple dot com>]

On Aug 9, 2004, at 11:58 AM, Dale Johannesen wrote:
> On Aug 9, 2004, at 9:20 AM, Mike Stump wrote:
>
> > This fixes the builds on x86 darwin.  They've been broken since:
>
> The following bit changes behavior on non-Darwin targets (consider
> PLUS (MINUS (UNSPEC  const) const)
> so I don't believe I can approve it

I think someone familiar with the canonical form on i386 should be able 
to say if that is safe.  I suspect all such instances about have 
already been collapsed so that that form never appears.  Even if it 
did, the assembler should accept it.  The only remaining issue would be 
if the outputer would write it out with the correct spelling...  I 
could not coax the compiler to form such a spelling...

Also, you used the _other_ lisp notion above, instead of the more 
canonical notation:

(plus (minus (unspec ...) const) const)  ...

I mention that least someone be thrown by it...  I bet your 
mind/fingers work in C.

Oh, and least someone wonder why we accept (plus (minus ...) const) on 
darwin, it was accepted before by darwin before the change that broke 
darwin and is actually used.