Patch: darwin specific

[Per Bothner]

Mike Stump wrote:
 > The usual rules are that the first to submit wins.

No, that is not the usual rule.

> Is it reasonable to address this second aspect (identifier size) of this
> patch later, or should the patch be rejected until such time as that
> second part is done?  That is the choice.

There have been suggestions that it would be better/cleaner to make
decl nodes accessible from rtl, and associate the needed information
with decls, rather than identifiers.  In that case, the question
becomes moot.  I don't understand the issues well enough to comment.

In any case, if there is agreement that a target-specific identifier
hook makes sense, I can easily enough modify it after it is checked
in to fit in my framework.
-- 
	--Per Bothner
per@bothner.com   http://www.bothner.com/per/