ObjC improvements and ObjC++

[Stan Shebs]

Now that Apple's developers are happily banging away on Panther
and Xcode and GCC 3.3, we have an opportunity to get local work
on Objective-C into FSF GCC 3.4, and to get started on Obj-C++
as well.

However, the local changes are lengthy and somewhat intertwined,
and it seems like excessive makework to try to separate and patch
them into the trunk one at a time; so I'd like to have a relatively
short-lived objc-improvements-branch to stage the changes, test
them away from Apple's other local changes and on other platforms,
with GNUstep, etc, and then move them into the trunk en masse. 99%
of the basic ObjC changes will be in gcc/objc; there are a couple
to generic C frontend code, and one dubious change to calls.c that
will need extra special attention.

ObjC++ will be more complicated.  There is some generification
needed to the objc/ code, and Zem will be working on new code
needed to handle ObjC syntax in the recursive descent parser.
We will also need to work up a spec of just exactly what ObjC++
is expected to do; we have an empirical understanding that says
"C++ and ObjC objects are mutually oblivious", and "plain ObjC
(modulo C) and plain C++ code both work unchanged", but that
needs to defined more precisely.  It's not clear at this point
whether ObjC++ can make the 3.4 schedule, but we do have the
cycles now and want to take maximum advantage of them while they
exist.

Stan