GNU libl4 overhaul

[Marcus Brinkmann]

Hi,

I gave GNU libl4 a lift.  Now it supports cleanly different interfaces
in parallel.  In particular, you can enable or disable the official L4
interface and our GNU interface.  This will eventually, when all bugs
have been fixed, allow us to use idl4 with this beast.  It might also
allow you to compile third party L4 software on our platform.
However, there are some issues to sort out before that will really
work (mainly the system call stubs, which we don't initialize lazily).

Anyway, this was a very destabilizing change, so some bugs have surely
cropt in.  In fact, many cropt in but I fixed a lot before I committed
these changes.  While reviewing the implementation I also found some
old bugs, and also updated the interfaces to the latest version of the
API.  powerpc needs some more work, though (I didn't test that target
at all).

For me, the code as it is there compiles and runs.  However, that code
doesn't exercise all of the API.  That's why I believe some bugs are
still there.  In particular, I wouldn't put my hand into fire for
string item handling.  For now, it was important to get the changes
into the repository so that we can work within small increments.

From now on, the way is wide open to develop the GNU libl4 interface,
which is currently not really so far from the official interface at
all, at least from first glance.  People who work with the code will
surely have suggestions, and I am interested in those.

The tags for this change are marcus-{before,after}-libl4-rewrite.

What else?  Well, this was only a warm up.  Next step is to get back
to the implementation of the Hurd, of course.  Oh, what else for GNU
libl4 you asked?  Easy: texinfo docs, and regression tests (we can
just emulate the system calls) to make sure that no silly bugs crop
in, like shifting a value twice.  Both are a lot of work, but it is
work that can be done by anybody with enough interest and motivation.

Thanks,
Marcus