POC: split init into /hurd/startup and /hurd/init

[Justus Winter]

Hi,

this is my proof of concept patch series splitting /hurd/init into
/hurd/startup and /hurd/init.

[PATCH 1/9] Add proc_set_init_task, make runsystem pid 1

Unchanged, just rebased to master.

[PATCH 2/9] startup: rename init to startup
[PATCH 3/9] Add a minimalist init program

The new /hurd/init is tiny, it just spawns a process that brings the
system up and reaps children. I just noticed that I failed to include
the scripts in the patch series, but I hope the overall idea comes
across.

[PATCH 4/9] startup: do not pass signals on to the child

I'm not sure why this was ever done, if anyone can shed some light on
this, I'd like to hear it ;)

Fwiw I *think* this might be related to
https://savannah.gnu.org/bugs/?20656

[PATCH 5/9] startup: remove unused function run_for_real
[PATCH 6/9] startup: be more specific in the shutdown message
[PATCH 7/9] startup: fix the declaration of the *_server functions

Cleanups.

[PATCH 8/9] poc /servers/startup

Right now /hurd/startup uses a single port for everything. I believe
that this is bad practice. It is hard to change though, since every
server interested in startup notifications actually requests
/hurd/startups message port and sends notification requests
there. This kind of PID based service discovery is unhurdish, so it
seems prudent to bind /hurd/startup to /servers/startup instead.

I think I figured out why /hurd/startup implements the msg protocol
explicitly, it has no signal thread. I actually wonder how the other
"bootstrap" servers deal with this.

[PATCH 9/9] XXX fix build

Fix build.

Comments?
Justus