Re: [patch] Fix Hurd's kill(pgid, sig)

[Marcus Brinkmann]

At Fri, 10 Feb 2006 04:20:34 +0200,
Guillem Jover <guillem@hadrons.org> wrote:
> So before getting the hurd, glibc and the hurd built again, I'd like
> to know if this is an acceptable solution (given that it adds a new
> interface! ;), and we most probably would include it in Debian for
> mass testing.

It seems to me that there is no difference in definition of what an
"essential task" is for init right now, and for the purposes of
excluding them from being killed by kill -1.  So, we already have an
interface to mark a process as essential, it's called
"startup_essential_task".

The remaining problem is that startup_essential_task goes to init, not
to proc.  This suggests either of two things:

1) init should do the killing.  OR

2) proc should ask init for the list of essential tasks.

I can see an argument for either.  Whatever is chosen, it will be a
startup_* interface, and that means that it can only be used by root,
which is good.  With your version, any process can register itself as
essential, which is not so good.

Thanks,
Marcus