Re: [PATCH 3/4] ptrace: reintroduce __ptrace_detach() as a callee of ptrace_exit()

[Oleg Nesterov <oleg@redhat.com>]

On 02/05, Roland McGrath wrote:
>
> > > Given its content, this function now better belongs in ptrace.c, I think.
> >
> > I don't completely agree... This helper imho has nothing to do with
> > ptracing, except it does __ptrace_unlink(). But OK, I will move it
> > if you prefer.
>
> Obviously where it goes is not a big deal.  But I think it's clear that it
> has everything to do with ptrace and nothing to do with anything else.
> It resolves a situation that can only arise because of ptrace magic.

OK, OK, I will move it.

> > In that case we should export task_detached().
>
> Or just recognize that this trivial wrapper around == -1 has little
> value two lines away from a place where = -1 is done explicitly.
> Really, the "abstraction" is more confusing than not in this function, IMHO.

Well, yes. The only problem it is not easy to grep for this check
without a helper.


(And I still hope we can change the rules sometimes, I think there
 is no good reason to have task_detached() or EXIT_DEAD tasks on
 ->children list. But this is offtopic.)

Oleg.