[PATCH] Avoid a pipe read in aio result scheduling

[Jeremy Allison]

On Mon, Aug 22, 2016 at 12:09:29PM -0700, Jeremy Allison wrote:
> On Mon, Aug 22, 2016 at 05:27:35PM +0200, Volker Lendecke wrote:
> > Hi!
> > 
> > Attached find a patchset that is supposed to improve our async I/O
> > result scheduling. For a more verbose explanation I'd recommend the
> > commit message for [PATCH 04/23].
> > 
> > Jeremy, I did not look deeply enough into your threaded messaging
> > tevent code to see whether the tevent_thread_proxy_schedule()
> > implementation can be based upon tevent_threaded_schedule_immediate().
> > 
> > Thanks metze for very deep review and insights.
> > 
> > Comments?
> 
> Oh it removes the asys stuff ! Very cool (deleted code is always
> good :-).
> 
> I'll take some time and go over this really carefully asap (next
> day or so).

Just got through the tevent core changes.

Still reviewing - but the tevent changes are *amazingly* small
and clean for the change they make - great job !

Solves a different problem than the tevent_thread_proxy_schedule()
code does though I think - this new code changes the core
tevent loop scheduling whereas the tevent_thread_proxy_schedule()
stuff is essentially built on top of existing tevent code as
a utility function to pass memory back and forth.

So I don't feel like the older code is obsoleted :-) :-).

Once I've finished going through and pushed I'll see if
I can build the tevent_thread_proxy_schedule() on top
of the new core functionality to save having the extra
pipe().