How to know if all pending tevent_req requests have completed?

[Jeremy Allison]

On Thu, Feb 23, 2017 at 03:58:36PM +1100, Amitay Isaacs wrote:
> On Thu, Feb 23, 2017 at 11:39 AM, Jeremy Allison <jra at samba.org> wrote:
> 
> > On Thu, Feb 23, 2017 at 11:18:41AM +1100, Amitay Isaacs wrote:
> > > Hi tevent experts,
> > >
> > > tevent_req_poll() allows to wait for a specific request.
> > >
> > > If I want to create n requests and wait for all of them to finish, is
> > there
> > > an easy way to do it?
> > >
> > > Here is an example computation:
> > >
> > > run_event_send()
> > >    ... run scripts one after the other ...
> > >
> > >    if (timer expires) {
> > >       run_debug_send()
> > >       tevent_req_done()
> > >    }
> > >
> > > In this case, run_debug_send() computation is *not* hanging off the
> > > run_event_send() request.
> > > The calling code needs to know that run_event computation has timed out,
> > > but doesn't really care whether debugging computation is over or not.
> > The
> > > debug program can run in the background and log useful debug information
> > > without affecting the caller.
> >
> > Can't you use tevent_wakeup_send() to set a timeout on the
> > tevent context, and then check the state of each req you've
> > saved off in a list somewhere by using tevent_req_is_in_progress() ?
> >
> 
> Well, the requests could created any at level, so keeping track of them
> might not be feasible without layering violations.
> 
> I was really hoping that event context keeps track of active requests and I
> can query the event context if the requests list (or queue) is empty or not.

The event context does keep track, but different backends
may do so in different ways - so there's no generic 'are
any events pending' call (well, there's tevent_common_have_events(),
but that's an internal tevent call meant for simple backends
to use, so you can't use that :-).

The layering violation is your real problem here.