Re: [dbus-cplusplus-devel] dispatcher timeout

[Theodore A. R. <ta...@gm...>]

On Mon, Dec 6, 2010 at 2:06 PM, Vince Laviano <vla...@go...> wrote:
> Hi Ted,
> On Mon, Dec 6, 2010 at 11:59 AM, Jason Glasgow <jgl...@go...> wrote:
>>
>> see main() for how we attach the dbus-c++ dispatcher to the glib main
>> loop.
>>
>> http://git.chromium.org/gitweb/?p=cashew.git;a=blob;f=src/main.cc;h=d7f64fe5f37b99e91d7eb2ef9e4566f01833e521;hb=HEAD#l102
>>
>> -Jason
>
> dbus-c++ provides several flavors of dispatcher.
> The vanilla one (DBus::BusDispatcher) does implement its own event loop,
> entered by DBus::BusDispatcher::enter and exited by
> DBus::BusDispatcher::leave.
> There is a separate dispatcher for glib integration:
> DBus::Glib::BusDispatcher.  When using the glib dispatcher, you still need
> to attach it, but, since it's using the glib main loop, the enter and leave
> routines are no-ops. You instead create a glib main loop and enter and leave
> the loop using the usual glib calls: g_main_loop_run and g_main_loop_quit.
>  See the cashew link that Jason provided for an example of how to use the
> glib dispatcher.
> The add_timeout, rem_timeout, add_watch, and rem_watch methods on the
> default dispatcher seem promising at first glance, but the conversation
> quoted below suggests that you've already tried them with no luck.
> Andreas says in the thread below:
>> Are you sure the C backend of DBus provides such a feature? Could you name
>> a function?
> The C backend for D-Bus is meant to allow integration of D-Bus events into
> an existing external event loop. So, it provides the ability for code
> implementing this event loop to provide a callback function for libdbus to
> call when libdbus wants to add a timeout that will be managed by this
> external event loop (and similarly for watches on fds that libdbus want the
> external event loop to manage).  However, one would expect that the external
> event loop will... um... be an event loop, meaning that its public API
> allows timeouts and watches to be created, managed, and removed by arbitrary
> clients, not just by libdbus through the dbus binding.  IMO, if the default
> dispatcher purports to be a real event loop (and it's not clear that it
> does, since it's intentionally very simple), it should allow timeouts and
> watches to be managed.  It should do this just because it's an event loop,
> not because it's part of a D-Bus binding.

That's exactly the (false) assumption I was making about BusDispatcher.

> Ted: An alternate approach would be to ignore the default dispatcher, look
> at how DBus::Glib::BusDispatcher is implemented, and follow the same model
> to implement DBus::EventLibThatTedPrefersInsteadOfGlib::BusDispatcher.
> Vince

Will have a look at that and see where it takes me.

Thanks for the help guys. I think that will get me on the right track.

Ted Roth