DBus in the kernel?

[Kimmo Hämäläinen]

On Thu, 2010-01-07 at 15:58 +0100, ext Thiago Macieira wrote:
... 
> > srsly? theres no way a byte array an be marshalled 'in place' ?
> > 
> > if so thats a pretty big FAIL. I cant see anything in the wire protocol
> > that prevents in place marshalling of byte arrays though.
> 
> The protocol allows for that, but the API doesn't.
> 
> It requires knowing that the byte array remains valid throughout the lifetime 
> of the DBusMessage object.
> 
> If you do:
>   DBusMessage *msg = dbus_message_new_something ();
>   dbus_message_fictitious_append_bytearray_nocopy (msg, array, len);
>   dbus_connection_send (conn, msg);
>   dbus_message_unref (msg);
> 
> The message may still not be sent by the time dbus_connection_send returns (if 
> the socket cannot be written to, for example). The event loop will be 
> responsible for sending it later, when the socket can accept data.
> 
> We don't know how long later that will be. And there's no way of notifying 
> that the data has been sent, so can be erased.

If there is a reply to the message (and often there is reply/error), you
would know when it's not needed anymore. You could use PendingCall or a
filter to receive the reply.

-Kimmo