Re: Async communication
- From: Daniel Nilsson <daniel oden homeip net>
- To: gtk-list gnome org
- Subject: Re: Async communication
- Date: Mon, 9 Aug 2004 08:04:31 -0400
On Sat, Aug 07, 2004 at 12:49:35AM +0200, Soeren Sandmann wrote:
> Daniel Nilsson <daniel oden homeip net> writes:
>
> > Any thoughts of a easier way to do this ? Could I use the glib
> > GIOChannel somehow or coud the glib Asynchronous Queues help in any
> > way ?
>
> Since you can select() on the serial port, you can avoid threads by
> having a function called every time something interesting happens on
> the file descriptor. The GLib mainloop has the functionality for this,
> but I wouldn't suggest using the GIOChannel; I think they are way too
> complex.
Thanks to everyone who responded to my question, I understand that
it's difficult to understand all the requirement when you haven't seen
the program...
I think what I'm hearing is basically the same idea from several
people, separate out the GUI thread from the worker thread and use
some synchronization mechanism (async queues for example) between the
actual worker and the GUI. Unfortunately I wasn't expecting the
communication portion of my program to have to block for such long
periods of time so I didn't do a great job of separating GUI and
workers in the beginning of my project. I'll know better next time...
I prototyped a solution to the problem last night though which seems
to work, I created another serial communication thread which runs
every 100mS using g_timeout_add. This thread checks an incoming queue
a (GQueue) for request and spawns of a background thread that runs and
completes the request, using GQueue for the incoming requests works
fine as far as I can tell since none of the functions run with
g_timeout_add runs concurently anyway and all the incoming requests
will come from such functions. When the background thread is done, the
results are pushed onto an async queue and then error checked and
moved to the final destination pointer. The requesting state machine
will only see a NULL pointer for the result until an error free answer
to the communition request is available. Seems to work fine in my
limited testing so far.
Thanks Daniel
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]