On Wed, 2005-10-12 at 15:48 +0200, Mikkel Kamstrup Erlandsen wrote:
> Good catch! We generally have to make sure any gtk stuff called from
> initialize () is surrounded by gtk.threads_{enter,leave}... I don't know
> if this is done...
>
> When we implement async handlers at some point writers of such handlers
> should also note that any gtk interaction should be surrounded by
> gtk.threads_{enter,leave}.
>
> I'll pu this in AsyncHandler __doc__-strings.
>
> Cheers
> Mikkel
>
That was probably the cause of the Async Xlib reply.
Concerning this, couldn't we surround the module.initialize() with
threads enter/leave in the module_loader.py ? Or would that render
threads useless ?
By the way, the network places isn't the only one using gconf or gtk
methods while called in a thread, i think mozilla and epiphany have
gconf usage in global scope to retreive the browser preference. Must
these also be placed in threads ?
I hate threads.
Raf
> On Wed, 2005-10-12 at 14:06 +0100, Joao Inacio wrote:
> > the network places module hangs in the test, and this makes it work for me:
> >
> > def _scan_network_places(self):
> > + gtk.threads_enter()
> > client = gconf.client_get_default()
> > if not client.dir_exists(NETWORK_PLACES_GCONF):
> > + gtk.threads_leave()
> > return
> >
> > dirs = client.all_dirs(NETWORK_PLACES_GCONF)
> > + gtk.threads_leave()
> >
> >
> > On 10/11/05, Raphael Slinckx <raphael slinckx net> wrote:
> > > On Tue, 2005-10-11 at 10:04 +0200, Mikkel Kamstrup Erlandsen wrote:
> > > > >
> > > > > The concerning fact was that the window was slow to come up. Digging
> > > > > through the code I tracked this down to the signals emitted from the
> > > > > ModuleLoader. I gave them the same priority as default GTK events.
> > > > > Seemingly this was way to aggressive. Changing this to
> > > > > gobject.PRIORITY_DEFAULT_IDLE did the trick.
> > > > >
> > > > >
> > > > > Raphael:
> > > > > You modifications to my initial patch are most welcome. They work out
> > > > > fine for me... I didn't know that python had a thread safe Queue ready
> > > > > to plug in. That is a valuable trick!
> > > > >
> > >
> > > Hi !
> > >
> > > Ok, i'm ready to put this in CVS, the idle_priority makes indeed the UI
> > > appear smoothly, all handlers are ported, i cleaned some debug messages,
> > > and added the blocking module in CVS, but it will of course not be
> > > installed (only EXTRA_DIST'ed). I also set the priority to 0 so it does
> > > not clutter testing.
> > >
> > > Now the question is: Should we release something before we integrate
> > > this and the async handler patches ? Mikkel and me thought it was a good
> > > idea, so we release often, release early. We will then merge the async
> > > module patch and async handler patch and make release again.
> > >
> > > Nigel, Mikkel is that ok ? Do we have critical things to fix before
> > > making a new release ?
> > >
> > > Raf
> > > PS. Attached is the new patch for async module, to be applied to clean
> > > CVS
> > >
> > >
> > > -----BEGIN PGP SIGNATURE-----
> > > Version: GnuPG v1.4.1 (GNU/Linux)
> > >
> > > iD8DBQBDS+4Wd7A94rlhffIRArAhAJ48TgNYRCU1TQxm64uWHtPoIbp6+gCggQY8
> > > HMVUdL3dqEQFDOF5JS1y7DY=
> > > =ddwc
> > > -----END PGP SIGNATURE-----
> > >
> > >
> > > _______________________________________________
> > > deskbar-applet-list mailing list
> > > deskbar-applet-list gnome org
> > > http://mail.gnome.org/mailman/listinfo/deskbar-applet-list
> > >
> > >
> > >
> > >
> >
> >
> > --
> > João Inácio
> > jcinacio gmail com
> > _______________________________________________
> > deskbar-applet-list mailing list
> > deskbar-applet-list gnome org
> > http://mail.gnome.org/mailman/listinfo/deskbar-applet-list
> >
>
>
Attachment:
signature.asc
Description: This is a digitally signed message part