Re: GTK+ at the UX Hackfest

On Wed, 2010-03-03 at 15:48 +0800, Li Yuan wrote:
> On Tue, 2010-03-02 at 12:27 +0000, Bastien Nocera wrote:
> > On Tue, 2010-03-02 at 12:22 +0800, Li Yuan wrote:
> > > Mathias made the patch
> > > (http://bugzilla-attachments.gnome.org/attachment.cgi?id=118981 ) to
> > > make at-spi support XSettings.
> > > 
> > > >From a11y side, gail and atk-bridge/atk-adaptor, as GTK+ modules, should
> > > be able to be loaded at any time. If there is no registry daemon
> > > running, for CORBA version, at-spi-registryd will be started through
> > > Bonobo activation (atk-bridge does this job); for D-Bus version,
> > > at-spi2-registryd should be started through D-Bus activation
> > > (atk-adaptor does this job).
> > > 
> > > There are also gnome_accessibility_module_shutdown functions for these
> > > A11Y modules. Maybe we need a unified way to shutdown GTK+ modules.
> > 
> > Right, it seems the discussion we had with Willie was with outdated or
> > incorrect data, so let's step back and look at the original problem
> > again.
> 
> Ah, sorry I didn't say it clearly. I didn't mean there is no problem to
> unload the modules. I meant we could use the
> gnome_accessibility_module_shutdown functions to disable accessibility.

I was actually talking about the instant-load for the modules when the
XSettings change.

> > We'd like to be able to enable/disable a11y in applications without a
> > session restart, and, if possible, when deactivated, that the modules
> > are unloaded and all a11y related functionality is torn down.
> 
> As the previous mail I sent, it is possible to enable accessibility
> without session restart. For disable, it depends on how we do it. It is
> possible to deregister the applications from registryd in
> gnome_accessibility_module_shutdown function, but unloaded the modules
> will cause applications crash. 

It would be nice if you could test that instant-on a11y works as
expected (module loading, application registering) and let us know
whether it works.

Shutting down is always going to be a bit more complicated, but the
overhead of the "leaks" should be small enough that it wouldn't matter
for the short term.

Do you want to look into that?

Cheers