Re: [orca-list] RFP - Guadalinex a11y edition

[trev saunders gmail com]

Hi,

> From: Samuel Thibault <samuel thibault ens-lyon org>
>
> > Luke Yelavich, le Fri 11 Jun 2010 08:45:22 +1000, a =E9crit :
> > > What is needed, is a way to signal all existing applications using GTK t=
> o retroactively load the atk/gail GTK modules.
> > =
>
> > I guess it'd be far easier to convince gtk people to do that than
> > keeping at-spi-registryd enabled by default.
>
> I really doubt it.
>
> Right now in order to load gail/atk-bridge modules, there isn't
> anything on GTK like "load accessibility modules" as in the deprecated
> libgnome-ui.
>
> In order to load the modules you use GTK_MODULES, an environment
> variable used to load any general purpose module during gtk_init.
>
> So, in order to be accepted a procedure like this, it should be a
> general method to load any general purpose module during runtime.
>
> You also have the problem of how signal gtk applications on
> runtime. You would require a IPC to do that, so the best candidate
> would be DBUS. This would mean add a new dependency to gtk+.

I'm tempted to try to do this with signals, its kind of ugly, but I believe could work.  My idea is basically 
have a file that contains a list of new modules you'd like gtk to load, the app could find out the name of 
this file from an enviroment variable or it could be determined at build time, whatever.  We'd put this file 
in $HOME/.new_gtk_moduels or something again the details don't really matter, but it should probably be in 
$HOME somewhere.  Then when you want a gtk application to load the modules specified there you can just send 
it a signal.  Now that I think about this you could probably do the same thing with a fifo, without the 
signals part.  While its far less enterprisey it seems reasonable, and fairly easy to implement.  While dbus 
could work it seems to me atleast like a really big hammer for this problem.

Trev