Re: bonobo activation environment matching and DISPLAY

["Gustavo J. A. M. Carneiro" <gjc inescporto pt>]

A Seg, 2004-08-23 às 16:35, Bill Haneman escreveu:
> On Fri, 2004-08-20 at 22:15, Gustavo J. A. M. Carneiro wrote:
> > ...> > We're wondering what the best fix is - should we just special case the
> > > > matching of DISPLAY in bonobo-activation, rather than trying to enforce
> > > > consistent values of DISPLAY among desktop processes?  Or should we call
> > > > gdk_get_display and re-normalize the environment of the client before
> > > > doing a bonobo-activation query?
> > > 
> > > 
> > > Nautilus does this early in main(), before registering to b-a to
> > > canonicalize the DISPLAY env var:
> > > 
> > > 	/* Need to set this to the canonical DISPLAY value, since
> > > 	   thats where we're registering per-display components */
> > > 	bonobo_activation_set_activation_env_value ("DISPLAY",
> > > 						    gdk_display_get_name (gdk_display_get_default()));
> > 
> >   This is for the old "bonobo activation environment" API, which works
> > with an artificial environment.  The new bonobo:environment property of
> > <oaf_server> matches against the real unix process environment. 
> > Therefore, to achieve the same effect with this method one one have to
> > do the same as in nautilus, but replacing
> > bonobo_activation_set_activation_env_value with setenv (or g_setenv).
> 
> Is this the right thing to do?  It seems to me that since we are mucking
> about with the env, it would be safer to use the
> set_activation_env_value rather than actually do a setenv() in the
> client.
> 
> What we ended up doing in at-spi is using a new bonobo-activation
> environment variable, which we call AT_SPI_DISPLAY, and create from
> DISPLAY at runtime in both the server and in querying clients.  In this
> way we can control the correlation between DISPLAY and at-spi-registry
> more effectively.  For instance, in the course of discussing this
> internally we concluded that indeed we wanted one at-spi-registryd per
> x-server (but not per-screen), at least in most scenarios, therefore we
> needed to strip the screen number off of DISPLAY in some cases anyway.
> 
> I assume that your note above about the bonobo:environment property
> doesn't mean that the "artificial environment", i.e. the constructed
> req_env, is deprecated...  let me know if I've missed something.  

  Yes, I think it is deprecated, but I defer to Michael on this one. 
Michael, what's your opinion on this?

-- 
Gustavo João Alves Marques Carneiro
<gjc inescporto pt> <gustavo users sourceforge net>
The universe is always one step beyond logic.