GNOME Bugzilla – Bug 144086
logout prompt inaccessible
Last modified: 2004-12-22 21:47:04 UTC
Using gnome built with Gnopernicus 0.9.4-11 on 31 May 2004 -Enable gnopernicus speech -Go to Launch->Log out Notice that gnopernicus doesnt acknowledge the logout screen at all. There is no spoken indication of how to change the logout options. Note that speech still works for the everything else at the same time.
Is our logout doing some kind of a grab here? The logout dialog is supposed to avoid grabs if the a11y gconf key is set.
Using the environment variable GNOPERNICUS_LOG set to at-spi I have determined that no events are sent to gnopernicus. Using gdb I have determined that the function spi_atk_bridge_focus_tracker is called in the application gnome-session.
I have transferred to at-spi/registry as that is where we must look to determine the cause of the problem.
I have stepped through gnome-session in gdb and I find that spi_atk_focus_tracker() is called. When spi_atk_bridge_get_registry() is called registry_died is TRUE. bonobo_activation_activate_from_id returns a non-zero registry. When Accessibility_Registry_registerApplication the function impl_accessibility_registry_register_application is NOT called. When Accessibility_Registry_notifyEvent is called it returns a Bonobo Exception. So it seems that alll attempts to call to the registry from gnome-session do not succeed. Bill, Do you have any ideas?
*** Bug 131460 has been marked as a duplicate of this bug. ***
This is a regression, but I have no idea what the cause is. It seems like an activation problem to me.
because the dup (131460) is a blocker for GOK logout, I am raising the priority of this bug.
I do not see this on Solaris.
This _may_ be similar to 144920, which was in turn related to an older gdm bug regarding the DISPLAY environment. bonobo-activation isn't really smart when it comes to matching the DISPLAY environment string, so for instance :0 != :0.0 this can cause weird activation failures when attempting to talk to the registry daemon. If gnome-session somehow has an unusual DISPLAY environment setting at startup time, this could account for the problem.
Not seen on Solaris.
Created attachment 30789 [details] [review] patch to use canonicalized display minus screen num for registry activation.
Believed to be fixed in CVS.