* Fix for bug #679013 - AtspiCollection should be implemented for
all containers
* Plug ref count leaks (BGO#679285).
* Only add items to the cache on children-changed if the children are
included in the event.
* Only send PropertyChange signals used for caching, absent listeners.
Especially:
* Only create a directory and a socket when requested.
Currently, atk-bridge clutters XDG_RUNTIME_DIR with directories for sockets,
which is especially annoying since gtk currently does not call
atk_bridge_adaptor_cleanup, so the directories never go away. This change do
not really solve the problem--it only hides it from users who do not really
need AT-SPI--but, nevertheless, there is no reason to create a directory or
socket if nothing has requested it, so doing this lazily makes sense.
* Don't create a (non-readable) subdirectory for the socket when root.
If an application is running as root but within a user's runtime directory,
then creating a temporary directory will result in the directory being owned
by root, and the user will not have permission to search the directory to
access the socket, so an AT running as the normal user will not be able to
connect to the application running as root.
Fixes regression introduced by the fix for BGO#678348.
* Create a library (libatk-bridge), currently with atk_bridge_adaptor_init()
and atk_bridge_adaptor_cleanup() functions, to facilitate linking directly
into, ie GTK+ 3.
* Only register events at beginning if an AT is listening (BGO#678475).
* Use XDG_RUNTIME_DIR to hold the socket, rather than a potentially secure
directory hard-coded under /tmp (BGO#678348).
* Fix various compiler warnings and build errors.
with at-spi-atk. It makes the assumption that at-spi2-atk is used with
gtk3 only.
2.5.2:
* Fix for bug 677211 - The collection interface's MATCH_ANY fails for states
2.3.91:
* Have GetIndexInParent() return a signed int, per the spec.
* Send a DoAction reply message before invoking atk (works around
atk_action_do_action potentially not returning right away for gtk).
2.3.90:
* Have GrabFocus return a bool, per the spec, rather than a uint32.
* Fix a potential crash when emitting a signal if the ATK implementor
misbehaves.
New/updated Khmer, Malayalam, Norwegian Nynorsk, Hindi translations