[PATCH] Enabling the Orca screen reader freezes Lightdm

[Kyle]

Steps to reproduce:

1. Enable Orca in /etc/lightdm/lightdm-gtk-greeter.conf by uncommenting the reader line and setting its value to orca
2. Start the lightdm service in the standard way for your distro, usually something like

sudo systemctl start lightdm

3. Once the greeter appears, press f4 to enable the screen reader.

The problem:
Orca starts and speaks "screen reader on," but then goes completely silent. The greeter appears to be completely frozen at this point, and a reboot appears to be the only fix. Orca can't be toggled off, and no keyboard input is accepted. Orca also refuses to read anything on the screen, indicating that it also may be frozen. Attempting a login without speech also fails at this point, which is what leads me to believe that the greeter is either hanging or has crashed.

This behavior seems consistent on Arch as well as Sonar, which is based on Manjaro, an Arch derivative.

[pvagner (peter-v)]

This accessibility related issue has been introduced in revision 313 when results of bug #1366534 were applied.
Before comitting this we had to kill at-spi immediatelly before user session has been launched which quite non standard was easily worked around. Unfortunatelly the change introduced with revision 313 breaks accessibility even while the greeter is showing. For the time being I am reverting to r312 but I'll be trying seperating at-spi related stuff from indicators work within r313 so we can hopefully move forward with this. I don't expect to bring some new improvements I just hope for accessible greeter session again.

[pvagner (peter-v)]

Further more by some tryal and error I have just commented out the line where NO_AT_BRIDGE environment variable is set and voila it now works fine for me. And there is no at-spi process behind as far as I can tell.

[pvagner (peter-v)]

I've reverted a couple of things introduced in R313.
Greeter no longer sets NO_AT_BRIDGE. If copy of at-spi2-core has been started by the greeter it-self child processes orca and ATK don't spawn a new instance of at-spi2. I assume by knowing this we are not risking further related issues.
To terminate child processes I have added back call to g_spawn_close_pid . Otherwise orca hangs while trying to kill it and I assume it might be similar with other apps e.g. on screen keyboard when they are toggled on / off during a single greeter session.

[Sean Davis (bluesabre)]

Thanks for the patch! This seems to work as described, included in the following commit.
http://bazaar.launchpad.net/~lightdm-gtk-greeter-team/lightdm-gtk-greeter/trunk/revision/367