Hi. I have been testing this bug yesterday and today.
After all this tests, I can conclude that the problem is related with a race condition.
First: the original environment where this bug was detected is not valid. I was only able to reproduce it with a11y enabled, something not suggested on the description of the bug.
Second: I don't see any relation between this bug and onboard.
About why I think that it is related to a race condition:
* Sometimes just at login the panel-service got stuck on a call to atk_get_root
* Sometimes it starts fine, but just when you launch orca, if crashes (on that g_closure_invoke)
* Sometimes it starts fine, and doesn't crash when you launch orca, but crashes when you interact with the panel (ie: logout)
* Sometimes it works fine, without any problem.
Not sure about the percentage for any of those options, but after all my tests, lets say 20/30/30/20
Recently Neil asked me to make some changes on when the a11y stuff is loaded due bug 817691, also related with race
conditions. At that moment we made just a little change, by moving when the a11y is initialized, in order to keep the changes as small as possible.
Finally, FWIW, need to say that the process of re-launch the panel-service works fine. It can crash or stop, but the new panel-service works fine (also with Orca), so we don't enter on the usual mode of panel-service crashing each time it tries to do something.
Hi. I have been testing this bug yesterday and today.
After all this tests, I can conclude that the problem is related with a race condition.
First: the original environment where this bug was detected is not valid. I was only able to reproduce it with a11y enabled, something not suggested on the description of the bug.
Second: I don't see any relation between this bug and onboard.
About why I think that it is related to a race condition:
* Sometimes just at login the panel-service got stuck on a call to atk_get_root
* Sometimes it starts fine, but just when you launch orca, if crashes (on that g_closure_invoke)
* Sometimes it starts fine, and doesn't crash when you launch orca, but crashes when you interact with the panel (ie: logout)
* Sometimes it works fine, without any problem.
Not sure about the percentage for any of those options, but after all my tests, lets say 20/30/30/20
Recently Neil asked me to make some changes on when the a11y stuff is loaded due bug 817691, also related with race
conditions. At that moment we made just a little change, by moving when the a11y is initialized, in order to keep the changes as small as possible.
Finally, FWIW, need to say that the process of re-launch the panel-service works fine. It can crash or stop, but the new panel-service works fine (also with Orca), so we don't enter on the usual mode of panel-service crashing each time it tries to do something.