GNOME Bugzilla – Bug 339858
Can no longer run dogtail on VNC or physical server
Last modified: 2006-09-06 14:58:59 UTC
'Afternoon everyone, I'm running dogtail 0.5.4-1 on Fedora5 - I was seeing intermittent problems earlier today where sniff would cause dogtail to crash with the exception listed below. I was able to get around these failures with a reboot (logging out/back in did not help). But now - even rebooting - or reinstalling pyspi and dogtail won't help - can anyone suggest a way to clear this up? Thanks!, Len DiMaggio ldimaggi@redhat.com [root@dhcp83-93 dt_rpms]# python Python 2.4.2 (#1, Feb 12 2006, 03:59:46) [GCC 4.1.0 20060210 (Red Hat 4.1.0-0.24)] on linux2 Type "help", "copyright", "credits" or "license" for more information. >>> from dogtail import tree Detecting distribution: Red Hat/Fedora/derived distribution Warning: AT-SPI error: pre method check: add: Unknown CORBA exception id: 'IDL:omg.org/CORBA/COMM_FAILURE:1.0' Traceback (most recent call last):
+ Trace 67869
test = root.children
children.append (Node (self.__accessible.getChildAtIndex (i)))
AssertionError >>>
This bug looks related to 321273 and 336136 (unknown CORBA exception). The difference is that even after a reboot and a reinstall of dogtail and pyspi, I cannot get dogtail to run at all on this machine. I tried running the python watcher script referenced in 321273 on this machine - the results were: [root@dhcp83-93 Desktop]# python watcher.py Wed Apr 26 21:32:29 2006 00 gnome-session 01 metacity 02 nautilus 03 gnome-panel 04 bluez-pin 05 eggcups Warning: AT-SPI error: pre method check: add: Unknown CORBA exception id: 'IDL:omg.org/CORBA/COMM_FAILURE:1.0' 06 > /root/Desktop/watcher.py(38)?() -> print c.getName() (Pdb) q Traceback (most recent call last):
+ Trace 67871
print c.getName()
return self.dispatch_line(frame)
if self.quitting: raise BdbQuit bdb.BdbQuit
Note - this problem seems to be limited to VNC servers. I am able to get dogtail to run if I create a new instance of a VNC server or work on the physical console of the machine. But, the original/failing VNC server is still corrupted and cannot be used - even if I kill/restart it.
My mistake - I'm now seeing this problem - where simply typing "from dogtail import tree" into the python interpreter can show the error - on both a physical server and a VNC instance. And - it does not go away with a reboot... The only workaround is to keep making new VNC server instances.
Found a workaround - thanks to Dave Malcolm - deleting /tmp/orbit-root seems to clear up the problem...
The fix might also be realted to 321273, but if you delete orbit-$username on a local machine with a local user, things really get unstable. Firefox won't boot, control-panel crashes, and eclipse won't save anymore being some interesting behaviour ;) It all goes away when you reboot of course. Is there another way we can solve this in the interim while the bug is being fixed? We are using dogtail in our Frysk builds, and this error prevents the tests from passing?
Len, can you reproduce this using CVS versions of pyspi and dogtail?
Zack - just saw your message - I'll try this week...after July 4
After building PySPI and Dogtail from CVS, the problem is much improved. I still see a CORBA error now and then, but I'm able to get around it without deleting /tmp/orbit-* and /tmp/gconf-*
Len, would you consider this bug to be fixed?
This should be fixed now. If it isn't, feel free to reopen the report.
Sorry - I missed your comment on July 14 - yes, this seems to be fixed. Thanks!