GNOME Bugzilla – Bug 311729
occasional hangs in ORBit_small_invoke_stub
Last modified: 2006-01-03 23:25:50 UTC
I've now seen this twice in the past few days; I'm running basically HEAD everything with accessibility turned on. The traces are slightly different, but the top few frames seem to be the same, so I've filed them together.
First trace: (this one happened at login time)
+ Trace 62018
Second trace- this one happened when I was launching an app that opened a notification area icon.
+ Trace 62019
Well - it's sending a message to the at-spi-registryd and blocking for the reply. Why it's not replying (or dying) is up to at-spi-registryd. Also - it's best to use 't a a bt' if there are other threads (looks like there may well be in this case). Finally - at-spi-registryd is prolly blocking on some AT it's trying to talk to - what ATs do you have running ? what state are they in ?
| Well - it's sending a message to the at-spi-registryd and blocking for the | reply. Why it's not replying (or dying) is up to at-spi-registryd. Ah, OK. Bill? :) |Also - it's best to use 't a a bt' if there are other threads (looks like there | may well be in this case). Will do next time I see it. Have gotten out of the habit. | Finally - at-spi-registryd is prolly blocking on some AT it's trying to talk | to - what ATs do you have running ? what state are they in ? None; at least none that I've actively run/turned on- I had the accessibility framework turned on for some QA testing, and left it on, but the QA stuff is not running or nor has it even been run since my last reboot.
I don't see how the second trace can be at-spi-registryd's fault, but then again I am not 100% sure what is happenning in the part of the trace above Bonobo_Unknown_unref - why is the moniker code getting called? ORB re-enterancy?
sure - the 2nd is just re-enterancy during an at-spi call out I guess. Why the method is hanging, I have no idea. Need to see the other threads & other processes it might be talking to to see what they're up to.
So, yesterday I discovered a way to reproduce this fairly repeatably: * killall clock-applet * reload? (yes) and it would hang in the reload dialog. However, I rebuilt this morning to get latest code and a full trace, and now I can't reproduce. Grr. So, I'll try again later, but for now this is NEEDINFO, I guess?
Still can't duplicate. Closing.