[g-a-devel] Strange crashes with pyatspi

[Zack Cerza <zcerza redhat com>]

Hi all,

So I've been working on a port[0] of dogtail to pyatspi whenever I've 
been able to for the last few months, and I'm having one major problem: 
Sometimes the entire session just crashes. That's right, the whole 
session. :(

I've spent a lot of time trying to debug this to no avail. The furthest 
I've gotten is that sometimes at-spi-registryd gets a SIGABRT, and 
exits. The next time dogtail (or even accerciser) starts after this, the 
session goes *BOOM*. Even toggling 
/desktop/gnome/interface/accessibility to 'false', and then to 'true' 
after at-spi-registryd dies will trigger this.

The most reliable way I've found to reproduce the problem is:

1 Run accerciser.
2 Run sniff (from dogtail).
3 Here, one of them might hang.
4 Close both (^C or kill if necessary).
5 Here, at-spi-registryd dies.
6 Run sniff or accerciser.
7 Here, the session goes down, and you get GDM back.

I don't know much else about what's causing this, but it's definitely a 
blocker for ditching the mess that is pyspi (even though pyspi is very 
broken in F7). Any help I could get would be greatly appreciated.

Thanks,
Zack

[0] http://svn.gnome.org/viewcvs/dogtail/branches/pyatspi/