Orca SEVERE Firefox regression independent of (was Re: is now crashing Firefox 3.0)

[Willie Walker <William Walker Sun COM>]

Hi All:

I noticed Firefox crashing as well.  The Firefox folks seem to have 
introduced a severe regression between Saturday and Sunday:  Firefox 
crashes regardless of whether Orca is running or not.  The main crash in 
Firefox that I'm experiencing is when I just happen to click on a link.  
If you revert back to Saturday's nightly, the problem goes away:

http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2007-03-10-04-trunk/firefox-3.0a3pre.en-US.linux-i686.tar.bz2

As an aside...

I've noticed people talking about the perceived instability of Orca 
itself, and I feel compelled to make a slight perception adjustment.  
Orca typically keeps running along when working with well behaved 
applications.  The main barriers to accessibility are when something 
goes awry in the AT-SPI infrastructure or an application.   The two main 
things that can go awry are either an application hanging or an 
application crashing. 

Due to the implementation of the AT-SPI, if an application hangs, it can 
hang an assistive technology and/or the entire accessibility 
infrastructure (and thus the desktop).  Orca has some defensive 
mechanisms in it for hangs, but it cannot always detect all hangs.  As 
such, when an application is behaving poorly, we sometimes have the 
paradoxical requirement of needing to restart Orca to recover from poor 
application behavior.

In my experiences, crashing applications are due to bugs in the 
application and/or toolkit that the application uses.  That is, when an 
application crashes, I'm usually pretty confident that there's not a bug 
in Orca that caused the crash.  Instead, it's typically a legal AT-SPI 
call from Orca that's causing a bug in the application/toolkit/AT-SPI to 
manifest itself in a catastrophic way.  When faced with these crashes, 
however, we will do the following:

1) Track the problem down.  Can we avoid tickling the bug in the 
application/toolkit/AT-SPI?  If so, we try to implement defensive and 
avoidance strategies in Orca itself.  A recent extreme example includes 
not looking at tooltips because examining tooltips through legal AT-SPI 
calls would cause the application to crash.

2) Whether we can work around the problem or not, we log a bug with the 
application, toolkit, and or AT-SPI infrastructure team.  We also pester 
them to make sure they fix the problem.  In the event that they don't 
fix the problem, we go into their code and fix it for them and plead 
with them to take our patch.  This represents a considerable time sink 
for us, but someone has to do it.

From an end user standpoint, I realize a crash is a crash and you don't 
care.  But, *I* care about the perception people have of Orca itself.  
We're a very hard working team and it hurts a little bit when I see Orca 
taking the blame for failures in other components.  I do have to remind 
myself that's what happens when you are on top of the food chain - you 
get ill when your food sources become toxic.

The good thing is that maintainers of the application, toolkit, and 
AT-SPI infrastructure components really do care about accessibility.   
Moving forward, we're working step by step to encourage component 
maintainers to test their changes for accessibility regressions before 
committing them.  This represents a cultural shift, and is thus 
something that will not happen overnight.

Will
(Orca Project Lead)

Lorenzo Taylor wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> I just installed the latest nightly build of FireFox 3.0.  Imagine my
> surprise when I found out that Orca is now somehow crashing FireFox.
> This is the FireFox dated March 10 and Orca from svn as of a couple of
> days ago since nothing of consequence has been changed since then other
> than the staroffice script and some translations.  I have only gotton 1
> page to display without crashing, and that page has no links; it's just
> text, and only about 3 lines of text.  Other pages either crash
> immediately or wait until I move the flat review cursor or press the tab
> or arrow keys.  I blindly tried a google search without Orca and when I
> started Orca again I found everything worked as expected while Orca
> wasn't running.  However, once I started Orca again and found that the
> search results page had loaded, at least according to the title and then
> pressed a flat review key, FireFox crashed and burned.  Is this a known
> issue at this time that is in the process of being fixed in either
> FireFox or Orca in the near future?  I was really enjoying my new-found
> freedom to visit more sites using FireFox.
>
> Live long and prosper,
> Lorenzo
> - -- 
> I've always found anomalies to be very relaxing. It's a curse.
> - --Jadzia Dax: Star Trek Deep Space Nine (The Assignment)
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.3 (GNU/Linux)
>
> iD8DBQFF9R56G9IpekrhBfIRAlbGAKDPbIRyJDWIWRJdsk3uLSUlweQPkQCePPMu
> vN+6Jh/77JxgRsW/lbJbQv8=
> =mcqI
> -----END PGP SIGNATURE-----
> _______________________________________________
> Orca-list mailing list
> Orca-list gnome org
> http://mail.gnome.org/mailman/listinfo/orca-list
>