GNOME Bugzilla – Bug 619666
Orca makes openoffice become fragile when autoinsertions are removed
Last modified: 2012-10-05 02:45:22 UTC
Sometimes while orca is running if you do an action which causes an autoinsertion in openoffice writer to be removed you may find openoffice crashes. Unfortunately this one only seems to happen sometimes and while I give an example of how to reproduce I don't know whether this will work for others (I have seen a couple of different actions cause this but the common thing seems to be that the autoinsertion disappears). Steps to reproduce: Method 1: 1. Write the following line in openoffice writer and press enter twice: (I) Hello world Expected: After the first time you press enter orca should present the (II) being autoinserted and pressing enter again removes the (II) and orca should present this. Actual: After the second enter orca shows that the line has become blank but openoffice may crash. Method 2: 1. Type the following line in openoffice writer while orca is running: (I) Hello world 2. Now press enter, backspace, press j, press backspace a couple of times. Expected: Orca will just report all updates in the document. Actual: After the final backspace I list above openoffice may crash. As said this is not a 100% reproduceable bug. Extra information: I am using openoffice 3.2.0 and orca from git (done today afternoon in the UK, but earlier git versions have shown this. I have never observed this with orca 2.30.1 but that may be luck. I have also tried without orca running (limited testing without orca as I cannot see the screen except to tell when the desktop is showing IE. to detect when openoffice crashes placing me back at the desktop).
Mentally comparing OOo script changes from 2.30.1 and master, I suspect it is just luck. However, I would suspect that Orca asking for something which OOo isn't expecting is what triggers the problem. I'll try to reproduce it. Thanks for the report!
Michael, No luck reproducing it so far. I've tried a *bunch* off and on today with both 3.2 in Linux Mint (Ubuntu based) and 3.3 dev m78 in OpenSolaris. It's my new brain-clearing activity. <smile> Therefore: 1. Would you mind attaching your user-settings.py, orca-customizations.py (if one exists), and app-settings/soffice.py (if one exists)? One possibility is that the trigger is settings-specific. I want to replace my settings with yours and try again. 2. If you're comfortable doing so and happen to have the time, would you please try the latest developer snapshot? (http://download.openoffice.org/next/index.html) The snapshot build can co-exist with the stable build -- at least currently. If memory serves me from previous cycles, that ultimately changes. Anyhoo... In the past few months, the OOo a11y guys have been cranking out fixes. While it's admittedly a bit of a longshot, your issue might be fixed; and if it's not, that would be equally valuable to know (and report to the guys, once we can pin this down). Thanks!
Michael, disregard my previous request. I was looking at another issue and was able to reproduce the crash. Only once so far. But using my settings (obviously) and 3.3. This bug is going to live up to it's last three numbers. I can just tell....
(In reply to comment #3) > Michael, disregard my previous request. I was looking at another issue and was > able to reproduce the crash. Only once so far. But using my settings > (obviously) and 3.3. > > This bug is going to live up to it's last three numbers. I can just tell.... I have to say I think you may be right about this living up to its final three digits. When it started for me I seemed to have reliable runs of producing it but once it gets out of that it seems to be very hard to reproduce again until it decides to start. If I get another run of these crashes I will try and enable debugging and try and capture it happening (fingers crossed its not a time dependent one which the slight delay debugging adss will prevent it occuring). I had been quite busy over the last few days so hadn't really had the time to try and chase this one up.
Is this fixed by bug 640372?
(In reply to comment #5) > Is this fixed by bug 640372? There's an excellent chance. And if not, let's open a new bug and get new data. After all, Orca is significantly different, AT-SPI2 is significantly different, GNOME is significantly different, LibreOffice ....