Joined: May 25, 2003 Posts: 4752 Location: Santa Barbara, CA
Posted: Thu Dec 16, 2004 10:10 pm Post subject:
Yeah, I think the firewall here may be preventing my machine from being a seeder though I thought it worked in the past. I'll look into that.
It would definitely not surprise me if the 10.3 update introduced some problems for the VM. It introduced several subtle bugs for other Carbon apps I work on that were legacy...and the roots of the 1.3.1 Carbon VM definitely reach back into legacy code.
This particular crash is troublesome. I've seen it only on two of my machines and never on my main dev box, and its crash logs are notoriously bad at trying to peg just what the problem is. While it seems like it's in the appleevent processing stuff of the VM, really, who knows. I'm hoping that perhaps the lazy binding stuff may work.
Here's what I'm inserting in main() to get the appleevent tracings...I'm hoping the lazy binding might report what symbol is causing the signal from dyld in the appleevent handler, but for the life of me I can't reproduce the rapp or quit problems tonight. I never saw them more then once a day and, aside from the rapp event and quit, could never find any other trigger that would cause them. I've been trying to find a cause for this for a long time now and am still baffled
Can you verify that "Patch-9" plus this test patch fixes this bug? To install this patch, execute the following commands in a terminal:
cd /Applications/NeoOfficeJ.app
sudo tar zxvf "/path/to/downloaded/file"
This test patch is the same as testpatch2.tar.gz except that I reenabled support open and print document events.
Here's my current theory as to what is causing this bug. Since this bug first surfaced in early November 2004 and the occurrences have grown steadily since then, I think that the Mac OS X 10.3.6 update (which was released on November 5, 2004) is the cause of this bug. Most likely, the update has changed a library that Java 1.3.1 is trying to load a symbol from.
To fix this, testpatch3.tar.gz installs an Apple event handler and does not allow anything but open and print documents to be passed to the JVM's handler. Since the crash always occurs in the JVM's Apple event handler, preventing invocation of the its handler should, I hope, prevent crashing.
Can as many of you try this patch today and see if any crashing occurs? If we don't see any crashes in the next 24 hours, then I can release "Patch-10" that includes this fix.
I left the two machines at work running your patch from last night (testpatch2). I did not get Ed's torrent patch. I will check on their status when I go in later. I've had testpatch2 running on my frankenmac since 3AM at home, with 1 large file open. I have since opened a number of large docs (spreadsheet, powerpoint, text docs, and an RTF of the complete works of SHakespeare--5.7MB!) and it is still fine. With these large files it is using just over 1 Gig of VM (frankenmac has 1.5 Gigs installed). It is not crashing and is still very nimble even with all these files open, scrolling, Expose, hiding from the dock, etc. Very stable indeed for these past 8 hours.
I will install "testpatch3" now and see if I get the same results.
"testpatch3" seems to be pretty stable--I've opened a bunch of big docs again and it is doing fine. The only thing not happening is quit from the dock. Unless the program is taking a long time to quit, quitting from the dock doesn't seem to work (but maybe this was the case before). It quits from the menu fine though.
Following your recent post about vm, the real memory consumption with several large docs is around 100 MB.
I'll keep it running and let you know if anything changes, but so far so good. I really hope you found the bug this time!
"testpatch3" seems to be pretty stable--I've opened a bunch of big docs again and it is doing fine. The only thing not happening is quit from the dock. Unless the program is taking a long time to quit, quitting from the dock doesn't seem to work (but maybe this was the case before). It quits from the menu fine though.
Oops! I forgot to add code to handle 'quit' and 'oapp' Apple events. These two events, plus the 'odoc' and 'pdoc' Apple events, are the 4 Apple events that Neo/J actually needs to handle.
Anyway, I have uploaded a new test patch here that includes handling of the 'quit' and 'oapp' Apple events:
Well, some mildly bad news. I installed "testpatch4," did the usual, left it running in the background. Tried the quit from dock, which worked fine the first few times. This last time, however, it crashed. I quit from the dock, it hung for a while, then disappeared. I then received the "soffice.bin has crashed" message.
I wonder if "quit from dock" is unstable? It will be interesting to see what the bug is.
Tried the quit from dock, which worked fine the first few times. This last time, however, it crashed. I quit from the dock, it hung for a while, then disappeared. I then received the "soffice.bin has crashed" message.
I wonder if "quit from dock" is unstable? It will be interesting to see what the bug is.
Hmm. I had my crash without trying to quit - just be switching. Could it be that one of these four events is bugged? Or several, or that the dock just doesn't talk sweetly enough to the JVM?
I wonder if "quit from dock" is unstable? It will be interesting to see what the bug is.
From your crash log, Java's event handler for both the 'quit' and 'oapp' events that I added to testpatch4.tar.gz apparently crash the JVM. In your crash log, it crashes in my code when I explicitly invoke the JVM's event handler for these 2 events.
So, as I noted in bug 209, I uploaded testpatch5.tar.gz. In this patch, I still trap 'quit' events, but instead of invoking the JVM's handler, I bypass the JVM and call OOo code directly.
Since the patches so far have eliminated the 'rapp' crash, we just need to eliminate the "quit from Dock" crash and we should be set.
Hmm. I had my crash without trying to quit - just be switching. Could it be that one of these four events is bugged? Or several, or that the dock just doesn't talk sweetly enough to the JVM?
Two questions:
1. Does the crashed thread in your crash log look like this following:
Here's my current theory as to what is causing this bug. Since this bug first surfaced in early November 2004 and the occurrences have grown steadily since then, I think that the Mac OS X 10.3.6 update
I reported crashes some days ago... but I'm still using 10.3.5 (10.3.6 and 10.3.7 freeze pb12'' using an esternal monitor)
"testpatch5" seems to be running fine on all 3 machines (frankenmac 9500, iBook G4, and 450 Mhz PM tower). Also, the patch I left on the last 2 machines overnight ("testpatch2" I think) was fine before I upgraded to testpatch5. The older powermac had screen redraw issues only (because of its old ATI Rage card?). The iBook was stable and had no redraw issues.
NeoJ is quitting politely from the dock, and seems fine overall so far. I'll keep my eyes on it and alert you to any problems. But for now it seems like you've nailed the bug.
NeoJ is quitting politely from the dock, and seems fine overall so far. I'll keep my eyes on it and alert you to any problems. But for now it seems like you've nailed the bug.
I have "Patch-10" (which uses the testpatch5.tar.gz code) ready to go so if I don't hear of any crashes from you between now and tomorrow morning, I will release "Patch-10" tomorrow.
Joined: Jun 20, 2003 Posts: 104 Location: Birmingham, Alabama; and Amelia (Terni), Italy
Posted: Fri Dec 17, 2004 5:08 pm Post subject: Activating w/o dock works
I've been able to keep NOJ up and running pretty well by avoiding activating it from the dock. I've been trying lots of docs, into background and/or hiding it and activating it again, but all the time avoiding activating NOJ via the dock (I can activate NOJ via one of those OS 9-type menu bar app lister without experiencing the crash).
OS X 10.3.5, patch 9, NO test patches. Hope this confirms where you're looking for the problem. Cheers, _________________ Gib Henry
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum You cannot attach files in this forum You cannot download files in this forum
