| View previous topic :: View next topic |
| Author |
Message |
OPENSTEP
The One
Joined: May 25, 2003
Posts: 4752
Location: Santa Barbara, CA
|
 Posted: Sun Dec 19, 2004 7:57 pm Post subject: |
|
Might you be able to go back into the Console application and check to see if there was a newer crash log for the application? The patch11 log that's up there looks like it was the same as the post testpatch6 one (same date of 12-1 . If the app's freezing, it also won't leave behind a crashlog on us, so we may need to find a series of steps to reproduce the freeze.
I've seen the white windows that aren't properly flushed to screen a couple times in all of my usage of the app. I'm not quite sure what winds up being responsible for doing the port flush, the VM or us. Back in the old Neo/C quickdraw carbon code window flushing was a persistent problem, so it wouldn't surprise me if the Carbon 1.3.1 VM doesn't get it quite right all the time either.
ed |
|
| Back to top |
|
 |
naramsin
Sentinel
Joined: Mar 13, 2004
Posts: 27
|
| Posted: Sun Dec 19, 2004 8:19 pm Post subject: |
|
| OPENSTEP wrote: | Might you be able to go back into the Console application and check to see if there was a newer crash log for the application? The patch11 log that's up there looks like it was the same as the post testpatch6 one (same date of 12-1. If the app's freezing, it also won't leave behind a crashlog on us, so we may need to find a series of steps to reproduce the freeze.
ed |
Sorry about that Ed! I flushed the logs in Console so I won't make the same mistake again. I guess, like you said, it didn't generate a new log. What a drag!
Thanks, in any case, for all the work and for catching my mistake.
Cheers,
Roark |
|
| Back to top |
|
|
OPENSTEP
The One
Joined: May 25, 2003
Posts: 4752
Location: Santa Barbara, CA
|
| Posted: Sun Dec 19, 2004 8:40 pm Post subject: |
|
Oh, no worries. It's just a relief to find that it didn't leave a log. For about half an hour there I was preparing to commit ritual suicide upon seeing bug 209 potentially open again or at least burn a figure of that silly java mascot in effigy 
The spinning beachball is quite frustrating though. I did notice a deadlock with regards to the display change notifications at one point during my rapp debugging and attached to get traces but hadn't done anything about those yet. I wouldn't think that expose or the screen savers should be sending display manager notifications, though, so it may be unrelated.
ed |
|
| Back to top |
|
|
pluby
The Architect
Joined: Jun 16, 2003
Posts: 11949
|
| Posted: Sun Dec 19, 2004 8:42 pm Post subject: |
|
| naramsin wrote: | I left patch11 running on the 2 machines at work; NeoJ crashed on the iBook G4 but is still running on the Powermac 450 with many large files open. I uploaded the crash log to Bugzilla as "patch11crash.txt."
I had been away from the computer for an hour or so, came back to the iBook and tried to call NeoJ (which had been hidden) from the Dock, but it would not come forward. By hiding the frontmost apps I saw that NeoJ had a spinning beach ball and needed to be force quit. |
The fact that there is no new crash log and that you are getting the spinning beach ball sounds a lot like "Patch-10" behavior instead of "Patch-11" behavior.
Interestingly, the user who reported bug 267 made the following observation:
| Quote: | | I should add that you need to reboot after adding the patch, or the problem persists (even though NeoOffice self reports that the patch is applied). You may want to let users know about this. |
It could be that my patch installer does not force Mac OS X to reload the update libraries but, instead, Mac OS X keeps using the old libraries that have been memory-mapped in a previous Neo/J run.
Can you try rebooting and let me know if this help?
Patrick |
|
| Back to top |
|
|
OPENSTEP
The One
Joined: May 25, 2003
Posts: 4752
Location: Santa Barbara, CA
|
| Posted: Sun Dec 19, 2004 8:46 pm Post subject: |
|
Patrick, does the installer have any checks to make sure NeoJ isn't running at the time of patching? I know when I'm doing development I've been burned in the past by trying to copy over a dylib that's still in use by the application or gdb and finding the copy didn't take as the file was still open.
ed |
|
| Back to top |
|
|
pluby
The Architect
Joined: Jun 16, 2003
Posts: 11949
|
| Posted: Sun Dec 19, 2004 8:51 pm Post subject: |
|
| OPENSTEP wrote: | | Patrick, does the installer have any checks to make sure NeoJ isn't running at the time of patching? I know when I'm doing development I've been burned in the past by trying to copy over a dylib that's still in use by the application or gdb and finding the copy didn't take as the file was still open. |
No I don't do any checks if Neo/J is running, but I could add that fairly easily.
Patrick |
|
| Back to top |
|
|
OPENSTEP
The One
Joined: May 25, 2003
Posts: 4752
Location: Santa Barbara, CA
|
| Posted: Sun Dec 19, 2004 9:01 pm Post subject: |
|
That might be a good thing to add in. I can think of the following scenario:
1) User launches NeoJ
2) Patch update page launches in web browser, NeoJ opens behind web browser and kind of gets forgotten.
3) User downloads dimage which automounts, runs installer.
It could be easy to forget NeoJ is still running and there's a possibility the patch might be installed incorrectly. I think the same thing used to happen with Apple's Software Update. I noticed the most recent iCal update did have a check to make sure iCal was quit prior to installing the patch, but am unsure if open files were the reason for adding it.
ed |
|
| Back to top |
|
|
pluby
The Architect
Joined: Jun 16, 2003
Posts: 11949
|
| Posted: Sun Dec 19, 2004 10:01 pm Post subject: |
|
I don't think the patch installer is failing as the reporter of bug 267 confirmed that they quit Neo/J before running the patch installer but when they restarted Neo/J, the pre-patch behavior persisted. Only when they rebooted, did the post-patch behavior show up.
Might it be possible suspect that Apple has changed the way determine if their library memory mapping cache needs updating?
Patrick |
|
| Back to top |
|
|
OPENSTEP
The One
Joined: May 25, 2003
Posts: 4752
Location: Santa Barbara, CA
|
| Posted: Sun Dec 19, 2004 11:50 pm Post subject: |
|
Oooh...now I get it (sorry for being a dumbass). I don't think anything has changed but then again I don't really know. I suppose I could hook up to the darwin CVS repository to see if anything changed that's causing the libraries or resolved addresses to be kept memory resident. dyld is in the cctools project of the darwin repository.
ed |
|
| Back to top |
|
|
Max_Barel
Oracle
Joined: May 31, 2003
Posts: 219
Location: French Alps
|
| Posted: Mon Dec 20, 2004 6:24 am Post subject: dynamic library update |
|
I observed other cases (can't remenber which right now) where dylib were not reloaded after update, unless rebooting.
Stopping all process presumably using the lib does not help. I even tried once to flush the file system buffer by copying a big file to /dev/null. No way.
I never found the way to force the refresh of a library beside rebooting. |
|
| Back to top |
|
|
ovvldc
Captain Naiobi
Joined: Sep 13, 2004
Posts: 2352
Location: Zürich, CH
|
| Posted: Mon Dec 20, 2004 6:42 am Post subject: Re: dynamic library update |
|
| Max_Barel wrote: | I observed other cases (can't remenber which right now) where dylib were not reloaded after update, unless rebooting.
Stopping all process presumably using the lib does not help. I even tried once to flush the file system buffer by copying a big file to /dev/null. No way.
I never found the way to force the refresh of a library beside rebooting. |
Does this mean all of the earlier fixes weren't actually loaded and used until we reboot? That should explain the stickiness of that bug...
Now I don't reboot. Ever. Unless I update OSX, that forces a reboot..
But there must be some way that Apple gets libraries to flush - installing a new iTunes doesn't require a restart.. |
|
| Back to top |
|
|
jakeOSX
Ninja
Joined: Aug 12, 2003
Posts: 1373
|
| Posted: Mon Dec 20, 2004 7:27 am Post subject: |
|
not that i am knowlegable with the inner details of the patches.. .but.
when we went from X11 menus to aqua menus, that worked without a system reboot. i remember copy-paste also working without a system reboot. Maybe these were different types of changes then the ones you are seeing, but I thought I'd through that out anyway.
-j |
|
| Back to top |
|
|
pluby
The Architect
Joined: Jun 16, 2003
Posts: 11949
|
| Posted: Mon Dec 20, 2004 9:23 am Post subject: |
|
| naramsin wrote: | I had been away from the computer for an hour or so, came back to the iBook and tried to call NeoJ (which had been hidden) from the Dock, but it would not come forward. By hiding the frontmost apps I saw that NeoJ had a spinning beach ball and needed to be force quit.
Another observation, probably unrelated. Sometimes, especially when NeoJ is not running, if I drag a large .txt file to the dock icon, NeoJ opens and gives me an ASCII dialogue box that is blank (white). I cannot escape out of it and must force quit. When NeoJ is already open it will sometimes give me the ASCII dialogue with the buttons intact. Have you seen this before? |
Ed's comment about the display notifications reminded me that I had added support for Mac OS X "wake up" notifications a few patches ago. I looked at my code and, staring me in the face, was an obvious potential deadlock condition when Neo/J receives "wake up" notifications.
I should have a fix for this out today.
Patrick |
|
| Back to top |
|
|
ovvldc
Captain Naiobi
Joined: Sep 13, 2004
Posts: 2352
Location: Zürich, CH
|
| Posted: Mon Dec 20, 2004 9:54 am Post subject: |
|
| pluby wrote: | | Ed's comment about the display notifications reminded me that I had added support for Mac OS X "wake up" notifications a few patches ago. I looked at my code and, staring me in the face, was an obvious potential deadlock condition when Neo/J receives "wake up" notifications. |
Which reminds me.. When you print to a file (through native dialog), either as .ps or when you choose "save as PDF", the title is somehow appended to include " - NeoOffice/J Alpha 2". Assuming this is not intended behaviour, I filed bug 270 on this..
Ciao,
Oscar |
|
| Back to top |
|
|
pluby
The Architect
Joined: Jun 16, 2003
Posts: 11949
|
| Posted: Mon Dec 20, 2004 11:31 am Post subject: |
|
I have just fixed the "wake up" deadlock issue and I have also fixed bug 269 and posted "Patch-12".
You can download "Patch-12" from the following URL:
http://www.planamesa.com/neojava/patch.php
Patrick |
|
| Back to top |
|
|
|
