Joined: May 25, 2003 Posts: 4752 Location: Santa Barbara, CA
Posted: Fri Mar 02, 2007 12:46 pm Post subject:
Thanks for your patience and your help Ray. I sent a PM with the instructions on Shark for grabbing detailed samples and the simpler activity monitor Sample utility that should be able to work without installing additional software. Please feel free to let me know if there's anything I can to do help.
Given what we've already found (bugs in the PPC VM, performance issues with OOo drawing...) it wouldn't surprise me if somehow this is a confluence of a number of different factors. Grr.
I would still like to see an Activity Monitor sample and a sample document to reproduce. With that information, I can continue work this weekend. However, with no data I can't proceed.
My apologies for being so harsh to K-9 and rays. My intent was not to criticize either of you but to get real data that I can use to fix this bug. My issue isn't with you personally but the fact is that you two hold the keys to any data that may allow fixing of this before hundreds of thousands of people download Neo 2.1 at the end of this month. Because of this, time is extremely short and, on my end, the clock is ticking very fast.
Since this bug started with Apple's recent Java update, this worries me greatly. I am very upset because I see that not only this bug showed up, but many basic Java bugs showed up as well and the huge list of basic Java stuff that just plain broken in the latest Leopard developer seed is very worrisome.
In short, I am seriously worried that I may need to pull release of Neo 2.1 if Java is no longer stable. The Leopard problems worried me but I convinced myself that it was worth the gamble to wait and see. However, with the current Java update issues, it is only a matter of time before tens of thousands of people see the same problem as K-9 and rays.
Posted: Fri Mar 02, 2007 2:21 pm Post subject: 2.1 EA P0 on 10.4.8/G4 - OK for me
For what it's worth, I do NOT seem to have this problem.I have version 2.1 early access patch 0, OS X 10.4.8, on a 17" powerbook G4 1.33GHz with 1.5Gb RAM
Software update tells me that:
2007-02-16 22:02:21 +0000: Installed "Java for Mac OS X 10.4, Release 5" (5.0)
In terminal:
ferrabosco:~ max$ java -version
java version "1.5.0_07"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.5.0_07-164)
Java HotSpot(TM) Client VM (build 1.5.0_07-87, mixed mode, sharing)
NeoOffice Tools->Options->Java tells me I am using this version of the JRE (though I gather this is not necessarily the Java version that's drawing Neo's pixels)
I also have no problems docking toolbars, & I can cmd-Q quit without crashing.
Any tests you would like me to do on an 'unaffected' machine?
Joined: May 25, 2003 Posts: 4752 Location: Santa Barbara, CA
Posted: Fri Mar 02, 2007 2:25 pm Post subject:
Here's some more detailed instructions for using Shark. Really, the Activity Montior sample is good....but Shark is really what we need to do some serious detective work in figuring out what the app and VM are doing:
Here's a series of steps that don't require any additional software to be installed. They'll get something, but heck something is better than nothing. These are also available as Patrick mentioned at:
Select soffice.bin/NeoOffice from the list of applications
Click "Inspect" at the top (the blue "i")
Get Neo into that state where it's stuttering (if it's hanging, even better...)
Return to the Activity Monitor, click "Sample"
_Quickly_ return to Neo and start typing so it stutters. Unfortunatley, with the Activity Montior sample there is only 6 seconds to try to get it into that state for the information to be useful!
After the Sample is done, click "Save" to save out the call trace to disk and send it along. The "call trace" shows which functions are being called and how often they are being called.
The Shark profiler is part of the CHUD tools. This profiler is the mother of all profilers and is one of the best I've seen on any platform. It's a developer's tool, but it does allow for snapshots of the application with a much more informative level of detail than Sample. Shark profiles are what developers really use to track down these gnarly performance issues.
To capture a Shark profile, first you'll need to install the CHUD tools from:
I don't believe they require the Developer tools to be installed. They also have other fun things that can amuse even non-developers (such as the ability to turn off individual processor cores or caches...always fun to play with performance).
The little window that appears configures what information it's going to capture. I set it to do "Time Profile" (the default), then "Process" (to isolate it to capturing only Neo), and then choose "soffice.bin" from the process popup menu on the far right.
Switch back to Neo then get it to start lagging (e.g. typing really long sentences or something)
Switch back to Shark. Click "Start". Shark will record what the application is doing for 30 seconds by default.
Switch back to Neo and keep typing to keep it in that lagged state as much as possible during that 30 second window. Again, much easier if it's hung.
Wait for Shark to finish sampling and process. This brings up a window titled "Sample 1 - Time profile of soffice.bin". Save this file and attach it to a BZ issue.
If you need more time than 30 seconds to get the issue to occur (or just want to give more information...it's all statistically sampled, so more is always better!), in Shark go to "Config > Show Mini Config Editor" and increase the "Time limit" edit field.
It is a bit involved, I know, but Shark gives lot more detailed and useful information than the Activity Monitor one which sometimes doesn't give enough insight to really track down issues.
Joined: May 25, 2003 Posts: 4752 Location: Santa Barbara, CA
Posted: Fri Mar 02, 2007 2:27 pm Post subject: Re: 2.1 EA P0 on 10.4.8/G4 - OK for me
Max wrote:
Any tests you would like me to do on an 'unaffected' machine?
Nope, I can't think of any that would be that helpful aside from finding a way to make your machine screw up too It's one of those things that just seems to be isolated to particular machines, and we really need these time profiles on those machines to help track it down. Thanks for the offer though!
Joined: May 25, 2003 Posts: 4752 Location: Santa Barbara, CA
Posted: Fri Mar 02, 2007 2:31 pm Post subject:
pluby wrote:
Since this bug started with Apple's recent Java update, this worries me greatly. I am very upset because I see that not only this bug showed up, but many basic Java bugs showed up as well and the huge list of basic Java stuff that just plain broken in the latest Leopard developer seed is very worrisome.
I hear you. FWIW, it's not just Java that's horked in the latest Leopard seeds; there's a lot of breakages still in other portions of the operating system that affect other apps I've written. Due to NDA I can't go into details, but IMHO I think that if the latest seed is any indication there's still a ways to go (or a few system updates to wait for...).
Still, by now the AWT garbage in the VM should be mature and there shouldn't be this level of regressions
Joined: Apr 25, 2006 Posts: 2315 Location: Montpellier, France
Posted: Fri Mar 02, 2007 2:43 pm Post subject:
This bug really seems awful (I've just read the thread). Sorry, but I don't have any way to help you fix it either.
I see Patrick talking about a build-it-yourself NeoOffice. Rather worrying (that would probably mean the end of NeoOffice).
I know that NeoOffice Intel is stuck on Java 1.5 to circumvent an Intel-specific Java 1.4 bug, but unless it is a serious or crashing bug, plan B could be to revert to Java 1.4.x for all versions of NeoOffice until the Java 1.5 problems are sorted out by Apple ? Or is it missing from Leopard ?
P.S. I assumed this bug doesn't occur with Java 1.4 (I didn't see it mentioned in the thread). Please forgive me if it's also present with 1.4.x.
Sorry for arriving so late in this 3d, but I had some problem too after the last Java update, not only with NeoOffice but also with some other apps.
So checking deeply in the system I noticed that there were some permission problems in my disk.
So I started the iMac with TigerDVD inserted holding down "C" key, did a repairing of drive permissions, and magically everythinf start to works again.
I have read in some other websites (i.g. this one) that these problems are been quite common after the update: could be this the problem our friends are experiencing?
Hi all, here's my first post here ...
well I started with english 2.1 EA p.0 yesterday , my machine is the iMac G5 isight 1.9GHz with 1.5GB RAM, 10.4.8, and I don't seem to have this problem, too.
I have :
java version "1.5.0_07"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.5.0_07-164)
Java HotSpot(TM) Client VM (build 1.5.0_07-87, mixed mode, sharing)
I had no permission problem after the java update.
Let me shortly show these from the terminal too:
- typing on a doc in writer with no autospellcheck :
Sorry for arriving so late in this 3d, but I had some problem too after the last Java update, not only with NeoOffice but also with some other apps.
So checking deeply in the system I noticed that there were some permission problems in my disk.
So I started the iMac with TigerDVD inserted holding down "C" key, did a repairing of drive permissions, and magically everythinf start to works again.
I have read in some other websites (i.g. this one) that these problems are been quite common after the update: could be this the problem our friends are experiencing?
Thanks for the link. I would not be suprised if the last Apple update caused drawing issues for some small combination of graphics cards. Since the update, I now see errors like the following in my console.log. Also, my graphic artist friend's PowerMac's custom colorspaces now no longer work after the Apple update.
Code:
Mar 3 09:19:37 guibomacmini /Applications/iChat.app/Contents/MacOS/iChat: CMSCreateDataProviderOrGetInfo : Invalid colorspace type
Since it is clear that this problem only happens on some machines and there are real problems for some users in the latest Apple update, it may be a good idea to use a standard PC troubleshooting approach. Something like the following should narrow this down. Note that you don't necessary need to do them in any particular order:
1. Eliminate user-installed haxies and other conflicts by creating a new user account on the machine and seeing if you can reproduce the CPU problem in that account. If you cannot, start looking for those haxies (MenuExtra and Unsanity APE are the most common to start looking for).
2. Eliminate the OS as the cause by using Carbon Cloner or a similar tool to copy the user's OS to a firewire drive. Then, hook the firewire drive to a machine that is know to *not* have have this problem and boot from the firewire drive. If the bug now shows up, the OS installation is the culprit.
If neither of the above points to the user-installed haxies or the OS installtion as the cause, then an Activity Monitor sample of NeoOffice while it is in a high CPU usage state is what we'll need.
... the fact is that you two hold the keys to any data that may allow fixing of this before hundreds of thousands of people download Neo 2.1 at the end of this month. Because of this, time is extremely short and, on my end, the clock is ticking very fast.
I've attached a couple of samples taken from my home iBook G3 in bug 2126 which I hope may be of some use. This iBook currently does not yet have the patch from bug 2202 added so might give a 'baseline' system from which to work.
Other samples will follow tomorrow from the iMac G5.
Hope this helps. _________________ Ray Saunders
World Scout Bureau
Last edited by rays on Sun Mar 04, 2007 11:19 am; edited 1 time in total
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