|
History
| Log In
|
|
|
|
[
Permlink
| « Hide
]
Carsten Ullrich - 2006-07-25 15:07
Adding Philipp and Martin as watchers. Martin has some strange problems with iCMap, too, but only using Windows.
My problem occurs only on Windows and is like follows: I start iCMap which seems to load the exercise properly. Then the iCMap frame flashes with rendering the exercise and disappears. That is, the frame is still there but the components of the frame are missing.If you wait for a while they might reappear again. If you try to click in the window and do some action the frame is not able to repaint itself. It seems something is blocking.
This behaviour occured only on Windows. First I thought it's due to Java 1.5. For a while it was safe to deinstall all jdks and install a jdk 1.4.2_10 or higher. The other day this method did not work any more. Without any changes in the systrem I asked Philipp what to do and was about to demonstrate that behaviour, For some reason, I was unable to reproduce it. Strangely enough, the assembly tool has no such problems. Therefore I highly recommend to use profiler to investigate where iCMap looses loads of time and where it gets stuck. Currently, we have noc such insight! I suggest to use HPJMeter. There is also a profiler shipped with jdk1.5 but I don't know how to run it inside a Java WebStart application. See also Nicolas also reported DnD problems under Java 1.5 on Linux.
Do both tools work with JDK 1.5 under Windows? I can't confirm for JDK 1.5, because after mixing several JREs and JDKs of 1.4 and 1.5 it turned out to be a mess. I figured out that the best thing is to have one JDK. And since my impression was that 1.4.2_10 or higher was the one that worked I kept this. Currently I have 1.4.2_12. Honestly, after the mess I don't dare to install any other jdk because I fear the next demo will be a desaster. :-|
Maybe, we should collect configurations and compare when it's working:
I am using only I have a similar problem with the OLM, BOTH on Windows and on Linux but only with Java 1.5!
I tried to debug (step-by-step) but it seems to avoid the problem. My guess for the moment is a problem with threads and SWING in Java 1.5. My further guess is some problems with SwingWorker (from the EDU.oswego.cs package, as far as I know both the OLM and ICMAP are using it) because OLM behavior (ie display of GUI independent of data retrieval in separate threads) is not the same with 1.5 and 1.4. Also, I'm using TouchGraph which uses different thread mechanism to separate data retrival from GUI update and it seems to be - sometime - also responsible for the OLM to freeze. So if that helps and if it is indeed the source of the problem, thanks for any info on how to solve it!!! Nicolas I've solved the problem in the OLM.
Or to be honest, I found people who had the same problem and fixed it and so I used their patch. It was definitively a synchronisation deadlock with repainting in the TouchGraph library with Java 1.5 As far as I can see, it has nothing to do with the SwingWorker library. Sorry not to be able to be more clear about how the bug was solved, I think it is too much related to the way TouchGraph is implemented. Nicolas |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
|
Bug
Closed
Major
iCMap causes problems when using Java 1.5 and Linux
jdk 1.4.2_12: