Community
Participate
Working Groups
Build ID: I20080330-1350 Steps To Reproduce: one way to reproduce: 1. select "Project -> Clean" to recompile a large jave project with thousand java class files. 2. the build will start, and normally will take about 1 hour to finish without any problem in 3.4M6, but in 3.4M7, sometime the Eclipse itself will hang up, and even can not be killed by task manager. another way to reproduce 1. start the Eclipse build 2. Open an IE or FF, get some link and "save linke as" to download some file. 3. Eclipse will hang up itself and also the Windows xp system. More information: The system is Dell 530 with dual-core CPU.
I am just wondering whether this problem is caused by the newly added "Java compiler on multi-CPU machines" feature, ================================================================= Java compiler on multi-CPU machines You should see faster Java compile times when using machines with multiple CPUs. The Eclipse Compiler for Java (aka "ecj") is now taking advantage of the multi-threading capability of multi-CPU machines. The full build of a Java project also takes advantage of this support. =================================================================
Can you please attach a thread dump to this bug report? See http://wiki.eclipse.org/How_to_report_a_deadlock#Getting_a_stack_trace_on_Windows
OK, I will check and get thread dump later, as currently I am back using 3.4M6.
I can not reproduce this problem now, but in the past 2 days, it occurs every time I try :( I will use 3.4M7 in the following days and update and give more info in case it reproduces again.
Please use the nightly build from tonight. We just released a fix for bug 229480, so we should get better error reporting from now on.
Any luck trying to reproduce the problem with the latest build ?
I used 3.4M7 and also latest nightly builds during these days, but failed to reproduce this issue again, except only once. but during that hangup, it is not caused by Eclipse, but clearcase dynamic view hang up. Now to this problem, I prefer to think it is not caused by Eclipse, but by clearcase dynamic view, but not with 100% sure. Why not 100% sure is that I also considered this possibility, but the hangup each time during the first two days' usage of 3.4M7 while OK on 3.4M6 at the same time make me suspect it is caused by issues introduced in 3.4M7. Since 3.4M7 didn't show up any problem during this week, so I think the possibility is quite low, and as I this issue can not be reproduced, I can not give any further info. If you also think the possibility in Eclipse is low, you can close this MR as a non-issue...
This problem occured again, but it is impossible to get the dump of stack info, as all application will be hung up, especially those file read/write related operation based on my observation, i.e. If i try to save a file in notepad, the notepad will be hung up. The only thing can do is to press the power button of the PC to shutdown and then restart. yes, this problem is quite strange, and maybe is not caused by Eclipse, but my windows system or some other applications running on my desktop. Yet the point is that I never came to this during my previous experience of 3.4M6. and when the problem occurs, the phenomena is not the same to my clearcase dynamic view hang up. I am afraid I can not provide more info to this ...
Kent - you should explain disabling the concurrent compiler, as a possible variation in the steps.
Good idea - to rule out the changes we released recently, please add to your command line the system property : -Djdt.compiler.useSingleThread=true This will run the compiler in a single thread (as it was in 3.4M6). If your machine hangs up again then we can rule out our changes and focus on something else that might have changed recently.
Now I can almost figure out the root cause now, that it is not caused by Eclipse 3.4M7 or the multi-core CPU support feature, but by clearcase dynamic view issue. As in this week, the hang up issue only occurred about 2 or 3 times, while at the same time I more and more suspect its relationship with clearcase from some phenomena observed. Here I will try to explain this.. I use clearcase dynamic view for development, and in case the network or something else has problem, the clearcase will hang up. Though it is rare, this just happened before. And once the clearcase hang up, it will never be able to be killed by Windows Task Manager, and also Eclipse! The only way to shutdown the machine is to press the power button... So I used to think it is clearcase's problem the first time I got hangup using Eclipse 3.4M7. But then later when I got this problem every time I begin the compilation in 3.4M7, I was not sure what's wrong here. I also noticed there is a big difference to the pure clearcase issue I described above, which will only block clearcase itself and Eclipse, but not other running Windows applications, but this time, all other Windows app will hang up. Just considering this point, I reported this bug, saying that the problem is caused by 3.4M7. Then this week, the experience with Eclipse 3.4M7 is quite smooth, before I got the problem again several days ago. So I switched back to 3.4M6, but then, I also got the same problem once under *3.4M6*. I think this just is the proof that it is not the fault of Eclipse 3.4M7. Now I am developing on clearcase snapshot view, with Eclipse Build id: I20080510-2000. The problem doesn't occur till now. I think the root cause maybe is on some clearcase concurrent issue, or some other network problem? or 3.4M7 multi-thread compiler will increase this possibility to make this issue happen? OK, all these are just assumptions. Anyway, Eclipse 3.4M7 is OK and should not be responsible for this hang up issue. Please close this bug, sorry for the trouble if any.
Closing as per reporter feedback.
Following the reporter's advice, and considering the reasons he gives, Verified for 3.4 RC1.
